i Economic feasibility of lignocellulosic ethanol production with enzyme recycling by Oscar Rosales Calderon B. Eng., Universidad Nacional Autónoma de México, 2008 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Chemical and Biological Engineering) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) May 2014 © Oscar Rosales Calderon, 2014 ii Abs t ra ct Ethanol produced from lignocellulose is one of the most promising biofuels . However, the technology to produce lignocellulosic ethanol is still under development and needs to be improved to become economically viable. Enzymatic hydrolysis is one of the most expensive process stages , primarily due to high enzyme costs. Consequently , two cost reduction strategies were studied: optimization of hydrolysis conditions and enzyme rec ycle by adsorption. To optimize enzymatic hydrolysis, the changes in concentration of cellulases during the reaction must be determined. Protein concentration changes under hydrolysis conditions for C elluclast 1.5L and Novozyme 188, were studied in the absence of substrate . Novozyme 188 protein concentration decreased by 55 to 64% at 50°C after 92 h. A model describing Novozyme 188 protein concentration changes was developed and used to determine free and adsorbed cellulases concentrations. Glucose and xylose yields (58 to 89% conversion) during enzymatic hydrolysis were modeled as a function of enzyme loading, time, lignin content and solids concentration. The proposed model successfully describes hydrolysis of substrates with different lignin contents, linking pretreatment and hydrolysis . The effect of l ignin content, enzyme loading and hydrolysis time on enzyme recovery wa s evaluated, achieving 0 to 35% cellulases recycled. A mass balance of the enzyme recovery process was built and used to achiev e a uniform production of sugar. Based on experimental data and the proposed models, the production of ethanol with and without enzyme recycling was simulated in AspenPlus . The ethanol production process at different operating conditions was economically evaluated. The economic analysis showed that raw material expenses determine production costs , where biomass , caustic and enzyme expenses are the major contributors to the operating cost . The lowest production cost s ($1.86 and $2.13/ kg ethanol) were obtained at low enzyme loadings and mild pretreatment conditions. Sugar losses at severe pretreatment conditions have a significant negative effect on production costs : severe conditions increased production cost by 18 to 23% . Therefore, optimal hydrolysis conditions must be determined considering the entire process. The implementation of the enzyme recycling process decreased production cost s up to 14% depending on operating conditions, demonstrating the potential benefits of the enzyme recycling technology. iii Prefa ce A version of Chapter 2 has been accepted for publication: Rosales - Calderon, O; Trajano , HL; Posarac, D; Duff, SJB. Stabi lity of commercial glucanase and β - glucosidase preparati ons under hydrolysis conditions. I designed and performed all the experiments and wrote the manuscript. Dr. Duff , Dr. Posarac and Dr. Trajano provided guidance on the research and modeling, and review ed and edited the manuscript. Part of Chapter 3 has been submitted for publication: Rosales- Calderon, O; Trajano , HL; Posarac, D; Duff, SJB. “Modeling of pretreated wheat straw enzymatic hydrolysis as a function of hydrolysis time, enzyme concentration, a nd lignin concentration”. I designed and performed all the experiments, developed the model and wrote the manuscript. Dr. Duff , Dr. Posarac and Dr. Trajano provided feedback and guidance on the research, reviewed and edited the manuscript . A version of Chapter 4 will be submitted for publication. Rosales- Calderon, O; Trajano , HL; Posarac, D; Duff, SJB. “Evaluation of enzyme recovery by adsorption during hydrolysis of pretreated wheat straw”. All the experiments reported in Chapter 4 were performed by mys elf. I developed the mass balances presented and wrote the manuscript. Dr. Duff, Dr. Posarac and Dr. Trajano provided guidance on the research, data analysis, and reviewed and edited the manuscript . Part of Chapter 5 is being prepared for submission as a publication. Techno- economic evaluation of lignocellulosic ethanol production process at different operating conditions . I performed the mass balances, simulations, economic analyses and data analysis presented in Chapter 5 and wrote the manuscript . Dr. Duff, Dr. Posarac and Dr. Trajano provided guidance on the process design, data analysis, and reviewed and edited the manuscript . Part of Chapter 6 is being prepared for submission as a publication. Techno- economic evaluation of the enzyme recycling by adso rption technology in the production of lignocellulosic ethanol . I performed the simulations, economic analyses, data analysis and wrote the manuscript. Dr. Duff, iv Dr. Posarac and Dr. Trajano provided guidance on the process design, data analysis, and review ed and edited the manuscript . v Abstract ........................................................................................................................................... ii Preface ............................................................................................................................................ iii Table of contents ............................................................................................................................. v List of tables ................................................................................................................................... ix List of figures ............................................................................................................................... xiii Nomenclature ............................................................................................................................. xviii Acknowledgements ..................................................................................................................... xxv Dedication .................................................................................................................................. xxvi 1 Introduction ............................................................................................................................. 1 1.1 Background ...................................................................................................................... 1 1.2 Ethanol ............................................................................................................................. 3 1.3 Ethanol production ........................................................................................................... 4 1.4 Feedstock .......................................................................................................................... 5 1.4.1 Agricultural residues ................................................................................................. 6 1.4.2 Cellulose ................................................................................................................... 7 1.4.3 Hemicellulose ........................................................................................................... 8 1.4.4 Lignin ...................................................................................................................... 10 1.4.5 Extractives .............................................................................................................. 11 1.5 Lignocellulosic ethanol production process ................................................................... 12 1.5.1 Biomass pretreatment ............................................................................................. 14 1.5.2 Enzymatic hydrolysis .............................................................................................. 23 1.5.3 Enzyme recycling ................................................................................................... 30 1.5.4 Fermentation of sugars for the production of ethanol ............................................. 37 1.5.5 Ethanol separation and concentration stage ............................................................ 38 1.6 Economic studies for the production of lignocellulosic ethanol .................................... 39 1.7 Research objectives and thesis layout ............................................................................ 42 1.8 Research implications .................................................................................................... 43 2 Stability of commercial enzyme preparations under hydrolysis conditions .......................... 45 2.1 Introduction .................................................................................................................... 45 2.2 Methods .......................................................................................................................... 46 2.2.1 Enzymes .................................................................................................................. 46 2.2.2 Analysis of enzyme activity and protein concentration .......................................... 46 2.2.3 Feedstock ................................................................................................................ 48 T a b l e o f c o n t e n t s vi 2.2.4 Pretreatment ............................................................................................................ 49 2.2.5 Chemical analysis ................................................................................................... 50 2.2.6 Protein concentration stability ................................................................................ 52 2.3 Results and discussion .................................................................................................... 52 2.3.1 Commercial enzyme preparations stability ............................................................. 52 2.3.2 Stability of β - glucosidases from A. niger ............................................................... 55 2.3.3 Novozyme 188 stability .......................................................................................... 60 2.4 Conclusions .................................................................................................................... 68 3 Modeling of pretreated wheat straw hydrolysis as a function of time, enzyme concentration, and lignin concentration ................................................................................................................ 70 3.1 Introduction .................................................................................................................... 70 3.2 Methods .......................................................................................................................... 72 3.2.1 Enzymatic hydrolysis .............................................................................................. 72 3.3 Results and discussion .................................................................................................... 74 3.3.1 Hydrolysis experiments .......................................................................................... 74 3.3.2 Cellulases concentration during enzymatic hydrolysis ........................................... 77 3.3.3 Kinetic model for the enzymatic hydrolysis of lignocellulose ............................... 83 3.4 Conclusions .................................................................................................................... 97 4 Evaluation of enzyme recovery by adsorption during hydrolysis of pretreated wheat straw 99 4.1 Introduction .................................................................................................................... 99 4.2 Methods ........................................................................................................................ 101 4.2.1 Enzymatic hydrolysis conditions .......................................................................... 101 4.2.2 Enzyme recycling ................................................................................................. 102 4.2.3 Accessible liquid in substrate ............................................................................... 104 4.3 Results and discussion .................................................................................................. 105 4.3.1 Effect of hydrolysis time on enzyme recycling .................................................... 106 4.3.2 Enzyme recycling without cellulase supplementation .......................................... 110 4.3.3 Enzyme recycling with cellulases supplementation ............................................. 114 4.4 Conclusion .................................................................................................................... 117 5 Economic evaluation of the production of lignocellulosic ethanol ..................................... 119 5.1 Introduction .................................................................................................................. 119 5.2 Methods ........................................................................................................................ 120 5.2.1 Process and simulation description ....................................................................... 121 5.2.2 Capital cost estimation .......................................................................................... 134 5.2.3 Operating cost ....................................................................................................... 143 vii 5.3 Results and discussion .................................................................................................. 147 5.3.1 Economic evaluation of base case scenarios ........................................................ 147 5.3.2 Sensitivity analyses ............................................................................................... 152 5.3.3 Economic evaluation of the complete simulated proce sses .................................. 162 5.4 Conclusions .................................................................................................................. 174 6 Economic evaluation of the production of lignocellulosic ethanol with enzyme recycling by adsorption .................................................................................................................................... 176 6.1 Introduction .................................................................................................................. 176 6.2 Methods ........................................................................................................................ 177 6.2.1 Process and simulation description ....................................................................... 178 6.2.2 Enzyme recycling process .................................................................................... 181 6.2.3 Cost estimation ..................................................................................................... 181 6.3 Results and discussion .................................................................................................. 183 6.3.1 Economic evaluation of enzyme recovery at different hydrolytic times .............. 183 6.3.2 Sensitivity analysis for ethanol production with enzyme recovery ...................... 188 6.4 Conclusions .................................................................................................................. 189 7 Conclusions and recommendations ..................................................................................... 191 7.1 Conclusions .................................................................................................................. 191 7.2 Future work .................................................................................................................. 197 References ................................................................................................................................... 199 Appendix A: Statistical analysis ................................................................................................. 216 A.1 Novozyme 188 models residuals analysis ........................................................................ 216 A.2 Cellulose hydrolysis model residuals analysis ................................................................. 218 A.3 Xylan hydrolysis model residuals analysis ...................................................................... 220 A.4 Accessible liquid in substrate ........................................................................................... 221 Appendix B: Economic analysis and simulation of ethanol production process ........................ 222 B.1 Mass balances for lignocellulosic ethanol production ...................................................... 222 B.2 Effect of washing pretreated substrates on enzymatic hydrolysis .................................... 232 B.3 Base case scenarios reactors details .................................................................................. 233 B.4 Molecular sieve cost ......................................................................................................... 237 B.5 Raw material cost ............................................................................................................. 240 B.5.1 Biomass ...................................................................................................................... 240 B.5.2 Oxygen ....................................................................................................................... 240 B.5.3 NaOH ......................................................................................................................... 242 B.5.4 Enzymes ..................................................................................................................... 242 viii B.5.5 DAP and CSL ............................................................................................................ 244 B.5.6 Natural gas ................................................................................................................. 245 Appendix C: Economic analysis and simulation of ethanol production process with enzyme recycling ...................................................................................................................................... 246 C.1 Mass balances for lignocellulosic ethanol production with enzyme recycling ................ 246 C.2 Base case scenarios reactor details for ethanol production with enzyme recovery .......... 255 ix Table 1. Properties of ga soline and ethanol (Schwietzke et al., 2008). .......................................... 4 Table 2. Composition of some lignocellulosic materials reported as wt% of dry matter (DM) (Demirbas, 2005). ........................................................................................................................... 6 Table 3. Detailed compositional analysis of wheat straw and corn stover ..................................... 9 Table 4. Characteristics of various economic analyses. ............................................................... 40 Table 5. Composition of raw and pretreated biomass................................................................... 50 Table 6. HPLC operating conditions for carbohydrates determination. ....................................... 51 Table 7. Enzymatic hydrolysis conditions (pretreatment conditions: M=30 min, 6 wt% NaOH/dry biomass, 120°C and S=60 min, 10 wt% NaOH/dry biomass, 150°C). ....................... 73 Table 8. Kineti c parameters determined for hydrolysis of cellulose by fitting equation 38 to scenarios M20- 5 and M40- 5. ........................................................................................................ 89 Table 9. Kinetic parameters determined for the hydrolysis of cellulose by fitting equa tion 38 to scenarios M20- 10 and M40- 10. .................................................................................................... 92 Table 10. Kinetic parameters for the hydrolysis of xylan. ............................................................ 94 Table 11. Enzymatic hydrolys is experimental conditions at 5 wt% solid concentration. Pretreatment conditions: M=30 min, 6 wt% NaOH/dry biomass, 120°C and S=60 min, 10 wt% NaOH/dry biomass, 150°C. ........................................................................................................ 102 Table 12. Polysaccharides and lignin conversions during pretreatment. .................................... 124 Table 13. Arabinan, galactan and mannan conversion used for the hydrolysis process ............. 126 Table 14. Seed reactions and assumed conversions (Humbird et al., 2011). .............................. 128 Table 15. Fermentation reactions (Humbird et al., 2011). .......................................................... 129 Table 16. Characteristics of the beer (D1A, D1B) and rectification column (D2) estimated in AspenPlus for scenario M20- 5.................................................................................................... 132 Table 17. Characteristics of the beer (D1) and rectific ation column (D2) estimated in AspenPlus for scenario M20 - 10. ................................................................................................................... 134 Table 18. Economic parameters for the economic evaluation of lignocellulosic ethanol production ................................................................................................................................... 135 L i s t o f t a b l e s x Table 19. Number and capacity of reactors needed to achieve the required cellulose conversion using batch and continuous reactors ........................................................................................... 138 Table 20. Distillation columns and molecular sieve details for scenario M20 - 5D. .................... 142 Table 21. Separation equipment details for scenario M20- 10D ................................................. 143 Table 22. Raw material and utilities used in the economic analysis .......................................... 144 Table 23. Raw material costs considered in the sensitivity analysis of the lignocellulosic ethanol process ......................................................................................................................................... 146 Table 24. Economic results for the lignocellulosic ethanol process for the base case scenarios. Pretreatment conditions: M=30 min, 6 wt% NaOH/dry biomass, 120°C and S=60 min, 10 wt% NaOH/dry biomass, 150°C. ........................................................................................................ 148 Table 25. Ethanol prices reported in past economic analyses .................................................... 150 Table 26. Fractional factorial design scenarios and production cost. Pret reatment conditions: M=30 min, 6 wt% NaOH/dry biomass, 120°C and S=60 min, 10 wt% NaOH/dry biomass, 150°C...................................................................................................................................................... 153 Table 27. Production cost estimated with equation 51 and APEA for different conditi ons. ...... 158 Table 28. Production costs for all proposed scenarios at different raw material costs estimated with equation 51. ......................................................................................................................... 160 Table 29. Ethanol production costs for scenario M20- 5D. ......................................................... 162 Table 30. Ethanol production costs for scenario M20- 10D. ....................................................... 167 Table 31. Econom ic results for scenarios M20 - 5D and M20- 10D with the addition of water treatment and boiler and turbogenerator processes. .................................................................... 170 Table 32. Cost and conditions considered by Humbird et al. (2011). ........................................ 172 Table 33. Economic results for scenarios M20- 5D and M20- 10D with BT, under costs and considerations reported by Humbird et al. (2011). ..................................................................... 172 Table 34. Production costs as GGE compared to gasoline cost (Energy, 2013) and corn ethanol (USDA Livestock poultry and grain market news, 2014). ......................................................... 173 Table 35. Enzyme recycling scenari os economically evaluated. Pretreatment conditions: M=30 min, 6 wt% NaOH/dry biomass, 120°C and S=60 min, 10 wt% NaOH/dry biomass, 150°C. ... 177 Table 36. Sensitivity analysis for the ethanol produ ction with enzyme recycling. .................... 183 Table 37. Economic results for scenario M40 - 5 with and without enzyme recovery by adsorption at different hydrolytic times. ....................................................................................................... 184 xi Table 38. Economic results for scenario S20- 5 with and without enzyme recovery by adsorption at different hydrolytic times. ....................................................................................................... 185 Table 39 . Economic results for scenar io S40- 5 with and without enzyme recovery by adsorption at different hydrolytic times. ....................................................................................................... 186 Table 40. Production cost of scenario S40- 5 with and without enzyme recycling at different raw material costs .............................................................................................................................. 189 Table 41. Accessible liquid in substrate ..................................................................................... 221 Table 42. Material balance for scenario M20 - 5. ......................................................................... 224 Table 43. Mass balance for scenario M40 - 5. .............................................................................. 225 Table 44. Mass balance for scenario S20 - 5. ............................................................................... 226 Table 45. Mass balance for scenario S40 - 5. ............................................................................... 227 Table 46. Mass balance for scenario M20 - 10. ............................................................................ 228 Table 47. Mass balance for scenario M4 0- 10. ............................................................................ 229 Table 48. Mass balance for scenario S20 - 10. ............................................................................. 230 Table 49. Mass balance for scenario S40 - 10. ............................................................................. 231 Table 50. Reactor details for scenario M20 - 5. ............................................................................ 233 Table 51. Reactor details for scenario M40 - 5. ............................................................................ 234 Table 52. Reactor details for scenario S20 - 5. ............................................................................. 234 Table 53. Reactor details for scenario S40 - 5. ............................................................................. 235 Table 54. Reactor details fo r scenario M20- 10. .......................................................................... 235 Table 55. Reactor details for scenario M40 - 10. .......................................................................... 236 Table 56. Reactor details for scenario S20 - 10. ........................................................................... 236 Table 57. Reactor details for scenario S40 - 10. ........................................................................... 237 Table 58. Constants for the molecular sieve desiccant (Gas processors suppliers associat ion, 2004). .......................................................................................................................................... 238 Table 59. Enzyme cost for different scenarios, considering an enzyme cost contribution of $1/gal ethanol ......................................................................................................................................... 243 Table 6 0. Natural gas composition ............................................................................................. 245 Table 61. Material balance for scenario M40 - 5- R5. ................................................................... 247 Table 62. Material balance for scenario M40 - 5- R24. ................................................................. 248 Table 63. Material balance for scenario M40 - 5- R48. ................................................................. 249 xii Table 64. Material balance for scenario S20 - 5- R24. .................................................................. 250 Table 65. Material balance for scenario S20 - 5- R48. .................................................................. 251 Table 66. Material balance for scenario S20 - 5- R5. .................................................................... 252 Table 67. Material balance for scenario S20 - 5- R24. .................................................................. 253 Table 68. Material balance for scenario S20 - 5- R48. .................................................................. 254 Table 69. Reactor details for scenario M40 - 5- R5 ....................................................................... 255 Table 70. Reactor details for scenario M40 - 5- R24 ..................................................................... 255 Table 71. Reactor detail s for scenario M40 - 5- R48 ..................................................................... 256 Table 72. Reactor details for scenario S20 - 5- R24 ...................................................................... 256 Table 73. Reactor details for scenario S20 - 5- R48 ...................................................................... 257 Table 74. Reactor details for scenario S20 - 5- R5 ........................................................................ 257 Table 75. Reactor details for scenario S40 - 5- R24 ...................................................................... 258 Table 76. Reactor details for scenario S40 - 5- R48 ...................................................................... 258 xiii Figure 1. Greenhouse gas emissions reductions by the use of bioethanol produced from corn (process fueled by natural gas or biomass) and cellulosic materials (Wang et al., 2007). ............. 3 Figure 2. Partial structure of cellulose (Rowell et al., 2005). ......................................................... 8 Figure 3 . Structure of xyloglucan in flowering plants; principal component of the hemice lluloses. In blue backbone β - D- glucans; in red α - D- xylose; in black α- D- galactose and in brown α- L-fucose residues (Ochoa - villarreal et al., 2012). .............................................................................. 9 Figure 4. Structural model for gymnosperm lig nin. Linkages to additional lignin chains are indicated (L- containing circles) (Ruiz - Dueñas and Martínez, 2009) ........................................... 11 Figure 5. Schematic flowsheet for the conversion of biomass to ethanol. ................................... 13 Figure 6. Schematic diagram of the pretratment effect on lignocellulose. ................................... 14 Figure 7. Initial chemical reactions involved in oxygen deligni fication. ..................................... 20 Figure 8. Superoxide induce the cleavage of carbon –carbon bonds (Gierer et al., 2001) ............ 21 Figure 9. Reduction of oxygen to water that occurs during oxygen delignification (Gierer, 1997)........................................................................................................................................................ 22 Figure 10. General mechanism of the enzymatic hydrolysis of cellulose by endo - glucanases, exo-glucanases and β - glucosidase (Divne et al., 1998; Helmich et al. to be publish, Kleywegt et al., 1997). ............................................................................................................................................ 26 Figure 11. Hydrolysis profiles of steam - exploded Douglas - fir over time at different enzyme loadings (FPU/g cellulose), 2 wt% DM, and 45°C (Lu et al., 2002). ........................................... 28 Figure 12. Enzyme recycling e xperimental diagram. ................................................................... 32 Figure 13. Hydrolysis yield achieved in each recycling round by adsorption for the hydrolysis of alkali pretreated wheat straw at 2 wt% of DM solids concentration, 20 FPU/ g cellulose, 150 rpm and 50°C using a commercial enzyme preparation from Sunson Group Ningxia (Qi et al., 2011)........................................................................................................................................................ 34 Figure 14. Schematic diagram of the ethanol separation process (Aden et al., 20 02; Humbird et al., 2011; Kazi et al., 2010; Kumar and Murthy, 2011). ............................................................... 39 L i s t o f F i g u r e s xiv Figure 15 . Protein calibration curves. Protein calibration curves for P - cellulase and P- β -glucosidase; and BSA . .................................................................................................................. 48 Figure 16. Ground raw and pretreated wheat straw ...................................................................... 48 Figure 17. Reactor used to conduct the oxygen delignification .................................................... 49 Figure 18. High- performance liquid chromatography (HPLC) equipment used to determine sugars concentration...................................................................................................................... 51 Figure 19. Celluclast 1.5L and Novozyme 188 protein concentration as a function of time at 50°C. Celluclast 1.5L and Novozyme 188 protein concentrat ion at different initial concentrations (6 g/L , 3 g/L and 1.4 g/L ) over time at 50°C. Line s are added to assist in visualizing trends. ......................................................................................................................... 53 Figure 20. Cocktails protein concentration at 20 and 50°C. Ce lluclast 1.5L and Novozyme 188, Celluclast 1.5L and Novozyme 188 protein concentration over time at 20 and 50°C. Lines are added to assist in visualizing trends. ............................................................................................. 54 Figure 21. Schematic aggregation of proteins. Idealized schematic representation of a general overall pathway to the formation of a ggregates (Morris et al., 2009). ......................................... 56 Figure 22. P- β - glucosidase concentration (Panel A), normalized concentration (Panel B) at 50°C, experimental data ([B 0]=0.317 g/L , 0.612 g/L and 1.388 g/L ) and 1st order kinetic model, equation 6. ..................................................................................................................................... 59 Figure 23. Effect of pretreated biomass on Novozyme 188 protein concentration. Novozyme 188 protein concentration at 50°C over time in the absence and presence of p retreated biomass: substrate M and S. Line added to assist with visualization. .......................................................... 61 Figure 24. Proposed kinetic model fit on Novozyme188. Novozyme protein concentration an d normalized concentration at 50°C, experimental data ([N 0]=0.36 g/L , 0.71 g/L and 1.44 g/L ) and the additives model, equation 10. ................................................................................................. 64 Figure 25. Proposed kinetic model considering the presence of additives in Novozyme188. Novozyme 188 protein normalized concentration at 50°C, experimental data ([N 0]=0.36 g/L , 0.71 g/L and 1.44 g/L ) and the additives model, equation 14. ..................................................... 67 Figure 26. Cell ulose conversion during hydrolysis for substrate M and S . ................................. 75 Figure 27. Cellulose and xylan conversion during enzymatic hydrolysis at 5 wt% (panel A) and 10 wt% D M solids concentration. Pretreatment conditions: M=30 min, 6 wt% NaOH/dry biomass, 120°C and S=60 min, 10 wt% NaOH/dry biomass, 150°C.. ......................................... 77 xv Figure 28. Glucose concentration during enzymatic hydrolysis from experimenta l data and equation 38. Cellulases concentration in solution calculated from the total protein concentration measured and Novozyme 188 protein concentration calculated from equation 14. ..................... 81 Figure 29. Sch ematic representation of overall pathway for the enzymatic hydrolysis of cellulose by cellulase.................................................................................................................................... 83 Figure 30. Xylose concentration during enzymatic hyd rolysis from experimental data and proposed kinetic model, equation 46. Pretreatment cond itions: M=30 min, 6 wt% NaOH/dry biomass, 120°C and S=60 min, 10 wt% NaOH/dry biomass, 150°C. .......................................... 97 Figure 31. Schematic diagrams of enzyme r ecycling methodology without and with addition of fresh cellulase at the start of the 2 nd hydrolysis round. The addition of fresh cellulase to enzymatic hydrolysis was used to evaluate the technical feasibility of a continuous process. .. 104 Figure 32. Cellulases distribution on fresh substrate after adsorption process ........................... 106 Figure 33. Percent of recycled cellulases achieved in each scenario at differ ent times referred to initial amount of cellulases loaded in the first round. ................................................................. 109 Figure 34. Cellulose and xylan conversion after 48 h as a percentage of the oretical maximum for the first and second hydrolysis rounds. Perce nt cellulases recycled by adsorption is calculated from equation 47 and cellulase recovery is defined by equation 49. .......................................... 111 Figure 35 . Total protein concentration (cellulases and Novozyme 188) at the end of the 1 st and 2nd hydrolysis round (48 h hydrolysis time), and Novozyme 188 protein concentration predicted by equation 14 for 48 h hydrolysis time. .................................................................................... 113 Figure 36. Glucose, xylose and total enzyme concentration at the end of the 1st and 2nd hydrolysis rounds at different recycling times (t1, t2, t3). Experime ntal concentrations of glucose , xylose and total protein concentrations from Figure 28. ............................................. 116 Figure 37. Mass balance of the primary components for 48 h enzymatic hydrolysis and enzyme recycling for scenario S40. ......................................................................................................... 117 Figure 38. Proposed process for the production of lignocellulosic ethanol. ............................... 120 Figure 39. Pretreatment, enzymatic hydrolysis and fermentation stages in the AspenPlus simulation for the production of lignocellulosic ethanol. Pretreatment reactor (RDELIG), rotary drum vacuum filter (RDVF1), hydrolysis reactor (RHYD), seed reactor (RSEED), sugar loss reactor (RLOSS), fermentation reactor (RFERMNT), pumps (P1 - P8), tanks (T1- T3), heaters (H1- H4), flashes (F1- F3), diammonium phosphate (DAP) and corn steep liquor (CSL). ......... 123 xvi Figure 40. Separation process simulation for scenario M20- 5. .................................................. 130 Figure 41. Simulated separation process for scenario M20- 10. .................................................. 133 Figure 42. Hydrolysis batch schedule for scenario M40- 5 ......................................................... 137 Figure 43. Fermentation batch schedule for scenario M20- 5 ..................................................... 140 Figure 44. Enzyme, biomass and caustic costs contribution to the total raw material expenses. 149 Figure 45. Analysis of p - values for the selection of relevant variables in the ethanol production economy obtained with JMP 10. ................................................................................................. 155 Figure 46. Production cost calculated with APEA and predicted with equation 51 ( ), 5% significance curve obtained with JMP 10. .................................................................................. 157 Figure 47. Effect of evaluated variables on the production cost of ethanol. .............................. 158 Figure 48. Process stage contribution to the TDC in scenario M20- 5D. .................................... 164 Figure 49. Breakdown of cost for the production of lignocellulosic ethanol in scena rio M20- 5D..................................................................................................................................................... 166 Figure 50. Instal led equipment cost distribution per process for the ethanol production in scenario M20- 10D. ..................................................................................................................... 168 Figure 51. Breakdown of cost for the production of lignocellulosic ethanol in scenario M20- 10D..................................................................................................................................................... 169 Figure 52. Process diagram for the production of lignocellulosic ethanol with enzyme recovery...................................................................................................................................................... 178 Figure 53. Simulation for the production of lignocellulosic ethanol with enzyme recovery. Pretreatment reactor (RDELIG), rotary drum vacuum filter (RDVF1), hydrolysis reactor (RHYD), adsorption reac tor (RADS), solid- liquid separators (FI1 and FI2), seed reactor (RSEED), sugar loss reactor (RLOSS), fermentation reactor (RFERMNT), pumps (P1 - P8), tanks (T1- T2), heaters (H1- H4), flashes (F1 - F4), diammonium phosphate (DAP) and corn steep liquor (CSL)........................................................................................................................................... 180 Figure 54. Production cost f or evaluated scenarios without and with enzyme recycling at different hydrolytic times. ........................................................................................................... 187 Figure 55. Re siduals for the predicted P - β - glucosidases concentration (equation 6) ([B 0]=0.32 g/L, 0.61 g/L and 1.39 g/L ). ........................................................................................................ 216 Figure 56. Residuals for the predicted P - β - glucosidases concentration (equation 10) ([B 0]=0.32 g/L, 0.61 g/L and 1.39 g/L ). ........................................................................................................ 217 xvii Figure 57. Residuals for the predicted P - β - glucosidases concentration (equation10) ([B 0]=0.32 g/L, 0.61 g/L and 1.39 g/L ). ........................................................................................................ 217 Figure 58. Residuals for predicted glucose concentration (equation 38) .................................... 218 Figure 59. Residuals for predicted glucose concentration (equation 38 with the a ddition of the effectiveness factor) .................................................................................................................... 219 Figure 60. Residuals for predicted glucose concentration (equation 38 with parameters obtained at 10 wt% solids concentration) .................................................................................................. 219 Figure 61. Residuals for predicted xylose concentration (equation 46 with parameters obtained at 5 and 10 wt% solids concentration) ............................................................................................ 220 Figure 63. Flowsheet of the pretreatment and enzymatic hydrolysis mass balance. .................. 222 Figure 64. Washing effect on the enzymatic hydrolysis of pretreated wheat straw. Poorly (0.4 L per batch) and extensively (4 L per b atch) pretreated biomass wash. ........................................ 233 Figure 65. Flowsheet of the pretreatment and enzymatic hydrolysis with enzyme recycling. ... 246 xviii $ US Dollars A Additives in Novozyme 188 AFEX Ammonia fiber/freeze explosion pretreatment AICc Akaikes` information criterion APEA Aspen Process Economic Analyzer ARP Ammonia recycling percolation B6 CO2 separation unit in AspenPlus simulation B β - glucosidase BC Biomass cost ($/ ton DM) variable in the mathematical expression proposed for the ethanol production cost prediction BSA Bovine Serum Albumin BT By- product boiler and turbogenerator system C1 Enzyme loading selector unit in Aspen Plus simulation C2 Solids concentration selector unit in Aspen Plus simulation CSS MS correction factor for saturation (dimensionless) CT MS correction factor for temperature (dimensionless) C Cellulose CBD Cellulose- binding domain in cellulases CBH Cellobiohydrolases or exo- glucanases CC Cellulose conversion (%) CCD Cellulases catalytic domain Ce Cellobiose N o m e n c l a t u r e xix CE* Cellulose- cellulase complex CISOLID Solid phase in AspenPlus simulation Cm Average change of cellulose conversion over change of time for substrate M Cm20 Average change of cellulose conversion over change of time for M20 at different solids concentrations Cm40 Average change of cellulose conversion over change of time for M40 at different solids concentrations Cs Average change of cellulo se conversion over change of time for substrate S CSL Corn steep liquor CSTR Continuous stirred- tank reactor C/X Average ratio of change in cellulose and xylan conversions with respect to time D* Denatured enzyme or protein structure D1 Beer column in scenario M20- 5 D1A and D1B Beer columns in scenario M20- 10 D2 Rectification column DM Dry matter Dmin Molecular sieve bed minimum diameter (m) DA Dilute acid pretreatment DAP Diammonium phosphate E10 90% gasoline - 10% ethanol blend E85 15% gasoline - 85% ethanol blend EA Cellulases concentration in the bulk liquid after adsorption (g/L) Ed Deactivated cellulases EH Cellulases concentration in the bulk liquid at the end of the hydrolysis (g/L) EL Mass of cellulases loaded for hydrolysis bef ore recycling process (g) ERe Mass of cellulases recycled (g) xx E Cellulases EC Enzyme cost ($/kg enzyme) variable in the mathematical expression proposed for the ethanol production cost prediction EG Exo- glucanases EL Enzyme loading (FPU/g cellulose) v ariable in the mathematical expression proposed for the ethanol production cost prediction FL Lignin factor (g cellulase/g lignin) F1 Flash unit to removed oxygen after pretreatment F Xylan FE* Cellulases- xylan complex FFV Flexible- fuel vehicles FI1 Solid residue and hydrolytic liquor separation unit FI2 Separation unit after adsorption during the enzyme recycling process G Glucose H Constant of the m olecular sieve particle ( h2/m 3 ) HMF Hydroxymethylfurfural HPLC High- performance liquid chromatog raphy IRR Internal rate of return k 1 Rate constant of β- glucosidase unfolding (1/ h) k 2 Rate constant of aggregate formation (D2∗) (1/ h) k 3 Rate constants of cellulases - cellulose complex formation (L/h g) k 3x Rate constants of cellulases - xylan complex formation (L/ g h) k 4 Rate constants of cellobiose production (1/h) k 4x Rate constants of xylose production (1/h) k 5 Rate constants of glucose production (1/h) k 1’ Rate constant for protein unfolding ( 1/ h) k 2’ Rate constant for protein refolding (L / g h) xxi k 1’’ Rate constant for protein u nfolding ( 1/ h) in the presence of additives k 2’’ Rate constant for protein refolding (L / g h) in the presence additives k - 3 Rate constants of cellulases desorption from cellulose (1/h) k - 3x Rate constants of cellulases desorption from xylan (1/h) k d Rate constants of cellulases deactivation (L/g h) K e Equilibrium constant for the production of cellobiose (g/L) K ex Equilibrium constant for the production of xylose (g/L) k n+j Rate constant of the aggregates formation ( 𝐷𝑛+𝑗∗ ) (1/ h) k m Rate constant of the aggregates formation ( 𝐷𝑚∗ ) (1/ h) K Constant of the m olecular sieve particle (h2/m 2) LMTZ MS length of the mass transfer zone (m) Ls Length of the saturation zone (m) LT Total MS bed height (m) L Lignin Le Bed length (m) LHW Liquid hot water pretreatment M Substrate M mc Dimensionless constant that relates protein and additives concentration in Novozyme 188 mf Mass flow rate entering molecular sieve (kg/h) MIXED Liquid phase in AspenPlus simulation MM Millions MS Molecular sieve unit N Novozyme 188 protein concentration (g/L) NC NaOH cost variable in the mathematical expression for the ethanol production cost ($/ Ton) NREL National Renewable Energy Laboratory xxii OD Oxygen delignification ∆PD MS total design pressure drop (kg/m2) P- β - glucosidase Lyophilized powder β- glucosidase from A. niger P- cellulase Lyophilized powder cellulases from T. reesei P Pressure (kg/m 2) PT Pretreatment conditions (°C) variable in the mathematical expression proposed for the ethanol production cost prediction q Volumetric flow rate entering MS (m 3 /h) rG Glucose kinetic rate during enzymatic hydrolysis RADS Adsorption reactor RDELIG Pretreatment reactor RDVF Rotary drum vacuum filter RDVF1 Unit for t he separation of pretreated solids from the pretreatment broth RFERMENT Fermentation reactor RHYD Hydrolytic reactor RLOSS Sugar losses reactor RSEED Seed reactor S Substrate S SAA Soaking in aqueous ammonia pretreatment SC Solids concentration (wt % DM) variable in the mathematical expression proposed for the ethanol production cost prediction SHF Separate hydrolysis and fermentation SP SO2 impregnation pretreatment SS Mass of desiccant in the MS (kg) SSF Simultaneous saccharification and f ermentation t Time (h) TA Total protein concentration after adsorption for enzyme recycling (g/L) xxiii TE Total protein concentration (g/L) TH Total protein concentration (cellulases, β- glucosidase and cocktail proteins) at the end of hydrolysis (g/L) TDC Total direct cost ($) v0 Volumetric flow entering reactor (L/h) Vadj Adjusted velocity in MS (m/h) Ve Superficial velocity in MS (m/h) Vmax Maximum superficial velocity (m/h) V Reactor volume (L) W r Amount of water removed during the MS adsorption period (kg) WT Water treatment system X Xylose XC Xylan conversion (%) Z MS mass transfer zone coefficient (m) Subscript 0 Initial concentration i i number of monomeric units in an aggregate (i<m) j j number of monomeric units in an aggregate (j<m) m m number of monomeric units in an a ggregate (m≥2) (n+i+j<m) n n number of monomeric units in an aggregate (n≥2) (n<m) R Reactor number in the hydrolysis train xxiv Units ADT Air dry ton of pulp product the weight of the pulp product is corrected to reflect the weight that the pulp would be i f the pulp were composed of 10% water and 90% fibre . CBU Cellobiase units FPU Filter- paper units GGE Gasoline gallon equivalent ODT Oven dry ton RCF Relative centrifugal force SCF Standard cubic foot ton Short ton Greek symbols ∆ Denotes change of any changeable quantity μ Viscosity (kg/m h) η Effectiveness factor ρB MS bulk density (kg/m 3 ) ρ Density (kg/m 3 ) xxv A c k n o w l e d g e m e n t s I would like to express my gratitude to my supervisors Dr. Duff, Dr. Posarac and Dr. Trajano for the opportunity to pursue a Ph. D. and to be part of their research group. I greatly appreciate the time and effort dedicated to the competition of my thesis and to develop my professional skills, but also for the good moments and laughs during these years. Next, I would like to acknowledge my committee members, Dr. Smith and Dr. Arantes for their criticism and valuable feedback to improve my research. Special thanks to my thesis examiners Dr. Mark Martinez and Dr. Jack N. Saddler. I would like to t hank all my lab mates, Sam Li, Derek Pope, David Kuan, Swati Yewalkar and Tong Wu, for their help and for creating an enjoyable environment to work in. My thanks go to all the faculty and staff of the Department of Chemical and Biological Engineering for their advice and assistance. I would also like to thank the Natural Science Engineering Research Council (NSERC) for providing the necessary funding for this research and to the Consejo Nacional de Ciencia y Tecnologia (CONACYT) for funding provided in t he first years of the program . Especial thanks to Dr. Costas Bas ín , Dr. Gómez Puyou and Dr. Duran Moreno for all their support to start my Ph. D. Next, I wish to thank my office mates, Amin Aziznia and especially Max Nodwell, for his advice and help at the beginning of my research and program . Very special thanks to my girlfriend, Erika Sato, for all her help, advice, energy, time and love during the complet ion of this project. I could have not achieved my goals without her constant support and love. I want to especially thank my parents, Ulises and Dolores, who provided me with all the necessary skills and attributes to pursue and reach all my goals , thank you for all your love. I would also like to thank my grandfathers, Rafael and Carmen, w ho have encouraged and made me smile in the hardest moments. Finally, thanks go to all my family , who have had a significant impact on my life and who have gave me endless happiness . xxvi Dedicated to my parents, my family and in memory of my grandpa Rafael & to Erika D e d i c a t i o n 1 1 I n t r o d u c t i o n 1.1 Background In the last century, the global population increased from 1.6 to 7.1 billion (Lutz et al., 2004; U.S. Department of Commerce, 2014) , increasing the energetic demand for transportation and industrial process es. In order to supply the required energy , there is a need for a large and stable source of energy. Currently , the main source of energy is petroleum, which is non-renewable and therefore , limited. The socio- political pressure to reduce greenhouse gas emissions combined with unstable oil prices make it imperative to find alternative sources of energy (Wang et al., 2007; Wyman, 1999) . In order to encourage the use and improvement of alternative energy sources, governments and organizations around the world have started to introduce plans to encourage the development of biofuels, solar or geothermal energy. For example, the European Union has made a commitment to raise the use of renewable energy to 20% by 2020 and to increase the fraction of biofuels in transport fuel to 10% by 2020 (Piebalgs, 2007) . The final solution to replace petroleum as the main source of energy and to decrease the production of contaminants will most likely consist of different alternative sources of energies. Bioethanol refers to ethanol produced from biomass and is an essential part of the solution to decrease the dependence on fossil fuels and environmental problems . Bioethanol can be mixed with gasoline and used as fuel in the transportation sect or. Thus, the use of bioethanol can reduce our dependency on fossil fue ls by reducing our consumption of gasoline. The US Department of Agriculture and Department of Energy have estimated that the US has the resource potential to produce over 1 billion tons of biomass annually, thus accounting for close to 30% displace ment of current American fossil fuel usage (about 80 billion gal) (Gray et al., 2006) . Bioethanol can be produced in different parts of the world due to the large biomass availability which makes bioethanol an attractive fuel for countries 2 without domestic petroleum reserves (Singh et al., 2010; Wyman, 1999) . Consequently , bioethanol offers political and economical security to countries without petroleum resources. The transportation sector produce s one- third of the total greenhouse gas emissions, consequently, the use of ethanol as fuel may have a significant impact on worldwide greenhouse emissions (Singh et al., 2010; Tao et al., 2013; Wyman, 1999) . Carbon dioxide (CO2) is a by- product of the etha nol production from biomass and is also produced during the burning of gasoline - bioethanol blends. However , the CO2 released to the atmosphere is consumed by growing biomass via photosynthesis . Therefore, it has been proposed that the net life cycle CO 2 emissions from bioethanol production and use is equal to zero (Lynd et al., 1991) . It is impor tant to note that predictions of life cycle CO2 emissions depend heavily on the assumed production technology . Multiple studies have discussed the positive and negative impact s of bioethanol from ecological and energetic point of view. These reviews have reported both favorable and unfavorable judgements for the production of bioethanol. The different opinions derive from the variety of conditions and assumptions used in each study. Consequently , the results can only be compared using conditioning factors , and, in some cases their comparison is impossible (Quirin et al., 2004; Singh et al., 2010) . These studies do not cover the full range of possible feedstocks and geographies, producing different conclusions about the economic and environmental benefits of bioethanol (Von Blottnitz and Curran, 2007) . Nonetheless, life cycle impact studies of bioethanol production have shown that all bioethanol production is mildly to strongly beneficial from a climate protection and a fossil fuel conservation perspective (Von Blottnitz and Curran, 2007) . Figure 1 presents the reductions in greenhouse gas emission for the production of ethanol from corn, using natural gas or biomass (e.g. corn syrup) to fuel the process , and from lignocellulos ic materials (e.g. wood, grass) . In general, the information available indicates that bioethanol is a viable alternative biofuel but its range of benefit s depends on the feedstock, process, geography, gasoline price and other variables. 3 Figure 1. Greenhouse gas emissions reductions by the use of bioethanol produced from corn (process fueled by natural gas or biomass) and cellulosic materials (Wang et al., 2007) . 1.2 Ethanol Ethanol performs well in cars , in its pure form in dedicated engines or in a mixture with other fuels. Ethanol has been introduced as a fuel in Brazil, the United Sta tes of America and some European countries, and it is expected that in 20 years, ethanol will be the primary renewable biofuel used in the transportation sector (Hahn - Hägerdal et al., 2006) . The use of ethanol either neat or blended increases the octane rating and the f uel’s oxygen content thus facilitating complete combustion. The high octane number of ethanol enables a higher compression ratio leading to greater engine efficiencies (Wyman, 1999) . The most commonly used blends are E85 (15% gasoline - 85% ethanol) and E10 (90% gasoline - 10% ethanol) (Demirbas, 2005) . Today, E10 is sold in every state of the United States of America . In fact, more than 95% of U.S. gasoline contains up to 10% ethanol (U.S. Department of Energy, 2013a) . In Canada, it is require d that the gasoline produce d or import ed has at least 5% (v/v) of renewable fuels (Environment Canada, 2011) . Bioethanol has the potential to replace large amounts of gasoline if manufacturing companies produced more flexible - fuel vehicles (FFV) that can use the E85 blend. Some important chemical properties of ethanol and gasoline are shown in Table 1. 100% 72% 48% 14% Gasoline Corn ethanol (fuelled by natural gas) Corn ethanol (fuelled by biomass) Lignocellulosic ethanol 4 Table 1. Properties of gasoline and ethanol (Schwietzke et al., 2008) . Che m ic al prope rti e s Gasol i ne Ethanol Density (kg/m 3 ) 725 790 Lowe r Heating Value (MJ/L) 35 23 Octane 87 115 Oxygen content (wt% ) 0 35 The efficiency of ethanol and gasoline as fuels in the transportation sector is difficult to compare due to the number of variables that must be considered and controlled, e.g. season and car condition. However, the higher density and lower energy content of ethanol shows that the ethanol volume required to produce the same energy of gasoline is greater (Schwietzke et al., 2008) . Another way to compare the gasoline and ethanol is with the gasoline gallon equivalent (GGE), which i s the amount of alternative fuel that equals the energy content of one liquid gallon of gasoline ; pure ethanol is equivalent to 1.5 GGE. 1. 3 Ethanol p roduction The production of ethanol from biomass can be classified into two generations based on the substrates used. First generation ethanol is produced from crops such as sugarcane, beet, corn, wheat, barley, rye, sorghum, and cassava that contain starch or sucrose. This technology is in place, and great majority of commercial production of bioethanol is from starch or sucrose . Second generation ethanol is produced from lignocellulosic biomass (lignocellulosic ethanol), such as stalks, leaves and husks of corn plants, wood chips, wheat straw , and energy crops such as switchgras s (Coppola et al., 2009) . Processes for production of advanced bioethanol from lignocellulosic materials by biochemical conve rsion are now operated on a large scale : by DONG Energy (Larsen et al., 2012), Abengoa (Abengoa Bioenergy), Chemtex/ Beta Renewables (Beta Renew ables) and DuPont (DuPont Biofuels Solutions ) have built demonstration plants . Since first generation feedstock is also used as human and animal food, the prices of food and land could increase with first generation ethanol consumption. In contrast, 5 lignocellulosic materials are available at low er prices as they are not part of the food chain, and their production reduces environmental risks like soil degradation, and water and air pollution relative to first generation biofuels (Gnansounou and Dauriat, 2010) . Second generation technology is undergoing thorough optimization to increase its cost - performance . Consequently, the second generation technology has a great potential due to its low substrate cost and high availability. Due to the complex structure of lignocellulosic materials, the process required to produce ethanol from lignocellulose is more expensive and complex than that required for conversion of sucrose or starch to ethanol (Coppola et al., 2009) . Ethanol production cost s can be reduced in different ways such as linking the process to biorefineries, or by co- production of heat and electricity through combustion of its own waste residues (Coppola et al., 2009) . Lignocellulosic materials offer a range of advan tages and opportunities for improvement that food crops cannot offer . Therefore, this thesis focused on the production of bioethanol from lignocellulosic materials. Commercial production of lignocellulosic ethanol require s reductions in processing cost s. In order to understand the difficulties in converting lignocellulose to ethanol, the next section presents a brief int roduction to lignocellulosic materials properties and components . 1.4 Feedstock Lignocellulosic biomass is the most abundant renewable organic component in the world, its annual production has been estimated in 1*10 10 ton worldwide (Alvira et al., 2010). The total potential bioet hanol production from crop residues and wasted crops is approximately 491 billion L/year, about 16 times higher than the current world bioethanol production (Balat et al., 2008). Lignocellulose consists of three main components: cellulose , hemicelluloses and lignin, and other minor components such as resins, salts, phenolic s, and fatty acids . The two major components (cellulose and hemicelluloses) are composed of chains of sugar molecules. These chains can be disrupt ed in order to produce monomeric sugars, however, these 6 reactions are more complicated than the reactions for the conversion of starch to sugars (Galbe and Zacchi, 2002) . Second generation ethanol can be produced from different kinds of raw materials : wood, municipal solid waste, waste - paper, agricultural residues and energy crops . The composition of differe nt lignocellulosic materials is shown in Table 2. Table 2. Composition of some lignocellulosic materials reported as wt % of dry matter (DM) (Demirbas, 2005) . M at er i al Cel l ul ose ( wt % DM) Hem i ce ll ulos e (w t % DM ) Li gni n (w t % DM) Ash (wt % DM) Ext r ac t i ve s (w t % DM) Algae (green) 20- 40 20- 50 - - - Cotton 80- 95 5 - 20 - - - Grasses 25 - 40 25 - 50 10- 30 - - Hardwoods 43 - 47 25 - 35 16 - 24 0.4- 0.8 2- 8 Softwoods 40- 44 25 - 29 25 - 31 0.4- 0.6 1- 5 Corns stover 39 - 47 26 - 31 3 - 5 12- 16 1- 3 Wheat straw 37 - 41 27 - 32 13 - 15 11- 14 5 - 9 Newspapers 40- 55 25 - 40 18- 30 - - 1.4.1 Agricultural residues Agricultural residues are advantageous because they are abundant, inexpensive, do not compete with current farmland, and do not require deforestation for new land. Agricultural residues are therefore likely to be the earliest feedstock adopted for lignocellulosic ethanol production, followed by the use of forest residues and, finally , the development of energy crops (Mabee and Saddler, 2010). Based on these advantages, agricultural waste s are the focus of this thesis . Agricultural activity in Canada produces millions of tons of biomass each year and has the potential to offer feedstocks for bioenergy and specific bioproducts while improving the rural economy (Wood and Layzell, 2003) . Ethanol from agricultural residues, forest and mill 7 residues, and energy crops may account for 6% to 60% of Canadian gasoline consumption, dependi ng upon effectiveness of conversion technology and access to feedstocks . In particular, conversion of a gricultural residues to ethanol could replace 0.7% to 11.9% of Canadian gasoline consumption (Mabee and Saddler, 2010). One specific limitation to the use of agricultural residues is that s ome biomass must be used as tillage to prevent soil erosion and maintain soil nutrient and organic matter levels, therefore only 20–70% of the biomass is available for use (Wood and Layzell, 2003) . The first lignocellulosic residues used by the Canadian biofuel industry are wheat straw, co rn cobs, and corn stover (Mabee and Saddler, 2010). In Denmark, Inbicon has built an advanced biorefinery which converts wheat straw i nto ethanol, lignin pell ets and C5 molasses (Larsen et al., 2012). Iogen currently operates a demonstration plant in Ottawa using wheat straw, while Greenfield Ethanol has developed a pilot plant in Chatham, Ontario that uses corn residues including cobs and stover (Mabee and Saddler, 2010). The largest mass of Canadian agricultural residues is wheat straw (20.6 M DM T/yr) (Wood and Layzell, 2003) and thus wheat straw is the focus of this thesis. 1.4.2 Cellulose Cellulose, (C6 H10O5 )n, is the most abundant organic chemical on earth. It is a fi brous, tough, water - insoluble substance found in the protective cell walls of plants, particularly in stalks, stems, trunks and all woody portions of plant tissue. Cellulose is a glucan polymer of D -glucopyranose units, which are linked together by β - (1→4)- glucosidic bonds. The building block for cellulose is cellobiose (Rowell et al., 2005) , a two monomer sugar (Figure 2). Hydrogen bonding occurs between cellulose molecules, resulting in parallel bundles of 36 cellulose chains called microfibrils. 8 OHH OHOHOOHOH OHOHOOHOH OHOHOOHOH OHOHOOHOH Figure 2. Partial structure of cellulose (Rowell et al., 2005) . The number of repeating sugar units is called degree of polymerization, and it is equal to the molecular weight of cellulose divided by the molecular weight of one molecular unit. Cellulose can be divided into crystalline and non- crystalline, accessible and non- accessible. As the packing density of cellulose increases, crystalline regions are form ed. The remaining portion has a lower packing density and is referred to as amorphous cellulose (non-crystalline). The layers of molecular chains are held together by weak van der Waals forces. The accessibility of cellulose refers to the availability of cellulose to water, microorganisms, enzymes, etc. The surfaces of crystalline cellulose are accessible but the core of the crystalline cellulose is non- accessible. Most non- crystalline cellulose is accessible but part of the non- crystalline cellulose is covered with both hemicelluloses and lignin (Rowell et al., 2005) . Rowell et al. (2005) presented a more thorough review on the structure and properties of cellulose that can be consulted for more information. 1.4.3 Hemicellulose Hemicellulose is the second most common polysacc haride in nature and represent s about 20–35% of lignocellulosic biomass. Hemicellulose is a mixture of polymerized monosaccharides such as glucose, mannose, galactose, xylose, arabinose, 4- O- methyl glucuronic acid and galacturonic acid residues. The structure of the principal component of the hemicellulose is illustrated in Figure 3 . Cellobiose 9 Figure 3 . Structure of xyloglucan in flowering plants ; principal component of the hemicelluloses. In blue backbone β - D- glucans; in red α- D- xylose; in black α- D- galactose and in brown α- L- fucose residues (Ochoa- villarreal et al., 2012). The polyssacharides contained in hemicelluloses are xylan, manna n, arabinan and galactan, which are composed of xylose, mannose, arabinose and galactose, respectively. Xylose is the predominant sugar in the hemicellulose of most hardwood. Arabinose can constitute a significant amount of the hemicelluloses in agricultu ral residues and herbaceous crops. The composition of two major agricultural residues, wheat straw and corn stover , are shown in Table 3. Table 3. Detailed compositional analysis of wheat straw and corn s tover A g r ic ult ur a l residue Co mpo sit io n ( w t % DM ) Ref e r e nc e Cellulo se He mic e llulo se To t a l lig nin Pro t e in Ash Xyla n Ara bina n Ga la c t a n Ma nna n Wheat straw 37.6 19.5 2.8 1.1 0.6 14.5 3.8 6.4 (Lee et al., 2007) Corn stover 37.5 21.7 2.7 1.6 0.6 18.9 4.7 6.3 10 Hemicelluloses crosslink s cellulose microfibrils together to form a network (Sivakumar et al., 2010). Hemicellulose is soluble in alkali and is easily hydrolyzed by acids . The detailed structure of hemicellulose has not been determined and only the ratios of sugars in this polysaccharide have been reported (Rowe ll et al., 2005) . 1.4.4 Lignin Lignin imparts rigidity to the cell walls and acts as a binder between plant cells, creating a composite material that is strongly resistant to compression, impact and bending. In addition, it also imparts resistance to biologic al degradation. Lignin is an aromatic polymer with a complex and diverse structure; monomer units appear to repeat randomly . Lignin is generally classified into three major groups based on their monomeric units: guaiacyl lignin in softwoods (gymnosperms), guaiacyl–syringyl lignin in hardwoods (angiosperms), and guaiacyl–syringyl –p- hydroxypheny l lignin in grasses (gramineae) (Lin and Lebo, 2001). The frequency of these groups may vary according to the morphologi cal location of lignin, plant species, and method of isolation. A possible lignin structure is shown in Figure 4 , w here lignin is a three- dimensional polymer connected by several acid- resistant C- C linkages. Ligni n is only partly degraded by oxidation and its behavior is in stark contrast with other native polymers such as protein, cellulose, hemicelluloses, and starch, which are generally hydrolyzed to monomeric units (Higuchi, 2006) . 11 Figure 4. Structural model for gymnosperm lignin. Linkages to additional lignin chains are indicated (L- containing circles) (Ruiz - Dueñas and Mar tínez, 2009) When lignocellulosic materials are used for the production of ethanol , the main by- product is lignin. Lignin can be used as a solid fuel for production of heat and/or electricity for which there are no foreseeable market limits (Galbe and Zacchi, 2002) . 1.4.5 Extractives Extractives are low molecular weight organic materials (ar omatics, terpenes or alcohols), some of which may be toxic to ethanol fermenting organi sms, or cause deposits in some pre -treatments. Extractives are generally soluble in neutral organic solvents or water. Some extractive compounds could be sold as chemicals (e.g. antioxidants) having a higher value than ethanol, but costs of purification ar e unknown. Extractives only represent around 4- 10 % 12 of the total weight of dry wood, and the extractive contents vary with plant species , geographical site and season (Ibrahim, 1998) . 1. 5 Lignocellulosic ethanol production p rocess Ethanol is currently produced from sugar cane and starch- containing materials (first generation), where the conversion of starch to ethanol includes a liquefaction step to solubilise starch and a hydrolysis step to produce glucose . The resulting glucose is then easily fermented. Although there a re similarities between the starch and the lignocellulosic process , the techno- economic challenges facing the former are significant (Hahn- Hägerdal et al., 2006) . Conversion of lignocellulose to ethanol is difficult due to: (1) the resistant na ture of biomass to breakdown due to lignin presence , (2) the variety of sugars which are released from hemicellulose and cellulose which make it necessity to find or genetically engineer organisms to efficien tly ferment these sugars, (3) the difficulty to efficiently breakdown cellulose and hemicellulose and (4) the costs for collection and storage of low density lignocellosic feedstocks (Balat et al., 2008). There are several opt ions for a lignocellulose to ethanol process, but regardless of which is chosen, the following features must be assessed in comparison with establ ished sugar- or starch- based ethanol production (Balat et al., 2008) . • Efficient depolymerization of cellulose and hemicellulose to soluble sugars. • Efficient fermentation of a mixed - sugar hydrolysate containing six- carbon (hexoses) and five - carbon (pentoses) sugars. • Advanced process integration to minimize energy demand. • Lower the lignin content of feedstock. One of the advantages of bioconversion with lignocellulose is the opportunity to create a biorefinery, producing valuable co- products in addition to bioethanol (Balat et al., 2008). In Figure 5 a flowsheet of one of the process configuration s proposed for the conversion of lignocellulosic materials to ethanol is presented . This process , named separate hydrolysis and fermentation (SHF), consists of four major unit operations: (1) pretreatment, (2) enzymatic 13 hydrolysis (conversion of polyssacharides to sugars b y the use of enzymes) , (3) fermentation (conversion of sugars to ethanol) , and (4) product separation/distillation. Figure 5. Schematic flowsheet for the conversion of biomass to ethanol. Another process conformation generally proposed is the combination of enzymatic hydrolysis of pretreated lignocelluloses and fermentation into one step, termed simultaneous saccharification and fermentation (SSF). In this conformation, glucose produced by enzymes is consumed immediately by the fermenting microorganism. This rapid consumption of sugars is advantageous as it reduces the inhibitory effects of cellobiose and glucose on enzymes by keeping the concentration of these sugars low . However, the difference between optimum temperatures for hydrolyzing enzymes and fermenting microorganisms limits the potential of SSF. The optimum temperature for the hydrolytic enzymes ( cellulases) is usually between 45 and 50°C, whereas Saccharomyces cerevisiae (fermenting microorganism) has an optimum temper ature between 30 and 35°C and is pract ically inactive above 40°C (Taherzadeh and Karimi, 2007) . The SSF media is more complex (e.g. addition of nutrients like such as carbon, oxygen, nitrogen, phosphor us, and sulfur ) than the media used for separate enzymatic hydrolysis (Olsson and Hahn- Hägerdal, 1996) . For this reason, the SHF configuration w ill use in this study to evaluate the possible recycling of enzymes after hydrolysis. As each unit operation is important to understand the entire process dynamics, in the next sections a brief explanation of each operation step is presented. PretreatmentEnzymatic hydrolysisFermentation SeparationBiomassEthanolWater, lignin, wastesLignin14 1.5.1 Biomass p retreatment Due to the robust and complex structure of lignocell ulosic biomass, pretreatment must be completed prior to enzymatic hydrolysis . The first step in the bioconversion of lignocellulosic ethanol production is biomass size reduction and pretre atment (Hahn-Hä gerdal et al., 2006) . Pretreatment plays a key role in the overall efficiency of the hydrolysis and fermentation steps since the purpose of pretreatment is to remove structural and compositional impediments to hydrolysis. With the appropriate pretreatment technology, the enzymatic hydrolysis rate and the yield of fermentable sugars will significantly increase (Mosier et al., 2005) . A schematic of pretreatment ’s effect on lignocel lulose is presented in Figure 6 . Figure 6. Schematic diagram of the pretratment effect on lignocellulose. Lignin impedes enzymatic hydrolysis of carbohydrates by blocking the access of enzymes to cellulose and hemicelluloses, and by binding enzymes . Cellulose crystallinity, accessible PretreatmentCellulose LigninHemicellulose15 surface area, the heterogeneous character of biomass particles, and cellulose sheathing by hemicelluloses contribute to the recalcitrance of lignocellulos ic biomass to hydrolysis (Mosier et al., 2005) . The choice of pretreatment technology also impacts how much of the biomass’ lignin, ash, and other components enter the solution and how they must be subsequently recovered (Yang and Wyman, 2008) . A successful pretreatment must exhibit the following features : (1) improve production of sugars during enzymatic hydrolysis, (2) avoid degradation or loss of carbohydrate, (3) avoid formation of inhibitors to hydrolysis and fermentation, and ( 4) be cost effective (Balat et al., 2008). A wide varity of pretreatment technologies have been proposed for the production of lignocellulosic ethanol, however, more research is still required to achieve more effective and commercially- viable technologies. In the next section, different pretreatment options and some of their characteristics are discussed. 184.108.40.206 Pretreatment methods Pretreatment can be carried out by physical, chemical, biological and physicochemical methods. The objective of physical or mechanical pretreatment s is to reduce biomass particle size, crystallinity, degree of polymerization and to increase its specific s urface area. This final aspect is key because it represents the area available for enzyme- substrate interactions and is influenced by pore size and the hemicellulose shielding effect (Laxman and Lachke, 2009) . In most cases, the power consumption needed to reach high digestibility during enzymatic hydrolysis is prohibitively high. The energy required can be even higher than the theoretical energy content available in biomass (Galbe and Zacchi, 2007) . Given the high energy requirements of milling and the low associated hydrolysis conversions obtained, it is likely that mechanical treatment by itself is not economically feasible, thus mechanical methods are often proposed in conjunction with other pretreatment methods (Alvira et al., 2010). 16 220.127.116.11.1 Chemical pretreatments Chemical pretreatments are generally effective in solubilising a large fraction of lignin while leaving behind much of the hemicellulose in an insoluble polymeric form and opening up the crystalline cellulosic substrate (Laxman and Lachke, 2009) . Alkaline pretreatment breaks the bonds between lignin and carbohydrates and disrupts the lignin structure, making the carbohydrates more accessible to enzymatic attack. As it acts mainly by delignification, it is more effective on agricultural residues and herbaceous crops than on wood materials, as the latter generally contains more lignin. Alkali pre treatment may be carried out at ambient conditions, but pretreatment time s are typically on the scale of hours or days rather than minutes or seconds (Balat et al., 2008). There are many types of acid ic pretreatments using sulphuric acid, hydrochloric acid, peracetic acid, nitric acid, or phosphoric acid. Dilute acid pre treatments at moderate temperatures using sulphuric or phosphoric acid can convert lignocellulosic biomass (including the hemicellulose fraction) to soluble sugars. The limitations of dilute sulfuric acid pretreatment include: corrosion necessitating the use of expensive materials of construction, neutralization of acid prior to enzymatic hydrolysis generates salts that must be disposed of, and formation of degradation products that inhibit fermentation . There are primarily two types of dilute acid pretreatment processes: a) low solids loa ding (5–10 wt% ) at high temperatures (T>160°C ) under continuous- flow and b) high solids loading (10–40 wt% ) at lower temperature (T<160°C ) under batch processes. Depending on the substrate and the conditions used, approximately 80 to 95% of the hemicellulo se can be recovered and 36 to 53% delignification can be achieved with this pretreatment (Balat et al., 2008; Torget et al., 1996) . Another approach is to use an organic or aqueous –organic solvent mixture. Numerous solvent mixtures can be utilized, including methanol, ethanol, acetone, ethylene glycol and tetrahydrofurfuryl alcohol, to solubilize lignin and generate pretreated lignoc ellulose materials suitable for enzymatic hydrolysis (Alvira et al., 2010) . This method breaks the internal lignin and hemicellulose bonds and separates lignin and hemicellul ose into fractions that can potentially be converted to useful products (Laxman and Lachke, 2009) . Compar ed 17 to other chemical pretreatments, the main advantage of these organosolv process es is the recovery of relatively pure lignin as a by - product. Removal of solvents from the system by using appropriate extraction and separation techniques, e.g., evaporation and condensation, is necessary. The solvents may be recycled to reduce operational costs. For economic reasons, of all the possible solvents, low - molecular weight alcohols with low boiling points such as ethanol and methanol are favored (Alvira et al., 2010). 18.104.22.168.2 Biological pretreatments Biological pretreatments employ microorganisms such as brown, white and soft - rot fungi which degrade lignin and hemicellulose and very little cellulose. Lignin degradation by white - rot fungi, the most effective biological pretreatment microo rganism, occurs through the action of lignin degrading enzymes such as peroxidases and laccases (Alvira et al., 2010). Phanerochaete chrysosporium, a white - rot fungus, is the most commonly used organism for delignification. This organism degrades 48.6% of lignin, 5.3% of cellulose, and 19.7% of hemicellulose in grape cluster stems over the course of 10 to 12 days (Laxman and Lachke, 2009) . Biological pretreatments are considered to be environmentally friendly and low energy as they are performed at low temperatures and need no chemicals. However, the rate of biological pretreatment processes is far too low for industrial use, and some material is lost as these microorganisms also consume hemicellulose and cellulose (Galbe and Zacchi, 2007) . 22.214.171.124. 3 Physicochemical pretreatment Physicochemical pretreatments combine physical and chemical methods. Some of the most important physicochemical pretreatments are flow through pretreatment or liquid hot water (LHW) , SO2- steam explosion , ammonia fiber/freeze exp ansion (AFEX) and ammonia recycling percolation (ARP). Liquid hot water process pass es water maintained in the liq uid state at elevated temperatures (190 to 200 oC) through biomass beds for 12 to 24 min. Between 4 % and 22% of the cellulose, 35% to 60% of the lignin and all of the hemicellulose 18 is removed (Mosier et al., 2005; Wyman et al., 2005) . The use of water in this pretreatment reduces the need for neutralization and conditioning chemicals since acid is not added. Monitoring and control of the pH of this system is required to minimize the hydrogen - ion concentration during pretreatment , prevent hydrolysis of cellulose and hemicellulose to oligosaccharides and monosaccharides, and avoid acid- catalyzed degradation of monosaccharides to degradation products (Weil et al., 1998) . SO2- steam explosion is a hydrothermal pretreatment in which biomass is subjected to SO 2 and pressurised steam for a period of time ranging from seconds to several minutes, and then suddenly depressurised. The addition of SO 2 or dilute acid to the steam explosion improve s enzymatic hydrolysis, decreases the production of inhibitory compounds, and lead s to more complete rem oval of hemicellulose. Al though lignin is not removed, it is thought that lignin properties are altered substantially. Weak acids such as acetic acid are generated during steam explosion from the acet yl groups present in hemicellulose while formic and levulinic acid are generated from degradation of furfural and hydroxymethylfurfural (HMF). Many different phenolic compounds are generated from lignin breakdown and the phenolics formed depend upon the raw materials (Alvira et al., 2010) . The necessary water wash after SO2- steam explosion to remove acid reduces overall sugar yields (Laxman and Lachke, 2009) . Limitations include destruction of a portion of the hemicellulose fraction, incomplete disruption of biomass structure, and the gener ation of inhibitory compounds. AFEX pretreatment is effective in improving cellulose digestion with the advantage of ammonia being recyclable due to its high volatility (Alvira et al., 2010). AFEX is a process in which ground, wetted lignocellulos e is placed in a pressure vessel with liquid ammonia (NH3 ) at a loading of about 1 –2 kg NH 3 /kg DM. Pressures exceeding 12 atm are required to operate at ambient temperature. The AFEX process pretreats lignocellulosic materials with liquid ammonia under pressure and then rapidly releases pressure. In this process cellulose is decrystallised, hemicellulose is prehydrolys ed, lignin is altered, the fiber structure is greatly disrupted, and the trace amounts of ammonia that remain can serve as a nitrogen source in the subsequent fermentation (Yang and Wyman, 2008) . Ammonia recovery is feasible due to its high volatility, however, the complexity and costs of ammonia recovery may limit the 19 commercial potential of AFEX pr etreatment. Improved understanding of the morphological changes and chemical compounds formed during AFEX may further improve process performance (Laxman and Lachke, 2009) . ARP pretreatment passes aqueous ammonia (5– 15 wt % of DM ) at elevated temperatures (80–180°C) through a packed bed of biomass. The outlet stream is then separated and ammonia is recycled in the effluent . Ammonia in aqueous solution and at high temperature breaks down lignin via the ammoniolysis reaction, but has virtually no effect on carbohydrates (Yang and Wyman, 2008) . A major challenge for ARP is to reduce liquid loadings to keep energy costs low . An alternative configuration to ARP, soaking in a queous ammonia (SAA), is being developed with this target in mind (Alvira et a l., 2010; Yang and Wyman, 2008) . Despite the wide range of pretreatment technologies proposed, research is still required to achieve more effective and commercially viable pretreatment methods (J ørgensen et al., 2007) . Most of the proposed pretreatments are under development and are not commercially available. Due to the large number of pretreatment technologies and substrate options, it is difficult to select the optimal technology for the eth anol production as each pretreatment has a different effect on biomass structure. The pretreatment technology effectiveness in the process has generally been evaluated in terms of lignin removal and improvements in enzymatic hydrolysis yields (Galbe and Zacchi, 2007; García - Cubero et al., 2009; Kim and Lee, 2007; Panagiotou and Olsson, 2007; Ruffell et al., 2010) . However, the effect of the pretreatment opera ting conditions on ethanol costs has not been thoroughly studied. 126.96.36.199.4 Oxygen Delignification Oxygen delignification, also known as wet oxidation, is an oxidative pretreatment method, which employs oxygen or air as catalyst. In the past two decades, this technology has 20 become an important part of modern pulp bleaching operations (Yang et al., 2003) . Oxygen delignification is used by the pulp and paper industry to remove lignin, which is one of the main pretreatment goals in the lignocellulosic ethanol industry . Oxygen delignification fractionates lignocellulose into a cellulose rich solid and a liquid rich in hemicelluloses and lignin. Additionally, oxygen delignification can easily be carried out as a continuous process providing advantages at an industrial scale (Schmidt and Thomsen, 1998) . The effectiveness of oxygen delignification depe nds on five operating parameters: retention time, reaction temperature, pulp consistency, alkali concentration, and oxygen partial pressure. Previous work has shown that an increase in any of these parameters will significantly enhance the degree of delign ification achieved. The oxidation is performed for 10–60 min at temperatures from 100 to 200°C and at pressures from 5 to 12 bar O 2. The addition of oxygen at temperatures above 170°C makes the process exothermic reducing the total energy demand (Gierer, 1997) . However, it has been reported by Charles et al. (2003) , that the increase in partial pressure of oxygen between 40 psig and 70 psig has a little effect on delignification. I ncreases in temperature from 125 to 165°C and in NaOH concentration from 1 to 3 g of NaOH/g DM , significantly improve lignin removal. During the first twenty minutes of the delignification, the reaction rate of delignification is high, however, between 20 and 60 min, the reaction rat e decrease and the lignin removal is dramatically reduced. Although the exact mechanism of lignin removal is not yet fully understand, Gierer (1997) , proposed that the degradation of lignin in an alkaline –oxygen medium a rises from the interactions of hydroxide and oxygen with the phenoxy hydroxyl group, as shown in Figure 7 . CH3OHOCH3NaOHCH3O-OCH3-e -CH3OOCH3O2+e -CH3OOCH3OO-+CH3OOCH3OO-+ OO-CH3OOCH3 Figure 7. Initial chemical reactions involved in oxygen delignification . 21 In this reaction, the phenolic hydroxyl group reacts with NaOH to form a phenolate ion, which then reacts with oxygen and superoxides to form the intermediate hydroperoxide. The hydroperoxide s have been proposed to form oxetanes , diooxiranes, muconic acid, and carbonyl structures that subsequently induce fragmentation of the aromatic ring of lignin and side chain scission as shown in Figure 8. The formation of the oxi ranes and muconic acids will introduce hydrophilic groups into the lignin structure, thus increasing its solubility in aqueous solutions . Such reactions have been shown to ultimately lead to the degradation of lignin to carboxylic acid, CO2 and water (Yang et al., 2003) . Although the chemistry of oxygen delignification has been attributed to phenolic units in lignin, non- phenolic structures have also been proposed to be involved in the process. The degradation of etherified phenolic units may begin with the benzylic oxidation of lignin resulting in the formation of α- carbonyl groups. The presence of an α - carbonyl group in the side chain of lignin has been shown to significantly increase the reactivity of non - phenolic units under oxygen del ignification conditions by inducing side chain cleavage reactions (Yang et al., 2003) . Figure 8. Superoxide induce the cleavage of carbon –carbon bonds (Gierer et al., 2001) 22 Gierer et al. (1997) proposed a reaction mechanism for the formation of the superoxides necessary for the oxidati ve cleavage of the phenolic rings present in lignin ( Figure 9) . These same superoxides are also thought to be responsible for the degradation of carbohydrate. Figure 9. Reduction of oxygen to water that occurs during oxygen delignification (Gierer, 1997) . Oxygen delignification has been proven to be an efficient method for solubilisation of hemicelluloses and lignin and to increase digestibility of cellulose. During oxygen delignification, the crystalline structure of cellulose is opened and hemicellulose is solubilised (Banerjee et al., 2009) . Oxygen delignification of wheat straw, sugarcane bagass e, corn stover, rice husk, softwood and clover - ryegrass mixture has been reported (Banerjee et al., 2009; Biswas et al., 2014; Charles et al., 2003; Klinke et al., 2002) . One of the advantage of this pretreatment is the low to non- existent production of f urfural and HMF , fermentation inhibitors, during the delignification of wheat straw (Bjerre et al., 1996; Klinke et al., 2003) , compared to steam explosion (490 and 80 mg/L of furfural and HMF, respectively , for poplar) (Oliva et al., 2003) and LHW methods (around 961 and 4,666 mg/L of furfural and HMF, respectively , for corn fiber) (Alvira et al., 2010; Dien et al., 2006) . In addition, Na2CO3 supplementation has shown to decr ease formation of inhibitory compounds by maintaining pH in the neutral to alkaline range (Alvira et al., 2010). It has been reported that lignin irreversibly adsorps cellul ases, decreasing enzymatic hydrolysis performance (Berlin et al., 2006; Nakagame et al., 2011; Pareek et al., 2013) . Due to the negative lignin effect on enzymatic hydrolysis, a pretreatment that can remove lignin or alter its structure may be advantageous for the process. It has been suggested that cellulases are adsorbed on lignin via hydrophobic interactions, ionic bond interactions, and hydrogen bond interactions (Nakagame et al., 2010) . Among these interactions, the 23 hydrophobic character of lignin has been identified as the major reason for the adsorption of cellulases. Oxygen delignification has been reported to be capable of achieving 65% degree of delignification with wheat straw and 96% recovery of cellulose (65% cellulose conversion during enzymatic hydrolysis ) with 70% of hemicelluloses removal (Palonen et al., 2004). Biswas et al. (2014) evaluated the influence of oxygen delignifi cation conditions on the enzymatic hydrolysis yields, reporting that the maximum theoretical conversion of cellulose, 87.4%, at the range of conditions evaluated was achieved at 185°C. It has been reported that the oxygen delignification increases the hydrophilic groups in the lignin structure, decreasing its hydrophob icity and augmenting its solubility in aqueous solution (Asgari et al., 1998; Gierer, 1985; Lee, 2011) . Therefore, it is expected that oxygen delignification not only reduces the amount of lignin in biomass but also reduce s the adsorption of cellulase on lignin by altering its structure. Increase in caustic loading, temperature or reaction time in oxygen delignification has a positive effect on the enzymatic hydrolysis of softwood where a kappa number (a measure of lignin content) decreased from 19 to 10, increased cellulose conversion from 69% to 92% (Charles et al., 2003) . Nonetheless, even when the increasing severity of the oxygen delignification seems to improve enzymatic hydrolysis conversion, its impact in the ethanol production cost is difficult to determine. Therefore, in the present work, the effect of mild and severe oxygen delignification conditions is determined considering the entire process economy. Oxygen delignification is used for pretreatment in the present study as it is a powerful technology to remove lignin from lignocellulose, produce s a small amount of in hibitory compounds , and is already used at industrial scales (Alberta- Pacific (AL - PAC) Forest Industries, 1995; FlowTec Industrietechnik GmbH, 2013; Linde group, 2013) . 1.5.2 Enzymatic h ydrolysis The second step of ethanol pro duction is hydrolysis, during which cellulose is converted into fermentable glucose ((C6 H10O5 )n+nH 2O→nC 6 H12O6 ). This reaction is catalysed by acid or 24 enzymes. In the case of enzymatic hydrolysis without preceding pretreatment, sugar yields <20% are obtained, whereas yields after pretreatment often exceed 90% (Hamelinck et al., 2005a) . There are several advantages and disadvantages of acid and enzymatic hydrolyses. Enzymatic hydrolysis is carried out under mild conditions, whereas acid hydrolysis requires high temperature and low pH, which results in corrosive conditions. While it is possible to obtain cellulose hydrolysis close to 100% by enzymatic hydrolysis, it is difficult to achieve such high yields with acid hydrolysi s. Furthermore, several inhibitory compounds are formed during acid hydrolysis, while the inhibitor production in enzymatic hydrolysis is very low to non- existing (Taherzadeh and Karimi, 2007) . However , enzymatic hydrolysis requires hours or days while acid hydrolysis can be completed within minutes. The price of enzymes is much higher than sulfuric acid used for acid hydrolysis. Finally, sugar production inhibits enzymatic hydrolysis . In order to overcome this problem, SSF has been developed, in which the sugars released from the hydrolysis are directly consumed by the present microorganisms. As mentioned earlier, since fermentation and hydrolysis usually have different optimum temperatures , SHF is still considered a viable option (Taherzadeh and Karimi, 2007) . Enzymatic hydrolysis is compatible with many pretreatment options. Many experts see enzymatic hydrolysis as key to cost - effective ethanol production in the long run. Though acid processes are more technically mature, enzymatic p rocesses have comparable projected costs and the potential for cost reductions as technology improves (Hamelinck et al., 2005a) . For these reasons, enzymatic hydrolysis is the most popular method of hydrolysis for bioethanol production research and was therefore investigated in the present project. Enzymatic hydrolysis of lignocellu losic materials is a slow process because cellulose hydrolysis is hindered by structural parameters, such surface area and cellulose crystallinity (Balat et al., 2008). The enzymatic degradation of solid cellulose is a complicated process that takes place at a solid - liquid phase boundary, where enzymes are the mobile components. Generally, enzymatic cellulose degradation is characterized by a continuous ly decreasing rate. This has been explained most often by the rapid hydrolysis of the readily accessible 25 fraction of cellulose, followed by significant product inhibition, and slow inactivation of absorbed enzyme molecules (Balat et al., 2008) . According to the traditional enzyme classification system, cellulase s (cellulolytic enzymes) are divided into three classes; endo- 1,4- β - D- glucanases (EG) (EC 188.8.131.52), which hydrolyse internal β- 1,4- glucosidic bonds randomly in the cellulose chain; exo - 1,4- β - D- glucanases or cellobiohydrolases (CBH) (EC 184.108.40.206), which move progressively along the cellulose chain and cleave off cellobios e units from the ends; and 1,4- β - D- glucosidases (B) (EC 220.127.116.11), which hydrolyse cellobiose to glucose and also cleave glucose units from cellooligosaccharides. In Figure 10, a simplified mechanism for the hydrol ysis of cellulose by endo- glucanases, exo- glucanases and β - glucosidases is presented. All the enzymes work synergistically to hydrolyse cellulose by creating new accessible sites, removing obstacles and relieving product inhibition (Jørgensen et al., 2007) . 26 Figure 10. General mechanism of the enzymatic hydrolysis of cellulose by endo- glucanases, exo- glucanases and β - glucosidase (Divne et al., 1998; Helmich et al. to be publish, Kleywegt et al., 1997) . Hemicellulose consists of a mixture of polymerized monosaccharide s chains with diverse side groups and c onsequently , the hemicellulolytic enzyme system is more complex . The hemicellulase system includes: endo- 1,4- β - D- xylanases (EC 18.104.22.168), which hydrolyse internal bonds in the xylan chain; 1,4- β - D- xylosidases (EC 22.214.171.124), which attack xylooligosaccharides from the non- reducing end and liberate xylose; endo - 1,4- β - D- mannanases (EC 126.96.36.199), which clea ve internal bonds in mannan and 1,4- β - D-mannosidases (EC 188.8.131.52), which cleave mannooligosaccharides to mannose. The side groups ar e removed by a number of enzyme: α- D- galactosidases (EC 184.108.40.206), α- L-arabinofuranosidases (EC 220.127.116.11), α - glucuronidases (EC 18.104.22.168), acetyl xylan esterases 27 (EC 22.214.171.124) , and feruloyl and p- cumaric acid esterases (EC 126.96.36.199) (Jørgensen et al., 2007) . The mechanism of enzymatic cellulose hydrolysis is not yet well understood. The complexity of the system, which arises from the concerted action of several cellulolytic enzymes acting on a heterogeneous solid substrate, makes experimental kinetic and mechanistic studies difficult. As a result, the rate determinin g features of the cellulase–cellulose system have yet to be identified, and the physical and chemical details of the enzyme –substrate interaction remain to be elucidated. The hydrolysis of cellulosic materials depends on the presence of endo- glucanases, exo-glucanases and β - glucosidase in proper ratios . If any one of these enzymes is present in less than the required amount, the others will be inhibited or will lack sufficient substrates to act upon. The hydrolysis rate increa ses with increasing temperatur e, however, since the catalytic activity of enzymes is related to its s tructure, deformation of the enzyme structure at high temperature can inactivate or destroy the enzyme. To strike a balance between increased activity and increased deactivation, it is preferable to conduct enzymatic hydrolysis at approximately 40 to 50° C (Lee, 2007) . Cellulases dosages of 10–30 FPU/g cellulose are often used in laboratory studies because they result in high glucose yields in a reasonable time (48–72 h) (Talebnia et al., 2010). Enzy me loading varies with pretreatment type and conditions, and concentration of raw materials . In Figure 11, the effect of enzyme loading on hydrolysis efficiency is illustrated by the results presented by Lu et al. (2002). It can be seen that hydrolysis yield increases with increasing enzyme loading, however, due to high enzyme costs, high enzyme loadings may not be viable . Enzyme loading ha s been frequently optimized by minimiz ing enzyme loading for maximum sugar production (Kumar and Wyman, 2009a; Shi et al., 2011) . However , these studies have been carried out at a single pretreatment conditions and a constant solid concentration with a constant reaction time and thus the number and size of hydrolytic reactors and associated capital cost have not been considered as part of the optimization . In the present study, the enzymat ic hydrolysis at different conditions is studied to understand the effect of hydrolytic variables on the production cost of ethanol. 28 Figure 11. Hydrolysis profiles of steam - exploded Douglas - fir over time at different enzyme loadings (FPU/g cellulose), 2 wt % DM , and 45°C (Lu et al., 2002) . β - glucosidase is, to some extent, inhibited by glucose, in consequence, the inhibition of β -glucosidase results in accumulation of cellobiose which is a potent inhibitor of the cellobiohydrolases (CBH). The competitive product inhibition of CBH can be partially overcome by addition of an excess of β - glucosidase (Jørgensen et al., 2007) . Some pretreatment products inhibit the enzyme system: cellulases are inhibited by formic acid, whereas compounds li ke vanillic acid, syringic acid, syringylaldehyde and formic acid cause significant inhibition of xylanases (Jørgensen et al., 2007) . Lignin is a major obstacle in achieving efficient hydrolysis as it shields cellulose chains and irreversibly adsorbs cellulases (Jørgensen et al., 2007; Palonen et al., 2004) . The effect s of lignin on enzymatic hydrolysis were demonstrated by Nakagame et al. (2011); Nakagame's study found that addition of isolated lignin to the hydrolysis of pretreated Avicel decreased cellulose conversion by as much as 49% . It was reported in an early study that enzyme cost represented about 19% to 23% of the total ethanol production cost , while pretreatment and fermentation account for 25 and 10% of the production cost respectively (Gregg et al., 1998) . Recent estimates for the cost contribution of enzymes range from $0.10/gal of ethanol to $0.40/gal of ethanol (Klein -Marcuschamer et al., 2012). The major enzyme manufacturers Genencor and Novozymes announced new commercial - grade cellulase cocktail s with an enzyme cost contribution of approximately $0.50/gal ethanol (Humbird et al., 2011) . However , Klein - Marcuschamer et al. (2012) estimated much higher enzyme costs of $0.68/gal of ethanol, assuming maximum 29 theoretical hydrolysis and fermentation yields, and $1.47/gal of ethanol, using saccharification and fermentation yields reported in the literature. These estimates all emphasize the need to reduc e enzyme cost s and usage in hydrolysis. There are several methods of decreas ing cellulases expenses : (i) decreasing enzyme loading (i.e., gram of enzyme used per gram of cellulose) by increasing reactivity of pretreated biomass, optimizing hydrolysis conditions and/or by recycling enzymes, (ii) increasing cellulase performance (unit per gram of cellulase) through cellulase engineering, and (iii) decreasing cellulase production costs (dollar per gram of cellulase) (Zhu et al., 2009) . Given the importance of enzymatic hydrolysis to process cost, hydrolysis optimization and cost reduction are the primar y focus of the present study. Hydrolysis cost r eduction was studied following two strategies: identifying the dominant hydrolysis variables to select the most advantageous operating conditions and determining whether enzyme recycling could significantly reduced the cost of ethanol from lignocellulose . Several techno- economic analyses have provided valuable information regarding the influence of enzyme and feedstock costs on ethanol production (Anex et al., 2010; Huang et al., 2009; Humbird et al., 2011; Kazi et al., 2010; Klein- Marcuschamer et al., 2012; Wingren et al., 2003) . However, most of these studies were developed using specific delignification and sugar yields for specific conditions, making it difficult to assess the impact of individual pretreatment and hydrolysis variables. Changes to one stage propagate through the entire process, therefore, a model which can describe the impact of time, enzyme loading and lignin content on glucose production during hydrolysis and can be incorporated into simulation software, is needed . Integrating such a model into process simulations would support the economic evaluation, detailed sensitivity analyses and optimization of the ethanol process (Bansal et al., 2009) . For the industrial production of lignocellulosic ethanol, it has been proposed that enzymatic hydrolysis should be conducted at solid concentration greater than 10 wt % because these conditions have the potential to improve process economics by increasing product concentration and decreasing capital costs (Hodge et al., 2008; Huang et al., 2009; Sassner et 30 al., 2008). However, operating hydrolysis at solid concentrations above 10–15 wt % has been technically difficult, especially at laboratory scale. The initial viscosity of the material at these concentrations is very high, making mixing difficult and inadequ ate. Consequently, mixing power consumption in stirred tank reactors becomes high. In pilot scale plants, 15–20 wt % DM has often been reported as the maximum that can be handled (Jørgensen et al., 2007) . Special reactor s that can handle a high solid concentration are being investigated in order to carry out enzymatic hydrolysis at industrial scales (National Renewable Energy Laboratory (NREL), 2011). Hydrolysis conversion decreases with solids concentration, as reported for the hydrolysis of filter paper and corn stover (Hodge et al., 2008; Kristensen et al., 2009) . Therefore, although enzymatic hydrolysis at high solid concentration decrease s the hydrolysis capital cost, it also decreases ethanol production. An alternate strategy to reduce enzymatic hydrolysis costs is discussed in the next section. 1.5.3 Enzyme rec ycling One of the most promising methods to decrease enzyme loading is enzyme recycling. Enzyme recycling has been proposed to be performed after hydrolysis , fermentation, simultaneous saccharification and fermentation and distillation. Recycling after hy drolysis is advantageous due the relatively fresh enzymes recycled. Enzymes can be recovered after SHF or SSF, where enzymes inhibition is reduced due to the consumption of sugars during fermentation, however, the produced ethanol also deactivate enzymes. The recycling of enzymes after distillation is possible when distillation is performed at low temperatures under vacuum (Lindedam et al., 2013) . Enzymes have been shown to remain remarkably active after lignocellulosic hydrolysis, which opens the possibility of recy cling enzyme s to reduce enzyme costs (Tu and Saddler, 2010). It has been shown that enzymes can be recovered from both the supernatant and the hydrolyzed residue after hydrolysis. The possibility of recovering enzymes by reusing the insoluble residues after hydrolysis has been studied by different researchers. By recycling the insoluble residues, the production of sugars in the second hydrolysis round increased by 20% with respect to th e first round (Weiss et al., 2013) . The hydrolysis yield was found to decay 31 exponentially with each recycling round. In Weiss’ study , the recycling of the solid residues was analysed by model the impact of the amount of solids recycled, wash of solids and enzyme supplementation on the hydrolysis yields. The fraction of solids recycled and makeup of enzyme were reported to have the major impact on glucose yield. Under this methodology, the solids concentration and lignin content increases with each round of recycling leading to the deactivation of the enzyme and to an increase in reactor volume. Jin et al. (2012) have proposed a new configurati on which may solve the accumulation of solids by recycling the solids residues in a reactor cascade train. However, more research is needed to evaluate the negative effect of lignin in the reaction and the benefits of this configuration in a continuous pro cess. There are two methods for recycling enzymes from the supernatant . The first is by ultrafiltration, which is an effective way to recover enzymes and to continuously remove products that could potentially inhibit hydrolysis reactions (Lu et al., 2002) . One study on the economics of ultrafiltration for ethanol production from corn stover using AFEX pretreatment reported cost savings of approximately 15% (Steele et al., 2005) . In Steele’s work, e nzyme recycling was determined based on enzyme activity, where the enzyme activity recycled was expressed as a ratio of enzyme activity after recovery to initial activity. No further information about the mass balance or the assumptions made to use activity to determine the amount of enzyme recycled , such as the difference between EG, CHB and B individual and combined activity, were provided. In the study by Steele et al. (2005) , the pretreatmen t and hydrolysis stages wer e simulated in AspenPlus based on experimental results (60% enzym e activity recycled and 15 FPU/ g of cellulose enzyme loading). The simulation was used to estimate the production cost of ethanol for a 50 million ga l ethanol/ year plant with and without enzyme recycling by ultrafiltration. From this study, ultrafiltration appears to be a viable option for recovering cellulase after hydrolysis; however, the method is limited by its cost and the eventual fouling of the filter membrane . The second option is recycling by adsorption, where endo- and exo- glucanases (from now on referred to as cellulases) can be recycled by adsorbing free cellulases onto newly - added substrates after hydrolysis (Zhu et al., 2009) . During enzymatic hydrolysis of cellulose, 32 cellulases adsorb on substrate and then gradually desorb as hydrolysis progresses. Some cellulases remain free in the solution (free cellulases), while others are bound to the residual substrates (both cellulose and lignin) (Tu et al., 2009b) . By exploiting the natural affinity of cellulases for cellulose, enzymes in the liquid phase are effectively recovered by adsorption onto fresh added substrates. As shown in Figure 12, the solid and liquid phases are separated after hydrolysis ; the liquid phase contains sugar s, cellulases and β - glucosidase. When f resh substrate is added to hydrolytic liquor, cellulases are adsorbed onto its surfaces. Once cellulases are adsorbed, the substrate is separated from the liquid phase to start a new hydrolysis round. The hydrolytic liquor containing unadsorbed enzymes and sugars is used for fermentation. Figure 12. Enzyme recycling experimental diagram. In past enzyme recovery studies, adsorption processes were conducted at low temperatures (4- 25°C for 1.5 t o 2.5 h) (Qi et al., 2011; Tu et al., 2007a; Tu et al., 2007b; Tu et al., 2009b; Tu and Saddler, 2010) to limit cellulase activity and study cellulase adsorption in the absence of the hydrolysis reaction. The Langmuir model was used to predict the adsorption of cellulase to softwoods with reasonable accuracy at low temperatures (4° C) (Lu et al., 2002; Tu et al., 2007b) . However, the adsorption process is most likely to be carried out at hydrolytic temperature at industrial scales in order to reduce energy costs. Therefore, in order to simulate the enzyme recycl ing process at an industrial scale, enzyme adsorption on fresh substrate was evaluated at 50°C in this study. 33 The potential for cellulase s to be recycled by adsorption and subsequently used to produce high cellulose conversions (65% to 93%) has been demonstrated by several researchers for different substrates (Lu et al., 2002; Steele et al., 2005; Tu et al., 2007a) . The different conversions obtained in these works are believed to be due to the differences in pretreatment technologies, substrates, and hydrolysis conditions used. For example , when steam- exploded Douglas- fir pre treated with hot alkali peroxide (8% lignin content) was hydrolysed with the preparation Celluclast 1.5 L at 20 FPU/g cellulose, for 24 h at 45°C and 2 wt % solids concentration, 93% cellulose conversion was obtained in the second hydrolysis round by recycling the enzymes (Lu et al., 2002). In another study, an ethanol- pretreated mixed softwood (6 % lignin content) was hydrolyzed with Celluclast 1.5L at 20 FPU/g cellulose, for 24 h at 45°C and 2 wt % solid s concentration, 70% cellulose conversion was obtained in the second hydrolysis round (Tu et al., 2007a) . Enzyme recovery from pretreated wheat straw was studied by Qi et al. (2011). In this study, enzyme recovery by adsorption was used to recycle enzyme from the liquid phase while enzymes in solid phase were recovered by washing the solid residue s with buffer to desorb and reuse enzymes. The cellulose and xylan hydrolysis yields reported for the second hydrolysis round were 90% as shown in Figure 13. However, in subsequent rounds, the hydrolysis yields (50% and 20%) decreased , possibly due to loss of cellulases by irreversible adsorption to lignin. From a process operating perspective a successful enzyme recycling process should aim to produce a uniform flow of sugars but to date, this has not been achieved. 34 Figure 13. Hydrolysis yield achieved in each recycling round by adsorption for the hydrolysis of alkali pretreated wheat straw at 2 wt % of DM solids concentration, 20 FPU/g cellulose, 150 rpm and 50°C using a commercial enzyme preparation from Su nson Group Ningxia (Qi et al., 2011). It is difficult to evaluate the amount of cellulases that can be recycled as only total enzyme concentration, the sum of cellulases and β - glucosidase concentration, can be measured. Since cellulases adsorb onto cellulose and lignin while β - glucosidase does not adsorb onto the substrate (Tu et al., 2007a; Tu and Saddler, 2010) , many researchers (Lu et al., 2002; Qi et al., 2011; Tu et al., 2007b; Tu et al., 2009) have assumed that any change in the total protein concentration in the liquid phase during hydrolysis is due to the adsorption of cellulases on the substrate. This implicitly assumes that total protein concentration in solution is constant at hydrolysis conditions (Lu et al., 2002; Qi et al., 2011; Tu et al., 2007b; Tu, Pan, & Saddler, 2009) . Consequently, the reported enzyme recovery (equation 48, pp 109) has been calculated from the difference in total protein concentration in solution before and after adsorption of enzymes onto fresh substrate. Based on these assumptions, enzyme recovery is equal to the amount of cellulases re covered. Using these assumptions , Tu et al. (2007a) reported an enzyme recovery of 85% when using surfactants to improve enzyme desorption from ethanol pretreated lodgepole pine . Qi et al. (2011) reported an enzyme recovery of 75% when rec overing enzymes from the liquid phase by adsorption onto alkali pretreated wheat straw . However, protein concentrations of widely used enzyme cocktails at hydrolysis conditions have not been demonstrated to be constant. If protein concentration is 0 20 40 60 80 100 R0 R1 R2 R3 Hydrolysis yield (%) Enzyme recycling rounds 35 not constant, the previously reported enzyme recoveries are inaccurate. Given its importance, changes in protein concentration at hydrolysis conditions were studied in the present project. Enzyme recycling performance is generally reported as a function of the en zyme activity remaining after recycling . Lu et al. (2002) reported that recycled enzymes retained 78 and 47% of their initial activity in the second and third rounds of hydrolysis of steam - exploded Douglas- fir extracted by hot alkali peroxide . The decrease in enzyme activity may be caused by irreversible cellulases adsorption on lignin or by enzyme deactivation due to high temperature or sugar - enzyme interactions. In other reports, enzyme recycling was evaluated by hydrolysis yields obtained in subsequent hydrolysis rounds . Tu et al. (2007b) reported a hydrolysis yield of 60% in the second hydrolysis round using steam exploded lodgepole pine ; the reduction in yield is likely due to the loss of activity observed by Lu et al. (2002). Tu and Saddler (2010) have also shown that hydrolysis yields decrease for each round of recycling : after four recycling rounds, yields from steam - exploded lodgepole pine and ethanol-pretreated lodgepole pine decreased from 100% to 10%, and from 100 to 80% , respectively. One of the key differences between steam - exploded lodgepole pine and the ethanol -pretreated lodgepole pine was the lignin content, which was 46% and 15% , respectively. The amount of cellulase recycled depends on substrate lignin content (Jørgensen et al., 2007; Lu et al., 2002). However, while lignin content may dictate hydrolysis yield, the fundamental differences between pretreatments and their effect on lignin structure may also cause the observed differences. Lu et al. (2002) showed that lignin is one of the major obstacles to both hydrolysis and enzyme recycling. Cellulases remain adsorbed on the residual material, rich in lignin, reducing the amount of cellulases available for recycling . The cellulose- binding domain (CBD) in cellulases is hydrophobic and probably participates in the lignin binding (Berlin et al., 2005) . Also, it has been demonstrated that there are lignin- binding sites on the cellulases catalytic domain (CCD). The CCD may significant ly contribute to overall cellulase- lignin binding. Consequently, weak lignin - binding cellulases might be engineered without compromising cellulose binding (Berlin et al., 2005) . Given this information, i t is clear that the amount of cellulase s recycled by adsorption depends on the lignin content of the substrate 36 (Jørgensen et al., 2007) . Enzyme recycling has been studied at a single lignin compos ition (Tu et al., 2007a; Tu et al., 2009b) , which does not allow the study of the impact of lignin on enzyme recycling. In other studies, enzyme recovery has been studied for substrates pretreated with different technologies, resulting in substrates with different lignin compositions (Kristensen et al., 2007; Lu et al., 2002) . However , as each pretreatment has a unique effect on lignin structure, it is still difficult to analyze the lignin content effect on enzyme recovery. The economic viability of enzyme recovery by adsorption in the ethanol production has not been demonstrated. It has been reported that enzyme recovery by adsorption may reduce enzyme expenses by 80% (Tu and Saddler, 2010) . Nonetheless, in order to accurately evaluate the viability of this technology, the sugar losses, capital and operational costs for enzyme recovery need to be taken into account along with the other process stages. To evaluate these factors a complete process simulation and economic analysis is needed. Xue et al. (2012) have shown that by recycling enzymes and supplementing the same enzyme loading in all hydrolysis, the hydrolysis yields increase in each recycling round by the accumulation of enzymes. However, a uniform production of sugars during the enzyme recycling process has to be achieved by supplementing cellulases to make up the same initial loading considering the recycled enzymes . In this way, the enzyme requirements can be reduced by producing the same amount of sugars. If this goal is achieved, the economic viability of the enzyme recycling process can be evaluated. In recent years, the enzyme recycling after distillation has gain attention. Lindedam et al. (2013) studied enzyme recycling after SSF, SHF and distillation configura tion at laboratory and pilot - plant scale using wheat straw . In this process, the broth after hydrolysis and fermentation is used for new hydrolysis rounds. Lindedam showed that the recovered enzyme activity after hydrolysis and fermentation is greater when first hydrolysis was conducted at low temperatures or short residence times . The enzyme activity in the whole slurry, using four enzyme preparations, after SHF was between 15 to 40% at 50°C and after SSF was between 15 to 60% at 50°C in the laboratory. These activities were achieved by recycling the whole slurry after fermentation, which represents a problem because ethanol is 37 being completely recycled. By recycling only the supernatant, enzyme activity after SHF and SSF was around 0 to 20% of the initial activity . This shows that the enzymes remind active after fermentation. The endo- glucanases and exo- glucanases activity potentially available for recycling in the whole slurry after SSF, in a pilot plant working at 25 wt % DM , was approximately 40 to 50% of the initial activity. Nonetheless, the distillation step (55°C under vacuum) to separate ethanol from the enzymes broth decrease d enzyme activity by 5 to 50% , depending on the enzyme preparation. Eckard et al. (2013) reported that the activity of enzymes available for recycle increased after hydrolysis and fermentation by the use of surfactants or casein, which increase the desorption of cellulases and the hydrolysis yields. Howe ver, the use of these compounds can also significantly increase the ethanol cost. One of the goals of this project is to develop a mass balance and simulation of the enzyme recycling process. Due to the simpler composition of the hydrolytic liquor relative to the fermentation broth, the enzyme recycling by adsorption after hydrolysis was the focus of this thesis 1.5.4 Fermentation of sugars for the production of ethanol An important limitation in the commercialization of lignocellulosic ethanol is the lack of microorganisms capable of efficiently fermenting both pentose and hexose sugars with high yields and high rates. For commercial ethanol production, the ideal microorganism should have a high ethanol yield and productivity, tolerance to inhibitors and high c oncentrations of ethanol, and the ability to ferment sugar s at high temperatures (Talebnia et al., 2010). Bacteria, yeast, and fungi ferment carbohydrates to ethanol under oxygen free conditions . According to reactions 1 and 2, the theoretical maximum yield of fermentation is 0.51 kg ethanol and 0.49 kg carbon dioxide per kg sugar (Hamelinck et al., 2005a) : 3𝐶5𝐻10𝑂5 → 5𝐶2𝐻5𝑂𝐻 + 5𝐶𝑂2 1 𝐶6𝐻12𝑂6 → 2𝐶2𝐻5𝑂𝐻 + 2𝐶𝑂2 2 The best known microorganisms for ethanol production from hexoses are the yeast Saccharomyces cerevisiae and the bacterium Zymomonas mobilis which offer high ethanol yields (90–97 % of the theoretical) (Talebnia et al., 2010). However, the native strains of S. 38 cerevisiae and Z. mobilis do not ferment xylose. Other microorganisms known to ferment xylose to ethanol, such as enteric bacteria and the yeasts Pichia stipitis, Candida sheha tae, and Pachysolen tannophilus are characterized by low ethanol yields (Chandel et al., 2007; Lin and Tanaka, 2006) . Genetic engineering and new screening technologies have yielded bacteria and yeast capable of fermenting both glucose and xylose, although fermentation of mannose, galactose and arabinose remain problematic (Hamelinck et al., 2005a) . The fermentation of all five major sugars produced during enzymatic hydrolysis (glucose, xylose, mannose, galactose and arabinose) may be possible in the near term using genetically -engineered yeast or bacteria. Referent to this thesis, the fermentation of sugars from hydrolyzed wheat straw has been widely studied with yeasts, bacteria and fungi, usually grown as pure cultures . P. stipitis (Nigam, 2001), Kluyveromyces marxianus (Tomás - Pejó et al., 2009) and recombinant strains of S. cerevisiae (Panagiotou and Olsson, 2007) , are the most widely studied yeasts f or ethanol fermentation based on wheat straw hydrolyzed as feedstock. 1.5.5 Ethanol separation and concentration stage The separation step is considered to be technically mature and its operational performance is well know n. A schematic of the distillation p rocess frequently proposed for the production of lignocellulosic ethanol is shown in Figure 14. Ethanol distillation is accomplished in two columns where the first column, called the beer column, removes most of the dissolved CO 2 and water. The second column, the rectification column, concentrates the ethanol to a near azeotropic composition, 95.6 wt% ethanol at 1 atm (Humbird et al., 2011). To produce anhydrous ethanol (99.5 wt % ), water is r emoved with a molecular sieve adsorption unit. Molecular sieves are materials composed of microporous substances t hat are characterized by their ability to retain defined types of chemical species on their surface. These materials packed into a vessel mak e it possible to separate ethanol from ethanol - water mixtures by adsorption mechanisms at high pressure (Bastidas and Gil, 2010) . The pressure 39 swing adsorption (PSA) process has succeeded on industrial scale to produce anhydrous ethanol and therefore, it was selected to carry out ethanol separation in the present project. Beer columnRectification columnMolecular Sieve AdsorptionFermentationLignin and waterEthanol product99.5% (wt)WaterCO2 Figure 14. Schematic diagram of the ethanol separation process (Aden et al., 2002; Humbird et al., 2011; Kazi et al., 2010; Kumar and Murthy, 2011) . 1. 6 Economic studies for the production of lignocellulosic ethanol In order to look at the potential overall benefits of the ligno cellulosic ethanol pr ocess, it is very important to perform a techno- economic analysis of the whole process. To focus research efforts on the most influential steps , the process must be modeled at an appropriate level of detail using relevant and consistent assumptions (Gnansounou and Dauriat, 2010). Published lignocellulosic ethanol techno- economic analyses often simplify the process stages and variables using a small number of pathways for a narrow range of feedstock s. In addition, not all of the assumptions are explicitly stated which makes it extremely difficult to compare different techno- economic evaluations. Finally, in most published studies, information about a number of economic factors such as project life, interest payment, internal return ratio, and raw material cost varied as part of sensitivity analyses are limited. The ethanol production costs obtained in past economic analysis range from $0.64 to $1.94/ kg ethanol (Humbird et al., 2011; Kazi et al., 2010; 40 Klein - Marcuschamer et al., 2010; Kumar and Murthy, 2011; Piccolo a nd Bezzo, 2009; Sassner et al., 2008; Tao et al., 2011; Wingren et al., 2003; Wingren et al., 2005) . The differences in estimated ethanol costs are caused by differences in assumptions and process configurations , as shown in Table 4. Table 4. Characteristics of various economic analyses. Fe ed s tock Pr et reat men t Simu l ati on platf orm Eth an ol price ($/k g eth an ol ) Ref eren ce Corn stover DA SuperPro Designer 1.53 (Klein - Marcuschamer et al., 2010) Grass straw DA, dilute alkali, LHW, steam explosion Super Pro Designer 1.03 to 1.27 (Kumar and Murthy, 2011) Corn stover DA AspenPlus 0.72 (Humbird et al., 2011) Hardwood DA AspenPlus 1.11 to 2.05 (Piccolo and Bezzo, 2009) Salix, corn stover, spruce SP AspenPlus 1.11 (Sassner et al., 2008) Switchgrass AFEX, DA, LHW, Lime, SAA, SP AspenPlus 0.78 to 1.36 (Tao et al., 2011) Spruce SP AspenPlus 0.78 (Wingren et al., 2005) Spruce SP AspenPlus 0.64 to 0.79 (Wingren et al., 2003) Corn Stover AFEX, DA, LHW AspenPlus 1.14 to 1.94 (Ka zi et al., 2010) The hydrolysis residence time, enzyme loading, and solids concentration during the enzymatic hydrolysis influence not only hydrolysis yield but also all of the downstream stages. Thus the impact of these variables on the final ethanol c ost is difficult to evaluate without considering the entire production process. Although several techno- economic analyses have provided valuable information regarding the influence of enzyme and feedstock cost on ethanol production (Anex et al., 2010; Hua ng et al., 2009; Humbird et al., 2011; Kazi et al., 2010; Klein- Marcuschamer et al., 2012; Wingren et al., 2003) , most 41 economic studies have been developed assuming specific sugar yields obtained at specific hydrolysis conditions, complicating efforts to assess the impact of individual variables on the overall process (Aden et al., 2002; Huang et al., 2009; Kazi et al., 2010; Sassner et al., 2008) . Given the important effect s of pretreatment and hydrolysis on the downstream stages , a model that links these two stages and models hydrolysis at different conditions can be an important tool to evaluate the economics of lignocellulosic ethanol production (Zhang et al., 2010). Integrating such a model into process simulations woul d integrate the hydrolytic time into the economic study, instead of limiting analysis to specific hydrolytic times (generally 36 and 96 h) as in previous studies (Aden et al., 2002; Huang et al., 2009; Kazi et al., 2010; Wingren et al., 2003) . Therefore, in the present study, ethanol production from different hydrolytic conditions will be studied to understand the contributions of hydrolytic variables to ethanol cost. Enzyme cost has been reported to be one of the major barriers for the lignocellulosic ethanol commercialization (Kristensen et al., 2007) . However, the importance of enzyme cost in the ethanol economy has not been reflected in the techno- economic analyses reported. This is believed to be caused by an optimistic enzyme cost assumed in these studies. Enzyme cost contribution to ethanol production cost of $0.10 to $0.34/gal of ethanol have been assumed (Aden et al., 2002; Galbe et al., 2007; Humbird et al., 2011; Sassner et al., 2008; Wingren et al., 2003) . However , Klein - Marcuschamer et al. (2012) have reported that the normally assumed enzyme cost contributions are comparable to soy protein production costs, a commonly and inexpensively produced protein. In consequence, the contribution of enzyme expenses to ethanol production cost is expected to be greater tha n previously reported . As mentioned previously , the economic feasibility of enzyme recovery by adsorption has not been thoroughly examined. Tu and Saddler (2010) estimated the cost of enzymes with recycling by dividing the cost of cellulase enzymes ($0.50/gal ethanol) by the number of hydrolysis rounds possib le with recycling. Based on this calculation, Tu and Saddler (2010) reported that enzyme recovery by adsorption may reduce enzyme expenses by 80%. Nonetheless, to accurately evaluate the viability of this technology, the sugar loss, equipment and operational costs associated with implementation of enzyme recovery must be accounted 42 for. A complete techno - economic analysis for the production of lignocellulosic ethanol with enzyme recovery by adsorption was prepared in the present study for the first time . 1. 7 Research objectives and thesis layout Ethanol produced from lignocellulose is a promising biofuel that can reduce our consumption of gasoline and CO 2 emissions. Nonetheless, the process for the production of ethanol from lignocellulose is still in development. In order to successfully commercialize lignocellul ose ethanol its production cost must be reduced to make it competitive with other fuels. Enzymatic hydrolysis process is one of the most expensive stages due to high enzyme costs. In this work, two cost reduction strategies were studied: enzymatic hydrolys is optimization within the context of the process economics and enzyme recovery by adsorption. Chapter 2 presents a study of protein concentration changes at hydrolysis conditions for the commercial cocktails most commonly used in the hydrolysis research: Celluclast 1.5L and Novozyme 188. A model that predicts Novozyme 188 protein concentrations changes at hydrolysis conditions was developed. Chapter 3 describes enzymatic hydrolysis performance at different conditions: enzyme loadings, pretreatment severity and solids concentrations. A kinetic model for predicting glucose yields during hydrolysis at different enzyme loadings, reaction times, solids concentration, and lignin content was proposed. Enzyme recovery by adsorption was experimental ly evaluated in Chapter 4. In this study, enzyme recycling was carried out at different hydrolytic times for different enzyme loadings and lignin concentrations. A material balance of the enzyme recycli ng process was buil t considering Novozyme 188 protein concentration changes, substrate porosity , and solids moisture content, thus enabling a mass- based determination of cellulase recovery. 43 In Chapter 5, an AspenPlus simulation of the process was built from the material balance obtained in Chapter 3 and 4. This simulatio n was economically evaluated with the Aspen Process Economic Analyzer (APEA). The production costs for the scenarios studi ed in Chapter 3 were determined, exposing the variables that control the production cost of ethanol . A sensitivity analysis of enzyme, caustic, and biomass price was carried out. The results were used to obtain an expression to predict ethanol production cost from enzyme loading, solids concentration in hydrolysis, pretreatment conditions, and enzyme, caustic, and biomass prices. A detai led separation stage was prepared for the scenarios with the lowest production cost s; detailed economic analysis of these scenarios was conducted. Chapter 6 presents the simulations of the process for the production of lignocellulosic ethanol with the a ddition of enzyme recovery by adsorption technology under different conditions. The viability of the enzyme recovery technology was determined by comparing the production cost of ethanol with and without enzyme recovery. The scenario in which enzyme recovery caused the largest production cost reduction was selected for sensitivity analysis. The viability of enzyme recovery is th us determined along with opportunities to improve enzyme recovery performance. In chapter 7, the conclusions obtained from this w ork are summarized and recommendations for further studies are given. 1.8 Research implications The model predicting Novozyme 188 protein concentration has significant implications since it has long been assumed that protein concentration was constant at hydrolysis conditions. Predicting Novozyme 188 protein concentration during hydrolysis enables determination of the distribution of cellulases during the reaction and thus adsorption of cellulases. This information can be used to optimize enzyme loadings and develop more accurate hydrolysis models. The Novozyme 188 protein model also allowed, for the first time, construction of a mass balance of the enzyme recycling process. With this information, the amount of 44 cellulases that can be recycled relative to the initial loading can be calculated. The presented methodology will support future enzyme recycling studies. The proposed model for the glucose yield during hydrolysis is advantageous because it can predict the hydrolysis of pretreated biomass at a wide rang e of conditions, such as enzyme loading, solids concentration, lignin content and time. This model is attractive because it requires only 5 parameters, while past hydrolysis models required up to 16 parameters. By implementing the negative effect of lignin in the hydrolysis yield, this process creates a bridge between pretreatment and enzymatic hydrolysis offering the opportunity to optimize these processes as a whole. The presented model allowed the study of the pretreatment conditions effect on downstream stages and illustrated the vital importance of pretreatment. The economic results obtained in this work show that enzymatic hydrolysis conditions cannot be optimized without considering the entire process . The economic analysis for the ethanol production with and without enzyme recycling for a wide range of conditions showed the importance of lignin in the enzyme recycling process . Enzyme recovery was better when substrates with low lignin content were used. Experimentally, the enzyme recycling applied at mild pretreatment conditions and low enzyme loadings results in extremely low percentage of cellulases recycled. However, the results of the economic analysis showed that enzyme recycling implementation reduced production cost of ethanol at high enzyme loadings and severe pretreatments. The high cost of enzymes is one of the major determinants of the ethanol production cost . Consequently, enzyme recycling must be improved to increase its benefits to the ethanol economy. The present project provided valuabl e information for lignocellulosic ethanol research and industry as it determined variables that control enzymatic hydrolysis and enzyme recycling and more importantly their impact on production cost s at an industrial scale. 45 2.1 Introduction Enzymatic hydrolysis is a complex reaction in which cellulases, most commonly from the fungus Trichoderma reesei, adsorb on cellulose to release glucooligomers, cellobiose, and glucose. As cellobiose inhibits cellulases action, β - glucosidase is added to the hydrolysis to cleave cellobiose to glucose (Lu et al., 2002). However, as lignocellulose also contains non -hydrolysable lignin to which cellulases a dsorb irreversibly, some cellulases are unproductively bound to lignin reducing the cellulases performance during hydrolysis (Farinas et al., 2010; Jørgensen et al., 2007; Zhou et al., 2009b) . Because of these factors, enzymatic hydrolysis requires a high enzyme l oad, which increases hydrolysis costs . One of the key economic factors and thus impediments to lignocellulosic ethanol production and commercialization is the amount and cost of enzyme s needed for hydrolysis (Lynd et al., 2008; Qi et al., 2011) . Consequently, enzymatic hydrolysis has been the target of numerous studies (Jørgensen et al., 2007; Taherzadeh and Karimi, 2007) . Several studies have attempted to increase the efficiency of the enzymes or to minimize the amount of enzyme needed (Arantes and Saddler, 2011; Shen and Agblevor, 2008b) . The recycling of cellulases is a relatively new technology that has been gaining popularity due to its potential to decrease enzyme requirements (Eckard et al., 2013; Jin et al., 2012; Lindedam et al., 2013; Qi et al., 2011; Weiss et al., 2013) . In order to quantify the amount of enzymes that can be rec ycled, it is necessary to understand the concentration changes of cellulases and β - glucosidase as well as proteins and compounds contained in the enzyme preparations generally used in the hydrolysis. The commercial enzyme cocktails from Novozyme, Celluclast 1.5L and Novozyme 188, have been widely used in the enzymatic hydrolysis (Arantes and Saddler, 2011; Hu et al., 2013; Kristensen et al., 2007; Qi et al., 2011; Rodrigues et al., 2012; Shi et al., 2011; Zheng et al., 2009) and enzyme recycling research (Eckard et al., 2013; Jin et al., 2012; Lindedam et al., 2013; Lu et 2 S t a b i l i t y o f c o m m e r c i a l e n z y m e p r e p a r a t i o n s u n d e r h y d r o l y s i s c o n d i t i o n s 46 al., 2002; Qi et al., 2011; Steele et al., 2005; Tu et al., 2007a; Tu et al., 2007b; Tu and Saddler, 2010; Weiss et al., 2013) . Many of these studies rely upon the assumption of constant protein concentration at hydrolysis con ditions; however this assumption has not been evaluated. Enzyme a nd protein structures determine solubility and activity and can change with pH, temperature, or concentration (Morris et al., 2009) . Therefore, the protein concentration in solution depends on system conditions and must be considered in enzyme adsorption and recycling studies. The objectives of this chapter are to evaluate Celluclast 1.5L and Novozyme 188 protein stability in solution and to model changes in protein concentration at enzymatic hydrolysis conditions in order to support the subsequent development of a kinetic model of the enzymatic hydrolysis and a mass balance of the enzyme recycling process . 2.2 Methods 2.2.1 Enzymes Lyophilized powder cellulases from Trichoderma reesei (P- cellulase, Cat. No. C8546, Sigma Aldrich) and β - glucosidase from Aspergillus niger (P- β - glucosidase, Cat. No. 9033- 06- 1, Sigma Aldrich) were used as standards . Commercial enzyme cocktails from Novozyme were also used: cellulase from T. reesei , Celluclast 1.5L ( 129.3 mg protein/mL, 30.7 CBU/mL, 63.8 F PU/mL) and β - glucosidase derived from A. niger , Novozyme 188 ( 102.2 mg protein/mL, 626.4 CBU/mL) . All enzymes were stored at 2°C until use. 2.2.2 Analysis of enzyme activity and protein concentration Cellulases activity was measured following the National R enewable Energy Laboratory (NREL) filter paper assay (Adney and Baker, 2008) and reported in filter - paper units ( FPU) per milliliter of solution . β - glucosidase activity was measured using the meth od described by Woods and Bhat (1988) and reported in cellobiase unit s (CBU). 47 Protein concentration was measured by the Bio- Rad protein assay, based on the colorimetric method of Bradford (Bio - Rad Laboratories, 2010). This is a dye- binding assay in which the absorbance shift of the dye Coomassie Brilliant Blue G - 250 is proportional to the protein concentration. As the Coomassie blue dye binds to primarily basic and aromatic amino acid residues, especially arginine, it is important to select the correct reference protein to accurately quantify protein (Zhu et al., 2009) . Therefore, the potential of P - cellulase, P- β -glucosidase, and the widely used Bovine Serum Albumin (BSA) as calibration standards was studied. It has been reported that components such as phenols can interfere with Coomassie blue dye (Existentes and Zaia, 1998) . In order to avoid any interference from the phenols and other components produced during pretreatment, pretreated biomass was extensi vely washed (Appendix B) . The washed biomass was placed in buffer and i ncubated at hydrolysis conditions for 4 hrs. The response obtained from the buffer - biomass solution and the buffer were similar . Calibration standards with a final v olume of 10 mL were prepared with Type II water (Purelab) for each enzyme concentration at room temperature. Samples of 0.5 mL were taken and centrifuged (RCF 16,904 g, 10 min). The absorbance spectrum was measured for each sample from 200 to 700 nm (UV - 1800, Shimadzu). The spectrum curves of the two enzymes were identical with three peaks at 264, 312 and 595 nm. Protein content was measured at 595 nm in the present study as it presented a higher and more defined response than those at 264 and 312 nm, and consequently increased measurement sensitivity. The response of the BSA was different from the P - cellulase and P- β - glucosidase as shown in Figure 15, most likely due to differences in their amino acid compositio ns (Zhu et al., 2009) . As the use of lyophilized cellulases as a calibration standard increases the accuracy in the protein qua ntification (Zhu et al., 2009) , P- cellulase and P- β - glucosidase were used as reference instead of the widely used BSA. 48 Figure 15. Protein calibration curves. Protein calibration curves (solid lines) for P - cellulase and P- β - glucosidase ( ); and BSA ( ). 2.2.3 Feedstock Wheat straw stored at room conditions was ground to pass through a 1 mm mesh sieve and stored at 4°C. The ground wheat straw is shown in Figure 16. The ground straw was w armed to room temperature prior to use. The moisture content of the biomass was determined by drying the sample at 105°C to constant weight. The composition of the raw substrate presented in Table 5 was determined as described in the chemical analysis section. Figure 16. Ground raw ( left ) and pretreated (right) wheat straw 0.0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1 1.2 Absorbance (595 nm) Protein concentration (g/L) 49 2.2.4 Pretreatment Oxygen delignification pretreatment of wheat straw was conducted in a 1 L bench top reactor (PARR 4520) shown in Figure 17. Figure 17. Reactor used to conduct the oxygen delignification The reactor configuration allowed control of temperature, reaction time, caustic concentration, and oxygen partial pressure. The reactor was charged at a concentration of 4 wt% dry wheat straw and 6 - 10 wt % NaOH (dry biomass), with a total mass of 500 g. The reactor was sealed and purged with nitrogen in order to remove oxygen. The vessel was heated to the desired temperature and then oxygen was bubbled through the reactor at 1 L/min (100 psig) . A mixing speed of 180 rpm was maintained during the entire process. The reaction was stopped by placing the reactor in ice and bringing it to atmospheric pressure. The p retreated biomass was then filtered and washed using a Buchner funnel and Whatman No. 4 filter paper. The pretreated biomass was extensively washed with around 4 L of water per batch in order to avoid interferences from the phenolic components in the pretr eatment liquor during the protein determination. The composition of the raw and pretreated substrates Pressure gauge Gas inlet Off - gas valve Heating jacket Delignificat ion reactor 50 is shown in Table 5, where the moisture content is percent weight of water in the substrate at a wet basis . The selected conditions represent the mild and severe conditions of the range of conditions tested by our group in past studies. Substrate M and S refer to substrate produced at mild (M= 30 min, 6 wt% NaOH/ dry biomass, 120°C ) and severe (S= 60 min, 10 wt% NaOH/ dry biomass, 150°C ) pretreatment conditions, respectively. Table 5. Composition of raw and pretreated biomass . F e e dst o c k Oxy g e n d elig nif ic a t io n co ndit io ns Co mpo sit io n ( w t % o f DM ) Mo ist ur e co nt e nt ( w t % of w et ma t t e r ) Subst r a t e Gluc a n Xy la n Ara bina n Ga la c t a n Ma nna n Lig nin Wheat straw Raw 35. 8 20.1 3.2 0.9 0.8 15.8 6.9 R Pretreated 30 min, 6 wt % NaOH/ dry biomass, 120°C 49.9 23.6 3.2 0.7 1.2 9.0 81.7 M Pretreated 60 min, 10 wt % NaOH/ dry biomass, 150°C 55.3 24.2 2.2 0.3 0.0 4.7 82.0 S 2.2.5 Chemical analysis The composition of raw and pretreated substrate were characterized according to the NREL procedure for carbohydrate and lignin determination (Sluiter et al., 2010) . Sugars were quantifie d using high- performance liquid chromatography (HPLC) equipped with an ion exchange PA1 (Dionex) column, a pulsed amperometric detector with a gold electrode, and a Spectra AS 3500 autoinjector (Dionex DX - 500, Dionex, CA) with fucose as internal standard. The HPLC equipment used is shown in Figure 18. 51 Figure 18. High- performance liquid chromatography (HPLC) equipment used to determine sugars concentration The Dionex DX600 HPLC system was used with the operating conditions outlined in Table 6. Table 6. HPLC operating conditions for carbohydrates determination. O pe r at ing Condit i on Column Temperature 30°C System Pressure 900- 1200 psia pH 10- 13 Sample Injection Volume 25 μL Total Retention Time 46 minutes 52 2.2.6 Protein concentration stability Protein stability was tested by quantifying protein concentration in solution over the course of 80 h. Acetate buffer (50 mM, pH 4.8) supplemented with 0.02% w/v tetracycline and 0.015% w/v cyclohexamide was placed in shakers (150 rpm) in 250 mL Erlenmeyer flasks at 20 or 50°C for one hour in order to achieve thermal equilibrium before the desired amount of pure enzymes or commercial enzyme preparation was added to a 50 mL final vol ume. Samples of 0.5 mL were withdrawn periodically and then centrifuged (RCF 16,904 g, 10 min). The protein concentration of the supernatant was measured. The Novozyme 188 protein stability in the presence of pretreated biomass was tested by quantifying t he cocktail’s protein concentration in solution over the course of 120 hours. Pretreated wheat straw (substrate S or M) was placed in buffer at 5 wt % solids concentration. The mixture was placed in a shaker (150 rpm) in 250 mL Erlenmeyer flasks at 50 oC bef ore the desired amount of Novozyme 188 was added to a 50 mL final volume. Solutions of Novozyme 188 in buffer were run in parallel as controls. Samples of 0.5 mL were withdrawn periodically and then centrifuged (RCF 16,904 g, 10 min) prior determining prot ein content. 2. 3 Results and discussion 2.3.1 Commercial enzyme preparations stability The enzyme system commonly used for enzymatic hydrolysis is a mixture of cellulases and β - glucosidase. Cellulases adsorb onto the substrate, primarily cellulose and lignin, whi le β -glucosidase has been reported to either not adsorb to the substrate (Tu et al., 2007a; Tu and Saddler, 2010), or to have a much lower adsorption on lignin than cellulases (Kumar and Wyman, 2009b) . Many hydrolysis studies have used Celluclast 1.5L and Novozyme 188 as sources of cellulases and β - glucosidase due to their high efficiency in hydrolyzing lignocellulosic materials. Even though these cocktails contain multiple proteins, enzymes and compounds, numerous studies have assumed that any change in the total protein 53 concentration during hydrolysis is due to the adsorption of cellulases on substrate (Arantes and Saddler, 2011; Lu et al., 2002; Qi et al., 2011; Tu et al., 2007b; Tu et al., 2009a) . By this reasoning, the protein concentration of Celluclast 1.5L and Novozyme 188 in the liquid phase should remain constant when substrate is absent. The validity of this assumption was tested by monitoring the protein concentration of Celluclast 1.5L and Novozyme 188 (1 FPU:5 CBU) at 2.2, 1.1, and 0.5 FPU/mL (6, 3 and 1.4 g/L protein concentration, respectively), for 78 hours without substrate at 50°C , a common hydrolysis temperature. T he highest cellulase activity has been reported at 40 to 50°C and 4 to 5 pH (Farinas et al., 2010; SERVA Electrophoresis, 2013) , therefore, these are the most common conditions used for enzymatic hydrolysis. As shown in Figure 19, the protein concentration decreased by 30 to 45% dependent on the initial enzyme loading. The protein aggregation observed in our experiments may increase at industrial conditions due to mechanical mixing (Kiese et al., 2008), therefore, protein aggregation has to be considered and addressed when scaling up the process. Figure 19. Celluclast 1.5L and Novozyme 188 prot ein concentration as a function of time at 50°C. Celluclast 1.5L and Novozyme 188 protein concentration at different initial concentrations (6 g/L , 3 g/L and 1.4 g/L ) over time at 50 °C. Lines are added to assist in visualizing trends. Farinas et al. (2010) reported the activity of endo- glucanases and β - glucosidases from a wild-type strain of A. niger . All the enzymes showed a high stability at 37°C, however, the 0 1 2 3 4 5 6 7 0 20 40 60 80 Protein concentration (g/L) Time (h) 54 activity of β - glucosidase decreased by approximately 28, 20 and 45% after 48, 72 and 96 h incubation at 50°C, respectively. In the same study, a 40 and 60% loss in activity after 24 and 96 h incubation at 50°C was also reported for endo- glucanases, which points to temperature as the responsible variable. In our work, precipitate was observed in the centrifuged samples, prior to protein quantification and it appeared that the amount of precipitate increased with incubation time. Given our observations and Farinas's previous work, we hypothesized that high temperatures promote protein aggregation causing the concentration of protein in the liquid phase to decrease. To confirm that aggregation is a consequence of temperature, Celluclast 1.5L , Novozyme 188, and a mixture of the two were incubated at 20°C and 50°C. The protein concentration was monitored for four days as shown in Figure 20 . Figure 20. Cocktails protein concentration at 20 and 50°C. Celluclast 1.5L and Novozyme 188 ( ), Celluclast 1.5L ( ) and Novozyme 188 ( ) protein concentration over time at 20 and 50°C. Lines are added to assist in visualizing trends. At 20°C the protein concentration remained practically co nstant for Celluclast 1.5L , Novozyme 188 and the mixture of the two cocktails ( Figure 20). However at 50°C a 34% drop in the total protein concentration was observed. Celluclast 1.5L suffered an 18% decrease in its concentration at 50°C after 4 days, demonstrating remarkable stability at high 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 1 2 3 4 Protein concentration (g/L) Time (Days) 20 ° C 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 1 2 3 4 Time (Days) 50 ° C 55 temperatures. The loss in Celluclast 1.5L protein concentration is similar to the 14% precipitation reported by Chylenski et al. (2012) for a cellula se preparation from T. reesei after 124 h at 50°C. In contrast, the protein concentration of Novozyme 188 decreased by 77% in 4 days at 50°C. This decrease in concentration was accompanied by an increase in the quantity of solid matter on the bottom of the flasks, indicating protein precipitation. It is important to note that the arithmetic sum of the individual Celluclast 1.5L and Novozyme 188 protein concentrations at 50°C (i nitial protein concentrations of Celluclast 1.5L and Novozyme 188, 1.61 g/L and 1.06 g /L respectively ) is very similar to the total protein concentration in the solution of Celluclast 1.5L and Novozyme 188 at 50 °C ( initial protein concentration of 2.70 g /L) for the period of time monitored. The assumption of constant protein concentrat ion during hydrolysis at 50°C does not hold due to the apparent precipitation of proteins from Novozyme 188, possibly due to aggregation. Dekker and Assay (1986) reported that Novozyme 188 lost 17 and 72% of its β - glucosidase activity after 48 and 120 h incubation at 50°C. Chundawat et al. (2011) reported a β -glucosidase protein composition of 7.7% in Novozyme 188, with large amounts of amylases (43.6%) and other proteins (42.1%). Therefore, to evaluate if the loss of β - glucosidase activity reported by Dekker and Assays (1986) is caused by β - glucosidase aggregation and precipitation , the stability of P - β - glucosidase from A. niger in solution at 50°C was examined next. 2.3.2 Stability of β - glucosidases from A. niger A representation o f the overall pathway generally proposed for the protein aggregation (fibrillation) is presented in Figure 21. Protein unfolding can be caused by pH changes, organic solvents, heat, protein concentration, shaking, or the presence of other proteins or chemical compounds (Lehninger et al., 2005) . It is hypothesized that native enzyme or protein monomer undergoes a conformational change denominated denaturation. The protein denaturation can, in some cases, lead to the exposure of “stick y” hydrophobic areas. These areas increase the propensity of the monomer to aggregate or “stick” to each other, causing it to become “active” (Morris et al., 2009) . The active monomers or unfolded proteins begin to 56 aggregate forming oligomers that ultimately lead to insoluble fibrils or amorphous aggregates. In short, when proteins suffer a structure change (denaturation) due to external stress, “sticky” areas in the protein can be exposed. These areas make denatured proteins aggregate to form insoluble macromolecules. Figure 21. Schematic aggregation of proteins. Idealized schematic representation of a general overall pathway to the formation of aggregates (Morris et al., 2009) . In his study of the thermal denaturation of almond β - glucosidase using differential scanning calorimetry (DSC), Tanaka (1991) found evidence indicating that β - glucosidase undergoes an irreversible structural transformation from its native co nformation to a denatured form between pH 4 - 8 and 50- 90°C . Therefore, given the changes in Novozyme 188 protein concentration in this study at similar temperatures, the β - glucosidases in Novozyme 188 may suffer a similar conformational change leading to ag gregation and subsequently precipitation. Mechanisms such as monomer addition (Oosawa and Kasai, 1962) , reversible association (Eisenberg, 1971) , prion aggregation (Eigen, 1996) , and a minimalistic 2- step model (Watzky et al., 2008) have been used to describe the role of protein aggregation in neurodegenerative diseases (Morris et al., 2009) . In order to develop a model of β -glucosidase aggregation, we utilized advances made in the field of heat induced β–lactoglobulin aggregation and precipitation, a phenomenon well - studied by the food industry (Verheul et al., 1998) . 57 The proposed protein aggregation reaction consists of two steps: the denaturation and the aggregation steps (Verheul et al., 1998) . The denaturation step accounts for β - glucosidase thermal denaturation, which was assumed to be an irreversible reaction given Tanaka's (1991) conclusions: 𝐵𝑘1→ 𝐷∗ 3 The native β - glucosidase (B) changes its structure as a result of high temperature, producing a denatured form of the enzyme (D*) that then aggregates in a series of irreversible aggregation reactions. It was assumed that only denatured enzymes participate i n the aggregation step. These aggregation reactions can be summarized as: 𝐷∗ + 𝐷∗𝑘2→ 𝐷2∗ 4 𝐷𝑛∗ + 𝐷𝑗∗𝑘𝑛+𝑗�⎯� 𝐷𝑛+𝑗∗ 𝐷𝑖∗ + 𝐷𝑛+𝑗∗ 𝑘𝑚�� 𝐷𝑚∗ 𝐷𝑛∗ ,𝐷𝑗∗, 𝐷𝑛+𝑗∗ , 𝐷𝑖∗ and 𝐷𝑚∗ are denatured β - glucosidase polymers consisting of n, j, i (j and i>1)( n and m≥2) (n+i+j<m) and m monomeric units and are treated as equivalent species. The aggregation step is comprised of hundreds of autocatalytic reactions (Morris et al., 2009) where denatured polymers adsorb on the aggregate increa sing its size and surface area. In consequence, the number of active spots to which denatured polymers can “stick”, increase s with the growing aggregate. Therefore, the aggregation of each polymer catalyze s the next aggregation reaction and, consequently, the entire aggregation process. In consequence, the aggregates 𝐷𝑛∗ ,𝐷𝑗∗, 𝐷𝑛+𝑗∗ , 𝐷𝑖∗ rapidly grow to form the insoluble species 𝐷𝑚∗ . According to the Bodenstein principle (Helfferich, 2003) , a steady- state situation will be reached quickly once the aggregation reaction starts; the rate at which D* is formed in the denaturation step will equal the rate at which it disappear s in the aggregation step (Roefs and De Kruif, 1994) . Therefore, the thermal denaturation step (equation 3) is slower than the aggregation step and is thus the rate limiting step. Under this consideration, our analysis assumes only native enzymes to be soluble. From this reaction, assuming first order kinetics, the β - glucosidase rate of disappearance is given by: 58 −𝑑[𝐵]𝑑𝑡= 𝑘1[𝐵] 5 Integrating this equation and taking [B]=[B 0] (g/L) at t=0 h gives: [𝐵] = [𝐵0]𝑒−𝑘1𝑡 6 Solutions of P - β - glucosidase at different concentrations were prepared using powdered P - β -glucosidase and incubated at 50°C. Soluble protein concentration was monitored and, as shown in Figure 22, protein concentration dec reased by as much as 84% over 98 h. The protein concentrations shown in Figure 22 were used to fit equation 5. The measured initial protein concentrations were also used in the model fitting process. The rate constant was determined by least squares regression , obtaining a k 1 of 0.024±0.001 h- 1. The proposed model had a good fit to the aggregation of P - β - glucosidase experimental data with a R 2= 0.97 as shown in Figure 22A. The analysis of the residuals showed that 78% of the predicted P - β -glucosidase concentrations are within 11% difference from the experimental data as shown in Appendix A (Figure 55 ). In agreement with the model, the aggregation rate of P - β -glucosidase is independent of the initial protein concen tration as shown in Figure 22B. 59 Figure 22. P- β - glucosidase concentration (Panel A), normalized concentration ( Panel B) at 50°C, experimental data ([B 0]=0.317 g/L ( ), 0.612 g/L ( ) and 1.388 g/L ( )) and 1st order kinetic model, e quation 6 ( ). Based on the presented model, 68% of the initial c oncentration of P - β - glucosidase is lost after 48 h incubation at 50°C , however, Dekker and Assay (1986) reported an activity loss of 17% for Novozyme 188. The reported loss in activity is incongruous with the large loss of P -β - glucosidase observed after 48 h incubation; therefore, β - glucosidase in Novozyme 188 seems to be more stable than P- β - glucosidase. As both β - glucosidase come from A. niger, the 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 0 20 40 60 80 100 Protein concentarion (g/L) Time (h) A 0 0.2 0.4 0.6 0.8 1 0 20 40 60 80 100 B/B₀ Time (h) B 60 apparent higher stability of β - glucosidase in Novozyme 188 may be caused by the stabilizing effect of the other proteins or compounds in Novozyme 188. 2.3.3 Novozyme 188 stability Understanding Novozyme 188 protein concentration changes at hydrolysis conditions can offer valuable information about cellulases adsorption- desorption process es. These enzymes adsorb onto substrate, while β - glucosidase has been reported to not adsorb to the substrate (Tu et al., 2007a; Tu and Saddler, 2010) . However, the interaction of Novozyme 188 with pretreated biomass has not been reported. For this reason, Novozyme 188 protein concentration was monitored in the presence of pretreated wheat straw with 4.7% (substrate S) and 9.0% (substrate M) lignin compositions. From Figure 23, the protein concentration profile s during the incubation of Novozyme 188 at 50°C in the absence or presence of pretreated biomass were similar and independent of pretreated biomass lignin composition. These results confirm that the proteins and enzymes contained in Novozyme 188 do not adsorb to substrate regards of lignin concentration. Our results agrees with the reported small to negligible adsorption of β - glucosidases from Novozym e 188 to pretreated biomass (Haven and Jørgensen, 2013) . Although the β - glucosidases in Novozym e 188 do not adsorb on the substrate, it was reported by Haven and Jørgensen (2013) that β - glucosidases in Cellic® CTec2 adsorb on different substrates. Therefore, the behaviour of β - glucosidases, cellulases and proteins in other cocktails is expected to be different from what has been reported here. 61 Figure 23. E ffect of pretreated biomass on Novozyme 188 protein concentration. Novozyme 188 protein concentration at 50°C over time in the absence ( ) and presence of pretreated biomass: substrate M ( ) and S ( ). Line added to assist with visualization . The changes in total protein concentration in solution during enzymatic hydrolysis are most probably caused by the Novozyme 188 protein aggregation and the adsorption- desorption of cellulases in the substrate. The protein mass balance in solution during hydrolysis is given by [𝑇𝐸] = [𝐸𝐿] + [𝑁] 7 where the total protein concentration [T E] (g/L) is defined by the protein concentration of Celluclast 1.5L in solution [ EL] (g/L) and Novozyme 188 protein concentration [N] (g/L) as shown in Figure 20. Hu et al. (2013) reported that the major cellulases monocomponents within Celluclast 1.5L , on a protein weight basis, were T. reesei Cel7A (CHB I), Cel6A (CHB II), Cel7B (EG I), and Cel5A (EG II) comprised approximately 56%, 12%, 5%, and 6% of the total protein, respectively. However, additional cellulases may be present in the cocktail as the T. reesei genome also includes Cel12 (EG III), Cel61 (EG IV) and Cel45 (EG V), as reported by Martinez et al. (2008) . From this work, it can be assumed that Celluclast 1.5L is primarily composed of cellulases . Therefore, the concentration of cellulases in solution [E L] (g/L), and consequently those adsorbed on substrate, can be indirectly determined by subtracting Novozyme 188 protein concentration [N] (g/L) from the total enzyme concentration [T E] (g/L) measured during hydrolysis. This approximation is possible 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0 20 40 60 80 100 120 140 Protein concentration (g/L) Time (h) 62 only if the Novozyme 188 protein concentration can be predicted during enzymatic hydrolysis. In order to model the aggregation of Novozyme 188, it was assumed that the majority of proteins and enzymes in Novozyme 188 have the same thermal stability. This assumption is supported by the results presented in Figure 20, where 80% of the initial Novozyme 188 protein concentration is lost after 4 days incubation at 50°C showing a consistent response to temperature. As mentioned, some compounds present in Novozyme 188 may stabilize the proteins and enzymes at high temperatures. As A. niger has been reported to produce chaperones (Guillemette et al., 2007; Lu et al., 2 010), it was assumed that the proteins and enzymes in Novozyme 188 are stabilized by chaperones. Chaperones are proteins that bind to non- native conformations of proteins and assist them to reach their native structure. Most chaperones bind to exposed hydrophobic surfaces of non- native species and thereby stabilize the protein or enzyme against aggregation. After being released from the chaperone, the protein can fold correctly, or rebind to the chaperone again until a native conformation is reached (Liberek et al., 2008; Schlieker et al., 2002) . The thermal denaturation of protein in Novozyme 188 can be described using the same assumptions and considerations used for the aggregation of P - β - glucosidase. However, due to the presence of chaperones, a reversible reaction was added to the model to account for the stabilizing effect of chaperones: 𝑁 𝑘2′ �⎯⎯⎯� 𝑘1′ �⎯⎯⎯⎯� 𝐷∗ 8 where N refers to the native proteins in Novozyme 188 and 𝐷∗ refers to the denatured proteins which interact with chaperones that regenerate their native protein structure. Therefore, the r ate of disappearance of native proteins is: −𝑑[𝑁]𝑑𝑡= 𝑘1′ [𝑁]− 𝑘2′ [𝐷∗] 9 The amount of protein precipitated is given by [D∗] = [N0] - [N] (g/L), where [N 0] is the initial Novozyme 188 protein concentration (g/ L). Solving this equation taking [N]=[N 0] (g/L), at t=0 h yields: 63 [N] =[N0]�𝑘2′ +𝑘1′e−t�𝑘2′ +𝑘1′ ��𝑘2′ +𝑘1′ 10 The rate constants in equation 10 were determin ed by least squares regression to be k1′= 0.017±0.003 h- 1 and 𝑘2′ =0.011±0.003 h- 1. The experimental initial concentrations were used for the model fitting. Equation 10 successfully predicts the Novozyme 188 protein concentration in solution at the tested hydrolysis conditions with a correlation coefficient of 0.96, as shown Figure 24A. The predicted concentrations over and underestimate the protein concentration by a maximum of 20% from the experimental data . The residuals analysis can be seen in Appendix A (Figure 56 ). Therefore, the proposed model (equation 10) has a good fit to the experimental data in the range of concentrations tested. 64 Figure 24. Proposed kinetic model fit on Novozyme188. Novozyme protein concentration (panel A) and normalized concentration (panel B) at 50°C , experimental data ([N 0]=0.36 g/L ( ), 0.71 g/L ( ) and 1.44 g/L ( )) and the additives model, equation 10 ( ). However, it is clear from Figure 24B that the protein aggregation from Novozyme 188 exhibits a peculiar behavior whi ch is not captured by equation 10: the rate of aggregation was slower at higher initial concentrati ons. These results contradict the aggregation behavior reported for other systems of proteins (Roefs and De Kruif, 1994; Verheul et al., 1998) , where the rate of aggregation and precipitation increases with initial concentration (Wang, 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 0 20 40 60 80 100 Protein concentarion (g/L) Time (h) A 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 20 40 60 80 100 [N]/[N0] Time (h) B 65 2005) . Under severe stress conditions, protein aggregation occurs even in the presence of chaperones, because the substantial increase of misfolded proteins cannot be buffered by the limited chaperone c apacity (Schlieker et al., 2002) . Therefore, it is unlikely that the increasing protein stability at increasing Novozyme 188 concentrations is caused by the chaperones. Other compounds have been explored for their ability to achieve similar stabilizing effects during denaturation. A common method to stabilize proteins is the use of additives such as sugars and polyols, salts, and surfactants which have chaperone - like effects. These additives stabilize proteins and suppress aggregation by changing proteins’ environmental properties (Wang, 2005) . Novozyme 188 is a mixture of compounds selected to improve β - glucosidase performance and stability. As Novozyme 188 is a commercial product, its preparation and composition are not public knowledge, making it difficult to fully understand its behavior. However, the presence of additives in Novozyme 188 may explain the increasing Novozyme 188 stability at high protein concentrations. If additives are present in the cocktail, their concentration increases at the same rate as Novozyme 188 protein. Therefore, the additives effect in the system becomes stronger as concentration increases. The hypothetical presence of additives was included in the reversible reaction of the proposed model: 𝑁 k2′′, 𝐴0�⎯⎯⎯⎯⎯� k1′′�⎯⎯⎯⎯⎯� 𝐷∗ 11 The denatured proteins ( D∗ ) are influenced by the additives ( A) to regenerate their native structure (N). Therefore the rate of disappearance and formation of native pr otein is now: −𝑑[𝑁]𝑑𝑡= 𝑘1′′[𝑁]− 𝑘2′′[𝐴0][𝐷∗] 12 The concentration of additives [A 0] (g/L) was assumed constant and proportional to the initial Novozyme 188 protein concentration [N 0] (g/L), as both come from the same solution. The initial additive concentration was assumed to be given by [A0] = mc[N0], where m c is a dimensionless constant. Using these relations, equation 12 becomes: 66 −𝑑[𝑁]𝑑𝑡= k1′′[𝑁] − k2′′𝑚𝑐([𝑁0]− [𝑁])[𝑁0 ] 13 Solving this equation using 𝑘𝑠 = 𝑘2′′ mc and taking [N]=[N 0] (g/L) at t=0 h yields: [𝑁] =[𝑁0] �𝑘𝑠[𝑁0] + 𝑘1′′𝑒−𝑡�𝑘1′′+𝑘𝑠[𝑁0]��𝑘1′′ + 𝑘𝑠[𝑁0] 14 The rate constants in equation 14 were determined by least squares regression to be k 1′′= 0.016± 0.002 h- 1 and k s=0.008± 0.003 Lg- 1h- 1. The value of constant k 1′′ is similar to the value obtained for k 1′ in equation 10, thus, only the reverse reaction is affected by the additives presence making it depend on cocktail loading. Equation 14 successfully predicts the concentration of Novozyme 188 in solution with a correlation coefficient of 0.98, as shown in Figure 25. The residual analysis showed that all the predicted protein concentrations differ from the experimental data by 15% ; the residual plot is provided in Appendix A (Figure 57) . W hen time tends to infinity, equation 14 indicates that the protein concentration of Novozyme 188 will reach equilibrium and a constant concentration in solution. After monitoring the protein concentration for 92 hrs, it was not possible to conclude whether the protein concentration had reached equil ibrium. Hydrolysis times of 1 to 4 days have been proposed for industrial operations therefore , monitoring the protein concentration for longer reaction times was not investigated given the industrial emphasis of this project. Based on the residual analysis for equation 10 and 14, it was concluded that the model with the additives consideration (equation 14) more accurately represents the protein aggregation observed. This conclusion agrees with the second order Akaikes` information criterion (AICc) (Hurvich and Tsai, 2007) calculated for equation 10 (- 28) and equation 14 (- 37), where the larger negative number reflects the better fit to the experimental work. 67 Figure 25. Proposed kinetic model considering the presence of additives in Novozyme188. Novozyme 188 protein normalized concentration at 50°C , experiment al data ([N 0]=0.36 g/L ( ), 0.71 g/L ( ) and 1.44 g/L ( )) and the additives model, equation 14 ( ). The aggregation of Novozyme 188 predicted with the additives model presented in Figure 25 is in better agreement with the experimental results where the Novozyme 188 stability increases at high loadings. The model has a good fit to the data at [ N0]= 0.71 and 1.44 g/L, however, the model poorly predicted the protein aggregation at [ N0]=0.36 g/L, as can be seen in Figure 25. At low cocktail loadings ( [N 0]=0.36 g/L) , the effect of a dditives in the system is likely to be minimal. Nonetheless, the additives concentration, as well as their effect on the system, increases at higher cocktail loadings ([N 0]= 1.44 g/L). Therefore, the additives stabilizing effect can be seen only at high pr otein loadings. This behavior is similar to results published where the aggregation of lysozyme is reduced by increasing concentrations of additives such as glycine ethylester, 2- methoxyethanol, KCl, spermine among others (Kudou et al., 2003; Shiraki et al., 2005) . The similarity of our results wi th previous additive studies supports our hypothesis that the increasing stability of Novozyme 188 protein concentration at high concentrations is due to the effect of additives. 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 20 40 60 80 100 [N]/[N0] Time (hrs) 68 2.4 Conclusions The successful commercialization of lignocellulosic ethanol re sts upon optimization of the enzyme cocktails and hydrolysis conditions thus enzyme adsorption to biomass, minimum enzyme loadings, and enzyme recycling has been examined in numerous studies. These studies have assumed that the protein concentration at hydrolysis conditions remains constant and that the most significant change in protein concentration during enzymatic hydrolysis is due to adsorption of cellulases on biomass. In this study, this fundamental assumption was tested employing two widely used commercial enzyme preparations: Celluclast 1.5L and Novozyme 188. The constant protein concentration assumption was found to be invalid when a 46% protein loss of a solution of Celluclast 1.5L and Novozyme 188 at 50 ° C was observed after four days incubation. Loss of Novozyme 188 protein concentration was determined to be the main cause of this decrease as Celluclast 1.5L protein concentration only slightly decreased (18% after 4 days) at 50 ° C. The protein loss in Novozyme 188 is likely due to heat - induced denaturation which promotes protein aggregation and ultimately precipitation. The aggregation of pure β - glucosidase in solution was monitored and successfully modeled by assuming the thermal denaturation reaction was the rate limiting step. However, the high precipitation of pure β - glucosidase observed does not agree with reported l oss of β -glucosidase activity in Novozyme 188. The apparently higher stability of β - glucosidase in the cocktail is believed to be caused by the presence of chaperones in the cocktail. The aggregation of proteins from Novozyme 188 was modeled considering the presence of chaperones in the cocktail and their effect on the system. The experimental data exhibited increasing protein stability at high protein loadings which was not predicted by the model. This behavior was proposed to be caused by additives. The presence of additives was included in the Novozyme 188 aggregation model producing a model which successfully describes the decrease in aggregation rate with increasing cocktail loading. The observed Novozyme 188 protein concentration changes are in agreeme nt with previous studies of the aggregation of lysozyme in presence of different additives. 69 The thermal instability of Novozyme 188 has not been previously reported and has significant implications for the study of enzymatic hydrolysis. The total enzyme concentration has been monitored during hydrolysis in past studies to examine cellulases adsorption- desorption process es. However, given the Novozyme 188 protein aggregation observed in this work, the adsorption- desorption information obtained is not accura te. By combining the measurement of total enzyme concentration with the model in this paper, cellulases concentration changes can be accurately determined during enzymatic hydrolysis. Cellulases concentration profiles will lead to greater understanding of the complex adsorption- desorption process during hydrolysis. Concentration profiles of cellulases can be used to minimize the amount of cellulases that do not adsorb and hydrolyse cellulose and consequently decrease the ethanol production cost. Cellulases concentration profiles during hydrolysis can also help to select the time when the concentration of cellulases is higher, thus increasing the amount of recyclable enzymes. The proposed model will enable mass balances that can be used to accurately determine the amount of cellulases that can be recycled by adsorption from the liquid phase as well as economic analysis of such technology for ethanol production. 70 3 .1 Introduction In order to make lignocellulosic ethanol production commercially attractive, several unit operations need to be improved. Of these, enzymatic hydrolysis is one of the most important due to high enzyme costs (Galbe and Zacchi, 2002) . Optimization of enzymatic hydrolysis will be supported by the development of kinetic models that accurately quantify cellulose conversion, hydrolysis rate and cellulase deactivation (Zhang et al., 2010) . Two approaches have been used to model the kinetics of enzymatic hydrolysis: empirical and mechanistic. Empirical models relate factors using math ematical correlations without providing insight into underlying mechanisms. These models are not applicable outside the conditions under which they were developed. Nonetheless, they are easy to develop, correctly predict hydrolysis yields , and can be used to optimize reaction conditions (Peri et al., 2007) . Zhou et al. (2009) developed an empirical model to predict g lucose production during hydrolysis depending on enzyme preparation composition. This model was used to optimize the composition of an enzyme preparation from Trichoderma viride for maximum sugar production from the hydrolysis of steam - exploded corn stover . Using hydrolytic data from 147 poplar wood samples, O'Dwyer et al. (2008) developed a model capable of predicting sugar conversions as a function of cellulose crystallinity index, lignin content, and acetyl content. Mechanistic models, developed from mechanisms, physical parameters, background theories, and assumptions, provide fundamental insight into the mechanisms of enzymatic hydrolysis. Mechanistic models can contain numerous differential equations and parameters and are thus challenging to develop and manage (Zhang et al., 2010) . A mechanistic model describing SSF of pretreated spruce and sugar cane as a function of cellulose crystallinity and degree of polymerization was devel oped by Ljunggren (2005) . Despite the model’s good fit to 3 M o d e l i n g o f p r e t r e a t e d w h e a t s t r a w h y d r o l y s i s a s a f u n c t i o n o f t i m e , e n z y m e c o n c e n t r a t i o n , a n d l i g n i n c o n c e n t r a t i o n 71 experimental data the large number of parameters, eighteen in total, required is a significant disadvantage. Shao et al. (2009) modeled SSF to evaluate feeding frequency and reactors performance using nineteen paramet ers. Liao et al. (2008) modeled enzymati c hydrolysis of different types of manure fibers using enzyme activity to account for enzyme concentration, which limits its application in process simulations. Zheng et al . (2009) developed a model to predict glucose concentration during hydrolysis at different enzyme a nd solids loadings and background glucose and cellobiose concentrations. Due to the complex solution of this model and the sixteen parameters its implementation in engineering software is challenging. Shen and Agblevor (2008a, 2008b) proposed a simple mechanistic model requiring only three parameters but determined the individual values of parameters for each initial enzyme concentration tested instead of entire set of conditions. According to Zhang et al. (2010) , these parameters should be an enzyme property in dependent of enzyme concentration. Using the model proposed by Shen and Agblevor (2008 a, 2008b), Zhang et al. (2010) determined parameters using hydrolysis data at multiple enzyme loadings to predict the hydrolysis of pretreated wheat straw. Shen and Agblevor (2008a, 2008b) assumed enzyme deactivation (EG, CBH and B) was due to the formation of an ineffective cellulose - or hemicellulose-enzyme complex while Zhang et al. (2010) assumed enzyme deactivation to be the result of end product inhibition (i.e., sugar - enzyme interactions). Neither of these studies modeled non- productive enzyme adsorption on lignin. Zheng et al. (2009) demonstrated the significant impact of lignin on hydrolysis rates by determining model parameters with and without incorporating enzyme adsorption by lignin in their model. Based on this analysis, Zheng et al. (2009) concluded that enzyme l oss due to non- pro ductive and irreversible adsorption to lignin should be considered in the development of a reliable and powerful model of enzymatic hydrolysis. Nonetheless, the performance of Zheng’s model at different lignin concentrations was not evaluated. Enzymatic hydrolysis is a complex heterogeneous process in which cellulose is hydrolyzed by enzymes known as cellulase s to glucose. There are a number of obstacles that diminish enzyme performance and decrease hydrolysis conversion. Such obstacles include the well -known inhibitory effect of cellobiose on cellulase, enzyme diffusion limitations and the 72 presence of lignin, which shields cellulose from enzymes and irreversibly adsorbs enzy mes (Shen and Agblevor, 2008b). Numerous techno- economic analyses have provided valuable information regarding the influence of enzyme and feedstock cost on ethanol production (Anex et al., 2010; Huang et al., 2009; Humbird et al., 2011; Kazi et al., 2010; Klein- Marcuschamer et al., 2012; Wingren et al., 2003) . However, most studies estimated ethanol costs f or a single operating condition using specific lignin conte nt and sugar yields, making it difficult to assess the impact of individual pretreatment and enzymatic hydrolysis variables on the overall process. T o address this need, a model that describes the impact of residence time, enzyme loading, solids concentration and lignin content on enzymatic hydrolysis, and can be incorporated into process simula tions is developed in the present work. 3 .2 Methods The substrates and methodologies reported in the Chapter 2 were used in the present section. 3.2.1 Enzymatic hydrolysis Hydrolysis was performed in 250 mL Erlenmeyer flasks (50 mL total reaction volume) at 50°C in an incubator at 150 rpm. The reaction was carried out in 50 mM acetate buffer (pH 4.8) supplemented with 0.02% w/v tetracycline and 0.015% w/v cyclohexamide to prevent microbial contamination. Mixed w heat straw and buffer were preheated to reach thermal equilibrium prior to enzyme addition. Samples of 0.5 mL were taken during the course of hydrolysis and centrifuged (RCF 16,904 g, 10 min). A 0.1 mL sample of the s upernatant was used to determine protein content and the remaining supernatant was kept at - 20°C until analyzed for sugars. In order to study the effect of cellulase s, hydrolysis time, solids loading and lignin content on enzymatic hydrolysis experiment s were conducted at a range of conditions, considering 73 substrates obtained at mild (M= 30 min, 6 wt % NaOH/ dry biomass, 120°C ) and severe (S= 60 min, 10 wt % NaOH/ dry biomass, 150°C ) pretreatment conditions. The hydrolysis conditions are presented in Table 7 . The experimental enzymatic hydrolysis at each condition was carried out in triplicate. Table 7. Enzymatic hydrolysis conditions (pretreatment conditions: M= 30 min, 6 wt % NaOH/ dry biomass, 120°C and S= 60 min, 10 wt % NaOH/ dry biomass, 150°C ). Scenario Total Solids (wt % DM) Substrate Cellulase loading (FPU/ g cellulose) M20- 5 5 M 20 M40- 5 40 S20- 5 S 20 S40- 5 40 M20- 10 10 M 20 M40- 10 40 S20- 10 S 20 S40- 10 40 The cellulose conversion (CC) (%) and xylan conversion (XC) (%) are defined as: 𝐶𝐶 =𝐺𝑙𝑢𝑐𝑜𝑠𝑒 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑑(𝑔) ∗ 0.90𝐶𝑒𝑙𝑙𝑢𝑙𝑜𝑠𝑒 (𝑔)∗ 100 15 𝑋𝐶 =𝑋𝑦𝑙𝑜𝑠𝑒 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑑(𝑔) ∗ 0.88𝑋𝑦𝑙𝑎𝑛 (𝑔)∗ 100 16 In order to compare the glucose production rates obtained at 5 and 10 wt % solids concentration and 20 and 40 FPU/g cellulose, the change of cellulose conversion (∆CC) over the change of time (∆t) for substrates M and S at different solids concentration was calculated. 𝐶𝑚 = �∆𝐶𝐶∆𝑡 𝑀20−5∆𝐶𝐶∆𝑡 𝑀40−10�𝐴𝑣𝑒𝑟𝑎𝑔𝑒 17 74 𝐶𝑠 = �∆𝐶𝐶∆𝑡 𝑆20−5∆𝐶𝐶∆𝑡 𝑆40−10�𝐴𝑣𝑒𝑟𝑎𝑔𝑒 18 𝐶𝑚20 = �∆𝐶𝐶∆𝑡 𝑀20−5∆𝐶𝐶∆𝑡 𝑀20−10�𝐴𝑣𝑒𝑟𝑎𝑔𝑒 19 𝐶𝑚40 = �∆𝐶𝐶∆𝑡 𝑀40−5∆𝐶𝐶∆𝑡 𝑀40−10�𝐴𝑣𝑒𝑟𝑎𝑔𝑒 20 In a similar way, the cellulose and xylan conversion rate in one scenario were compared using the average ratio of change in cellulose (∆CC) and xylan (∆XC) conversions with respect to time (∆t). 𝐶𝑋= �∆𝐶𝐶∆𝑡∆𝑋𝐶∆𝑡�𝐴𝑣𝑒𝑟𝑎𝑔𝑒 21 3 . 3 Results and discussion 3.3.1 Hydrolysis experiments The hydrolysis experiments were carried out at multiple conditions (Table 7) in order to assess how the solids and enzyme loading affect reaction rate and cellulose conversion. Cellulose conversions during enzymatic hydrolysis are shown in Figure 26. 75 Figure 26. Cellulose conversion during hydrolysis for substrate M and S. For equal solids loadings, higher cellulose conversions w ere observed at 40 FPU/g cellulose compared to 20 FPU/g cellulose for substrat es M and S. The exception to this trend is scenario S20- 5, where 20 FPU/g was sufficient to hydrolyse all available cellulose. However , in scenario S40- 5 maximum conversion was r eached more quickly (after 10 h) indicating an increase in reaction rate but not in cellulose conversion. When cellulose concentration and 0 20 40 60 80 0 20 40 60 80 Cellulose converison (%) Time (h) Subs t r at e M M20- 5 M40- 5 M20- 10 M40- 10 0 20 40 60 80 100 0 20 40 60 80 Cellulose converison (%) Time (h) Subs t r at e S S20- 5 S40- 5 S20- 10 S40- 10 76 enzyme loading were increased proportionately, the reaction rate ( Cm =1.08, Cs =1.12) and cellulose conversion (Figure 26) were similar. At equal enzyme l oading, the reaction rates observed for substrate M (high lignin content) when working at 5 and 10 wt % solids concentration were similar (Cm20=1.25, Cm40=1.08) . However, lower conversions were achieved at high solids concentrations (e.g. CCM20-5 =69% , CCM20-10 =58%) which may be due to limited enzyme mobility or elevated sugar concentrations (Bansal et al., 2009; Hodge et al., 2008) . In each scenario, the calculated xylan and cellulose reaction rates were similar and proportional ( C/X=0.94).Therefore, xylan and cellul ose conversions can be linearly related as shown in Figure 27. This behaviour is in agreement with the results reported by Hu et al. (2011) at multiple enzyme loadings. As shown in Figure 27, the regression equation has a negative y- intercep t, which indicates faster hydrolysis of xylan . This suggests a greater amount of xylan relative to cellulose on the substrate's surface at the start of the reaction . 77 Figure 27. Cellulose and xylan conversion during enzymatic hydrolysis at 5 wt % (panel A) and 10 wt % DM solids concentration (panel B) . Pretreatment conditions: M= 30 min, 6 wt % NaOH/ dry biomass, 120°C and S= 60 min, 10 wt % NaOH/ dry biomass, 150°C . 3.3.2 Cellulases concentration during enzymatic hydrolysis The protein mass balance in solution during hydrolysis is given by equation 7 where Celluclast 1.5L is mainly composed of cel lulases. Concentration of cellulases in solution [ EL] was indirectly determined by subtracting the Novozyme 188 protein concentration [N] (g/L) from total enzyme concentration [T E] (g/L) measured during hydrolysis. The protein 0 20 40 60 80 100 0 20 40 60 80 100 Cellulose conversion (%) Xylan conversion (%) M20- 5 M40- 5 S20- 5 S40- 5 A 0 20 40 60 80 100 0 20 40 60 80 Cellulose conversion (%) Xylan conversion (%) M20- 10 M40- 10 S20- 10 S40- 10 B CC = 0.9916*XC - 4.2386 R² = 0.9565 CC = 1.0725*XC - 4.2182 R² = 0.9365 78 concentration from Novozyme 188 can be determined using equation 14. The Novozyme 188 protein, cellulases and total enzyme concentrations during enzymatic hydrolysis are shown in Figure 28. 0 10 20 30 Glucose concentration (g/L) M20 - 5 0.0 0.4 0.8 1.2 0 8 16 24 32 40 48 56 64 72 80 Protein concentration (g/L) Time (h) 0 10 20 30 Glucose concentration (g/L) M40 - 5 0.0 0.8 1.6 2.4 0 8 16 24 32 40 48 56 64 72 80 Protein concentration (g/L) Time (h) 79 0 10 20 30 Glucose concentration (g/L) S 20- 5 0.0 0.4 0.8 1.2 1.6 0 8 16 24 32 40 48 56 64 72 80 Protein concentration (g/L) Time (h) 0 10 20 30 Glucose concentration (g/L) S40 - 5 0.0 0.8 1.6 2.4 0 8 16 24 32 40 48 56 64 72 80 Prtotein concentration (g/L) Time (h) 80 0 10 20 30 40 50 Glucose concentration (g/L) M20 - 10 0.0 0.8 1.6 2.4 0 8 16 24 32 40 48 56 64 72 80 Glucose concentration (g/L) Time (h) 0 10 20 30 40 50 Glucose concentration (g/L) M40 - 10 0.0 1.0 2.0 3.0 4.0 5.0 6.0 0 8 16 24 32 40 48 56 64 72 80 Protein concentration (g/L) Time (h) 81 Figure 28. Glucose concentration during enzymatic hydrolysis from experimental data ( ) and equat ion 38 ( ). Cellulases concentration ( ) in solution calculated from the total protein concentration ( X ) measured (line added to assist in visualizing trends) and Novozyme 188 protein concentration ( ) calculated from equation 14. 0 10 20 30 40 50 Glucose concentration (g/L) S 20 - 10 0.0 0.8 1.6 2.4 0 8 16 24 32 40 48 56 64 72 80 Protein concentration (g/L) Time (h) 0 10 20 30 40 50 Glucose concentration (g/L) S 40 - 10 0.0 2.0 4.0 6.0 0 8 16 24 32 40 48 56 64 72 80 Protein concentration (g/L) Time (h) 82 For each scenario, it can be seen that during the first two hours of enzymatic hydrolysis the concentration of cellu lases decreased by 86 to 100%. As the reaction continues, cellulases desorb at different rates in each scenario to reach constant concentration after 24 to 48 h; the desorption rate appears to depend on lignin content and initial enzyme loading. A larger amount of cellulases desorb from substrate S than from substrate M, most likely due to the low lignin content of substrate S . Severe pretreatment conditions may decrease the hydrophobic character of lignin to a gre ater extent than mild conditions, leading to a decrease in cellulases adsorption. The mechanism of oxygen delignification is extremely complex and is not fully understood (Suzuki et al., 2006) . Consequently, most efforts to characterize the effect of oxygen delignification on biomass have focused on modelling and predict ing the solubilisation of lignin, cellulose and hemicellulose (Ahring et al., 1998; Kindsigo and Kallas, Lissens et al., 2004). Greater efforts are needed to understand the effect of oxygen delignification severity on lignin structure . In scenario S20- 10, the amount of cellulases loaded may be insufficient to saturate the substrate since all cellulases remain adsorbed during the reaction. At high enzyme loadings (S40- 10), the amount of cellulases in solution is more than enough to saturate the substrate and therefore the amount of free cellulases in solution increased. Increasing enzyme loading also increased cellulase desorption rate. These results suggest that enzyme loading can be optimized by loading sufficient cellulases to saturate the available cellulose surface while minimizing excess cellulases. The β-glucosidases ratio used in this thesis (1 FPU:5 CBU) is high compare to other works (1 FPU:2 CBU). The selected β - glucosidases loaded was used to avoid any cellobiose inhibition. However, b y optimizing cellulases loadin g, the β- glucosidases usage can be minimized without affecting the consumption of cellobiose . Most of the loaded cellulases ( 61 to 100%) remain adsorbed after 76 h of hydrolysis, probably due to irreversible adsorption onto lignin. Therefore, lignin plays an important role in the successful adsorption of cellulases on cellulose, the hydrolysis rate, and cellulose conversion and thus lignin content should be considered when calculating enzyme loading and must be included in hydrolysis models. 83 3.3.3 Kinetic model for the enzymatic hydrolysis of lignocellulose 3. 3. 3.1 Enzymatic hydrolysis of cellulose Hydrolytic enzymes work synergisticall y to hydrolyse cellulose by creating new accessible sites, removing obstacles, and relieving product inhibition. In Figure 29, a simplified pathway for the hydrolysis of cellulose by cellulases and β - glucosidases is presented . Figure 29. Schematic representation of overall pathway for the enzymatic hydrolysis of cellulose by cellulase. Cellulases were treated as having a single combined effect on the insoluble substrate as both enzymes adsorb onto substrate (Pribowo et al., 2012) ; together they are represented as E . The CelluloseEndoglucanases Exoglucanases β-glucosidaseAdsorptionComplexationCatalysisCellobioseGlucoseCellulase84 surface and structure of the insoluble substrate were assumed to be homogeneous (Shen and Agblevor, 2008a). Thus enzymatic hydrolysis can be described by the equations: 𝐸 + 𝐶𝑘3�𝑘−3�⎯� 𝐶𝐸∗𝑘4→ 𝐸 + 𝐶𝑒 22 𝐶𝑒 + 𝐵𝑘5→ 𝐺 + 𝐵 23 The active E in solution is adsorbed onto cellulose active sites (C) to form the enzyme -cellulose complex ( CE*) from which cellobiose (Ce) is released. In equation 22, k3 (L/h g), k-3 (1/h) and k 4 (1/h), are the rate constants o f enzyme adsorption, enzyme desorption and cellobiose production, respectively. Shen and Agblevor (2008a, 2008b) and Zhang et al. (2010) described enzymatic hydrolysis using a version of e quation 22 in which glucose is directly produced from cellulose instead of cellobiose. This assumption implies that all cellobiose produced is converted to glucose. Shen and Agblevor, and Zhang’s model s assume a single combined effect of all enzymes and thus E refer red to both cellulases and β -glucosidase. However, as the enzyme preparations used by Shen and Agblevor, and Zhang contained cellulases and β - glucosidase in unknown proportions (similar to the crude cellulase powder used by Zhang et al. (2010) ), the β - glucosidase concentrations may have been insufficient to efficiently convert all cellobiose as assumed. Therefore, a degree of cellobiose inhibition may have been present in the reaction. In the present work, β - glucosidase is supplemented independently using Novozyme 188 in order to control β - glucosidase concentration during hydrolysis and avoid cellobiose inhibition. In equation 23, cellobiose is converted to glucose (G) by β - glucosidases (B), where k 5 (1/h) is the glucose production rate constant. The enzyme - cellulose complex formation rate and substrate mass balance are shown in equations 24 and 25. 𝑑 [𝐶𝐸∗]𝑑𝑡= 𝑘3[𝐶][𝐸] − 𝑘−3[𝐶𝐸∗]− 𝑘4[𝐶𝐸∗] 24 [𝐶] = [𝐶0]− [𝐶𝐸∗]− [𝐶𝑒] − [𝐺] 25 where [ C0] is the initial concentration of cellulose using an anhydro correction of 0.90 (g/L) (Sluiter et al., 2010). Substituting equation 25 into 24 and applying the steady state condition ( 𝑑[𝐶𝐸∗] 𝑑𝑡≈ 0) (Shen and Agblevor, 2008b) according to the Michaelis- Menten scheme. This assumption has been widely and successfully used to simplify complex hydrolysis reactions 85 and generate extremely good fit s to experimental data (Bansal et al., 2009; Shen and Agblevor, 2008a; Shen and Ag blevor, 2008b; Zhang et al., 2010) . The enzyme - cellulose complex concentration is [𝐶𝐸∗] =[𝐸]([𝐶0]− [𝐶𝑒]− [𝐺])[𝐸] + 𝐾𝑒 26 𝐾𝑒 =𝑘−3 + 𝑘4𝑘3 27 where Ke is the equilibrium constant (g/L). From e quation 22 and 23, the cellobiose and glucose production rate are : 𝑑[𝐶𝑒]𝑑𝑡= 𝑘4[𝐶𝐸∗]− 𝑘5[𝐶𝑒][𝐵] 28 𝑑[𝐺]𝑑𝑡= 𝑘5[𝐶𝑒][𝐵] 29 For industrial lignocellulosic ethanol production, cellobiose inhibition will most likely be eliminated, either by using a balanced enzyme cocktail or by adding excess of β - glucosidase (Li et al., 2008). In agreement with the expected industrial operating conditions and in order to avoid cellobiose inhibition, β - glucosidase was supplemented in excess . The observed cellobiose concentrations during hydrolysis at 5 wt % solids concentration were approximately zero and ranged from 0 to 8% of the glucose concentration at 10 wt % solids concentration. Therefore, cellobiose concentration was assumed to be negligible during the reaction and to remain constant over time ( 𝑑[𝐶𝑒] 𝑑𝑡≈ 0). Substituting equation 29 in 28, yields in equation 30: 𝑑[𝐺]𝑑𝑡= 𝑘4[𝐸]([𝐶0]− [𝐺])[𝐸] + 𝐾𝑒 30 If the concentration of β - glucosidase is not enough to efficiently convert cellobiose to glucose, then the cellobiose conversion rate must be implemented in the model together with the β - glucosidase inhibition and precipitation rate, significantly complicating the mathematical solution of the model. Equation 30 is similar to previously published models (Shen and Agblevor, 2008a; 2008b) , however, in the present model, term E refers only to cellulases (EG and CBH), assuming an excess of β - glucosidase and not to the system EG, 86 CBH and β - glucosidase. Theref ore, the cellulase mass balance during enzymatic hydrolysis is: [𝐸0] = [𝐸] + [𝐶𝐸∗] + [𝐸𝐿∗] + [𝐸𝑑] 31 where [ E0] (g/L) is the initial concentration of cellulase, [ EL* ] (g/L) is the concentration of the enzyme- lignin complex formed by the irreversible non- productive adsorption of enzyme on lignin and [ Ed] (g/L) represents the concentration of deactivated cellulases due to temperature (Chylenski et al., 2012; Farinas et al., 2010) , enzyme - glucose (Lee, 2007; Maurer et al., 2012) or enzyme - xylooligomers (Qing et al., 2010) interactions. Due to β -glucosidase supplementation, cellobiose inhibition during hydrolysis was negligible and thus does not affect [E d] . Figure 28 showed that a significant fraction of enzyme was irreversi bly adsorbed during the first two hour s of reaction for all tested scenarios, which agrees with Zheng’s conclusions (Zhang et al., 2010) . Given this observation, it was assumed that the enzyme - lignin complex (EL*) is formed before hydrolysis starts and remains approximately constant throughout the reaction. Therefore, the effective initial concentration of cellulase available for the reaction, [ ER] (g/L), is given by [𝐸𝑅] = [𝐸0]− [𝐸𝐿∗] 32 The enzyme - lignin complex formation will depend on the concentration of lignin [ L] (g/L) in the system and was defined as [𝐸𝐿∗] = [𝐿] ∗ 𝐹𝐿 3 3 where FL is a lignin factor that represents the amount of e nzyme adsorbed on lignin (g cellulase/g lignin). This factor has been explored in the past by our group (Pope, 2011) , However, by being able to follow the cellulases concentration in solution (Figure 28), it is possible to assume that the cellulases adsorption on lignin mainly occurs at the start of the reaction and is defined by the lignin adsorption ca pabilities . Despite the large amount of cellulases that are irreversibly adsorbed on lignin, past models did not consider non -productive enzyme adsorption on lignin (Shen and Agblevor, 2008a ,2008b; Zhang et al., 2010). With th e addition of the lignin factor, the proposed model accounts for lignin content which is defined by pretreatment conditions. Therefore, the proposed kinetic model links pretreatment and enzymatic hydrolysis. Combining e quations 31, 32, and 33 : 87 [𝐸𝑅] = [𝐸0] − [𝐿] ∗ 𝐹𝐿 = [𝐸] + [𝐶𝐸∗] + [𝐸𝑑] 34 Therefore, the change in concentration of active cell ulases is: 𝑑[𝐸]𝑑𝑡= −𝑑[𝐸𝑑]𝑑𝑡 3 5 The rate of cellulase deactivation (Ed) was assumed to follow a second order reaction with a rate constant kd (L/g h) (Shen and Agblevor, 2008a, 2008b; Zhang et al., 2010) : 𝑑[𝐸]𝑑𝑡= −𝑘𝑑[𝐸]2 3 6 Integrating equation 36 with the initial condition [ E] = [ ER] at t=0 , gives 𝐸 =𝐸𝑅1 + 𝐸𝑅𝑘𝑑𝑡 3 7 Substituting equation 37 and 32 in 30, and integrating it with the initial condition [ G] =0 at t=0 , yields [𝐺] = [𝐶0]�1− �1 +𝐾𝑒𝑘𝑑([𝐸0]− [𝐿]𝐹𝐿)𝑡𝐾𝑒 + ([𝐸0]− [𝐿]𝐹𝐿)�−𝑘4𝐾𝑒𝑘𝑑� 38 Equation 38 predicts the concentration of glucose during enzymatic hydrolysis based on hydrolysis time, cellulase loading and system lignin content. Following equation 38, when time tends to infinity, the concentration of glucose equals the initial concentration of cellulose. However, due to the inhibiti on of the enzymes, thermal deactivation, substrate characteristics and substrate depletion , complete cellulose conversion may not be achieved (Arantes and Saddler, 2011). Therefore, a more detailed modelling of enzyme inhibition and deactivation, as well as the substrate characteristics are required to overcome the limitations of equation 38. The model is one of the first to include lignin content as a key factor in diminished cellulose conversion. In addition, the proposed model has a simple mathematic solution and requires only 4 parameters which faci litates its use in process simulations and optimizations. 88 3. 3. 3.2 Cellulose hydrolysis model fitting and rate constant determination The proposed kinetic model was fit to the experimental data obtained from scenarios M20- 5 and M40- 5. The determined kinetic parameters are presented in Table 8 and the fit of the kinetic model to experimental results is shown in Figure 28. The calculated lignin factor is the same order of magnit ude as the maximum adsorption capacity of lignin experimentally measured by Kumar et al. (2009b) for multiple pretreatment technologies (57 - 127 mg enzyme/g lignin). The kin etic model has a good fit to the experimental data as reflected by the high correlation factor (R 2=0.99). Analysis of the residuals showed that the model underestimated glucose concentration by 0 to 17% in the first 2 h of reaction after which glucose was overestimated by 0 to 7%. In order to determine the model predictive capacity at different lignin contents, the model was used to predict glucose yields for scenarios S20- 5 and S40- 5. The model underestimates the production of glucose when the reaction sta rted to decelerate in the scenarios S20- 5 and S40- 5 (Figure 28). This may be caused by the lower lignin content in substrate S which increases the accessibility of cellulose and the reaction rate. The model predict ed the glucose production of substrate S (4.7% lignin content) with a high correlation coefficient (R 2= 0.93) . The predicted glucose concentrations were 0 to 20% lower than the experimental values in t he first 20 h of the reaction, after which the differenc es dropped to 0 to 10%. Residuals plots are shown in Appendix A ( Figure 58A). The predicted and experimental glucose concentrations are shown in Figure 28. 89 Table 8. Kinetic parameters determined for hydrolysis of cellulose by fitting equation 38 to scenarios M20- 5 and M40- 5. P arame t er Sym bol Uni t s Parame t er s det e rmi ne d from M20- 5 and M40 - 5 Equili brium constant K e g/ L 16.16± 0.25 Enzyme desorption rate constant k 4 1/ h 2.58±0. 43 Enzyme deactivation rate constant k d L/ g h 0.33± 0.19 Lignin factor FL g cellulase/ g lignin 0.09±0. 06 Correlation coefficient R2 0.99 K i ne ti c model apply to S20- 5 and S40 - 5 Correlation coefficient R2 0.93 K i ne ti c model apply at 10 w t % sol i ds conc e ntr ati on Effectiveness factor η 0.79±0.10 Correlation coefficient R2 0.93 At 10 wt % solids concentration, the predicted concentration of glucose during hydrolysis was overestimated by 57% on average (fit not shown). Similar expressions proposed by Shen and Agblevor (2008a, 2008b) and Zhang et al. (2010) were developed at a relatively low solids concentrations (1 to 5 wt %) and were not evaluated at higher solid concentrations. The model proposed by Zheng et al. (2009) successfully predicted hydrolysis yields for solid concentrations of 4 to 12 wt %, lik ely because the model had seven independently determined adsorption parameters and nine parameters obtained by model fitting . However like the model in this paper, Zheng’s model more accurately predicted glucose concentration at low solids concentration. This was attributed to product inhibition and mass transfer limitations but this behavior was not discussed further by Zheng et al. (2009) . Due to the high density, water mobility at 10 wt % solids concentration is limited. Consequently, the mass transfer of water soluble compounds is thought to be restricted. Diffusion limitations restrict the attack of enzymes on new catalytic, resulting in a non uniform concentration system. Spots with high enzymes concentration and sugars production results in enzyme inhibition (Hodge et al., 2008). It has been proposed that glucose reduces enzyme adsorption to lignocellulosic 90 substrates during high solid hydrolysis, leading to a decrease in the hydrolysis yield (Kristensen et al., 2009) . Pribowo (2014) reported that g lucose reduces enzyme adsorption at glucose concentrations above 100 g/L .In agreement with Pribowo’s results, Maurer et al. (2012) reported that glucose does not affect the adsorption properties of cellulases at low glucose concentration. At the glucose concentrations obtained in this thesis, glucose inhibition of cellulases adsorption is unlikely . Therefore, it is likely that e nzyme diffusion/adsorption and inhibition are the limiting factor of hydrolysis at high solids concentration. To account for the decrease in the overall reaction rate as a result of diffusion limitations, a diffusion factor (η) was defined : 𝜂 =Actual hydrolysis rate Hydrolysis rate if diffusion is infinitively fast=𝑟𝐺𝑟𝐺𝑠 3 9 If soluble compounds mobility in the system is not limited, then the reaction rate without diffusion limitations (rGs) is given by equation 30. Therefore, introducing equation 30 in equation 39, the diffusion limited reaction rate (rG) is: 𝑟𝐺 = 𝜂(𝑟𝐺𝑠) = 𝜂 �𝑘4[𝐸]([𝐶0]− [𝐺])[𝐸] + 𝐾𝑒� 40 The diffusion factor can also be included in equation 38, where 𝜂 = 1 for reaction conditions without diffusion limita tions. The diffusion factor was calculated by least squares regression . This factor accounts is an empirical factor which represents the limited mobility of the liquid phase at high solids concentrations that leads to gradients of concentration, enzyme deactivation and accumulation of cellobiose . Equation 38 using Ke, k2, kd and FL from Table 8, was fit to hydrolysis data at 10 wt % solids concentration; the resulting diffusi on factor has a value of 0.79. With the introduction of 𝜂, equation 38 has a good fit to the experimental data (R2=0.93) for the hydrolysis at 10 wt % solids concentrations as shown in Figure 28. The model overestimated glucose production by 27 to 60% in the first 10 h of hydrolysis in scenarios M20- 10 and S20- 10, however cellulose conversion during this period is less than half of the maximum conversion. After this period, the predicted glucose concentrations are within 4 to 20% of the measured concentrations for all scenarios operating at 10 wt % solids concentration. The diffusion limitations in heterogeneous reactions under conditions where mixing was inadequate have been studied with the fractal like kinetics which considers the 91 diffusion of particles in restrictive spaces or fractal surfaces (Kopelman, 1986) . The fractal systems has been studied with Monte Carlo simulations (Berry, 2002; Kopelman, 1986) , and, it has only been applied to the hydrolysis of cellulose with an empirical equation (Väljamäe et al., 2003) . The impl ementation of the fractal kinetic to the actual model is object of future investigations. Kristensen et al. (2009) observed a linear decrease in cellulose conversion at increasing solids concentrations for the hydrolysis of filter paper. This behaviour is in agreement with the addition of the effectiveness factor. Rosgaard et al. (2007) reported a decrease in the hydrolytic broth viscosity over time during hydrolysis of steam - pretreated barley straw; in the f irst 10 h of hydrolysis broth viscosity decreased by approximately 80% . This decrease in viscosity at the start of the reaction may be related to the model’s overestimation of glucose production during the first 10 h, after which, predicted sugar concentra tion is in better agreement with experimental data. The second order Akaike’s information criterion (AICc) meas ures the quality of a model and its fit to a set of data. AICc can be used to select the model that fits better a set of data with a minimum number of parameters (Hurvich and Tsai, 2007) . AICc was calculated for the proposed model with (AICc=266) and without (AICc=674) the addition of the effectiveness factor. Based on these results, it was concluded that the addition of the effectiveness factor to the proposed model substantially improves the predictive capabilities of equation 38. In order to better understand hydrolysis limitations at high solids concentrations equation 38 was independently fit to the hydrolysis data at 10 wt % solids concentration t o study the effect of diffusion limitation s on the parameters proposed. The kinetic parameters were determined by fitting equation 38 to scenarios M20- 10 and M40- 10, the obtained parameters are shown on Table 9 . The determined parameters were then used to calculate glucose production in scenarios S20- 10 and S40- 10. Experimental and predicted glucose productions in S20- 10 and S40- 10 also had a good correlation (R2= 0.94) . Most of the predicted concentrations differ ed from the experimental data by 20% while 9% of the predicted glucose concentrations differed from experimental concentrations by 20 to 40% ( Figure 60) . 92 Table 9 . Kinetic parameters determined for the hydrolysis of cellulose by fitting equation 38 to scenarios M20- 10 and M40- 10. P arame t er Sym bol Uni t s Parame t er s det er mi ne d from M20 - 10 and M40 - 10 Equilibrium constant K e g/ L 17.07±5.94 Enzyme desorption rate constant k 4 1/ h 1.79±0.49 Enzyme deactivation rate constant k d L/ g h 0.39±0.19 Lignin factor FL g cellulase/ g lignin 0.11±0.10 Correlation coefficient R2 0.99 K i ne ti c model apply to S20- 10 and S40 - 10 Correlation coeff icient R2 0.94 Correlation coefficients obtained at 5 and 10 wt % solids concentration were similar. However, it was found that K e increases and k4 decreases as solids concentration increases. The decrease in k 4 is believed to be due to the accumulation of cellobiose , which inhibits cellulases. Cellobiose accumulation at 10 wt % solid concentration may be caused by the high concentrations of glucose at certain points in the system which inhibits β - glucosidase (Kristensen et al., 2009) . An increase in Ke is due to an increase in k-3 or a decrease in k3. An increase in k-3 represents an increase in the desorption of cellulase s from cellulose or an increase in the non - productive cellulase- lignin complex formation (Maurer et al., 2012; Väljamäe et al., 1998) . A decrease in k3 reflects the difficulty of cellulases to reach and effectively adsorb on cellulose , this is in agreement with the diffusion limitations expected at 10 wt % solid s concentration (Bansal et al., 2009) . The deactivation constant (kd) obtained at 10 wt % solid s concentration is slightly higher than at 5 wt % solid s concentration. This may be caused by the high enzyme -xylooligomer (Qing et al., 2010) and enzyme - glucose (Maurer et al., 2012) interactions at 10 wt % solids concentration where high sugar concentrations a re obtained. The lignin factor 93 estimated at 10 wt % solid concentration is slightly greater than the lignin factor calculated at 5 wt % solids. N onetheless, both lignin factors are of the same magnitude than the reported by Kumar et al. (2009b) . Differences in the parameters determined at 5 and 10 wt % solids concentration suggest diffusion limitations are the primary barrier to an efficient hydrolysis at high solids concentrations. 3. 3. 3. 3 Enzymatic hydrolysis of xylan Since xylose is the second major sugar produced from lignocellulose, it must be introduced in the economic analysis of the ethanol production process. T o model xylan yields during hydrolysis, the assumptions considered in the development of equation 38 were used. It was also assumed that xylan hydrolysis does not interfere with cellulose hydrolysis. The hydrolysis of xylan is represented by: 𝐸 + 𝐹𝑘3𝑥��𝑘−3𝑥�⎯�𝐹𝐸∗𝑘4𝑥�� 𝐸 + 𝑋 41 where active cellulases (E) in solution are adsorbed onto xylan active sites (F) to form the enzyme - xylan complex ( FE*) from which xylose (X) is released. In equatio n 41, k 3 x (L/h g), k - 3 x (1/h) and k4x (1/h), are the rate constants of enzyme adsorption, enzyme desorption and xylose production, respectively . EG is the only cellulase previously show n to have activity on xylan (specifically EG I) (Lawoko et al., 2000) , therefore, xylan is primarily hydrolysed by cellulase EG present in Cellulast 1.5L . This hypothesis is supported by the enzyme recycling results reported in C hapter 4, where high xylan hydrolysis yields were obtained by recycling only cellulases without xylanases. This is discussed in great detail in section 4.3.2 (pp 110). Since the individual concentration of EG in Celluclast 1.5L has not been measured, total cellulases concentration was used to model xylan hydrolysis. The enzyme - cellulose complex format ion rate and substrate mass balance are shown in equations 42 and 43. 𝑑 [𝐹𝐸∗]𝑑𝑡= 𝑘3𝑥[𝐹][𝐸]− 𝑘−3𝑥[𝐹𝐸∗]− 𝑘4𝑥[𝐹𝐸∗] 42 94 [𝐹] = [𝐹] − [𝐹𝐸∗]− [𝑋] 43 where [ F0 ] is the initial concentration of xylan (g/L). Substituting equation 43 into 42 and applying the steady state condition ( 𝑑[𝐹𝐸∗] 𝑑𝑡≈ 0) (Shen and Agblevor, 2008b), the enzyme -xylan complex concentration is : [𝐹𝐸∗] =[𝐸]([𝐹0]− [𝑋])[𝐸] + 𝐾𝑒𝑥 44 𝐾𝑒𝑥 =𝑘−3𝑥 + 𝑘𝑥4𝑘3𝑥 45 where K ex is the equilibrium constant (g/L). Substituting equation 37 and 32 in 44, and integrating it with the initial condition [ X ] =0 at t=0 , yields [𝑋] = [𝐹0]�1− �1 +𝐾𝑒𝑥𝑘𝑑([𝐸0]− [𝐿]𝐹𝐿)𝑡𝐾𝑒𝑥 + ([𝐸0]− [𝐿]𝐹𝐿)�−𝑘4𝑥𝐾𝑒𝑥𝑘𝑑� 46 Equation 46 w as fit to the experimental data for hydrolysis of xylan in substrate M at 5 and 10 wt % solids concentration independently . The value of k d and FL calculated for equation 38 (Table 9 ) were used to determined Kex and k4x as cellulases deactivation occurs in the same system. Estimated parameters for the hydrolysis of xylan are presented in Table 10. Table 10. Kinetic parameters for the hydrolysis of xylan . P arame t er Sym bol Uni t s Parame t er s det er mi ne d from M20 and M 40 5 wt % solids concentration 10 wt % solids concentration Equilibrium constant K ex g/ L 17.89±1.57 31.17± 3.14 Enzyme desorption rate constant k 4x 1/ h 3.63±0.32 3.24±0.30 Correlation coefficient R2 0.98 0.96 Equation 46 has a good fit to the experimental data as shown in Figure 30 and by the high correlation coefficients shown in Table 10. The residuals analysis showed that 11% of the 95 predicted xylose concentrations differed from experimental data by 20 to 66% , while the rest of the predicted concentrations differed by 20% . Normalized residuals plots are provided in Appendix A ( Figure 61) . The implementation of the diffusion factor to equation 46 predicted with similar accuracy the hydrolysis of xylan in scenarios M20 - 10, M40- 10 and S40- 10. However, it has a poor fit to the experimental data in scenario S20- 10 (R2=0.48), where all the cellulases remained adsorbed during the entire reaction, Figure 28. This may be due to the higher affinity of cellulases and xylanases for xylan than for cellulose (Qing and Wyman, 2011), especially in scenario S20 - 10, where the amount of enzymes loaded is not enough to saturate the substrate. Due to the lack of information about the xylanases system and the hydrolysis of xylan , it is difficult get a conclusion from the results obtained. It was considered that the diffusion factor do not capture the entire process behind the hydrolysis of xylan. Consequently, the values presented in Table 10 were used to model the production of xylose during hydrolysis. 0 5 10 15 20 0 8 16 24 32 40 48 56 64 72 80 Xylose concentration (g/L) Time (h) M20 - 5 0 5 10 15 20 0 8 16 24 32 40 48 56 64 72 80 Xylose concentration (g/L) Time (hrs) M40 - 5 96 0 5 10 15 20 0 8 16 24 32 40 48 56 64 72 80 Xylose concentration (g/L) Time (h) S 20 - 5 0 5 10 15 20 0 8 16 24 32 40 48 56 64 72 80 Xylose concentration (g/L) Time (hrs) S 40 - 5 0 5 10 15 20 0 8 16 24 32 40 48 56 64 72 80 Xylose concentration (g/L) Time (hrs) M20 - 10 0 5 10 15 20 0 8 16 24 32 40 48 56 64 72 80 Xylose concentration (g/L) Time (hrs) M40 - 10 97 Figure 30 . Xylose concentration during enzymatic hydrolysis from experimental data ( ) and proposed kinetic model, e quat ion 46 ( ). Pretreatment conditions: M= 30 min, 6 wt % NaOH/ dry biomass, 120°C and S= 60 min, 10 wt % NaOH/ dry biomass, 150°C . Due to the high cost of enzymes, optimization of enzyme loading can be highly advantageous for the overall process economy (Arantes and Saddler, 2011) . Therefore, t he proposed hydrolysis models offer the possibility to optimize the enzyme loading as they can be used to model different pretreatment and hydrolysis conditions. 3.4 Conclusions Reducing the costs of enzymatic hydrolysis is essential to the commercialization of lignocellulosic ethanol. Kinetic models of enzymatic hydrolysis help frame hydrolysis studies and support the development of techno - economic analyses. Enzymatic hydrolysis of pretreated wheat straw was conducted under a range of conditions to support the 0 5 10 15 20 0 8 16 24 32 40 48 56 64 72 80 Xylose concentration (g/L) Time (hrs) S 20 - 10 0 5 10 15 20 0 8 16 24 32 40 48 56 64 72 80 Xylose concentration (g/L) Time (hrs) S 40 - 10 98 development of a model that describes the impact of residence time, enzyme loading, solids concentration and lignin content on enzymatic hydrolysis and that can be easily implemented in process simulations. In agreement with past studies, it was found that lignin is a major barrier to efficient enzymatic hydrolysis reaction. Consequently, it is imperativ e to account for the effects of lignin in enzymatic hydrolysis models. Moreover, enzyme loadings should be defined by both the cellulose and lignin content of the substrate. Lignocellulosic enzymatic hydrolysis is a complex process that involves numerous steps. However, by applying assumptions based on experimental observations and industrial hydrolysis conditions, the system was simplified and a kinetic model relating the production of glucose to residence time, enzyme loading, and lignin content was obta ined. Due to its relatively simple form, the use of only five parameters, and the inclusion of lignin content, the proposed model can be easily used in simulation software for techno - economic analysis and can thus support studies examining the effect of hy drolysis time, pretreatment conditions, and enzyme loading on enzymatic hydrolysis as well as on fermentation and separation processes. Applying the assumptions used to develop the cellulose hydrolysis model, a similar equation was proposed to predict t he hydrolysis of xylan in the same system. The predicted xylan hydrolysis model fit the experimental data. This model will allow the modeling of xylan hydrolysis in process simulations and subsequent economic analysis. 99 4.1 Introduction Given the high cost of enzymes, enzyme recycling technology is a promising process for decreasing ethanol production costs (Galbe and Zacchi, 2002) . Enzyme s can be recovered after hydrolysis, fermentation, SSF or distillation. The recycling of enzymes after distillation is possible when distillation is performed at low temperatures under vacuum (Eckard et al., 2013; Lindedam et al., 2013) . Lindedam et al. (2013) studied the enzyme recycling after SSF, SHF and distillation configuration at laboratory and pilot- plant scale usin g wheat straw. The enzymes activity in the supernatant after SHF and SSF was up to 20% of the initial activity at laboratory scale. This shows that the enzymes remain active after the SHF or SSF processes, however , the recycling of the whole slurry, with t he produced ethanol, is problematic to develop a continuous process . The potentially activity of cellulases available for recycling in the whole slurry after SSF, in a pilot plant, was around 40 to 50% of the initial activity (Lindedam et al., 2013) . The addition of the distillation, step necessary to separate ethanol from the fermentation broth, decreased the enzyme activity in the fermentation broth by up to 50% . Since the objective of this chapter is to develop a mass balance of the enzyme recycling process, and cons idering the relatively simplicity of the hydrolytic broth when comparing with the fermentation media, the recovery of the enzymes after hydrolysis was selected as the f ocus of this thesis. Enzyme recovery after hydrolysis can be performed by ultrafiltrat ion, recycling solid residues with enzymes and adsorption of enzymes on fresh substrate. By recovering the insoluble residues and place them with new substrate in new hydrolysis round, the production of sugars in the second hydrolysis increased by 20% with respect to the first hydrolysis (Weiss et al., 2013) . Under this methodology, the solids and lignin concentration in hydrolysis increase with each round, leading to enzyme deactivation and large reactor 4 E v a l u a t i o n o f e n z y m e r e c o v e r y b y a d s o r p t i o n d u r i n g h y d r o l y s i s o f p r e t r e a t e d w h e a t s t r a w 100 volume requirements. Jin et al. (2012) have proposed a new configuration which may solve the accumulation of solids by recycling the solid, however, the effect negative of the increasing lignin concentration in different systems has to be evaluated. T he use of ultrafiltration to recover enzymes using the AFEX pretreatment reported cost savings of approximately 15% (Steel e et al., 2005) . Despite the reported savings, this method is limited by its cost and the eventual fouling of the filter membrane. The potential to recycle enzymes by adsorption and obtain high cellulose conversions (48-93%) in up to three subsequent rounds of hydrolysis has been demonstrated by several researchers using a variety of substrates (Kristensen et al., 2007; Lu et al., 2002; Qi et al., 2011; Steele et al., 2005; Tu et al., 2007a) . Unfortunately, β - glucosidase cannot be simultaneously recycled by adsorption as it either does not adsorb to the substrate (Tu et al., 2007a; Tu and Saddler, 2010) or adsorbs to a far lesser degree than cellulases (Kumar and Wyman, 2009b) . Therefore fresh β - glucosidase must be added at the beginning of each round of hydrolysis to avoid build- up of cellobiose and subs equent end - product inhibition of cellulase (Qi et al., 2011). The effectiveness of enzymatic hydrolysis and enzyme recycling depends on factors such as substrate lignin content after pretreatment, hydrolysis time, and enzyme loading. Cellulase activity during hydrolysis and the amount of cellulase available for recycling depends upon lignin content (Lu et al., 2002) as lignin irreversibly absorbs cellulases (Farinas et al., 2010; Jørgensen et al., 2007; Nakagame et al., 2010; Zhou et al., 2009b) . The amount of cellulases available for recycle changes with time as they adsorb onto the substrate and gradually desorb as cellulose is depleted (Lu et al., 2002; Maurer et al., 2012) . Cellulases desorption is proportional to the amount of adsorbed cellulases, which is defined primarily by enzyme loading (Maurer et al., 2012). In short, predicting optimum recycling conditions is a complex endeavor. In order to evaluate the technological and economic viability of enzyme recycling for ethanol production, it is necessary to determine the amount of enzyme that can be recycled as well as recycling’s potential to provide a uniform flow of sugar as part of a continuous process. The 101 objectives of this work are to determine the fraction of enzymes that can be recycled by developing refined material balances of both β - glucosidase and cellulases, and to evaluate the performance of recycled cellulases in subsequent hydrolysis rounds. Finally, the model for the Novozyme 188 protein precipi tation presented in C hapter 2 and empirical data on enzyme recycling will be combined to determine if it is possible to use enzyme recycling and batch hydrolysis to achieve a uniform mass flow rate of sugars from hydrolysis . 4.2 Methods The substrates, enzymes an d methodologies followed for the pretreatment, enzyme determination, substrate characterization (section 2.2, page 46) , enzymatic hydrolysis and cellulases concentration (section 3.2, page 72 ) presented in past sections were applied in the present chapter. 4.2.1 Enzymatic hydrolysis conditions The cellulose conversion achieved in each of the scenarios tested is presented in Figure 26, page 75. Due to the higher cellulose conversion obtained at 5 than at 10 wt % solids concentration, operating at 5 wt % solids concentration was hypothesized to be more advantageous than operate at 10 wt % solids concentration. Given these results, enzymatic hydrolyses in this chapter were carried out at 5 wt % solids concentration. The influence of reaction time on enzyme recycling was studied by conducting hydrolysis for three different times (t1, t2, t3 ). Based on the production of glucose during enzymatic hydrolysis shown in Figure 28, the selected times were chosen to correspond with different reaction stages: high conversion rate, deceleration of conversion rate, and approaching maximum conversion. Enzymatic hydrolysis conditions used in this chapter are summarized in Table 11. 102 Table 11. Enzymatic hydrolysis experimental conditions at 5 wt % solid conc entration. Pretreatment conditions: M= 30 min, 6 wt % NaOH/ dry biomass, 120°C and S=60 min, 10 wt % NaOH/ dry biomass, 150°C . Sc e nari o Hydr ol ysi s tim e (h ) M20- 5 24 48 72 M40- 5 5 24 48 S20- 5 12 24 48 S40- 5 5 24 48 4.2.2 Enzyme recycling A schematic diagram of enzyme recycling methodology is shown on Figure 31 . To interrupt enzymatic hydrolysis at the selected reaction time, the filtrate and residual substrate were separated using a glass microfiber membr ane (Whatman GF/A). Samples of 0.5 mL were taken before and after filtration and centrifuged (RCF 16,904 g, 10 min) to withdraw the supernatant. A 0.1 mL sample of supernatant was used to determine protein content and remaining supernatant was kept at - 20°C for future compositional analysis. Fresh substrate equivalent to the amount initially added was suspended in the filtrate liquor in a 250 mL flask in the shaker at 50°C and 150 rpm for 20 min; sugar produced during adsorption was included in mass balance calculations. In previous works, enzyme adsorption was carried out at 4 or 25°C for 1.5 to 2.5 h (Qi et al., 2011; Tu et al., 2007a; Tu et al., 2007b; Tu et al., 2009b; Tu and Saddler, 2010) . However, as shown in the Chapter 3, the majority of cellulases are adsorbed on substrate in the first hour of enzymatic hydrolysis; therefore, it was believed that 20 min was sufficient to achieve high enzyme adsorption. An additional 103 consideration is that the low temperature adsorption process implies a high energy cost to cool the hydrolysis liquor for adsorption and then reheat it for subsequent saccharification and fermentation, to avoid these costs during the economic analysis in chapter 5, the adsorption process was conducted at 50°C. After adsorption, filtration with Whatman GF/A was repeated. Samples of 0.5 mL were taken befor e and after filtration to determine protein and sugar concentration in the liquor and to calculate the mass of enzymes adsorbed on the substrate. The fresh substrate with enzymes bound on its surface was then used for a second round of enzymatic hydrolysis . In the first set of experiments, the second hydrolysis was carried out by suspending the fresh substrate with the adsorbed cellulases in fresh buffer. Fresh β - glucosidase was added to the second hydrolysis at the same loading as the first hydrolysis as shown in Figure 31A. In this set of experiments, each r ound of hydrolysis lasted 48 h for all scenarios. Sugar and enz yme concentration in the liquid phase were determined at the end of the second round of hydrolysis. In order to implement enzyme recycling at an industrial scale, it is necessary to determine the conditions required to maintain uniform sugar and enzyme concentrations at the end of each hydrolysis round. In the second set of recycling tests, the second hydrolysis was carried out by suspending the substrate with adsorbed cellulases in buffer with sufficient fresh β -glucosidase and cellulases added to match the initial enzyme concentration in the first hydrolysis (Figure 31 B). Sugar and enzyme concentration in the liquid phase were determined at the end of the second hydroly sis. 104 Figure 31. Schematic diagrams of enzyme recycling methodology without (A) and with (B) addition of fresh cellulase at the start of the 2 nd hydrolysis round. The addition of fresh cellulase to enzymatic hydrolysis was used t o evaluate the technical feasibility of a continuous process. 4.2.3 Accessible liquid in substrate Cellulases and β - glucosidase have been reported to be similarly sized (Grethlein, 1985; Workman and Day, 1982) . The primary components in Celluclast 1.5L are exo - glucanases (CBH I, CBH II) and endo- glucanases (EG I, EG II) (Hu et al., 2013) , which have a molecular weight of 52- 62 kDa (Pribowo et al., 2012) while β - glucosidase has a weight of 80- 120 kDa (Alftrén and Hobley, 2013; Pribowo et al., 2012) . The molecular weight of Novozyme 188 components fall in a similar range, 50- 120 kDa (Alftrén and Hobley, 2013) . Enzymatic HydrolysisFiltrationRe-adsorptionFiltrationSugars, β-glucosidase and unrecovered cellulasesβ-glucosidaseand bufferSugars, β-glucosidaseand unbound cellulasesSubstrateSolid residueSubstrate and bound cellulasesEnzymatic HydrolysisFiltrationRe-adsorptionFiltrationSugars, β-glucosidase and unrecovered cellulasesβ-glucosidaseand bufferSugars, β-glucosidaseand unbound cellulasesSubstrateSolid residueSubstrate and bound cellulases CellulasesAB105 As Novozyme 188 components do not adsorb to the substrate ( section 2.3.3, page 60) and have similar molecular weights, Novozyme 188 was used to estimate the fraction of liquid contained in substrate that is inaccessible to large molecules such as cellulases. Solutions of Novozyme 188 at three different protein concentrations, similar to those use d during enzymatic hydrolysis, were prepared ( 0.34, 1.71, and 2.65 g/L). A 5 g sample of wet substrate S (82.0% moisture content) was added to 5 mL of each Novozyme 188 solution and placed in a shaker at 20°C and 150 rpm together with control flasks conta ining 5 mL of each Novozyme 188 solution. Total protein concentration in each flask was monitored for one hour. The protein mass in the Novozyme 188- substrate and control flasks is equal but the enzyme - accessible liquid in the substrate causes the measured protein concentration in the Novozyme 188- substrate flasks to decrease. The concentration differences were used to calculate the volume of enzyme- accessible liquid in substrate. 4. 3 Results and discussion The mass of cellulases recycled is determined from the difference in cellulases mass in solution before and after cellulases adsorption on fresh substrate. However, dilution of enzymes by the liquid contained in the substrate must be considered in this calculation as the pretreated substrates have a moistu re content of approximately 82%. The decrease of total protein concentration due to dilution can be easily mistaken by the reduction due to adsorption. However, not all liquid in the substrate is accessible to enzymes due to the substrate pore size distrib ution. The diameter of cellulases and β - glucosidase has been reported to be approximately 5.9 nm (Grethlein, 1985; Workman and Day, 1982) ; pores smaller than this are inaccessible to enzymes. For example, Grethlein et al. ( 1985) reported that only 65% of the pore volume in pretreated mi xed wood were accessible to 5.1 nm solutes. The pore size distribution varies with biomass and pretreatment (Grethlein, 1985) and cannot be easily predicted a priori . In order to estimate the percent of enzyme accessible liquid in the substrate, the dilution of Novoz yme 188 protein content by the substrate liquid was quantified at three different concentrations. From these tests, it was determined that 82% 106 of the biom ass liquid is enzyme accessible; these results are presented in the Appendix A, section A.4 (page 221) . After cellulases adsorption, the substrate is separated from the hydrolysis liquor in order to start a new round of hydrolysis. The liquid contained in the substrate also contains non-adsorbed cellulases (Figure 32) as well as sugars, β - glucosidase and proteins from Novozyme 188. These sugars and enzymes must also be taken into account in mass balance calculations. Figure 32. Cellulase s distribution on fresh substrate after adsorption process 4.3.1 Effect of hydrolysis time on enzyme recycling In Chapter 3, it was found that cellulases concentration in solution varied with time and with pretreatment conditions. Substrates M and S exhibited different cellulases adsorption profiles during hydrolysis. The different cellulases adsorption profiles were likely due to their different lignin contents. Substrate M, which contains more lignin, adsorbed a larger amount of cellulases than substrate S. In addition, the desorption rate of cellulases from substrate S was faster than in the case of substrate M. Given these differences, the fraction of cellulases 107 that can be recovered by adsorption at three different hydrolysis times was measured for each substrate (Figure 33). The percent of recycled cellulases is defined as: Recycled cellulases (%) = 100EL − EReEL 47 where EL (g) is the mass of cellulases initially loaded and E Re (g) is the mass of cellulases recycled. The mass of cellulases during the reaction was calculated from the concentration of cellulases in solution, considering the change in Novozyme 188 protein concentration. Cellulases recycling using new commercial cocktails, such as Cellic® CTec2 , is challenging since β- glucosidase is mixed together with cellulases and other proteins , complicating determination of the cellulases adsorption. Moreover, the adsorption of the β- glucosidase contained in Cellic® CTec2 complicates the system (Haven and Jørgensen, 2013) . Therefore, the determination of the cellulases concentration changes will be more difficult. For M40, S20, and S40 the percent of recycled enzyme increased with time, likely due to the gradual desorption of cellulases from the substrate back to the liquid phase. Pretreatment severity and enzyme loading clearly affect the amount of enzymes recycled. Significantly more enzymes were recycled in S20 (0 to 19% recycled cellulases ) than in M20 (0 to 2% recycled cellulases), likely due to the low lignin content of substrate S (4.7% lignin content). A greater fraction of cellulases were recycled in scenarios M40 and S40 (5 to 35%) than in scenarios M20 and S20 (0 to 19%). Doubling enzyme loading saturates the substrate surface with cellulases, increasing the amount of cellulases in solution available for rec ycling. High enzyme loadings ( e.g. 40 FPU/g cellulose) enable faster hydrolysis than low loadings ( e.g. 20 FPU/g cellulose) but m ay result in increased operating costs. Although the percent of recycled cellulases increases with enzyme loading, the total mass of cellulases lost is also greater thus the increase in enzyme loading may not be justifiable. Finally, the very low enzyme re cycled in M20 demonstrates that enzyme recycling is not always advantageous and thus each pretreatment/hydrolysis configuration must be evaluated before implementing enzyme recycling. 108 109 Figure 33. Percent of r ecycled cellulases achieved in each scenario at different times referred to initial amount of cellulases loaded in the first round. The mass fraction of enzymes that can be recycled has not been previously reported, making it difficult to compare our resu lts with past studies. In past enzyme recycling by adsorption studies, enzyme recycling performance was reported as a function of cellulases activity after recycling (Lu et al., 2002; Steele et al., 2005) , hydrolysis yields obtained in subsequent round of hydrolysis (Tu et al., 2007a; Tu et al., 2009b) or the difference in total protein concentration before and after adsorption (Qi et al., 2011; Tu et al., 2007a) . Enzyme recovery for different substrates has been reported in some studies and is defined as: Enzyme recovery (%) = 100[TH]− [TA][TH] 48 where [T H] (g/L) is the total protein conc entration (cellulases, β - glucosidase and cocktail proteins) at the end of hydrolysis and [T A] (g/L) is the total protein concentration after adsorption. Enzyme recovery (%) refers to total protein concentration which includes cellulases, β - glucosidase and cocktail proteins . Tu et al. (2007a) reported an enzyme recovery of 88% for ethanol - pretreated mixed softwood (EPMS) after 24 h of hydrolysis. Similarly, Qi et al. (2011) achieved 74.8% enzyme recovery for dilute alkali pretreated wheat straw after 48 h of hydrolysis. In the present work, the enzyme recovery percent (equation 48) 110 was not used, instead, the percent of cellulases recovered after hydrol ysis by adsorption is reported : Cellulases recovery (%) = 100[EH]− [EA][EH] 49 where [E H] (g/L) is the cellulases concentration in the bulk liquid at the end of the hydrolysis and [E A] (g/L) is the cellulases concentration in the bulk liquid after adsorption process . Conditions in scenario S20 (enzyme loading, hydrolysis time, lignin content) are the closes t to those reported by Qi et al. (2011) ; a cellulases recovery of 56% was achieved in our study as shown in Figure 34. Although cellulases recovery refers only to the recovered cellulases by adsorption and excludes cellulases and Novozyme 188 proteins contained in the substrate liquid , it is in close agreement with Qi’s results (Qi et al., 2011) . 4.3.2 Enzyme recycling without cellulase supplementation Enzyme recycling was initially carried out without adding fresh cellulases to the s econd hydrolysis in order to investigate the performance of the recycled enzyme during hydrolysis. Conversions obtained in the first and second rounds of hydrolysis, each after 48 h , and the percent of cellulases recycled are shown in Figure 34. 71% 81% 2% 0% 10% 29% Cellulose conversion Xylan conversion Recycled cellulases Cellulase recovery M20 111 Figure 34. Cellulose and xylan conversion after 48 h as a percentage of theoretical maximum for the first ( ) and second ( ) hydrolysis rounds. Percent cellulases recycled by adsorption is calculated from equation 47 and cellulase recovery is defined by equation 49. For both substrates, increasing enzyme loading did not substantially increase cellulose and xylan conversion during the first hydrolysis. This is in agreement with Arantes and Saddler's work which showed that cellulose accessibility is the reaction limiting factor (Arantes and 84% 87% 18% 70% 29% 40% Cellulose conversion Xylan conversion Cellulase recycled Cellulase recovery M40 75% 84% 16% 56% 33% 41% Cellulose conversion Xylan conversion Cellulase recycled Cellulase recovery S 20 91% 89% 32% 70% 71% 57% Cellulose conversion Xylan conversion Cellulase recycled Cellulase recovery S 40 112 Saddler, 2011). As substrate S has less lignin than substrate M, cellulose in substrate S is expected to be more accessible, which may explain the greater hydrolysis yields from substrate S. However, this does not mean that severe pretreatment conditions are the most advantageous conditions for ethanol production, since increasing severity also increases sugar loss. During pretreatment 10.3% and 25.8% of glucan, and 24.3% and 42.1% of xylan were solubilised under condition M and S, respectively. In addition, 2% to 2.5% of the glucose released by the first hydrolysis was lost during filtration. Optimization of glucose production with enzyme recycling must consider both of these losses. Enzyme recycling also provided unexpected insight into the activity of the enzymes in Novozyme's Celluclast 1.5L cocktail. In scenario S40, a relatively high cellulose conversion was a chieved during the second hydrolysis using only recycled cellulases. Seventy eight percent of the first hydrolysis yield was obtained in the second hydrolysis using only 32% of the initial cellulases. This indicates that the recycled enzyme reversibly adso rbs on cellulose and exhibits high activity. Lu et al. (2002) reported that 93% of the first hydrolysis yield was obtained in the second hydrolysis of steam - explo ded Douglas- fir pretreated by hot alkali peroxide by recycling 78% of the initial enzymes activity recovered from enzymes in solution and solid residue. This shows that a fraction of the initial enzyme loading can achieve a high hydrolysis yield in the second hydrolysis round. However, due to cellulases loss, cellulases need to be supplemented to achieve similar conversions in the first and second hydrolysis. The major cellulases in Celluclast 1.5L ( CBH I, CBH II, EG I and EG II) have been reported to reversibly adsorb (Goyal et al., 1991; Hu et al., 2013; Pribowo et al., 2012; Várnai et al., 2011; Zhang and Lynd, 2004) . The high xylan yields during the second round (Figure 34) are also interesting since T. reesei xylanases are unlikely to be responsible for the hydrolysis of xylan, as these enzymes do not adsorb in significant quantities to substrate and their activity is lost between 24 and 72 h (Pribowo et al., 2012) . Therefore, xylan hydrolysis during the second round is likely due to EG, specifically EG I, as it is the only cellulase enzyme previously shown to have activity on xylan (Lawoko et al., 2000) . Deeper characterization of the highly active recycled enzymes may yield valuable information for improvement of enzyme performance. 113 Key to the ultimate success of enzyme recycling is controlling enzyme concentration during each hydrolysis round; total enzyme concentration after each round of hydrolysis is shown in Figure 35 along with Novozyme 188 protein concent ration as predicted by equation 14. Figure 35 shows that the total enzyme concentration in solution at the end of the second hydrolysis was approximately equal to the predicted Nov ozyme 188 protein conce ntration. This suggests that all the recycled cellulases remained adsorbed onto the solid residue substrate (rich in lignin) at the end of the second round. This may be due to the adsorption-desorption of cellulases from cellulose and especially by the ir reversibly adsorption of cellulases on lignin. Figure 35. Total protein concentration (cellulases and Novozyme 188) at the end of the 1 st ( ) and 2nd ( ) hydrolysis round (48 h hydrolysis time), and Novozyme 188 protein concentration ( ) predicted by equation 14 for 48 h hydrolysis time. At the start of the first round of hydrolysis round some cellulases are adsorbed irreversibly on lignin, and thus do not participate in the hydrolysis reaction (A rantes and Saddler, 2011). A portion of the cellulases is reversibly adsorbed by the accessible cellulose (Maurer et al., 2012) and once the accessible cellulose is saturated, the balance of cellulases remain in solution. As the reaction proceeds, more lignin and cellulose are exposed and more cellulases are adsorbed. Once most of the accessible cellulose is hydrolyzed, the reaction approaches maximum conversion and cellulases concentration in solution becomes approximately constant. Without supplementation, the recycled cellulases are distributed between cellulose and lignin at the start of the second hydrolysis. As hydrolysis proceeds, cellulases desorb from cellulose and are then re - adsorbed to accessible cellulose or unsaturated lignin. Therefore, the recycled cellulases are ultimately, co mpletely adsorbed onto lignin, leaving 0.0 0.4 0.8 1.2 1.6 M20 M40 S20 S40 Enzyme concentration (g/L) 114 primarily Novozyme 188 in the liquid phase at the end of the second hydrolysis. This suggests that cellulases recycling for more than two rounds of hydrolysis is not viable for the presented system. It was reported for other systems (different substrates and pretreatments), that hydrolysis yield decreases with each recycling round (Lu et al., 2002; Qi et al., 2011; Tu et al., 2007b) . Based on the decrease in enzyme activity in each recycling round reported by Lu et al. (2002), it seems that, as described in this work, the hydrolysis yield decreases in each round due to cellulases loss caused by the presenc e of lignin. The difference in cellulases loss among systems may be caused by the different lignin structures. Therefore, fresh cellulases must be supplemented prior to each round to ensure consistent cellulases, β -glucosidase, and sugars concentrations at the start and end of each hydrolysis round. 4.3.3 Enzyme recycling with cellulase s supplementation Fresh cellulases and β - glucosidase were added prior to the second hydrolysis round in order to achieve the same enzyme loading as at the start of the first roun d. The masses of recycled cellulases adsorbed on substrate and in the liquid contained in fresh substrate were considered when calculating the mass of cellulases required for supplementation. Glucose, xylose and total protein concentrations after the fir st and second hydrolysis for scenarios S20 and S40 are shown in Figure 36. Concentration profiles of glucose, xylose, and total protein from Figure 28 are shown as lines. In Figure 36, the final glucose, xylose, and total protein concentrations at the end of the first and second hydrolysis are within the standard error of the concentration profiles of the uninterrupted hydrolysis. Cellulases and β - glucosidase supplementation was also performed for scenario M40 with similar results. Although almost no enzyme is recycled after the first round in scenario M20 (Figure 33 and Figure 34), the recycling procedure (filtration, adsorption, filtration, re - suspension with supplementation) was performed. In this scenario, the second hydrolysis is performed with essentially 100% fresh cellulases . The glucose, xylose, and total protein concentrations at the end of the first and second hydrolysis for all the reaction times tested were equal and concurred with the results reported in Figure 28 for non- interrupted hydrolysis . 115 Since glucose inhibits cellulases, it may be that not all the cellulases recycled by adsorption are active. Consequently, the activity at the start of the first and second hydrolysis could differ. Nonetheless, the glucose, xylose, and total protein concentrations at the end of both rounds at all tested times were equal and consistent with previous results for uninterrupted hydrolysis. Therefore, the initial activity in both hydrolysis rounds was considered to be similar. This confirms that a continuous production of a constant mass of glucose and xylose was possible despite filtration and the other interruptions or losses associated with the recycling procedure. These results demonstrate that the models in Chapters 2 and 3 can be used to predict the progress of enzyme recycling and that enzyme recycling by adsorption can be used to achieve consistent sugar production through rounds of recycling. 116 Figure 36. Glucose ( ), xylose ( ) and total enzyme ( ) concentration at the end of the 1st and 2nd hydrolysis rounds at different recycling times (t1, t2, t3). E xperimental concentrations of glucose ( ), xylose ( ) and total protein ( ) concentrations from Figure 28. Based on the experimental results and model assumptions, a mass balance of the enzyme recycling process was constructed. For the first time, cellulases distribution during hydrolysis and recovery was determined. The mass balance for scenario S40 with enzyme recycling after 48 h hydrolysis is shown in Figure 37. This mass balance shows that implementing enzy me recycling reduces the amount of cellulases fed to hydrolysis (required enzyme loading=cellulases in stream 4 + stream 6=0.43 g cellulases) by 36% (100*(cellulases in 0 0.5 1 1.5 2 2.5 3 3.5 4 0 5 10 15 20 25 30 0 8 16 24 32 40 48 56 64 72 80 Protein in solution (g/L) Sugars (g/L) Time (h) S 20 0.8 1.3 1.8 2.3 2.8 3.3 3.8 0 5 10 15 20 25 30 0 8 16 24 32 40 48 56 64 72 80 Protein in solution (g/L) Sugar (g/L) Time (h) S 40 t1 t2 t3 t1 t2 t3 117 stream 4)/0.43 g cellulases). In addition, the amount of Novozyme 188 fed to the system is reduced by 6% as some Novozyme 188 is recycled in the liquid contained in the substrate after adsorption. Most importantly, this mass balance can be used to simulate a continuous process for the production of ethanol with enzyme recycling and support e conomic evaluations of enzyme recycling. St r e am (g) 1 2 3 4 5 6 7 8 9 Water 43.66 221.14 192.77 28.37 149.99 3.16 181.51 4.03 177.48 Glucose 5.55 4.83 0.72 5.67 0.12 5.55 Xylose 2.48 2.16 0.32 2.54 0.06 2.48 Cellulases 0.27 0.11 0.15 0.28 0.43 0.16 0.27 Novozyme 188 0.08 0.07 0.01 0.15 0.16 0.08 0.08 Cellulose 5.32 5.32 0 5.32 0.86 0.86 0 Xylan 2.33 2.33 0 2.33 0.38 0.38 0 Lignin 0.46 0.46 0 0.46 0.46 0.46 0 Figure 37. Mass balance of the primary compone nts for 48 h enzymatic hydrolysis and enzyme recycling for scenario S40. 4.4 Conclusion Given the high cost of enzymes used for the production of lignocellulosic ethanol, the potential to reduce operating costs by recycling enzymes is a tantalizing proposition. To assess the viability of such an approach, an assessment of the amount and activity of recyclable enzymes as well as the conditions needed for a continuous process is needed. Enzymatic HydrolysisFiltrationRe-adsorptionFiltrationWaterPretreatmentEnzymesFermentation123 456789Solid residues118 Novozyme 188 protein precipitation, the dilution effect of liquid in t he substrate and the volume of enzyme accessible pores within the substrate were accounted for when determining the amount of cellulases recycled by adsorption. T he amount of cellulases recycled increased with decreasing lignin content, increasing cellulas es loading, and increasing hydrolysis time. The recycled cellulases were shown to be remarkably active during the second hydrolysis round; 71% cellulose conversion was obtained using 32% recycled cellulases. However, it was found that very little cellula ses remained in solution after the second hydrolysis round, likely due to irreversible adsorption onto lignin. Therefore cellulases supplementation is needed to support continuous enzyme recycle. Using a model of Novozyme 188 protein precipitation and ac counting for the liquid within the biomass, the amount of fresh cellulases and β - glucosidase required to achieve equal initial enzyme loading for each hydrolysis round was successfully calculated. By duplicating the initial conditions for each hydrolysis r ound, equal cellulose and xylan conversions were achieved during each hydrolysis round. Therefore, it was concluded that a uniform production of sugars can be achieved during the enzyme recycling process . Mass balances of such a process were developed for future techno - economic evaluations to identify optimum process conditions such as pretreatment severity, enzyme loading and recycle time and to evaluate the economic viability of enzyme recycling. 119 5 .1 Introduction Many pretreatment process alternatives for the production of ethanol from lignocel lulosic materials have been pro posed, making difficult to compare the advantages of each technology (Alvira et al., 2010). Enzymatic hydrolysis has been studied for a large number of substrates, pretreatments and hydrolytic conditions (Koo et al., 2011; Rosgaard et al., 2007; Yu et al., 2011), and as result, pretreatment and hydrol ysis have been significantly improved (Arantes and Saddler, 2011; Koo et al., 2011; Kristensen et al., 2009; Rosgaard et al., 2007; Tu and Saddler, 2010). Nonetheless, optimization of pretreatment and enzymatic hydrolysis is incomplete without considering their effect on fermentation and separation stages and ethanol cost. Lignocellulosic ethanol production requires further improvements to compete with gasoline prices (von Sivers and Zacchi, 1996) . Therefore, it is important to perform techno- economic modeling and analysis of the entire lignocellulosic ethanol process (Huang et al., 2009) . Reported e nzyme cost s vary from a cost contribution to the production of lignocellulosic ethanol from $0.10 to $0.35/gal ethanol (Aden et al., 2002; Galbe et al., 2007; Klein-Marcuschamer et al., 2010; Kumar and Murthy, 2011) . These enzyme costs are considered optimistic by Klein - Marcuschamer et al. (2012), who estimate a contribution of enzymes to ethanol cost as high as $1.47/gal ethanol using saccharification and ferm entation conversions reported in literature . Moreover, most of the published economic analyses are carried out at one set of conditions for enzymatic hydrolysis and pretreatment conversion, making it difficult to evaluate the effect of individual variables on the production cost. The identification of the variables that control the process economy and their interconnection is highly relevant to focusing research efforts on the most significant areas . 5 E c o n o m i c e v a l u a t i o n o f t h e p r o d u c t i o n o f l i g n o c e l l u l o s i c e t h a n o l 120 Pretreatment and enzymatic hydrolysis o perating conditions that result in the highest hydrolysis yields have been used to model enzymatic hydrolysis (80 to 100% cellulose conversion) (Galbe et al., 2007; Kazi et al., 2010; Piccolo and Bezzo, 2009; Sassner et al., 2008; Wingren et al., 2005) ; however, these conditions may not correspond with the optimal operating conditions since their selection does not contemplate economic factors . Besides, the use of conversion and yields does not allow the study of the impact of variables like residence time, solids loading, enzyme loading or temperature , on production cost. Therefore, the p roposed kinetic model for enzymatic hydrolysis will be used to economically evaluate the process for the production of ethanol . By analysing process economic at different conditions, we will be able to determine the effect of individual variables on the pr oduction cost and to identify the bottle necks in the process . A sensitivity analysis will be conducted to evaluate the impact of the primary contributor cost to the ethanol cost. 5 .2 Methods The process for the conversion of lignocellulose to ethanol propos ed in this study is illustrated in Figure 38. The production of ethanol , at the conditions shown in Table 7 (page 73) , are economically evaluated in this section. The optimal configuration among the proposed scenarios was identified base d on the ethanol production cost. O2WaterNaOHBiomassFlash Filtration FermentationHydrolysisWet OxydationPret. LiquorO2Water EnzymeSeedCO2DistillationEthanolLiquor and ResidueMolecular sieve Figure 38. Proposed process for the production of lignocellulosic ethanol. The separat ion step is a complex system, the design of which depends on multiple variables such as flow rate or ethanol concentration at the inlet, and thus its design is a very demanding 121 process. Therefore, considering the complexity of the distillation train design and the number of scenarios evaluated, pretreatment, hydrolysis and fermentation stages were simulated in AspenPlus and economically evaluated in APEA . The cost of the separation process was estimated and added to the production cost by scaling the separa tion cost reported by Humbird et al. (2011) to the flow rate and concentration of ethanol entering the separation stage. A sensitivity analysis for each evaluated scenario was carried out to understand t he effect of enzyme, NaOH , and biomass prices on production cost. The sensitivity analysis results were used to select the process conditions with the lowest production cost for which a detailed separation train was built in AspenPlus. The economics of the complete process simulations of the most promising scenarios were then analyzed in APEA. 5.2.1 Process and simulation description The composition of w heat straw is shown in Table 5 (page 50) and the material balances for the process at the studied conditions are detailed in Appendix B. The balance of the biomass is made up of extractives, protein and uronic acids. It was assumed that biomass is transported to the feedstock handling area fo r size reduction and storage before being fed to pretreatment . The cost given to biomass includes all of these operations. Previous reports suggest that the optimum plant size will have the capacity of 2000 to 4000 ton DM/day (Huang et al., 2009) . The plant capacity in this study was set to 2000 ton DM /day which is the most commonly assumed size in past economic analyses (Hamelinck et al., 2005b; Humbird et al., 2011; Kazi et al., 2010; Klein- Marcuschamer et al., 2010; Sassner et al., 2008; Wingren et al., 2005) . Aspen Plus V7.3.2 simulation software from Aspen Tech (Aspen Technology Inc., 2011a) was used to simulate the process. Physical property models in AspenPlus were used for most compounds, however, polysaccharides, lignin, proteins and fermentation nutrients were modeled using the physical properties reported by Humbird et al. (2011). Substreams were used to introduce both solid and liquid phases to the simulation: MIXED was set for compounds in liquid phase and CISOLID for solid compounds (Aspen Technology Inc., 2011a). Experimental mass balances were used to build the simulation; these mass balances 122 are provided in Appendix B. As no experimental work was performed for the fermentation, seed and separation stages, AspenPlus was used to solve the mass and energy balances. The fermentation stage was defined using reactions and conversions reported by Humbird et al. (2011), which is the most recent and most complete techno - economic analysis reported. Distillation units were modeled using the simulator operations in AspenPlus and molecular sieves were simulated as separation units. The resulting simulation can be seen in Figure 39. 123 Figure 39. Pretreatment, enzymatic hy drolysis and fermentation stages in the AspenPlus simulation for the production of lignocellulosic ethanol. Pretreatment reactor (RDELIG), rotary drum vacuum filter (RDVF1), hydrolysis reactor (RHYD), seed reactor (RSEED), sugar loss reactor (RLOSS), fermentation reactor (RFERMNT), pumps (P1 - P8), tanks (T1- T3), heaters (H1 - H4), flashes (F1- F3), diammonium phosphate (DAP) and corn steep liquor (CSL). 124 188.8.131.52 Pretreatment stage Oxygen delignification was modeled with a yield reactor (RDELIG) in AspenPlus. During oxygen delignification, the crystalline structure of cellulose is opened, a fraction of cellulose and hemicellulose are solubilised, and lignin is decomposed to CO2, H2O, and carboxylic acids (Banerjee et al., 2009) . In this process the biomass slurry is heated in a mixer tank , oxygen is injected to the slurry line and feed to the oxygen delignification reactor . After pretreatment , biomass at high pressure is released to a blow tank and then fed to the washer. The RDELIG works at the two conditions described previously : mild (M) and severe (S) (Table 5 ). Biomass composition before and after pret reatment were determined using the chemical analysis method presented in C hapter 2. Pretreatment mass balances (section B.1, page 222) used to simulate lignin and polysaccharides re moval at condition M and S (Table 12). It was assumed that sugar oligomers produced during pretreatment are not fermentable due to the difficulty to recover sugars from the pretreatment liquor and to the possible presence of fermentation inh ibitors. However , research is being conducted to recover sugars from the pretreatment liquor (Djioleu et al., 2012; Pérez et al., 2008) . Table 12. Polysaccharides and lignin conversions during pretreatment . % Conve r t e d to Produc t Reac ti on Condi ti on M Condi ti on S Glucan to glucose oligomer 10.3 25.8 Xylan to xylose ol igomer 24.3 42.1 Arabinan to arabinose oligomer 37.0 66.9 Galactan to galactose oligomer 54.1 82.5 Mannan to mannose oligomer 1.0 100.0 Lignin to soluble lignin 63.2 85.6 Overall hemicelluloses to oligomers 23.3 48.6 Pretreatment conditions in RDELIG are based on the operating temperature: 120°C for condition M and 150°C for condition S. The corresponding conversions , water and NaOH flow rates automatically update for each condition. The selection of operating condition is 125 based on a “Calculator block” written in FORTRAN. Sodium hydroxide (50% NaOH, stream 60) and pure oxygen (stream 9) were fed to RDELIG. The oxygen flow rate was kept constant during pretreatment experiments; therefore, the oxygen flow rate to RDELIG is constant. After pretreatment, pressure is released and the oxygen is removed in a flash unit (F1) simulated using a Flash2 AspenPlus unit , which represents the blow tank. In this flash, the pretreatment liquor is cooled to 50°C and 1 atm. Pretreated solids are separated with a Sep unit (RDVF1) which separates the inlet stream into any number of outlet streams based on separation ratios. Although the pretreated biomass was extensively washed to avoid any interference from the pretreatment liquor with the protein measurement during experiments , a wash step was not included in the sim ulation. The inhibitor effect of the compounds produced during pretreatment on the hydrolysis was studied by comparing the glucose production during hydrolysis using poorly washed pretreated biomass (≈0.4 L per batch) and well - washed biomass (≈4 L per batch). From results obtained (Figure 64), it was concluded that the pretreatment liquor does not affect the enzymatic hydrolysis and consequently, an extensive washing stage is not required. Furthermore, t he production of furfural and HMF, fermentation inhibitors, in the delignification of wheat straw has been reported to be low to non- existent (Bjerre et al., 1996) . The pretreated biomass is perfu nctorily wash ed during the solid- liquid separation after pretreatment, which is performed in a rotary drum vacuum filter (RDVF), as detail in section 184.108.40.206. From the Sep unit (RDVF1), pretreated biomass at 82 wt % moisture content is recovered and fed to hydrolysis reactor. 220.127.116.11 Enzymatic hydrolysis A simple RCSTR unit from AspenPlus was used to simulate the continuous production of sugars from the enzymatic hydrolysis process in the reactor RHYD. This unit was modeled based on the proposed models for the enzymatic hydrolysis considering a batch system. The models proposed for the production of glucose (equation 38, page 87) , xylose (equation 46, page 94) and Novozyme 188 protein aggr egation (equation 14, page 66) are not supported in the RCSTR unit. Therefore, a “Calculator block ” written in FORTRAN was used to introduce the proposed expressions in the simulation. 126 [𝑁] =[𝑁0] �𝑘𝑠[𝑁0] + 𝑘1′′𝑒−𝑡�𝑘1′′+𝑘𝑠[𝑁0]��𝑘1′′ + 𝑘𝑠[𝑁0] 14 [𝐺] = [𝐶0]�1− �1 +𝐾𝑒𝑘𝑑([𝐸0]− [𝐿]𝐹𝐿)𝑡𝐾𝑒 + ([𝐸0]− [𝐿]𝐹𝐿)�−𝑘4𝐾𝑒𝑘𝑑� 38 [𝑋] = [𝐹0]�1− �1 +𝐾𝑒𝑥𝑘𝑑([𝐸0]− [𝐿]𝐹𝐿)𝑡𝐾𝑒𝑥 + ([𝐸0]− [𝐿]𝐹𝐿)�−𝑘4𝑥𝐾𝑒𝑥𝑘𝑑� 46 Equation 38 and 46 were used to describe the production of g lucose and xylose, respectively, during hydrolysis. Experimental arabinan, galactan and mannan conversions are shown in Table 13 , for 5 and 10 wt % solids concentration and were used to simulate hydrolysis of arabinan, galactan and mannan. Residence time in RHYD defines the concentration of glucose, xylose and Novozyme 188 protein conce ntrations in the RHYD outlet stream. Table 13. Arabinan, galactan and mannan conversion used for the hydrolysis process % Conve r t e d to produc t Sol i d conc e nt r at ion ( w t % ) 5 10 Arabinan to arabinose 61.7 62.4 Galactan to galactose 21.3 11.1 Mannan to mannose 22.1 10.7 Conversions and kinetics parameters corresponding to the desired operating conditions are automatically selected when solid s concentration and enzyme loading are defined. Cellulase s are supplemented from Celluclast 1.5L and β - glucosidases from Novozyme 188, which were assumed to be purchased from a n external supplier. Celluclast 1.5L (streams: 15- 1 and 15- 2) and Novozyme 188 (streams: 16- 1 and 16- 2) are fed to RHYD in separate s treams: streams 15- 1 and 16- 1 f eed enz yme at 20 FPU/g cellulose and streams 15 - 2 and 16 - 2 fe ed enzyme at 40 FPU/g cellulose. E nzyme loading is defined by selecting stream 17 - 1 (20 FPU/g cellulose) or 17- 2 (40 FPU/g cellulose) in the Selector unit, C1. The same methodology is implemented 127 in Selector unit, C2 to define solids concentration in RHYD; streams 20- 1 and 20- 2 provide sufficient water to the reactor for 5 and 10 wt % solids concentration, respectively. Enzyme and water flow rate to RHYD are automatically defined by a “Calculator block” b ased on the amount of pretreated biomass entering RHYD. 18.104.22.168 Fermentation and seed process The fermentation arrangement proposed by Humbird et al. (2011), consisting of seed and fermentation processes, was used in the present study. In the seed process, microorganisms for the fermentation reaction are grown. In our process, 10% of the hydrolytic liquor (stream 27) is fed to the seed reactor (RSEED) to provide the inoculum volume needed for fermentation. The RSEED reactor was modeled as an RStoic unit in AspenPlus since this unit allow s the use of chemical reactions and conversion to simulate production of the fermentation organisms ( Table 14). Recombinant Zymomonas mobilis bacterium is used as the organism performing fermentation . Z. mobilis was assumed to simultaneously ferment glucose and xylose to ethanol. The minor hemicellulosic sugar arabinose is also fermented to ethanol with the same yield as xylose (Hum bird et al., 2011). Corn steep liquor (CSL) (stream 29 and 39) and diammonium phosphate (DAP) (stream 30 and 38) were define d in the simulation as reported by Humbird et al. (2011). CSL (stream 29) and DAP (stream 30) are fed to RSEED as nutrients for Z. mobilis . RSEED operates at 32°C with a 24 h residence time. CSL is fed to RSEED at 0.5 wt % and DAP at 0.67 g/L of the whole slurry (Humbird et al., 2011). The reactions and conversions used in RSEED are shown in Table 14. 128 Table 14. Seed reactions and assumed conversions (Humbird et al., 2011). Re ac ti on Conve r si on (%) Glucose → 2 Ethanol + 2 CO2 95.0 Glucose + 0.047 CSL + 0.018 DAP → 6 Z. mobilis + 2.4 H 2O 2.0 Glucose + 2 H 2O → 2 Glycerol + O2 0.4 Glucose + 2 CO 2 → 2 Succinic Acid + O2 0.6 3 Xylose → 5 Ethanol + 5 CO2 85.0 Xylose + 0.039 CSL + 0.015 DAP → 5 Z. mobilis + 2 H 2O 1.9 3 Xylose + 5 H 2O → 5 Glycerol + 2.5 O2 0.3 3 Arabinose → 5 Ethanol+ 5 CO2 0.85 Arabinose + 0.015 DAP + 0.3087 Protein → 5 Z. mobilis + 2 H 2O 0.019 3 Arabinose + 5 H 2O → 5 Glycerol + 2.5 O2 0.003 3 Arabinose + 5 CO 2 → 5 Succinic acid + 2.5 O 2 0.015 The inoculum from RSEED (stream 36) is mixed with the hydrolytic liquor coming from RHYD (stream 28). As sugar may be lost to side products by contaminating microorganisms during the seed and fermentation process, 3% of the sugars available for fermentation is assumed to be lost due to contamination, which is the sugar loss tolerated by the corn ethanol industry (Humbird et al., 2011). Sugar losses were modeled by a RStoic unit (RLOSS), where 3% of the sugars are converted to lactic acid (Humbird et al., 2011). The hydrolytic liquor (stream 37) is separated to feed 3% of the hydrolytic liquor and inoculum (stream 40) to RLOSS where all fed sugars are converted to lactic acid . RLOSS outlet stream 41 is mixed with the hydrolytic liquor (stream 42) and fed to the fermentation reactor (RFERMENT) which operates at 1 atm and 32 °C. CSL (stream 39) and DAP (stream 38) are fe d to RFERMENT at 0.25 wt % and 0.33 g/L of whole slurry, respectively. The reactions and conversion used in RFERMENT are shown in Table 15. Mannose and galactose were considered unfermentatble while arabinose follows the xylose stoichiometry, with 85% conversion to ethanol. CO2 produced during ferm entation is removed in RFERMENT, which was simulated by a Flash2 unit (F3). 129 Table 15. Fermentation reactions (Humbird et al., 2011) . Re ac ti on Conve r si on (%) Glucose → 2 Ethanol + 2 CO 2 95.0 Glucose + 0.047 CSL + 0.018 DAP → 6 Z. mobilis + 2.4 H 2O 2.0 Glucose + 2 H 2O → 2 Glycerol + O 2 0.4 Glucose + 2 CO 2 → 2 Succinic Acid + O2 0.6 3 Xylose → 5 Ethanol + 5 CO 2 85.0 Xylose + 0.039 CSL + 0.015 DAP → 5 Z. mobilis + 2 H2O 1.9 3 Xylose + 5 H 2O → 5 Glycerol + 2.5 O 2 0.3 22.214.171.124 Distillation The separation process simulated for scenario s M20- 5 and M 20- 10 are shown in Figure 40 and Figure 41. The propose d distillation train consists of a beer column, rectification column and molecular sieve. Preliminary specifications of distillation columns w ere estimated using the DSTWU unit from in AspenPlus. This unit provides an initial estimate of the min imum number of theoretical stages, the minimum reflux ratio, the feed stage, and the produc ts split. With this information, the rigorous calculation of the distillation columns was performed using the RadFrac model of AspenPlus, which is based on mass, equilibrium, summation, and heat (MESH). Energy consumption estimation was based on the thermal energy required by the heat exchangers and reboilers. In scenario M20- 5, fermentation broth is fed to two beer columns simulated as RadFrac units (D1A, D1B). Due to the large fermentation flow rate (stream 49) , two identical beer columns in parallel were selected to carry out the first separation step. 130 Figure 40. Separation p rocess simulation for scenario M 20- 5. The details of the dis tillation columns in scenario M20- 5 are shown in Table 16. All solids are removed in the beer columns bottoms along with 98% of the water from the fermentation broth poor in ethanol (stream 49) producing stream 75, with a mass fraction of ethanol of 0.38. The solid residues (stream 56) , which consist primarily of lignin, can be separated and used to produce process electricity or process steam (Humbird et al., 2011). No provisions were made in the AspenPlus simulation for further processing of lignin. The possible use of lignin to produce energy was evaluated as part of the economic analysis (Section 126.96.36.199, page 169). In order to remove most of the CO 2 from the fermentation broth, 3.2% of the overhead product from the beer column is separated as vapor; this removes 89% of the entering CO2. Less than 1% of the ethanol fed to the distillation train is lost with this stream . The remaining overhead product (35.7 % w/ w ethanol) is condensed and fed to the rectification column (D2) modeled as a RadFrac unit. A dual vessel pressure swing adsorber (PSA) was selected to produce anhydrous ethanol (99.5% w/w) from the rectification column effluent . The typical PSA cycle includes a production step in which the ethanol - water mix from the rectification column flows into one 131 of the PSA vessel s from the top at a high pressure. During this step water is adsorbed while the ethanol passes through the column and is collected at the bottom of the bed. After the production step, the bed must be regenerated ( water desorbed ) and prepared for the next cycle. In order to operate continuously, two columns are required. With this arrangement, one column carries out the production step while the second column is regenerated. During the regeneration, the pressure in the bed is reduced while some water is desorbed. This step is referred as depressurization. In the next regeneration step, water is desorbed from the bed under vacuum conditions. Near the end of the regeneration step, a portion of product vapor (99.5% w/w ethanol) is used to purge the vessel to remove the remaining adsorbed water. Then, the vessel is re- pressurized with product ethanol vapor from the operat ing vessel. The adsorbent bed has then completed its pressure swing cycle and is ready to enter a new production step (Simo et al., 2008) . The ethanol stream (stream 60) used to regenerate the PSA unit (MS) is fed to D2 to minimize ethanol losses. The characteristics of the r ectification column (D2) are shown in Table 16. The overhead product from D2 (92 % w/w ethanol) is fed to MS, which is modeled as a Sep unit (MS). The molecular sieve recovers 85% of the fed ethanol as product . Ethanol separation in MS is carried out at 1.6 atm and 116°C, while the sieve regeneration process is conducted at 0.15 atm and 32°C , as reported by Humbird et al. (2011) . The energy required for the regeneration process was provided by heaters H7 and H8 while the required pressure changes were provided using pump P9. The product stream (99.5% w/w ethanol) from the MS is cooled and pumped to the final Sep unit, B6 , so that the remaining CO2 can be removed. Product stream 79 is then sent to storage. 132 Table 16. Characteristics of the b eer (D1A, D1B) and rectification column (D2) estimated in AspenPlus for scenario M20- 5. De s i gn Cons i der ati on D1A, D1B D2 Feed temperature (°C) 100 Stream 55: 112 Stream 67: 78 Column pressure (atm) 2 1.2 Feed Tray 2 Stream 67: 11 Stream 55: 14 Tray type Sieve Sieve Number of t rays 11 29 Reflux r atio 3 3.4 Column diameter (m) 4.95 3.34 Enz ymatic hydrolysis in scenario M20- 10 is conducted at 10 wt % solid s concentration, consequently , the ethanol concentration in the fermen tation broth (steam 49) is higher than in scenario M20- 5. Enzymatic hydrolysis at 10 wt % solids concentration also decreases the flow rate of the fermentation broth entering the distillation train. As a result, in contrast with scenario M20- 5, only one bee r column (D1) was required in scenario M 10- 10 as shown in Figure 41. In order to remove CO2 from the fermentation broth, 3.9% of the overhead products are removed as vapor , removing 80.5% of CO2. The beer column (D1) recovers 99.5% of the ethanol (stream 49) in the overhead products (stream 51) and removes solids and 99.5% of water in the bottom products (stream 52). 133 Figure 41. Simulated separation p rocess for scenario M 20- 10. The characteristics of the distillation columns in scenario M20- 10 are shown in Table 17. The outlet stream from D1 (stream 51 ) is fe d to the rectification column (D2) , as shown in Figure 41, and contains 82.9% w /w ethanol concentration. The m olecular sieve regeneration stream is f ed to column D2 to reduce ethanol losses (stream 56) . The rectification column (RadFrac unit) removes 66.5% of the fed water in the bottom products and generates an overhead stream rich in ethanol. This ethanol rich stream is fed to the molecular sieve (MS). The D2 overhead stream (stream 53) contains 91.7% w/w ethanol and 0.80% w/w CO2. Fifteen percent of the ethanol fed to the molecular sieve is recycled to D2 (stream 56 ) after the MS regeneration step. MS operates at the same conditions as in scenario M20- 5. Stream 60 is cooled to 38°C and fed to a Sep unit (B6) to remove the remaining CO2. Product stream 79 (99.5% w/w ethanol) is then sent to storage. 134 Table 17. Characteristics of the b eer (D1) and rectification c olumn (D2) estimated in AspenPlus for scenario M20- 10. De s i gn Cons i der ati on D1 D2 Feed temperature (°C) 50 Stream 78: 71 Stream 59: 78 Column pressure (atm) 2 1.2 Feed tray 8 Stream 59: 6 Stream 78: 14 Tray type Sieve Sieve Number of t rays 19 18 Reflux r atio 2.28 3.24 Column diameter (m) 5.03 2.94 5.2.2 Capital c ost estimation The process was defined in APAE as shown in Table 18. These specifications are in accordance with the economic parameters published in multiple lignocellulosic economic analyses (Hamelinck et al., 2005b; Humbird et al., 2011; Kazi et al., 2010; Klein-Marcuschamer et al., 2010; Sassner et al., 2008; Wingren et al., 2005) . The accelerated depreciation method was used as proposed by Humbird et al. (2011) . 135 Table 18. Economic parameters for the economic evaluation of lignocellulosic ethanol production F acil it y t ype Parame t er Ref e re nc e Length of start up period 12 weeks (Humbird et al., 2011) Number of weeks per period 52 Operating life of plant 10 periods (Kumar and Murthy, 2011) Operating hours per period 8,000 (Piccolo and Bezzo, 2009) (Sassner et al., 2008) Project location North America Currency USD Internal rate of return 10% (Klein - Marcuschamer et al., 2010) (Piccolo and Bezzo, 2009) Start of basic engineering 01- Jan - 2012 Tax rate per period 35% (Piccolo and Bezzo, 2009) (Humbird et al., 2011) 188.8.131.52 Unit map ping in AspenP lus The units in AspenP lus simulations are representations of real processes; consequently, to assign a real cost to each piece of equipment, it is necessary to “m ap” them into APEA. In this process, the equipment that best represents the simulation unit is selected from the APEA database. The selected equipment is then sized and defined to estimate its cost. A summary of the considerations applied to the simulation equipment characterization is presented in the following section. 136 184.108.40.206 Mapping reactors Considering the high temperatures and pressures used in the delignification process, a pr essure/vacuum vessel jacketed tank unit from the APEA database was selected as the pretreatment reactor (Gallego, 2002). Due to the pretreatment 's harsh conditions, stainless steel 316 was selected for reactor construction, with a 95% working volume and a L/D of 2 (Fassel, 1974) . Due to the mild conditions used in the hydrolysis and fermentation, agitated enclosed tanks were selected from the APEA database to represent the hydrolysis and fermentation reactors. S tainless steel 304 was selected for hydrolysis and fermentation reactors with 75 kWh agitators as reported by NREL (Humbird et al., 2011). The p retreatment process was modeled as a continuous process (Schmidt and Thomsen, 1998) . The delignification reactor was sized using the volumetric flow rate and the experimental residence time (Table 5, page 9) . The hydrolysis reactor system was modeled as a system of batch reactors with the enough reactors to achieve a continuous process that matched the flow rates estimated in the simulation. The base reactor size used in the hydrolysis system was 1,000,000 gallons with a 950,000 gal working volume as proposed by Humbird et al. (2011). The number of reactors required was calculated assuming a 3 hour idle phase to clean the reactor between batches. The time to fill and draw the reactor was calculated based on the inlet flow rate. The reaction times were set to achieve cellulose conversions close to the maximum conversion based on the information presented in Figure 28. In Figure 42, the batch schedule for scenario M 40- 5 is detailed . In this scenario, 10 reactors of 1,000,000 gal and one of 600,000 gal were required to achieve the continuous production of sugars that matched the flow rates in the simulation. The 600,000 gal reactor requires a different fill - draw time and non- working time in accordance with its size , similar to the other reactors, a 3 h cleaning period was assumed. 137 Figure 42. Hydrolysis batch schedule for scenario M40- 5 138 Using the flow rate estimated in the Asp enPlus simulation, a comparison was made of the number of reactors required in a system of batch versus continuous reactors. The residence time for in each continuous reactor was estimated assuming ideal stirred tanks conditions (Brethauer and Wyman, 2010; de Gooijer et al., 1996) : where V (L) is the reactor volume, v0 (L/h) is the volumetric flow rateentering the reactor, the concentration of glucose entering the R reactor number in the train (R>0) is [G R] (kg/L), the glucose concentration in the outlet stream of the R reactor that enters the R+1 reactor in the train is [G R+1 ] (kg/L) and r G (kg/L*h) is the glucose reaction rate (equation 30, page 85). Using the defined reactor capacity (1,000,000 gal), the number of reactors in the train was set to match the cellulose conversion and glucose production achieved in the batch system. The number of reactors needed in each system is presented in Table 19. Table 19. Number and capacity of reactors needed to achieve the required cellulose conversion using batch and continuous reactors Number and capacity of reactors 1,000,000 gal 800,000 gal 600,000 gal 500,000 gal 300,000 gal Total Scenario CSTR Batch CSTR Batch CSTR Batch CSTR Batch CSTR Batch CSTR Batch M20- 5 15 15 1 15 16 M40- 5 11 10 1 11 10 S20- 5 13 13 13 13 S40- 5 7 7 1 1 8 8 M20- 10 8 9 1 1 9 10 M40- 10 8 9 1 1 9 10 S20- 10 6 7 1 7 7 S40- 10 6 7 1 7 7 Due to the long reaction times needed to achieve high cellulose conversion during hydrolysis, the volume and number of reactors estimated in both systems were similar. Therefore, the differe nces in capital cost of a batch or continuous system was assumed to have little effect on the ethanol cost for the reaction times considered. To account for the high density of the hydrolytic broth during the first hours of the reaction at high solids conc entrations, a 𝜏 =𝑉𝑣0=[𝐺𝑅]− [𝐺𝑅+1]−𝑟𝐺 50 139 continuous process was selected to estimate the cost of the hydrolysis stage. Due to the high density of the hydrolytic media at 10 wt % solid s concentration, 250,000 gallon reactors capable of handling the dense media were used to carry out hydrolysis for the first two hour s (Aden et al., 2002; Humbird et al., 2011) , after which the density is expected to decrease due to the action of the enzymes (National Renewable E nergy Laboratory, 2011; Rosgaard et al., 2007) . Once the density decreases due to the action of enzymes, the reaction continues in stirred 1,000,000 gal tanks until desired cellulose conversion is reached. The hydrolytic media is then sent to the fermenta tion stage. To the best of our knowledge, there are no published reports of continuous fermentation systems operating at the large scales required for this project (Brethauer and Wyman, 2010) . Therefore , the fermentation system was designed as a batch system . Reactors of 1,000,000 gallons were used to cover the 36 h residence time required for the fermentation , as defined by Humbird et al. (2011). The number of reactors required was calculated assuming a 3 hou r idle phase to clean the reactor between batches. The time to fi ll and draw the reactor was calculated based on the inlet flow rate. In Figure 43, the batch schedule for scenario M 20- 5 is detailed for where reactors 12 reactors of 1,000,000 gal were required to achieve a continuous process that match ed the flow rate s estimated in the simulation. The last reactor in this train has a non- working time due to the slightly shorter fill- draw period used to match the flow rate estimated in the AspenPlus simulation. A batch system was also used for the seed reactor train. The configuration was modeled on the one proposed by Aden et al. (2002), which used fi ve reactors of increasing volume and a total residence time of 24 h . 140 Figure 43. Fermentation batch schedule for scenario M20- 5 141 APEA can estimate the cost of vessels with a m aximum diameter of 4.5 m and a L/D ratio of 3 (for pressures 0 - 250 psi) (Aspen Technology, 2011; Aspen Technology Inc., 2011b) . Therefore, the cost of the proposed hydrolysis and fermentation reactors cannot be estimated with APEA. Instead, the Mulet, Corripio, and Evans method (Seider et al., 2009) was employed to estimate the cost of the hydrolysis and fermentation reactors. 220.127.116.11 Mapping separators A rotary drum vacuum filter (RDVF) with wash was selected from the APEA database to separate the pretreated wheat straw from the pretreatment liqu or after oxygen delignification . Aqua Merik specified a rotary drum filter with a maximum capacity of 272 m 3 /h and capable of achieving up to 97% solids recovery (Aqua Merik, 2010) . Due to the harsh delignification conditions, stainless steel 316 was selected as the RDVF material. The number of rotary drum filters was calculated based on the delignification outlet flow rate and the reported RDVF maximum capacity. P retreated biomass moisture content reported in Table 5 was used to model the RDVF unit (RDVF1) in the simulation. A solid belt conveyor was added to move pret reated wheat straw from RDVF to the hydrolysis reactor. The RDVF characteristics for all base case scenarios can be seen in Appendix A. 18.104.22.168 Mapping distillation columns and molecular sieve When a distillation column from AspenP lus simulator is mapped into A PEA, it is separated into five units: condenser, condenser accumulator, reboiler, reboiler pump, and distillation tower. Carbon steel was the material of construction used for each unit. The size and capacity of each of these units is estimated within APEA . A dual vessel temperature swing adsorber (Humbird et al., 2011) from the APEA library was selected to represent the molecular sieve unit. The molecular sieve was sized following the methodology proposed by the Gas Processors Supplie rs Association (GPSA) (2004) using the information provided by Humbird et al. (2011) and Gebreyohannes (2007) . 142 In scenario M20- 5D two beer columns in parallel , each one with a 6 m diameter were selected to carry out the separation process. Initially a single beer column with a 9 m diameter was specified. However, no reports about an operating distillation column with such a large dimension were found. Therefore, two columns in parallel were selected as the most viable configuration. The reported tangent to tangent height estimated by APEA is calculated from the tray stack height plus 4.5 m (disengagement and liquid reservoir height) . The details of the beer and rectification co lumns and molecular sieve for scenario M 20- 5D are shown i n Table 20. Table 20. Distillation columns and molecular sieve details for scenario M 20- 5D . Uni t Num be r of equi pm e nt Equi pm e nt det a il s Beer column 2 6 m dia., 9.9 m tall, 11 trays, 609 mm spacing Rectification column 1 3.4 m dia., 20.8 m tall, 29 trays, 609 mm spacing Molecular sieve 4 paired vessels 3.1 m dia., 12.7 m tall, 59, 298 kg 4Å, 1/8” bead The molecular sieve vessels and sieve (also referred as desiccant) needed for the ethanol separation were estimated following the method presented by Gas Processors Suppliers Association (2004). The molecular sieve calculations for scenario M20- 5 D are provided in Appendix B. From the inlet flow rate, two paired vessels were selected from the APEA database for the molecular sieve equipment. The adsorption step is carried out in one vessel while the sieve in the second vessel is regenerated. Assuming a price of $9 50 (2007) for 150 kg of 4Å, 1/8” bead molecular sieve (Gebreyohannes, 2007) and the mass of sieve needed per vessels (Table 20), the sieve cost was determined to be $393,152 per vessel . In scenario M20- 10D, the high solids concentration used for the enzymatic hydrolysis , increased the ethanol concentration and decreased the flow rate of the fermentation liquor entering the separation train. Therefore, only one beer column was required to separate ethanol from the solid residues and the majority of the water. The molecular sieve cost was estimated following the sa me methodology described above. The equipment details are 143 shown in Table 21 . Considering an adsorption period of 12 h, the estimated desicca nt cost for scenario M20- 10D was $325,300 per vessel . Table 21. Separation equipment details for scenario M20- 10D Uni t Num be r of equi pm e nt item s Equi pm e nt det ail s Beer column 1 5.0 m dia., 14.7 m tall, 19 trays, 609 mm spacing Rectification column 1 3.1 m dia., 14.1 m tall, 18 trays, 609 mm spacing Molecular sieve 4 paired vessels 2.7 m dia., 13.5 m tall, 49, 059 kg 4Å, 1/8” bead 5.2.3 Operating cost The cost assigned to raw materials and utilities consumed by the ethanol process a re presented in Table 22. 144 Table 22. Raw material and utilities used in the economic analysis P arame t er Pri c e Uni t s Sour c e s Raw mat er i al Wheat straw 57.63 $/t on DM (Leistritz et al., 2006; U.S. Department of Energy, 2001) Oxygen 302 $/h (Wilcox, 2004) NaOH 50% diaphragm grade 420- 850 $/t on (Chang, 2006) Celluclast 1.5L and Novozyme 188 4 to 8 $/ kg enzymes (Humbird et al., 2011; Klein-Marcuschamer et al., 2012) Diammonium phosphate (DAP) 0.4888 $/lb (Humbird et al., 2011) Corn steep liquor (CSL) 0.0281 $/lb (Humbird et al., 2011) Ut i li ti e s Steam, 50 psig 5.50 $/1000 kg (Seider et al., 2009) Steam, 150 psig 8.80 $/1000 kg Steam, 450 psig 12.10 $/1000 kg Cooling water 0.013 $/m 3 Process water 0.13 $/m 3 Electricity 0.04 $/kWh Natural gas (1020 Btu/SCF) 2.70 $/1000 SCF Details of the raw material cost determination are presented in A ppendix B. 5 .2.3.1 Base case scenarios In the base case scenarios, the raw material costs considered as most realistic were used in the economic analysis. The lower limit of the caustic soda prices ($420/ ton) given in Table 22, was used as the caustic price for each base case scenario. Biomass cost, including the feedstock handling costs, was set to $57.63/ ton DM, while an enzyme cost of $7/kg of 145 enzyme. More details about the raw material cost determination are presented in Appendix B: B.5 Raw material cost . In order to facilitate the analysis, the total enzyme price was assigned to the cellulase stream in APEA. The assumed DAP and CSL costs are presented in Table 22. In the base case scenarios the pretreatment, enzymatic hydrolysis and fermentation stages were simulated and economically evaluated with APEA. The process production cost s of the separation stage were estimated by scaling the cost reported b y Humbird et al. (2011). Recovery costs were estimated from the flow rate and concentration of ethanol in the fermentation stage outlet using an economy of scale factor of 0.6 (Seider et al., 2009) . The selected enzymatic hydrolysis residence times were chosen to achieve maximum cellulose conversion in each scenario in the shortest residence time as shown in Figure 28. 5 .2.3.2 Sensitivity analyses Enzyme and biomass prices are not fixed and their value , in this and other studies, are simple estimates. Sensitivity analysis of the enzyme and biomass costs can help to understand the process viability in different market scenarios. Caustic soda expenses are an important variable in the proposed process configuration due to the large amount of NaOH required during delignification. Pope (2011) reported that caustic cost accounts for approximately 19% of the final ethanol sell ing price. Caustic soda prices are highly variable (Abreu, 2012), and consequently , the change in caustic prices w as also evaluated in the sensitivity analysis. The minimum and maximum cost of NaOH was set at $370 and $850/ ton NaOH 50% (Abreu, 2012), for the sensitivity analysis . A NaOH 50% diaphragm grade solution was assumed in this study, as it is the most convenient and economical form of caustic soda used in the industry (Dow, 2010) . As agricultural residues are not widely sold, their commercial cost is unknown. A biomass price range of $30- 80/ ton DM was considered in this study, where the minimum cost is in close agreement with the biomass collection, processing, storage, and transportation cost. Enzyme prices are still speculative sin ce hydrolytic enzyme s are not yet produced industrially . Due to the variation in assumed enzyme costs in past studies , the range of enzyme cost evaluated in the sensitivity analysis was defined using the cost reported by Klein - Marcuschanmer et al. (2012) ($1/gal ethanol , equal to $5/kg enzyme ) and Genencor 146 and Novozymes ( $0.50/gal ethanol , equal to $7/kg enzyme ) (Humbird et al., 2011). An enzyme cost range of $5 to $9 per kg enzyme , which contains the mentioned prices , was used in this study. More details about the enzyme cost selection are presented in Appendix B: B.5.4 Enzymes . Enzyme costs are normally reported as a function of ethanol production ($/gal ethanol) . However , enzyme cost per mass of enzy me, calculated from the reported $/gal ethanol, may change depending on the process configuration assumed , as shown in Appendix B. E nzyme prices were reported as $/kg enzyme in the present study , in order to standardize enzyme and ethanol cost, helping to clarify the enzyme cost role in the process . Table 23 summarizes the c austic, enzyme and biomass prices used for the sensitivity analysis. Table 23. Raw material costs considered in the sensitivity analys is of the lignocellulosic ethanol process Raw Mat er i al Base case scenari o Sens it i vi ty analys is cost s Uni t s Biomass 57.6 30.0, 5 7 .6, 80 .0 $/ ton DM NaOH 50% 420 370, 420, 850 $/t on Enzyme 7 5, 7 , 9 $/ kg enzyme 5 .2.3. 3 Lignocellulosic ethanol production proc ess mathematical modeling To understand the effect s of the enzyme loading, solids concentration, pretreatment conditions, and enzyme, biomass and NaOH costs on the ethanol production cost investigated in our work , a total of 216 scenarios needs to be economically evaluated. A fractional factorial design was appl ied to reduce the number of scenarios required as part of the sensitivity analysis. The fractional factorial design was carried out in the statistic software JMP 10 (SAS Institute Inc., 2012) . Discrete numeric factors with three levels were used to describe the cost of biomass, caustic and enzyme . Discrete numeric factors with two levels describe the pretreated conditions (condition M or S), enzyme loading (20 or 40 FPU/g cellulose) and solids concentration (5 or 10 wt% ). A third order interaction was used to carry out the fractional factorial design , resulting in 64 scenarios. The resulting scenarios were evaluated with APAE to determine their corresponding e thanol production cost. These 147 scenarios together with the base case scenarios were used to obtain a mathematical expression for predicting the ethanol production cost as a function of pretreatment severity (PT, °C), enzyme loading (EL, FPU/g cellulose), solid s concentration (SC, wt% ), enzyme cost (EC, $/ kg ethanol ), biomass cost (BC, $/ ton DM) and caustic cost (NC, $/t on NaOH 50% ). The production cost obtained from AspenPlus were modeled within JMP 10 (SAS Institute Inc., 2012) with a third order interaction. The selection of the variables and variables interactions that define the production cost was carried out using the p- value obtained from the statistical analysis. A p - value of 0.05 is considered on the borderl ine of statistical significance, while valu es under 0.01 are considered statistically significant. P - values below 0.005 are considered highly statistically significant indicating that the observation is highly unlikely to be the sole result of random chance (Institute of work and health, 2005) . Therefore, p - values can indicate the variables with the highest impact in the production cost. In consequence, only variables that were highly statistically significant (p<0.005) were used to model the production cost. The fit of the proposed equation was evaluated by analysing the normalized residuals. 5 . 3 Results and discussion 5.3.1 Economic evaluation of base case scenarios The economic results of the base case scenarios obtained from the APEA analysis with the scaled recovery costs are shown in Table 24. The reactor size and number of reactors proposed in each base case scenario is detailed in Appendix B. 148 Table 24. Economic results for the lignocellulosic ethanol process for the base case scenarios . Pretreatment conditions: M= 30 min, 6 wt % NaOH/ dry biomass, 120°C and S=60 min, 10 wt % NaOH/ dry biomass, 150°C . S cen ari o M20 - 5 M40 - 5 S20- 5 S40- 5 M20 - 10 M40 - 10 S20- 10 S40- 10 Total cap ital cost (MM$ /year ) 23.8 21.2 20.5 18.2 14.6 14.2 13.2 13.0 Total operating cost (MM$ /year) 191.2 256.4 213.3 257.2 184.4 250.9 199.1 254.1 Total raw materials cost (MM$ / year) 149.9 211.4 162.5 213.3 146.9 208.5 160.2 211.1 Total utilities cost (MM$ / year) 15.9 15.9 15.9 15.8 15.8 15.8 15.8 15.8 Ethanol produced (kg/h) 14,414 16,085 12,734 12,939 11,692 13,955 10,171 11,491 Fermentable sugars concentration at fermentation in let (w/w) 0.028 0.031 0.032 0.033 0.046 0.055 0.054 0.060 Ethanol concentration at fermentation outlet (w/w) 0.014 0.015 0.016 0.017 0.023 0.028 0.027 0.030 Production cost ($/kg ethanol) 1.86 2.16 2.30 2.66 2.13 2.38 2.61 2.91 Operating cost, particularly raw material s cost, is the driving variable in ethanol economics. As shown in Figure 44, enzymes cost is, in most of the scenarios, the main material expense, accounting for 45% of the total raw material expenses on average. In scenarios S20- 5 and S20- 10, caustic was the primary raw material expense, as the enzyme loading was lower (1.2 kg enzyme broth/kg caustic) than in the rest of the other scenarios (1.3 to 2.5 kg enzyme broth/kg caustic) . Another reason for the high caustic impact on scenarios S20- 5 and S20- 10, is the high biomass loss suffered at severe pretreatment condition (S) (52.1% biomass degraded) compared with the loss at mild conditions (M) (35.7% biomass degraded ). Biomass and caustic are the other major raw material expenses and their importance varies with the operating conditions . Therefore, enzyme loading and price seems to be the dominant factors in the ethanol production. The economic analyses reveal that pretreatment has a 149 significant impact on enzyme loading and hydrolysis efficiency therefore although the effect of pretreatment on ethanol costs is indirect and difficult to determine, it has a major impact on project economics . Figure 44. Enzyme, biomass and caustic costs contribution to the total raw material expenses. The production cost of ethanol estimated for all base case scenarios is higher than most of the ethanol costs reported previousl y as shown in Table 25. The difference is due to a number of factors including feedstock type , raw material prices , solids concentration, hydrolysis configuration, pretreatment type and operating conditions (Table 25) . In this study, raw material expenses are one of the major controlling variables of ethanol cost thus assumptions about raw material costs signif icantly affect the outcome of the economic analyses . To illustrate this difference, consider enzyme prices assumed in past studies: Sassner et al. (2008) used an enzyme cost of $3.02/10 6 FPU, Humbird et al. (2011) used a enzyme contribution of $0.34/gal ethanol , Kumar and Murt hy (2011) assumed an enzyme cost of $0.52/kg of enzyme broth similar to the $0.51/kg enzyme broth used by Kazi et al. (2010). In this thesis, a higher enzyme cost of $ 7/ kg enzyme ($1 2.35/10 6 FPU or $0.63/kg enzyme broth (80 g protein/L) ) was used in the base case scenarios considering the enzyme cost reported by Genencor and Novozymes (Humbird et al., 2011; Kumar and Murthy, 2011) . In order to estimate the enzyme cost per enzyme mass , the flow rate of enzymes and the concentration of enzymes in the preparation or broth used are required . However , this information is not 0 10 20 30 40 50 60 70 M20- 5 M40- 5 S20- 5 S40- 5 M20- 10 M40- 10 S20- 10 S40- 10 Contribution to total raw material expenses (%) Scenario Enzyme Biomass NaOH 150 always reported in the economic analyses, consequently , only the enzyme cost of the mentioned studies were estimated o n a $/kg of enzyme basis. Table 25. Ethanol prices r eported in past economic analyses Pr et reat men t Hyd rol ys i s con f i gu rati on NaOH 50% ($/ t on ) Eth an ol price ($/k g eth an ol ) Ref eren ce DA SHF 109 1.53 (Klein - Marcuschamer et al., 2010) DA, dilute alkali, LHW, steam explosion SSF 408 1.03 to 1.27 (Kumar and Murthy, 2011) DA SHF- SSF 136 0.72 (Humbird et al., 2011) DA SSF - 1.11 to 2.05 (Piccolo and Bezzo, 2009) SP SSF 218 1.11 (Sassner et al., 2008) AFEX, D A, LHW, Lime, SAA, SP SHF- SSF - 0.78 to 1.36 (Tao et al., 2011) SP SHF, SSF 118 0.78 (Wingren et al., 2005) SP SHF, SSF 263 0.64 to 0.79 (Wingren et al., 2003) AFEX, DA, LHW SHF- SSF - 1.14 to 1.94 (Kazi et al., 2010) OD SHF 420 1.86 to 2.91 This work - no reported. Caustic cost in the present study ($420/ ton NaOH 50%) is higher than those reported in past studies. Only the cost reported by Kuma r and Murthy (2011) is in agreement with the cost assumed in the present study (Table 25 ). Estimated capital costs are within the range of capital cost s reported for different process configurations : $113M (Wingren et al., 2005) , $325- $364M (Tao et al., 2011), $115M (Piccolo and Bez zo, 2009) , $91- $115M (Kumar and Murthy, 2011) , $187- 202M (Huang et al., 2009) , $340M (Klein - Marcuschamer et al., 2010) and $442M (Humbird et al., 2011). However, the capital cost does not have a major impact on the production price. 151 The impact of enzyme loading on production cost was studied by comparing scenarios at different enzyme loadings. From Table 24, we can see that i ncreasing the enzyme loading from 20 to 40 FPU/g cellulose reduced capital cost s by 3 and 11% at 5 and 10 wt % solids concentration, respectively. Maximum cellulos e conversion was reached more quickly at high enzyme loadings ( Figure 28), which decreases the reaction time and number of hydrolytic reactors needed. Moreover, at similar solids concentration, higher cellulose conversions were achieved at 40FPU/g cellulose (Figure 26) , with the exception of scenarios S20- 5 and S40- 5, where 20 FPU/g cellulose is believed to be the optimal enzyme loading . Ethanol flow rate increased by 12 to 19% , causing a reduction in the distillation cost when increasing enzyme loading from 20 to 40 FPU/g cellulose . Operating costs increase by 26 to 36% when enzyme loading is doubled. Although, the ethanol flow rate increase d and capital cost decreased by doubling enzyme loading, production cost s increased by 11 to 21%, showing the key role of enzyme expenses in the process economy. A more pronounce effect of enzyme loading on operating cost was observed at mild pretreatment conditions than at severe conditions, as reported in Table 24. As mentioned, higher cellulose losses occur at severe pretreatment condition s, which reduced the total mass of cellulose in hydrolysis, and similarly, the mass of enzymes. A smaller enzyme flow rate i s then required to hydrolyse substrate S than substrate M . As a result, the impact of enzyme loading on operating cost is more important at mild than at severe pretreatment conditions . Increasing solids concentration from 5 to 10% decreased capital cost s by 28 to 39% . High solids concentrations decrease volumetric flow rate s, reducing the number and size of hydrolysis, fermentation and distillation pieces of equipment. Operating at h igh solids concentration also reduced operating cost (1 to 4%) by decreasing the process water usage. Nonetheless, as high solids concentration limits enzyme mobility (Bansal et al., 2009) , lower hydrolysis yields are obtained at 10 wt % solids concentration. As a result, the ethanol production decreased by 11 to 20% when operating at 10 wt % solids concentration. Despite the significant reduction in capital cost , and the small decrease in operating costs, production costs increased by 9 to 18% operating at high solids concentration. Although operation at low solids concentration may be more economically favorable , the number of pieces of 152 equipment nee ded at low solids concentrations may be unfavorable as it increases the risks of contamination. Capital cost decreased by 8 to 14% when the pretreatment severity was increased . More pretreatment reactors are needed to operate at severe conditions (substrate S). Nonetheless, the rapid hydrolysis of substrate S reduces the number of hydrolysis reactors leading to a net decrease in capital cost. Operating cost s increase by 8% operating at severe pretreatment conditions and 20 FPU/g cellulose , however, opera ting costs were unaffected by pretreatmen t conditions when 40 FPU/g cellulose was used for hydrolysis . At high enzyme loadings, enzyme expenses dominate the production cost, reducing the impact of pretreatment on the process economy. The ethanol production decreases by 12 to 20% with increasing pretreatment severity , which is explained by the increasing sugar loss at severe conditions. In consequence, production cost s increase by 18 to 23% when working at condition S. 5.3.2 Sensitivity analyses The scenarios obtained from the fractional factorial design were economically evaluated. In these scenarios, the pretreatment, enzymatic hydrolysis and fermentation stages were evaluated with APEA, while the recovery stage costs were scaled from those reported by Humbird et al. (2011). Production costs obtained for the factorial design and base case scenarios are shown i n Table 26. 153 Table 26. Fr actional factorial design scenarios and production cost. Pretreatment conditions: M= 30 min, 6 wt % NaOH/ dry biomass, 120°C and S= 60 min, 10 wt % NaOH/ dry biomass, 150°C . Sc e na r io Enz y me co st ($ /kg enzy me ) B io ma ss ($ / t o n DM ) N a O H 50 % ($ / t o n) APEA Pro duc t io n co st ($ /kg etha no l) E qua t io n 51 Pro duc t io n co st ($/kg etha no l) N o r ma liz e d residua l (%) M20- 5 5 30 370 1.45 1.42 2.3 5 30 420 1.49 1.45 2.8 5 58 420 1.68 1.62 3.3 5 80 370 1.79 1.73 3.5 5 80 850 2.19 2.05 6.4 7 58 370 1.82 1.82 0.2 7 58 420 1.86 1.86 0.3 9 30 370 1.82 1.85 -1.8 9 30 850 2.21 2.30 -3.9 9 80 850 2.56 2.61 -2.3 M40- 5 5 30 370 1.62 1.61 0.6 5 30 420 1.66 1.64 0.7 5 30 850 1.97 1.93 2.3 5 58 420 1.83 1.82 0.4 5 80 850 2.28 2.24 1.7 7 58 370 2.12 2.12 0.3 7 58 420 2.16 2.16 0.1 9 30 370 2.28 2.24 1.9 9 58 850 2.81 2.86 -2.1 9 80 370 2.59 2.55 1.4 S20- 5 5 30 370 1.74 1.81 -4.3 5 30 420 1.81 1.90 -4.6 5 30 850 2.48 2.65 -6.6 5 58 420 2.03 2.07 -2.2 5 80 850 2.87 2.96 -3.1 7 58 370 2.12 2.21 -4.3 7 58 420 2. 3 0 2.31 -0.8 9 30 370 2.08 2.25 -7.9 9 30 850 3.71 3.21 13.4 9 80 370 2.47 2.56 -3.6 S40- 5 5 30 370 2.03 2.01 1.2 5 30 420 2.11 2.09 0.7 5 58 420 2.32 2.27 2.2 5 58 850 2.98 3.02 -1.4 5 80 370 2.42 2.32 3.9 7 58 370 2.58 2.51 2.8 7 58 420 2.66 2.61 2.1 9 30 370 2.71 2.63 2.9 9 30 850 3.45 3.60 -4.5 9 80 850 3.83 3.92 -2.3 154 Sc e na r io Enz y me co st ($ /kg enzy me ) B io ma ss ($ / t o n DM ) N a O H 50 % ($ / t o n) APEA Pro duc t io n co st ($ /kg etha no l) E qua t io n 51 Pro duc t io n co st ($/kg etha no l) N o r ma liz e d residua l (%) M20- 10 5 30 370 1.61 1.62 -0.5 5 58 420 1.90 1.90 -0.2 5 80 370 2.04 2.07 -1.4 7 58 420 2.13 2.14 -0.5 9 30 370 2.07 2.06 0.6 9 80 370 2.49 2.50 -0.3 9 80 850 2.98 2.95 1.0 M40- 10 5 30 370 1.76 1.82 -3.4 5 30 850 2.17 2.14 1.4 5 58 420 2.00 2.10 -5.1 5 80 850 2.52 2.58 -2.3 7 58 420 2.38 2.43 -2.5 7 80 370 2.49 2.59 -3.8 9 30 370 2.52 2.45 3.0 9 30 850 2.93 2.89 1.2 9 80 850 3.28 3.34 -1.7 S20- 10 5 30 370 2.03 2.02 0.4 5 58 420 2.39 2.35 1.7 5 80 370 2.51 2.46 2.1 5 80 850 3.45 3.30 4.4 7 58 420 2.61 2.59 0.8 9 30 370 2.46 2.45 0.2 9 30 850 3.39 3.42 -0.8 9 80 850 3.88 3.86 0.4 S40- 10 5 30 370 2.20 2.21 -0.7 5 30 850 3.03 3.05 -0.8 5 58 420 2.52 2.55 -1.0 7 30 370 2.58 2.54 1.5 7 58 420 2.91 2.88 0.8 7 80 850 3.84 3.89 -1.3 9 30 850 3.79 3.81 -0.5 9 80 370 3.40 3.28 3.3 Production costs reported in Table 26 were used to develop an equation capable to predict production cost s from operating conditions and raw material prices . This expression was developed within JM P 10 (SAS Institute Inc., 2012); p- values obtained for the variables and their relations are shown in Figure 45. The evaluated variables are: pretreatment temperature (PT: 120 and 150°C), enzyme loading (EL : 20 and 40 FPU/g ce llulose), solids concentration (SC: 5 and 10 wt % solids concentration), enzyme cost (EC: $5 to $7/kg enzyme ), NaOH cost (NC: $370 to $850/T on) and biomass cost (BC: $30 to $80/ ton DM). 155 Figure 45. Analysis of p - values for the selection of relevant variables in the ethanol production economy obtained with JMP 10. 156 The highly statistically significant variables (p - value<0.005) were selected to develop a model of ethanol economics. Based on the results presented i n Figure 45 , the terms selected for ethanol cost modeling are : NC, PT, EC, SC, BC, EL, NC- PT, EC- EL, NC- EC, and SC-BC. The large t- ratio obtained for caustic cost (NC) reflects its significant impact on the ethanol production cost . These results are in agreement with those reported by Tao et al. (2011), which demonstrates the substantial impact of different pretreatments on ethanol cost. As expected, raw material cost s (EC, NC, BC) are very important to project economics since production cost depends primarily on operating costs . Enzyme s are one of the major contributors to raw material cost and therefore , more efforts are needed to minimize enzyme usage (EL) and especially enzyme price (EC). The substantial contribution of enzyme cost to ethanol cost has not been fully reflected in past analys es due to the optimistic assumed enzyme prices (Humbird et al., 2011; Kumar and Murt hy, 2011; Sassner et al., 2008; Wingren et al., 2005) . The enzyme cost s assumed in many studies were deemed extremely optimistic by Klein - Marcuschamer et al. (2012). Thus, the impact of enzyme cost on ethanol cost may be higher than previously rep orted. P- values show the interaction between raw material costs and process stages: NC- PT, EC- EL, NC- EC, and SC- BC. These relationships emphasize the need to optimize pretreatment and hydrolysis as a whole . The selected terms were fit by least squares in J MP 10 to obtain an expression to predict ethanol production cost (PC: $/kg ethanol) : 𝑃𝐶 = 1.6540 + 0.1356 ∗ �𝑆𝐶 − 7.52.5�+ 0.3273 ∗ �𝑃𝑇 − 13515�+ 0.1390 ∗ �𝐸𝐿 − 3010�+ 0.1358 ∗ 𝐸𝐶 + 0.1896 ∗ �𝐵𝐶 − 5525�+ 0.3165 ∗ �𝑁𝐶 − 610240�+ 0.1295 �𝑃𝑇 − 13515� �𝑁𝐶 − 610240�+ 0.0219(𝐸𝐶 − 6.5872) �𝐸𝐿 − 3010�+ 0.0151(𝐸𝐶 − 6.5872) �𝑁𝐶 − 610240�+ 0.0322 �𝑆𝐶 − 7.52.5� �𝐵𝐶 − 5525� 51 Production costs obtained with equation 51 are compared with the production costs generated from APEA in Table 26. Equation 51 had a good fit to the APEA results (R2=0.98) as shown in Figure 46. 157 Figure 46. Production cost calculated with APEA and predicted with equation 51 ( ), 5% significance curve ( ) obtained with JMP 10. The predictive capabilities of equation 51 were tested by comparing production costs estimated from APEA and eq uation 51 for scenarios not used in the development of equation 51. The results are shown in Table 27. 158 Table 27. Production cost estimated with equation 51 and APEA for different conditions. S cen ari o Enzyme cost ($/ k g enzy me) Biomas s ($/ t on DM ) NaOH 50% ($/ t on ) APE A Prod u cti on cost ($/k g eth an ol ) Equ ati on 51 Prod u cti on cost ($/k g eth an ol ) Nor mal i zed resi d u al (% ) M20- 5 9 30 370 1.82 1.84 - 1.3 M40- 5 7 80 370 2.26 2.25 0.2 S20- 5 9 30 420 2.16 2.34 - 8.2 S40- 5 5 58 850 2.98 3.07 - 3.0 M20- 10 7 58 420 2.13 2.13 - 0.4 M40- 10 5 58 850 2.36 2.43 - 2.9 S20- 10 7 80 370 2.73 2.69 1.5 S40- 10 7 30 850 3.41 3.45 - 1.1 Equation 51 predicts the ethanol costs obtained with APEA within 8% as shown in Table 27. In consequence , equation 51 can be used to study the effect of each variable in the process as shown in Figure 47. Figure 47. Effect of evaluated variables on the production cos t of ethanol. From the results shown in Figure 47 and the parameters of equation 51, it is clear that increasing raw material cost s increases the sensitivity of ethanol cost to hydrolysis and pretreatment conditions . High enzyme and caustic prices increased the impact of pretreatment conditions on ethanol cost. Therefore, mild pretreatment conditions (substrate M), low solids concentration (5 wt% ) and enzyme loadings (20 FPU/g cellulose) m ay be more advantageous for the process , since these conditions reduce the impact of raw material costs on production costs . Operating under these conditions may help to protect the process economy against unstable raw material price s. However , optimal pre treatment conditions 159 must exist as extremely mild conditions lead to high lignin contents that decrease ethanol yields. Similarly, optimum enzyme loadings must exist as extremely low enzyme loading s results in unattractive cellulose conversion (Lu et al., 2002). Based on the results presented in Table 26 and Table 27, it was concluded that equation 51 can be used to estimate the production cost of ethanol at different conditions and raw material costs for the proposed process configuration. Equation 51 was used to estimate the production costs for all of the studied scenarios (Table 7 ) across the full range of raw material costs proposed (Table 23) as shown i n Table 28. 160 Table 28. Production costs for all proposed scenarios at different raw material costs estimated with equation 51. E n zyme cost ($/ kg enzyme ) Biomass ($/t on DM ) NaOH 50% ( $/ t on ) Produ ction cost ($/kg ethan ol) M20 - 5 M40 - 5 S20 - 5 S40 - 5 M20 - 10 M40 - 10 S20 - 10 S40 - 10 5 30 370 1.42 1.61 1.81 2.01 1.62 1.82 2.02 2.21 5 30 420 1.45 1.64 1.90 2.09 1.66 1.85 2.10 2.30 5 30 850 1.73 1.93 2.65 2.84 1.94 2.14 2.85 3.05 5 58 370 1.59 1.78 1.98 2.18 1.87 2.06 2.26 2.46 5 58 420 1.62 1.82 2.07 2.27 1.90 2.10 2.35 2.55 5 58 850 1.91 2.10 2.82 3.02 2.18 2.38 3.10 3.29 5 80 370 1.73 1.93 2.13 2.32 2.07 2.26 2.46 2.66 5 80 420 1.76 1.96 2.21 2.41 2.10 2.29 2.55 2.74 5 80 850 2.05 2.24 2.96 3.16 2.38 2.58 3.30 3.49 7 30 370 1.63 1.92 2.03 2.32 1.84 2.13 2.24 2.53 7 30 420 1.67 1.96 2.12 2.41 1.88 2.17 2.33 2.62 7 30 850 2.02 2.31 2.93 3.22 2.22 2.52 3.14 3.43 7 58 370 1.81 2.10 2.20 2.49 2.08 2.38 2.48 2.77 7 58 420 1.85 2.14 2.30 2.59 2.12 2.42 2.57 2.87 7 58 850 2.19 2.48 3.10 3.40 2.47 2.76 3.38 3.67 7 80 370 1.95 2.24 2.34 2.64 2.28 2.58 2.68 2.97 7 80 420 1.99 2.28 2.44 2.73 2.32 2.62 2.77 3.06 7 80 850 2.33 2.62 3.24 3.54 2.67 2.96 3.58 3.87 9 30 370 1.85 2.24 2.25 2.63 2.06 2.45 2.45 2.84 9 30 420 1.90 2.29 2.35 2.73 2.10 2.49 2.55 2.94 9 30 850 2.30 2.69 3.21 3.60 2.51 2.89 3.42 3.81 9 58 370 2.02 2.41 2.42 2.81 2.30 2.69 2.70 3.09 9 58 420 2.07 2.46 2.52 2.91 2.35 2.74 2.80 3.19 9 58 850 2.47 2.86 3.39 3.78 2.75 3.1 4 3.67 4.05 9 80 370 2.17 2.55 2.56 2.95 2.50 2.89 2.90 3.28 9 80 420 2.21 2.60 2.66 3.05 2.55 2.94 3.00 3.39 9 80 850 2.61 3.00 3.53 3.92 2.95 3.34 3.86 4.25 Average 1.94 2.23 2.52 2.82 2.21 2.50 2.80 3.09 161 Numerous studies have identified the most promising pretreatment conditions based on delignification efficiencies and digestibility of the resulting substrate (Galbe and Zacchi, 2007; Kim and Lee, 2007; Lu et al., 2002) . Moreover, economic studies often select pretreatment conditions based on the pretreated substrates digestibility or because they have been widely reported (Kazi et al., 2010; Kumar and Murthy, 2011; Piccolo and Bezzo, 2009) . However , in this study, it was found that despite the low lignin content and high cellulose conversions obtained for substrate S, production costs for pretreatment condition S were higher than for pretreatment condition M due to sugar losses during pr etreatment. Therefore, typical pretreatment conditions that results in high delignification and hydrolysis yields may not be the most advantageous. The lowest production cost in Table 28 was obtained for scenario M20- 5 ($1.42/kg ethanol) using the lowest raw material costs proposed in the sensitivity analysis. This production cost was achiev ed using an enzyme cost of $5/kg enzyme, which corresponds to the enzyme cost assumed by NREL ($0.34/gal ethanol) (Humbird et al., 2011). This enzyme cost corresponds to approximately half the cost used by Klein - Marcuschamer et al. (2010), and is considered an optimistic cost that may be achieved in the future . The caustic cost used in scenario M20-5 ($ 370/ ton NaOH 50% ) from Table 28 is still higher than most of the costs presented in Table 25. The biomass cost used in this scenario ($30/t on DM) is considered as ideal since the biomass collection, p rocessing, storage and transportation cost is estimated on $34.1/t on DM (Leistritz et al., 2006) . A cost of this magnitude implies that farmers are not paid for the agricultural residues. Scenario M20- 5 has the lowest production cost over the whole range of conditions and cost tested, while scenario M20- 10 yielded the second lowest production costs. These results are unexpected because, as mentioned earlier, high delignification and high hydrolysis yields were favored in past studies. Optimizing pretreatment conditions requires a decrease in raw material usage and sugar losses. Scenario M40- 5 had the third lowest production costs, but, scenarios M20- 5 and M20- 10, due to their low enzyme loading, have a lower impact to the enzyme price fluctuations , which is important since enzyme prices are undefined . For these 162 reasons, detailed distillation trains for scenarios M20- 5 and M20- 10 were built in AspenPlus to carry out detailed economic analyses. 5.3.3 Economic evaluation of the complete simulated process es Scenarios M20- 5 and M20- 10 resulted in the lowest p roduction costs at all the evaluated raw material costs; consequently, a detailed separation train was buil t in AspenPlus for both scenarios. The separation stage was added to the pretreatment, hydrolysis and fermentation simulation. These simulations were then used to do a detailed economic analysis in APEA. The utilities and raw materials costs considered in this analysis are defined in the methods section (page 143 ). 5. 3. 3.1 Economic analysis of scenario M20 - 5 including simulated separation stage (M20- 5D) The simulated p retreatment, enzymatic hydrolysis and fermentation stages are the same as those reported for the base case scenarios M20- 5, therefore, scenario M20- 5 with simulated distillation equipment stage will be referred as scenario M20- 5D . The distillation train for scenario M20- 5D is detailed in the methods section (page 129 ). Economic results for scenario M20- 5D obtained from APEA are summarized in Table 29. Table 29. Ethanol production costs for scenario M20- 5D. S cen ari o M20 - 5D Total capital cost ( MM$ /year ) 18.5 Total operating cost ( MM$ /year) 201.4 Total Raw Materials Cost ( MM$ /Year) 147.9 Total Utilities Cost (MM$ /Year) 25.2 Ethanol fl ow rate (kg/h) 14,299 Production cost ($/kg ethanol ) 1.9 4 163 Production cost for scenario M20- 5D is 4% higher than scenario M20- 5 (Table 28). Scenarios M20- 5 and M20- 5D differ in the method used to estimate the separation cost (distillation and MS train). Separation stage costs in scenario M20- 5 were estimated by equation 51, which was developed by scaling up the separation cost reported by Humbird et al. (2011). On the other hand, the recovery cost for scenario M20- 5D was based on the simulated separation process presented in section 22.214.171.124 (pp 129) . The production cost s differ due to the higher energy expenses in scenario M20- 5D than in scenario M20- 5. Another factor that may contribute to the different results is that M20- 5D produces 0.8% less ethanol than M20- 5, as M20- 5D accounts for ethanol losses during distillation. Due to the large number and size of hydrolytic and fermentation reactors needed in this scenario, the reactor costs account for 7 2% of the total direct cost (TDC) as shown i n Figure 48. Therefore, ethanol cost can be re duced by decreasing the number of reactors required during hydrolysis and fermentation. This may be done by increasing solids concentration, enzyme loading or enzyme and f ermentative organism performance. However, high solids concentrations decrease cellulose conversion (Figure 26), increasing ethanol cost (Table 24). High enzyme loadings increase reaction rates but also ethanol cost ( Table 24). Therefore, hydrolysis at high enzym e loadings is not viable unless enzyme prices are significantly reduced. By increasing the performance of the hydrolytic enzymes and fermentative organisms, maximum cellulose and sugar conversion may be reached faster . This will decrease residence times and the number of reactors required . However, the cost of the hydrolytic enzymes will remain an issue for the production of ethanol, and the economic benefit s of more active enzymes will depend on their production cost . 164 Figure 48. Process stage contribution to the TDC in scenario M20- 5 D. The estimated contribution of pretreatment to the TDC ( MM$ 18.5) (Figure 48) in this work is lower than the 26% (MM$ 29.9) reported by Humbird et al. (2011) . This is likely due to the different pretreatment technologies used as well as the contribution of hydrolysis and fermentation to the TDC (28% , MM$ 31.3 ) in Humbird’s work . In this thesis, a SHF configuration was use d (MM$ 71.1 hydrolysis and fermentation capital cost) , while Humbird’s process included a 24 h hydrolysis followed by SSF . As a result, Humbird's process uses fewer hydrolysis and fermentation reactors than the process in this study . Moreover, Humbird et al. (2011) reported a reactor cost of $849,700 for a 1,000,000 gal reactor, while a reactor of equal capacity was estimated to cost $1,771,200 in the present work . Other published economic analyses have assumed the equipment costs reported in a n earlier NREL report (Aden et al., 2002): $730 ,500 (2011) for a 1,000,000 gal reactor. Consequently , their TDC distribution is similar to that reported by Humbird (Kazi et al., 2010; Kumar and Murthy, 2011; Piccolo and Bezzo, 2009; Tao et al., 2011) . Our results are in agreement with the TDC distribution reported by Wingren et al. (2003) , where pr etreatment, hydrolysis, fermentation and distillation have a TDC contribution of 21% (MM$ 12.9) , 40% (MM$ 24.9) , 16% (MM$ 10.1) and 21% (MM$ 6.1) , respectively. The large hydrolysis and fermentation TDC contribution repor ted by Hamemelinck et al. (2005b) is also in accordance with our results. Pretreatment 19% Hydrolysis 37% 0% Fermentation 36% Separation 8% 165 The contribution of raw materials, utilities and capital costs to the production cost are shown in Figure 49 . Operating charges include: operating supplies , laboratory charges, plant overhead (charges during production for services, facilities, payroll overhead, etc.) , and general and administrative costs (administrative salaries/expenses, R&D, product distribution and sales costs). At an internal rate return (IRR) of 10%, the annualized capital cost has a small impact on the production cost, compared to operating cost s. Enzyme, biomass and caustic cost are the major factors that determined ethanol cost, contributing to 66% of the production cost. It is clear that decreasing raw material costs is essential to process economics. The high raw material cost contribution obtained in this work is in agreement with reports by Galbe et al. (2007) (57%), Humbird et al. (2011) (54%), Piccolo and Bezzo (2009) (57%) and Wingren et al. (2003) (61%). Due to the frequent changes in caustic prices and the wide range of variation, caustic cost is difficult to control and predict (Chang, 2006) . The use of cheaper sources of caustic should be considered. Another possible option to decrease caustic expenses is by increasing temperature, taking into account the possible production of steam from lignin, which ca n reduce energy costs. The price of b iomass is still an estimate, consequently, more studies are needed to define its cost (Leistritz et al., 2006; Talebnia et al., 2010) . In previous studies, substrate costs had a major contribution to the production cost: 32% in Galbe et al. (2007) , 35% in Humbird et al. (2011), 35% in Kazi et al. ( 2010), 29% in Klein - Marcuschamer et al. (2010), 40% in Tao et al. (2011) and 31- 32% in Wingren et al. (2003; 2005) . The low biomass cost contribution (19%) reported here is likely due to the use of a higher enzyme cost ($6/kg enzyme) relative to other studies (Arifeen et al. , 2007; Gnansounou and Dauriat, 2010; Humbird et al., 2011; Kazi et al., 2010) . This enzyme cost seems realistic based on the information given by Genencor and Novozymes ( $0.50/gal ethanol) (Humbird et al., 2011) and Klein -Marcushamer et al. (2012) ($1/gal ethanol). More details about the assumed enzyme costs are provided in Appendix B. Enzyme cost is the major raw material cost in scenario M20 - 5D; therefore, greater efforts to decrease enzyme production costs are needed. Anot her option to decrease enzyme expenses is by enzyme recovery, where cellulases are recycled by 166 adsorption on fresh substrate to start a new hydrolysis round. The economic viability of this technology has not been demonstrated and it is evaluated in Chapter 6. Figure 49. Breakdown of cost for the production of lignocellulosic ethanol in scenario M20-5D 5. 3. 3.2 Economic analysis of scenario M20 - 10 including simulated separation stage (M20- 10D) Pretreatment, hydrolysis and fermentatio n stages simulated for the base case scenario M20 -10 were used for t he analysis of scenario M20 - 10D. The separation stage (distillation and MS train) was simulated and added to these stages to evaluate the entire process economy (scenario M20- 10D). The details of the simulated separation stage are described in the Steam 11% Process water 1% Enzyme 28% Biomass 19% Other Raw Materials 2% NaOH 17% Oxygen 1% Annualized capital cost 8% Labor and Maintenance 4% Operating charges 9% 167 methods section. The economic results for scenario M20- 10D with the simulated distillation train are shown in Table 30. Table 30. Ethanol product ion costs for scenario M20- 10D. S cen ari o M20 - 10D Total capital cost ( MM$ /year ) 12.6 Total operating cost ( MM$ /year) 177.3 Total raw m aterials cost (MM$ / year) 145.2 Total utilities cost (MM$ / year) 9.3 Ethanol flow rate (kg/h) 11,633 Production cost ($/kg ethanol) 2.06 Operating costs for M20 - 10D are $0.07/kg ethanol lower than for M20- 10, which is likely due to the use of process streams to meet heating and cooling requirements in scenario M20-10D. The use of hot streams to cover the process energy requirements reduced operating cost by 4%. Relative to M20- 5D, a greater amount of heating can be performed using process streams in M20- 10D because flow rate is lower and thus the energy required for distillation is lower . The ethanol p roduction cost estimated for scenarios M20 - 10 and M20- 10D differ by 3.3%, indicating that equation 51 successfully predict s the production cost of ethanol . The contribution of each process stage to TDC is shown in Figure 50. By increasing solids concentration from 5 to 10 wt % in the hydrolysis stage, the TDC was reduced by 30 %. Hydrolysis contribution to capital cost is lower in M20 - 10D than in M20- 5D , as result of the high solids concentration, and associated reduction in the number of reactors . Consequently , separation and pretreatment capital costs become more important to the project economics at high solids concentrations. Nonetheless, hydrolysis and fermentation capital costs are still the primary equipment expenses , and the number of reactors required must be reduced. 168 Figure 50. Installed equipment cost distribution per process for the ethanol production i n scenario M20- 10D. The contribution of utilities, r aw materials and capital costs to the production cost are shown in Figure 51. The cost distribution in scenario M20- 10D is similar to that obtained for scenario M20- 5D. B y doubling solids concentration in hydrolysis, the contribution of raw materials to the ethanol price increased. On the other hand, annualized capital cost, operating charges and steam contributions to ethanol cost decreased. Under scenario M20- 10D, the process economy is more vulnerable to raw material cost s and market changes. H eating costs decreased by 55% in scenario M20 - 10D compared to M20- 5D. The presented results encourage the efforts to reduce the biomass and enzyme prices in order to successfully commercializ e lignocellulosic ethanol. Pretreatment 20% Hydrolysis 32% 0% Fermentation 38% Separation 10% 169 Figure 51. Breakdown of cost for the production of lignocellulosic ethanol in scenario M20-10D 5. 3. 3. 3 Expanding the scope of ethanol production In the present section, the possibility of incorporating a water treatment system (WT) and by -product boiler and turbogenerator (BT) facility to scenarios M20- 5D and M20- 10D is examined. In order to account for the WT and BT costs, the considerations, capital and operational costs reported by Humbird et al. (2011) were scaled to have a gross estimate of the WT and BT impact on the ethanol cost . BT operating and capital costs were estimated assuming the solid residue separated in beer columns (D1, D1A and D1B) is burned to produce steam. WT cost s were estimated assuming that the water recovered from the separation stage can be recovered, treated and recycled to pretreatment . Capital and operating costs were estimated using an economy scale factor of 0.6 (Seider et al., 2009) . Steam 5% Process water 1% Enzyme 32% Biomass 22% Other Raw Materials 1% NaOH 19% Oxygen 1% Annualized capital cost 7% Labor and Maintenance Cost 3% Operating charges 9% 170 Using the BT information and considerations provided by Humbird et al. (2011), and the amount of solid residues recovered in M20- 5D, the production of steam by BT was estimated to be 283,000 kg/h at 454°C and 6.2 MPa . Half of the steam produced is used to meet distillation heat requirements and the balance is used to produce 31.2 MW of electricity. From the produced electrici ty, 10.52 MW is used to cover process requirements and the remainder is sold to the grid at $0.057/kWh, price set by Humbird et al. (2011). In scenario M20- 10D, 285,000 kg/h of steam is produced, 48% of which is used to cover distillation requirements . The remaining steam generates 48 MW. This is sufficient to meet process requirements (10.12 MW) the balance is sold to the grid. Based on the water treatment configuration reported by Humbird et al. (2011), it was assumed that 72% of the process water is recycled. Using these assumptions, the economic results for scenario M20- 5D and M20- 10D with the addition of WT and BT are shown on Table 31. Table 31. Economic results for scenarios M20- 5D and M20- 10D with the addition of water treatment and boiler and turbogenerator processes. Sc e nari o M20- 5D wit h BT M20 - 5D wit h BT and WT M 20- 10D wit h BT M20 - 10D wit h B T and WT Total capital cost ( MM$ /year ) 25.8 41.3 20.7 34.3 Total operating cost ( MM$ /year) 192.3 199.4 170.9 176.9 Total raw m aterials Cost (MM$ / year) 147.9 145.9 145.2 143.1 Total utilities cost (MM$ / year) 13.4 13.4 0.01 0.01 Production cost ($/kg ethanol ) 1.82 2.02 1.87 2.08 BT implementation increased capital cost by 39% for M20 - 5D and by 64% for M20- 10D. However, by generating process steam and selling electricity, operating costs decrease by 5% and 4%, in scenarios M20- 5D and M20- 10D, respectively. G eneration of steam and electricity reduced production cost s in M20- 5D by $0.11/kg ethanol and by $0.19/kg ethanol in M20- 10D. The use of generated steam reduced utilities expenses by 47% and 99% for 171 M20- 5D and M20- 10D, respectively. Thus, converting solid residues to steam and electricity can significantly reduce ethanol production costs and a detailed analysis of the optimal process configuration and conditions to convert lignin to steam will be the focus of future work . In line with this conclusion, it has been reported that the Danish Inbicon demonstration plant for the production of ethanol from wheat straw operating includes a combined heat and power plant (Larsen et al., 2012). Production costs obtained for scenario M20- 5D and M20-10D with BT are in agreement with cost s reported by Kazy et al. (2010) and Piccolo and Bezzo (2009) . WT addition increased capital costs by 60% and 66% for scenarios M20 - 5D and M20- 10D with BT, respectively. Due to the low process water contribution to production costs, the addition of WT increased the ethanol cost by $0.20 and $0.21/kg ethanol from scenarios M20- 5D and M20- 10D with BT, respectively. T he WT viability for the process must be discussed based on its potential to have a positive environmental impact. It has been mentioned repeatedly that the cost of hydrolytic reactors, fermentors, enzyme and caustic assumed in the present work are higher than in other economic analys es (Humbird et al., 2011; Sassner et al., 2008; Wingren et al., 2003) . In order to evaluate the effect of the assumed costs, the raw mater ial and reactor costs reported by Humbird et al. (2011) (Table 32) , and used in numerous analyses (Kazi et al., 2010; Kumar and Murthy, 2011; Piccolo and Bezzo, 2009; Tao et al., 2011) , were applied to scenarios M20 - 5D and M20- 10D with BT. In this study, a cellulase loading of 40 mg protein/g cellulose results in cellulose conversions of 69% and 85% after 76 h for M20- 5D and M20- 10D, respectively. Humbird et al. ( 2011) considered an enzyme loading of 20 mg protein/g cellulose , for which a cellulose conversion of 90% after 84 h was used. In order to match the high enzyme performance assumed by Humbird in this work, it was assumed that the same cellulose conversions achieved in M20-5D and M20- 10D could be obtained using half of the enzyme loaded. 172 Table 32. Cost and conditions considered by Humbird et al. (2011) . P arame t er Val ue NaOH 50% cost $135.65/t on Enzyme cost $0.34/gal of ethanol ($4.83/kg enzyme) Reactor (1,000,000 gal) purchase cost $849,660 each (2010) Enzyme loading 20 mg protein/g cellulose The economic results for M20- 5D and M20 - 10D with BT under the assumptions presented in Table 32 are shown in Table 33. Capital costs decreased by 2 3% and 25% for M20- 5D and M20- 10D, respectively, due to the decrease in reactor cost. Operating cost s decreased by 40 and 44% for scenarios M20- 5D and M20- 10D with BT, respectively. The low raw material costs reduced production cost s by 38% and 44% for M20 - 5D and M20- 10D with BT, respectively. These p roduction costs are in agreement with ethanol prices reported in most economic studies (Kazi et al., 2010; Klein- Marcuschamer et al., 2010; Kumar and Murthy, 2011; Piccolo and Bezzo, 2009; Sassner et al., 2008) . Table 33. Economic results for scenarios M20- 5D and M20- 10D with BT , under costs and considerations reported by Humbird et al. (2011). Sc e nari o M20 - 5D wit h BT M20 - 10D wi t h BT Total capital cost ( MM$ /year ) 14.4 17.6 Total operating cost ( MM$ /year) 117.1 97.1 Total Raw Materials Cost ( MM$ / year) 82.3 79.6 Total Utilities Cost (MM$ / year) 13.4 0.01 Production cost ($/kg ethanol ) 1.13 1.05 Production cost for M20 - 5D and M20- 10D wi th the assumed costs from Table 33 are the lowest obtained in this study. Even so, they are 46 to 57% higher than those reported by 173 Humbird et al. (2011). The difference is likely due to the different cellulose conversions: in this thesis, the experimental conversions were lower than the 90% cellulose conversion reported by Humbird et al. In the present study, a SHF process was used; in c onsequence, the number of reactors required is larger than in other studies which applied the SSF configuration. Production costs obtained in this study presented as g asoline gallon equivalent (GGE), are shown in Table 34. Table 34. Production cost s as GGE compared to gasoline cost (U.S. Department of Energy, 2013b) and corn ethanol (USDA Livestock poultry and grain market news, 2014) . I t em $/ GGE Gasoline 3.3 Corn ethanol 3.7 M20- 5D 8.7 M20- 10D 9.2 M20- 5D with BT 8.2 M20- 10D with BT 8.4 M20- 5D with BT under NREL parameters 5.1 M20- 10D with BT under NREL parameters 4.7 Production costs for scenarios with and without BT implementation are double the current gasoline cost. Only by implementing the raw material and equipment cost s reported by Humbird et al. (2011) are production costs low enough to compete with gasoline and corn ethanol prices . The ethanol cost reported by Humbird et al. (2011) is equivalent to $3.2/GGE, which is lower than gasoline and corn ethanol prices presented in Table 34 . Ethanol prices in this range should be considered targets. Ethanol prices of this magnitude would be highly commercially attractive and do not reflect the reality, where lignocellulosic ethanol is not yet widely industrially produced. Economic analysis of the production of ethanol must consider unfavorable material and equipment prices in accordance to expected industrial costs. More 174 efforts are needed to increase hydrolysis yields and reduce raw material expenses in order to commercialize lignocellulosic ethanol. 5 .4 Conclusions Different operating conditions based on experimental results and models for the production of glucose and xylose from lignocellulose were economically evaluated . These scenarios were evaluated using raw material cost s considered to be in agreement with expected industrial prices . From the base case scenarios, it was shown that operating costs are the major contributor to production cost. Therefore, use and prices of raw material are the controlling variables of ethanol economics. Unexpectedly, operating conditions that result in high delignificatio n and hydrolysis yields led to high production costs. Mild pretreatment conditions and low enzyme loadings yielded the lowest production cost s (scenario M20- 5 and M20- 10), emphasizing the importance of minimiz ing sugar losses during pretreatment. Pretreatment and hydrolysis optimization must be done considering the entire process economy and not only individual yields. An equation capable of predicting production cost s at different pretreatment conditions, enzyme and solids loading in hydrolysis, and enzym e, caustic and biomass costs was developed. The proposed equation successfully predict ed production cost s for the proposed process at different conditions , when compared with the results obtained from APEA . This expression was used to c onduct a sensitivity analysis on raw material prices . Analysis of this expression revealed the impact of pretreatment in the process economics . Although pretreatment is not a major contributor to the production cost, its impact on the downstream process makes it a significant variable in the process economics , closely followed by raw material costs. Due to the high level of uncertainty in biomass and enzyme prices , and highly variable caustic prices , production costs from the sensitivity analysis range d from $ 1.42 to $4.16/ kg ethanol. Scenario M20- 5 had the lowest production cost at all raw material costs tested. Mild conditions minimize the impact of raw materials market changes . Nonetheless, optimal 175 conditions must exist as extremely low enzyme loadings or mild pretreatmen t conditions will produce economically unattractive hydrolysis yields. High enzyme prices make enzyme usage a key factor in the process . Enzyme cost has previously been determined by its contribution to the process as $/gal ethanol . Enzyme prices per mass calculated from enzyme contribution costs depend on the process configuration. As a result, the enzyme cost per mass widely varies from one study to another . To better compare enzyme prices and ethanol costs between studies, costs should be standardized by reporting enzyme costs per mass of enzyme, protein, or enzyme broth. The conversion of solid residues to steam and electricity has a positive effect on the process economics. The by- product’s high value for the ethanol industry encourages the research of new technologies to convert solid residue to valuable by- products . The addition of a water treatment process increases production cost s; consequently, its viability will depend on the environmental advantages that water treatment offer s. The lowest p roduction cost estimated in this study is double the current gasoline price therefore lignocellulosic ethanol production i s not commercially competitive under the assumed conditions and raw material cost s. To decrease ethanol cost it is necessary to address k ey points in the process. Pretreatment conditions have to be optimized to reduce caustic requirements and sugar losses . The number and size of reactors required during hydrolysis and fermentations must also be reduced. Enzyme loading s and prices have to be reduced, this can be done by reducing lignin content (without increasing sugar loss es), or by enzyme recovery. 176 6 E c o n o m i c e v a l u a t i o n o f t h e p r o d u c t i o n o f l i g n o c e l l u l o s i c e t h a n o l w i t h e n z y m e r e c y c l i n g b y a d s o r p t i o n 6 .1 Introduction The conversion of l ignocellulose to ethanol is a complex and expensive process (Coppola et al., 2009) . In order to make lignocellulosic ethanol commercially attractive its production cost must be reduced. The contribution of e nzyme expenses to production costs in past economic analyses depended upon the enzyme price and process configuration assumed. In the previous chapter, it was demonstrated th at enzyme expenses are one of the major contributor to production costs. Due to the recalcitrant lignocellulose composition, high enzyme loadings and relatively long incubation periods are typically required to achieve high cellulose conversions (Lu et al., 2002). Therefore, different options to decrease enzyme loadings and enzymatic hydrolysis cost have been studied. One of the most promising ways to decrease enzyme usage and cost is by enzyme recycling . During enzymatic hydrolysis , cellulases adsorb on cellulose and then gradually desorb as the reaction advances. By taking advantage of this absorption, cellulases in solution after hydrolysis can be adsorbed onto fresh substrate and used to start new hydrolytic cycles (Qi et al., 2011). Due to the irreversible adsorption of cellulases on lignin, the presence of lignin is thought to be the most important factor in the enzyme recycling process, as it diminish es the amount of enzymes that can be recycled (Lu et al., 2002). Enzyme recycling by adsorption has been demonstrated to be experimentally possible by multiple authors (Kristensen et al., 2007; Lu et al., 2002; Qi et al., 2011; Steele et al., 2005; Tu et al., 2007a) but the benefits that this technology can offer are unknown. To accurately evaluate the viability of this technology, sugar loss es, equipment and operational cost s involved in the enzyme recovery process must be taken into account, as well as all the process stages. In the present chapter , enzyme recovery by adsorption was simulated and economically evaluated. 177 6 .2 Methods The production of lignocellulosic ethanol from a process with and without the implementation of enzyme recycling by adsorption at different conditions was simulated in AspenPlus and economically evaluated with APAE . Production cost reduction caused by the enzy me recovery implementation was the criterion to evaluate enzyme recovery viability. The results of the enzyme recycling experiments in Chapter 4 were used to simulate and economically evaluate the enzyme recycling at conditions presented in Table 11. Scenario M20- 5 with enzyme recycling after 24, 48 and 72 h, and scenario S20- 5 with enzyme recycling after 12 h were discarded from this economic analysis. In these scenarios, the amount of enzyme recycled was inconsequential (<5% cellulases recovery), reason why the y were omitted . The scenarios economically evaluated in this chapter are presented in Table 35 . Based on glucose concentration changes during hydrolysis (Figure 28), the selected times were chosen to correspond to different reaction stages: high conversion rate, deceleration of conversion rate, and approaching maximum conversion. Table 35. Enzyme recycli ng scenarios economically evaluated. Pretreatment conditions: M= 30 min, 6 wt % NaOH/ dry biomass, 120°C and S= 60 min, 10 wt % NaOH/ dry biomass, 150°C . S cen ari o Sub strate Enzyme Load i n g (FPU/ g cell u l os e) Recycl i n g ti me (h) M40- 5- R5 M 40 5 M40- 5 - R24 24 M40- 5 - R48 48 S20- 5- R24 S 20 24 S20- 5 - R48 48 S40- 5 - R5 40 5 S40- 5 - R24 24 S40- 5 - R48 48 Similar to the economic evaluations presented in C hapter 5, pretreatment, enzymatic hydrolysis and f ermentation stages were first simulated in AspenPlus and economically 178 evaluated in APEA. The cost of the separation stage was estimated and added to the process by scaling up the separation cost reported by Humbird et al. (2011) , based on the flow rate and concentration of ethanol at the inlet of the separation stage. A sensitivity analysis was carried out for the most promising scenario to study the effect of enzyme, caustic and biomass prices on product ion costs and enzyme recycling savings . 6.2.1 Process and simulation description The pretreated biomass composition is shown in Table 5. The same assumptions used to build the ethanol production without enzyme recycling mass balances were used for the enzyme recycling simulation . The process for the lignocellulosic ethanol production with enzyme recovery is shown in Figure 52. O2WaterNaOHBiomassFlash Filtration FermentationWet OxydationPret. LiquorO2Water EnzymeSeedCO2DistillationLiquor and ResidueEnzyme adsorptionFiltrationFiltration HydrolysisSolid residueEthanolMolecular Sieve Figure 52. P rocess diagram for the production of lignocellulosic ethanol with enzyme recovery. This process was modeled with AspenPlus V7.3.2 software from Aspen Tech (Aspen Technology Inc., 2011a). Physical property models in AspenPlus were used for all the common compounds, however, physical models reported by Humbird et al. ( 2011) were used to model polysaccharides and lignin properties . Mass balances built from the experimental recycling work presented in C hapter 4 were used to simulat e the process, and are summarized in Appendix C. AspenP lus was used to solve the mass and energy balances for 179 the fermentation and separation stages as no experimental work was produced for these stages. The resulting simulation is provided in Figure 53. 180 Figure 53. Simulation for the production of lignocellulosic ethanol with enzyme recovery. Pretreatment reactor (RDELIG), rotary drum vacuum filter (RDVF1), hydrolysis reactor (RHYD), adsorption reactor (RADS), solid- liquid separators (FI1 and FI2), seed reactor (RSEED), sugar loss reactor (RLOSS), fermentation reactor (RFERMNT), pumps (P1 - P8), tanks (T1- T2), heaters (H1- H4), flashes (F1 - F4), diammonium phosphate (DAP) and corn steep liquor (CSL). 181 The assumptions and considerations used for the pretreatment, hydrolysis, seed and fermentation process in the previous chapter were maintained, t herefore, only the enzyme recovery process is detailed in this chapter . 6.2.2 Enzyme recycling process Pretreated biomass and the hydrolytic liquor (stream R1), containing cellulases, are fed to the adsorption reactor (RADS) , modeled with a RStoic unit. The amount of cellulases adsorbed on pretreated biomass in each scenario was manually set as the conversion of cellulases in liquid phase (L) to cellulase s in solid phase (S). The amount of cellulase s adsorbed during experimental work was used to model the cellulases adsorption. RADS operates at the experiment al conditions: 50°C and 20 min residence time. After adsorption , solid and liquid phase s are separated in a Sep unit (FI2). The solids and adsorbed cellulases (stream R2) are recovered and fed to hydrolytic reactor (RHYD). RHYD was modeled as described in Chapter 5. Following experimental biomass moisture content after separation ( represented as FI2), 10% of stream R1 is fed to RHYD with the biomass . The remaining liquor recovered from Fl2 is sent to the fermentation stage . Fresh cellulases, β - glucosidase and water are fed to RHYD. After enzymatic hydrolysis, the hydrolytic liquor is recovered from separator unit FI1 and fed to RADS to recover free cellulase. It was assumed that all solid residues were separated, and , based on the experimental solid residues moisture content, around 2.3 to 2.5% of the hydrolytic liquor is lost with the solid residues. 6.2.3 Cost estimation APEA was used to economically evaluate pretreat ment, enzymatic hydrolysis and fermentation stage s in all scenarios. The assumptions used for the economic analysis presented in last chapter were used in the present analysis. 182 126.96.36.199 Unit map ping in AspenP lus The units in AspenP lus are a representation of real process; consequently, to assign a cost and equipment to each unit , it is necessary to “ map” them into APEA. Reactor mapping methodology followed in C hapter 5 was also applied to map pretreatment, hydrolysis and fermentation reactors in this analysis. 188.8.131.52 Mapping separators In order to separate pretreated biomass after delignification, a rotary drum vacuum filter (RDVF1) with wash was selected as a unit to separate and rinse pretreated biomass. The rotary drum filter was sized following the same methodology as in Chapter 5. A moisture content of 82% was measured for the pretreated biomas s, which was used to set the RDVF1 separation efficienc y. Single vacuum rotary drum dryer (AspenTech, 2011) equipment listed in the APEA library was used to estimate RDVF1’s cost . A solid belt conveyor was added t o RDVF1 to transport the pretreated biomass to the hydrolysis stage. FI1 and FI2 separator s were mapped as rotary drum vacuum filters without washing to avoid enzyme and sugar dilution. The number of rotary drum filter s was selected based on flow rate. 6.2.3. 3 Operating cost s Costs assigned to raw materials and utilities are the same as those in Chapter 5 for the base case scenarios. All the assumptions mentioned in the previous chapter were used in the current economic analysis. Enzyme, biomass and caustic prices for the base case scenarios are shown in Table 22. In order to evaluate the effect of enzyme recycling on process enzyme expenses, an enzyme cost of $7/kg enzyme wa s assigned to the Novozyme 188 183 stream, based on the β - glucosidase concentration of the cocktail (7.7% of the protein content) reported by Chundawat et al. (2011) . Therefore, the Celluclast 1.5L (streams 15 - 1 and 15- 2) and Novozyme 188 (streams 16 - 1 and 16- 2) costs are $0.89/kg Celluclast 1.5L and $0.05/kg Novozyme 188. 184.108.40.206 Sensitivity analysis A sensitivity analysis was performed for the enzyme recovery scenario that resulted in the biggest decrease in production cost s. Enzymes , caustic and biomass costs were selected for the sensitivity analysis based on their impact on the process economics reported in Chapter 5. In the present analysis , only the lower and upper range of raw materia l prices , shown in Table 36, were used in the sensitivity analysis. Table 36. Sensitivity analysis for the ethanol production with enzyme recycling. Raw Mat er i al Base case scenari o Sens it i vi ty analys is cost s Uni t s Biomass 57.6 30, 80 $/ ton DM NaOH 50% 420 370, 850 $/t on Enzyme 7 5, 9 $/kg enzyme 6. 3 Results and discussion 6.3.1 Economic evaluation of enzyme recovery at different hydrolytic times Economic results for ethanol production with e nzyme recycling were compared to results for ethanol production without enzyme recycling to evaluate enzyme recovery viability. The characteristics and number of reactor s for all the evaluated scenarios are provided in Appendix C. Economic results for scen arios M40- 5 with and without enzyme recycling , considering base raw material costs , are shown in Table 37. 184 Table 3 7. Economic results for scenario M40 - 5 with and without enzyme recovery by adsorption at different hydrolytic times. No enzy me recov ery Enzyme re covery Scen ari o M40 - 5 M40 - 5 M40 - 5 -R5 M40 - 5 -R24 M40 - 5 -R48 Hydrolysis time (h) 24 48 5 24 48 Total capital cost (MM$ /year ) 20.5 23.2 21.3 22.0 24.8 Total operating cost (MM$ /year) 254.0 256.0 246.0 240.0 236.2 Ethanol flow rate (kg/h) 15,622 16,933 10,857 15,246 16,465 Enzyme recycled (%) 0 0 7 13 17 Production cost ($ /kg) 2.21 2.07 3.08 2.15 1.98 Due to the short hydrolysis time and low enzyme recycling efficiency , the production cost of scenario M40- 5 - R5 is the highest production cost in Table 37. This is caused by the low cellulose conversion (44%) obtained after 5 h of reaction, compared with conversions after 24 h (67%) and 48 h (69%) . Since M40- 5- R5 ’s production cost is higher than the other scenarios, M40- 5- R5 is considered unviable. Enzyme recovery implementation reduced operating cost by 6% in scenario M40- 5- R24. In this scenario, ethanol production decrease d by 2% and capital cost increased by 7% , the net effect on enzyme recovery was that the production cost decreased by 3% . By recycling enzyme s after 48 h of hydrolysis ( M40- 5-R48), capital cost increased by 7%. Due to sugar losses during solid and liquid separation, ethanol production decreased by 3% compared with the process without recycling . Operating costs decreased by 8% and the final production cost decrease d by 5%. S avings were achieved by reducing operating costs and m aintaining the ethanol production. E nzyme recycling imp lementation results in conservative savings for scenario M40 - 5. Economic results for scenario S20- 5 with and without enzyme recycling are shown in Table 38. 185 Table 38. Economic results for scenario S20- 5 with and without enzyme recovery by adsorption at different hydrolytic times. No enzy me recov ery Enzyme re covery Scen ari o S20- 5 S20- 5 S20 - 5 -R 24 S20 - 5 -R 48 Hydrolysis time (h) 24 48 24 48 Total capital cost ( MM$ ) 183.9 205.1 197.0 220.0 Total operating cost (MM$ /year) 210.4 213.3 198.3 193.2 Ethanol flow rate (kg/h) 11,395 12,734 11,144 12,422 Enzyme recycled (%) 0 0 9 20 Production cost ($ /kg) 2.53 2.30 2.44 2.16 Enzyme recovery implementation in scenario S20- 5- R24 increased capita l cost by 9%, while reducing operating cost by 6% and ethanol production by 2% (2.2% sugar losses during solids separation) which leads to a small production cost reduction of 3%. Enzyme recovery implementation again resulted in small benefits for the proce ss. By recycling the enzyme after 48 h of hydrolysis (S20- 4- R48) the capital cost increased by 7 % , due to additional equipment introduced to the process . Sugar losses decreased ethanol production by 2% , similar to scenario S20- 5- R24. Nonetheless, due to the small decrease in the ethanol produced, the production cost decreased by 6 % . Cellulose conversions for scenarios S40- 5- R5, S40- 5- R24 and S40- 5- R48 were 62%, 80% and 87%, respectively. Economic results for scenario S40- 5- R5 are similar to the ones obtained for S20- 5- R5, showing that high production cost s result after short hydrolysis times due to low cellulose conversion ( Table 39) . These results indicate that ethanol production at low hydrolysis yields is not viable, and that optimal conditions must exist. As capital cost has a smaller impact on ethanol costs than operating cost , longer hydrolytic times (i.e. more reactors) may be preferable as they result in low er production costs. Nonetheless , the operation of a large number of high capacity reactors may not be technically viable as it increases the risk of contamination. By recycling enzyme after 24 h of hydrolysis (S40- 5 -R24), capital cost is 9% higher than for scenario S40- 5. Ethanol production in scenario S40-186 5- R24 is similar to that obtained for scenario S 40- 5. Enzyme recovery decreased the operating cost by 1 5%, and production cost by 12%. In scenario S40- 5 - R48, the biggest increase in capital cost was obtained. Despite this, the significant decrease in oper ating cost (18%) led to a production cost reduction of 14%, the largest economic benefit obtained. Table 3 9. Economic results for scenario S40- 5 with and without enzyme recovery by adsorption at different hydrolytic times. No enz y me recov ery Enzyme re covery Scen ari o S40- 5 S40- 5 S40 - 5 -R5 S40 - 5 -R24 S40 - 5 -R48 Hydrolysis time (h) 24 48 5 24 48 Total capital cost (MM$ /year ) 18.2 18.9 19.2 19.9 21.9 Total operating cost (MM$ /year) 257.2 267.8 248.9 218.8 220.0 Ethanol flow rate (kg /h) 12,939 13,877 9,325 12,689 13,515 Enzyme recycled (%) 0 0 6.73 35 36 Production cost ($ /kg) 2.66 2.58 3.59 2.35 2.24 Tu and Saddler (2010) reported a production cost reduction of 8.6% with the addition of Tween 80 (surfactant) and enzyme recycling (approximately 50% enzyme recovered ). In this study, it was assumed that enzyme recycling implementation does not incur additional capital and operational costs . The greatest cellulase recovery achieved was 35.4 and 36.1% for scenarios S40- 5 - R48 and S40- 5- R24 (Figure 54 ). Furthermore, surfactants were not used in this study; even so, 12 and 14% production cost reductions were achieved in these scenarios. Similar savings were obtained in scenario S40- 5 - R48 and S40- 5- R24, where after 24 h of hydrolysis, cellulases and glucose concentrations remain approximately constant ( Figure 28). Glucose and cellulases concentration after 24 and 48 h of hydrolysis are similar , which can explain the proximity of the cellulases recoveries and savings obtained in scenarios S40- 5-R48 and S40- 5- R24. Due to the longer hydrolysis time in scenario S40- 5 - R48, the number of reactors required was double the number of reactors needed in scenario S40 - 5- R24. However , because of the low impact of capital cost on the process ec onomics, the production cost reductions obtained for scenarios S40- 4- R48 and S40- 4- R24 were similar. 187 Figure 54. Production cost for evaluated scenarios with out ( ) and with enzyme recycling ( ) at different hydrolytic times. The p ercent of cellulases recover ed are shown on top of each scenario. A larger amount of cellulases were recycled at 40 FPU/g cellulose than at 20 FPU/g cellulose, behavior that agrees with the results shown in Figure 28, where higher cellulases concentrations are present in solution for scenarios working at 40 FPU/g cellulose. Although, high enzyme loadings favor enzyme recycling, the production cost s at high enz yme loadings were much higher than those obtained for other scenarios , such as M40- 5 - R24, M40- 5- R48 or S20- 5- R48, as shown in Figure 54. Therefore, increasing the enzyme loading to improve enzyme recycling is not economically viable. Enzyme recovery was more successful using substrate S at both 20 and 40 FPU/g cellulose , achieving a 3 to 14% reduction in capital cost. Substrate S is likely more suitable for enzyme recycling by adsorption than substrate M due to it s lower lignin content, which reduces cellulases losses due to adsorption on lignin (Farinas et al., 2010; Jørgensen et al., 2007; Nakagame et al., 2010; Zhou et al., 2009b) . Nonetheless, even whe n enzyme recovery decreased the ethanol cost using substrate S, the ethanol cost in scenario M20- 5 without enzyme recovery was lower (Table 24). Scenario M20- 5 produced the lowest ethanol cost in this work but this scenario is unattractive for enzyme recovery since less than 2% cellulase recovery was achieved in the experimental work as shown in Figure 33. From these results it 0.0 0.5 1.0 1.5 2.0 2.5 3.0 M40- R24 M40- R48 S20- R24 S20- R48 S40- R24 S40- R48 Production cost ($/kg ethanol) No recovery Recovery 12.5% 16.9% 8.7% 20.3% 35.4% 36.1% 188 is clear that the technical feasibility and eco nomic benefits of enzyme recycling depend heavily on process configuration and operating conditions. It is also possible that the lowest ethanol production cost could be obtained without enzyme recovery , as in this study. In order to increase the benefits of the enzyme recycling by adsorption technology on the process economics, the amount of cellulases must be increased. The addition of surfactants , such as Tween 80, during hydrolysis has been shown to improve both enzymatic hydrolysis and enzyme recovery (Tu and Saddler, 2010). The effect of surfactants on ethanol economics will depend on the quantity of surfactants needed (Tween 80 price is estimated in $1.00/kg (Tu and Saddler, 2010)), and the hydrolysis and enzyme recovery enhancements obtained. In addition, the low process water ( around 1%) and capital cost (7 to 8%) contributions to production cost reported in Chapter 5, suggest that the addition of a washing process to increase cellulases recovery from solid residues would help to decrease production costs (Maurer et al., 2012). The benefits of these technologies and their viability have yet to be evaluated. 6.3.2 Sensitivity analysis for ethanol production with enz yme recovery The economic results obtained in the previous section indicate that scenarios S40- 5- R24 and S40- 5- R48 are the most promising scenarios for implementing enzyme recovery since the highest production cost reductions (12 and 14%) were obtained under these conditions. Due to the long hydrolysis time at which S40 - 5 - R48 operates , more hydrolytic reactors are needed in this scenario than in S40- 5- R24. The operation of a large number of high capacity reactors complicates process control and substantial ly increases the risk of contamination during saccharification. Therefore, S40 - 5- R24 is thought to be technically more manageable, and was selected for the sensitivity analysis. The production cost s obtained for scenarios S40- 5 and S40- 5- R24 are shown on Table 40. 189 Table 40. Production cost of scenario S40- 5 with and without enzyme recycling at different raw material costs Enzym e cost ($/ kg enzym e ) Biom as s ($/ t on DM ) NaO H 50% ($/t on) Prod uc ti on cost ($/ kg ethanol ) Dif f e r e nc e i n produc ti on cost (%) S40 - 5 S40 - 5 - R2 4 5 30 370 2.01 1.83 8.8 5 30 850 2.84 2.58 9.1 5 80 370 2.3 2 2.22 4.1 5 80 850 3. 16 2.97 5.9 9 30 370 2.63 2.28 13.2 9 30 850 3. 60 3.03 15.8 9 80 370 2.95 2.67 9.4 9 80 850 3. 92 3.42 12.7 These results show that enzyme recycling implementation reduces production cost s for scenario S40- 5- R24 at different raw material costs . The reductions in production cost increase as enzyme and caustic costs increases. However, high biomass costs decrease the enzyme recovery benefits . Therefore, considering the high cost of NaOH , enzyme recycling seems to be beneficial when oxygen delignificati on is used. Implementation of e nzyme recycling is beneficial for scenario S40 - 5 - R24 at all the raw material cos ts evaluated. The benefits of enzyme recycling are highly dependent on raw material costs and its implementation will depend on the conditions at which the process operates. 6 .4 Conclusions Enzyme recycling by adsorption reduces cellulases loading, therefore, enzyme recycling was expected to have a positive effect on ethanol econom ics. Our results shows that enzyme recovery is viable at all the conditions economically evaluated in this analysis. However, experimentally, there are certain conditions at which enzyme recycling is not attractive due to the low cellulases recovery achieved. It was shown that short hydrolysis times increase ethanol costs due to the resulting low cellulose conversion. The economic benefits of enzyme recycling are minor when enzyme expenses are not the major contributor to raw material costs. However, with the actual estimated cost of cellulases, the enzyme recycling 190 technology can be a powerful strategy to decrease hydrolysis expenses. It was also observed that sugar losses during solid and liquid separation negatively affect production costs, consequently , optimization of the separation process is required . The greatest reduction in production cost was obtained at high enzyme loadings using substrate with low lign in content as high enzyme loadings are favorable to achieving high cellulase recoveries and ethanol cost reductions. Nonetheless, for the present system, it was more economically viable to operate at low enzyme loadings with a high lignin content substrate, conditions not favorable for enzyme recycling . Optimal operating conditions may not be suitable for enzyme recycling implementation . There are several potential strategies to increase the benefits of enzyme recovery on lignocellulosic ethanol production. Sugar losses during solid- liquid separation ha ve a significant impact on ethanol costs; therefore the separation process must be improve d. Other options include the use of surfactants to increase enzyme recoveries , and the recovery of cellulases from soli d residues by washing the solid to desorb enzymes from lignin. W ashing solid residues can also reduce sugar losses by recovering sugars in the liquid contained in the solids . Considering that enzyme cost is one of the controlling parameters in the ethanol economics, and due to the estimated high enzyme costs, enzyme recovery by adsorption is a promising technology and requires further improvements . 191 7 C o n c l u s i o n s a n d r e c o m m e n d a t i o n s 7 .1 Conclusions The overall goal of this project is to minimize lignocellulosic ethanol production cost s by optimizing operating conditions and recycling enzymes after hydrolysis . In order to prepare an economic analysis of lignocellulosic ethanol production, with and without enzyme recycling, enzymatic hydrolysis was experimentally evaluated. During the development of this work, several bottlenecks in the production of lignocellulosic ethanol were identified. Commercial enzyme preparations have been widely used to study enzymatic hydrolysis due to their high enzyme ac tivity and performance. Nonetheless, due to the lack of information about their composition and the complexity of the hydrolysis reaction, it has been difficult to draw concrete conclusions about enzymatic hydrolysis mechanism s. Adsorption of cellulases on substrate during hydrolysis has been studied through total protein concentration changes measured during hydrolysis. This approach assumed that any change in total protein concentration is caused by the cellulases adsorption onto substrate. However, it w as shown in Chapter 2 that Novozyme 188 suffer s a substantial decrease in protein concentration, probably due to protein structural changes at high temperatures . It was hypothesized that the high hydrolysis temp erature causes proteins and enzymes in Novozyme 188 to unfold leading to their aggregation and precipitation. Protein losses must be taken into account not only when working with Novozyme 188, but for any enzyme preparation. By considering and modeling protein concentration changes during hydrolysis, it is possible to more accurately determine cellulases distribution in solid and liquid phase. This information can be used to optimize enzyme loadings or to select optimum hydrolysis times for enzyme recycling. The characterization of the most widely used enzyme preparations will significantly improve our knowledge of the hydrolysis reaction . Different mechanistic models have been proposed to describe glucose yield during enzymatic hydrolysis. Reported models often require a large number of parameters and have a complex 192 solution which makes them difficult to manage. The model proposed in this project has the advantages of being a simple expression requiring few parameters . Moreover, this model links the pretre atment and hydrolysis processes by predicting the hydrolysis of substrates with different lignin contents . This makes the simultaneous optimiz ation of pretreatment and hydrolysis operating conditions possible . These stages have been reported to significantly affect ethanol economics. The negative effect of lignin during hydrolysis was introduced to the model by a lignin factor that describes the amount of cellulases adsorbed on lignin. Most of the cellulases irreversibly adsorb on lignin at the start of the reaction, which points to lignin as one of the major barriers to an efficient hydrolysis. The lignin factor was determined by fitting the proposed model to experimental data. Similar values for the adsorption of cellulases on lignin have been determined experimentally for biomass subjected to different pretreatments . The experimental determination of the lignin factor can be used to increase our knowledge of lignin- enzyme interactions and improve the proposed model. Cellulases are thought to adsorb on lignin via hydrophobic interactions, ionic bond interactions, and hydrogen bond interactions. As lignin is the major obstacle in the production of lignocellulosic ethanol , the characterization of the lignin structure is vital to improve the enzymes efficiency . Given the large range of pretreatment technologies, more research is needed to identify the effect of these t echnologies on lignin structure. Industrial enzymatic hydrolysis has been proposed to be conducted at 10 wt % solids concentration by various authors. However, high solids concentrations decrease hydrolysis yields due to diffusion limitations. Therefore, the development of reactors capable of handling high solids concentration without compromising hydrolysis yields is imperative. High solids concentration reactors are being investigated and developed by NREL . However , these reactors are in the pilot plant stage and more research is needed to accelerate their industrial application . Enzyme recycling by adsorption has been considered in various s tudies as a promising technology for reducing ethanol costs. Some enzyme recycling studies have evaluated enzyme recovery efficiency by activity or hydrolysis yields. However , it is advisable to report the mass of recycled enzymes instead, in order to economically evaluate this 193 technology and facilitate the comparison of different enzyme preparations . In order to accurately determine enzyme recovery , it is necessary to take into account the substrate’s pore distribution and moisture content , especially when working at 5 wt % or higher solids concentration, as these factors can interfere with enzyme recovery determination. In our results, lignin was shown to be the major impediment to efficient enzyme recycling as it irreversibly adsorbs cellulases. Higher enz yme recoveries were achieved by increasing enzyme loadings, although, this also increases production cost s. Hydrolysis operation at high enzyme loadings , even when implementing enzyme recycling , seems to be unviable for the process. By predicting the concentration of cellulases in solution using measured total protein concentration and predicting Novozyme 188 concentration, it was observed that after a quick adsorption stage, cellulases are slowly desorbed . Once the desorption process ends, cellulases concentration in solution remains constant for the rest of the hydrolysis. Consequently, enzyme recoveries obtained during this period were similar ( e.g. S40- 5- R24 and S40- 5- R48). This period of time seems to be the optimal hydrolysis time for enzyme recovery, as the highest sugar concentrations are achieved during this period . A mass balance of the enzyme recovery process was reported for the first time , demonstrating the possibility of monitoring the concentration changes of major components in hydrolysis. The amount of cellulases recovered from the liquid phase after hydrolysis in this work is similar to those reported in past studies. However, t he amount of cellulases recovered relative to the initial cellulases loaded determined from the mass balance has n ot been reported in past studies, making them difficult to compare with our results. The reported mass balances were successfully used to implement this technology in simulations of the ethanol production process . The methodology presented can be used to e conomically evaluate the viability of the enzyme recycling in different systems. The addition of surfactants is another option to increase hydrolysis and enzyme recovery yields by increasing the cellulases’ desorption. However, due to the high contribution of raw materials to production costs, the addition of another relatively expensive chemical must be evaluated. The estimation of the cost of the hydrolysis and fermentation reactor s is challenging due to the reactor sizes required in the process . More information is required to standardize reactor 194 costs since hydrolysis and fermentation capital costs account for approximately 68% of the total capital cost. One way to reduce the number of reactors in the process is by operating at high solids concentrations. However, the lowest production cost s obtained in this analysis were obtained operating at both 5 and 10 wt % solids concentration. Given the technical difficulties in operating a large number of reactors , e.g. preventing contamination in continuous proce sses, operating few er reactors is likely technically more advantageous. The modelling of the fermentation stage with a kinetic model can significantly improve the economic analysis. Therefore, the implementation of a kinetic model to simulate the fermentation will be the focus of future work. Raw material cost s were found to control the ethanol production cost . Due to their importance, raw material costs must be investigated in detail. Biomass cost is unknown; however, a biomass cost of $50 to $60/ ton DM has been used in several studies. Caustic cost has to be taken into account when using oxygen delignification and other pretreatment s. Independent of the pretreatment technology used, pretreatment of raw materials has been shown to be one of the controll ing variables in the determination of production cost . Pretreatment chemical costs are defined by market changes. Greater detail about the concentration and physical properties of pretreatment chemicals will improve the analysis and comparison of different process configurations. Enzyme expenses were shown to be the major contributor to production cost. T he importance of enzyme expenses has not been reported, mainly because of poorly defined and optimistic assumptions of enzyme costs. Enzyme costs have been generally reported in $/gal ethanol , however, this particular form of reporting enzyme cost can result in different enzyme cost s per mass of enzyme, depending on the process configuration. T he widely assumed enzyme cost of $0.34/gal ethanol may result in $/g enzyme close to the expected if cellulases were produced as inexpensively as soy protein. Therefore, in agreement with Klein - Marcushamer et al. (2012), more realistic and detailed enzyme cost s must be implemented in future economic analysis. If the importance of enzyme cost is not recognized, it will be difficult to make the necessary increases in the research efforts to improve enzyme s production and reduce enzyme s prices. 195 Due to the negative effect of lignin on enzymatic hydrolysis, pretreatm ent and hydrolysis conditions that result in high hydrolysis yields have been identified as the best operating conditions for the production of ethanol. However , in the present study, the lowest production costs were obtained using at the mild pretr eatment condition. Severe pretreatment conditions are required to produce substrates with low lignin contents , such as substrate S. These severe conditions also increase sugar losses; as a result, the production of ethanol from substrate S was lower than for subs trate M. Our results show ed the significant negative impact of sugar losses on the process economics. From the results obtained in this thesis, the range of pretreatment operating conditions can be narrowed in order to identify the optimal conditions to maximize lignin removal and minimize sugar losses. Sugar losses during pretreatment can be reduced b y recovering sugars from the pretreatment liquor or by decreasing the pretreatment temperature . To reduce sugar losses and production costs , pretreatment and hydrolysis conditions must be optimized simultaneously . Consequently, the hydrolytic model developed, which links pretreatment and hydrolysis process, is a valuable tool to optimize operating conditions. The economic evaluation of enzyme recycling technol ogy showed that this technology can reduce the ethanol production costs . The magnitude of the production cost reduction achieved with the enzyme recycling process is primarily related to the amount of enzymes recovered. Production cost reductions are also to some extent determined by operating conditions, caustic and biomass costs, and process configuration. Minor production cost reductions were obtained for scenarios in which enzyme expense was not the controlling economic variable. The sensitivity analysis of the process with enzyme recycling show ed that the positive impacts of this technology increase at high enzyme and caustic prices. Based on recent information , enzyme prices are expected to be higher than the optimistic enzyme costs considered in many reports ; consequently, enzyme recycling is a valuable tool to reduce operating costs. The largest production cost reductions estimated in this project were obtained by recovering approximately 36% of the loaded cellulases. These savings are higher than those estimated by Tu and Saddler (2010). In their work, the production cost saving was estimated assuming 196 a higher enzyme recovery and surfactants addition. E nzyme recycling seems to be more advantageous than previously estimated, and consequently , more improvements to this technology are needed. One of the cr ucial points during the enzyme recycling process is the sugar losses during solid- liquid separation which decrease the production of ethanol . In order to increase enzyme recycling benefits, the separation process must be improved to reduce sugar losses. By washing solid residues after separation and using the washing liquor in new hydrolysis rounds, enzyme recovery can be improved and sugar loss reduced. D ue to the low impact of capital cost s and process water expenses on production costs, the addition of a washing process appears to be a highly promising alternative to increase enzyme recycling benefits . The addition of surfactants is another option to increase hydrolysis and enzyme recovery yields by increasing the cellulases’ desorption. However, due to t he high contribution of the raw materials to the production costs, the addition of another relatively expensive chemical must be evaluated. Despite the positive effect of implementing enzyme recycling under certain conditions, it is important to note that the lowest production cost achieved in this project was obtained at conditions for which enz yme recycling was ineffective. Therefore, it may be possible that the optimal operating conditions for the process are not compatible with the enzyme recycling technology. Enzyme recycling implementation must be evaluated at the process conditions selected. This project illustrates the complexity of the lignocellulosic ethanol production process, and the need to consider all stages when optimizing this process . In order to make lignocellulosic ethanol competitive with other fuels, it is necessary to focus research efforts on the economic variables that control the ethanol economics. The link between e xperimental and economic analyses must be reinforced to effectively improve the production of lignocellulosic ethanol. 197 7 .2 Future wor k This research experimentally measured the protein concentration changes, hydrolysis of lignocellulose and enzyme recycling. Simulation and economic analysis of the production process of ethanol with and without enzyme recycling were developed to investigate the impact of different variables o n the ethanol cost. Further research needs to be done to improve the accuracy of the models proposed, performance of enzyme recycling, and to reduce the production costs of ethanol . The following is a set of recommendations for future study. • Changes in protein concentration at hydrolysis conditions were modeled for Novozyme 188. In order to extent our understanding on the enzymes and protein stability, enzyme activity and protein concentrations must be simultaneously measured. • The β- glucosidase behaviour at hydrolysis conditions may vary depending on the microorganism used for its production. Therefore, protein stability of different commercial cocktails has to be evaluated. • In order to improve the proposed model for the hydrolysis of cellulose , the value of the lignin factor must be experimentally determined . In this way, the number of estimated parameters in the model will be reduced. Similarly, the implementation of fractal kinetics in the model may improve the modeling of hydrolysis at high solids concentration. • Optimization of the cellulases and β - glucosidase loading at conditions that result in low production cost may result in larger savings. • The determination of the main components in Celluclast 1.5, during enzyme recycling, can lead to the identification of the most active cellulases in the cocktail . This information can help improve the recycling process and cocktail composition. • Improvement in the enzyme recycling performance could greatly decrease production cost. Consequently, the implementation of surfactants and solid residues washing to improve recycling performance must be experimentally and economically evaluated. 198 • Milder pretreatment conditions and lower enzyme loadings need to be evaluated to identify optimal process conditions. • The implementation of fermentation models in the simulation can increase the sensitivity of the economic analyses at different sugar concentrations and reaction times. • A pinch point analysis to minimise energy consumption can significantly reduce the production cost of ethanol . Based on our results, the impleme ntation of a boiler and turbogenerator to produce steam and energy from lignin can reduce ethanol cost. Therefore , a detail analysis of lignin processing must be performed. 199 Abengoa Bioenergy. Lignocellulosic bioethanol plant. Abreu A. 2012. Growing caustic soda stocks change world trade. ICIS Chem Bus . Aden A, Ruth M, Ibsen K, Jechura J, Neeves K, Sheehan J, Wallace B, Montague L, Slayton A, Lukas J. 2002. Lignocellulosic biomass to ethanol process design and economics utilizing co - current dilute acid prehydrolysis and enzymatic hydrolysis for corn stover. Springfield. Adney B, Baker J. 2008. Measurement of cellulase activities. Natl Renew Energy Lab:1 –8. Ahring BK, Licht D, Schmidt AS, Sommer P, Thomsen AB. 1998. Production of ethanol from wet oxidised wheat straw by Thermoanaerobacter mathranii. Bioresour Technol 68: 3 –9. Alberta- Pacific (AL - PAC) Forest Industries. 1995. Pulp mill cuts costs and chlorine with oxygen delignification. Alft rén J, Hobley TJ. 2013. Covalent immobilization of β - glucosidase on magnetic particles for lignocellulose hydrolysis. Appl Biochem Biotechnol 169:2076–87. Alvira P, Tomás - Pejó E, Ballesteros M, Negro MJ. 2010. Pretreatment technologies for an efficient bi oethanol production process based on enzymatic hydrolysis: A review. Bioresour Technol 101:4851–4861. Anex RP, Aden A, Kazi FK, Fortman J, Swanson RM, Wright MM, Satrio JA, Brown RC, Daugaard DE, Platon A, Kothandaraman G, Hsu DD, Dutta A. 2010. Techno- economic comparison of biomass - to- transportation fuels via pyrolysis, gasification, and biochemical pathways. Fuel 89:S29–S35. Aqua Merik. 2010. Rotary drum filters. Aqua Merik . http://www.aquamerik.com/catalogue/produits.cgi?category=filtration_filtreatambo urrotatif&listtype=current&lg=eng. Arantes V, Saddler JN. 2011. Cellulose accessibility limits the effectiveness of minimum cellulase loading on the efficient hydrolysis of pretreated lignocellulosic substrates. Biotechnol Biofuels 3:1 –16. Arifeen N, Wang R, Kookos IK, Webb C, Koutinas A a. 2007. Process design and optimization of novel wheat - based continuous bioethanol production system. Biotechnol Prog 23:1394–403. R e f e r e n c e s 200 Asgari F, Argyropoulos DS, Events C. 1998. Fundamentals of oxygen delignification. Part II . Functional group formation/elimination in residual kraft lignin. Can. J. Chem. 1615: 606–1615. Aspen Technology. 2011. Aspen Process Economic Analyzer V7.3.1. Burlington,. Aspen Technology Inc. 2011a. Aspen plus: user guide volume1. Burlington. Vol. 1. Aspen Technology Inc. 2011b. Aspen capital cost estimator V7.3.1. Burlington. AspenTech. 2011. Aspen Icarus Reference Guide. Burlington. Balat M, Balat H, Öz C. 2008. Progress in bioethanol processing. Prog. Energy Combust. Sci. 34: 551–573. Banerjee S, Sen R, Pandey R a., Chakrabarti T, Satpute D, Giri BS, Mudliar S. 2009. Evaluation of wet air oxidation as a pretreatment strategy for bioethanol production from rice husk and process optimization. Biomass Bioenerg 33:1680–1686. Bansal P, Hall M, Realff MJ, L ee JH, Bommarius AS. 2009. Modeling cellulase kinetics on lignocellulosic substrates. Biotechnol Adv 27:833– 48. Bastidas PA, Gil ID. 2010. Comparison of the main ethanol dehydration technologies through process simulation. 20th Eur. Symp. Comput. Aided Pr ocess Eng.1–4. Berlin A, Balakshin M, Gilkes N, Kadla J, Maximenko V, Kubo S, Saddler J. 2006. Inhibition of cellulase, xylanase and beta - glucosidase activities by softwood lignin preparations. J Biotechnol 125: 98–209. Berlin A, Girkes N, Kurabi A, Tu M, Kilburn D, Saddler J. 2005. Weak lignin- binding enzymes a novel approach to improve activity of cellulases for hydrolysis of lignocellulosics. Appl Biochem Biotechnol 121: 6 3–170. Berry H. 2002. Monte carlo simulations of enzyme reactions in two dimensions: fractal kinetics and spatial segregation. Biophys J 83:1891–901. Beta Renewables. Lignocellulosic bio - ethanol. http://www.betarenewables.com/ Bio- Rad Laboratories. 2010. Bio- Rad protein assay. Bio -rad. California. Biswas R, Uellendahl H, Ahring BK. 2014. Wet explosion pretreatment of sugarcane bagasse for enhanced enzymatic hydrolysis. Biomass Bioenerg 61: 04–113. Bjerre AB, Bjerring A, Ploger A, Schmidt AS. 1996. Pretreatment of wheat straw using combined wet oxidation and alkaline hydrolysis resultri ng in convertible cellulose and hemicellulose. Biotechnol Bioeng 49: 568–577. 201 Von Blottnitz H, Curran MA. 2007. A review of assessments conducted on bio- ethanol as a transportation fuel from a net energy, greenhouse gas, and environmental life cycle perspective. J. Clean. Prod. 15: 607–619. Brethauer S, Wyman CE. 2010. Review: Continuous hydrolysis and fermentation for cellulosic ethanol production. Bioresour Technol 101:4862–74. Chandel AK, Es C, Rudravaram R, Narasu ML, Rao V, Ravindra P. 2007. Economics and environmental impact of bioethanol production technologies : an appraisal. Biotechnol Mol Biol 2:14–32. Chang J. 2006. Indicative chemical prices A - Z (ICIS). Chem Mark Report . Charles N, Mansfield SD, Mirochnik O, Duff SJB. 2003. Effect of oxygen deli gnification operating parameters on downstream enzymatic hydrolysis of softwood substrates. Biotechnol Prog 19: 606–11. Chemical engineering plant cost index. 2013. Economic Indicators. http://www.che.com/pci/. Chundawat SPS, Lipton MS, Purvine SO, Uppugundla N, Gao D, Balan V, Dale BE. 2011. Proteomics- based compositional analysis of complex cellulase- hemicellulase mixtures. J Proteome Res 10:4365–72. Chylenski P, Felby C, Østergaard Haven M, Gama M, Selig MJ. 2012. Precipitation of Trichoderma reesei commercial cellulase preparations under standard enzymatic hydrolysis conditions for lignocelluloses. Biotechnol Lett 34:1475–82. Coppola F, Bastianoni S, Østergård H. 2009. Sustainability of bioethanol production from wheat with recycled residues as evaluat ed by emergy assessment. Biomass Bioenerg 33:1626–1642. Dekker RFH, Assays E. 1986. Kinetic, inhibition, and stability properties of a commercial p-0- gluwsidase (cellobiase) preparation from Aspergillus niger and its suitability in the hydrolysis of lignocellulose. Biotechnol Bioeng 2 8 :1438–1442. Demirbas A. 2005. Bioethanol from cellulosic materials: a renewable motor fuel from biomass. Energy Sources 27: 3 7–337. Dien BS, Li X - L, Iten LB, Jordan DB, Nichols NN, O’Bryan PJ, Cotta M a. 2006. Enzymatic saccharification of hot - water pretreated corn fiber for production of monosaccharides. Enzym Microb Technol 3 9 :1137–1144. Divne C, Ståhlberg J, Teeri TT, Jones T a. 1998. High- resolution crystal structures reveal how a cellulose chain is bound in the 50 A long tunnel of cellobiohydrolase I from Trichoderma reesei. J Mol Biol 275: 309–25. 202 Djioleu AC, Arora A, Martin EM, Smith JA, Pelkki MH, Carrier DJ. 2012. Sugar recovery from the pretreatment/ enzymatic hydrolysis of high and low specific gravity poplar clones. Agric Food Anal Bacteriol 2:121–131. Dow. 2010. Caustic soda solution handbook. www.dowcaustic.com DuPont Biofuels Solutions. Cellulosic ethanol. http://biofuels.dupont.com/cellulosic -ethanol/neva Eckard AD, Muthukumarappan K, Gibbons W. 2013. Enzyme r ecycling in a simultaneous and separate saccharification and fermentation of corn stover: a comparison between the effect of polymeric micelles of surfactants and polypeptides. Bioresour Technol 132: 02–9. Eigen M. 1996. Prionics or the kinetic basis of prion diseases. Biophys. Chem. 63:A1 –18. Eisenberg H. 1971. Glutamate dehydrogenase: anatomy of a regulatory enzyme. Acc. Chem. Res. 306: 379–385. Environment Canada. 2011. Federal renewable fuels regulations: overview. Minist. Environ. https://www.ec.gc.ca/ energie- energy/default.asp?lang=En&n=828C9342- 1. Existentes M, Zaia CTB V. 1998. Deterninacao de proteinas totais via espectrofometria: vantagens e desvantagens dos metodos existentes. Quim. Nova 21: 787–793. Farinas CS, Loyo MM, Baraldo A, Tardioli PW, Net o VB, Couri S. 2010. Finding stable cellulase and xylanase: evaluation of the synergistic effect of pH and temperature. N Biotechnol 27:810–815. Fassel M. 1974. Reactor for wet oxydation of organic matter. United States of America Patent. FlowTec Industri etechnik GmbH. 2013. DualOx enhanced process. http://www.flowtec.at/index.php/en/content/view/full/1361. Galbe M, Sassner P, Wingren A, Zacchi G. 2007. Process engineering economics of bioethanol production. Adv Biochem Engin/Biotechnol 108: 3 3–327. Galbe M, Zacchi G. 2002. A review of the production of ethanol from softwood. Appl Microbiol Biotechnol 59: 618–28. Galbe M, Zacchi G. 2007. Pretreatment of lignocellulosic materials for efficient bioethanol production. Adv Biochem Engin/Biotechnol 108:41–65. 203 Gallego JP. 2002. The use of wet oxidation and PACT ® for the treatment of propylene oxide/styrene monomer (PO/SM) industrial wastewaters at the Repsol PO/SM plant in Tarragona, Spain. Tarragona. García - Cubero M a T, González - Benito G, Indacoechea I, Coca M, Bolado S. 2009. Effect of ozonolysis pretreatment on enzymatic digestibility of wheat and rye straw. Bioresour Technol 100: 608–13. Gas processors suppliers association. 2004. Engineering data book. Oklahoma. Gebreyohannes S. 2007. Process design and economic evaluation od an ethanol production process by biomass gasification. Bahir Dar University. Gierer J. 1985. Chemistry of delignification. Wood Sci. Technol. 312: 89–312. Gierer J. 1997. Formation and involvement of superoxide and hydroxyl radicals in TCG bleaching processes: A review. Holzforschung 51 : 34–46. Gierer J, Reitberger T, Yang E, Yoon B - H. 2001. Formation and Involvement of Radicals in Oxygen Delignification Studied By the Autoxidation of Lignin and Carbohydrate Model Compounds. J Wood Chem T echnol 21: 3 3–341. Gnansounou E, Dauriat A. 2010. Techno- economic analysis of lignocellulosic ethanol: A review. Bioresour Technol 101:4980–91. De Gooijer CD, Bakker W a. M, Beeftink HH, Tramper J. 1996. Bioreactors in series: An overview of design proce dures and practical applications. Enzym Microb Technol 18:202–219. Goyal A, Ghosh B, Eveleigh D. 1991. Characteristics of fungal cellulases. Bioresour Technol 36: 37–50. Gray K a, Zhao L, Emptage M. 2006. Bioethanol. Curr. Opin. Chem. Biol. 10: 41–6. Gregg DJ, Boussaid A, Saddler JN. 1998. Techno- economic cellulose evaluation of a generic wood- process: effect of increased yields and enzyme recycle. Bioresour Technol 63: 7 –12. Grethlein HE. 1985. The effect of pore size distribution on the rate of enzymatic hydrolysis of cellulosic substrates. Nat Biotechnol 3:155–160. Guillemette T, van Peij NNME, Goosen T, Lanthaler K, Robson GD, van den Hondel C a MJJ, Stam H, Archer DB. 2007. Genomic analysis of the secretion stress response in the enzyme - producing cell factory Aspergillus niger. BMC Genomics 8:158–75. 204 Hahn- Hägerdal B, Galbe M, Gorwa- Grauslund MF, Lidén G, Zacchi G. 2006. Bio- ethanol the fuel of tomorrow from the residues of today. Trends Biotechnol. 24: 5 9–56. Hamelinck CN, Hooijdonk G Van, Faaij AP. 2005a. Ethanol from lignocellulosic biomass: techno- economic performance in short - , middle- and long- term. Biomass Bioenerg 28: 3 4–410. Hamelinck CN, Hooijdonk G Van, Faaij AP. 2005b. Ethanol from lignocellulosic biomass: techno- economic performance in short - , middle- and long- term. Biomass Bioenerg 28: 3 4–410. Haven MO, Jørgensen H. 2013. Adsorption of β - glucosidases in two commercial preparations onto pretreated biomass and lignin. Biotechnol Biofuels 6:165. Helfferich FG. 2003. Kinetics of homogeneous multistep reactions. Ed. R. G. Compton 2nd ed. Amsterdam: Elseiver Science. Helmich KE, Banerjee G, Bianchetti CM, Gudmundsson M, Sandgren M, Walton JD, Jr. GNP. Crystal structure of beta - d- glucoside glucohydrolase from Trichoderma reesei. To be Publ. http ://www.rcsb.org/pdb/explore/explore.do?structureId=4I8D. Higuchi T. 2006. Look back over the studies of lignin biochemistry. J Wood Sci 52: –8. Hodge DB, Karim MN, Schell DJ, McMillan JD. 2008. Soluble and insoluble solids contributions to high- solids enz ymatic hydrolysis of lignocellulose. Bioresour Technol 99:8940–8. Hu J, Arantes V, Pribowo A, Saddler JN. 2013. The synergistic action of accessory enzymes enhances the hydrolytic potential of a “cellulase mixture” but is highly substrate specific. Biotechnol Biofuels 6:112–24. Hu J, Arantes V, Saddler JN. 2011. The enhancement of enzymatic hydrolysis of lignocellulosic substrates by the addition of accessory enzymes such as xylanase: is it an additive or synergistic effect? Biotechnol Biofuels 4: 36. Huang HJ, Ramaswamy S, Al - Dajani W, Tschirner U, Cairncross RA. 2009. Effect of biomass species and plant size on cellulosic ethanol: A comparative process and economic analysis. Biomass Bioenerg 33:234–246. Humbird D, Davis R, Tao L, Kinchin C, Hsu D, Aden A. 2011. Process design and economics for biochemical conversion of lignocellulosic biomass to ethanol. Natl Renew Energy Lab. Golden. Hurvich CM, Tsai C. 2007. Regression and time series model selection in small samples. Biometroika 76:297–307. 205 Ibrahim M. 1998. Clean fractionation of biomass steam explotion and extraction; Virginia Polytechnic institute and state university. Institute of work and health. 2005. Statistical significance. Work. https://www.iwh.on.ca/wrmb/statistical- significance. Jin M, Gunawan C, Uppugundla N, Balan V, Dale BE. 2012. A novel integrated biological process for cellulosic ethanol production featuring high ethanol productivity, enzyme recycling and yeast cells reuse. Energy Environ. Sci. 5: 7168. Jørgensen H, Kristensen JB, Felby C . 2007. Enzymatic conversion of lignocellulose into fermentable sugars: challenges and opportunities. Biofuel Bioprod Biorefin 1: 19–134. Kazi FK, Fortman JA, Anex RP, Hsu DD, Aden A, Dutta A, Kothandaraman G. 2010. Techno- economic comparison of process te chnologies for biochemical ethanol production from corn stover. Fuel 89:S20–S28. Kiese S, Papppenberger A, Friess W, Mahler H. 2008. Shaken , Not Stirred : Mechanical Stress Testing of An IgG1 Antibody. J. Pharm. Sci. 97:4347–4366. Kim TH, Lee YY. 2007. Pretreatment of corn stover by soaking in aqueous ammonia at moderate temperatures. Appl Biochem Biotechnol 136–140:81–92. Kindsigo M, Kallas J. Degradation of lignins by wet oxidation: model water solutions . Proc Estonian Acad Sci Chem 8:132–144. Klein - Marcuschamer D, Oleskowicz - Popiel P, Simmons BA, Blanch HW. 2012. The challenge of enzyme cost in the production of lignocellulosic biofuels. Biotechnol Bioeng 109: 083–7. Klein - Marcuschamer D, Oleskowicz - Popiel P, Simmons B a., Blanch HW. 2010. Technoeconomic analysis of biofuels: A wiki - based platform for lignocellulosic biorefineries. Biomass Bioenerg 34:1914–1921. Kleywegt GJ, Zou JY, Divne C, Davies GJ, Sinning I, Stâhlberg J, Reinikainen T, Srisodsuk M, Teeri TT, Jones T a. 1997. The crystal structure of the catalytic core domain of endoglucanase I from Trichoderma reesei at 3.6 A resolution, and a comparison with related enzymes. J Mol Biol 272: 383–97. Klinke HB, Olsson L, Thomsen a B, Ahring BK. 2003. Potential inhibitors from wet oxidation of wheat straw and their effect on ethanol production of Saccharomyces cerevisiae: wet oxi dation and fermentation by yeast. Biotechnol Bioeng 81: 73 –47. Klinke HB, Ahring BK, Schmidt AS, Thomsen AB. 2002. Characterization of degradation products from alkaline wet oxidation of wheat straw. Bioresour Technol 82:15–26. 206 Koo B - W, Treasure TH, Jameel H, Phillips RB, Chang H - M, Park S. 2011. Reduction of enzyme dosage by oxygen delignification and mechanical refining for enzymatic hydrolysis of green liquor - pretreated hardwood. Appl Biochem Biotechnol 165:832–44. Kopelman R. 1986. Rate processes on fractals : theory, simulations, and experiments. J Stat Phys 42:185–200. Kristensen JB, Felby C, Jørgensen H. 2009. Yield- determining factors in high- solids enzymatic hydrolysis of lignocellulose. Biotechnol Biofuels 2:11. Kristensen JB, Börjesson J, Bruun MH, Tjerneld F, Jørgensen H. 2007. Use of surface active additives in enzymatic hydrolysis of wheat straw lignocellulose. Enzym Microb Technol 40:888–895. Kudou M, Shiraki K, Fujiwara S, Imanaka T, Takagi M. 2003. Prevention of thermal inactivation and aggregation of lysozyme by polyamines. Eur J Biochem 27 0:4547–4554. Kumar D, Murthy GS. 2011. Impact of pretreatment and downstream processing technologies on economics and energy in cellulosic ethanol production. Biotechnol Biofuels 4:27. Kumar R, Wyman CE. 2009a. Effect of enzyme supplementation at moderate cellulase loadings on initial glucose and xylose release from corn stover solids pretreated by leading technologies. Biotechnol Bioeng 102:457–67. Kumar R, Wyman CE. 2009b. Access of cellulase to ce llulose and lignin for poplar solids produced by leading pretreatment technologies. Biotechnol Prog 25:807–819. Larsen J, Haven MØ, Thirup L. 2012. Inbicon makes lignocellulosic ethanol a commercial reality. Biomass Bioenerg 46: 36– 5. Lawoko M, Nutt A, He nriksson H, Gellerstedt G, Henriksson G. 2000. Hemicellulase activity of aerobic fungal cellulases. Holzforschung 54: 97–500. Laxman RS, Lachke AH. 2009. Bioethanol lignocellulosic biomass. In: . Handb plant -based biofuels, pp. 121–139. Lee D, Owens VN, B oe A, Jeranyama P. 2007. Composition of herbaceous biomass feedstocks. South Dakota State University. Lee M. 2011. Lignin model compound approach to modelling oxygen delignification reaction kinetics. University of British Columbia. Lee S. 2007. Ethanol fr om lignocellulosics. In: . Handb Altern Fuel Technol. Taylor and Francis group, pp. 344–372. 207 Lehninger A, Nelson DL, Cox M. 2005. Lehninger principles of biochemistry. Ed. Albert L. Lehninger 4th ed. W.H. Freeman & Company. Leistritz FL, Stowers MD, Mcdona ld WF. 2006. Preliminary feasibility analysis for an integrated biomaterials and ethanol biorefinery using wheat straw feedstock. North Dakota. Li B, Shen Y, Li B. 2008. Quasi- steady- state laws in enzyme kinetics. J Phys Chem 112: 311–21. Liao W, Liu Y, W en Z, Frear C, Chen S. 2008. Kinetic modeling of enzymatic hydrolysis of cellulose in differently pretreated fibers from dairy manure. Biotechnol Bioeng 101:441–51. Liberek K, Lewandowska A, Zietkiewicz S. 2008. Chaperones in control of protein disaggregation. EMBO J 27: 3 8–35. Lin SY, Lebo SE. 2001. Lignin. In: John Wiley and Sons, I, editor. Kirk -Othmer Encycl Chem Technol, pp. 1–32. Lin Y, Tanaka S. 2006. Ethanol fermentation from biomass resources: current state and prospects. Appl Microbiol Biotechnol 69: 627–42. Linde group. 2013. Oxygen delignification: cost efficient bleaching of wood pulp. Processes . http://www.lindeus.com/en/processes/cleaning_polishing_grinding/bleaching/oxygen_delignification/index.html. Lindedam J, Haven MØ, Chylenski P, Jørgensen H, Felby C. 2013. Recycling cellulases for cellulosic ethanol production at industrial relevant conditions: potential and temperature dependency at high solid processes. Bioresour Technol 148: 80–8. Lissens G, Klinke H, Verstraete W, Ahring B, Thomsen AB. 2004. Wet oxidation pre -treatment of woody yard waste: parameter optimization and enzymatic digestibility for ethanol production. J. Chem. Technol. Biotechnol. 79:889–895. Ljunggren M. 2005. Kinetic analysis and modeling of enzymatic hydrolysis and S SF. Cellulose 2:1 –6. Lu X, Sun J, Nimtz M, Wissing J, Zeng A - P, Rinas U. 2010. The intra- and extracellular proteome of Aspergillus niger growing on defined medium with xylose or maltose as carbon substrate. Microb Cell Fact 9:23–36. Lu Y, Yang B, Gregg D, Saddler JN, Mansfield SD. 2002. Cellulase adsorption and an evaluation of enzyme recycle during hydrolysis of steam - exploded softwood residues. Appl Biochem Biotechnol 98- 100: 641–54. 208 Lutz W, Sanderson WC, Scherbov S. 2004. The end of world population gr owth in the 21st century new challenges for human capital formation and sustainable development Earthscan. London: James & James. Vol. 44. Lynd LR, Laser MS, Bransby D, Dale BE, Davison B, Hamilton R, Himmel M, Keller M, McMillan JD, Sheehan J, Wyman CE. 2008. How biotech can transform biofuels. Nat Biotechnol 26:169–72. Lynd LR, Cushman JH, Nichols RJ, Wyman CE. 1991. Fuel ethanol from cellulosic biomass. Science 251: 318–1323. Mabee WE, Saddler JN. 2010. Bioethanol from lignocellulosics: Status and perspe ctives in Canada. Bioresour Technol 101:4806–13. Martinez D, Berka RM, Henrissat B, Saloheimo M, Arvas M, Baker SE, Chapman J, Chertkov O, Coutinho PM, Cullen D, Danchin EGJ, Grigoriev I V, Harris P, Jackson M, Kubicek CP, Han CS, Ho I, Larrondo LF, de Le on AL, Magnuson JK, Merino S, Misra M, Nelson B, Putnam N, Robbertse B, Salamov A a, Schmoll M, Terry A, Thayer N, Westerholm - Parvinen A, Schoch CL, Yao J, Barabote R, Barbote R, Nelson MA, Detter C, Bruce D, Kuske CR, Xie G, Richardson P, Rokhsar DS, Luca s SM, Rubin EM, Dunn- Coleman N, Ward M, Brettin TS. 2008. Genome sequencing and analysis of the biomass- degrading fungus Trichoderma reesei (syn. Hypocrea jecorina). Nat Biotechnol 26: 553–60. Maurer SA, Bedbrook CN, Radke CJ. 2012. Cellulase adsorption and reactivity on a cellulose surface from flow ellipsometry. Ind Eng Chem Res 51: 1389–11400. Morris AM, Watzky MA, Finke RG. 2009. Protein aggregation kinetics, mechanism, and curve- fitting. Biochim Biophys Acta 1794: 375–397. Mosier N, Wyman C, Dale B, Ela nder R, Lee YY, Holtzapple M, Ladisch M. 2005. Features of promising technologies for pretreatment of lignocellulosic biomass. Bioresour Technol 96: 673–86. Nakagame S, Chandra RP, Kadla JF, Saddler JN. 2011. The isolation, characterization and effect of l ignin isolated from steam pretreated Douglas - fir on the enzymatic hydrolysis of cellulose. Bioresour Technol 102:4507–17. Nakagame S, Chandra RP, Saddler JN. 2010. The effect of isolated lignins, obtained from a range of pretreated lignocellulosic substrates, on enzymatic hydrolysis. Biotechnol Bioeng 105:871–9. National Renewable Energy Laboratory (NREL). 2011. High solids enzymatic hydrolysis reactors. Natl Renew Energy Lab. Colorado. 209 Nigam JN. 2001. Ethanol production f .rom wheat straw hemicellulose hy drolysate by Pichia stipitis. J Biotechnol 87:17 –27. O’Dwyer JP, Zhu L, Granda CB, Chang VS, Holtzapple MT. 2008. Neural network prediction of biomass digestibility based on structural features. Biotechnol Prog 24: 83–92. Ochoa- villarreal M, Aispuro - hernández E, Vargas - arispuro I, Martínez - téllez MÁ. 2012. Plant cell wall polymers : function, structure and biological activity of their derivatives. Polymerization. Creative Commons Attribution, p. 66.. Oliva JM, Sáez F, Ballesteros I, González A, Negro MJ, Manzanares P, Ballesteros M. 2003. Effect of lignocellulosic degradation compounds from steam explosion pretreatment on ethanol fermentation by thermotolerant yeast Kluyveromyces marxianus. Appl Biochem Biotechnol 105 - 108: 41–53. Olsson L, Hahn- Hägerdal B . 1996. Fermentation of lignocellulosic hydrolysates for ethanol production. Enzym Microb Technol 18: 3 2–331. Oosawa F, Kasai M. 1962. A theory of linear and helical aggregations of macromolecules. J Mol Biol 4:10–21. Palonen H, Thomsen AB, Tenkanen M, S chmidt AS, Viikari L. 2004. Evaluation of wet oxidation pretreatment for enzymatic hydrolysis of softwood. Appl Biochem Biotechnol 117: –17. Panagiotou G, Olsson L. 2007. Effect of compounds released during pretreatment of wheat straw on microbial growth and enzymatic hydrolysis rates. Biotechnol Bioeng 96:250–258. Pareek N, Gillgren T, Jönsson LJ. 2013. Adsorption of proteins involved in hydrolysis of lignocellulose on lignins and hemicelluloses. Bioresour Technol 148: 70–7. Pérez J a., Ballesteros I, Bal lesteros M, Sáez F, Negro MJ, Manzanares P. 2008. Optimizing liquid hot water pretreatment conditions to enhance sugar recovery from wheat straw for fuel - ethanol production. Fuel 87: 3640–3647. Peri S, Karra S, Lee YY, Karim MN. 2007. Modeling intrinsic ki netics of enzymatic cellulose hydrolysis. Biotechnol Prog 23: 6 6–37. Piccolo C, Bezzo F. 2009. A techno- economic comparison between two technologies for bioethanol production from lignocellulose. Biomass Bioenerg 33:478–491. Piebalgs A. 2007. Renewables make the difference. European Commissioner for Energy . 210 Pope D. 2011. Assessing the potential for bioethanol production from oxygen delignified corn stover and wheat straw . University of British Columbia. Pribowo A, Arantes V, Saddler JN. 2012. The adsorpti on and enzyme activity profiles of specific Trichoderma reesei. Enzym Microb Technol 50:195–203. Pribowo YA. 2014. Enzyme - substrate interactions and their influence on enzyme recycling strategies as a way of reusing cellulases . University of British Columbia. Qi B, Chen X, Su Y, Wan Y. 2011. Enzyme adsorption and recycling during hydrolysis of wheat straw lignocellulose. Bioresour Technol 102 :2881–9. Qing Q, Wyman CE. 2011. Supplementation with xylanase and β - xylosidase to reduce xylo-oligomer and xylan inhibition of enzymatic hydrolysis of cellulose and pretreated corn stover. Biotechnol Biofuels 4:18. Qing Q, Yang B, Wyman CE. 2010. Xylooligomers are strong inhibitors of cellulose hydrolysis by enzymes. Bioresour Technol 101: 9624–30. Quirin M, Gatner SO, Pehnt M, Reinhardt GA. 2004. CO2 Mitigation through biofuels in the transport sector. Heidelberg. Rodrigues AC, Leitão AF, Moreira S, Felby C, Gama M. 2012. Recycling of cellulases in lignocellulosic hydrolysates using alkaline elution. Bioresour Technol 110: 526–33. Roefs SP, De Kruif KG. 1994. A model for the denaturation and aggregation of beta -lactoglobulin. Eur J Biochem 226:883–9. Rosgaard L, Andric P, Dam- Johansen K, Pedersen S, Meyer AS. 2007. Effects of substrate loading on enzymatic hydrolysis and viscosity of pretreated barley straw. Appl Biochem Biotechnol 143:27–40. Rowell RM, Pettersen R, Han JS, Rowell JS, Tshabalala MA, Service F. 2005. Cell wall chemistry. In: Rowell, editor. Handb wood Chem wood Compos . Florida: CRC Press, pp. 37–74. Ruffe ll J, Levie B, Helle S, Duff S. 2010. Pretreatment and enzymatic hydrolysis of recovered fibre for ethanol production. Bioresour Technol 101:2267–72. Ruiz - Dueñas FJ, Martínez AT. 2009. Microbial degradation of lignin: how a bulky recalcitrant polymer is efficiently recycled in nature and how we can take advantage of this. Microb. Biotechnol. 2:164–77. SAS Institute Inc. 2012. JMP 10 modeling and multivariate methods. Cary, NC: Institute Inc. 211 Sassner P, Galbe M, Zacchi G. 2008. Techno - economic evaluation of bioethanol production from three different lignocellulosic materials. Biomass Bioenerg 32:422–430. Schlieker C, Bukau B, Mogk A. 2002. Prevention and reversion of protein aggregation by molecular chaperones in the E. coli cytosol: implications for their applicability in biotechnology. J Biotechnol 96:13–21. Schmidt AS, Thomsen AB. 1998. Optimization of wet oxidation pretreatment of wheat straw. Bioresour Technol 64:139–151. Schwietzke S, Ladisch M, Russo L, Kwant K. 2008. Gaps in the research of 2nd Generation transportation biofuels. Golden. Vol. 41. Seider WD, Seader JD, Lewin DR. 2009. Product and process design principles: synthesis, analysis and design 2nd. ed. New York: McGraw - Hill. SERVA Electrophoresis. 2013. Cellulases. Vol. 49. http://fe.rmutl.ac.th/2012/wp-content/uploads/Cellulase.pdf. Shao X, Lynd L, Wyman C, Bakker A. 2009. Kinetic modeling of cellulosic biomass to ethanol via simultaneous saccharification and fermentation: Part I. Accommodation of intermittent feeding and analysis of staged reactors. Biotechnol Bioeng 102: 59–65. Shen J, Agblevor FA. 2008a. Kinetics of enzymatic hydrolysis of steam - exploded cotton gin waste. Chem Eng Commun 195: 107–1121. Shen J, Agblevor FA. 2008b. Optimization of enzyme loading and hydrolytic time in the hydrolysis of mixtures of cotton gin waste and recycled paper sludge for the maximum profit rate. Biochem Eng J 41:241–250. Shi J, Ebrik M a, Yang B, Garlock RJ, Balan V, Dale BE, Pallapolu VR, Lee YY, Kim Y, Mosier NS, Ladisch MR, Holtzapple MT, Falls M, Sierra - Ramirez R, Donohoe BS, Vinzant TB, Elander RT, Hames B, Thomas S, Warner RE, Wyman CE. 2011. Application of cellulase and hemicellulase to pure xylan, pure cellulose, and switchgrass solids from leading pretreatments. Bioresour Technol 102: 1080–8. Shiraki K, Kudou M, Sakamoto R, Yanagihara I, Takagi M. 2005. Amino acid esters prevent thermal inactivation and aggregation of lysozyme. Biotechnol Prog 21: 640–3. Simo M, Brown CJ, Hlavacek V. 2008. Simulation of pressure swing adsorption in fuel ethanol production process. Comput Chem Eng 32:1635–1649. Singh A, Pant D, Korres NE, Nizami A - S, Prasad S, Murphy JD. 2010. Key issues in life cycle assessment of ethanol production from lignocellulosic biomass: Challenges and perspectives. Bioresour Technol 101: 5 03–12. 212 Sivakumar G, Vail DR, Xu J, Burner DM, Lay JO, Ge X, Weathers PJ. 2010. Bioethanol and biodiesel: Alternative liquid fuels for future generations. Eng Life Sci 10:8 –18. Von Sivers M, Zacchi G. 1996. Ethanol from lignocellulosics: A review of the economy. Bioresour Technol 56:131–140. Sluiter A, Hames B, Ruiz R, Scarlata C, Sluiter J, Templeton D. 2010. Determination of structural carbohydrates and lignin in biomass determination of structural carbohydrates and lignin in biomass. Natl Renew Energy Lab. Colorado. Steele B, Raj S, Nghiem J, Stowers M. 2005. Enzyme recovery and recycling following hydrolysis of ammonia fiber explosion – treated corn stover. Appl Biochem Biotechnol 121: 901–910. Suzuki H, Cao J, Jin F, Kishita a., Enomoto H, Moriya T. 2006. Wet oxidation of lignin model compounds and acetic acid production. J. Mater. Sci. 41: 591–1597. Taherzadeh MJ, Karimi K. 2007. Enzyme - based hydrolysis processes for ethanol from lignocellulosic materials: a review. Bioresour 2: 707–738. Talebnia F, Karakashev D, Angelidaki I. 2010. Production of bioethanol from wheat straw: An overview on pretreatment, hydrolysis and fermentation. Bioresour Technol 101:4744–53. Tanaka A. 1991. Differential scanning calorimetric study of the thermal denaturation of almond beta- glucosidase. Agric Biol Chem 55:2773–2776. Tao L, Aden A, Elander RT, Pallapolu VR, Lee YY, Garlock RJ, Balan V, Dale BE, Kim Y, Mosier NS, Ladisch MR, Falls M, Holtzapple MT, Sierra R, Shi J, Ebrik M a, Redmond T, Yang B, Wyman CE, Hames B, Thomas S, Warner RE. 2011. Process and technoeconomic analysis of leading pretreatment technologies for lignocellulosic ethanol production using switchgrass. Bioresour Technol 102: 1105–14. Tao L, Tan ECD, Mccormick R, Zhang M, Aden A, He X, Zigler BT. 2013. Techno-economic analysis and life- cycle assessment of cellulosic isobutanol and comparison with cellulosic ethanol and n- butanol. Biofuel Bioprod Biorefin 8: 30–48. Tomás - Pejó E, Oliva JM, González a., Ballesteros I, Ballesteros M. 2009. Bioethanol production from wheat straw by the thermotolerant yeast Kluyveromyces marxianus CECT 10875 in a simultaneous saccharification and fermentation fed- batch process. Fuel 88:2142–2147. Torget R, Hatzis C, Hayward TKAY, Hsu T. 1996. Optimization of reverse - flow, t wo -temperature, dilute - acid pretreatment to enhance biomass conversion to ethanol. Appl Biochem Biotechnol 57:85 –100. 213 Tu M, Chandra RP, Saddler JN. 2007a. Evaluating the distribution of cellulases and the recycling of free cellulases during the hydrolysis of lignocellulosic substrates. Biotechnol Prog 23: 398–406. Tu M, Chandra RP, Saddler JN. 2007b. Recycling cellulases during the hydrolysis of steam exploded and ethanol pretreated Lodgepole pine. Biotechnol Prog 23:1130–7. Tu M, Pan X, Saddler JN. 2009a. Adsorption of cellulase on cellulolytic enzyme lignin from lodgepole pine. J Agric Food Chem 57: 7771–8. Tu M, Saddler JN. 2010. Potential enzyme cost reduction with the addition of surfactant during the hydrolysis of pretreated softwood. Appl Biochem Bio technol 16 1:274–87. Tu M, Zhang X, Paice M, MacFarlane P, Saddler JN. 2009b. The potential of enzyme recycling during the hydrolysis of a mixed softwood feedstock. Bioresour Technol 100: 6407–15. U.S. Department of Commerce. 2014. United States Census Bur eau. http://www.census.gov/en.html. U.S. Department of Energy. 2001. Biomass Multi - year program plan. Energy Efficiency and Renewable Energy (EERE) . U.S. Department of Energy. 2013a. Ethanol blends. Altern Fuels Data Cent . http://www.afdc.energy.gov/fuels/ ethanol_blends.html. U.S. Department of Energy. 2013b. Alternative fuel price. Altern Fuels Data Cent . http://www.afdc.energy.gov/fuels/prices.html. Union gas. 2012. Chemical composition of natural gas. Union gas Ltd. http://www.uniongas.com/aboutus/aboutng/composition.asp. USDA Livestock poultry and grain market news. 2014. National weekly ethanol summary. http: www.ams.usda.gov/LPSMarketNewsPage Väljamäe P, Kipper K, Pettersson G, Johansson G. 2003. Synergistic cellulose hydrolysis can be described in terms of fractal - like kinetics. Biotechnol Bioeng 84:254–7. Väljamäe P, Sild V, Pettersson G, Johansson G. 1998. The initial kinetics of hydrolysis by cellobiohydrolases I and II is consistent with a cellulose surface - erosion model. Eur J Biochem 253:469–75. Várnai A, Viikari L, Marjamaa K, Siika - aho M. 2011. Adsorption of monocomponent enzymes in enzyme mixture analyzed quantitatively during hydrolysis of lignocellulose substrates. Bioresour Technol 102: 220–7. 214 Verheul M, Roefs SPFM, Kruif KG De. 1998. Kinet ics of heat - induced aggregation of beta -lactoglobulin. J Agric Food Chem 8561: 96–903. Wang M, Wu M, Huo H. 2007. Life - cycle energy and greenhouse gas emission impacts of different corn ethanol plant types. Environ Res Lett 2:024001. Wang W. 2005. Protein aggregation and its inhibition in biopharmaceutics. Int J Pharm 289:1 –30. Watzky M, Morris AM, Ross ED, Finke RG. 2008. Fitting yeast and mammalian prion aggregation kinetic data with the Finke - Watzky two - step model of nucleation and autocatalytic growth . Biochemistry 47:10790–10800. Weil J, Brewer M, Hendrickson R, Sarikaya A, Ladisch MR. 1998. Continuous pH monitoring during pretreatment of yellow poplar wood sawdust by pressure cooking in water. Appl Biochem Biotechnol 70- 72: 99–111. Weiss N, Börjesso n J, Pedersen LS, Meyer AS. 2013. Enzymatic lignocellulose hydrolysis: Improved cellulase productivity by insoluble solids recycling. Biotechnol Biofuels 6: 5. Wilcox R. 2004. Oxygen costs. http://people.clarkson.edu/~wwilcox/Design/oxycost.htm#_ftn2. Wing ren A, Galbe M, Zacchi G. 2003. Techno- economic evaluation of producing ethanol from softwood: comparison of SSF and SHF and identification of bottlenecks. Biotechnol Prog 19: 109–17. Wingren A, Galbe M, Zacchi G. 2005. Effect of reduction in yeast and enzyme concentrations in a simultaneous- saccharification - and- fermentation – based bioethanol process. Appl Biochem Biotechnol 121- 124: 85–499. Wood SM, Layzell DB. 2003. A canadian biomass inventory: feedstocks for a bio- based economy. BIOCAP Canada Foundation. Wood TM, Bhat KM. 1988. Methods for measuring cellulase activities. Methods Enzymol 160: 87–112. Workman WE, Day DF. 1982. Purification and properties of β - Glucosidase from Aspergillus terreus. Appl. Environ. Microbiol. 44:1289–1295. Wyman CE, Dale BE, Elander RT, Holtzapple M, Ladisch MR, Lee YY. 2005. Coordinated development of leading biomass pretreatment technologies. Bioresour Technol 96:1959–66. Wyman CE. 1999. Biomass ethanol: technical progress, opportunities, and commercial challenges. Annu. Rev. Energy Environ. 24:189–226. 215 Xue Y, Jameel H, Park S. 2012. Strategies to recycle enzyme and their impact on enzymatic hydrolysis for bioethanol production. Bioresour 7: 602–615. Yang B, Wyman CE. 2008. Pretreatment : the key to unlocking low- cost cellulosic ethanol. Biofuels Bioprod Bior 2: 6 –40. Yang R, Lucia L, Ragauskas AJ, Jameel H. 2003. Oxygen delignification chemistry and its impact on pulp fibers. J Wood Chem Technol 23:13–29. Yu Z, Jameel H, Chang H - M, Park S. 2011. The effect of delignification of forest biomass on enzymatic hydrolysis. Bioresour Technol 102: 9 83–9. Zhang YHP, Lynd LR. 2004. Toward an aggregated understanding of enzymatic hydrolysis of cellulose: noncom plexed cellulase systems. Biotechnol Bioeng 88: 797– 24. Zhang Y, Xu J - L, Xu H - J, Yuan Z - H, Guo Y. 2010. Cellulase deactivation based kinetic modeling of enzymatic hydrolysis of steam- exploded wheat straw. Bioresour Technol 101:8261–6. Zheng Y, Pan Z, Zha ng R, Jenkins BM. 2009. Kinetic modeling for enzymatic hydrolysis of pretreated creeping wild ryegrass. Biotechnol Bioeng 102: 558–69. Zhou J, Wang Y - H, Chu J, Luo L - Z, Zhuang Y - P, Zhang S - L. 2009a. Optimization of cellulase mixture for efficient hydrolys is of steam - exploded corn stover by statistically designed experiments. Bioresour Technol 100:819–25. Zhou Y, Su S, Song M, Nie H - L, Zhu L - M. 2009b. Improving the stability of cellulase by immobilization on chitosan - coated magnetic nanoparticles modified with α - ketoglutaric acid. Bioinforma Biomed Eng:1 –4. Zhu Z, Sathitsuksanoh N, Zhang P. 2009. Direct quantitative determination of adsorbed cellulase on lignocellulosic biomass with its application to study cellulase desorption for potential recycling. Anal yst 134:2267–72. 216 A.1 Novozyme 188 models residuals analysis The residuals were normalized with the experimental data defined as : 𝑁𝑜𝑟𝑚𝑎𝑙𝑖𝑧𝑒𝑑 𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙 𝑖 =𝐸𝑥𝑝𝑒𝑟𝑖𝑚𝑒𝑛𝑡𝑎𝑙 𝑣𝑎𝑙𝑢𝑒 𝑖 − 𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑 𝑣𝑎𝑙𝑢𝑒 𝑖𝐸𝑥𝑝𝑒𝑟𝑖𝑚𝑒𝑛𝑡𝑎𝑙 𝑣𝑎𝑙𝑢𝑒 𝑖 52 The residuals for the experimental data and equation 6 are shown in Figure 55. Figure 55. Residuals for the predicted P - β - glucosidases concentration (equation 6) ([B 0]=0.32 g/L ( ), 0.61 g/L ( ) and 1.39 g/L ( )). The residuals for the experimental data and equation 10 are shown in Figure 56. - 0.2 - 0.1 0 0.1 0.2 0.3 0.4 0 0.5 1 1.5 Normalized residues Predited protein concentration (g/L) A p p e n d i x A : S t a t i s t i c a l a n a l y s i s 217 Figure 56. Residuals for the predicted P - β - glucosidases concentration (equation 10) ([B 0]=0.32 g/L ( ), 0.61 g/L ( ) and 1.39 g/L ( )). The residuals for the experimental data and equ ation 10 are shown in Figure 57 . Figure 57. Residuals for the predicted P - β - glucosidases concentration (equation 10) ([ B0]=0.32 g/L ( ), 0.61 g/L ( ) and 1.39 g/L ( )). - 0.3 - 0.2 - 0.1 0 0.1 0.2 0.0 0.5 1.0 1.5 Normalized residues Predicted protein concentration (g/L) - 0.2 - 0.1 0 0.1 0.2 0.0 0.5 1.0 1.5 Normalized residues Predicted protein concentration (g/L) 218 A.2 Cellulose hydrolysis model residuals analysis Using the normalized residuals defined in equation 52, the normalized residuals obtained for equation 38 at 5 and 10 wt % solids concentrations are shown in Figure 58. Figure 58. Residuals for predicted glucose concentration (equation 38) Normalized residuals for equation 38 with the addition of effectiveness factor are shown in Figure 59, for 10 wt % solids co ncentrations. - 0.8 - 0.6 - 0.4 - 0.2 0 0.2 0.4 0 5 10 15 20 25 30 Normalized residuals Predicted glucose concentration (g/L) M20- 5 M40- 5 S20- 5 S40- 5 A - 1.6 - 1.2 - 0.8 - 0.4 0 0 10 20 30 40 50 60 Normalized residuals Predicted glucose concentration (g/L) M20- 10 M40- 10 S20- 10 S40- 10 B 219 Figure 59. Residuals for predicted glucose concentration (equation 38 with the addition of the effectiveness factor) Normalized residuals for equation 38 with the kinetic parameters calculated specifically at 10 wt % solids concentration is shown in Figure 60, for 10 wt % solids concentrations. Figure 60. Residuals for predicted glucose concentr ation (equation 38 with parameters obtained at 10 wt % solids concentration) - 1.2 - 0.8 - 0.4 0 0.4 0 10 20 30 40 50 Normalized residuals Predicted glucose concentration (g/L) M20- 10 M40- 10 S20- 10 S40- 10 - 0.6 - 0.4 - 0.2 0 0.2 0.4 0 10 20 30 40 50 Normalized residuals Predicted glucose concentration (g/L) M20- 10 M40- 10 S20- 10 S40- 10 220 A.3 Xylan hydrolysis model residuals analysis Using the normalized residuals definition in equation 52, ar e shown in Figure 61, for the normalized residuals obtained for equation 46 at 5 and 10 wt % solids concentrations. Figure 61. Residuals for predicted xylose concentration (equat ion 46 with parameters obtained at 5 and 10 wt % solids concentration) - 0.4 - 0.2 0 0.2 0.4 0.6 0.8 0 5 10 15 20 Normalized residuals Predicted xylose concentration (g/L) M20- 5 M40- 5 M20- 10 M40- 10 S20- 5 S40- 5 S20- 10 S40- 10 221 A.4 Accessible li quid in substrate The protein concentration in the stock and biomass - stock solutions with different initial protein concentrations over time at 20°C and 150 rpm are shown in Table 41 . Table 41. Accessible liquid in substrate Time (min) Stock solution Biomass- stock solution Protein concentration (g/L) 0 0.34 1.74 2.62 0.20 1.09 1.58 15 0.33 1.67 2.63 0.22 0.99 1.54 30 0.35 1.71 2.71 0.22 0.97 1.53 60 0.34 1.71 2.65 0.21 1.02 1.55 Average 0.34 1.71 2.65 0.21 1.02 1.54 Protein mass (g) 1.70 8.53 13.27 1.92 9.21 14.03 Based on this information and knowing that 5 mL of stock was added to the biomass - stock solutions, the mass of protein in each flask was calculated. Knowing the mass of protein in each flask, the volume required in the biomass - stock to obtain the measured protein concentrations was determined ( 8.33 mL), from where 5 mL comes from the stock. Therefore, 3.33 mL of the 4.07 mL of water in the substrate were accessible for the proteins, representing 82% of the liquid in the substrate. 222 B.1 Mass balances for lignocellulosic ethanol production Mass balances were built from experimental data collected in the pretreatment and enzymatic hydrolysis of wheat straw for all the scenarios studied. The process diagram followed is shown in Figure 63. Figure 62. Flowsheet of the pretreatment and enzymatic hydrolysis mass balance. The composition and amount of biomass before and after pretreatment was determined following the Chemical analysis described in Chapter 2. A component referred to as “O ther” was introduced to close the mass balance. The comp onent “Other ” accounts for compounds , such as proteins, uronic acids and some extractives that were not characterized . Monomeric sugars contained in pretreated liquor were not measured. Nonetheless, they were introduced to the mass balances by considering that all removed polysaccharides during pretreatment were solubilised as oligomers. The amount of non- hydrolyzed polysaccharides after enzymatic hydrolysis was calculated from the initial concentration of polysaccharides and A p p e n d i x B : E c o n o m i c a n a l y s i s a n d s i m u l a t i o n o f e t h a n o l p r o d u c t i o n p r o c e s s 1 2 4 6 Pretreatment 8 10 11 12 13 14 Enzymatic hydrolysis 15 16 17 20 24 Fermentation Biomass Water Cellulases β-glucosidase Water NaOH Separation 223 the produced monomeric sugar s. In this calculation a polyssacharides to sugars factor of 0.9 and 0.88 was used for hexo - and pentosugars, respectively . The pretreated liquor removed from pretreated substrate in the separation unit ( Figure 63) w as calculated based on the pretreated biomass moisture content after separation . Lignin, ash, extractives, other, cellulases and β - glucosidase can be found i n the liquid (L) and solid (S) phase. Material balances for all scenarios considering 76 h of hydrolysis reaction are presented in Table 7 are detailed in Table 42 to Table 49. 224 Table 42. Material balance for scenario M20 - 5. P has e Com pone nt (g) 1 2 4 6 8 10 11 12 13 14 15 16 17 20 24 Liquid H2O 476.07 1.47 477.54 1.20 478.74 478.74 0.17 478.57 421.01 57.56 1.76 0.83 2.59 184.45 243.71 Glucose 0.00 0 0 0 5.16 Xylose 0.00 0 0 0 2.79 Arabinose 0.00 0 0 0 0.29 Galactose 0.00 0 0 0 0.02 Mannose 0.00 0 0 0 0.04 Lignin 2.00 2.00 1.76 0.24 0.24 Glucose olig. 0.74 0.74 0.65 0.09 0.09 Xylose olig. 0.98 0.98 0.86 0.12 0.12 Arabinose olig. 0.24 0.24 0.21 0.03 0.03 Galactose olig. 0.10 0.10 0.09 0.01 0.01 Mannose olig. 0 0 0 0 0 Extractives 1.47 1.47 1.29 0.18 0.18 Ash 0.43 0.43 0.43 0 0 Other 1.21 1.21 1.06 0.15 0.15 NaOH 1.20 1.20 1.20 1.20 1.06 0.14 0.14 Cellulase 0.26 0.26 0 B- glucosidase 0.09 0.09 0.03 Solid Glucan 7.17 7.17 7.17 6.44 6.44 6.44 1.79 Xylan 4.02 4.02 4.02 3.05 3.05 3.05 0.59 Arabinan 0.65 0.65 0.65 0.41 0.41 0.41 0.16 Galactan 0.19 0.19 0.19 0.09 0.09 0.09 0.07 Mannan 0.16 0.16 0.16 0.15 0.15 0.15 0.12 Lignin 3.17 3.17 3.17 1.17 1.17 1.17 1.17 Extractives 1.47 1.47 1.47 0.00 0.00 0.00 0 Ash 0.80 0.80 0.80 0.38 0.38 0.38 0.38 Other 2.43 2.43 2.43 1.22 1.22 1.22 1.22 B- glucosidase 0.06 Cellulase 0.26 225 Table 43. Mass balance for scenario M4 0- 5. P has e Com pone nt (g) 1 2 4 6 8 10 11 12 13 14 15 16 17 20 24 Liquid H2O 476.07 1.47 477.54 1.20 478.74 478.74 0.17 478.57 421.01 57.56 3.51 1.67 5.18 181.51 243.24 Glucose 0 0 0 0 6.07 Xylose 0 0 0 0 2.97 Arabinose 0 0 0 0 0.29 Galactose 0 0 0 0 0.02 Mannose 0 0 0 0 0.04 Lignin 2.00 2.00 1.76 0.24 0.24 Glucose olig. 0.74 0.74 0.65 0.09 0.09 Xylose olig. 0.98 0.98 0.86 0.12 0.12 Arabinose olig. 0.24 0.24 0.21 0.03 0.03 Galactose olig. 0.10 0.10 0.09 0.01 0.01 Mannose olig. 0 0 0 0 0 Extractives 1.47 1.47 1.29 0.18 0.18 Ash 0.43 0.43 0.43 0 0 Other 1.21 1.21 1.06 0.15 0.15 NaOH 1.20 1.20 1.20 1.20 1.06 0.14 0.14 Cellulase 0.52 0.52 0 B- glucosidase 0.19 0.19 0.08 Solid Glucan 7.17 7.17 7.17 6.44 6.44 6.44 0.97 Xylan 4.02 4.02 4.02 3.05 3.05 3.05 0.43 Arabinan 0.65 0.65 0.65 0.41 0.41 0.41 0.16 Galactan 0.19 0.19 0.19 0.09 0.09 0.09 0.07 Mannan 0.16 0.16 0.16 0.15 0.15 0.15 0.12 Lignin 3.17 3.17 3.17 1.17 1.17 1.17 1.17 Extractives 1.47 1.47 1.47 0 0 0 0 Ash 0.80 0.80 0.80 0.38 0.38 0.38 0.38 Other 2.43 2.43 2.43 1.22 1.22 1.22 1.22 B- glucosidase 0.11 Cellulase 0.52 226 Table 44. Mass balance for scenario S 20- 5. P has e Com pone nt (g) 1 2 4 6 8 10 11 12 13 14 15 16 17 20 24 Liquid H2O 474.47 1.47 475.94 2.00 477.94 477.94 0.17 477.77 434.11 43.66 1.45 0.69 2.14 136.63 181.67 Glucose 0 0 0 0 4.77 Xylose 0 0 0 0 2.17 Arabinose 0 0 0 0 0.15 Galactose 0 0 0 0 0.01 Mannose 0 0 0 0 0.00 Lignin 2.71 2.71 2.47 0.25 0.25 Glucose olig. 1.85 1.85 1.68 0.17 0.17 Xylose olig. 1.69 1.69 1.54 0.15 0.15 Arabinose olig. 0.43 0.43 0.39 0.04 0.04 Galactose olig. 0.15 0.15 0.14 0.01 0.01 Mannose olig. 0.16 0.16 0.14 0.01 0.01 Extractives 1.47 1.47 1.34 0.13 0.13 Ash 0.37 0.37 0.37 0 0 Other 1.60 1.60 1.45 0.15 0.15 NaOH 2.00 2.00 2.00 2.00 1.82 0.18 0.18 Cellulase 0.22 0.22 0 B- glucosidase 0.08 0.08 0.03 Solid Glucan 7.17 7.17 7.17 5.32 5.32 5.32 1.03 Xylan 4.02 4.02 4.02 2.33 2.33 2.33 0.42 Arabinan 0.65 0.65 0.65 0.21 0.21 0.21 0.08 Galactan 0.19 0.19 0.19 0.03 0.03 0.03 0.03 Mannan 0.16 0.16 0.16 0 0 0 0 Lignin 3.17 3.17 3.17 0.46 0.46 0.46 0.46 Extractives 1.47 1.47 1.47 0 0 0 0 Ash 0.80 0.80 0.80 0.44 0.44 0.44 0.44 Other 2.43 2.43 2.43 0.83 0.83 0.83 0.83 B- glucosidase 0.05 Cellulase 0.22 227 Table 45. Mass balance for scenario S4 0- 5. P has e Com pone nt (g) 1 2 4 6 8 10 11 12 13 14 15 16 17 20 24 Liquid H2O 474.47 1.47 47 5.94 2.00 477.94 477.94 0.17 477.77 434.11 43.66 2.90 1.38 4.28 134.19 181.33 Glucose 0 0 0 0 5.14 Xylose 0 0 0 0 2.29 Arabinose 0 0 0 0 0.15 Galactose 0 0 0 0 0.01 Mannose 0 0 0 0 0 Lignin 2.71 2.71 2.47 0.25 0.25 Glucose olig. 1.85 1.85 1.68 0.17 0.17 Xylose olig. 1.69 1.69 1.54 0.15 0.15 Arabinose olig. 0.43 0.43 0.39 0.04 0.04 Galactose olig. 0.15 0.15 0.14 0.01 0.01 Mannose olig. 0.16 0.16 0.14 0.01 0.01 Extractives 1.47 1.47 1.34 0.13 0.13 Ash 0.37 0.37 0.37 0 0 Other 1.60 1.60 1.45 0.15 0.15 NaOH 2.00 2.00 2.00 2.00 1.82 0.18 0.18 Cellulase 0.43 0.43 0 B- glucosidase 0.16 0.16 0.07 Solid Glucan 7.17 7.17 7.17 5.32 5.32 5.32 0.69 Xylan 4.02 4.02 4.02 2.33 2.33 2.33 0.31 Arabinan 0.65 0.65 0.65 0.21 0.21 0.21 0.08 Galactan 0.19 0.19 0.19 0.03 0.03 0.03 0.03 Mannan 0.16 0.16 0.16 0 0 0 0 Lignin 3.17 3.17 3.17 0.46 0.46 0.46 0.46 Extractives 1.47 1.47 1.47 0 0 0 Ash 0.80 0.80 0.80 0.44 0.44 0.44 0.44Other 2.43 2.43 2.43 0.83 0.83 0.83 0.83 B- glucosidase 0.09 Cellulase 0.43 228 Table 46. Mass balance for scenario M20 - 10. P has e Com pone nt (g) 1 2 4 6 8 10 11 12 13 14 15 16 17 20 24 Liquid H2O 476.07 1.47 477.54 1.20 478.74 478.74 0.17 478.57 421.01 57.5 6 1.76 0.83 2.59 55.53 114.95 Glucose 0 0 0 0 4.41 Xylose 0 0 0 0 2.05 Arabinose 0 0 0 0 0.29 Galactose 0 0 0 0 0.01 Mannose 0 0 0 0 0.02 Lignin 2.00 2.00 1.76 0.24 0.24 Glucose olig. 0.74 0.74 0.65 0.09 0.09 Xylose olig. 0.98 0.98 0.86 0.12 0.12 Arabinose olig. 0.24 0.24 0.21 0.03 0.03 Galactose olig. 0.10 0.10 0.09 0.01 0.01 Mannose olig. 0 0 0 0 0 Extractives 1.47 1.47 1.29 0.18 0.18 Ash 0.43 0.43 0.43 0 0 Other 1.21 1.21 1.06 0.15 0.15 NaOH 1.20 1.20 1.20 1.20 1.06 0.14 0.14 Cellulase 0.26 0.26 0.00 B- glucosidase 0.09 0.09 0.04 Solid Glucan 7.17 7.17 7.17 6.44 6.44 6.44 2.47 Xylan 4.02 4.02 4.02 3.05 3.05 3.05 1.24 Arabinan 0.65 0.65 0.65 0.41 0.41 0.41 0.15 Galactan 0.19 0.19 0.19 0.09 0.09 0.09 0.08 Mannan 0.16 0.16 0.16 0.15 0.15 0.15 0.14 Lignin 3.17 3.17 3.17 1.17 1.17 1.17 1.17 Extractives 1.47 1.47 1.47 0 0 0 0 Ash 0.80 0.80 0.80 0.38 0.38 0.38 0.38 Other 2.43 2.43 2.43 1.22 1.22 1.22 1.22 B- glucosidase 0.06 Cellulase 0.26 229 Table 47. Mass balance for scenario M40 - 10. P has e Com pone nt (g) 1 2 4 6 8 10 11 12 13 14 15 16 17 20 24 Liquid H2O 476.07 1.47 477.54 1.20 478.74 478.74 0.17 478.57 421.01 57.56 3.51 1.67 5.18 52.58 114.49 Glucose 0 0 0 0 5.26 Xylose 0 0 0 0 2.25 Arabinose 0 0 0 0 0.29 Galactose 0 0 0 0 0.01 Mannose 0 0 0 0 0.02 Lignin 2.00 2.00 1.76 0.24 0.24 Glucose olig. 0.74 0.74 0.65 0.09 0.09 X ylose olig. 0.98 0.98 0.86 0.12 0.12 Arabinose olig. 0.24 0.24 0.21 0.03 0.03 Galactose olig. 0.10 0.10 0.09 0.01 0.01 Mannose olig. 0 0 0 0 0 Extractives 1.47 1.47 1.29 0.18 0.18 Ash 0.43 0.43 0.43 0 0.00 Other 1.21 1.21 1.06 0.15 0.15 NaOH 1.20 1.20 1.20 1.20 1.06 0.14 0.14 Cellulase 0.52 0.52 0 B- glucosidase 0.19 0.19 0.10 Solid Glucan 7.17 7.17 7.17 6.44 6.44 6.44 1.70 Xylan 4.02 4.02 4.02 3.05 3.05 3.05 1.06 Arabinan 0.65 0.65 0.65 0.41 0.41 0.41 0.15 Galactan 0.19 0.19 0.19 0.09 0.09 0.09 0.08 Mannan 0.16 0.16 0.16 0.15 0.15 0.15 0.14 Lignin 3.17 3.17 3.17 1.17 1.17 1.17 1.17 Extractives 1.47 1.47 1.47 0 0 0 0 Ash 0.80 0.80 0.80 0.38 0.38 0.38 0.38 Other 2.43 2.43 2.43 1.22 1.22 1.22 1.22 B- glucosidase 0.09 Cellulase 0.52 230 Table 48. Mass balance for scenario S 20- 10. P has e Com pone nt (g) 1 2 4 6 8 10 11 12 13 14 15 16 17 20 24 Liquid H2O 474.47 1.47 475.94 2.00 477.94 477.94 0.17 477.77 434.11 43.66 1.45 0.69 2.14 40.46 85.64 Glucose 0 0 0 0 4.05 Xylose 0 0 0 0 1.64 Arabinose 0 0 0 0 0.15 Galactose 0 0 0 0 0 Mannose 0 0 0 0 0 Lignin 2.71 2.71 2.47 0.25 0.25 Glucose olig. 1.85 1.85 1.68 0.17 0.17 Xylose olig. 1.69 1.69 1.54 0.15 0.15 Arabinose olig. 0.43 0.43 0.39 0.04 0.04 Galactose olig. 0.15 0.15 0.14 0.01 0.01 Mannose olig. 0.16 0.16 0.14 0.01 0.01 Extractives 1.47 1.47 1.34 0.13 0.13 Ash 0.37 0.37 0.37 0 0 Other 1.6 0 1.60 1.45 0.15 0.15 NaOH 2.00 2.00 2.00 2.00 1.82 0.18 0.18 Cellulase 0.22 0.22 0 B- glucosidase 0.08 0.08 0.03 Solid Glucan 7.17 7.17 7.17 5.32 5.32 5.32 1.67 Xylan 4.02 4.02 4.02 2.33 2.33 2.33 0.89 Arabinan 0.65 0.65 0.65 0.21 0.21 0.21 0.08 Galactan 0.19 0.19 0.19 0.03 0.03 0.03 0.03 Mannan 0.16 0.16 0.16 0 0 0 0 Lignin 3.17 3.17 3.17 0.46 0.46 0.46 0.46 Extractives 1.47 1.47 1.47 0 0 0 0 Ash 0.80 0.80 0.80 0.44 0.44 0.44 0.44 Other 2.43 2.43 2.43 0.83 0.83 0.83 0.83 B- glucosidase 0.04 Cellulase 0.22 231 Table 49. Mass balance for scenario S40 - 10. P has e Com pone nt (g) 1 2 4 6 8 10 11 12 13 14 15 16 17 20 24 Liquid H2O 474.47 1.47 475.94 2.00 477.94 477.94 0.17 477.77 434.11 43.66 2.90 1.38 4.28 38.02 85.29 Glucose 0 0 0 0 4.48 Xylose 0 0 0 0 1.75 Arabinose 0 0 0 0 0.15 Galactose 0 0 0 0 0.00 Mannose 0 0 0 0 0.00 Lignin 2.71 2.71 2.47 0.25 0.25 Glucose olig. 1.85 1.85 1.68 0.17 0.17 Xylose olig. 1.69 1.69 1.54 0.15 0.15 Arabinose olig. 0.43 0.43 0.39 0.04 0.04 Galactose olig. 0.15 0.15 0.14 0.01 0.01 Mannose olig. 0.16 0.16 0.14 0.01 0.01 Extractives 1.47 1.47 1.34 0.13 0.13 Ash 0.37 0.37 0.37 0.00 0.00 Other 1.60 1.60 1.45 0.15 0.15 NaOH 2.00 2.00 2.00 2.00 1.82 0.18 0.18 Cellulase 0.43 0.43 0.00 B- glucosidase 0.16 0.16 0.08 Solid Glucan 7.17 7.17 7.17 5.32 5.32 5.32 1.29 Xylan 4.02 4.02 4.02 2.33 2.33 2.33 0.79 Arabinan 0.65 0.65 0.6 5 0.21 0.21 0.21 0.08 Galactan 0.19 0.19 0.19 0.03 0.03 0.03 0.03 Mannan 0.16 0.16 0.16 0 0 0 0 Lignin 3.17 3.17 3.17 0.46 0.46 0.46 0.46 Extractives 1.47 1.47 1.47 0 0 0 Ash 0.80 0.80 0.80 0.44 0.44 0.44 0.44Other 2.43 2.43 2.43 0.83 0.83 0.83 0.83 B- glucosidase 0.07 Cellulase 0.43 232 B.2 Ef fect of washing pretreated substrates on enzymatic hydrolysis The results from the hydrolysis of substrate M and S poorly (0.4 L of water per batch ) and extensively (with 0.4 L of water per batch ) washed are presented in Figure 64. 0 20 40 60 80 100 0 20 40 60 80 Cellulose conversion (%) Time (h) M20 - 5 0 20 40 60 80 100 0 20 40 60 80 Cellulose conversion (%) Time (h) M40 - 5 0 20 40 60 80 100 0 20 40 60 80 Cellulose conversion (%) Time (h) S2 0 - 5 0 20 40 60 80 100 0 20 40 60 80 Cellulose conversion (%) Time (h) S4 0 - 5 0 20 40 60 80 100 0 20 40 60 80 Cellulose conversion (%) Time (h) M20 - 10 0 20 40 60 80 100 0 20 40 60 80 Cellulose conversion (%) Time (h) M40 - 10 233 Figure 63. Washing effect on the enzy matic hydrolysis of pretreated wheat straw. Poorly (0.4 L per batch) ( ) and extensively (4 L per batch) ( ) pretreated biomass wash . Lines are added to assist in visualizing trends. B. 3 Base case scenarios reactors details The characteristics of the reactors used in the economic analysis of production of ethanol process without enzyme recycling are shown below. Table 50. Reactor details for scenario M20 - 5. Un i t Resi d en ce ti me (h) Numb e r of equ i p men t Reacto r volu me (gal ) Vessel Diamete r (m) Vessel Heigh t (m) Delignification reactor 0.5 3 96,300 6.1 12.3 Hydrolysis reactor 48 15 1,000,000 13.4 26.8 1 661,100 11.7 23.4 Fermentor 36 12 1,000,000 13.4 26.8 1st Seed reactor 24 2 20 0.6 0.6 2nd Seed reactor 2 300 0.9 1.8 3th Seed reactor 2 3,000 1.8 4.6 4th Seed reactor 2 31,000 4.3 9.8 5th Seed reactor 2 310,000 9.4 20.4 0 20 40 60 80 100 0 20 40 60 80 Cellulose conversion (%) Time (h) S2 0 - 1 0 0 20 40 60 80 100 0 20 40 60 80 Cellulose conversion (%) Time (h) S4 0 - 1 0 234 Table 51. Reactor details for scenario M40 - 5. Un i t Resi d en ce time (h) Numb e r of equ i p men t Reacto r volu me (gal ) Vessel Diamete r (m) Vessel Heigh t (m) Delignification reactor 0.5 3 96,300 6.1 12.3 Hydrolysis reactor 30 11 1,000,000 13.4 26.8 Fermentor 36 12 1,000,000 13.4 26.8 1st Seed reactor 24 2 20 0.6 0.6 2nd Seed reactor 2 300 0.9 1.8 3th Seed reactor 2 3,000 1.8 4.6 4th Seed reactor 2 31,000 4.3 8.6 5th Seed reactor 2 300,000 9.0 18.0 Table 52. Reactor details for scenario S20 - 5. Un i t Resi d en ce time (h) Numb e r of equ i p men t Reacto r volu me (gal ) Vessel Diamete r (m) Vessel Heigh t (m) Delignification reactor 1 5 115,600 6.5 13.1 Hydrolysis reactor 48 12 1,000,000 13.4 26.8 1 647,200 11.6 23.2 Fermentor 36 9 950, 000 13.4 26.8 1 883,200 12.9 25.7 1st Seed reactor 24 2 20 0.6 0.6 2nd Seed reactor 2 200 0.9 1.8 3th Seed reactor 2 2,400 1.8 3.6 4th Seed reactor 2 24,000 8.2 16.4 5th Seed reactor 2 230,000 8.2 16.4 235 Table 53. Reactor details for scenario S40 - 5. Un i t Resi d en ce time (h) Numb e r of equ i p men t Reacto r volu me (gal ) Vessel Diamete r (m) Vessel Heigh t (m) Delignification reactor 1 5 115,600 6.5 13.1 Hydrolysis reactor 24 7 1,000, 000 13.4 26.8 1 209,400 8.0 15.9 Fermentor 36 9 1,000,000 13.4 26.8 1 883,200 12.9 25.7 1st Seed reactor 24 2 20 0.6 0.6 2nd Seed reactor 2 200 0.8 0.2 3th Seed reactor 2 2,400 1.8 3.6 4th Seed reactor 2 24,000 3.9 7.7 5th Seed reactor 2 230,000 8.2 16.4 Table 54. Reactor details for scenario M20 - 10. Un i t Resi d en ce time (h) Numb e r of equ i p men t Reacto r volu me (gal ) Vessel Diamete r (m) Vessel Heigh t (m) Delignification reactor 0.5 3 96,300 6.1 12.3 Hydrolysis reactor 48 8 1,000,000 13.4 26.8 2 250,000 8.4 16.9 Fermentor 36 7 1,000,000 13.4 26.8 1 579,000 11.2 22.3 1st Seed reactor 24 2 20 0.6 0.6 2nd Seed reactor 2 200 0.9 1.8 3th Seed reactor 2 2,000 1.7 3.4 4th Seed reactor 2 33,000 4.3 8.6 5th Seed reactor 2 310,000 9.1 18.1 236 Table 55. Reactor details for scenario M40 - 10. Un i t Resi d en ce time (h) Numb e r of equ i p men t Reacto r volu me (gal ) Vessel Diamete r (m) Vessel Heigh t (m) Delignification Reactor 0.5 3 96,300 6.1 12.3 Hydrolysis reactor 48 8 1,000,000 13.4 26.8 2 250,000 8.4 16.9 Fermentor 36 7 1,000,000 13.4 26.8 1 579,000 11.2 22.3 1st Seed reactor 24 2 20 0.6 0.6 2nd Seed reactor 2 200 0.9 1.8 3th Seed reactor 2 2,000 1.7 3.4 4th Seed reactor 2 33,000 4.3 8.6 5th Seed reactor 2 310,000 9.1 18.1 Table 56. Reactor detail s for scenario S20 - 10. Un i t Resi d en ce time (h) Numb e r of equ i p men t Reacto r volu me (gal ) Vessel Diamete r (m) Vessel Heigh t (m) Delignification reactor 1 5 115,600 6.5 13.1 Hydrolysis reactor 48 6 1,000,000 13.4 26.8 1 250,000 8.4 16.9 1 352,500 9.5 18.9 Fermentor 36 6 1,000,000 13.4 26.8 1 320,400 9.2 18.3 1st Seed reactor 24 2 20 0.6 0.6 2nd Seed reactor 2 200 0.9 1.8 3th Seed reactor 2 2,500 1.8 3.6 4th Seed reactor 2 25,000 3.9 7.8 5th Seed reactor 2 230,000 8.2 16.4 237 Table 57. Reactor details for scenario S40 - 10. Un i t Resi d en ce time (h) Numb e r of equ i p men t Reacto r volu me (gal ) Vessel Diamete r (m) Vessel Heigh t (m) Delignification reactor 1 5 115,600 6.5 13.1 Hydrolysis reactor 48 6 1,000,000 13.4 26.8 1 250,000 8.4 16.9 1 611,800 11.4 22.8 Fermentor 36 6 1,000,000 13.4 26.8 1 320,400 9.2 18.3 1st Seed reactor 24 2 20 0.6 0.6 2nd Seed reactor 2 200 0.8 0.2 3th Seed reactor 2 2,500 1.8 3.6 4th Seed reactor 2 25,000 3.9 7.8 5th Seed reacto r 2 230,000 8.2 16.4 B.4 Molecular sieve cost Molecular sieve vessels were designed following the methodology given by the “Gas processors suppliers association” hand book (Gas processors suppliers association, 2004) . The first step is to calculate the superficial velocity with a modified Ergun equation: ∆𝑃𝐿𝑒= 𝐻 ∗ 𝜇 ∗ 𝑉𝑒 + 𝐾 ∗ 𝜌 ∗ 𝑉𝑒2 5 3 where ∆P is the total pressure drop ( kg/m 2), Ve is the superficial velocity ( m/ h), Le is the bed length (m), μ is the viscosity (kg/m h ), ρ is the density (kg/m 3 ) and H and K are constants which are shown in Table 58. 238 Table 58. Constants for the molecular sieve desiccant (Gas processors suppliers association, 2004). P art ic le Type H (h 2 / m 3 ) K (h 2 / m 2 ) 1/8" bead (4x8 mesh) 5.49E-04 2.66 E-07 1/8" extrudate 7.08E-04 3.71 E-07 A 1/8” bead (4x8 mesh) was used in the present study. From the simulation result s, the viscosity and density of the inlet stream 59 ( 91.9 wt% ethanol, 7.4 wt% water and 0.73 wt % CO2) are 0.043 kg/m h and 1.291 kg/m3 , respect ively. The maximum allowable ∆P/Le reported is 761.20 kg/m 3 , therefore, solving e quation 53 for Ve considering the maximum allowable ∆P/L, a maximum superficial velocity (Vmax) of 3,161.08 m/h was determined. Once Vmax is estimated, the bed minimum diameter (Dmin) (m) can be calculated. 𝐷𝑚𝑖𝑛 = �4𝑞𝜋𝑉𝑚𝑎𝑥�0.5 54 𝑞 =𝑚𝑓60𝜌 5 5 where q is the inlet volumetric flow rate (m 3 /h ) and mf the mass flow rate (kg /h ). The mass flow e ntering the molecular sieve in M20- 5 is 18, 308.78 kg/h , resulting in a Dmin of 2.39 m. The nearest standard diameter to Dmin is then selected, in this case it was 2.8 m (Gas processors suppliers association, 2004) . With this information the adjusted velocity (Vadj) (m/h ) was calculated: 𝑉𝑎𝑑𝑗 = 𝑉𝑚𝑎𝑥 �𝐷𝑚𝑖𝑛𝐷𝑠𝑒𝑙𝑒𝑐𝑡𝑒𝑑�2 5 6 The next step is to calculate the mass of desiccant (S S) (kg ) required in the separation process where typicall y 8 to 12 h adsorption periods are used (Gebreyohannes, 2007) . During the adsorption period, water is adsorbed on the sieve, producing a rich ethanol stream as product. In order to regenerate the molecular sieve, in the regeneration step, water is desorbed from the sieve passing hot ethanol t hrough the bed at low pressures. Molecular sieves have the capacity to hold approximately 13 pounds of water per 100 pounds of sieve. New sieve s have an effective capacity near 20%, however, a 13% effective capacity was used, as it represents the approximate capacity of a 3 to 5 year old sieve (Gas processors suppliers association, 239 2004). The mass of desiccant required in the process is calculated by dividing the amount of water to be removed during the cycle by the effective capacity. 𝑆𝑠 =𝑊𝑟0.13𝐶𝑠𝑠𝐶𝑇 5 7 where Wr is the amount of water to be remo ved in the adsorption period (kg ). For scenario S20- 5 the amount of water to be adsorbed in a 12 h period was 15,417.50 kg . Since the feed stream is water saturated, the relative humidity is 100%, so C SS is 1.0 (Gas processors suppliers association, 2004) . CT values of 0.7 and 1 were used for temperatures below 21°C and above 87°C, respectively. The mass of desiccant (S S) calculated was 118,596.26 kg per vessel. In the next step , the bed height was determined considering that a molecular sieve column has a saturation zone and a mass transfer zone. The length of the saturation zone (L s) (m) can be calculated using equation 58. 𝐿𝑠 =4𝑆𝑆𝜋 𝐷𝑎𝑑𝑗2𝜌𝐵 58 Molecular sieves have a bulk density of 672.78 to 736.85 kg / m3 for spherical particles (Gas processors suppliers association, 2004) . In consequence, a bulk density (ρB) of 672.78 kg/m3 was used. The length of the saturati on zone obtained was 39.30 m . The length of the mass transfer zone ( LMTZ) can be estimated as follows: 𝐿𝑀𝑇𝑍 = 𝑍 �𝑉𝑎𝑑𝑗12.18�0.3 5 9 For the for 1/8 inch sieve selected, the value of Z is 0.52 m (Gas processors suppliers association, 2004), the resulting mass transfer zone was 0.84 m . The total bed height was the sum of these two values, and it should be no less than the vessel diameter or 1.8 m . The total bed height (LT) obtained was 40.1 m . Next, the total bed pressure drop is checked. The ∆P/L for the selected diameter, D selected, is adjusted using the following approximation: �Δ𝑃D𝐿𝑇�𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑≅ (761.20 kg/m^3) �𝑉𝑎𝑑𝑗𝑉𝑚𝑎𝑥�2 60 where the total bed height is L=LS + LMTZ. The total design pressure drop is then calculated, total design pressure values are recommended to be between the range 3,515 to 5,625 kg/m 2. The total design pressure drop ( ∆PD) is important because the operating pressure drop can 240 double the design value over three years. A high pressure drop plus the bed weight can crush the sieve. If the design pressure drop exceeds 5,625 kg/m 2, the bed diameter should be increased and the sieve amount and vessel dimensions recalculated (Gas processors suppliers association, 2004). The pressure drop obtained with the parameters calculated was 30,555 kg/m 2, which is greater than the recommended pressure and therefore , the Dselected was reselected. A final pressure drop of 4,078 kg/m 2 was obt ained with a bed diameter of 3.4 m and a bed height of 20 .7 m for scenario S20- 5. However, due to the vessel height, an arrangement of two pressure swing adsorption in parallel with 3.1 m diameter and a bed height of 12.7 m was selected . B. 5 Raw material cost B.5.1 Biomass Collection, processing , storage, and transportation costs of biomass between the field and pretreatment reactor was assumed to be $34.13/ ton DM (Leistritz et al., 2006) . Based on the feedstock supply models provided by the U.S. Department of Energy (U.S. Department of Energy, 2001), the grower payment was considered to be $23.50/ ton DM, giving a f inal raw biomass cost of $57.63/t on DM. Raw wheat straw had 7.35 % moisture content; therefore, the raw material price used as a stream price in the simulation was set to $53 .39/ ton wet biomass. B.5 .2 Oxygen The oxygen flow rate measured during the oxygen delignification experiment was 1 L O2/min (100.86 kg O 2 / ADT). The experimental oxygen flow rate was high enough to assure saturation and avoid any oxygen limitations. Therefore, oxygen flow rate used in the simulation and economic analysis was the typical oxygen flow rate used in the industry, 15-24 O2 kg/ ADT (Linde group, 2013) . This flow rate was used for both delignification 241 conditions as biomass loading was constant in both conditions. The upper value of the oxygen flow rate was selected for this study: 2204 ton DMday=2222.2 ADTday∗24 O2 kgADT=53333.3 O2 kgday=58.79 O2 tonday The usual commercial arrangement in the industrial gas business is to supply product over the fence. A company provides the necessary gas flow to the process for a fixed monthly facility fee. To exemplify the oxygen cost estimation methodology used in this project , a reference capital cost of 26.5 million dollars per month plant which requires a 1000 ton O2/day was assumed (Wilcox, 2004) . The cost was updated from October 2004 to October 2011 using the Chemical Engineering plant cost index (Chemical engineering plant cost index, 2013) . 𝐶𝑜𝑠𝑡 = $26,500,000 (2004) ∗585.7 (2011)444.2 (2004)= $34,941,580 (2011) An economy of scale of 0.6 was assumed to scale the plant from 1000 ton O2/day to 58.79 ton O2/day. Monthly capital cost = $34,941,580 ∗ �58.79 O2 Ton/day1000 O2 Ton/day�0.6= $6,381,462 A facility fee of 2.75% of the capital per month, which includes a capital charge, all maintenance and repair costs, back - up systems and back - up supply , was used to estimate the oxygen cost. Power costs, which are proportional to production rate, are passed through to the customer. Annual capital fee = $6,381,462 ∗ 0.0275 ∗ 12 months = $2,105,882 A reference of 15 kWh/28.32 m 3 of oxygen was used (Wilcox, 2004) . The oxygen density 1.33 kg/m 3 was calculated at normal temperature and pressure. 242 Electricity anual cost = 53333.3 O2 kgday∗1 O2m31.33 O2kg∗15 kWh28.32 O2 m3∗0.04 $kWh∗365 days1 yr= $309,897 This cost is the final electricity bill to pa y for the oxygen flow rate required on the pretreatment step. The total annual oxygen cost is then: Total anual cost = $2,105,882 + $309,897 = $2,415,780 B.5 .3 NaOH NaOH used for the economic evaluation was set to $420/ ton of 50% diaphragm grade solution, whi ch is the lower price in the range reported in literature given in Table 22. However, different prices were used for the sensitivity analysis in order to study the effect of different caustic soda prices in the overall process. B.5 .4 Enzymes The cost of the enzymes and, in consequence , its relevance to the lignocellulosic ethanol process is still debate d (Klein - Marcuschamer et al., 2012). Due to the lack of public information about enzyme cost, the researchers have had to make estimates. Most estimated enzyme prices have been reported as dollars per gallon of ethanol produced. This make it difficult to compare enzyme costs since ethanol production depends on many factors besides enzyme expenses , like feedstock price, pretreatment , proce ss configuration, operating conditions or ethanol yield (Klein - Marcuschamer et al., 2012). Klein - Marcushamer et al. (2012) proposed an enzyme cost of $1/gal ethanol as a value that smoothes out the effects of assumptions at optimistic or conservative extremes. In the pr esent work, it was assumed that Celluclast 1.5L is composed mainly of cellulases, and, Novozyme 188 has a β- glucosidase protein composition of 7.7%. Therefore, this percent of protein 243 content in Novozyme 188 was considered as enzyme and a price was assigned to this fraction . Using the enzyme concentration in the broth fed to hydrolysis reactor and an enzyme cost contribution of $1/gal ethanol (Klein - Marcuschamer et al., 2012), the enzyme cost per enzyme mass was estimated for eight scenarios, as shown in Table 59. These results show that enzyme cost s per enzyme mass when calculated from the enzyme cost contribution per gallon of ethanol varies widely with operating conditions. The wide range of enzyme costs obtained emphasizes the inconsistency of enzyme costs assumed in the literature. Table 59. Enzyme cost for different scenarios , considering an enzyme cost contribution of $1/gal ethanol Sc e nari o M20- 5 M40- 5 S20 - 5 S40 - 5 M20- 10 M40- 10 S20 - 10 S40 - 10 $/ kg enzyme 4.4 2.5 4.7 2.4 3.6 2.1 0.4 2.1 Humbird et al. (2011) estimated a enzyme cost contribution of $0.34/gal of ethanol ($4.83/kg enzyme) , however, it was mentioned in the same report that the major enzyme manufacturers Genencor and Novozymes announced a new commercial - grade cellulase enzyme with an enzyme cost contribution of approximately $0.50/gal ethanol. The ethanol production achieved by Humbird et al. (2011) is 21,675.34 kg ethanol/h, where a enzyme broth flow rate of 13,838.14 kg /h is fed to hydrolysis process . The reported enzyme broth flow rate referred to the broth produced in an on- site enzymes production plant . This enzyme broth contained 520.86 kg/h of enzyme and other components. The enzyme cost ($/kg enzyme) corresponding to a n enzyme to ethanol cost contribution of $0.50/gal is : $0.50 USDgal ethanol∗1 gal6.46 lb∗1b0.45kg=$0.17 kg ethanol $0.17 USDkg ethanol∗21675.34 kg ethanol/h520.86 kg enzyme/h =$7.10 kg enzyme The calculated enzyme cost ( $7.1/kg enzyme ) is in close agreement with the commercial textile grade cellulase (Spezyme C P, $5 per kg ), which is based on Genencor International 244 estimates (Leistritz et al., 2006) . From the $7.1/kg enzyme cost and the highest enzyme cost presented in Table 59 (scenario S20- 5) , an enzyme cost range of $5 to $9/kg enzyme was used in this work to evaluate the impact of enzyme price on lignocellulosic ethanol production. The enzyme cost range used in the sensitivity analysis contains the mentioned enzyme costs ($ 5 and $7/ kg enzyme ) which were considered to be the most realistic enzyme prices . Enzy me cost of $7 /kg enzyme was selected as base case scenario. B.5 .5 DAP and CSL DAP and CSL prices reported by Humbird et al. (2011) were used in the present study. The costs were updated from 2007 to 2011 using the chemical engineering plant cost index (Chemical engineering plant cost index, 2013) : $0.025(2007)𝐶𝑆𝐿 𝑙𝑏∗585.7 (2011)525.4 (2007)=$0.028 (2011)𝐶𝑆𝐿 𝑙𝑏 $0.439 (2007)𝐷𝐴𝑃 𝑙𝑏∗585.7 (2011)525.4 (2007)=$0.489 (2011)𝐷𝐴𝑃 𝑙𝑏 245 B.5.6 Natural g as Natural gas composition reported by Union G as (2012) (Table 60 ) was used to calculate the natural gas cost and its energy density. Table 60. Na tural gas composition Co mp on en t Typ i cal anal ysi s (mol ar fracti on ) Methane 0.952 Ethane 0.025 Propane 0.002 Isobutane 0.0003 Butane 0.0003 Isopentane 0.0001 Pentane 0.0001 Hexane 0.0001 Nitrogen 0.013 Carbon dioxide 0.007 Oxygen 0.0001 A standard cubic f oot (SCF) is equal to a cubic foot of volume at 60°F and 1 atm. A natural gas density of 0 .02 kg/SCF natural gas was calculated u sing AspenP lus and natural gas composition shown in Table 60. The final nat ural gas cost and energy density were $0.13/kg and 50.66 M Btu/kg gas natural, respectively. 246 C.1 Mass balances for lignocellulosic ethanol production with enzyme recycling Mass balances for the enzyme recycling process were built from experimental data. The process diagram used for all scenarios is shown in Figure 65. Figure 64. Flowsheet of the pretreatment and enzymatic hydrolysis with enzyme recycling. The material balances built for the lignocellulosic ethanol process with enzyme recycling followed the same assumptions considered in the construction of the process wit hout enzyme recovery presented in Appendix B. Mass balances for the process with enzyme recycling are shown in Table 61 to Table 68 . A p p e n d i x C : E c o n o m i c a n a l y s i s a n d s i m u l a t i o n o f e t h a n o l p r o d u c t i o n p r o c e s s w i t h e n z y m e r e c y c l i n g 1 2 4 6 Pretreatment 8 10 11 12 13 14 Enzymatic Hydrolysis 15 16 18 24 Fermentation Separation Adsorption Separation Separation 21 19 23 22 17 20 247 Table 61. Material balance for scenario M40 - 5- R5. P has e Co mp o ne nt (g ) 1 2 4 6 8 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Liquid H2O 476.07 1.47 477.54 1.20 478.74 478.74 0.17 478.57 421.01 57.56 3.29 1.54 4.82 295.67 32.54 206.93 243.65 5.54 238.11 263.13 Glucose 4.04 0.44 4.13 0.09 4.04 3.60 Xylose 2.12 0.23 2.17 0.05 2.12 1.89 Arabinose 0.31 0.03 0.32 0.01 0.31 0.28 Galactose 0.02 0 0.02 0 0.02 0.02 Mannose 0.04 0 0.04 0 0.04 0.04 Lignin 2.00 2.00 1.76 0.24 0.27 0.03 0.03 0 0.03 0.24 Glucose olig. 0.74 0.74 0.65 0.09 0.10 0.01 0.01 0 0.01 0.09 Xylose olig. 0.98 0.98 0.86 0.12 0.13 0.01 0.01 0 0.01 0.12 Arabinose olig. 0.24 0.24 0.21 0.03 0.03 0 0.00 0 0 0.03 Galactose olig. 0.10 0.10 0.09 0.01 0.01 0 0.00 0 0 0.01 Mannose olig. 0 0 0 0 0 0 0 0 0 0 Extractives 1.47 1.47 1.29 0.18 0.20 0.02 0.02 0 0.02 0.18 Ash 0.43 0.43 0.43 0 0 0 0 0 0 0 Other 1.21 1.21 1.06 0.15 0.16 0.02 0.02 0 0.02 0.14 NaOH 1.20 1.20 1.20 1.20 1.06 0.14 0.16 0.02 0.02 0 0.02 0.14 Cellulase 0.49 0.00 0.49 0.46 0.01 0.50 0.01 0.48 0.45 B- glucosidase 0.00 0.17 0.17 0.17 0.01 0.18 0 0.17 0.16 Solid Glucan 7.17 7.17 7.17 6.44 6.44 6.44 6.44 6.44 3.12 3.12 0 0 Xylan 4.02 4.02 4.02 3.05 3.05 3.05 3.05 3.05 1.34 1.34 0 0 Arabinan 0.65 0.65 0.65 0.41 0.41 0.41 0.41 0.41 0.16 0.16 0 0 Galactan 0.19 0.19 0.19 0.09 0.09 0.09 0.09 0.09 0.07 0.07 0 0 Mannan 0.16 0.16 0.16 0.15 0.15 0.15 0.15 0.15 0.12 0.12 0 0 Lignin 3.17 3.17 3.17 1.17 1.17 1.17 1.17 1.17 1.17 1.17 0 0 Extractives 1.47 1.47 1.47 0 0 0 0 0 0 0 0 0 Ash 0.80 0.80 0.80 0.38 0.38 0.38 0.38 0.38 0.38 0.38 0 0 Other 2.43 2.43 2.43 1.22 1.22 1.22 1.22 1.22 1.22 1.22 0 0 B- glucosidase 0 0 0 0 0 0 0 0 0.01 0.01 0 0 Cellulase 0 0 0 0 0 0 0.03 0.03 0.03 0.03 0 0 248 Table 62. Material balance for scenario M40 - 5- R24. P has e Co mp o ne nt (g ) 1 2 4 6 8 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Liquid H2O 476.07 1.47 477.54 1.20 478.74 478.74 0.17 478.57 421.01 57.56 3.08 1.57 4.64 295.32 32.54 207.14 243.43 5.66 237.77 262.79 Glucose 5.84 0.64 5.98 0.14 5.84 5.20 Xylose 2.91 0.32 2.98 0.07 2.91 2.59 Arabinose 0.31 0.03 0.32 0.01 0.31 0.28 Galactose 0.02 0 0.02 0 0.02 0.02 Mannose 0.04 0 0.04 0 0.04 0.04 Lignin 2.00 2.00 1.76 0.24 0.27 0.03 0.03 0 0.03 0.24 Glucose olig. 0.74 0.74 0.65 0.09 0.10 0.01 0.01 0 0.01 0.09 Xylose olig. 0.98 0.98 0.86 0.12 0.13 0.01 0.01 0 0.01 0.12 Arabinose olig. 0.24 0.24 0.21 0.03 0.03 0 0 0 0 0.03 Galactose olig. 0.10 0.10 0.09 0.01 0.01 0 0 0 0 0.01 Mannose olig. 0 0 0 0 0 0 0 0 0 0 Extractives 1.47 1.47 1.29 0.18 0.20 0.02 0.02 0 0.02 0.18 Ash 0.43 0.43 0.43 0.00 0 0 0 0 0 0 Other 1.21 1.21 1.06 0.15 0.16 0.02 0.02 0 0.02 0.14 NaOH 1.20 1.20 1.20 1.20 1.06 0.14 0.16 0.02 0.02 0 0.02 0.14 Cellulase 0.46 0 0.46 0.39 0.01 0.46 0.01 0.45 0.39 B- glucosidase 0 0.18 0.18 0.13 0.01 0.13 0.00 0.13 0.12 Solid Glucan 7.17 7.17 7.17 6.44 6.44 6.44 6.44 6.44 1.63 1.63 0 0 Xylan 4.02 4.02 4.02 3.05 3.05 3.05 3.05 3.05 0.71 0.71 0 0 Arabinan 0.65 0.65 0.65 0.41 0.41 0.41 0.41 0.41 0.16 0.16 0 0 Galactan 0.19 0.19 0.19 0.09 0.09 0.09 0.09 0.09 0.07 0.07 0 0 Mannan 0.16 0.16 0.16 0.15 0.15 0.15 0.15 0.15 0.12 0.12 0 0 Lignin 3.17 3.17 3.17 1.17 1.17 1.17 1.17 1.17 1.17 1.17 0 0 Extractives 1.47 1.47 1.47 0 0 0 0 0 0 0 0 0 Ash 0.80 0.80 0.80 0.38 0.38 0.38 0.38 0.38 0.38 0.38 0 0 Other 2.43 2.43 2.43 1.22 1.22 1.22 1.22 1.22 1.22 1.22 0 0 B- glucosidase 0 0 0 0 0 0 0 0 0.06 0.06 0 0 Cellulase 0 0 0 0 0 0 0.06 0.06 0.06 0.06 0 0 249 Table 63. Material balance for scenario M40 - 5- R48. P has e Co mp o ne nt (g ) 1 2 4 6 8 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Liquid H2O 476.07 1.47 477.54 1.20 478.74 478.74 0.17 478.57 421.01 57.56 2.92 1.59 4.51 294.85 32.54 207.30 243.37 6.08 237.29 262.31 Glucose 6.36 0.70 6.52 0.16 6.36 5.66 Xylose 3.13 0.35 3.21 0.08 3.13 2.79 Arabinose 0.31 0.03 0.32 0.01 0.31 0.28 Galactose 0.02 0 0.02 0 0.02 0.02 Mannose 0.04 0 0.04 0 0.04 0.04 Lignin 2.00 2.00 1.76 0.24 0.27 0.03 0.03 0 0.03 0.24 Glucose olig. 0.74 0.74 0.65 0.09 0.10 0.01 0.01 0 0.01 0.09 Xylose olig. 0.98 0.98 0.86 0.12 0.13 0.01 0.01 0 0.01 0.12 Arabinose olig. 0.24 0.24 0.21 0.03 0.03 0 0 0 0 0.03 Galactose olig. 0.10 0.10 0.09 0.01 0.01 0 0 0 0 0.01 Mannose olig. 0 0 0 0 0 0 0 0 0 0 Extractives 1.47 1.47 1.29 0.18 0.20 0.02 0.02 0 0.02 0.18 Ash 0.43 0.43 0.43 0 0 0 0 0 0 0 Other 1.21 1.21 1.06 0.15 0.16 0.02 0.02 0 0.02 0.14 NaOH 1.20 1.20 1.20 1.20 1.06 0.14 0.16 0.02 0.02 0 0.02 0.14 Cellulase 0.43 0.00 0.43 0.35 0.01 0.44 0.01 0.43 0.34 B- glucosidase 0.00 0.18 0.18 0.10 0.01 0.10 0 0.10 0.09 Solid Glucan 7.17 7.17 7.17 6.44 6.44 6.44 6.44 6.44 1.20 1.20 0 0 Xylan 4.02 4.02 4.02 3.05 3.05 3.05 3.05 3.05 0.52 0.52 0 0 Arabinan 0.65 0.65 0.65 0.41 0.41 0.41 0.41 0.41 0.16 0.16 0 0 Galactan 0.19 0.19 0.19 0.09 0.09 0.09 0.09 0.09 0.07 0.07 0 0 Mannan 0.16 0.16 0.16 0.15 0.15 0.15 0.15 0.15 0.12 0.12 0 0 Lignin 3.17 3.17 3.17 1.17 1.17 1.17 1.17 1.17 1.17 1.17 0 0 Extractives 1.47 1.47 1.47 0 0 0 0 0 0 0 0 0 Ash 0.80 0.80 0.80 0.38 0.38 0.38 0.38 0.38 0.38 0.38 0 0 Other 2.43 2.43 2.43 1.22 1.22 1.22 1.22 1.22 1.22 1.22 0 0 B- glucosidase 0 0 0 0 0 0 0 0 0.09 0.09 0 0 Cellulase 0 0 0 0 0 0 0.08 0.08 0.08 0.08 0 0 250 Table 64. Material balance for scenario S20 - 5- R24. P has e Co mp o ne nt (g ) 1 2 4 6 8 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Liquid H2O 474.47 1.47 475.94 2.00 477.94 477.94 0.17 477.77 434.11 43.66 1.33 0.64 1.97 221.52 28.10 152.39 181.81 3.95
UBC Theses and Dissertations
Economic feasibility of lignocellulosic ethanol production with enzyme recycling Rosales Calderon, Oscar 2014
Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.
- 24-ubc_2014_september_rosales-calderon_oscar.pdf [ 4.02MB ]
- JSON: 24-1.0167226.json
- JSON-LD: 24-1.0167226-ld.json
- RDF/XML (Pretty): 24-1.0167226-rdf.xml
- RDF/JSON: 24-1.0167226-rdf.json
- Turtle: 24-1.0167226-turtle.txt
- N-Triples: 24-1.0167226-rdf-ntriples.txt
- Original Record: 24-1.0167226-source.json
- Full Text