EXTRACTIVE AMMONIA (EA): A NOVEL AMMONIA-BASED PRETREATMENT
TECHNOLOGY FOR LIGNOCELLULOSIC BIOMASS
By
Leonardo da Costa Sousa

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Chemical Engineering – Doctor of Philosophy

2014

ABSTRACT
EXTRACTIVE AMMONIA (EA): A NOVEL AMMONIA-BASED PRETREATMENT
TECHNOLOGY FOR LIGNOCELLULOSIC BIOMASS
By
Leonardo da Costa Sousa
Large scale production of economically viable biofuels can only be achieved from widely
available resources, notably lignocellulosic biomass. This feedstock is largely composed by
complex carbohydrates, which can be enzymatically hydrolyzed and converted to fuels and
chemicals by fermentative organisms. Lignocellulosic biomass is recalcitrant to enzymatic
degradation and therefore, a pretreatment step is required for achieving acceptable sugar yields.
Recent research has improved our fundamental understanding about the physico-chemical events
occurring during ammonia pretreatment of biomass that correlate with enzymatic hydrolysis
yields. From this understanding, a novel Extractive Ammonia (EA) pretreatment was developed
and is presented herein for the first time. This technology allows the conversion of naturally
occurring cellulose I (CI) to cellulose III (CIII) and the selective extraction of lignin from the
plant cell wall with liquid ammonia. These extractives are collected in a separate stream as a
valuable byproduct. While CIII is known to improve enzymatic hydrolysis rates up to two fold
compared to native CI, lignin is acknowledged as a major inhibitor for both enzymes and
microbes. Other key events include ester-bonds cleavage via ammonolysis and hydrolysis
reactions during EA. Ester bonds play an important role in cross-linking lignin and carbohydrates
and therefore, their cleavage disrupts the complex cell wall architecture, allowing carbohydrates
to be more easily accessible by enzymes.

The effect of pretreatment variables on CIII conversion was studied using isolated cellulose from
commercial sources (Avicel). These studies revealed that samples of higher CIII crystallinity can
significantly increase enzymatic activity, which contradicts the current paradigm based on CI
results, which show that higher CI crystallinity reduces enzymatic activity of cellulases. In the
present work, EA performance was evaluated using corn stover (CS) as primary biomass
feedstock. An empirical model was created using statistical design of experiments to predict EA
pretreatment effectiveness on lignin extraction, ester-bond cleavage and sugar yields as a
function of pretreatment conditions. The results show that EA allows extraction of up to ~ 50 %
of the lignin present in corn stover, while cleaving about 70 % of the ester bonds. These effects,
in synergy with cellulose III formation, allow maximum monomeric glucose and xylose
conversions of 93 % and 79 %, respectively, using 15 mg/g glucan of enzyme for 24 h, at 1%
glucan loading enzymatic hydrolysis. At 15 % to 20% solid loadings, EA allows up to 2.7 fold
reduction of enzyme loading during enzymatic hydrolysis compared to AFEX

TM

-CS. The

benefits of EA on fermentation were also explored using the novel RaBIT process, which is
capable of doubling biofuel productivity while decreasing enzyme loading by 30 % compared
with traditional SHF. By coupling EA and RaBIT technologies was possible to efficiently
generate 191 g ethanol/Kg biomass in 48 h. using 7.5 mg/g glucan of enzyme loading.
The viability of the EA process depends on enzyme savings and on the utilization of lignin
streams. From this perspective, EA lignin extracts were fractionated by sequential precipitation
generating a fraction composed by ~92 % lignin, which represented about 32 % of the lignin
initially present in CS. Techno-economic evaluation of the EA biorefinery was also performed
and will be presented along with recommendations for further improvement of this technology.

Copyright by
LEONARDO DA COSTA SOUSA
2014

Dedicated to my dear daughter Maria Madalena and to my parents

v

ACKNOWLEDGEMENTS

I would like to express my gratitude to my research advisor, Prof. Bruce Dale, and to Prof.
Venkatesh Balan for accepting me in their team and for supporting me during my PhD.

A special acknowledgement goes to my Committee members Prof. Dan Jones, Prof. Carl Lira
and Prof. Dennis Miller for their support and suggestions to improve my work.

When I joined Prof. Dale’s lab in 2007, I found a vibrant group of people working in synergy to
overcome the challenges of lignocellulosic biomass conversion to fuels and chemicals. The work
presented herein would never be possible without the help of many of my colleagues, their
support and contribution. I would like to highlight the contribution of Shishir Chundawat,
Mingjie Jin, James Humpula, Nirmal Uppugundla, Vijay Bokade, Margie Magyer, Christa
Gunawan, Bryan Bals, Andrea Orjuela, Xiaoyu Tang and Prof. Albert Cheh to this work. I would
also like to highlight the contribution of various collaborators, notably Prof. Marcus Foston,
Prof. John Ralph, Ali Azarpira and Fachuang Lu, who brought their expertise in lignin
characterization to this work.
A special thanks goes to my colleagues Ming Woei Lau, Dahai Gao, Rebecca Ong and Cory
Sarks for the vibrant discussions, support, friendship and for providing me a pleasant work
environment.
I would also like to exend my gratitude to Pete Donald, Brandon Guthrie, Aaron Vigil and Lucas
Holcomb for their hard work and commitment. They were a vital part of this project, as they

vi

helped me building reactors, performing pretreatment experiments and composition analysis on
many of my samples.

I would like to thank the Great Lakes Bioenergy Research Center (GLBRC) and Fundacao para a
Ciencia e Tecnologia (FCT) for funding my PhD work.

A very important acknowledgment goes to those who supported me in the good and bad times
during my stay in US. Thanks to my friends Anchita, Durga, Alvaro, Ven, Mingjie, Shishir and
my ex-wife Xana for their support and encouragement.

Finally, I would like to express my gratitude to my family, especially to my parents Celeste
Sousa and Mateus Sousa, and to my daughter Madalena, for their love, support, motivation and
encouragement.

vii

TABLE OF CONTENTS

LIST OF TABLES ..................................................................................................................... xiii

LIST OF FIGURES .................................................................................................................... xv

CHAPTER 1 - INTRODUCTION.............................................................................................. 1
1.1. Project objectives ............................................................................................................... 1
CHAPTER 2 – LITERATURE REVIEW .................................................................................. 5
2.1. Lignocellulosic Biomass ..................................................................................................... 5
2.1.1. Cellulose ....................................................................................................................... 5
2.1.2. Hemicelluloses ............................................................................................................. 7
2.1.3. Pectins .......................................................................................................................... 9
2.1.4. Lignins........................................................................................................................ 11
2.2. Lignocellulosic Biorefinery ............................................................................................. 12
2.3. Importance of Pretreatment in the Lignocellulosic Biorefinery Context ................... 16
2.4. Important Pretreatment Technologies Available Today.............................................. 18
2.4.1. Physical Pretreatment .............................................................................................. 18
2.4.2. Solvent Fractionation................................................................................................ 18
2.4.3. Chemical Pretreatment ............................................................................................ 20
2.4.3.1. Acidic pretreatments ........................................................................................... 20
2.4.3.2. Alkaline pretreatments ........................................................................................ 22
2.4.3.3. Oxidative pretreatments ...................................................................................... 25
2.4.4. Biological pretreatment ............................................................................................ 26
2.5. Considerations for an ideal pretreatment technology .................................................. 27
2.6. Ammonia and Alkaline Pretreatment Chemistry ......................................................... 28
2.6.1. Reactions involving hydroxyl ions and lignocellulosic biomass............................ 28
2.6.1.1. Reactions with polysaccharides. ......................................................................... 28
2.6.1.2. Reactions with Lignin ......................................................................................... 30
2.6.2. Reactions involving ammonia and lignocellulosic biomass ................................... 32
2.6.2.1. Ammonolysis Reactions ...................................................................................... 32
2.6.3. Maillard Reactions .................................................................................................... 34
2.6.4. Rearrangement of cellulose crystalline state (Cellulose Iβ to IIII) ....................... 34
2.7. Major downstream processing bottlenecks in the biorefinery .................................... 37
2.7.1. “Solids effect” and enzyme loading ......................................................................... 37
2.7.2. Enzyme and microbial inhibition ............................................................................ 40

CHAPTER 3 - FACTORS THAT CONTRIBUTE TO CELLULOSE III CONVERSION
AND ASSOCIATED IMPACT ON ENZYMATIC DIGESTIBILITY OF CELLULOSE. 43
viii

3.1. Abstract ............................................................................................................................. 43
3.2. Introduction ...................................................................................................................... 44
3.3. Experimental .................................................................................................................... 46
3.3.1. Reaction apparatus ................................................................................................... 46
3.3.2. Cellulose III conversion studies ............................................................................... 48
3.3.2.1. Impact of sampling solvent on the conversion of ammonia-cellulose complex to
CIII ................................................................................................................................... 48
3.3.2.2. Effect of ammonia loading on CIII conversion ................................................. 49
3.3.2.3. Effect of temperature on cellulose III conversion ............................................. 49
3.3.2.4. Effect of ethanol:ammonia solvent system on CIII conversion ........................ 50
3.3.3. Preparation of cellulose standards .......................................................................... 51
3.3.4. X-ray powder diffraction (XRD) ............................................................................. 51
3.3.5. XRD data analysis ..................................................................................................... 52
3.3.6. Enzymatic hydrolysis ................................................................................................ 55
3.4. Results and Discussion..................................................................................................... 55
3.4.1. Effect of sampling media on cellulose III conversion ............................................ 55
3.4.2. Effect of ammonia loading on cellulose III conversion ......................................... 59
3.4.3. Effect of temperature and time on cellulose III conversion .................................. 63
3.4.4. Effect of solvent concentration on cellulose III conversion ................................... 68
3.5. Conclusion ........................................................................................................................ 72

CHAPTER 4 - EXTRACTIVE AMMONIA (EA) PRETREATMENT OF CORN
STOVER: REACTIONS IN ANHYDROUS AMMONIA. ..................................................... 75
4.1. Abstract ............................................................................................................................. 75
4.2. Introduction ...................................................................................................................... 76
4.3. Experimental .................................................................................................................... 78
4.3.1. Untreated corn stover ............................................................................................... 78
4.3.2. Extractive Ammonia (EA) pretreatment reactor construction ............................ 78
4.3.3. EA Pretreatment of corn stover............................................................................... 81
4.3.4. Experimental design for pretreatment.................................................................... 82
4.3.5. Composition analysis of plant cell wall components .............................................. 85
4.3.6. Enzymatic hydrolysis of EA pretreated corn stover .............................................. 87
4.3.7. X-Ray Diffraction (XRD) of EA pretreated corn stover ....................................... 87
4.4. Results and Discussion..................................................................................................... 88
4.4.1. Extraction of p-coumarate and ferulate from corn stover during EA
pretreatment ........................................................................................................................ 88
4.4.1.1. Analysis of ester-linked p-coumarate and ferulate in untreated and EA
pretreated corn stover....................................................................................................... 88
4.4.1.2. Response surface analysis: % cleavage of p-coumarate and ferulate as a
function of pretreatment conditions. ............................................................................... 90
4.4.2. Extraction of lignin from corn stover during EA pretreatment ........................... 96
4.4.2.1. Analysis of lignin extraction in untreated and EA pretreated corn stover ....... 96
4.4.2.2. Response surface analysis: % lignin extraction as a function of pretreatment
conditions ......................................................................................................................... 99
ix

4.4.3. Enzymatic hydrolysis of EA pretreated corn stover ............................................ 104
4.4.3.1. Enzymatic digestibility of EA pretreated corn stover ....................................... 104
4.4.3.2. Response surface analysis of enzymatic hydrolysis performance of EA
pretreated corn stover..................................................................................................... 112
4.4.4. Model integration and process optimization ........................................................ 120
4.5. Conclusion ...................................................................................................................... 122
APPENDIX ............................................................................................................................ 125
CHAPTER 5 – HIGH SOLID LOADING ENZYMATIC HYDROLYSIS OF
EXTRACTIVE AMMONIA (EA) PRETREATED CORN STOVER ................................ 129
5.1. Abstract ........................................................................................................................... 129
5.2. Introduction .................................................................................................................... 130
5.3. Experimental .................................................................................................................. 131
5.3.1. Untreated corn stover ............................................................................................. 131
5.3.2. Extractive Ammonia (EA) pretreatment reactor................................................. 132
5.3.3. Pretreatment of corn stover ................................................................................... 134
5.3.4. X-ray powder diffraction (XRD) ........................................................................... 135
5.3.5. Enzymatic hydrolysis for EA pretreatment optimization ................................... 136
5.3.6. Experimental design for high solid loading enzymatic hydrolysis ..................... 136
5.3.7. High solids loading enzymatic hydrolysis ............................................................. 139
5.4. Results and Discussion................................................................................................... 139
5.4.1. Evaluation of EA pretreatment on corn stover .................................................... 139
5.4.2. High solid loading enzymatic hydrolysis of EA corn stover ............................... 146
5.4.2.1. Definition of enzymatic hydrolysis conditions: fixed variables ....................... 146
TM
5.4.2.2. Empirical modeling of high solid loading enzymatic hydrolysis: AFEX -CS
......................................................................................................................................... 148
5.4.2.3. Empirical modeling of high solid loading enzymatic hydrolysis: EA-CS ....... 153
TM
5.4.2.4. Performance of high solid loading enzymatic hydrolysis on EA- and AFEX CS.................................................................................................................................... 158
5.5. Conclusion ...................................................................................................................... 169
APPENDIX ............................................................................................................................ 172

CHAPTER 6 - ISOLATION AND CHARACTERIZATION OF LIGNIN DERIVED
FROM THE NOVEL EXTRACTIVE AMMONIA (EA) PRETREATMENT. ................. 174
6.1. Abstract ........................................................................................................................... 174
6.2. Introduction .................................................................................................................... 175
6.3. Experimental .................................................................................................................. 177
6.3.1. Untreated corn stover ............................................................................................. 177
6.3.2. Extractive Ammonia (EA) pretreatment of corn stover ..................................... 177
6.3.3. EA extractives fractionation and production of lignin rich streams from EA
process ................................................................................................................................ 179
6.3.4. Enzymatic hydrolysis (EH) of EA pretreated corn stover .................................. 180
x

6.4. Lignin mass balance....................................................................................................... 181
6.4.1. Characterization of EA extractives fractions ....................................................... 182
6.4.1.1. Lignin characterization by NMR...................................................................... 182
6.4.1.2. Gel permeation chromatography (GPC) .......................................................... 183
6.4.1.3. Elemental analysis ............................................................................................ 184
6.4.1.4. Thermogravimetric analysis (TGA) and differential scanning calorimetry
(DSC) analysis ................................................................................................................ 184
6.5. Results and Discussion................................................................................................... 185
6.5.1. EA pretreatment process and lignin mass balance .............................................. 185
6.5.2. NMR characterization of EA extractives fractions from corn stover ................ 189
13
6.5.2.1. C NMR and 2D NMR analysis of lignin derived from EA pretreated corn
stover ............................................................................................................................... 189
31
6.5.2.2. P NMR analysis of lignin derived from EA pretreatment of corn stover .... 194
6.5.3. Gel Permeation Chromatography (GPC) ............................................................. 195
6.5.4. Elemental analysis ................................................................................................... 198
6.5.5. Thermogravimetric analysis (TGA) of the EA fractions..................................... 201
6.6. Conclusion ...................................................................................................................... 204
APPENDIX ............................................................................................................................ 206
CHAPTER 7 – CONVERSION OF EXTRACTIVE AMMONIA PRETREATED CORN
STOVER TO ETHANOL BY RABIT PROCESS................................................................. 208
7.1. Abstract ........................................................................................................................... 208
7.2. Introduction .................................................................................................................... 208
7.3. Experimental .................................................................................................................. 210
7.3.1. EA pretreated corn stover ...................................................................................... 210
7.3.2. Microorganisms and seed culture preparation .................................................... 210
7.3.3. Conventional separate enzymatic hydrolysis (EH) and fermentation (SHF) .... 211
7.3.4. Rapid Bioconversion with Integrated Enzyme Recycling Technology (RaBIT)212
7.4. Results and Discussion................................................................................................... 213
7.4.1. Conventional SHF ................................................................................................... 213
7.4.2. Optimization of enzyme loadings for the cycle 2 and the last step of RaBIT .... 215
7.4.3. Performance of RaBIT ........................................................................................... 219
7.4.4. Comparison of RaBIT and conventional SHF ..................................................... 222
7.5. Conclusion ...................................................................................................................... 227
CHAPTER 8 – EXTRACTIVE AMMONIA (EA) PRETREATMENT PROCESS DESIGN,
PRELIMINARY ECONOMICS AND FINAL CONSIDERATIONS. ................................ 229
8.1. Abstract ........................................................................................................................... 229
8.2. Introduction .................................................................................................................... 230
8.3. Experimental .................................................................................................................. 231
8.3.1. EA pretreatment process flow diagram (PFD) .................................................... 231
8.3.2. Material and energy balances around EA pretreatment .................................... 232
xi

8.3.3. Equipment sizing and cost ..................................................................................... 234
8.3.4. Techno-economic analysis ...................................................................................... 234
8.4. Results and Discussion................................................................................................... 236
8.4.1. Material and energy balances for the EA pretreatment model .......................... 236
8.4.2. Capital cost of EA pretreatment ............................................................................ 239
8.4.3. Material balance for enzymatic hydrolysis and fermentation and MESP
calculation .......................................................................................................................... 240
8.4.4. Performance of EA-based biorefinery .................................................................. 243
8.5. Conclusion ...................................................................................................................... 247
APPENDIX ............................................................................................................................ 250

REFERENCES .......................................................................................................................... 260

xii

LIST OF TABLES

Table 2-1 Composition of lignocellulosic materials in cellulose, hemicellulose, lignin and ash
[5-7]................................................................................................................................................. 5
Table 2-2 Optimum operating conditions of some of the leading pretreatment technologies
that maximize enzymatic digestibility of pretreated corn stover [67-70]. ............................. 23
Table 2-3 Effect of leading thermochemical pretreatments on the physicochemical
properties of corn stover and respective glucan and xylan conversion after enzymatic
hydrolysis. .................................................................................................................................... 24
Table 4-1 High and low limits defined for the experimental points using Box Benhken
design............................................................................................................................................ 83
Table 4-2 EA pretreatment conditions defined by a Box Benhken design. ........................... 83
Table 4-3 Compound-dependent parameters after mass spectrometry optimization. ......... 86
Table 4-4 Pretreatment conditions used for the Box Benhken design and the respective
number for reference of model fit plots below. ...................................................................... 126
Table 5-1 High and low limits defined for the experimental points using Box Benhken
design.......................................................................................................................................... 137
Table 5-2 Enzymatic hydrolysis conditions defined by a Box Benhken DOE. .................... 138
Table 5-3 Enzyme mixtures used during enzyme cocktail optimization for EA and
TM
AFEX -CS.............................................................................................................................. 173
Table 5-4 Enzyme optimization results for EA and AFEX-CS using enzyme mixtures 1 to 6.
Glucose (Glc), xylose, (Xyl) and arabinose (Ara) concentrations are presented here as a
measure of each enzyme cocktail performance after 72 h enzymatic hydrolysis. Highlighted
in bold letters is the enzyme mixture used for this study. ..................................................... 173
Table 6-1 Compositional analysis of fractions F1 to F5. ....................................................... 188
13

Table 6-2 Assignments and integration value of quantitative
C NMR spectra of
fractionated lignin from EA pretreated corn stover [186]. ................................................... 192
Table 6-3 Abundance of assigned hydroxyl groups of fractionated EA extractives from corn
31
stover, determined by P NMR [189]. ................................................................................... 195
xiii

Table 6-4 Mineral inorganic elements and CHNO analysis of fractions derived from the EA
process. ....................................................................................................................................... 200
Table 7-1 Design for the optimization of enzyme loadings (EL) for the Cycle 2 and the last
step.............................................................................................................................................. 215
Table 7-2 Compositions of unhydrolyzed solids from RaBIT and conventional SHF
processes..................................................................................................................................... 227
Table 8-1 Summary of equipment and capital cost for EA pretreatment. .......................... 239
Table 8-2 Ethanol and lignin yields for the optimized base case scenario of an EA
biorefinery. ................................................................................................................................ 242
Table 8-3 Economic considerations for the base case EA biorefinery (2007 US$). ............ 243
Table 8-4 Stream table for EA pretreatment process............................................................ 253
Table 8-5 Total energy duty of EA pretreatment. ................................................................. 253
Table 8-6 EA reactor specifications and cost. ........................................................................ 254
Table 8-7 Ammonia storage tank specifications and cost. .................................................... 254
Table 8-8 Ammonia evaporation column specifications and cost for ammonia recycling
system. ........................................................................................................................................ 254
Table 8-9 Distillation equipment specifications and cost for ammonia recycling system of
EA pretreatment. ...................................................................................................................... 255
Table 8-10 Pump specifications and cost for ammonia recycling system of EA
pretreatment. ............................................................................................................................. 256
Table 8-11 Compressor specifications and cost for ammonia recycling system of EA
pretreatment. ............................................................................................................................. 256
Table 8-12 Utility costs and pretreatment capital cost assumed for economic model of
a
biorefinery ................................................................................................................................ 257
Table 8-13 Regression coefficients for MESP prediction as a function of enzymatic
hydrolysis conditions for base case EA pretreated corn stover. ........................................... 258

xiv

LIST OF FIGURES

Figure 2-1 Graphical representation of the structure of cellulose microbibrils. Cellulose
microfibrils (green) are composed of chains of -(1,4)-linked D-glucose subunits. ...................... 7
Figure 2-2 Graphical representations of the major hemicellulose structures in grasses, i.e.,
Xyloglucan (left) and Arabinoxylan (right). Xyloglucan is composed by a glucan backbone
decorated by xylose residues (circled in red, Xyl). Arabinoxylan is composed by a xylan
backbone decorated with arabinose residues (circled in green, Arb) and glucuronic acid residues
(circled in blue). .............................................................................................................................. 8
Figure 2-3 Representation of the major pectin domains present in plants. .................................. 10
Figure 2-4 Representation of the general precursor of lignin biosynthesis in angiosperms. (pcoumaryl alcohol: R = R’=H ; coniferyl alcohol: R = H, R’= OCH3 ; and sinapyl alcohol: R = R’
= OCH3). ...................................................................................................................................... 11
Figure 2-5 Process flow diagram representing the unit operations for biological conversion of
lignocellulosic biomass into ethanol [1]. ...................................................................................... 13
Figure 2-6 Description of the various biorefinery process configurations and the number of tanks
required for each configuration. .................................................................................................... 14
Figure 2-7 Pretreatment choice affects all biorefinery related operations. This fact gives
pretreatment a central role affecting the design and economic viability of a lignocellulosic
biorefinery [1]. .............................................................................................................................. 17
Figure 2-8 Simplified diagram of proposed ferulate and diferulate cross linkages with lignin and
arabino-xylans. 1- Acetyl decoration in xylan backbone; 2- Ester linked ferulate to arabinoxylan; 3- Ester linked arabinose to lignin; 4- 5-5 ester linked ferulate dimer cross-linking
arabino-xyalan chains; 5 –Ether-linked ferulate to lignin............................................................. 31
Figure 2-9 Correlation between enzymatic digestibility of AFEX™ pretreated corn stover and
total feruloyl cleavage (amide + acid) during AFEX™ pretreatment [23]. .................................. 33
Figure 2-10 i) Representation of cellulose Iβ and IIII crystal structure, including hydrogen
bonding network and glucan chain organization. ii) Synergistic effect of endocellulases and
exocellulases in cellulose Iβ (red) and IIII (black) hydrolysis [105]. ........................................... 36
Figure 2-11 Glucose yields after enzymatic hydrolysis of the different cellulose substrates.
Enzymatic hydrolysis was performed using 1.5 FPU Spezyme CP cellulase/g glucan and 6.4 pNPGU Novo188 β-glucosidase /g glucan. Left and right bars represent 6 h and 24h enzymatic
hydrolysis data [105]..................................................................................................................... 37
xv

Figure 2-12 i) Solids effect observed for different time points of enzymatic hydrolysis. ii) Effect
of T2 relaxation time of restricted bulk water (RBW) on enzymatic digestibility after 24h. iii)
Effect of solids content on T2 relaxation time of RBW [2]. ......................................................... 39
Figure 2-13 i) Relationship between enzyme adsorption and initial solids content [3]. ii)
Maximum cellulase adsorption on cellulose I (red) and cellulose III (black) for three different
kinds of cellulases [105]. .............................................................................................................. 40
Figure 2-14 Contribution of AFEX™ decomposition products, ethanol and fermentation
metabolites on xylose consumption rate and cell growth of S.cerevisiae 424A(LNH-ST) [109]. 42
Figure 3-1 Instrumental apparatus for conversion of CI to CIII: 1) High-pressure mixing vessel
and mantle used to preheat ammonia to set-point temperature; 2) Temperature controller for
ammonia preheating vessel; 3) Water circulator used to heat the syringe pump to reaction setpoint temperature; 4) Syringe pump to pump the desired volume of ammonia; 5) Balance to
confirm the delivery of the desired ammonia loading; 6) Backpressure regulator; 7) Highpressure mixed reaction vessel and mantle for cellulose III conversion; 8) Temperature controller
for reaction vessel; 9) Solenoid valve for pressure control (nitrogen inlet); 10) Solenoid valve for
pressure control (nitrogen outlet), also used for venting the system; 11) Deep tube sampling port;
12) computer controls and data logging........................................................................................ 47
Figure 3-2 Example of the XRD spectra of cellulose I (CI), cellulose III (CIII) and amorphous
cellulose (AC). Amorphous subtraction method was applied to CI and cellulose CIII standards
for this study. The yellow area was subtracted from the XRD spectra of CI and CIII standards as
described in this figure to recreate the 100% crystalline CI and CIII spectra used for XRD spectra
deconvolution. ............................................................................................................................... 53
Figure 3-3 Example of an XRD peak deconvolution applied to a sample containing a mixture of
amorphous cellulose (AC), cellulose I (CI) and cellulose III (CIII). The model spectra (red line),
contrasting with the sample XRD spectra (black line) is presented in the top right corner. ........ 54
Figure 3-4 i) Glucose conversion of cellulose treated at 90 C with 6:1 (w/w) ammonia-tocellulose ratio for 20 min residence time, sampled to various solvents and to no solvent (air). The
control experiment corresponds to cellulose I (Avicel PH-101). Enzymatic hydrolysis was
performed using 4 mg/g glucan of Accellerase 1500 enzymes for 24h; ii. a) Crystallinity index
(CrI) of the various samples of ammonia treated cellulose sampled to different solvents and to no
solvent (air). CrI measurement was performed by amorphous subtraction method of the XRD
spectra showed in b)...................................................................................................................... 59
Figure 3-5 i) Glucose conversion of ammonia treated cellulose under various ammonia:cellulose
ratios, varying from 1:1 to 6:1 (w/w), compared to the control (Avicel PH-101). Enzyme loading
used in the experiment was 4 mg/g glucan for a residence time of 24h; ii) XRD spectra of
ammonia treated cellulose using various ammonia:cellulose ratios at 90 C for 10 min residence
time. .............................................................................................................................................. 62
xvi

Figure 3-6 i) Relative proportions of cellulose III (CIII), cellulose I (CI) and amorphous
cellulose (AC) during anhydrous ammonia treatment of Avicel PH-101 at 25 °C, as a function of
residence time (a) and respective XRD spectra used to calculate CIII, CI and AC contents at 25
°C (b). ii) Relative proportions of cellulose III (CIII), cellulose I (CI) and amorphous cellulose
(AC) during anhydrous ammonia treatment of Avicel PH-101 at 90 °C, as a function of residence
time (a) and respective XRD spectra used to calculate CIII, CI and AC contents at 90 °C. ........ 65
Figure 3-7 i) Comparison between amorphous content of Avicel PH-101 treated at various
temperatures and correspondent effect on 4h glucose conversion during enzymatic hydrolysis.
Enzymatic hydrolysis was performed with 7.5 mg/g glucan of Accelerase 1500 enzymes at 50
°C. ii) XRD spectra of Avicel PH-101 treated with phosphoric acid (PASC) and with ammonia at
-30, 25 and 90 °C. At 25 and 90 °C treatment, ammonia was evaporated at 25 °C in ethanol,
while at -30 °C ammonia was vacuum filtered and dried in the hood at ambient temperature. ... 66
Figure 3-8 Vapor-liquid equilibrium curve (P-xy) for ammonia/ethanol mixture at 120 °C. The
curve was predicted using NRTL model, estimating missing parameters using UNIFAC by
Aspen Plus (v.7.1) software. ......................................................................................................... 69
Figure 3-9 Relative proportions of cellulose I (CI), cellulose III (CIII) and amorphous cellulose
(AC) as a function of ammonia concentration and liquid-to-solid ratio during ammonia
treatment. The ammonia concentrations used in this study were 17 wt% (i), 33 wt% (ii), 50 wt%
(iii) and 60 wt% (iv). Ammonia treatment was performed at 25 °C for 30 min residence time
using 4:1, 6:1 and 10:1 liquid-to-solid ratios for all ammonia concentrations. The crystalline
composition of cellulose was correlated with 24h glucose conversion using 4 mg/g glucan of
enzyme for the various samples. ................................................................................................... 72
Figure 4-1 Schematic drawing of the EA pretreatment reactor (A.) and extractives collector (B.).
The pictures of the reaction cell (6) and of the stainless steel filter installed in the cap of the
reactor (6’) are highlighted. The EA reactor outlet connector (8) is attached to the extractives
collector inlet connector (8’) during operation. ............................................................................ 79
Figure 4-2 (A) Residual plots for the regression of ester-linked pCA % cleavage as a function of
pretreatment variables (temperature, time and NH3:BM ratio) and respective interaction terms.
Residuals and fitted values are expressed in (% cleavage). Regression terms were only
considered if their P value was smaller than 0.05. (B) Estimation of regression coefficients and P
values for the terms affecting ester-linked p-coumaric acid % cleavage. Parameters that
determine the goodness of the fit are also shown. ........................................................................ 92
Figure 4-3 (A) Residual plots for the regression of ester-linked FA % cleavage as a function of
pretreatment variables (temperature, time and NH3:BM ratio) and respective interaction terms.
Residuals and fitted values are expressed in (% cleavage). Regression terms were only
considered if their P value was smaller than 0.05. (B) Estimation of regression coefficients and P
values for the terms affecting ester-linked ferulic acid % cleavage. Parameters that determine the
goodness of the fit are also shown. ............................................................................................... 93
xvii

Figure 4-4 Contour plots describing % cleavage of ester-linked p-coumaric (A) and ferulic acid
(B) as a function of EA pretreatment conditions. Experimental values were calculated based on
the content of ester-linked p-coumaric and ferulic acids present in corn stover prior to EA
pretreatment. The contour plots represent a small area of the universe described by the respective
mathematical models, with fixed values of NH3:BM = 6/1 (I), time = 30 min (II) and
temperature = 115 C (III). ........................................................................................................... 96
Figure 4-5 Stacked bar chart comparing the mean values for the composition of EA pretreated
corn stover (EA-CS) with respect to the untreated corn stover (UT-CS). The pretreatment
condition used for this chart was temperature = 121.8 C , NH3:BM loading of 6.2:1 and
residence time of 32 min, which represents the highest severity within the boundary conditions
defined for EA pretreatment in this work. In this analysis, only structural carbohydrates from the
plant cell wall and Klason lignin were considered. Coefficients of variation were below 2.5 % for
all measurements. .......................................................................................................................... 98
Figure 4-6 (A) Residual plots for the regression of % lignin extraction as a function of
pretreatment variables (temperature, time and NH3:BM ratio) and respective interaction terms.
Residuals and fitted values are expressed in (% extraction). (B) Estimation of regression
coefficients and P values for the terms affecting % lignin extraction. Parameters that determine
the goodness of the fit are also shown. ....................................................................................... 100
Figure 4-7 Contour plots describing % lignin extraction as a function of EA pretreatment
conditions. Experimental values were calculated based on the lignin content in corn stover prior
to EA pretreatment. The contour plots represent a small area of the universe described by the
respective mathematical models, with fixed values of NH3:BM = 6/1 (I), time = 30 min (II) and
temperature = 115 C (III). ......................................................................................................... 103
Figure 4-8 Conversions to monomeric glucose and xylose after 24 h enzymatic hydrolysis of
various EA-CS samples (A1-C1). Glucan conversion of optimally pretreated AFEX-CS is
indicated by a green dashed line. XRD spectra from EA-CS as a function of pretreatment
conditions (A2-C2). EA conditions are located in the top left corner of each XRD spectra and
include temperature; NH3:BM loading and time, in this respective order. Cellulose I and III
controls are Avicel and ammonia treated Avicel, respectively................................................... 106
Figure 4-9 Correlation of % glucan conversion as a function of % lignin extraction. The
regression had a P value = 0.000 for the term % lignin extraction, suggesting that lignin
extraction is related to % glucan conversion. The regression (black line), 95 % confidence
interval (red line) and 95 % prediction interval (green line) are shown in the graphic. ............. 109
Figure 4-10 Linear regressions correlating glucan (A) and xylan (C) conversions with ferulic
acid cleavage, as well as correlating glucan (B) and xylan (D) conversions with p-coumaric acid
cleavage. Regression equations P values and R2 values are indicated in the bottom of each
graphic......................................................................................................................................... 111

xviii

Figure 4-11 (A) Residual plots for the regression of 24h % glucan conversion to monomeric
glucose as a function of pretreatment variables (temperature, time and NH3:BM ratio) and
respective interaction terms. Residuals and fitted values are expressed in (% conversion). (B)
Estimation of regression coefficients and P values for the terms affecting % glucan conversion.
Parameters that determine the goodness of the fit are also shown. ............................................ 114
Figure 4-12 (A) Residual plots for the regression of 24h % xylan conversion to monomeric
xylose as a function of pretreatment variables (temperature, time and NH3:BM ratio) and
respective interaction terms. Residuals and fitted values are expressed in (% xylan conversion).
(B) Estimation of regression coefficients and P values for the terms affecting % xylan
conversion. Parameters that determine the goodness of the fit are also shown. ......................... 116
Figure 4-13 Contour plots describing monomeric 24h % glucan (A) and % xylan (B)
conversions as a function of EA pretreatment conditions. Experimental values were calculated
based on the glucan and xylan contents of the EA pretreated biomass. Maximum glucan and
xylan loss during pretreatment was 6.4 % and 4.4 %, respectively, at maximum severity
conditions (i.e. temperature = 121.8 C, NH3:BM loading of 6.2:1 and residence time of 32 min).
The contour plots represent a small area of the universe described by the respective mathematical
models, with fixed values of NH3:BM = 6/1 (I), time = 30 min (II) and temperature = 115 C
(III). ............................................................................................................................................. 118
Figure 4-14 Model prediction fits to experimental data for 24h glucose conversion as a function
of EA conditions. ........................................................................................................................ 126
Figure 4-15 Model prediction fits to experimental data for 24h xylose conversion as a function
of EA conditions. ........................................................................................................................ 127
Figure 4-16 Model prediction fits to experimental data for lignin extraction as a function of EA
conditions. ................................................................................................................................... 127
Figure 4-17 Model prediction fits to experimental data for p-coumarate ester cleavage as a
function of EA conditions. .......................................................................................................... 128
Figure 4-18 Model prediction fits to experimental data for ferulate ester cleavage as a function
of EA conditions. ........................................................................................................................ 128
Figure 5-1 Picture of scaled-up EA pretreatment system, composed by 3 high-pressure reactors
of 600 mL each, ammonia delivery system and extractives collector. ....................................... 133
Figure 5-2 i) Effect of moisture content on enzymatic hydrolysis and lignin extraction
performances. Experiments were performed at 110 °C for 60 min residence time and 3:1
ammonia loading; ii) Effect of temperature on enzymatic hydrolysis and lignin extraction
performances. Experiments were performed with 10 % moisture content, 3:1 ammonia loading
and 60 min residence time; iii) Effect of ammonia:biomass ratio at 30 min residence time on
enzymatic hydrolysis and lignin performances. Experiments were performed with 10 % moisture
content at 120 °C; iv) Effect of ammonia:biomass ratio at 60 min residence time on enzymatic
xix

hydrolysis and lignin extraction performances. Experiments were performed with 10 % moisture
content at 120 °C. Enzymatic hydrolysis in all cases was performed with an enzyme loading of
15 mg/g glucan for 24 h at 50 °C and 250 RPM......................................................................... 142
Figure 5-3 Comparison between experimental observations and predicted observations by the
model presented in CHAPTER 4 for glucose, xylose and lignin extraction performances.
Pretreatment conditions were performed with a fixed biomass moisture of 10 %, 120 °C and 30
min residence time, varying ammonia:biomass ratio from 3:1 to 6:1. ....................................... 144
Figure 5-4 Process mass balance for the EA pretreatment condition chosen for this study.
Enzymatic hydrolysis conversion (%X) is a function of variables such as time, % solids loading
and enzyme loading during enzymatic hydrolysis. Empirical modeling was performed to
determine these relationships in this study and complete the material balance around
pretreatment and enzymatic hydrolysis. ..................................................................................... 146
Figure 5-5 (A) Residual plots for the regression of glucose conversion on AFEXTM-CS as a
function of enzymatic hydrolysis variables (time, solid loading and enzyme loading) and
respective interaction terms. Residuals and fitted values are expressed in (glucose conversion).
(B) Estimation of regression coefficients and P values for the terms affecting glucose conversion.
Parameters that determine the goodness of the fit are also shown. ............................................ 149
Figure 5-6 (A) Residual plots for the regression of xylose conversion on AFEXTM-CS as a
function of enzymatic hydrolysis variables (time, solid loading and enzyme loading) and
respective interaction terms. Residuals and fitted values are expressed in (xylose conversion). (B)
Estimation of regression coefficients and P values for the terms affecting xylose conversion.
Parameters that determine the goodness of the fit are also shown. ............................................ 151
Figure 5-7 (A) Residual plots for the regression of glucose conversion on EA-CS as a function
of enzymatic hydrolysis variables (time, solid loading and enzyme loading) and respective
interaction terms. Residuals and fitted values are expressed in (glucose conversion). (B)
Estimation of regression coefficients and P values for the terms affecting glucose conversion.
Parameters that determine the goodness of the fit are also shown. ............................................ 156
Figure 5-8 (A) Residual plots for the regression of xylose conversion on EA-CS as a function of
enzymatic hydrolysis variables (time, solid loading and enzyme loading) and respective
interaction terms. Residuals and fitted values are expressed in (xylose conversion). (B)
Estimation of regression coefficients and P values for the terms affecting xylose conversion.
Parameters that determine the goodness of the fit are also shown. ............................................ 157
Figure 5-9 Comparison of average experimental observations and model predictions for i)
glucose conversion and ii) xylose conversion for EA- and AFEXTM-CS. iii) Experimental
conditions formulated by the Box Benhken design, which were used for model formulation. .. 160

xx

Figure 5-10 Contour plots with model predictions for A) glucose and B) xylose conversion for
AFEXTM-CS, as a function of time, enzyme loading and solids loading during enzymatic
hydrolysis. Fixed variables were chosen to be at the “middle point” conditions. ...................... 162
Figure 5-11 Contour plots with model predictions for A) glucose and B) xylose conversion for
EA-CS, as a function of time, enzyme loading and solids loading during enzymatic hydrolysis.
Fixed variables were chosen to be at the “middle point” conditions. ......................................... 163
Figure 5-12 Kinetics of enzymatic hydrolysis performed at i) 15 % solids and ii) 20 % solids, for
enzyme loadings of a) 20 mg/ g glucan, b) 10 mg/ g glucan and c) 7.5 mg/ g glucan. The kinetics
was performed on EA- and AFEXTM-CS and accounts for both glucose and xylose conversion.
..................................................................................................................................................... 164
Figure 5-13 i) Terminal (120 h) glucose and xylose conversion as a function of solids loading
and enzyme loading; ii) Absolute improvement (AI) in glucose and xylose conversion observed
for EA-CS with respect to AFEXTM-CS; iii) Relative improvement (RI) .................................. 168
Figure 6-1 Schematic illustration of the EA pretreatment apparatus and process streams. ....... 178
Figure 6-2 Sugar conversion based on monomeric + oligomeric glucose and xylose during
enzymatic hydrolysis of AFEXTM and EA pretreated corn stover. ............................................. 186
Figure 6-3 Illustration of EA lignin processing and mass balance. Fraction F6 is the enzymatic
hydrolyzate, rich in glucose and xylose, which is usually fermented to ethanol in a
lignocellulosic ethanol biorefinery. Fractions F1-5 may be utilized for value addition within the
biorefinery, with special attention to fraction F3 and F5 that contain 32% and 51% of the lignin
originally present in untreated corn stover, respectively. ........................................................... 187
Figure 6-4 2D-HSQC NMR of fractions derived from EA pretreated corn stover................... 193
Figure 6-5 Gel permeation chromatograms of fractions F1-F5. Weight average molecular weight
(Mw) was calculated based on polystyrene standards. ................................................................ 197
Figure 6-6 Thermogravimetric curves for the various biomass fractions (F1-F5) derived from
the EA pretreatment of corn stover. The heating ramps for all the samples were set at 10 C/min.
..................................................................................................................................................... 202
Figure 6-7 DSC of F3 lignin showing the glass transition temperature at approximately 117°C.
Black and red curves (top and bottom) are sample replicates, while the green curve (middle) is
the average of the two replicates. ................................................................................................ 207
Figure 7-1 Diagram of the RaBIT process (adapted from Jin, M. et al. (2012) [4]).................. 213
Figure 7-2 Enzymatic hydrolysis (a) and fermentation (b) performance of conventional SHF on
EA pretreated corn stover. .......................................................................................................... 214
xxi

Figure 7-3 Optimization of the enzyme loadings (EL) for the Cycle 2 (a) and the last step (b)
during RaBIT process on EA CS. Experiment design was shown in Table 7-1. ........................ 217
Figure 7-4 Enzymatic hydrolysis performance (a), fermentation performance (b) and oligomeric
sugars hydrolyzed and consumed during fermentation (c) using the RaBIT process................. 220
Figure 7-5 Sugar conversions during enzymatic hydrolysis (a) and sugar-to-ethanol conversions
(b) for the processes of RaBIT and conventional SHF. .............................................................. 224
Figure 7-6 Comparison of ethanol yield (a), ethanol productivity (b) and timelines (c) for RaBIT
and conventional SHF processes. Ethanol productivity was calculated based on enzymatic
hydrolysis plus fermentation. ...................................................................................................... 225
Figure 7-7 Mass balance comparison of RaBIT and conventional SHF processes. .................. 227
Figure 8-1 Pie chart describing the heat consumption of the various unit operations during EA
pretreatment. The total heat duty is 9.08 GJ/ ton biomass. ......................................................... 238
Figure 8-2 Pie chart describing the heat consumption of the various unit operations during EA
pretreatment. The total electricity consumption is 89.59 kWh/ ton biomass.............................. 238
Figure 8-3 MESP values for biorefineries based on different pretreatment technologies. ........ 244
Figure 8-4 Sensitivity analysis chart for measuring the impact of variations in utility and capital
cost of EA pretreatment on the MESP. The impact of enzyme cost and lignin selling price on the
MESP are also presented. ........................................................................................................... 246
Figure 8-5 EA pretreatment process flow diagram. EA1 and EA2 work intermittently one after
the other, sharing the equipment used for ammonia recovery (highlighted in blue). This system of
two reactors (EA1 and EA2) is replicated five times to compose the full biorefinery, scaled for
processing 2000 tons/day of untreated corn stover. The ammonia recycling system duplicated to
be shared among the 10 reactors that are synchronized during operation. ................................. 251
Figure 8-6 Schedule for EA pretreatment operation in the biorefinery. The pretreatment unit is
composed by five systems of two reactors that load and unload (UL) intermittently. While
pretreatment reactors requires the utilization of five systems of two reactors, ammonia recycling
can be operated with two systems, each of them composed by a distillation column, an ammonia
evaporator, a compressor and a pump......................................................................................... 252
Figure 8-7 Residual plots and histograms for MESP prediction as a function of enzymatic
hydrolysis conditions (enzyme loading, solid loading and incubation time) for base case EA
pretreated corn stover. ................................................................................................................. 257

xxii

Figure 8-8 Optimization plots for defining enzymatic hydrolysis conditions that minimize MESP
($/gallon ethanol). % sugar conversions are also presented in the left side on the plots in blue
numbering. These values were calculated for base case EA pretreated corn stover................... 259

xxiii

CHAPTER 1 - INTRODUCTION

1.1. Project objectives
The increasing interest in exploring renewable sources of energy derives from several factors
which include increasing oil consumption, depleting oil reserves, environmental concerns due to
green-house gas (GHG) emissions, national security and energy independence issues. Among the
different renewable sources of energy, lignocellulosic biomass is considered to be the only
sustainable resource, capable of delivering liquid fuels and chemicals on a large scale [1].
Biochemical processing of lignocellulosic biomass to liquid fuels takes advantage of the fact that
the plant cell wall is composed of carbohydrate polymers (cellulose and hemicellulose), which
can be hydrolyzed by cellulolytic enzymes and fermented by microorganisms to produce fuel
molecules (e.g. ethanol, butanol). Since biomass is extremely recalcitrant to enzyme degradation,
efficient enzymatic hydrolysis cannot be obtained before a pretreatment step, which improves
accessibility of enzymes to the substrate. Ammonia Fiber Expansion (AFEX

TM

) is one the

leading pretreatment technologies available today, capable of improving enzymatic hydrolysis
yields from 3-10 folds, depending on the nature of the substrate [1]. This technology shows a
vast potential to solve major bottlenecks in the biorefinery. However, current biorefinery
scenarios, including AFEX

TM

-based scenarios, still have important issues that must be addressed

before turning the bioeconomy into a reality. For example, cellulolytic enzymes are expensive
and their price contributes in large extent to the operation cost of the biorefinery, which can only
be solved by decreasing enzyme utilization. One major problem observed during enzymatic
hydrolysis is related to the fact that industrially relevant conditions require cellulose loadings
1

higher than 6% (w/w), allowing the production of high concentration of fermentable sugars
which can translate to high concentrations of ethanol (>40g/L) after fermentation. Under these
conditions, sugar yields are severely reduced comparing to low cellulose loadings (e.g. 1%
(w/w)), possibly due to decreased water activity and lower hydration of the cellulose surface [2,
3]. As a consequence of this issue, a higher enzyme loading is required to effectively hydrolyze
lignocellulosic biomass at industrially relevant conditions. Solving this problem requires either
engineering the substrate during the pretreatment step, or engineering the enzymes to work
effectively under those conditions. In our perspective, little technological advances have been
observed in the pretreatment area, where relevant contributions to solve these important
bottlenecks remain scarce. In realistic terms, the properties of lignocellulosic biomass can be
manipulated and engineered by the pretreatment to improve downstream processing, allowing
existing microorganisms and enzymes to perform efficiently at industrially relevant conditions.
From this point of view, this work attempts to understand and be able to control key events
occurring during ammonia pretreatment of corn stover. Moreover, it proposes to modify the
current AFEX

TM

pretreatment methodology in order to maximize the potential of ammonia as a

pretreatment chemical and therefore be able to:

1) Reduce enzyme loading from ~20mg/g glucan to < 10mg/g glucan without highly affecting
enzymatic hydrolysis yields;

2) Improve enzymatic hydrolysis yields at higher solid loadings by minimizing the negative
impact of solids on enzymatic hydrolysis yields;
2

4) Create conditions to valorize lignin streams derived from pretreated biomass.

For these objectives, the creation of a novel pretreatment technology is proposed herein, where
naturally occurring cellulose I is modified to a more hydrophilic cellulose III, thereby improving
enzymatic hydrolysis rates at low and high solid loading. Also, due to the inhibitory role of
lignin toward enzymes and microbes, we propose to use this novel pretreatment technology to
extract lignin from the plant cell wall, thereby reducing enzyme and microbial inhibition during
downstream processing. This novel technology, denominated by Extractive Ammonia (EA)
pretreatment is proposed herein for the first time and consists of contacting lignocellulosic
biomass with liquid ammonia, at high liquid-to-solid ratio and relatively mild temperatures
(~120 °C) for short residence times. The effect of this treatment on the plant cell wall chemistry
will be determined by measuring 1) the impact of treatment conditions in cellulose III
conversion, 2) the degree of ester bond cleavage on the plant cell wall through ammonolysis and
hydrolysis reactions and also by 3) determining the extent of lignin extractability during EA
pretreatment.
The effect of plant cell wall modifications proposed herein on downstream processing will be
also evaluated by studying high solid loading enzymatic hydrolysis. Empirical modeling will be
performed using statistical design of experiments to correlate sugar yields with enzymatic
hydrolysis conditions, notably solids loading, enzyme loading and incubation time. The extracted
lignin and its potential for valorization to fuels and chemicals will be also evaluated by a detailed
physico-chemical characterization.

3

State-of-the-art fermentation technology will be used to test the novel EA pretreated substrate
performance. The novel Rapid Bioconversion with Integrated Enzyme Recycling Technology
(RaBIT) process has demonstrated to be capable of saving an extra 30 % of enzyme due to its
enzyme recycling capability on AFEX

TM

pretreated corn stover (AFEX

TM

-CS) [4]. Therefore,

by integrating a more digestible substrate to the RaBIT technology, we expect that enzyme
loading could be further reduced to lower levels.
The objectives proposed here and their relevance will be supported by techno-economic
evaluation. Major economics information will include calculation of the minimum ethanol
selling price (MESP), using current downstream technology coupled with EA processes, based
on experimental data and empirical predictive models. The feedstock of interest in this study will
be corn stover in all cases.
Finally, the information obtained from these studies will be used to formulate appropriate
recommendations to improve EA pretreatment technology and allow it to play a major role in
future biorefineries.

4

CHAPTER 2 – LITERATURE REVIEW

2.1. Lignocellulosic Biomass
The main components of the plant cell wall are cellulose, hemicellulose, lignin, protein and ash.
The ratio between these components varies among families of plants, like grasses, softwoods and
hardwoods (Table 2-1).

Table 2-1 Composition of lignocellulosic materials in cellulose, hemicellulose, lignin and ash
[5-7].
Lignocellulosic Materials Cellulose (%) Hemicellulose (%) Lignin (%) Ash (%)
Hardwoods

40-55

24-40

18-25

<1

Softwoods

45-50

25-35

25-35

<1

Grasses

25-40

35-50

10-30

1-12

These plant cell wall components form a complex interlinked structure that can provide both
support and a defense mechanism against invading microbes [6], [5], [8]. This is one of the
reasons why plant cell walls are highly recalcitrant towards fungal and bacterial enzymes and
why a pretreatment is required to disrupt this complex cell wall architecture prior to enzymatic
hydrolysis [9], [1]. The chemical structure of these components defines their role in the plant cell
wall architecture, which will be discussed herein.
2.1.1. Cellulose
Cellulose is the main polysaccharide in the plant cell wall. Its structure is identical for all plants;
however woody biomasses have a higher cellulose content compared to grasses. The primary
structure of cellulose is a linear (i.e., non-branched) chain consisting of -(1,4)-linked D-glucose
5

residues. Several of these chains are assembled in parallel through hydrogen bonding to form a
secondary structure, called cellulose microfibril [10]. In plants, these microfibrils are composed
by a more abundant fraction of ordered (crystalline) cellulose and a smaller fraction of less
ordered (amorphous) cellulose [11]. Crystalline cellulose occurs naturally in two different
polymorphs denominated by cellulose I and I [12]. These crystalline structures are organized
in parallel chains connected by hydrogen bonds that allow the formation of cellulose sheets. In
cellulose I and I there is no strong hydrogen bonding between the cellulose sheets. The only
difference between these two polymorphs of cellulose I is the relative displacement of the sheets
in the chain direction, resulting in triclinic and monoclinic unit cells for cellulose I and I ,
respectively. The only cellulose allomorph present in plants is cellulose I, while cellulose I can
be found in cell walls of bacteria and algae. In some organisms, notably in Valonia, is interesting
to see that cellulose is consistently composed of an eight-chain unit cell, accommodating both
native cellulose polymorphs, i.e. cellulose I and I [13].

6

Figure 2-1 Graphical representation of the structure of cellulose microbibrils. Cellulose
microfibrils (green) are composed of chains of -(1,4)-linked D-glucose subunits.

2.1.2. Hemicelluloses
Hemicelluloses are the second major polysaccharides found in plant cell walls and are mainly
composed of aldopentoses (i.e., arabinose and xylose) and play an important role in cross-linking
cellulose microfibrils. In grasses, the most abundant hemicelluloses are xyloglucans and
arabinoxylans.
Xyloglucan has a backbone similar to cellulose, decorated with xylose side-residues linked to
every 3 of 4 glucose monomers (Figure 2-2). This polymer is thought to either bind
spontaneously to adjacent cellulose microfibrils, maintaining their link, or bind covalently to
pectin polysaccharides [10]. Arabinoxylans have a -(1,4)-D-xylan backbone, branched with
arabinose residues. Glucuronic and ferulic acid branches are also present in the arabinoxylan
7

structure. Ferulic acid residues are responsible for cross-linking arabinoxylans to lignin. There
are between nine and ten ferulic acid ester-ether bridges every 100 C6-C3 lignin monomers [14],
[15].

Figure 2-2 Graphical representations of the major hemicellulose structures in grasses, i.e.,
Xyloglucan (left) and Arabinoxylan (right). Xyloglucan is composed by a glucan backbone
decorated by xylose residues (circled in red, Xyl). Arabinoxylan is composed by a xylan
backbone decorated with arabinose residues (circled in green, Arb) and glucuronic acid residues
(circled in blue).

In softwoods the major hemicelluloses are galactoglucomannan, glucomannan and arabinoxylan.
From these types galactoglucomannan is the most important in softwoods and it is composed of a
-(1,4)-linked

D-mannopyranose

and

D-glucopyranose

backbone

decorated

with

D-

galactopyranose residues, which are linked as single-unit side chains by -(1,6) bonds [14]. The

8

hydroxyl groups in the C2 and C3 positions of the backbone residues are partially substituted by
O-acetyl groups. On an average, is possible find one O-acetyl group per 3-4 hexose units.
In hardwoods the main hemicellulose component is O-acetyl-4-O-methyilglucurono--D-xylan.
As in softwoods and grasses, the backbone consists of β-(1,4)-D-xylan units, however in
hardwoods most of the hydroxyl groups at the carbons C2 and/or C3 are substituted by acetyl
groups. In addition, xylose units are substituted with -(1-2)-linked 4-O-methylglucuronic acid
residues, normally at every tenth xylose unit [16]. Unlike in softwoods and grasses, hardwood
xylans do not contain arabinose in their side chains.
2.1.3. Pectins
Pectins are a complex and heterogeneous group of polysaccharides with different domains that
are believed to be covalently linked together. The more important pectin domains include
rhamnogalacturan I and II, galactans and arabinans (Figure 2-3). Rhamnogalacturan I is a
polymer constituted by alternating residues of galacturonic acid and rhamnose, while
rhamnogalacturan II is constituted by a complex structure of eleven different sugar residues,
linked to a backbone chain of galacturonic acid molecules that is called homogalacturonan when
no side chains are attached. On the other hand, when side branches of xylose are covalently
bound to this chain, it is called xylogalacturonan.
Arabinans are branched polymers constituted by arabinose, that can be linked to a backbone of β(1,4)-D-galactan creating a new structure called arabinogalactan. The neutral arabinans and
arabinogalactans can be linked to the acidic pectins, promoting flexibility to the cell wall [10,
16].

9

Figure 2-3 Representation of the major pectin domains present in plants.

10

2.1.4. Lignins
In angiosperms, the lignin polymer is produced by dehydrogenative polymerization of three
different cinnamyl alcohols (p-coumaryl, coniferyl, and sinapyl alcohol) that differ in the degree
of methoxylation at the C3 and C5 positions of the aromatic ring (Figure 2-4).






Figure 2-4 Representation of the general precursor of lignin biosynthesis in angiosperms. (pcoumaryl alcohol: R = R’=H ; coniferyl alcohol: R = H, R’= OCH3 ; and sinapyl alcohol: R = R’
= OCH3).

When these alcohols are transformed in lignin polymers, are called p-hydroxyphenyl (H),
guaiacyl (G) and syringyl (S) units of the polymer, respectively. In addition to these three main
components, lignin polymers incorporate compounds that derive from incomplete biosynthesis of
components H, G and S, as well as other phenylpropanoid units such as acetates,
hydroxycinnamyl aldehydes and others [17].
A variety of chemical linkages, including ether and carbon-carbon bonds, connects the units of a
complex network of different compounds. This polymer complexity and heterogeneity depends
on the relative proportions of the three main monolignol units (H, G and S). Lignin from grasses
incorporates G and S units at comparable levels and more H units than dicots. In the case of
softwoods, lignin is essentially made of G units while in hardwoods is made of G and S units.
11

Lignin rich in G units has relatively more carbon-carbon bonds than lignin rich in S units. For
this reason, grasses and softwoods are less susceptible to Kraft delignification than hardwoods.
Kraft delignification is a process used by the paper industry during the pulping process, breaking
down the non-condensed ether -O-4-linkages in lignin using alkali agents (e.g. NaOH). The
carbon-carbon bonds are more resistant to chemical degradation, so they are not affected during
this process. In plants, lignification of the cell wall is an important defense mechanism to various
pathogens, notably fungi and bacteria, and helps provide the required strength to plant tissues.
Cell walls of the various plant organs and tissues contain different lignin contents as well as
different ratios of the three monolignols referred above [18].
2.2. Lignocellulosic Biorefinery
As a result of the organization and interaction between these polymeric structures, the plant cell
wall is naturally recalcitrant to biological degradation [19]. In the context of biological
conversion of lignocellulosic biomass to ethanol (Figure 2-5), a pretreatment step before
enzymatic hydrolysis is required to facilitate the access of the enzymes to its substrates,
enhancing the rate of hydrolysis by 3–10-fold, depending on the type of pretreatment and nature
of the substrate. Several economic models acknowledge that pretreatment is a major unit
operation in a lignocellulosic biorefinery, accounting for 16–19% of its total capital investment
[20] and being the second largest expense after the power plant generator [21]. For this reason it
is important to mature existing technologies and develop new methodologies to decrease the
economic impact of pretreatment on the overall biorefinery system. Enzymatic hydrolysis is
another important unit operation, where the polymeric carbohydrates are simplified into
fermentable sugars by a group of enzymes containing a gamut of activities that synergistically
12

hydrolyze cellulose and hemicelluloses [22]. In this process, achieving high sugar yields (>80%)
at high solid loadings (> 20%) using enzyme additions below 10mg/g glucan is one of the major
goals towards a commercially viable biorefinery. To achieve such goal, enzyme companies such
as Novozymes (Bagsvaerd, Denmark) and DuPont (Wilmington, DE, USA) have been working
on the development of cellulase and hemicellulase cocktails, with improved enzyme properties,
notably improved specific activity and thermal stability. In parallel, the development of
pretreatment technologies that perform critical structural modifications of the plant cell wall,
including reduction of crystallinity of cellulose, modification of cellulose crystalline structure,
cleavage of hemicellulose side chains and lignin-carbohydrate-complex (LCC) bonds, will also
contribute to higher enzyme activity and further reduction of enzyme loading [23].

Figure 2-5 Process flow diagram representing the unit operations for biological conversion of
lignocellulosic biomass into ethanol [1].

Enzymatic hydrolysis and fermentation can be performed in the biorefinery in two separate unit
operations, typically denominated by Separate Hydrolysis and Fermentation (SHF).
Alternatively, these two unit operations can be combined in one process denominated by
13

Simultaneous Saccharification and Fermentation (SSF). Finally, another processing possibility
involves microorganisms than can produce and secrete cellulolytic enzymes, which are able to
hydrolyze the pretreated biomass to fermentable sugars. These sugars are further converted to
fuel molecules by the same organisms in a process denominated by Consolidated Bioprocessing
(CBP).

Process
Unit Operations

SHF SSF CBP

Enzyme Production

Tank
1

Enzymatic Hydrolysis

Tank
2

Tank
1

Tank
1
Tank
2

Fermentation

Tank
3

Figure 2-6 Description of the various biorefinery process configurations and the number of tanks
required for each configuration.

These three different processes present advantages and disadvantages. In one hand, SHF allows
the operation of enzymatic hydrolysis and fermentation at optimal conditions for enzymes and
microbes. While enzymatic hydrolysis occurs typically at 50C, efficient microbial fermentation
occurs between 30C and 40C, depending on the organism. The optimal pH of both processes is
14

often slightly different, as fungal enzymes usually operate at pH between 4.5 and 5, while yeast
fermentation is usually efficient at pH 5.5. One major drawback of SHF is that the accumulation
of soluble sugars above certain concentrations inhibits cellulases and hemicellulases, thereby
decreasing the hydrolysis rate and final product yields. The SSF process does not allow the
concentration of sugars to accumulate during enzymatic hydrolysis since microorganisms
simultaneously consume these sugars as they are produced. Also, the SSF process is performed
in one single vessel, which has some economic impact due to capital investment reduction.
However, SSF processes operate under non-optimal conditions for both enzymatic hydrolysis
and fermentation, which usually need to be fine-tuned for maximizing ethanol yields. Also, the
ethanol produced by microorganisms is inhibitory to enzymes and therefore, some of the
advantages gained by eliminating the sugar inhibition may be lost by ethanol inhibition [24].
While SHF and SSF require external enzyme production and supplementation to the process,
CBP processes eliminate the need for dedicated vessels to produce enzymes. However, CBP
microorganisms are often susceptible to plant- and pretreatment-derived inhibitors that impact
the rate and yields of fuel production [25]. Developing robust CBP organisms and eliminating
possible plant-derived inhibitors from the fermentation media are two possible strategies that are
being explored to enable this type of process in a future biorefinery [25].
In most biorefinery models, the biomass components that do not participate in fuel production
via fermentation are subjected to waste treatment, of which results a solid stream that is used to
feed the power plant generator where electricity and heat are produced for the biorefinery.
However, one approach to possibly improve the economics of lignocellulosic biofuels is to
provide value addition to this waste stream. An obvious component of this solid waste stream
that has good potential for value addition is lignin. Lignin is the major renewable source of
15

phenolic compounds in nature and has multiple applications in the areas of biofuels, chemical
precursors, polymers and resins [26-28]. Electricity and heat production is a rather low-end
application for lignin and may not be the most economical solution for the biorefinery. The
lignin can be isolated from the plant cell wall in the front of the biorefinery during the
pretreatment process or downstream after fermentation and waste residue treatment (Figure 2-5).
Lignin removal before enzymatic hydrolysis provides several process advantages, as the
presence of lignin inhibits cellulolytic enzymes and microbes [29, 30]. Moreover, as lignin is not
a substrate for enzymes and microbes, its presence occupies non-productive volume in the
reactors, which represents an increase in capital investment.
One other form to reduce enzyme cost in the biorefinery is proposed by several authors and
consists on the implementation of enzyme recycling techniques to partially recover the added
enzymes for the following process cycle [31, 32]. This technique, associated with novel
bioprocessing methods such as Fast-SSF or Fast-SHF (also known as RaBIT processes), can
reduce enzyme loading up to 30% and presents improvements in biofuel productivity [4].
2.3. Importance of Pretreatment in the Lignocellulosic Biorefinery Context
In plants, the cell wall acts naturally as a physical protection against pathogens [33]. Some of
these pathogens carry hydrolytic enzymes that target different plant cell wall components
promoting the disruption of the fibers [33]. From the evolutionary perspective, the plant cell wall
structure developed naturally to avoid pathogen penetration. Substrate-related factors affecting
enzymatic digestibility are normally associated on a macroscale with cell wall porosity [34] and
biomass particle size [35]. On a microscale, these factors are associated with cellulose
crystallinity and degree of polymerization [36], hemicellulose side-chain branching [37], lignin
16

composition [18], and coumarate and ferulate crosslinking [18, 38] . For an efficient biological
conversion of cellulose into fermentable sugars, it is important to modify the physical and
chemical properties of the plant cell wall, which can be done through a pretreatment step.
Different pretreatments have varying chemistries and ways of interacting with the plant cell wall
components. Also, there is an extensive variety of lignocellulosic materials, ranging from
grasses, softwoods, and hardwoods, which also have different physical and chemical properties.
For this reason there is interdependence between pretreatment, the type of substrate, and the way
it is eventually processed.
Moreover, it is evident that the choice of pretreatment will also impact the physicochemical
properties of the pretreated biomass. These properties will also profoundly affect downstream
processes such as enzyme and microbial selection, preconditioning, by-product utilization, waste
residue handling, and ethanol recovery (Figure 2-7). Consequently, the choice of pretreatment
will have an economic impact on each one of the biorefinery stages. These aspects make
pretreatment the central unit operation in a biologically based sugar-platform for processing
lignocellulosic biomass into fuels and chemicals.

Figure 2-7 Pretreatment choice affects all biorefinery related operations. This fact gives
pretreatment a central role affecting the design and economic viability of a lignocellulosic
biorefinery [1].
17

2.4. Important Pretreatment Technologies Available Today
Several pretreatment technologies are available today, which can be classified, based on the
nature of the cell wall disruption, into four categories: 1) physical; second; 2) fractionation; 3)
chemical; and 4) biological pretreatment.
2.4.1. Physical Pretreatment
Physical pretreatments operate based on the principle of particle size reduction by mechanical
stress. This can be obtained by dry, wet vibratory, and compression based ball milling
procedures [39-41], increasing the enzyme performance by improving the surface area to volume
ratio and in some cases by reducing degree of polymerization and crystallinity of cellulose [36].
Although physical pretreatments are not sufficient to dramatically increase sugar conversions,
most other pretreatments require a minimal particle size reduction in order to be effective,
especially to overcome mass and heat transport problems. It is also important to notice that
beyond a certain particle size this type of pretreatment becomes economically unfeasible due to
the extensive energy requirements associated with the milling process.
2.4.2. Solvent Fractionation
This category of pretreatment technologies applies the principle of differential solubilization and
partitioning of various components of the plant cell wall, including cellulose, by the disruption of
the hydrogen bonding between microfibrils [42]. Several fractionation methodologies are
available today; however, the most attractive ones are the organosolv process, phosphoric acid
fractionation, and ionic liquids based fractionation.
Organosolv process uses organic solvents, usually alcohols in the presence of an acidic catalyst
to extract lignin from lignocellulosic biomass [43]. In this pretreatment, operating temperatures
18

range between 90 and 120°C for grasses and 155 and 220°C for woods, with a residence time
ranging from 25 to 100 min. Depending on the substrate, catalyst concentration can vary from
0.83 to 1.67% and alcohol concentration from 25 up to 74% (v/v). Lignol Innovation
Corporation, Vancouver, Canada, is investing in the organosolv process using ethanol as a
preferred solvent. In their process, the insoluble cellulose fraction is used to produce fermentable
sugars via enzymatic hydrolysis, while the liquid phase containing lignin, furfural, xylose, acetic
acid, and lipophilic extracts is processed to recover these compounds [44]. Also,
oligosaccharides are subjected to a mild acid hydrolysis to convert them into fermentable sugars
to produce more ethanol. The solvent is recovered by distillation and reused while the lignin is
used as an additive binder for other applications.
In phosphoric acid fractionation, a series of different solvent extractions using phosphoric acid,
acetone, and water at temperatures around 50°C are used to fractionate the plant cell wall into
amorphous cellulose, hemicellulose, lignin, and acetic acid. The fractionation is possible owing
to the preferential solubility of various plant cell wall components in different solvents. Using
this methodology, it is possible to decrystallize cellulose fibers, removing at the same time most
of the hemicelluloses and lignin [45]. This pretreatment has several advantages, since it can
operate at a low temperature and pressure (atmospheric pressure) giving at the same time high
yields of glucose and xylose [46]. However, the challenges facing this method have to do with
finding economical ways of recovering and purifying phosphoric acid (and other solvents) for
recycle at the end of each pretreatment batch.
Only recently ionic liquids were first applied as a pretreatment technology to convert
lignocellulosic biomass into fermentable sugars [47]. Ionic liquids have the capability to form
hydrogen bonds with cellulose at high temperatures, because of the presence of anions like
19

chloride, formate, acetate, or alkyl phosphonate. This pretreatment uses ionic liquids in a ratio of
1:10 (biomass:ionic liquid, w/w) and temperatures ranging from 100 to 150°C. The soluble
biomass is regenerated using antisolvents such as water, methanol and ethanol, and then
subjected to enzymatic hydrolysis to produce fermentable sugars. One important factor is the fact
that residual ionic liquids remaining in the biomass after regeneration could interfere with
hydrolytic enzyme activities (and downstream fermentation steps), affecting the final sugar and
biofuel yields [48]. After regeneration, ionic liquids may be recovered from antisolvents by flash
distillation and reused in the process [49]. Ionic liquids have tremendous potential for pretreating
lignocellulosic biomass, producing a substrate that can achieve more than 90% cellulose
digestibility [50]; however, the cost of these chemicals must be significantly reduced or their
recovery made highly efficient, in order to make this technology a serious candidate for an
industrially based biorefinery process.
2.4.3. Chemical Pretreatment
Most of the leading pretreatment technologies are represented in this category, which comprises
acidic, alkaline and oxidative based pretreatments. Optimal operational conditions for some
leading chemical pretreatments to maximize enzymatic digestibility of corn stover are
summarized in Table 2-2. In this category, most pretreatments differ in the types of chemistries
and mechanisms that are responsible for cell wall ultrastructural and chemical modifications
(Table 2-3) that result in improved enzyme accessibility and higher sugar yields.
2.4.3.1. Acidic pretreatments
Several acidic pretreatment technologies are available, such as concentrated and dilute acid,
steam explosion, and liquid hot water pretreatment. Although the chemistry governing these
20

pretreatments is quite similar, they operate using different methodologies. While dilute acid uses
an externally added acid catalyst, steam explosion, and liquid hot water pretreatments operate in
acidic conditions due to the fact that water behaves like an acid at high temperatures and also due
to liberation of acetic acid from hemicellulose under these conditions [51]. The acid is
responsible for chemical hydrolysis, targeting mainly hemicelluloses and lignin, solubilizing a
great part of these components from the plant cell wall structure and improving the enzyme
accessibility to cellulose. However, during this process several degradation products are formed,
mainly furfural, 5-hydroxymethylfurfural, phenolic acids and aldehydes, levulinic acid, and other
aliphatic acids which can inhibit both enzymatic hydrolysis and fermentation [52, 53].
Dilute-acid pretreatments typically use sulfuric acid as a catalyst for hemicellulose and lignin

solubilization at low acid concentrations (0.05–5%), and temperatures ranging from 160 to
220°C [54, 55], in order to minimize the formation of degradation products. Naturally,
minimizing the sugar degradation contributes to higher sugar yields at the end of the process.
However, this methodology requires extensive washing of the cellulose-rich slurry fraction after
pretreatment and/or detoxification of the hydrolyzates before fermentation [53, 56]. One
advantage of this pretreatment is that hemicellulose hydrolytic enzymes (e.g. xylanases) need not
be added during enzymatic hydrolysis, which can be considered a definite cost saving feature.
However, most microbes have been unable to ferment xylose from the pretreatment liquor
without substantial detoxification [56].
Steam explosion has been one of the most widely implemented pretreatment technologies, being
successfully applied to several types of lignocellulosic biomass (e.g. softwoods, hardwoods, and
agricultural residues). This pretreatment has been demonstrated in several variants, with and
without the presence of several chemical catalysts (e.g. sulfuric acid, sulfur dioxide, sodium
21

hydroxide, and ammonia) before steam explosion [57-61]. It operates at high temperature (160–
290°C) and pressure for a certain duration of time (ranging from a few seconds to several
minutes) before the pressure is explosively released [57-60].
Liquid hot water pretreatment uses water at high temperatures (160–230°C) and pressures
(>5 MPa), in order to maintain water in the liquid state [51]. Several methodologies have been
developed to promote an effective contact between the biomass and the liquid water, varying
from co-current, counter-current, and flow-through [62]. Recent variants of this pretreatment
allow better pH control between 4 and 7, minimizing the nonspecific degradation of
polysaccharides [63].
In addition, several other mineral and organic acids (nitric, carbonic, succinic, fumaric, maleic,
and citric acid) have been used successfully as pretreatment catalysts [64-66].
2.4.3.2.Alkaline pretreatments
This kind of pretreatment uses alkaline catalysts, such as calcium oxide (lime), ammonia, and
sodium hydroxide. Most of these chemicals specifically target hemicellulose acetyl groups and
lignin–carbohydrate ester linkages [75, 76]. These reactions help solubilize and extract lignin
from the biomass, reducing nonspecific binding during enzymatic hydrolysis [71]. However, cell
wall chemical and ultrastructural modifications still need to be understood for most alkaline
pretreatments in order to develop proper enzyme mixtures that can effectively hydrolyze both
cellulose and hemicellulose.

22

Table 2-2 Optimum operating conditions of some of the leading pretreatment technologies that maximize enzymatic
digestibility of pretreated corn stover [67-70].
Reaction
Pretreatment Temp.
Chemical
Chemical loading
Water loading
time
Specific notes
type
(°C)
used
(g/g dry biomass) (g/g dry biomass)
(min)
Sulfuric
Dilute acid
160
20
0.015
3
Batch process
acid
Sulfur dioxide
Sulfur
Soaked overnight in 3% acid
Steam
190
5
0.03
4
dioxide
solution before pretreatment
Explosion
Controlled pH
liquid
hot 190
15
5.25
Flow-through mode
water
Liquid ammonia added to moist
AFEX
90
5
Ammonia
1
0.6
biomass before heating reactor
Flow-through mode using 5
ARP
170
10
Ammonia
0.5
2.8
ml/min of ammoniacal solution,
15% (w/w)
Calcium
Lime
55
4 weeks
0.5
10
Purged with air
hydroxide

23

Table 2-3 Effect of leading thermochemical pretreatments on the physicochemical properties of corn stover and respective
glucan and xylan conversion after enzymatic hydrolysis.
%
%
%
% acetyl
%
%
Pretreatment
Cellulose
Cellulose
cellulose
hemicellulose
lignin
groups
glucan
xylan
type
CrI
DP
removed
removed
removed
removed
conversion conversion
Dilute acid
53
5-10
70-75
18
2700
55
92
93
Sulfur dioxide
a
a
ND
3-5
40
40-45
3000
55
87
78
steam explosion
Controlled pH
45
5-10
40
ND
5600
55
91
81
liquid hot water
AFEX
36
0
0
0
6600
30-35
96
91
ARP
26
1-5
50-60
75-85
4600
85-90
90
88
Lime
56
1-3
30-35
55-60
3200
90-95
94
76
Enzymatic hydrolysis was conducted using 15 FPU/g glucan cellulase loading (Spezyme CP, Genencor). Glucan and xylan
conversions were calculated relative to the initial composition of corn stover, including both monomeric and oligomeric sugars. The
cellulose crystallinity index for untreated corn stover is 50 units and the degree of polymerization is 7000 [54, 63, 67, 69, 71-74]. ND:
not determined; CrI: Crystallinity Index; DP: Degree of Polymerization.
a
Enzymatic hydrolysis was performed using Celluclast 1.5 L at 65 FPU/g mixture and 17 beta-glucosidase IU/g mixture.

24

TM

Ammonia fiber expansion (AFEX

) is a lower temperature process (60–140°C) where

concentrated ammonia is used as a catalyst (0.3–2 kg ammonia per kg of dry weight biomass).
Ammonia is added to prewetted biomass (0.6–2 kg moisture per kg of dry weight biomass) in a
high-pressure reactor and is cooked for 5–45 min before the pressure is rapidly released. The
volatility of ammonia allows it to be recovered and reused, leaving the dried biomass ready for
enzymatic hydrolysis [77]. Because ammonolysis reactions do not produce major inhibitors (e.g.
organic acids) for enzymes and microbes, it is possible to ferment and hydrolyze this substrate
without any sort of detoxification [78, 79]. However, AFEX

TM

treated biomass requires the

utilization of hemicellulose hydrolytic enzymes in addition to cellulases to produce fermentable
sugars. X-ray diffraction studies have also revealed that ammonia decrystallizes cellulose [80],
facilitating enzyme activity on this substrate.
Other ammonia based methodologies (e.g. Ammonia Recycle Percolation (ARP), Soaking in
Aqueous Ammonia (SAA), Supercritical Ammonia and Ammonia-Hydrogen Peroxide
pretreatments[81, 82] are available, differing mostly in the thermodynamic state of ammonia–
water mixtures, and ammonia concentrations. For example, ARP is carried out in a flow-through,
recycle mode by percolating ammoniacal solutions (5–15% concentration) through a packed bed
reactor with the biomass under pressure and high temperatures. During this process,
hemicellulose and lignin are removed from the biomass in the liquid phase.
2.4.3.3.Oxidative pretreatments
Oxidizing agents can also be used to remove lignin and hemicellulose from biomass, in order to
increase enzymatic digestibility of cellulose [83]. These agents can react selectively with lignin
aromatics and alkyl/aryl ether linkages or also target hemicellulose and cellulose, which will
25

contribute negatively to the final sugar hydrolysis yields. Several kinds of degradation products
may be formed (primarily aliphatic aldehydes and aliphatic organic acids) that have been found
to inhibit enzymatic hydrolysis of cellulose.
Alkaline wet oxidation is a form of oxidative pretreatment which operates under alkaline
conditions and high temperatures (170–220°C) using pressurized air/oxygen or hydrogen
peroxide as oxidant [84, 85]. The alkali in this pretreatment is sodium carbonate, which helps
solubilize hemicelluloses [84] and also minimizes the formation of furan based degradation
products that could inhibit enzymes.
2.4.4. Biological pretreatment
Biological pretreatments, as opposed to chemical based pretreatments, are non-energy intensive
processes [86], normally performed using fungi (e.g. white rot basidiomycetes) and certain
actinomycetes. During pretreatment, these organisms secrete extracellular enzymes like lignin
peroxidases and laccases that help remove a considerable amount of lignin from the biomass
[87]. However, biological processes require longer residence time (several hours to few days)
[88], which constitutes a serious limitation for its application in large-scale biorefinery related
operations. Also, most microbes typically consume some part of the carbohydrates available in
the biomass during growth, which will negatively impact the sugar yield at the end of the
process. Important advantages of this pretreatment may be more evident when less recalcitrant
genetically modified plant materials are available with higher carbohydrate content, requiring
milder pretreatments.

26

2.5. Considerations for an ideal pretreatment technology
The evaluation of a pretreatment should comprise not only technological factors, but also
economical and environmental criteria. Feedstock cost represents the highest variable cost in a
biorefinery [20]. For this reason, sugar yield contributes in a large extent to the overall
economical feasibility of the biorefinery. Also, sugar concentration is an important factor
contributing to both economical and environmental aspects of the biorefinery (e.g. volume of
reactors, size of product recovery operations, energy inputs, etc.). At the same time
environmental impacts owing to land use management factors [89] and waste residue production
would also affect the biorefinery economics. In the pretreatment point of view, energy
requirements are associated with chemical and water utilization, temperature, pressure, mixing
and chemical recovery during its operation. These parameters have economical and
environmental impacts; however, it is important not to place boundaries around the pretreatment
step only, because the pretreatment method will impact all other unit operations in a biorefinery
[90] (Figure 2-5). For example, one some pretreatments require preconditioning of hydrolyzates
because of inhibitory compounds produced during pretreatment [53], contributing to additional
costs and solid-waste management concerns. The type of chemicals utilized is also a criterion to
take into consideration. Besides the reactor volume, the construction material is also going to
affect the total capital investment for pretreatment. A less corrosive chemical with low toxicity
would contribute for reducing this cost as well as increasing safety and environmental benefits.
Lignin recovery is also an important criterion; because this component of the plant cell wall can
be used as a precursor to manufacture valuable chemicals [44]. This area of research is rapidly
evolving and could be considered a valuable option in the future, contributing to the economical
feasibility of a biorefinery. Moreover, some pretreatments have the potential to produce dry
27

pretreated substrates, which could benefit distributed processing facilities because of lower
transportation costs. Currently, biomass transportation and storage is considered a technical,
economical, and environmental concern [91] and should be addressed before choosing a certain
pretreatment technology for large-scale implementation. Some of these pretreatment
technologies (e.g. AFEX

TM

) also have the potential to process biomass for animal feed [92],

increasing its digestibility and reducing the feedstock requirements for animal production [91].
This application can create conditions for sustainable land use management, both for bioenergy
and feed production, promoting at the same time the desired environmental benefits of
lignocellulosic biofuels.
2.6. Ammonia and Alkaline Pretreatment Chemistry
2.6.1. Reactions involving hydroxyl ions and lignocellulosic biomass.
2.6.1.1. Reactions with polysaccharides.
Initial reactions consist of the solvation of hydroxyl groups by hydroxyl ions, causing the
biomass to swell. At higher temperatures, there are many chemical reactions that can occur. The
most important ones are considered [93]: (i) Dissolution of un-degraded polysaccharides; (ii)
Peeling of end-groups and the formation of alkali stable end-groups; (iii) Alkaline hydrolysis of
glucosidic bonds and acetyl groups and (iv) Degradation and decomposition of dissolved
polysaccharides and peeled monosaccharides.
Loss of polysaccharides and decrease in the degree of polymerization during alkali treatments
are in large extent due to peeling and hydrolytic reactions. These are the most important
reactions to monitor in order to preserve the integrity of the polysaccharides. At temperatures
28

around 100ºC, the degradation of polysaccharide chains starts from the existing reducing end
groups, which is commonly called the primary peeling reaction. At temperatures around 150ºC,
polysacharide chains start to be degraded by alkaline hydrolysis and the new reducing ends
formed in this process will be also subject to endwise peeling reactions (also known as secondary
peeling) [94].
Peeling reactions of polysaccharides involve the elimination of the reducing end groups, forming
a number of different carboxylic acid compounds. Sjöström [94] proposed a mechanism for
endwise peeling where monomeric sugars resulting from peeling reactions can be tautomerized
to a dicarbonyl compound that can be rearranged to yield isosaccharinic acids. Other degradation
products that can result from these reactions include lactic acid, 2-hydroxybutanoic acid and 2,5dihydroxypentanoic acid. Some of these compounds can be found in AFEX

TM

pretreated corn

stover, however they exist in small amounts due to low reaction temperatures and residence
TM

times during AFEX

operation [95].

Hemicelluloses are in general much more vulnerable to these types of chemical reactions in
alkali media than cellulose. Moreover, is also possible to see differences among the various kinds
of hemicellulose components. For example, xylans are more stable and less vulnerable than
glucomanans and arabinans. Interestingly, early cleavage of arabinose side-groups in softwood
xylans has a stabilizing effect against alkaline peeling, since an alkali-stable metasaccharinic
acid end group is formed after the loss of the arabinose side group [93].
The reactions that occur during endwise peeling terminate when competing reactions start to take
place. These reactions are called stopping reactions and are very important in preventing the
disruption of the polysaccharide fibers and their degradation [93]. They are initiated by a β29

hydroxy elimination at the C2 position, producing a tautomeric intermediate that is converted to
an alkali stable metasaccharinic acid group or C2-methylglyceric acid [94].
Other small molecular weight acids are also formed in alkali conditions, including acetic acid
and formic acid. Formic acid is a product of peeling reactions of polysaccharides while acetic
acid is formed by cleavage of the acetyl side chain groups present in grasses and hardwoods
xylans. Glucuronic acid side groups of xylan are also hydrolyzed during intensive alkaline
pretreatment [93, 94]. All these compounds are found in AFEX

TM

pretreated corn stover

[23].Therefore, minimization of peeling reactions would be important to minimize sugar
decomposition as well as formation of highly inhibitory organic acids to enzymes and
microorganisms.
2.6.1.2. Reactions with Lignin
In any discussion about lignin degradation, the most important aspect is the varied behavior and
stability of the possible bonds that exist in the lignin polymer. The most labile bonds in alkaline
conditions are ester bonds and aryl-ether bonds, which are mostly cleaved during kraft pulping
treatment. Also, aryl-alkyl or alkyl-alkyl bonds are typically destroyed in alkaline medium, but to
a lesser extent than the aryl-ether bonds. Diaryl ether and C-C bonds are normally stable in these
conditions, meaning they usually remain unaltered during alkali treatment.
In hardwoods and softwoods, the most common linkages are of the types - and -aryl ether,
which means these are the most important linkages to cleave to promote lignin degradation. As
- and -aryl ether linkages are relatively easy to cleave, their cleavage is independent of the
hydroxyl ion concentration and is normally the main reaction that occurs in the initial phase of
the alkaline pulping processes [93]. During alkaline treatment of biomass, a great variety of
30

compounds can be formed from lignin. Most are aromatic acids, aromatic aldehydes and
phenolic compounds, which can have an inhibitory effect on enzymes and microorganisms.
Ester linkages are abundantly present in grasses (e.g. corn stover), participating in ferulate and
di-ferulate cross-linking with hemicellulose and lignin [96]. These linkages are easily hydrolyzed
by hydroxyl ions resulting in the formation of ferulic acid and di-ferulates. These components
can be found in AFEX

TM

pretreated corn stover [23, 95], as well as their amide counterparts,

which result from competing reactions involving ammonia, i.e. ammonolysis reactions.

Figure 2-8 Simplified diagram of proposed ferulate and diferulate cross linkages with lignin and
arabino-xylans. 1- Acetyl decoration in xylan backbone; 2- Ester linked ferulate to arabinoxylan; 3- Ester linked arabinose to lignin; 4- 5-5 ester linked ferulate dimer cross-linking
arabino-xyalan chains; 5 –Ether-linked ferulate to lignin.

31

2.6.2. Reactions involving ammonia and lignocellulosic biomass
2.6.2.1. Ammonolysis Reactions
In the special case of AFEX

TM

pretreatment of lignocellulosic biomass, ammonia not only

participates in the production of hydroxyl ions, which can drive a variety of reactions as
mentioned above, but also acts as a reactant itself. Amidation of organic compounds via
ammonolysis occurs during AFEX

TM

pretreatment, as was evident when considerable amounts

of acetamide and phenolic amides were found in water extracts of AFEX

TM

treated corn stover

[16, 95].
Ammonolysis reactions can occur in aqueous solutions or in the vapor-phase and target mainly
ester linkages present in the plant cell wall. As mentioned above, this type of bonds are abundant
in grasses and can be found in arabinoxylan acetyl groups, ferulate and di-ferulate cross-linking
of lignin and hemicellulose as well as p-coumarate residues which occupy terminal sites in lignin
polymers. Ammonolysis products of these interactions yield amides that can be found in
TM

AFEX

pretreated corn stover [23, 95, 97]. The correlation between enzymatic hydrolysis

yields and extent of cleavage of these ester bonds is evident from Figure 2-9, proving that these
chemical interactions are important to maximize in order to have efficient enzymatic activity on
TM

AFEX

substrates. Therefore, understanding ammonolysis and hydrolysis reactions of

structural ester bonds in the plant cell wall is extremely relevant for improving ammonia-based
pretreatments of lignocellulosic biomass.

32

Figure 2-9 Correlation between enzymatic digestibility of AFEX
total feruloyl cleavage (amide + acid) during AFEX

TM

TM

pretreated corn stover and

pretreatment [23].

H. E. French et al. [98] has studied the effect of aqueous ammonia solutions in a variety of esters
containing different leaving groups with different molecular weights. It was found that the ratio
of ammonolysis to hydrolysis decreases with increasing molecular weight of the alcohols
associated with the different esters. Moreover, the reactivity decreases by increasing the
molecular weight of the ester, especially in the case where the leaving group had side chains
close to the ester bond (e.g. isopropyl acetate). The justification for this fact may relate to
stereochemical factors during the chemical reaction and the accessibility of ammonium ions
compared with hydroxyl ions to access the carbonyl group. Gordon Maxwell et al. [99] described
the catalytic effect of some solvents during ammonolysis reactions, ranging from glycerol
(higher catalytic activity) to butanol, methanol and dioxane (lower catalytic activity). According
to the authors, these differences may relate to the polarization of the nitrogen-hydrogen bond,
33

increasing the reactivity of ammonia as a nucleophilic agent. Also, P.M Williamson et al.[100]
and L.L. Fellinger et al.[101] described catalytic effects during ammonolysis reactions caused by
electrolytes, such as sodium and ammonium salts. AFEX

TM

pretreatment utilizes aqueous

ammonia as a reactant medium, and therefore is possible to accelerate ammonolysis reactions by
introduction of more advantageous solvent systems, coupled with the presence of inorganic salts.
Understanding the effect of solvents and salts during pretreatment will help to better design
ammonia-based pretreatment methodologies. Ammonolytic reactions target not only ester bonds,
but also react with phenolic compounds, alkyl halogens and alcohols.
2.6.3. Maillard Reactions
Ammonia (or amino compounds) can react with polysaccharides generating the so-called
Maillard reaction products. Maillard chemistry is typically composed by an extremely complex
network of reactions. More than fifty reaction products have been identified after reacting
ammonia with D-glucose, which belong to a variety of families of compounds such as pyrazines,
imidazoles, ketones, aldehydes and amides [102]. Some of the Maillard products identified in
TM

AFEX

treated corn stover were listed by Chundawat et al. (2009) [23].

2.6.4. Rearrangement of cellulose crystalline state (Cellulose Iβ to IIII)
Cellulose Iβ is the predominant allomorph of cellulose present in plants. Thermochemical
pretreatments are known to affect cellulose crystallinity at different levels by i) transforming
cellulose Iβ into amorphous cellulose, ii) increasing cellulose crystallinity Iβ due to degradation
of the amorphous regions of cellulose or by iii) converting it to different allomorphs. It has been

34

reported in the literature that a mild decrystallization of cellulose occured in conventional
TM

AFEX

[38], however there is no evidence of formation of other cellulose allomorphs.

Anhydrous ammonia is known to modify cellulose Iβ crystal structure to cellulose IIII (Figure
2-10). This modification is defined as the rearrangement of intracrystalline hydrogen bonds and
stacking interactions between glucose units. X-Ray and neutron diffraction have elucidated the
formation of an intermediate complex (i.e. ammonia-cellulose I complex) before cellulose IIII
conversion, where ammonia molecules trapped inside the cellulose crystal form hydrogen bonds
with neighboring hydroxyl groups from cellulose [103, 104]. Cellulose IIII is formed once
ammonia is removed from the intermediate complex, allowing a rewiring of hydrogen bond
network and adjustments on the packing of cellulose glucan chains. Our recent work has
demonstrated that this crystalline allomorph of cellulose enhances depolymerization rate by 2 to
2.5 folds, improving synergistic effects between endocellulases and exocellulases, especially
when endocellulases of GH family 5 and 7 are involved [105] (Figure 2-10).
From the different pretreatment technologies available today, ionic liquid pretreatment and
phosphoric acid pretreatment produce cellulose II [106] and amorphous cellulose [107]
respectively. Both pretreatments show the highest improvements in enzymatic digestibility
among the pretreatment technologies available in the literature. However, high cost of the
chemicals used in these methods and catalyst recovery bottlenecks are still major problems to
solve before they can be economically viable.

35

Figure 2-10 i) Representation of cellulose Iβ and IIII crystal structure, including hydrogen
bonding network and glucan chain organization. ii) Synergistic effect of endocellulases and
exocellulases in cellulose Iβ (red) and IIII (black) hydrolysis [105].
From our recent work, we concluded that by forming cellulose IIII, using liquid anhydrous
ammonia, enzymatic hydrolysis yields improve above cellulose II result for commercial enzyme
36

mixtures (Spezyme CP and Novozyme 188 by Genencor and Novozymes, respectively).
However, amorphous cellulose is still the most digestible form of cellulose available (Figure
2-11).

Figure 2-11 Glucose yields after enzymatic hydrolysis of the different cellulose substrates.
Enzymatic hydrolysis was performed using 1.5 FPU Spezyme CP cellulase/g glucan and 6.4 pNPGU Novo188 β-glucosidase /g glucan. Left and right bars represent 6 h and 24h enzymatic
hydrolysis data [105].

From this set of results, it is clear that cellulose crystallinity is the ultimate barrier for efficient
biomass degradation to fermentable sugars. Therefore, is important to define strategies to modify
TM

AFEX

pretreatment to produce cellulose IIII in an efficient, economic viable fashion.

2.7. Major downstream processing bottlenecks in the biorefinery
2.7.1.“Solids effect” and enzyme loading
“Solids effect” is defined as the decrease in sugar yields caused by increasing the initial solids
loading of enzymatic hydrolysis. This phenomenon is one of the most important bottlenecks to
solve in the biorefinery, since one important goal is to produce high concentrations of
37

fermentable sugars so they can be fermented to high concentrations of product (e.g. ethanol).
This will decrease the level of water handling in the biorefinery and decrease the size and energy
spent in downstream unit operations (e.g. fermentors and distillation column size).
Most, if not all, leading pretreatment technologies are not able to solve the “solids effect”
problem [3]. Several authors have been studying possible reasons for this phenomenon, of which
the reduction of water activity by the presence of solids gained particular attention [2]. In Figure
2-12 ii) is possible to observe that higher T2 relaxation times (a measure of water activity) are
associated with improvement of glucose yields after 24h of enzymatic hydrolysis. Water activity
decreases with increasing solid loading (Figure 2-12 iii)) and, as a consequence, diffusivity
coefficients for proteins and for sugars released during enzymatic hydrolysis are negatively
affected, thereby decreasing sugar yields at higher solid loading [2].
Kristensen, J.B. et al [3] suggested decrease in cellulase binding at high solid loadings as one of
the factors that contribute for the “solids effect”. In fact, is possible to observe a linear
correlation between enzyme binding and enzymatic conversion with respect to initial solids
loading (Figure 2-13 i)). From our previous work, we found that enzyme binding is not so
important to convert cellulose III into monomeric glucose (Figure 2-13 ii)), since is possible to
improve enzymatic digestibility by 2-2.5 folds with lower enzyme bound to cellulose. This
finding, coupled with the fact that the surface of cellulose III is more hydrophilic than cellulose I,
brings us to the hypothesis that is possible to minimize the effect of solids by converting
cellulose Iβ to cellulose IIII.

38

Figure 2-12 i) Solids effect observed for different time points of enzymatic hydrolysis. ii) Effect
of T2 relaxation time of restricted bulk water (RBW) on enzymatic digestibility after 24h. iii)
Effect of solids content on T2 relaxation time of RBW [2].

For conventional AFEX

TM

pretreatment, typical enzyme loadings to convert 80% of the

cellulose into monomeric glucose from corn stover range 20-30 mg enzyme/g glucan, at high
solids loading. Optimal enzyme loading has been decreasing as novel commercial enzyme
cocktails have been evolving to efficiently convert cellulose into monomeric glucose. However,
39

to be competitive in the biofuels area is important to look at the level of enzyme loading required
to degrade starch into glucose in the corn grain ethanol industry. Currently, is possible to
efficiently hydrolyze starch using ~2 mg enzyme/g glucan in an industrial setup. Our previous
results show that by converting cellulose I to III, enzymatic hydrolysis rates can increase 2 to 2.5
fold [105]. If this holds true for higher solids loading and if cellulose III reduces the effect of
solids during enzymatic hydrolysis, it may be possible to decrease enzyme loading to the single
digit region while hypothetically maintaining enzymatic hydrolysis yields around 80%.

i)

ii)

Figure 2-13 i) Relationship between enzyme adsorption and initial solids content [3]. ii)
Maximum cellulase adsorption on cellulose I (red) and cellulose III (black) for three different
kinds of cellulases [105].

2.7.2.Enzyme and microbial inhibition
Even though AFEX

TM

pretreatment is known to produce highly digestible substrates and

fermentable hydrolyzates without detoxification procedures, it does produce decomposition
products that inhibit enzymes and microbes to certain extent [108].
40

Jin, M. et al. [109] has identified the sources of inhibition from AFEX

TM

corn stover

hydrolyzate and quantified the contribution of each one of them to the final decrease in
fermentation performance of S.cerevisiae 424A (LNH-ST) (Figure 2-14). From this work, is
possible to observe that AFEX

TM

decomposition products reduce cell growth by 22% and

specific xylose consumption rate by 13%. Therefore, is important to identify major components
that contribute to this inhibition and study mechanisms to control the production of major
components during pretreatment.
Inhibition factors tend to be even more critical for the case of CBP, where a single organism is
able to produce enzymes, digest lignocellulosic biomass and produce ethanol simultaneously.
Recent work has demonstrated that AFEX

TM

corn stover inhibits the enzymatic hydrolysis

performance during CBP when compared with pure cellulose (Avicel) [110].
For the reasons mentioned above, it is clear that the presence of lignin and other AFEX

TM

decomposition products are negatively affecting major downstream processing operations,
especially in the case of CBP fermentation. CBP methodology is considered to be a long term
solution for the lignocellulosic biofuels industry; however, pretreatment methodologies as well
as microorganisms must be perfected before this technology becomes economically viable.

41

TM

Figure 2-14 Contribution of AFEX
decomposition products, ethanol and fermentation
metabolites on xylose consumption rate and cell growth of S.cerevisiae 424A(LNH-ST) [109].

Lignin has also been known to act as a natural inhibitor to both microbes and enzymes. The
recalcitrant nature of the lignin molecule has been largely explained by effects such as its role on
increasing non-productive binding of cellulases, restriction of enzyme accessibility to the
carbohydrates and inhibitory effects caused by the water soluble lignin-derived compounds [111113]. Also, authors have demonstrated that the inhibitory role of lignin is largely dependent on
the nature of the molecules that compose the lignin polymer. For example, phenolic hydrolyxyl
groups were found to be critical to the inhibitory effect of lignin to cellulolytic enzymes by Pan
et al. (2008) [30]. Therefore, reducing the effect of lignin inhibition, ideally through an
extraction process during pretreatment, would possibly improve downstream operations during
biofuel production and would also create a new source of revenue to the biorefinery due to the
high market potential of lignin.

42

CHAPTER 3 - FACTORS THAT CONTRIBUTE TO CELLULOSE III CONVERSION
AND ASSOCIATED IMPACT ON ENZYMATIC DIGESTIBILITY OF CELLULOSE.

3.1. Abstract
The crystalline nature of cellulose is considered a natural barrier to enzymatic action, preventing
efficient decomposition of cellulose to simple sugars during microbial attack. The crystalline
allomorph of plant-derived cellulose, i.e., cellulose Iβ (CI), can be altered by ammonia treatment
to form a more amenable substrate to enzymatic degradation, designated by cellulose III (CIII).
In this work, we studied several aspects of CIII conversion, including the effect of ammonia-tocellulose ratio, ammonia concentration in organic solvent, incubation time and incubation
temperature on the formation of CIII. To evaluate the crystal allomorph composition of
ammonia-treated samples, XRD spectra were processed based on a new strategy that combines
amorphous subtraction and peak deconvolution methods. Kinetic data show that the amorphous
content of cellulose increases substantially for the first few seconds of ammonia incubation.
After this early event, a slow reduction of the amorphous content is observed, being replaced by
the formation of CIII for extended incubation times. These observations suggest that incomplete
disruption of hydrogen bonding during formation of the ammonia-cellulose complex, promote a
disordered assembly of the cellulose chains upon evaporation of ammonia. Elevated
temperatures accelerate ammonia penetration into the cellulose fibers and formation of the
ammonia-cellulose intermediate, generating cellulose samples of higher CIII crystallinity after
30 min of incubation time. As CIII crystallinity increases, we also observed improvement of
enzymatic hydrolysis rates up to 2 folds compared to native CI. This observation contradicts the
43

current paradigm that supports the idea that lower cellulose crystallinity benefits enzymatic
degradation of cellulose. Furthermore, the utilization of co-solvents, such as ethanol, along with
ammonia seems to be compatible with CIII conversion, as no reversion to CI was observed when
the ammonia-cellulose intermediate is formed. From our results, cellulose treated with 60 wt% at
25 °C for 30 min incubation time, allow approximately two folds increase in enzymatic
hydrolysis rates as compared to CI. This solvent compatibility enables the reduction of operating
pressure at higher pretreatment temperatures, which can be economically advantageous in the
context of a lignocellulosic biorefinery.
3.2. Introduction
One of the major bottlenecks associated with biological conversion of lignocellulosic biomass to
fuels and chemicals, derives from the structural organization of cellulose in tightly packed
microfibrils, sustained by a complex network of hydrogen bonds [114-116]. This naturally
occurring crystalline arrangement of cellulose, denominated by cellulose Iβ (CI), is highly
recalcitrant to cellulases, representing the final barrier for sugar conversion [114]. Amorphous
cellulose (AC) and other non-native forms of crystalline cellulose are known to be more
vulnerable to cellulase activity, generating substantial improvements to the rate of enzymatic
hydrolysis [105]. For example, Chundawat et al. (2011) converted CI derived from Avicel PH101 into cellulose III (CIII) using anhydrous ammonia, thereby improving enzymatic hydrolysis
rates by two folds compared to the original CI [105]. Previously, Igarashi et al. (2007) [117] had
already observed that cellulose III promoted about five fold higher enzymatic hydrolysis rate
using Cel7A enzyme on algal-derived cellulose. The reasons for such improvement are far from
being totally understood, however several pieces of work have unveiled interesting phenomena
44

that contradicted some of the paradigms established for enzymatic hydrolysis of CI. For
example, Gao et al. (2013) [118] observed recently that cellulases tend to bind less effectively to
CIII than to CI, even though their specific activity is higher on CIII. Computational simulations
suggested that this enhancement is derived from favorable free energies during decrystallization
of individual cellulose chains, as well as intrinsic kinetic factors related to processivity of bound
enzymes to the CIII substrate.
Several authors have studied the factors that impact cellulose III conversion [119]. Liquid
anhydrous ammonia and primary amines have been successfully used as swelling agents of
cellulose, as they intercalate onto the crystalline structure of cellulose by hydrogen bonding [80,
120-122]. CIII is formed from these cellulose complexes of larger crystallite size upon removal
of the mentioned swelling agents [80, 120]. Ammonia has been the most commonly used
swelling agent for CIII conversion in the literature. Ammonia treatment has been performed at
various conditions that include its usage in supercritical state or liquid state at ambient pressures
[80, 117]. Other authors cycle temperature from -33 °C to 25 °C or higher prior to ammonia
evaporation [123]. However, all processing conditions referred by those authors are energy
intensive and do not benefit the implementation of CIII conversion in a lignocellulosic
biorefinery context. Ideally, the most favorable conditions would be performed at constant and
close to ambient temperatures, thereby reducing energy cost and capital cost due to milder
operating pressures.
The effect of temperature during ammonia evaporation has been considered a major factor
impacting conversion of CI to CIII [80]. Higher temperatures can improve significantly
crystallinity indices (CrI) of the treated cellulose samples, while ammonia treatment time has
been reported to reduce CrI [80] as well as the CI content of cellulose. To our knowledge, a
45

systematic understanding of how temperature, time and ammonia loading impact CIII conversion
is not yet been reported in the literature. Also, the impact of crystalline CIII composition on
enzymatic hydrolysis yields is a question that remains to be answered in a systematic manner.
Therefore, the work presented herein attempts to create analytic methodologies to measure
relative ratios of CI, CIII and AC in samples treated under various processing regimes, thereby
allowing understanding of the effect of cellulose allomorph composition on enzymatic hydrolysis
efficiency. Moreover, the effect of ammonia treatment variables (i.e., temperature, time,
ammonia loading and concentration in the presence of co-solvents) on cellulose allomorph
composition will be evaluated in detail in this work.
3.3. Experimental
3.3.1. Reaction apparatus
A reaction apparatus was designed and assembled to convert cellulose Iβ (CI) to cellulose IIII
(CIII) under controlled conditions. As shown in Figure 3-1, the apparatus consisted of two major
sections that include ammonia delivery (1-6) and reaction equipment (7-12). As it was important
to accurately deliver the appropriate amounts of ammonia at set-point temperature, an ammonia
preheating vessel (1 – Parr Instrument Company, Moline, IL, USA) equipped with temperature
controller (2) was installed on the top of a balance (5 – A&D, Inc., San Jose, CA, USA). This
balance was used to measure the displaced weight of ammonia from the preheating vessel to the
reactor by a jacketed syringe pump (4 – Teledyne Isco Model 500D, Lincoln, NK, USA). To
avoid heat losses during ammonia delivery, the syringe pump jacket was temperature controlled
by a circulating water bath (3 – VWR, Radnor, PA, USA). Ammonia was delivered to the
46

reaction vessel (7 – HEL, Inc., Hertfordshire, UK) at constant pressure with the help of a
backpressure regulator (6 – Parker, Cleveland, OH, USA). An online temperature PID controller
(8 – Julabo, Seelbach, Germany) was installed to maintain the desired temperature during
reaction time. Reaction pressure was controlled by HEL, Inc. software (12) actuating two
solenoid valves (9 and 10) for nitrogen inlet and outlet, respectively. Reaction temperature and
mixing controls were also activated via computer software designed by HEL, Inc. (12). Cellulose
suspensions were sampled through a valve connected to a deep tube (11) that was installed in top
of the reaction vessel.
3
90C

1

9
300 RPM

N2

Solenoid
Valve

10

4
NH3

6

Parr

Vent

300
RPM

N2

Solenoid
Valve

GC-TCD

HEL

7

Sampling
11

pH

SP 90 C
90°C
74 C

SP 90 C
90 C

2

12

8

5

Figure 3-1 Instrumental apparatus for conversion of CI to CIII: 1) High-pressure mixing vessel
and mantle used to preheat ammonia to set-point temperature; 2) Temperature controller for
ammonia preheating vessel; 3) Water circulator used to heat the syringe pump to reaction setpoint temperature; 4) Syringe pump to pump the desired volume of ammonia; 5) Balance to
confirm the delivery of the desired ammonia loading; 6) Backpressure regulator; 7) Highpressure mixed reaction vessel and mantle for cellulose III conversion; 8) Temperature controller
47

Figure 3-1 (cont’d)
for reaction vessel; 9) Solenoid valve for pressure control (nitrogen inlet); 10) Solenoid valve for
pressure control (nitrogen outlet), also used for venting the system; 11) Deep tube sampling port;
12) computer controls and data logging.

3.3.2. Cellulose III conversion studies
3.3.2.1. Impact of sampling solvent on the conversion of ammonia-cellulose complex to CIII
The ammonia-cellulose complex was formed in a high pressure reaction vessel (Figure 3-1, 7)
by contacting anhydrous ammonia with Avicel PH-101 (Sigma-Aldrich, St.Louis, MO, USA)
using 6:1 ammonia-to-celullose ratio, at 90 °C for 20 min residence time. The reaction occurred
under 1000 psi of nitrogen overpressure to assure all the loaded ammonia was in the liquid
phase. Suspensions of ammonia-cellulose complex in liquid ammonia were sampled into vessels
containing excess of various solvents at 25 °C. These vessels were kept in a temperature
controlled water bath to reduce temperature drop during ammonia evaporation. The solvents
used in this study were acetone, methanol, ethanol and water, in a proportion of approximately
30:1 (v/v) solvent-to-sample. Ammonia pressure was released from the top of the sampling
vessel and the resulting cellulose fibers were filtered using a Millipore (Billerica, MA, USA)
vacuum-filtration system equipped with a glass microfiber filter (Whatman GF/A, Kent, UK).
The treated cellulose samples were further washed with the respective solvents and dried
overnight in the hood to evaporate residual solvent. Samples were stored in zip sealed plastic
bags at 4°C prior to usage. XRD was performed in these treated samples within the first 48 hours
after reaction completion.

48

3.3.2.2. Effect of ammonia loading on CIII conversion
The ammonia-cellulose complex was formed using various loadings of anhydrous ammonia,
notably 1:1, 2:1, 3:1 and 6:1 ammonia-to-cellulose ratio, at 90 °C for 10 min residence time. The
microcrystalline cellulose used in this study was Avicel PH-101 (Sigma-Aldrich, St. Louis, MO,
USA). The ammonia-cellulose complex activation occurred under 1000 psi of nitrogen
overpressure to assure all the loaded ammonia was in the liquid phase. A suspension of that
intermediate complex was transferred to 100 % ethanol in a ratio of approximately 30:1 solventto-sample, which was maintained at 25 °C in a temperature controlled water bath. Ammonia
pressure was released from the top of the sampling vessel and the resulting cellulose fibers were
filtered using a Millipore (Billerica, MA, USA) vacuum-filtration system equipped with a glass
microfiber filter (Whatman GF/A, Kent, UK). The treated cellulose samples were further washed
with absolute ethanol and dried overnight in the hood to evaporate residual solvent. Samples
were stored in zip sealed plastic bags at 4 °C prior to usage. XRD was performed in these treated
samples within the first 48 hours after reaction completion.
3.3.2.3. Effect of temperature on cellulose III conversion
In these experiments, ammonia was added to a reaction vessel (Figure 3-1, 7) containing Avicel
PH-101 at set point temperature, using a syringe pump (Teledyne Isco Model 500D, Lincoln,
NK, USA). Ammonia-cellulose complexes were formed solely with anhydrous ammonia, using
6:1 ammonia-to-cellulose ratio, at various temperatures (ranging from -30 °C to 90 °C). All
incubations were performed for 30 min residence time and 1000 psi of nitrogen overpressure to
assure that all added ammonia was in liquid state and in contact with cellulose. In this study, it
was important to differentiate between the effects of temperature during ammonia-cellulose
49

complex formation and temperature of CIII conversion during ammonia evaporation. For this
purpose, ammonia-cellulose complexes formed at the various temperatures were sampled to a
vessel containing excess of 100 % ethanol at a constant temperature of 25 °C. The cellulose
suspensions in ethanol, obtained after ammonia evaporation from the sampling vessel, were
further filtered using a Millipore (Billerica, MA, USA) vacuum-filtration system equipped with a
glass microfiber filter (Whatman GF/A, Kent, UK). Residual ammonia was further removed by
flowing 100 % ethanol through the sample while applying vacuum to the filtration system.
Ammonia-treated cellulose samples were further dried in the hood overnight to evaporate
residual ethanol. Samples were stored in zip sealed plastic bags at 4 °C prior to usage. XRD was
performed in these treated samples within the first 48 hours after reaction completion.
3.3.2.4. Effect of ethanol:ammonia solvent system on CIII conversion
Avicel PH-101 was placed in the high pressure reactor (Figure 3-1, 7) in the presence of
calculated amounts of 100 % ethanol. Appropriate volumes of ammonia were loaded in the
reactor with a syringe pump at 25 °C. Several concentrations of ammonia in ethanol were tested,
ranging from 17 wt% to 60 wt% ammonia. Liquid-to-solid ratios were also varied for a fixed
ammonia concentration, ranging from 4:1 to 10:1 (w/w). Avicel PH-101 was incubated in these
ammonia solutions at 25 °C for 30 min residence time before being transferred through a
sampling valve to a vessel containing absolute ethanol at 25 °C. Ammonia was further
evaporated from the sampling vessel and the cellulose suspension was transferred to a vacuum
filtration system (EMD Millipore, Billerica, MA, USA), equipped with glass microfiber filters
(Whatman GF/A, Kent, UK). The treated cellulose was extensively washed with 100 % ethanol
to remove residual ammonia. Treated cellulose samples were air dried overnight in the hood to
50

evaporate residual ethanol prior to storage in zip sealed plastic bags at 4 °C. XRD was performed
in these treated samples within the first 48 hours after reaction completion.
3.3.3. Preparation of cellulose standards
High purity (>95 %) microcrystalline CI, purchased from Sigma-Aldrich (Avicel PH-101),
served as CI standard in this study. The CIII standard was produced from Avicel PH-101 using
liquid anhydrous ammonia, prepared in a high-pressure stirred batch reactor (Figure 3-1) at 90
°C for 30 min residence time, using 6:1 ammonia-to-cellulose ratio. The reactor pressure was
maintained constant at 1000 psi with nitrogen. After reaction completion, ammonia was slowly
evaporated from the reactor through a venting valve (Figure 3-1, 10). During this process, the
temperature of the reactor slowly decreased due to heat of vaporization of ammonia and was
stabilized at 25 °C with a heating controller. After reaching atmospheric pressure, the cellulose
sample was removed from the reactor, transferred to a flat container and placed overnight in the
hood to evaporate residual ammonia. No visual differences in color were observed after
ammonia treatment of cellulose at these conditions. The CIII standard was stored at 4 °C in a zip
sealed bag prior to usage. The amorphous cellulose (AC) standard was produced using
phosphoric acid treatment as described in the literature [124]. Phosphoric acid swollen cellulose
(PASC) was freeze dried and stored at 4 C in zip sealed bags prior to further usage.
3.3.4. X-ray powder diffraction (XRD)
XRD was performed on an X-ray powder diffractometer with beam parallelized by a Gobel
mirror (D8 Advance with Lynxeye detector; Bruker, Bruker AXS Inc., Madison, WI, USA).
CuKα radiation (wavelength = 1.5418 Å) was generated at 40 kV and 40 mA. Detector slit was
set to 2.000 mm. Sample was analyzed using a coupled 2θ/θ scan type with a continuous PSD
51

fast scan mode. 2θ started at 8.000° and ended at 30.0277° with increments of 0.02151°, while θ
started at 4.0000° and ended at 15.0138° with increments of 0.01075°. Step time was 1.000 sec
(i.e., 1025 total steps, effective total time 1157 sec per run). Cellulose samples (approximately
0.5 g) were placed in a specimen holder ring made of PMMA with 25 mm diameter and 8.5 mm
height, rotating at 5 degrees per minute during analysis.
3.3.5. XRD data analysis
To determine the relative amounts of the various crystalline allomorphs of cellulose (CI and
CIII) as well as the level of AC in a sample, a series of standards were produced as described
previously and further analyzed by XRD. As the CI and CIII standards contained a relatively
high content of AC, it was required to remove the amorphous portion of the XRD spectra,
thereby reproducing spectra correspondent to 100 % crystalline CI and CIII. For this purpose,
AC standard (PASC) was used in a technique designated by amorphous subtraction method
(ASM) [125, 126]. Figure 3-2 shows how the ASM method was applied to the CI and CIII
standards. The yellow area, representing the AC spectrum, was subtracted from the XRD spectra
of CI and CIII, generating new XRD spectra representative of 100 % crystalline CI and CIII.
These XRD spectra representing 100 % CI and CIII were further used in peak deconvolution of
several ammonia treated samples produced in this study. As example, Figure 3-3 shows a
sample containing a mixture of AC, CI and CIII, which was deconvoluted using standard spectra
of CI, CIII and AC. The deconvolution was performed using the following equation:

Model XRD Spectra =  · AC +  · CI + γ · CIII
Where,
52

(3-1)

AC = amorphous cellulose spectra; CI = 100 % cellulose I standard spectra; CIII = 100%
cellulose III standard spectra. The coefficients ,  and γ control the proportion of each crystal
in the mixture.

Figure 3-2 Example of the XRD spectra of cellulose I (CI), cellulose III (CIII) and amorphous
cellulose (AC). Amorphous subtraction method was applied to CI and cellulose CIII standards
for this study. The yellow area was subtracted from the XRD spectra of CI and CIII standards as
described in this figure to recreate the 100 % crystalline CI and CIII spectra used for XRD
spectra deconvolution.
The coefficients ,  and γ from the model XRD spectrum were determined using the least
squares method. The resulting model XRD spectra was further evaluated for statistical
2

significance, showing in all cases P values of 0.00 and R values above 0.99. In Figure 3-3 is
possible to observe an example of a model curve (in red) fitting the XRD spectra of a sample
containing unknown composition of CI, CIII and AC. The composition of a designated sample
containing CI, CIII and AC was analyzed based on the relative area occupied by each standard
53

2

XRD spectrum after deconvolution. For all the model fits reported in this manuscript, R values
were always higher than 0.99, demonstrating the ability of the model to fit XRD spectra obtained
for the various conditions.

Figure 3-3 Example of an XRD peak deconvolution applied to a sample containing a mixture of
amorphous cellulose (AC), cellulose I (CI) and cellulose III (CIII). The model spectra (red line),
contrasting with the sample XRD spectra (black line) is presented in the top right corner.

Peak deconvolution methods have been used in the literature to calculate the crystallinity index
of cellulose samples [126-129], however they are usually performed summing a combination of
Gaussian, Lorentzian or Voigt functions to fit XRD spectra of the sample. However, multiple
curves are required to describe a mixture of various cellulose allomorphs, and the high number of
degrees of freedom increases the number of possible combinations of intensity for the curves that
describe the XRD spectrum of the sample. The method described herein, eliminates this problem
54

in great extent, as the ratio between each peak is maintained constant within the standard XRD
spectrum of CI, CII and AC. Therefore, the method produces consistent results, favoring our
analysis, which is will be mainly a relative comparison between samples as opposed to a
discussion about absolute values.
3.3.6. Enzymatic hydrolysis
Cellulose samples were subjected to enzymatic hydrolysis at 1% glucan loading. The total
protein loadings used during reaction was 4.0 or 7.5 mg/ g glucan of Accellerase 1500 enzymes
(DuPont, formerly Genencor, Palo Alto, CA, USA), depending on the experiment. When 4 mg
enzyme/g of glucan was used, sugar yields were evaluated for 24h residence time, while for 7.5
mg/g glucan, enzymatic hydrolysis was evaluated for 4 h residence time. In all experiments, the
reaction pH was adjusted to 4.8 prior to incubation at 50 °C in an orbital shaking incubator set at
200 RPM (New Brunswick, Innova 44, Enfield, CT, USA). Monomeric glucose was quantified
using a Shimadzu HPLC system equipped with an Aminex HPX-87P carbohydrate analysis
column (Biorad, Hercules, CA) and Waters 410 refractive index detector (RID). Degassed HPLC
grade water with 5 mM H2SO4 was used as a mobile phase at 0.6 mL/min. Injection volume was
20 µL with a run time of 20 minutes.
3.4. Results and Discussion
3.4.1. Effect of sampling media on cellulose III conversion
The mechanism for converting CI to CIII proposed by Schuerch (1963) [130], involves the
penetration of anhydrous ammonia into the cellulose crystal, followed by the disruption of
hydrogen bonding, creating a crystalline complex with a larger unit cell than CI, designated by
55

ammonia-cellulose complex. After evaporating ammonia, the hydrogen bonding is rewired in a
completely different pattern, generating a new crystal designated by CIII. During this operation,
the heat of vaporization of ammonia is often large enough to reduce the temperature in the
surface of the cellulose fibers below 0 C. As described by Menachem Lewin et al. (1971) [80],
the temperature at which CIII is formed, i.e., during ammonia evaporation, can affect the
crystallinity of cellulose and therefore, is necessary a good control of this parameter to avoid
misleading experimental results during CIII formation. For this purpose, a “heat sink” composed
by a large enough volume of solvent was prepared. The ammonia-cellulose complex, suspended
in liquid ammonia, was sampled from the high-pressure reactor into a “heat sink” prior to
ammonia evaporation. The cellulose sample was further filtered and washed extensively with
solvent to remove residual ammonia before being dried in the hood overnight at room
temperature. For this experiment, acetone, ethanol, water and methanol were evaluated as
sampling solvents. The ammonia-cellulose complex was generated using Avicel PH-101 and
ammonia at 90 C with 6:1 ammonia-to-cellulose ratio for 20 min. The glass vials containing the
sampling solvents were also kept at 25 C in a temperature-controlled water bath. As ammonia
evaporated from the solvent, temperature did not decrease below 20 C for all cases and
therefore, the solvents tested herein could potentially be used to control temperature during
assembly of the CIII crystal. However, similarly to temperature, the presence of a solvent can
also affect the way CIII crystals are assembled and the final crystallinity of the sample.
Therefore, this work attempted to evaluate differences between the various solvents tested herein
in order to select the one that can offer the best environment for CIII conversion.
By performing XRD on the treated cellulose samples (Figure 3-4), it was possible to observe
differences in CIII conversion as a function of the sampling solvent used. CIII formation was
56

evaluated based on crystallinity index (CrI) measurements, which were performed according to
the amorphous subtraction method, as described earlier. The range of values available on the
literature for CrI of Avicel PH-101 can vary significantly from ~39 to ~90 %, depending on the
calculation method [126, 131]. Amorphous subtraction methods usually generate crystallinity
indexes between ~60 and ~75 % for Avicel PH-101 [126, 131], however the result obtained in
this work was 50.1%, which is slightly lower than most reports from the literature. As these
values of CrI are not consistent among authors, our discussion will be based on relative
differences between the various samples generated in this work, instead of focusing on
comparing absolute values with literature results. The XRD data shown in Figure 3-4 clearly
reveal the importance of the environment at which hydrogen bonding is formed during CIII
crystal assembly. From the organic solvents used in this study, ethanol was the one that allowed
higher conversion to CIII and consequently higher glucose conversion. Also, methanol seemed to
provide a better environment for CIII formation as compared to acetone and air, which generated
similar results. The reasons behind such differences are still to be determined and more work is
required to clarify this observation. As ethanol was the most effective sampling media for CIII
conversion, it was defined as the choice of sampling solvent for the subsequent experimental
work.
The most distinct result was generated by sampling the ammonia-cellulose complex into water.
This sample generated an undefined XRD pattern that could not be effectively deconvoluted by
the method described earlier in the materials and methods section, and therefore, it suggests that
this sample of low crystallinity cannot be described as a mixture of solely CI, CIII and AC
produced by phosphoric acid treatment. These results also suggest that water interferes with
rewiring of the hydrogen bonding once ammonia leaves the ammonia-cellulose complex,
57

creating a less ordered pattern of hydrogen bonds that is reflected by the amorphous-like XRD
spectrum (Figure 3-4 ii.b). Lewin et al. (1971) [80], performed a similar experiment with cotton
cellulose, reporting a complete reversion to CI after drying the sample at 65 °C. Others studies
conducted by Barry et al. (1936) [132] and Clark et al. (1937) [133] revealed that aqueous
ammonia treatment is able revert CIII back to CI, after drying at 105 °C. The cellulose type and
drying method used by these studies are two major differences contrasting with the work
presented herein, where Avicel PH-101 was used as the cellulose source and freeze drying was
used as the preferred method to remove water from the sample. The rationale behind this choice
of drying process was purposely to avoid reversion of CIII back to CI, as our preliminary
evaluation revealed that freeze drying does not allow reversion of fully converted CIII to CI and
has low impact on the final cellulose crystallinity [data not shown].
Enzymatic digestion of the water-sampled cellulose resulted in lower sugar yields compared with
cellulose sampled to organic solvents, even though the former presents lower crystallinity. This
observation is not concordant with the current paradigm that supports the idea that less
crystalline cellulose allows higher enzymatic hydrolysis yields [114]. The definition of AC as
being a sample with no ordered structural pattern can contribute to the misleading idea that all
AC is made alike. However, this data suggests that this paradigm can no longer be applied for
cellulose III in all instances. Another relevant observation supporting this idea is that higher CIII
crystallinity improved cellulose vulnerability to enzymatic digestion, which usually is not
observed for CI [114, 134, 135]. This topic will be further explored in this manuscript, where
new evidence supporting this idea will be presented and discussed.

58

Figure 3-4 i) Glucose conversion of cellulose treated at 90 C with 6:1 (w/w) ammonia-tocellulose ratio for 20 min residence time, sampled to various solvents and to no solvent (air). The
control experiment corresponds to cellulose I (Avicel PH-101). Enzymatic hydrolysis was
performed using 4 mg/g glucan of Accellerase 1500 enzymes for 24h; ii. a) Crystallinity index
(CrI) of the various samples of ammonia treated cellulose sampled to different solvents and to no
solvent (air). CrI measurement was performed by amorphous subtraction method of the XRD
spectra showed in b).

3.4.2. Effect of ammonia loading on cellulose III conversion
Ammonia loading is one of the key factors contributing to the effectiveness of ammonia-based
pretreatments on lignocellulosic biomass [136, 137]. Most of these pretreatments use aqueous
ammonia at elevated temperatures to improve the accessibility of enzymes to cellulose. Although
59

these pretreatments can partially decrystallize cellulose fibers [138], the presence of elevated
proportions of water inhibits modifications to the crystalline morphology of cellulose, notably
the formation of CIII. Efficient conversion of CI to a more easily digestible CIII crystal requires
contact of the cellulose fibers with liquid anhydrous ammonia for a period of time at a range of
temperatures (effect of temperature and time will be further discussed in this manuscript). The
ammonia-to-cellulose ratio can impact both the rate of liquid ammonia penetration to the
cellulose fibers and the number of hydrogen bonding sites that will be activated by ammonia.
Incomplete formation of the ammonia-cellulose complex will affect cellulose III conversion and
enzymatic hydrolysis performance. As ammonia usage has economic implications to the
biorefinery, is important to optimize ammonia loading to efficiently convert CI to CIII using
minimum ammonia-to-biomass ratio. To determine the minimum ammonia loading that allows
efficient CIII conversion, anhydrous ammonia was put in contact with microcrystalline CI
(Avicel PH-101) at various ammonia-to-cellulose ratios ranging from 1:1 to 6:1 (w/w). In Figure
3-5, it is possible to observe (i) the extent of CIII conversion by XRD and (ii) the respective
impact on enzymatic hydrolysis for this range of ammonia loadings. From the XRD spectra
presented in Figure 3-5 ii) it is evident that ammonia-to-cellulose ratios of 1:1 and 2:1 are not
sufficient to completely modify the native crystal state of cellulose in 10 min of ammonia
treatment. Efficient conversion of CI to CIII is only observed at ammonia-to-cellulose ratios
above or equal to 3:1. From the XRD spectra, it is also clear that the CIII peaks are sharper and
taller for 6:1 than for 3:1 ammonia loading, leading to a more effective CIII conversion. This
result suggests that higher ammonia loadings (e.g., 6:1) can improve the mass action responsible
for efficient penetration of ammonia in the cellulose crystal, being able to activate the ammoniacellulose complex at faster rate than the lower ammonia loadings (e.g., 3:1). It is possible that for
60

the lower ammonia loadings, some hydrogen bonding sites are not activated by ammonia within
the first 10 min of treatment, leading to a poorer alignment between cellulose chains during
ammonia evaporation which resulted in a product of lower CIII crystallinity. For ammonia-tocellulose ratios above 6:1 it was not possible to observe significant changes in CIII conversion
(data not shown), suggesting a saturation of the ammonia loading effect around 6:1. For this
reason, 6:1 was the maximum ammonia loading used in this study.
Other authors have successfully converted cellulose I to III using ammonia loadings between 2:1
and 3:1 for longer periods of time (30-90 min), while ramping the temperature from -75 C to 25
C or higher [123]. Also, literature reports show that is possible to achieve high CIII conversion
with ammonia in the supercritical state [139]. However, these techniques are not practical for an
industrial pretreatment method, as supercritical conditions and temperature fluctuations of those
amplitudes often require high energy demands. Therefore, we chose to deliver ammonia in
subcritical conditions, at set-point temperature and allow it to contact cellulose for the desired
residence time. This more practical technique showed to be effective on producing CIII, which
can be confirmed by the various XRD results presented in this study.
The enzymatic hydrolysis results from Figure 3-5 i) show a progressive increase of glucose
yields as a function of CIII conversion. Also, the results are concordant to our previous
observation, showing that higher CIII crystallinity favors improved enzymatic hydrolysis rates.
Sugar yields increased up to approximately two fold for the most crystalline sample of CIII (i.e.,
6:1), as compared to native CI (from ~32 % to ~ 59 %). This level of improvement has been
observed by other authors in the literature, notably Chundawat et al. (2011) [105], using similar
substrate and enzyme cocktail. However, in the literature it is possible to find a wide range of
results. For example, Igarashi et al. (2007) [117] observed even higher improvements when
61

applying purified enzymes on algae-derived CIII, resulting in approximately five folds higher
glucose conversion compared with algae-derived cellulose I. On the other hand, Mittal et al.
(2011) [123] recently showed significantly lower improvements, or even no improvements
during 24h enzymatic hydrolysis of CIII compared with CI. However, a different cellulase
cocktail and enzyme loading were used in that work, which could partially explain some of the
variability with respect to this and other reports from the literature.

Figure 3-5 i) Glucose conversion of ammonia treated cellulose under various ammonia:cellulose
ratios, varying from 1:1 to 6:1 (w/w), compared to the control (Avicel PH-101). Enzyme loading
used in the experiment was 4 mg/g glucan for a residence time of 24h; ii) XRD spectra of
ammonia treated cellulose using various ammonia:cellulose ratios at 90 C for 10 min residence
time.

62

3.4.3. Effect of temperature and time on cellulose III conversion
Temperature is another key factor contributing to CIII conversion as reported by a number of
authors in the literature [80, 123]. This factor is not only important during evaporation of
ammonia, but also plays an important role during formation of the ammonia-cellulose
intermediate [140]. In this work, the temperature during ammonia evaporation was set constant
at 25 °C. This temperature was controlled using absolute ethanol as the sampling solvent, which
was placed in a temperature-controlled water bath as described earlier. Ammonia treatment was
conducted at 25 °C and 90 °C to evaluate the effect of temperature in the process of CIII
conversion. These two temperatures are reasonably far apart from each other, allowing CIII
conversion differences to be noticeable. Also, 90 °C was the maximum temperature limit
observed for effective CIII conversion without thermally decomposing the sample.
Decomposition products derived from high temperature treatments could inhibit enzymes during
cellulose digestion [108] and affect the correlations between sugar yields and CIII conversion
performed in this study.
Residence time is also important for CIII conversion, as observed initially by Clark et al. (1937)
[133] and Lewin et al. (1971) [80]. The latter observed a decrease in equatorial CrI with
increasing time of ammonia treatment from 2 sec to 5 min. In his experiment, ammonia was
evaporated from cotton samples that were dried at 105 °C. Lewin et al. (1971) also described the
transition of CI to CIII as a function of treatment time, with increasing proportions of CIII for
larger residence times. However, these authors did not quantify the proportions of AC, CI and
CIII as a function of time and only allowed a maximum treatment time of 5 min. In Figure 3-6 it
is possible to clearly observe the transition between these various forms of cellulose as a function
of treatment time. For this experiment, Avicel PH-101 was treated with 100 % anhydrous
63

ammonia at 25 °C and 90 °C in order to evaluate CIII conversion profiles at low and high
temperatures. Figure 3-6 i) clearly shows an abrupt increase of CIII content within the first
minute of treatment, while CI was rapidly depleted from the sample at 25 °C. This observation
shows that CIII can be formed at 25 °C with no further heat requirement. It also confirms the
results obtained by Mittal et al. (2011), who observed full CIII conversion at 25 °C for Avicel
PH-101 and other sources of cellulose [123], including corn stover cellulose. Figure 3-6 i.a) also
shows a rapid increase of the amorphous content of cellulose (AC) in the first minute of
treatment, which slowly decreases with treatment time. One possible justification for this result
is that ammonia could fully activate some accessible areas of cellulose during the first minutes of
treatment, generating mainly CIII after ammonia evaporation. As ammonia could not access less
exposed regions of the cellulose microfibrils and activate the ammonia-cellulose intermediate,
they remained intact as CI. Other areas could be partially activated by ammonia through the
disruption of some hydrogen bonds, which during rapid ammonia evaporation could not align
properly to generate a crystalline structure, resulting in an amorphous XRD pattern. The results
obtained by Lewin et al. (1971) [80] also showed an increase of the amorphous content of cotton
in the first 5 min of treatment time. As 5 min was the maximum residence time tested by this
author, it may have not been long enough for ammonia to effectively penetrate well packed
cotton fibers and therefore, it was not possible to observe the expected reduction of the
amorphous content.
At 90 °C, a similar CIII conversion profile could be observed, with a rapid increase of the
amorphous content of cellulose, followed by its reduction for increasing treatment times. At the
same time, a rapid increase of CIII conversion was observed as CI was depleted from the sample.
However, a major difference between treatment at 25 °C and 90 °C is that at higher temperatures
64

CIII is converted at a much faster rate, generating a sample of higher crystallinity than at lower
temperatures. For example, the maximum amorphous content obtained at 90 °C was ~62 %,
comparing with ~66 % at 25 °C, both occurring at 30 sec of treatment. This result suggests that
temperature plays a role on promoting ammonia penetration in the cellulose fiber and creating
faster formation of the ammonia-cellulose complex. It is also experimental evidence supporting
previous computational simulation results obtained by Bellesia et al. (2011) [140] that clearly
suggests an increase of ammonia penetration rate with temperature.

Figure 3-6 i) Relative proportions of cellulose III (CIII), cellulose I (CI) and amorphous
cellulose (AC) during anhydrous ammonia treatment of Avicel PH-101 at 25 °C, as a function of
65

Figure 3-6 (cont’d)
residence time (a) and respective XRD spectra used to calculate CIII, CI and AC contents at 25
°C (b). ii) Relative proportions of cellulose III (CIII), cellulose I (CI) and amorphous cellulose
(AC) during anhydrous ammonia treatment of Avicel PH-101 at 90 °C, as a function of residence
time (a) and respective XRD spectra used to calculate CIII, CI and AC contents at 90 °C.

At 30 min residence time, the treated cellulose was totally depleted of CI and was composed
mainly by AC and CIII at both 25 °C and 90 °C (Figure 3-6). The main difference between the
two samples at this time point was that at 90 °C, the sample contained approximately 49 % of
CIII, while at 25 °C it contained about 41 % of CIII. The effect of temperature during ammonia
treatment of Avicel PH-101 was also reported by Mittal et al. (2011) [123], who also observed
increase of the CrI for samples treated at higher temperatures. However, a comprehensive kinetic
profile has never been presented in the literature showing the evolution of AC, CI and CIII as a
function of time and temperature.

Figure 3-7 i) Comparison between amorphous content of Avicel PH-101 treated at various
temperatures and correspondent effect on 4h glucose conversion during enzymatic hydrolysis.
Enzymatic hydrolysis was performed with 7.5 mg/g glucan of Accelerase 1500 enzymes at 50
°C. ii) XRD spectra of Avicel PH-101 treated with phosphoric acid (PASC) and with ammonia at
-30, 25 and 90 °C. At 25 and 90 °C treatment, ammonia was evaporated at 25 °C in ethanol,
while at -30 °C ammonia was vacuum filtered and dried in the hood at ambient temperature.
66

To evaluate the effect of temperature and crystalline composition of cellulose on enzymatic
hydrolysis, Avicel PH-101 was treated at -30, 25 and 90 °C for 30 minutes. Samples treated at 25
and 90 °C were transferred to ethanol at 25 °C prior to ammonia evaporation, while the sample
treated at -30 °C was vacuum filtered and dried in the hood overnight at ambient temperature.
The results presented in Figure 3-7 demonstrate that was possible to obtain cellulose samples
with various levels of amorphous content. While at 25 and 90 °C it was possible to deplete the
sample completely from native CI, being converted to CIII and AC, ammonia treatment at -30 °C
generated an undefined XRD pattern that could not be completely deconvoluted and quantified
as a mixture of AC, CI and CIII. This XRD pattern suggests that ammonia treatment -30 °C
promotes the formation of highly disordered cellulose. This fact was also observed by other
authors, namely Saafan et al. (1984) [141] and Mittal et al. (2011) [123], however the impact of
this structural modification of cellulose on enzymatic digestibility was never reported in the
literature. Supporting our previous observations, the results presented in Figure 3-7 also show
that ammonia-treated cellulose samples with higher CIII content are more vulnerable to
enzymatic activity. According to the current paradigm, cellulose of higher amorphous content
should be more digestible by cellulases [114, 134, 135], contradicting the results shown in
Figure 3-7. However the observations used to create the current theory were based on results
obtained in native cellulose (CI) and amorphous cellulose generated by ball milling or
phosphoric acid treatment [107, 142]. From Figure 3-7 it is possible to confirm that highly
disordered cellulose created by ammonia treatment at -30 °C is more digestible than untreated
Avicel (CI) composed by approximately 49.9 % of AC. However, the same figure shows that
PASC, composed only by AC, presents approximately 1.5-fold improvement with respect to the
ammonia treated sample at -30 °C. These observations suggest that different methods for
67

creating AC generate quite distinct responses to enzymatic activity. Future work is required to
understand the basis for these differences at the structural level of cellulose, including evaluation
of the level of hydrogen bonding, hydrophilicity and surface area. These studies also must be
correlated with enzyme-substrate interaction measurements, such as binding affinity and
catalytic activity of cellulases on the various amorphous substrates.
From the perspective of ammonia pretreatment of lignocellulosic biomass, it is important to treat
cellulose in order to obtain the highest possible content of CIII, which translates to a more
efficient enzymatic digestion by cellulases. This work clearly shows that various levels of CIII,
CI and AC content can be achieved by controlling pretreatment variables such as temperature,
time and ammonia loading.
3.4.4. Effect of solvent concentration on cellulose III conversion
CIII conversion has been traditionally performed using 100 % liquid anhydrous ammonia,
applying heat for relatively short periods of time (< 1h) [80]. As this method uses a liquefied gas,
i.e. ammonia, pressure can be considered a potential issue for economic and safety reasons. For
example, when operating at 120 °C, pressure can increase up to approximately 1300 psi to
maintain ammonia in the liquid state. To avoid such levels of operating pressure and possibly
improve the economics and safety of the process, would be important to test the usage of less
volatile co-solvents, preferably polar solvents that can allow CIII to be formed. As previously
discussed in this manuscript, water does not allow CIII to be formed and therefore, is not a viable
option for this process. Our previous work defined ethanol as a very compatible solvent to be
used as a heat sink during ammonia evaporation. Ethanol and ammonia are also chemically
compatible at temperatures below 150 °C while in the absence of catalysts [143, 144].
68

Ultimately, ethanol is a cheap chemical, widely available and is considered one of the major
biofuel products produced by lignocellulosic biorefineries. Therefore, this chemical could be
considered a good candidate to be used as a co-solvent in an ammonia-based pretreatment of
lignocellulosic biomass.
Figure 3-8 shows the vapor-liquid equilibrium (VLE) curve of ammonia-ethanol mixture at 120
°C. This curve was obtained using the Non-Random Two Liquid (NRTL) thermodynamic model,
estimating the missing parameters with UNIFAC model in ASPEN Plus (v.7.1). This VLE curve
can help us determine the thermodynamic state of a two component mixture (ammonia/ethanol)
as a function of pressure for a temperature of 120 °C. Analyzing this curve it is evident that is
possible to decrease the operating pressure below 750 psi for mixtures consisting of < 60 wt%
ammonia. However, is important to evaluate the range of concentrations of ammonia that allow

Pressure (psia)
250.0 500.0 750.0 1000.0 1250.0 1500.0

efficient CIII conversion.

0.0

P-xy for AMMONIA/ETHANOL
P-x 120.0 C
P-y 120.0 C

0.1

0.2

0.3
0.4
0.5
0.6
0.7
Liquid/Vapor Mass Fraction AMMONIA

0.8

0.9

1.0

Figure 3-8 Vapor-liquid equilibrium curve (P-xy) for ammonia/ethanol mixture at 120 °C. The
curve was predicted using NRTL model, estimating missing parameters using UNIFAC by
Aspen Plus (v.7.1) software.
69

To determine the impact of co-solvent on CIII conversion, Avicel PH-101 was put in contact
with various concentrations of ammonia in ethanol, ranging from 17 to 60 %, at different liquidto-solid ratios, ranging from 4:1 to 10:1 (Figure 3-9). These experiments were performed at 25
°C for 30 min residence time. The results show that both 17 and 33 wt% ammonia solutions are
not concentrated enough to allow effective penetration of ammonia in the cellulose crystals and
disrupt hydrogen bonding. Most of the cellulose (CI) remains intact after pretreatment under
such conditions. Using a 50 wt% ammonia solution was possible to observe some CIII
conversion (21-27 %), especially at 6:1 and 10:1 ammonia-to-cellulose ratios. At 4:1 there was a
slightly lower CIII conversion compared to the later, which may have occurred due to mass
action limitations. It is important to note that 4:1 liquid-to-solid ratio, using a 50 wt% ammonia
solution, represents 2:1 ammonia-to-cellulose ratio. From the results obtained in this experiment,
CIII conversion could only be obtained after increasing the concentration of ammonia from 33
wt% to 50 wt%. It is possible that not only ammonia mass action is responsible for the formation
of the ammonia-cellulose complex intermediate. For example, pH may be also responsible for
improving solvation of the cellulose crystal by ammonia and perhaps facilitate the penetration of
ammonia in the cellulose crystal. To prove such hypothesis, a new set of studies is required
where pH should be monitored and controlled during contact of CI with ammonia.
When the concentration of ammonia was increased to 60 wt%, CIII conversion increased
substantially for all liquid-to-solid ratios. As demonstrated in Figure 3-9 iv), CIII conversion did
not vary significantly within the range of liquid-to-solid ratios used in this study. These results
also show that not all the CI was activated by ammonia, since about 5 % of CI was still present
in the various treated samples with 60 wt% ammonia. Nevertheless, was possible to obtain
approximately 40 % of CIII and 55 % of AC in all these samples.
70

The enzymatic hydrolysis performance was also measured as a function of ammonia
concentration and liquid-to-solid ratio, being also presented in Figure 3-9. These results show a
trend comparable with previously reported data from this manuscript, where higher glucose
yields tend to be achieved with increasing content of CIII in the treated sample. For the best
condition tested herein, i.e. treatment with 60 wt% ammonia, our results show 55-57 % glucan
conversion in the first 24h of enzymatic hydrolysis with 4 mg/g glucan of enzyme loading,
independently of the liquid-to-solid ratio used. This result suggests that is possible to achieve an
increase of approximately 2 fold comparing with naturally occurring CI hydrolyzed under the
same conditions (compare with control in Figure 3-4). At 4:1 liquid-to-solid ratio, the ammonia
loading represents about 2.4 g/g of cellulose and an ethanol loading of 1.6 g/g of cellulose. In
this study, the effect of temperature was not evaluated for the ammonia-ethanol system, since the
major goal of this work was to determine the potential utilization of ethanol as a co-solvent
during CIII conversion. In this context, our results indicate that ethanol is a suitable co-solvent
for such application, generating a substrate rich in CIII and allowing enzymatic hydrolysis results
comparable to the ones obtained by a 10 min treatment using anhydrous ammonia at 25 °C. Also,
the utilization of ethanol enables the possibility of operating at lower pressures during ammonia
treatment, which could be a determinant factor for the industrial implementation of a
pretreatment technology based on CIII conversion of lignocellulosic biomass.

71

Figure 3-9 Relative proportions of cellulose I (CI), cellulose III (CIII) and amorphous cellulose
(AC) as a function of ammonia concentration and liquid-to-solid ratio during ammonia
treatment. The ammonia concentrations used in this study were 17 wt% (i), 33 wt% (ii), 50 wt%
(iii) and 60 wt% (iv). Ammonia treatment was performed at 25 °C for 30 min residence time
using 4:1, 6:1 and 10:1 liquid-to-solid ratios for all ammonia concentrations. The crystalline
composition of cellulose was correlated with 24h glucose conversion using 4 mg/g glucan of
enzyme for the various samples.

3.5. Conclusion
In this work, the key factors contributing to CIII conversion of Avicel PH-101 and its impact on
enzymatic digestibility of cellulose were studied. For this purpose, a hybrid methodology for
XRD analysis, consisting of the combination of amorphous subtraction and peak deconvolution
72

methods was successfully developed and implemented to determine relative amounts of CI, CIII
and AC. To better understand the impact of variables such as incubation temperature, ammonia
loading and residence time, it was also important to have constant temperature during ammonia
release, which was revealed to be challenging due to heat of evaporation of ammonia. For this
reason, another innovation was introduced in this work, consisting of the transfer of the
ammonia-cellulose complex to ethanol before ammonia was evaporated. This method allowed
the ammonia evaporation at relatively stable temperatures avoiding inconsistent and misleading
results.
CIII conversion requires the access of liquid ammonia to the crystalline structure of cellulose and
therefore, this process needs a minimum ratio of ammonia-to-cellulose to be effective. From our
experiments, 3:1 was the minimum ammonia loading that could totally convert CI into CIII and
AC. Higher ammonia loadings, such as 6:1 could improve CIII conversion by reducing the
amorphous content of the treated sample. From kinetic data using 6:1 ammonia-to-cellulose
ratio, we could observe that CI is converted to different proportions of AC and CIII, depending
on ammonia incubation time and temperature. In our experiments, the amorphous content of
treated cellulose increased in the first few moments of ammonia impregnation, possibly due to
incomplete activation of the ammonia-cellulose complex before ammonia evaporation. As time
progressed, ammonia could incorporate more effectively into the crystalline structure of cellulose
allowing a complete formation of the ammonia-cellulose complex, which was better converted to
CIII after ammonia evaporation. This resulted in a decrease of the amorphous content of treated
cellulose as a function of time. From enzymatic hydrolysis results, we concluded that higher CIII
content benefits sugar conversion from cellulose, contrasting with what has been reported for CI.
Also, the AC generated by ammonia treatment cannot be compared with AC from PASC, for
73

example. This fact suggests that not all amorphous cellulose sources are created equally and
possibly have very distinct properties. To maximize CIII conversion and reduce AC contents of
treated samples it is important to use higher temperatures of incubation for longer residence
times, with ammonia-to-cellulose ratios larger than 3:1. However, in the perspective of an
ammonia-based pretreatment of lignocellulosic biomass, higher severity translates into additional
processing costs. One of the most important processing issues when operating with ammonia at
elevated temperatures is the high pressure. To be able to decrease operating pressure while
effectively converting CI to CIII and improving enzymatic hydrolysis yields, ammonia was used
in solution with ethanol as co-solvent. From the results in this study, we observed that 60 wt%
ammonia in ethanol allowed efficient conversion of CI to CIII, enhancing enzymatic hydrolysis
rates by 2 fold compared to untreated Avicel PH-101.

74

CHAPTER 4 -

EXTRACTIVE

AMMONIA

(EA)

PRETREATMENT

OF

CORN

STOVER: REACTIONS IN ANHYDROUS AMMONIA.

4.1. Abstract
The development of a novel pretreatment technology, denominated as Extractive Ammonia (EA)
pretreatment, was inspired by recent understanding of ammonia- plant cell wall interactions and
mechanism of ammonia pretreatment of lignocellulosic biomass. This pretreatment technology is
proposed herein for the first time and is defined by the formation of a less recalcitrant crystal of
cellulose, i.e., cellulose III, and a highly selective lignin solubilization, followed by extraction. In
this work, a response surface design approach was used to develop a mathematical model that is
able to predict sugar conversions (glucan and xylan), lignin extraction yields and extent of ester
bond cleavage as a function of pretreatment variables (temperature, ammonia-to-biomass ratio
and time) for corn stover. From the experimental results it was possible to observe that sugar
conversions during enzymatic hydrolysis is highly correlated with cellulose III formation, lignin
extraction yields and ester bond cleavage. Process optimization predicted the condition that
maximize sugar conversion as being 121.8 C, 6.2:1 NH3:Biomass loading and 30.2 min
residence time. At this condition is possible to achieve 93 % of glucan conversion to glucose and
79 % xylan conversion to xylose in 24 h enzymatic hydrolysis using 15 mg protein/g glucan, at 1
% glucan loading. At the same processing conditions, the regression model predicts an extraction
of 48 % of the lignin initially present in corn stover. Operational details about this pretreatment
method and equipment design will be also highlighted here.

75

4.2. Introduction
From the understanding built in the past few years of how ammonia interacts with plant cell wall
components, it is clear that certain events have great impact in enzymatic digestibility of
lignocellulosic biomass. Recent work from Chundawat et al. [23, 105, 145] has suggested a
TM

mechanism of action for a leading ammonia pretreatment technology, designated as AFEX

.

During this pretreatment, moisture is added to the biomass, which allows an exothermal reaction
to occur when in contact with ammonia, generating a sudden increase of temperature that can go
up to 140 C. This interaction reduces the need for external heat during pretreatment, which
could be advantageous from the economic stand point. However, addition of water inhibits the
formation of cellulose III [80], which can be 2-5 times more digestible than native cellulose I,
when performing studies in Avicel [105, 117]. Therefore, AFEX

TM

pretreatment, similarly to

the other ammonia-based pretreatments proposed in the literature [146], is not able to transform
the crystalline structure of native cellulose, which is highly recalcitrant to cellulolytic enzymes.
Another important feature of ammonia as pretreatment chemical is the ability to cleave ester
bonds. These bonds are highly abundant in lignin-carbohydrate complexes (LCC) from grasses
and the cleavage of these cross-linkages is thought to help disrupt the complex network of
polymers that constitute the plant cell wall. Ammonolysis of ester bonds are known to be
facilitated by the presence of polar solvents such as water, which also can provide hydroxyl ions
capable of performing hydrolysis of these same bonds [95, 99]. However, little is known about
the conditions that favor these reactions in lignocellulosic biomass, especially in the absence of
water or other organic solvents.

76

Also according to Chundawat et al [23, 95, 145], ammonia is able to solubilize cell wall
components such as lignin and hemicellulose, which are condensed on the surface of the biomass
TM

after ammonia is vaporized by the sudden release of pressure during AFEX

pretreatment.

TEM visualization has demonstrated that this delocalization leaves behind porous structures and
delamination zones that are thought to expose cellulose to enzymatic activity, thereby
contributing to the increase of digestibility of AFEX

TM

pretreated biomass [145]. Lignin is

known to be a highly inhibitory component of the cell wall to enzymes and also to fermentative
microbes [29, 30]. Therefore, removal of lignin from the bioprocessing operations would
potentially benefit both enzymatic hydrolysis and fermentation.
To better integrate all these known benefits derived from ammonia interactions with the plant
cell wall, we propose to perform ammonia pretreatment in the absence of water, allowing
cellulose III to be formed. To allow this transformation, higher ammonia-to-biomass ratios are
required which can potentially benefit lignin solubilization and increase lignin removal from the
system. This novel pretreatment technology, denominated Extractive Ammonia (EA)
pretreatment, will be presented herein for the first time and this work will include the description
of the EA reactor and pretreatment methodology. This work will also present a global overview
of the important aspects of this new pretreatment technology. Using anhydrous ammonia in the
absence of water or other organic solvents, we will present how pretreatment conditions will
impact ester bond cleavage from ferulic and coumaric acids, and highlight the importance of this
event to enzymatic digestibility of corn stover. Further, lignin solubilization and extraction will
also be modeled as a function of pretreatment variables using a statistical design of experiments,
which will allow us to define the range of conditions that promote efficient lignin extraction.
77

Moreover, enzymatic digestibility will be studied as a function of pretreatment variables and
correlated with factors such as lignin extraction, cellulose III conversion and ester bond
cleavage. Optimization of pretreatment conditions that maximize sugar yields from corn stover
will be also presented in this study.
4.3. Experimental
4.3.1. Untreated corn stover
Untreated corn stover (UT-CS), a Pioneer 36H56 hybrid, was harvested in September 2009 in
Wisconsin (USA) and oven dried at 60 ºC for approximately 2 weeks. The biomass was further
passed through a 5 mm screen installed in a Christy hammer mill (Christison Scientific LTD,
England) and stored at 4 ºC in heat sealed bags prior to utilization. The moisture content of the
dried and milled corn stover was approximately 6 % on a wet weight basis. The biomass
composition analysis was performed using NREL protocols NREL/TP-510-42618 and
NREL/TP-510-42620. Based on this protocol, the untreated corn stover contained approximately
38 % glucan, 23 % xylan, 1 % galactan, 3 % arabinan, 14 % Klason lignin, 2 % acid soluble
lignin, 5 % ash and 15 % extractives (i.e. ethanol and water soluble compounds), on a dry weight
basis.
4.3.2. Extractive Ammonia (EA) pretreatment reactor construction
In Figure 4-1 is possible to visualize a schematic description of the EA equipment parts. This
equipment is composed of two parts, with distinct functions, i.e., the EA reactor (Figure 4-1 A),
where biomass is put in contact with ammonia for a desired temperature and residence time; and

78

the extractives collector (Figure 4-1 B), where the ammonia-soluble biomass components are
collected after filtration and also where ammonia is evaporated from the system.

Figure 4-1 Schematic drawing of the EA pretreatment reactor (A.) and extractives collector (B.).
The pictures of the reaction cell (6) and of the stainless steel filter installed in the cap of the
reactor (6’) are highlighted. The EA reactor outlet connector (8) is attached to the extractives
collector inlet connector (8’) during operation.

Extractive ammonia (EA) pretreatment reactors (Figure 4-1 A) were constructed using 33 mL
stainless steel cells from the Accelerated Solvent Extractor (ASE) system, manufactured by
Dionex (Thermo Scientific, Sunnyvale, CA, USA). These cells were modified appropriately to
79

be used independently from the ASE equipment and to be able to handle liquefied gases, such as
ammonia. The cells consist of two caps that screw into a threaded pipe. In the top cap, a
thermowell, a pressure gauge and a ball valve were installed. In the bottom cap, a sintered
stainless steel filter (80 m pore size) was installed to separate the solid biomass from the liquid
extractives during EA (6’). Also, the bottom cap holds a ball valve that is used to maintain the
reaction components in the cell during EA pretreatment.
The extractives collector (Figure 4-1 B) is composed of a sanitary tee tube of 1” diameter (11),
capped in the top with a sanitary fitting (9) that holds a deep tube (10), which is used to release
the extractives in the bottom of the collector after EA reaction is completed. During EA reaction,
the deep tube is fixed to the bottom of the EA reactor by joining together the “quick connectors”
8 and 8’. The side fitting of the collector holds a pressure gauge (12) and a 3-way valve (13),
followed by a metering valve (14). The 3-way valve was installed to permit a nitrogen inlet (16)
for pressurizing the collector and also to release the gases from the collector. The metering valve
function is to control the evaporation rate of the ammonia as well as the flow of gases through
the system.
This reactor system was designed to handle pressures up to 1500 psi and temperatures up to 140
C, holding approximately 3 g of biomass (dry weight) per reaction run (depending on the
biomass). This reactor system is flexible, as it can be adapted to several lengths of reactor pipe,
allowing the user to control the working reaction volume down to 1mL, if required. In our
design, five of these units were built in parallel and assembled in a permanent frame. In this
frame was also installed a customized heating block, made of aluminum that was machined to
perfectly enclose the five reactors (including the caps) simultaneously. The temperature was
controlled by a PID controller connected to the heating elements on the block and the respective
80

temperature sensors. This configuration allowed the possibility of performing small scale, highthroughput reactions for determination of optimum EA pretreatment conditions of various
substrates. However, the scope of this work will be centered on the understanding of important
factors that affect EA pretreatment effectiveness in both sugar conversions and lignin extraction
yields.
4.3.3. EA Pretreatment of corn stover
EA pretreatment was conducted in the tubular reactors described above and shown in Figure
4-1. The moisture in the corn stover was raised to 10 % (wet weight basis) by spraying
appropriate amounts of distilled water homogeneously throughout the corn stover (3.2 g on a dry
weight basis) before transferring it to the reactor cells (Figure 4-1, 6). The desired amount of
ammonia was loaded with a syringe pump (Harvard Apparatus - model PHD 2000, Holliston,
MA, USA) calibrated to deliver ammonia to a 0.01 g accuracy in the reactors, which were preheated to reaction temperature. Immediately after loading ammonia, the reactors were heated to
the desired temperature. The top fitting was connected to a nitrogen line (Figure 4-1, 1) and the
bottom fitting was connected to the extractives collector (Figure 4-1, 8 and 8’). The extractives
collector was pressurized with nitrogen to equalize the pressure developed in the reactor during
the course of pretreatment. After reaching the desired residence time, the valve between the
reactor and collector (Figure 4-1, 7) was opened and the extractives were filtered and drained
into the collector. Nitrogen overpressure of 1200 psi was applied to the reactor to maintain
ammonia in the liquid state before opening the outlet ball valve and metering valve connected to
the extractives collector (Figure 4-1, 13 and 14). The valve in the top of the reactor (Figure 4-1,
2) was kept opened to allow a pressure of approximately 1200 psi to be maintained in the system
81

for ~5 min. During this process, nitrogen was allowed to slowly flow through the system while
helping ammonia and the respective extractives to be filtered and drained to the collector. After 5
min, the nitrogen inlet valve was closed and the system was depressurized slowly (~3 to 5 min).
The pretreated biomass was transferred from the reactor to a tray, which was placed under the
hood overnight to remove any residual ammonia from the biomass. Note that the biomass was
fairly dry after EA pretreatment and this extra drying procedure was performed to stabilize the
biomass with respect to moisture and residual ammonia contents. After collecting the pretreated
biomass, the reactor was reconnected to the extractives collector and all the system lines were
cleaned with 70 % ethanol and 90 % acetone (both in water) to remove residual extractives from
the lines, which were drained to the extractives collector. After drying, the total weight and
moisture content of the EA pretreated corn stover (EA-CS) was measured with an analytical
balance and moisture analyzer A&D MX-50 (A&D Engineering, Inc., San Jose, CA, USA),
respectively. The samples were stored at 4 C in sealed bags prior to further usage in
compositional analysis and enzymatic hydrolysis studies.
4.3.4. Experimental design for pretreatment
A statistical design of experiments (DOE) was used in this study to evaluate the effect of
temperature, ammonia-to-biomass ratio (NH3:BM) and residence time on % lignin extraction
and ferulate/coumarate release during EA pretreatment of corn stover. The effect of the same
variables on the sugar release of EA-CS during enzymatic hydrolysis was also determined
herein. For this purpose, a Box Behnken experimental design was created using Minitab software
(Minitab Inc., State College, PA, USA), with two replicates, an alpha value of 1.15 and with high

82

and low values for each factor as shown in Table 4-1. The resulting experimental conditions and
respective experiment order are shown in Table 4-2.
A full quadratic response surface analysis was performed on the experimental results (notably
glucan and xylan conversions, % lignin extraction and ferulate/coumarate release during EA
pretreatment of corn stover) as a function of temperature, NH3:BM and residence time. All
interaction effects between factors were considered in this analysis and parameters were included
in the model based on their P value, as well as their influence on the predictive ability of the
model. The regression equations were used to predict the responses of the various effects as a
function of the pretreatment conditions within the boundaries set by the experimental design
shown in Table 4-1.

Table 4-1 High and low limits defined for the experimental points using Box Benhken
design.
Factor
Low
High
25
115
Temperature (C)
3
6
NH :BM
3

Time (min)

5

30

Table 4-2 EA pretreatment conditions defined by a Box Benhken design.
a
RunOrder
Time (min)
Temperature (C)
NH3:BM
1
25
6
30
2
25
6
30
3
70
4.5
17.5
4
25
3
5
5
25
6
5
6
25
3
5
7
70
4.5
17.5
8
25
6
5

83

Table 4-2 (cont’d)
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
a

115
115
70
115
70
70
25
70
115
115
115
70
115
115
25
70
70
70
70
70
18.25
70
121.75
70
70
121.75
70
18.25
70
70
70
70

3
3
4.5
3
4.5
4.5
3
4.5
6
6
6
4.5
6
3
3
4.5
4.5
4.5
2.775
4.5
4.5
4.5
4.5
4.5
6.225
4.5
4.5
4.5
4.5
4.5
2.775
6.225

NH3:BM – Ammonia to biomass ratio.

84

5
5
17.5
30
17.5
17.5
30
17.5
30
30
5
17.5
5
30
30
17.5
31.875
3.125
17.5
17.5
17.5
17.5
17.5
17.5
17.5
17.5
17.5
17.5
31.875
3.125
17.5
17.5

4.3.5. Composition analysis of plant cell wall components
UT-CS and EA-CS were subjected to compositional analysis according to the standard NREL
protocols NREL/TP-510-42618 and NREL/TP-510-42620. One exception for the analysis of
EA-CS was the absence of water and methanol extraction, to avoid removing water and ethanol
soluble lignin from the biomass and thereby overpredict % lignin extraction. Mass balances on
glucan, xylan, arabinan and lignin were performed around the EA pretreatment.
For determining ester-linked ferulic and coumaric acid contents remaining in the pretreated cell
walls, 40 mg of milled biomass (0.5 mm mesh size) was subjected to ethanol extraction by
adding 6 mL of 80% ethanol to a capped test tube. This mixture was incubated at 50 C for 1
hour with periodic mixing (every 15 min) to facilitate mass transfer. After incubation, the sample
was centrifuged for 4 min at 3500 rpm. The supernatant was aspirated from the tube prior to a
second addition of 6 mL of 80 % ethanol. The sample was gently vortexed at room temperature
and centrifuged for 4 min at 3500 rpm before the liquid was aspirated from the tube. To facilitate
drying, the sample was washed twice with 6 mL of 100 % ethanol. The extracted biomass was
left in a 50 C oven overnight to dry before digestion. Quantification of ester-linked
hydroxycinnamic acids in the biomass was performed using a 2 M sodium hydroxide (NaOH)
digestion as described in the literature [147, 148]. Diethylether extraction of the
hydroxycinnamic acids after NaOH digestion was not performed to avoid error propagation.
Instead, the alkaline mixture was filtered (0.22 m syringe filter) into a microcentrifuge tube and
diluted prior to quantification using LC-MS. LC-MS analyses were done using a QTRAP 3200
mass spectrometer (AB/Sciex) equipped with binary LC-20AD pumps (Shimadzu, Japan), a SILHTc autosampler and Ascentis Express C18 column (5 cm x 2.1 mm; 2.7 m particle size). The
85

analytes were eluted with a reversed phase gradient using Solvent A (0.15 % aqueous formic
acid) and Solvent B (methanol). Total solvent flow was maintained at 400 m/min, and a
gradient elution was performed using the following solvent compositions: initial, 5% B, held for
1 min; linear gradient to 18 % B at 8 min and then to 50 % B at 9 min; sudden increase to 99 %
B at 9.01 min and held until 10 min; and finally returned to initial condition at 10.1 min and held
until 12 min. Injection volume and column temperature were 5 L and 50 C, respectively.
Enhanced Product Ion (EPI) scans were generated for the [M+H]+ ions of coumaric and ferulic
acids using electrospray ionization in positive ion mode to identify abundant product ions. Parent
ions selected for fragmentation were the protonated molecular species of coumaric and ferulic
acids. This analysis was performed to select suitable product ions for Multiple Reaction
Monitoring (MRM) analysis. Source and gas parameters were all optimized for the standards to
generate signals with highest intensity. After optimization, source parameters were as follows:
curtain gas, 20 (arbitrary units); ion spray voltage, 4,500 V; temperature, 500oC; gas 1, 25
(arbitrary units); gas 2, 25 (arbitrary units). Compound dependent parameters after optimization
are shown in the table below:
Table 4-3 Compound-dependent parameters after mass spectrometry optimization.
Compound name
MRM transition
DP (V) EP (V) CE (V)
Coumaric acid
165 > 91
20
8
35
Ferulic acid
195 > 117
20
10
30
DP: Declustering potential; EP: Entrance potential; CE: Collision Energy

All mass spectrometric data were acquired and processed using Analyst v. 1.4.2 software from
AB/Sciex. The quantification of ester-linked hydroxycinnamic acids was performed in
quadruplicates for each EA-CS sample as well as the untreated corn stover control, using
appropriate standard curves. The results were further processed to calculate the percent of ester86

linked coumaric and ferulic acids that were cleaved during EA pretreatment, with respect to the
untreated corn stover.
4.3.6. Enzymatic hydrolysis of EA pretreated corn stover
Enzymatic hydrolysis was performed at 1 % glucan loading, using 15 mg of enzyme per gram of
glucan in 15 mL vials, incubated at 50 ºC, with pH 4.8 for 24 h in an orbital shaking incubator
(New Brunswick, USA). The enzymes utilized in this work were Cellic® CTec2 (138 mg
protein/ml,

batch

No.VCNI0001)

and

Cellic®

HTec2

(157

mg

protein/ml,

batch

No.VHN00001), generously provided by Novozymes (Franklinton, NC, USA). The protein
concentration for the enzymes was determined using the Kjeldahl nitrogen analysis method
(AOAC Method 2001.11, Dairy One Cooperative Inc., Ithaca, NY, USA). The enzyme ratios
utilized in this work were 50 % Cellic® CTec2 and 50 % Cellic® HTec2 in a dry protein weight
basis. These ratios were previously optimized to maximize total sugar conversion on EA
pretreated corn stover. After enzymatic hydrolysis, samples of the hydrolyzate were prepared for
glucose and xylose analysis using HPLC equipped with a Bio-Rad Aminex HPX-87H column
(Bio-Rad, Hercules, CA, USA) as previously described [88].
4.3.7. X-Ray Diffraction (XRD) of EA pretreated corn stover
XRD was performed on an X-ray powder diffractometer with beam parallelized by a Gobel
mirror (D8 Advance with Lynxeye detector; Bruker AXS Inc., Madison, WI, USA). Cuk
radiation (wavelenghth = 1.5418 Å) was generated at 40 kV and 40 mA. The detector slit was set
to 2.000 mm. Sample was analyzed using a coupled 2θ/θ scan type with a continuous PSD fast
scan mode. 2θ started at 8.000 and ended at 30.0277 with increments of 0.02151, while θ
started at 4.0000 and ended at 15.0138 with increments of 0.01075. Step time was 1.000 sec
87

(i.e., 1025 total steps, effective total time 1157 sec per run). EA-CS samples (approximately 0.5
g) were placed in the specimen holder ring made of PMMA, with 25 mm diameter and 8.5 mm
height, rotating at 5/min during analysis.
4.4. Results and Discussion
4.4.1. Extraction of p-coumarate and ferulate from corn stover during EA pretreatment
4.4.1.1. Analysis of ester-linked p-coumarate and ferulate in untreated and EA pretreated corn
stover
Similarly to other warm season grasses, maize contains a considerable amount of ester-linked
phenolic compounds in the plant cell wall [149, 150], notably hydroxycinnamic acids like pcoumaric (pCA) and ferulic acid (FA). pCA is known to be ester-linked to lignin [151] and there
are no evidences in the literature suggesting that this compound has an important role in crosslinking cell wall polymers. In the other hand, FA is described in the literature as being an
important bridge, cross-linking arabinoxylan polymers (in the case of diferulates) or even crosslinking arabinoxylans with lignin polymers [152, 153]. By cleaving these ester bonds, is possible
to disrupt the ultra-structure of the plant cell wall and therefore, is possible to improve enzyme
accessibility to carbohydrate substrates. Most chemical pretreatments have the ability to cleave
ester linkages, as both acidic and alkaline conditions result in hydrolysis reactions yielding
phenolic acids as the respective products. Ammonia based treatments like EA can also break
ester linkages through ammonolysis reactions that yield products like phenolic amides [23, 95,
97]. To evaluate the effect of pretreatment conditions on pCA and FA cleavage, a 2 M NaOH
treatment at room temperature was performed on UT-CS (control) as well as EA-CS pretreated
88

under the various conditions (see Box Benhken design on Table 4-2). From this analysis, UT-CS
showed a pCA content of 14.17 mg/g of dried corn stover and an ester-linked FA content of 5.66
mg/g of dried corn stover. pCA and ester-linked FA contents in corn stover vary significantly
across the various plant tissues. For example, Akin et al. (2006) [148] observed variations of
pCA from 1.5 mg/g (leaf-blade lamina) to 32.1 mg/g (stem rind) in a Pioneer 3085 corn hybrid.
In the same study, ester-linked FA varied from 2.5 mg/g (leaf-blade lamina) to 6.6 mg/g (stem
rind). The corn stover used in this study contains a milled mixture of the various fractions of the
plant and therefore, the measured amounts of esterified hydroxycinnamic acids are a weighted
average from the harvested plant. Literature reports show a large range of pCA and FA contents
for the whole corn plant, which depends on the hybrid and maturity of the plant [148, 150, 154].
The values measured in this study for pCA and FA in untreated corn stover were similar to the
numbers reported in the literature.
The esterified pCA and FA contents of EA-CS varied significantly with respect to the
pretreatment conditions. It is important to note that the measured ester-linked pCA and FA from
EA-CS, is a result of alkaline hydrolysis of previously extracted samples with ethanol. This
extraction method was used to remove residual pCA and FA that were cleaved by ammonia, but
not totally extracted from the biomass during pretreatment, and therefore provide means to
quantify the ester-bound hydroxycinnamic acids in EA-CS. The % cleavage of ester-linked
hydroxycinnamic acids after EA pretreatment increased as a function of pretreatment severity,
ranging from 5.70 % to 71.96 % for pCA and 0.13 % to 76% for FA (% loss referenced to the
ester-linked pCA and FA in UT-CS). This result suggests that under certain conditions, ammonia
is able to effectively cleave ester linkages in the absence of polar solvents, which are often
known to catalyze ammonolysis reactions [99]. To further understand how the EA pretreatment
89

variables affect ester bond cleavage and further extraction of from the plant cell wall during EA
pretreatment, a mathematical model based on statistical DOE was implemented. The results for
this model will be further discussed below.
4.4.1.2. Response surface analysis: % cleavage of p-coumarate and ferulate as a function of
pretreatment conditions.
To obtain a mathematical model describing how ester linkages are cleaved as a function of EA
conditions, 40 experimental data points were acquired for pCA and FA % cleavage and were
subjected to surface response analysis using a full quadratic regression. The terms considered for
the regression model were based on their P value (P < 0.05) and also based on the predictive
2

ability of the resulting model, considering R values, variance and residual analysis. The
statistical results obtained in this study are shown in Figure 4-2 and Figure 4-3, which include
2

residual plots, histograms, probability plots, regression coefficients and predicted R values for
% cleavage of pCa and FA as a function of the terms considered for each model. The residual
plots show that in both cases the residuals are mostly within  6 % range. The histograms do not
always show a perfect “bell” shaped curve centered in zero, however the greater fraction of the
residuals are present between  2.5%, as can be observed by the normal probability plots. The
2

goodness of the fits can be evaluated by measuring the R values shown in Figure 4-2 B and
2

Figure 4-3 B. The predicted R (R-Sq (pred)) value for the regression of % pCA cleavage was
98.11 %, while for % FA cleavage was 97.32 %, using the coefficients for the terms shown in
Figure 4-2 B and Figure 4-3 B. These values indicate the ability of the models to predict
2

2

responses for new observations, and since they are similar to the R and adjusted R values, the
90

models do not seem to overfit and have adequate predictive ability. The resulting response
surface equations for % pCA and % FA cleavage are given below:

% pCA Cleavage = (-24.7923) + (0.6028 X1) + (3.2131 X2) + (1.3931 X3) –
(0.001 X1X1) – (0.0192 X3X3) + (0.0053 X1X3) – (0.125 X2X3)

(4-1)

% FA Cleavage = (-14.2344) – (0.204 X1) + (2.771 X2) + (0.8119 X3) + (0.0048 X1X1)

(4-2)

– (0.0201 X3X3) + (0.0112 X1X3) – (0.0696 X2X3)
Where,
X1 = Temperature (C) [18.25 to 121.75 C]; X2 = NH3:BM [2.78 to 6.23]; X3 = Time (min)
[3.13 to 31.88 min].

Analyzing the regression coefficients for pCA % cleavage it is possible to observe that
temperature, NH3:BM and time are important factors that explain the variability of the
experimental

data

(i.e.,

P

<

0.05).

Interaction

factors

NH3:BM*NH3:BM

and

Temperature*NH3:BM did not seem to explain the variability of the experimental data, showing
P values above 0.05 and did not contribute to the improvement of the model. Therefore, these
terms were not considered in Equation (4-1).

91

Figure 4-2 (A) Residual plots for the regression of ester-linked pCA % cleavage as a function of
pretreatment variables (temperature, time and NH3:BM ratio) and respective interaction terms.
Residuals and fitted values are expressed in (% cleavage). Regression terms were only
considered if their P value was smaller than 0.05. (B) Estimation of regression coefficients and P
values for the terms affecting ester-linked p-coumaric acid % cleavage. Parameters that
determine the goodness of the fit are also shown.
92

Figure 4-3 (A) Residual plots for the regression of ester-linked FA % cleavage as a function of
pretreatment variables (temperature, time and NH3:BM ratio) and respective interaction terms.
Residuals and fitted values are expressed in (% cleavage). Regression terms were only
considered if their P value was smaller than 0.05. (B) Estimation of regression coefficients and P
values for the terms affecting ester-linked ferulic acid % cleavage. Parameters that determine the
goodness of the fit are also shown.
93

The model used to predict ester-linked FA % cleavage considered the same terms used in the
pCA % cleavage model. An important aspect that was taken in consideration during the
formulation of this model is related to the interaction term NH3:BM*time, that shows a P value
above 0.05 (Figure 4-3 B), suggesting that this term is not related to ester-linked FA % cleavage.
2

However, during optimization of the model, this term impacted the goodness of the fit (R ), the
2

predictive ability of the model (predicted R ) and residual values. Therefore, this term was
included in Equation (4-2).
To simplify this discussion and visualize the effects of the various pretreatment parameters on
the cleavage of ester-linked pCA and FA during EA pretreatment, graphical representations of
the two models are shown in Figure 4-4 A and Figure 4-4 B, respectively. The contour plot
from Figure 4-4 A-I predicts a gradual increase in ester bond cleavage from pCA with respect to
temperature for a fixed NH3:BM loading of 6:1. It is also possible to observe a positive effect of
pretreatment time on pCA % cleavage from the same counter plot. In Figure 4-4 A-II, it can be
observed an almost linear increase of pCA % cleavage with respect to temperature and
practically no dependence of NH3:BM loading for a fixed time of 30 min. At the same time,
Figure 4-4 A-III shows some dependence between pCA % cleavage and NH3:BM loading for
the lower residence times, while at the higher residence times (approaching 30 min)

this

dependence is significantly decreased (for a fixed temperature of 115 C). Taking in
consideration that this reaction is heterogeneous, is possible that this profile is a result of
transport phenomena, where a longer time is required for the ammonia to reach the inner pores of
the biomass particles for the lower NH3:BM loadings. However, detailed kinetic studies should
94

be performed to confirm this dependence and properly determine transport and reaction
parameters, which are not included in this study.
A similar reaction profile was achieved for FA cleavage. However, major differences between
rates of FA and pCA cleavage seem to be associated with kinetic parameters, notably activation
energy (Ea), as the temperature dependence profile is the most distinct between the two
reactions. For example, it is clear that the ester bonds linking FA to carbohydrates start to be
effectively cleaved at temperatures above 40 C, while ester-linked pCA is effectively cleaved
even at 20 C. Also, is visible that at higher temperatures, similar percentages of ester bonds are
cleaved in FA as in pCA. This observation supports the idea that the differences between these
two reactions could derive mainly from differences in Ea, as is clear that the rate of cleavage of
ester bonds from FA is more sensitive to temperature (Figure 4-4 A-I and B-I).

95

A

TimeContour
= 30 min.
115°C
Plot of p-coumaricTemperature
acid % loss vs=L/S,
Temperature

Contour Plot of p-coumaric acid % loss vs time,Contour
Temperature
Plot of p-coumaric acid % loss vs L/S, Temperature
Contour Plot of p-coumaric acid % loss vs time, L/S

5.5

NH3L/S
:BM

5.0

20

4.5

15

3.5

5

3.0
20

40

60

80

100

120

6.0
25

5.5
5.0

Hold Values
L/S 6

4.0

10

30

II

L/S

I

25

(min.)
Timetime

p-coumaric
acid %
loss
< 10
10 – 20
20 – 30
30 – 40
40 – 50
50 – 60
60 – 70
> 70

6.0

(min.)
Timetime

NH3:BM= 6/1

30

4.5

40

60

80

15
10

4.0
20

20

p-coumaric

p-coumaric

p- Coumaric
acid Acid
% loss
p-coumaric
< 10.0
Cleavage
(%)
acid %
10.0loss
– 20.0

IIIacid % loss
10
20
30
40
50
60

<
–
–
–
–
–
–
>

10
20
30
40
50
60
70
70

– 30.0
<20.0
10
30.0 – 40.0
– 50.0
10 –40.0
20
50.0 – 60.0
60.0 – 70.0
20 – 30
> 70.0
30 Hold
– Values
40
Temperature 115
40 – 50
50 – 60
60 – 70
> 70

Hold Values
time 30

5
100

120

3.0 3.5 4.0 4.5 5.0 5.5 6.0
Hold Values
3.5
Temperature
time 30
NH3L/S
:BM
Temperature (°C)
3.0
Contour Plot of
ferulic
acid %
loss vs time, Temperature
Contour Plot of
ferulic
acid
%
loss
vs L/S,
Temperature
Contour
Plot%
of ferulic
acid
% loss
vs time, L/S
NH
:BM=
6/1
Time
=
30
min.
Temperature
=L/S,
115°C
Contour
Plot
of p-coumaric
acid
loss vs
Temperature
3
20
40
60 ferulic80
100
120
acid
ferulic acid
ferulic acid %
Ferulic
Acid
30 I
6.0 II loss
30
% loss
p-coumaric
Temperature
III%<loss10.0
< 10.0
6.0
Cleavage
acid %(%)
loss
< 10.0
Temperature

5.5
5.0

20

4.5

15

4.0

10

–
–
–
–
–
–
>

10.0
20.0
30.0
40.0
50.0
60.0

NHL/S
3:BM

(min.)
Timetime

25

20.0
30.0
40.0
50.0
60.0
70.0
70.0

5.5
5.0
L/S

B

4.5

(min.)
Timetime

Temperature (°C)

4.0

3.0
20

40

60

80

Temperature

100

Temperature (°C)

120

20

40

60

80
100
3.5
Temperature (°C)
3.0
Temperature

20
15

10.0
20.0
30.0
40.0
50.0
60.0

–
–
–
–
–
–
>

20.0
30.0
40.0
50.0
60.0
70.0
70.0

10.0
20.0
30.0
40.0
50.0
60.0

Hold Values
time 30

5
120

–

20.0

< –1030.0
– 40.0
10 – –2050.0
– 60.0
20 – –3070.0
> 70.0
30 Hold
– Values
40
Temperature 115
40 – 50
50 – 60
60 – 70
> 70

10

3.5 HoldL/SValues
6

5

25

3.0 3.5 4.0 4.5 5.0 5.5 6.0

NH3L/S
:BM

Hold Values
time 30

40
60
80
100
120
Figure 4-4 Contour plots describing % cleavage20of ester-linked
p-coumaric
(A) and ferulic acid
Temperature
(B) as a function of EA pretreatment conditions. Experimental values were calculated based on
the content of ester-linked p-coumaric and ferulic acids present in corn stover prior to EA
pretreatment. The contour plots represent a small area of the universe described by the respective
mathematical models, with fixed values of NH3:BM = 6/1 (I), time = 30 min (II) and
temperature = 115 C (III).

4.4.2. Extraction of lignin from corn stover during EA pretreatment
4.4.2.1. Analysis of lignin extraction in untreated and EA pretreated corn stover
The lignin present in EA-CS was quantified using the NREL protocol, excluding both ethanol
and water extraction steps, as these solvents can remove lignin components from the pretreated
biomass. If we included these steps in the composition analysis method, we would be
overpredicting lignin extraction during EA pretreatment. The untreated corn stover was analyzed
96

according to the original protocol and showed 14 % Klason lignin and 2 % acid soluble lignin
contents. The % lignin extraction was calculated based on the lignin content determined before
and after pretreatment.
During EA pretreatment of corn stover, several components of the biomass were extracted, of
which lignin was the most abundant. Lignin extraction values ranged from 16.1 to 45.5 wt% of
the original lignin present in untreated corn stover. Ammonia is known to be a good solvent for
lignin extraction and has been used for lignocellulose pulping, usually in aqueous solutions with
and without potassium hydroxide [155]. It has also been used in the supercritical state for
pulping feedstocks like bamboo or yellow poplar [156]. A recent study from Zhongguo Liu and
co-workers (2012) [157], showed that by using aqueous ammonia at concentrations around 30
wt% and temperatures of 120 C was possible to extract about 61.7 wt% of the lignin from
Miscanthus. From the same study, a temperature of 180 C achieved a delignification of 76.9
wt%. However, such conditions also promoted the decomposition and extraction of carbohydrate
components, notably cellulose (~88 to ~94 wt% recovery) and hemicellulose (~39 to ~69 wt%
recovery). The technology proposed in this manuscript (i.e., EA pretreatment) attempts to
maximize lignin extraction while preserving the structural carbohydrates intact in the solid
phase, which can then be converted to fermentable sugars during enzymatic hydrolysis. In
Figure 4-5 it is possible to compare the composition of EA-CS referenced with respect to UTCS. In this chart, EA pretreatment was performed under conditions that maximize lignin
extraction within the boundaries of this study and the limitations of the equipment, i.e.,
temperature = 120 C, NH3:BM = 6:1 and time = 30 min. From these results, it is clear that
while lignin was the major component removed from corn stover, structural carbohydrate
97

composition was not significantly reduced during EA pretreatment (< 5 % reduction).
Comparing these results to the study by Liu and co-workers (2012) [157], where lignin from
Miscanthus was extracted using aqueous ammonia, EA pretreatment reveals lower effectiveness
for biomass delignification (~44 wt% compared with ~62 wt% at similar processing conditions).
However, an important advantage is that anhydrous ammonia appears to be more selective to
lignin extraction than aqueous ammonia. For example, Liu and co-workers (2012) [157] showed
cellulose and hemicellulose recoveries of 94.40.9 % and 69.21.0 %, respectively, while the
lowest glucan and xylan recoveries obtained during EA pretreatment were 93.6  0.2 % and 95.6
 0.4 %, respectively. Other authors that used aqueous ammonia solutions for delignification,
often in supercritical state, observe lower recoveries of hemicellulose than cellulose [158]. This
trend was not observed when treating corn stover with anhydrous ammonia under sub-critical
conditions, as both xylan and glucan recoveries were approximately equal.
Composition of EA-CS with respect to UT-CS
100%
16%

16%

% Composition

80%

3%

9%
3%

23%

Lignin

22%

60%

40%

Extracted

Arabinan

Xylan

38%
36%

Glucan
20%
20%

15%

Others

0%

UT-CS

EA-CS

Figure 4-5 Stacked bar chart comparing the mean values for the composition of EA pretreated
corn stover (EA-CS) with respect to the untreated corn stover (UT-CS). The pretreatment
98

Figure 4-5 (cont’d)
condition used for this chart was temperature = 121.8 C , NH3:BM loading of 6.2:1 and
residence time of 32 min, which represents the highest severity within the boundary conditions
defined for EA pretreatment in this work. In this analysis, only structural carbohydrates from the
plant cell wall and Klason lignin were considered. Coefficients of variation were below 2.5 % for
all measurements.

4.4.2.2. Response surface analysis: % lignin extraction as a function of pretreatment
conditions
A mathematical model was formulated based on 40 experimental points expressing the %
delignification at various pretreatment conditions (see Table 4-2 for the experimental design). A
surface response analysis using a quadratic regression model was performed using the terms with
a P value below 0.05 and also other terms that improved the predictive ability of the model,
2

considering predicted R values, variance and residual analysis of the regression. The statistical
results obtained in this study are shown in Figure 4-6. The residual plots in Figure 4-6 A show
that the residuals for the regression model were mostly within  5 % range. The predicted R

2

2

value of this regression was 84.83 %, which was relatively similar to the R (90.30 %) and
2

adjusted R (88.54 %) values. These results suggest that the model has good ability to predict
unknown observations and is not overfitting experimental data.
The residuals are randomly spread around the regression line, implying the absence of a
systematic lack-of-fit. The histogram does not show a perfect “bell shape” curve, however most
of the residuals have a value close to zero. The highest number of residuals are located in the
negative region of the histogram, however many of these negative residuals correspond to
99

replicates of the middle point condition, which fit below the regression curve as it can be
observed in the residual (“versus fits”) plot from Figure 4-6 A.

Figure 4-6 (A) Residual plots for the regression of % lignin extraction as a function of
pretreatment variables (temperature, time and NH3:BM ratio) and respective interaction terms.
100

Figure 4-6 (cont’d)
Residuals and fitted values are expressed in (% extraction). (B) Estimation of regression
coefficients and P values for the terms affecting % lignin extraction. Parameters that determine
the goodness of the fit are also shown.

Figure 4-6 B shows the terms that were considered in the mathematical model describing %
lignin extraction as a function of EA pretreatment conditions. The constant and NH3:BM terms
have a P value higher than 0.05, suggesting that are not related to % lignin extraction. However,
they were included in the regression model as they impacted substantially the goodness of the fit
2

to the experimental data and the predictability of the model (predicted R ). The interaction terms
temperature*time and NH3:BM*time were not considered in the model as their P value was
higher than 0.05 and they did not improve the fit of the model to the experimental data. As a
result, the mathematic equation that best described % lignin extraction as a function of
pretreatment conditions is as follows:

% Lignin Extraction = (4.27832) - (0.21178 X1) + (5.48885 X2) + (1.433314 X3) +
(0.0048 X1X1)

- (0.96579 X2X2) - (0.03028 X3X3) +

(4-3)

(0.07482 X1X2)
Where,
X1 = Temperature (C) [18.25 to 121.75 C]; X2 = NH3:BM [2.78 to 6.23]; X3 = Time (min)
[3.13 to 31.88 min].

101

A series of contour plots expressing % lignin extraction as a function of pretreatment conditions
are shown in Figure 4-7. These contour plots represent a region of the universe described by
Equation (4-3), with fixed parameters at 120 C, NH3:BM loading of 6:1 and 30 min of
residence time. Figure 4-7 I shows the increase in delignification as a function of temperature
and time, where a minimum delignification of ~11% (at 18.25 C and 3 min) and a maximum
delignification of ~48% (at 120 C and 25 min) were achieved according to the regression
model. Kinetic models based on ammonia pulping experiments often show three phases of
delignification, with distinct rates and temperatures at which delignification occurs. Lignin is a
heterogeneous polymer, with various levels of hydrophilicity, molecular weights and diverse
chemical linkages. Therefore, the decomposition and solubilization of these various fractions of
lignin require distinct conditions. In contrast with most pulping methods that use hydroxyl ions
as a primary reactant, EA pretreatment uses anhydrous ammonia, which acts as a solvent and a
reactant. Ammonia is expected to cleave ester bonds in the biomass but not the abundant ether
bonds that support the lignin structure. Pulping methods that use high concentrations of hydroxyl
ions are able to cleave ether bonds, reducing the lignin molecular weight and thereby facilitating
solubilization as well as extraction [159]. Lower molecular weights and hydrophilic lignins
should also be more susceptible to ammonia solubilization and it is possible that these are the
native properties of the lignin extracted at lower temperatures and residence times. However,
future research should be performed to confirm this hypothesis. p-Coumarate ester bond
cleavage during EA pretreatment occurs at mild conditions, as can be observed in Figure 4-4 A.
This fact suggests that even at these conditions ester-linked components of the lignin are being
cleaved and solubilized by ammonia. As a general observation, is important to mention that in
102

the literature, lignin from grasses such as rice straw or wheat straw can be delignified with
milder temperatures and shorter residence times compared to wood [155, 156, 160]. This
difference is often attributed to a higher porosity observed in grasses that facilitates the
penetration of the pulping chemicals and differences in lignin chemistry [155, 160]. These
distinct properties of grass lignins compared with lignins from woods are becoming more evident
in recent years. For example, recent work developed by Jose C. del Rio and coworkers (2012)
[161] about characterization of lignin from wheat straw, showed a substantial level of tricin units
incorporated in the lignin, which has not been observed in wood lignin.

Contour Plot of Lignin %loss vs L/S, Temperature

Contour Plot of Lignin %loss vs time, Temperature
Contour Plot of Lignin %loss vs L/S, Temperature Contour Plot of Lignin %loss vs time, L/S
3

Time = 30 min.

NH :BM= 6/1

5.5

25

5.0

20
L/S

(min.)
Timetime

II
6.0

15

5.5

Hold Values
L/S 6

5.0

L/S

3.0

20

40

60

80

100

Temperature
Temperature
(°C)

120

<
30 –
35 –
40 –
>

20

%loss

30
35
40
45
45

30 –
35 –

4.5
20

40

Hold Values
time 30

15

Hold Values
time 30

5
60

80

100

4.0Temperature (°C)

120

35
40

<40 –3045
> 45
30 – 35
Hold Values
35 Temperature
– 40 115
40 – 45
> 45

10

3.5

5

Lignin
Lignin
Lignin
%loss
30
Extraction <(%)

Lignin

III%loss

25

4.5
4.0

10

30

(min.)
Timetime

6.0

I

NH3:BM

30

Temperature = 115°C

Lignin
%loss
< 30
30 – 35
35 – 40
40 – 45
> 45

3.0 3.5

Temperature

4.0 4.5 5.0 5.5 6.0

NHL/S
3:BM

3.5 % lignin extraction as a function of EA pretreatment
Figure 4-7 Contour plots describing
conditions. Experimental values were calculated based on the lignin content in corn stover prior
3.0 represent a small area of the universe described by the
to EA pretreatment. The contour plots
respective mathematical models, with fixed values of NH3:BM = 6/1 (I), time = 30 min (II) and
20
40
60
80
100 120
temperature = 115 C (III).
Temperature

Figure 4-7 II and III shows the expected profile, where delignification is improved with
increasing temperature and NH3:BM loading. These contour plots show that efficient
delignification (>45 %) occurs at temperatures ranging between 115 and 120 C with residence
times above 15 min and NH3:BM loadings above 5.5:1. This result suggests that delignification
103

is favored by high liquid-to-solid ratios and higher temperatures, which consequently generates
higher operating pressures (~1200 psi), increasing pretreatment capital cost. However, the effect
of delignification on downstream processing, including enzymatic hydrolysis and fermentation
must be evaluated before a balanced economic assessment is done. Moreover, lignin valorization
is an area of great economic potential [26, 162] and the development of new technologies for
lignin upgrading could be instrumental towards the feasibility of this process.
4.4.3. Enzymatic hydrolysis of EA pretreated corn stover
4.4.3.1. Enzymatic digestibility of EA pretreated corn stover
Enzymatic digestibility of the EA-CS was evaluated as a function of pretreatment conditions. In
this work, monomeric glucan and xylan conversions after 24h enzymatic hydrolysis were
calculated based on the composition of the EA pretreated biomass. However, the carbohydrate
loss is not significant in most of the pretreatment conditions, reaching a maximum of 6.4 ± 0.2 %
glucan and 4.4 ± 0.4 % xylan for the most severe conditions (i.e. Temperature = 121.8 C ,
NH3:BM loading of 6.2:1 and residence time of 32 min). The variables temperature, NH3:BM
loading and residence time are known to significantly impact the physicochemical properties of
ammonia treated cell walls [145]. By varying ammonia pretreatment conditions it is possible to
generate a wide variety of properties, which include cellulose III conversion, degree of
delignification, level of ester bond cleavage, porosity and level of inhibitory compounds for
microbes and enzymes [23, 95, 105, 145]. These properties have significant effects on enzymatic
digestibility of the plant cell wall and therefore, pretreatment can be manipulated to create
favorable conditions for enzymatic action. Not all these properties will be discussed and
104

analyzed here, however, it is important to discuss those aspects that are unique in this novel
ammonia-based pretreatment, notably cellulose III conversion and degree of delignification. The
importance of ester bond cleavage will be also discussed herein, as it shows a good correlation to
enzymatic digestibility of corn stover.
From Figure 4-8 (A1, B1 and C1) it is possible to observe glucan and xylan conversions (24h
enzymatic hydrolysis) achieved by EA-CS pretreated in a wide range of temperatures, NH3:BM
loadings and residence times. These conditions were established by a Box Behnken design, as
described previously. The same EA pretreated samples were subjected to XRD analysis to
differentiate the crystalline state of cellulose. From these analyses, a series of spectra were
generated and are presented in Figure 4-8 (A2, B2 and C2). A substantial amount of work has
been done on cellulose III, where improvements of 2-5 fold have been observed for monomeric
glucose conversions in isolated cellulose (e.g. Avicel). However, to our knowledge there are no
reports of the effect of cellulose III conversion on the enzymatic hydrolysis of lignocellulosic
biomass, including corn stover, which will be discussed herein.

105

A
1
VV
IV
IV

Xylan
Xylan
Xylan

C
1

1

AFEX™

V

V

IV

IV

IV

III

III

III

II

II

II

I

I

I

IIII

0

II

2
2020

B

V

III
III

00

Glucan
Glucan
Glucan

20

40
60
24h Conversion (%)

80

0

100

% Conversion

25 °C; 6:1;

20

40
60
24h Conversion (%)

80

100

0

% Conversion

2

20

40

60

80

100

% Conversion

2

V

70 °C; 6.225:1;
17.5 min.

V

121.75 °C; 4.5:1;
17.5 min.

V

25 °C; 6:1;
5.0 min.

IV

70 °C; 4.5:1;
31.875 min.

IV

115 °C; 6:1;
30.0 min.

IV

25 °C; 3:1;
30.0 min.

III

70 °C; 4.5:1;
17.5 min.

III

115 °C; 6:1;
5.0 min.

III

25 °C; 3:1;
5.0 min.

II

70 °C; 4.5:1;
3.125 min.

II

115 °C; 3:1;
30.0 min.

II

18.24 °C; 4.5:1;
17.5 min.

I

70 °C; 2.5:1;
17.5 min.

I

115 °C; 3:1;
5.0 min.

I

40 min.6060
40
8080
30.0
24hConversion
Conversion(%)
(%)
24h

100
100

Control Cellulose III

Control Cellulose III

Control Cellulose III

Control Cellulose I

Control Cellulose I

Control Cellulose I

5

10

15

20

2θ

25

30

5

10

15

20

2θ

25

30

5

10

15

20

25

30

2θ

Figure 4-8 Conversions to monomeric glucose and xylose after 24 h enzymatic hydrolysis of various EA-CS samples (A1-C1).
Glucan conversion of optimally pretreated AFEX-CS is indicated by a green dashed line. XRD spectra from EA-CS as a function of
pretreatment conditions (A2-C2). EA conditions are located in the top left corner of each XRD spectra and include temperature;
NH3:BM loading and time, in this respective order. Cellulose I and III controls are Avicel and ammonia treated Avicel, respectively.
106

One of the major bottlenecks when analyzing cellulose crystallinity from lignocellulosic biomass
is that there is a lack of quantitative methods to measure amorphous cellulose, cellulose I and III
when they are simultaneously present in same sample. Besides the peak convolution problem in
techniques like XRD, the major difficulty is associated with the presence of lignin and other
amorphous components that compose lignocellulosic biomass, which may vary significantly
between samples that were treated under different regimes. Biomass bleaching followed by FTRaman is a potential technique that can help solve this problem in the future [163, 164].
However, as this method is still not developed for cellulose III, it could not be applied in the
present research. Despite the analytical limitations, it is possible to perform a qualitative
evaluation of the effects that each pretreatment condition had on the crystallinity of the
pretreated sample by XRD analysis. From the results shown in Figure 4-8, it was possible to
conclude that cellulose III is formed for all the conditions within the boundaries of this study.
One of the important requirements of cellulose III conversion is that the biomass must be in
contact with liquid ammonia, in the absence of moisture [80]. However, our work shows that
even at an NH3:BM ratio of 2.5:1 and 10% moisture in the corn stover, some level of cellulose
III conversion occurred (Figure 4-8 B2-I), revealed by the presence of a peak at 2θ = ~20.5
(highlighted by a red dashed line). Nevertheless, it is clear that higher liquid-to-solid ratios
accelerate the formation of cellulose III in corn stover by analyzing the XRD spectra in Figure
4-8 A2-II to A2-V and C2-I to C2-IV. As was previously observed in Avicel (CHAPTER 1),
lower temperatures in the vicinity of 25 C are sufficient to observe cellulose III conversion,
however it was clear that higher temperatures favor the rate at which cellulose III is converted.
For example, at 6:1 NH3:BM loading cellulose I was not fully transformed into cellulose III at
107

25 C and 5 min residence time (Figure 4-8 A2-IV), while at 115 C, total conversion to
cellulose III is shown by the disappearance of a peak at 2θ = ~22.5 (highlighted by a grey
dashed line) (Figure 4-8 D2-III). Also at higher temperatures, it is clear that the XRD peaks
from cellulose III become sharper. This phenomenon is a result of higher cellulose III conversion
(as discussed in CHAPTER 1) and of the lower lignin content of samples that were pretreated at
higher temperatures (Figure 4-7). The combination of these two properties (i.e., cellulose III
formation and lignin removal) was expected to improve enzymatic digestibility of lignocellulosic
biomass and, indeed, these results suggest that samples with higher levels of cellulose III
conversion and lower lignin content are the most susceptible to enzymatic digestion of the
carbohydrates, as demonstrated by the sugar conversions from Figure 4-8 (A1-C1). The lowest
glucan and xylan conversions obtained in this study were 38.7 % and 22.1%, respectively, which
corresponded to an EA pretreatment performed at 25 C, with 3:1 NH3:BM loading and 5 min of
residence time (Figure 4-8 A-II). From the XRD spectra, this sample has low crystallinity and
contains a mixture of cellulose I and III from the presence of convoluted peaks in the region of
2θ between ~20.5 and ~22.5. The highest sugar yields were observed at higher temperatures and
liquid–to-solid ratios, from EA conditions C-IV and C-V. Both conditions had similar enzymatic
hydrolysis results, showing 89.0 % and 88.1 % of glucan conversion, respectively. Xylan
conversions were 74.3 % and 73.4 %, respectively. The XRD spectra of these samples confirm
total cellulose III conversion and lower levels of amorphous components in the biomass
compared to other conditions, which shows that a substantial amount of lignin is removed during
EA pretreatment at higher temperatures. Even though XRD data does not allow a quantitative
correlation between cellulose III conversion and sugar yields, lignin correlations with enzymatic
108

hydrolysis performance can be measured to support our hypothesis that lignin removal is an
important factor that enhances enzymatic hydrolysis yields in EA pretreatment. Figure 4-9
correlates % glucan conversion with % lignin extraction, where we observe a positive correlation
2

between these two variables. The R shown in Figure 4-9 measures the variability in the data set
that is accounted by the linear regression model in the same figure. Therefore, this linear model
can explain 52.9 % of the variability in the experimental points. This result can be explained by
the fact that lignin removal is not the only factor contributing to higher cellulose digestibility;
however, is certainly a factor that contributes positively to glucan conversion, as the P value for
this regression term was 0.000. The P value is the observed significance level of the test, i.e., the
of % Glucan
Conversion
with
Lignin
Extraction
hypothesis that there is a Correlation
correlation between
glucan conversion
(%)
and%
lignin
extraction
(%).
% Conv Glucan 24 hr = 21.80 + 1.459 Lignin %loss

(%)
Glucan
Conversion
% Glucan
Conversion

100

Regression
95% CI
95% PI

80

S
R-Sq
R-Sq(adj)

60
40
20

Y = 21.80 + 1.459 X

0
15

20

10.9057
52.9%
51.7%

R2 = 52.9%
P = 0.000

25 30 35 40 45
% Lignin
Extraction
Lignin
Extraction
(%)

50

Figure 4-9 Correlation of % glucan conversion as a function of % lignin extraction. The
regression had a P value = 0.000 for the term % lignin extraction, suggesting that lignin
extraction is related to % glucan conversion. The regression (black line), 95 % confidence
interval (red line) and 95 % prediction interval (green line) are shown in the graphic.

As the P value is lower than the significance level (i.e., 0.05), we can reject the null hypothesis
that states that there is no significant correlation between the two variables in the test. The
109

correlation between glucan conversion and lignin extraction was previously observed by other
authors when performing other pretreatments of lignocellulosic biomass [157, 165].
Another factor that is important for sugar conversion is the level of ester bond cleavage in the
biomass. The cleavage of ferulic acid and p-coumaric acid by ammonia are critical events for
disrupting the ultra-structure of the plant cell wall, including cross-links between lignin and
carbohydrates. In AFEX

TM

pretreatment, these reactions help to release cell wall components

that are further solubilized and deposited in the surface of the biomass [145]. In EA pretreatment,
is likely that these same components are being extracted from the cell wall, creating pores and
better access routes for enzymes to target the carbohydrates, similarly to AFEX

TM

[145].

Previous work has correlated enzymatic digestibility with ferulate and coumarate cleavage for
TM

AFEX

[23], however this correlation was not yet been confirmed under operating conditions

such as the ones used during EA pretreatment. Therefore, the results obtained for ferulate and
coumarate cleavage during EA pretreatment were correlated with glucan and xylan conversions
(Figure 4-10).
In all cases presented in Figure 4-10, it is possible to observe good linear correlation with R

2

values between 73.5 % and 93.1 %. The P values for all linear regressions were 0.000, which
means that the terms ferulic acid and p-coumaric acid % cleavage are related with glucan and
xylan conversions. It is also possible to observe that p-coumaric acid cleavage seems to be a
2

better indicator of sugar conversion than ferulic acid, given the higher R values and better fit to
the linear regression (Figure 4-10 B and D). This result is derived from the fact that the points
of low ferulic acid cleavage do not follow the linear trend of the remaining points and therefore,
110

a quadratic regression would result in a better fit for these correlations (Figure 4-10 A and C).
However, the main purpose of these correlations is to prove the importance of such events during
ammonia pretreatment, providing insight to the scientific community about the aspects that most
contribute to biomass recalcitrance, so they can be addressed properly in future research.
Fitted Line Plot

Fitted Line Plot

% Conv Glucan 24 hr = 49.88 + 0.5917 ferulic acid % loss

80

S
R-Sq
R-Sq(adj)

Glucan
Conversion
% Conv
Glucan 24 hr (%)

Glucan
Conversion
% Conv
Glucan 24 hr (%)

100

% Conv Glucan 24 hr = 35.50 + 0.7720 p-coumaric acid % loss

A

60
40
20

R2 = 80.8%
P = 0.000

Y = 49.88 + 0.59 X
0
10

80

S
R-Sq
R-Sq(adj)

B

4.17641
93.1%
92.9%

60
40
20

R2 = 93.1%
P = 0.000

Y = 35.50 + 0.77 X

20

30

40

50

60

70

80

Xylan
Conversion
% Conv
Xylan 24 hr (%)

S
R-Sq
R-Sq(adj)

C

60
40
20

R2 = 73.5%
P = 0.000

Y = 35.12 + 0.59 X

0
0

0

Fitted
Line
acidCleavage
%
lossPlot (%)
Ferulicferulic
Acid
% Conv Xylan 24 hr = 35.12 + 0.5941 ferulic acid % loss

100

Xylan
Conversion
% Conv
Xylan 24 hr (%)

80.8%
80.3%

0
0

80

100
6.95857

10

20

30

40

50

60

70

% loss
Ferulicferulic
AcidacidCleavage
(%)

20

30

40

50

60

70

80

100
8.60174
73.5%
72.9%

80

S
R-Sq
R-Sq(adj)

D

60
40
20

R2 = 90.4%
P = 0.000

Y = 19.63 + 0.80 X

0

80

10

Fitted
Line
Plot
p-coumaric
acid
% loss
p-Coumaric
Acid
Cleavage
(%)
% Conv Xylan 24 hr = 19.63 + 0.8018 p-coumaric acid % loss

0

10

20

30

40

50

60

70

80

p-coumaric acid % loss
p-Coumaric
Acid Cleavage (%)

Figure 4-10 Linear regressions correlating glucan (A) and xylan (C) conversions with ferulic
acid cleavage, as well as correlating glucan (B) and xylan (D) conversions with p-coumaric acid
cleavage. Regression equations P values and R2 values are indicated in the bottom of each
graphic.

111

5.11282
90.7%
90.4%

4.4.3.2.Response surface analysis of enzymatic hydrolysis performance of EA pretreated corn
stover
A response surface analysis was also performed on glucan and xylan conversion as a function of
pretreatment conditions. The same Box Behnken experimental design used in lignin and ester
bond cleavage analysis was applied for sugar conversion. Enzymatic hydrolysis results were
obtained after 24 h incubation with an optimized enzyme cocktail as described in the Methods
section. Figure 4-11 shows the residual plots, histograms and the coefficients associated with the
terms that better describe glucan conversion with respect to EA pretreatment conditions. From
the residual plots, we can observe that the experimental points are relatively dispersed and
evenly cover all the range of fitted values. There is no recognizable pattern in the residual plot
(versus fits), however there is a region between 45 and 70% conversion where the all residuals
are negative. This indicates that in that region the model overpredicts glucan conversion.
Nevertheless, the residuals vary between -6 and 7 % and the highest residuals were observed
between 5 and 70% glucan conversion. The histogram shows a good “bell shape” curve centered
on zero, which means that most of the values fitted the model well. The regression coefficients
used in this model are presented in Figure 4-11 B and were the coefficients that gave a better fit
to the experimental data. Most of the terms used in this model have a P value below 0.05 with
the exception of the constant (P value = 0.327) term and time (P value = 0.079). This suggests
that these two terms do not relate to glucan conversion. However, the absence of these terms in
the mathematical model impacts negatively the fit to experimental data. Considering these terms,
2

2

the predicted R value for this model is 94.73 %, which is similar to the R (96.64 %) and
2

adjusted R (96.03 %) values. This result suggests that the model has a good fit to the
112

experimental data. The resulting model equation that predicts 24h % glucan conversion as a
function of EA pretreatment variables is as follows:

24h % Glucan Conversion = (-8.4167) - (0.7285 X1) + (12.3293 X2) + (0.2021 X3) (4-4)
(0.0025 X1X1) - (1.1312 X2X2) + (0.0028 X1X3)
Where,
X1 = Temperature (C) [18.25 to 121.75 C]; X2 = NH3:BM [2.78 to 6.23]; X3 = Time (min)
[3.13 to 31.88 min].

113

Figure 4-11 (A) Residual plots for the regression of 24h % glucan conversion to monomeric
glucose as a function of pretreatment variables (temperature, time and NH3:BM ratio) and
respective interaction terms. Residuals and fitted values are expressed in (% conversion). (B)
Estimation of regression coefficients and P values for the terms affecting % glucan conversion.
Parameters that determine the goodness of the fit are also shown.

114

The same methodology was applied to develop a regression model that predicts xylan conversion
with respect to pretreatment variables. Figure 4-12 shows the residual plots and histogram for
the regression, as well as the coefficients from the terms used in this model. From the residual
we can observe that the residuals show a small range, between -4 and 4%. This is a good
indication that all the predicted values are close to the experimental values. The residuals are
randomly spread; however there is a slight tendency to observe greater residual values at higher
xylan conversions. One other remark about this model is that, in the region between 40 and 50 %
xylan conversion, there are no experimental data to fit the curve. Therefore, the model could not
be properly tested in that region of the curve, which could lead to a flawed prediction. However,
as this is a relatively small area of xylan conversions, the model was still considered acceptable.
The histogram shows a profile close to a normal distribution, which indicates the model
predictions are more often closer to the experimental results than the opposite. The terms used in
the regression model are presented in Figure 4-12 B, as well as the respective coefficients. Most
of the terms have a P value below 0.05, with the exception of NH3:BM and NH3:BM*NH3:BM,
2

which were included in the model due to the fact that their absence decreases the predicted R

value and increases the amplitude of the residuals. Nevertheless, these terms seem to be
unrelated to 24h % xylan conversion from the P value analysis. For this model, the P value for
2

2

the F test was 0.000 and the predicted R value was 96.78 %, which was close to the R (98.15
2

%) and adjusted R (97.74 %) values. Therefore, these results suggest that the model shows a
good fit to experimental data, as well as good ability to predict new observations.

115

Figure 4-12 (A) Residual plots for the regression of 24h % xylan conversion to monomeric
xylose as a function of pretreatment variables (temperature, time and NH3:BM ratio) and
respective interaction terms. Residuals and fitted values are expressed in (% xylan conversion).
(B) Estimation of regression coefficients and P values for the terms affecting % xylan
conversion. Parameters that determine the goodness of the fit are also shown.

116

The resulting model equation that predicts 24h % xylan conversion as a function of EA
pretreatment variables is as follows:

24h % Xylan Conversion = (-19.8537) - (0.9504 X1) + (6.6245 X2) + (0.6163 X3) (4-5)
(0.004 X1X1) - (0.4878 X2X2) - (0.0144 X3X3) + (0.0033 X1X3)
Where,
X1 = Temperature (C) [18.25 to 121.75 C]; X2 = NH3:BM [2.78 to 6.23]; X3 = Time (min)
[3.13 to 31.88 min].

To visualize the effects of EA pretreatment variables on both 24h glucan and xylan conversions a
series of contour plots were prepared, as shown in Figure 4-13. The plots cut a section at fixed
values of (I) NH3:BM = 6:1, (II) time = 30 min and (III) temperature = 115 C, respectively.
Figure 4-13 A shows the dependence of 24h % glucan conversion with respect to temperature
and time. As EA pretreatment time and temperature increases, glucan conversion also increases
progressively, achieving values above 90 % conversion after 25 min, at 120 C. A similar profile
is observed with respect to temperature and NH3:BM loading. However, NH3:BM does not seem
to impact glucan conversion very significantly for loadings above ~ 4:1, at fixed temperature and
30 min of residence time. At high temperature (115 C), we can observe an increase of glucan
conversion with respect to time and also NH3:BM, and it is possible to achieve conversions
above 90 % using NH3:BM ratios of 4:1 and 30 min or residence time. Similar conversions can
117

be achieved at higher NH3:BM loadings and lower residence times as shown in Figure 4-13 AIII.

Contour Plot of % Conv Glucan 24 hr_1 vs time, Temperature
Contour Plot of % Conv Glucan 24 hr_1 vs L/S, Temperature
Contour Plot of % Conv Glucan 24 hr_1 vs time, L/S
3

Time = 30 min.

6.0

I

5.0

NH3:BM

20

L/S

time

25

Time (min.)

II

5.5

15

4.5

Hold Values
L/S 6

4.0

10

3.5

5

3.0
20

40

60

80

100

120

20

40

Temperature = 115°C

Contour Plot of % Conv Glucan 24 hr_1 vs time, L/S
% Conv
% Conv
30 IIIGlucan
Glucan
Glucan 24
% C onv
24 hr_1
hr_1
30
Conversion
Glucan 24(%)
< 40
hr_1 < 40
40 – 50
25
50 – 60
– 50
<40 40
60 – 70
50 – 60
25
70 – 80
40
–
50
20
80 – 90
60 – 70
> 90
50 –70 60
– 80
Hold Values
20
– 90
15
60 –80 70
time 30
> 90
70 – 80
Hold Values
10
80
– 90 115
15
Temperature
> 90
5
time

30

% Conv
Glucan
24 hr_1
< 40
40 – 50
50 – 60
60 – 70
70 – 80
80 – 90
> 90

Time (min.)

NH :BM= 6/1

time

A

60

80

10

100

120

3.0 3.5 4.0 4.5

Temperature
Temperature (°C)

Temperature (°C)
Temperature

5.0 5.5 6.0

Hold Values
Temperature 115

NHL/S
3:BM

5

30 min.
Temperature
=6.0
115°C
4.0
4.5
5.0% Conv
5.5 Xylan
% Conv
Xylan6/1
24 hr vs time, Temperature
Contour Plot ofTime
% Conv=Xylan
24 hr 3.0
vs L/S,3.5
Temperature
Contour
Plot of
24 hr vs time, L/S
BContour Plot ofNH
3:BM=
Contour Plot of % Conv Xylan 24 hr vs time, Temperature
Xy lan
II% Conv
24 hr

5.5

NHL/S
3:BM

Timetime(min.)

25
20
15
10

<
–
–
–
–
–
>

20
30
40
50
60

5.0
4.5
4.0

30

30

20
30
40
50
60
70
70

25

25
20

3.5 Hold Values
L/S 6

5

(min.)
Timetime

6.0

time

I

30

40

60

80

Temperature

100

Temperature (°C)

120

15

20.0
30.0
40.0
50.0
60.0

<
–
–
–
–
–
>

20.0
Xylan
24–hr 30.0
30.0
40.0
50.0
60.0

40

60

Hold Values
time 30

10

8010 100

Temperature

Temperature (°C)

120

3.0 3.5

–

40.0

< –2050.0
– 60.0
20 – –3070.0
> 70.0
30 Hold
– Values
40
Temperature 115
40
– 50
50 – 60
60 – 70
> 70

5
20

% Conv

Xylan
Xy lan 24 hr
% Conv
< (%)
20.0
Conversion

20.0
30.0
40.0
50.0
60.0
70.0
70.0

15

3.0
20

20

% Conv

IIIL/S
Xy lan 24 hr

4.0 4.5 5.0 5.5 6.0
L/S

Hold Values
L/S 6

NH3:BM

5

Figure 4-13 Contour plots describing monomeric 24h % glucan (A) and % xylan (B)
20
40
60
80
100 120
conversions as a function of EA pretreatment conditions. Temperature
Experimental values were calculated
based on the glucan and xylan contents of the EA pretreated biomass. Maximum glucan and
xylan loss during pretreatment was 6.4 % and 4.4 %, respectively, at maximum severity
conditions (i.e. temperature = 121.8 C, NH3:BM loading of 6.2:1 and residence time of 32 min).
The contour plots represent a small area of the universe described by the respective mathematical
models, with fixed values of NH3:BM = 6/1 (I), time = 30 min (II) and temperature = 115 C
(III).

From Figure 4-13 B-I, it is possible to observe that xylan conversion varies significantly with
respect to temperature, but not as significantly with respect to time. Monomeric xylan

118

conversions can reach values above 70% for temperatures of approximately 90 C and residence
times of 25 min. However, the reaction time can be reduced to approximately 10 min by
increasing the temperature, while achieving similar monomeric xylan conversions. It is important
to note that the maximum experimental value for xylan conversion was 74.3  0.59 % and the
region > 70 % xylan conversion is already near the maximum achieved during enzymatic
hydrolysis. In Figure 4-13 B-II, xylan conversion is predicted as a function of temperature and
NH3:BM ratio. As observed before, temperature is a major factor contributing to the
effectiveness of pretreatment, as demonstrated by the gradual increase in xylan conversion. The
NH3:BM ratio does not play a significant role at lower temperatures, however at higher
temperatures, high NH3:BM ratios are important to maximize xylan conversion. For example, at
~90 C and 6:1 NH3:BM ratio is possible to achieve xylan conversions above 70 %, however, by
increasing temperature to 120 C the same conversions can be achieved at 3.5:1 NH3:BM
loading. From Figure 4-13 B-III, it can be observed that at 115 C, both time and NH3:BM
loading contribute to improve xylan conversions from ~50 % to > 70 %. At this temperature, it is
possible to achieve > 70 % monomeric xylan conversions at 3.5 NH3:BM ratio and 30 min
residence time. The same conversion can be obtained with progressively lower residence time as
we increase ammonia loading during EA pretreatment. From this analysis, we can observe that
while maximum glucan conversion is possible in a very narrow range of conditions,
maximization of xylan conversion can be obtained in a much broader range, overlapping the
conditions that maximize glucan conversion.
119

4.4.4. Model integration and process optimization
Finally, it is important to integrate the knowledge obtained for the various elements studied
herein and elaborate on the predicted conditions that maximize monomeric glucan and xylan
conversions, as well as lignin extraction.
The results presented in this work clearly define ester bond cleavage and lignin extraction as
major events that relate to improved digestibility of EA-CS. However, from our analysis, ester
bond cleavage does not necessarily correlate with higher lignin extraction [results not shown].
This fact results from the presence of variables that are important to extract lignin, such as
NH3:BM ratio, which do not impact ester bond cleavage in such a high extent (within the
boundaries of this study). Cellulose crystallinity index (CrI) coupled with cellulose III formation
are also important factors that drive monomeric sugar conversion during enzymatic hydrolysis.
Though cellulose III conversion does not seem to depend on lignin extraction, CrI of the biomass
is highly dependent of the presence of lignin, since lignin is a major amorphous component in
the biomass. Lignin negatively impacts enzymatic hydrolysis [30, 165] and therefore, higher
crystallinity of the biomass sample often benefit its digestibility if cellulose is not being
decrystalized. One big advantage of EA compared to major lignin extraction methods is that it is
highly selective to lignin, leaving the carbohydrates behind for further processing to fermentable
sugars during enzymatic hydrolysis. This property is extremely advantageous, because lignin
extraction by EA pretreatment will not carry other detrimental events such as loss of
hemicellulose, which would affect the final sugar yields.
From the model equations determined in this study, it is possible to calculate maximum
monomeric sugar conversion and the associated lignin extraction yields. Also, the conditions that
120

will allow this can be determined within the boundaries of our experimental design. Therefore,
theoretical results show that the maximum monomeric glucan conversion (24h of enzymatic
hydrolysis) for EA pretreatment is about 93% and xylan conversion is approximately 79%. The
pretreatment conditions that allow maximum sugar yields is 121.8 C, 6.2:1 NH3:BM loading
and 30.2 min of residence time, according to the regression model. At this condition, the model
also predicts a lignin extraction of approximately 48%.
Choosing a 3:1 ammonia loading while maintaining the same temperature of 121.8 C and 30.2
min of residence time, the model predicts a slight decrease in glucan conversion to ~87 %, while
xylan conversion drops to ~68 %. Sugar yields may not be impacted greatly according to the
model; however, lignin extraction drops significantly to ~29% yield by reducing NH3:BM
loading to 3:1.
The models presented in this work probably cannot be generalized to all maize varieties, since is
possible to observe high variability between biomass samples, including within the same species.
They also cannot be generalized to all enzymatic hydrolysis time points and glucan loadings.
However, the models provide useful insights about how sugar yields and lignin extraction
depend on pretreatment variables. The models also show how ester bond cleavage depends on
the operating conditions during pretreatment and how they correlate to sugar yields. More
importantly, they are a useful tool to help optimize pretreatment conditions and prepare
economic analyses for a future EA-based biorefinery.
Further studies are required to fully unveil the mechanism by which EA pretreatment operates.
These studies might include imaging techniques, porosity analysis and proper measurement of
cellulose III conversion in pretreated biomass. However, the study presented herein provides
121

clear guidelines about some of the elements that contribute to EA pretreatment effectiveness,
which could be potentially used to further improve this pretreatment methodology of
lignocellulosic biomass.
Even though this pretreatment method constitutes a novel approach to ammonia pretreatment,
with beneficial results for both enzymatic hydrolysis and lignin extraction, some drawbacks are
present and should be considered for future work. One important topic to consider is the elevated
operating pressure during this pretreatment, reaching 1200 psi for the optimal conditions. Highpressure systems are common in industry; however process economics and ammonia safety
would benefit from decreasing operating pressure, which can be achieved to a certain extent by
the addition of a polar co-solvent during pretreatment, which does not compromise cellulose III
conversion (e.g. ethanol).
4.5. Conclusion
EA pretreatment is a novel ammonia-based pretreatment technology that shows significant
potential to decrease the recalcitrance of lignocellulosic biomass toward enzymatic digestion.
This benefit is created by the conversion of cellulose III, disruption of important cross-links
present in the biomass (e.g. ester-linked ferulates and coumarates), followed by lignin
solubilization and extraction. All these factors show good correlation with sugar yields,
suggesting that a mechanistic understanding of EA process should involve a deeper analysis of
how these events are interdependent and synergize to facilitate biomass deconstruction. The
combination of these features is unique from this pretreatment technology and has been proposed
here for the first time.

122

A statistical design of experiments was also implemented to predict monomeric glucan and xylan
conversions, lignin extraction yields and ester bond cleavage in corn stover as a function of EA
pretreatment variables (i.e., temperature, NH3:BM ratio and time). The resulting regression
models were evaluated based on the goodness of the fit, residual plots and histograms, before
being considered reliable. The regression models show that ester bond cleavage in the plant cell
wall does not depend on NH3:BM loading as much as temperature or time (within the boundaries
of this study). In the other hand, lignin extraction and sugar yields increase significantly with
increasing values of temperature, NH3:BM loading and time. Sugar yields are maximized at
121.8 C, 6.2:1 NH3:BM loading and 30.2 min of residence time, according to our model,
yielding ~93 % monomeric glucan conversion and 79 % monomeric xylan conversion after 24h
enzymatic hydrolysis at 1 % glucan loading. Under these pretreatment conditions, the model
also predicts extraction of approximately 48 % of the lignin present in untreated corn stover,
which could be further upgraded to fuels and chemicals. Characterization of this lignin stream is
important to define its value and potential for further usage in the biorefinery. The valorization of
lignin will be instrumental to offset some of the negative aspects related to this technology,
which impact the economics of the process in comparison to other ammonia pretreatments (e.g.
AFEX™). A major drawback of this pretreatment is the fact that it operates at high pressure, up
to 1200 psi at the most severe conditions tested herein. However, it is possible to develop
strategies to further improve the EA method and decrease the operating pressure by the usage of
co-solvents, such as ethanol. Higher ammonia loadings are also another important drawback,
which will require attention in future research. Detailed economic evaluation and process
123

engineering studies should be performed in future research to define strategies for minimizing
ammonia loading and operating pressure during EA pretreatment, while achieving similar levels
of ester bond cleavage, lignin solubilization and cellulose III conversion.

124

APPENDIX

125

Table 4-4 Pretreatment conditions used for the Box Benhken design and the respective
number for reference of model fit plots below.
Temperature NH3:BM
time
Experiment Number
18.25
25
25
25
25
70
70
70
70
70
115
115
115
115
121.75

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

4.5
3
3
6
6
2.775
4.5
4.5
4.5
6.225
3
3
6
6
4.5

17.5
5
30
5
30
17.5
3.125
17.5
31.875
17.5
5
30
5
30
17.5

% Conversion

Glucose Conversion: Model Fits
100
90
80
70
60
50
40
30
20
10
0

Experimental

1

2

3

4

5

6

7

Model

8

9 10 11 12 13 14 15

Experiment Number
Figure 4-14 Model prediction fits to experimental data for 24h glucose conversion as a function
of EA conditions.
126

% Conversion

Xylose Conversion: Model Fits
100
90
80
70
60
50
40
30
20
10
0

Experimental

1

2

3

4

5

6

7

Model

8

9 10 11 12 13 14 15

Experiment Number
Figure 4-15 Model prediction fits to experimental data for 24h xylose conversion as a function
of EA conditions.

Lignin Extraction: Model Fits

% Lignin Extraction

60

Experimental

Model

50
40
30
20
10
0
1

2

3

4

5

6

7

8

9 10 11 12 13 14 15

Experiment Number
Figure 4-16 Model prediction fits to experimental data for lignin extraction as a function of EA
conditions.
127

p-Coumarate Ester Cleavage: Model Fits
80

Experimental

Model

% Cleavage

70
60
50
40
30
20
10
0
1

2

3

4

5

6

7

8

9 10 11 12 13 14 15

Experiment Number
Figure 4-17 Model prediction fits to experimental data for p-coumarate ester cleavage as a
function of EA conditions.

Ferulate Ester Cleavage: Model Fits
80

Experimental

Model

% Cleavage

70
60
50
40
30
20
10
0
1

2

3

4

5

6

7

8

9 10 11 12 13 14 15

Experiment Number
Figure 4-18 Model prediction fits to experimental data for ferulate ester cleavage as a function
of EA conditions.
128

CHAPTER 5 –

HIGH

SOLID

LOADING

ENZYMATIC

HYDROLYSIS

OF

EXTRACTIVE AMMONIA (EA) PRETREATED CORN STOVER

5.1. Abstract
In this work, the performance of EA pretreated corn stover (EA-CS) during enzymatic hydrolysis
will be compared with AFEX

TM

-CS at high solid loadings and evaluated as a function of 1)

solid loading, 2) enzyme loading and 3) hydrolysis time. A statistical design of experiments (Box
Benhken) was applied for these three variables and a full quadratic regression was performed to
determine model equations that statistically represent monomeric glucose and xylose conversions
as a function of those variables and their respective interaction parameters. To fulfill the
feedstock demand for these experiments, a larger scale EA pretreatment reactor was also
constructed. This new EA system operation was validated for corn stover and the results were
successfully compared with the model predictions defined in CHAPTER 4 for sugar conversion
and lignin extraction. The appropriate EA pretreatment condition was defined based on sugar
yields and lignin extraction. Enzymatic hydrolysis analysis was performed using model
TM

predictions and the benefits of EA pretreatment were compared with AFEX
analysis, EA-CS presents better performance than AFEX

TM

. From this

at high solid loadings, where 95 %

glucose and 85 % xylose conversions can be achieved for 30 mg/g glucan of enzyme loading at
15 % solids, comparing with 85 % and 70 % obtained for AFEX

TM

-CS, respectively. The most

significant benefit of EA pretreatment was observed at 7.5 mg/g glucan enzyme loadings for 15
% solid loading, where absolute improvements reached approximately 23 % and 15 % for
129

glucose and xylose conversion, respectively, compared with AFEX

TM

-CS. Under these

conditions, EA-CS can reach approximately 88 % and 76 % glucose and xylose conversion,
respectively. This result can only be achieved by AFEX

TM

-CS when enzyme loadings are above

20 mg/g glucan, which represents a 2.7 fold increase in enzyme loading.

5.2. Introduction
The novel Extractive Ammonia (EA) pretreatment is known to improve enzymatic hydrolysis of
lignocellulosic biomass (CHAPTER 4) by extracting lignin from the plant cell wall, while
simultaneously converting naturally occurring cellulose I (CI) to cellulose III (CIII). Modern
pretreatment technologies are required to enhance enzymatic hydrolysis at high solid loadings, as
commercial processes require high concentration of sugars in the hydrolysate to potentiate high
biofuel concentrations after fermentation. This requirement is important to facilitate biofuel
separation and purification. For example, economically viable distillation of ethanol in a
biorefinery requires a minimum concentration of 4 wt% ethanol after fermentation [166] and to
achieve this concentration, a minimum of 6 % glucan loading (assuming a corresponding 4 %
xylan loading) is required during enzymatic hydrolysis [167]. However, high solid loading
enzymatic hydrolysis has its own bottlenecks, as sugar yields tend negatively affected by
increasing solids loading [168-170], a phenomenon usually described as the “solids effect” [170].
Though authors have correlated the “solids effect” with transport phenomena, water activity,
product inhibition, lignin inhibition and pretreatment decomposition products inhibition [2, 169,
170], the reason behind the “solids effect” is still not well understood. In the scope of this work,
enzymatic hydrolysis at high solid loadings will be studied for EA pretreated corn stover (EA130

CS). The enzymatic hydrolysis profile will be evaluated as a function of solids loading, enzyme
loading and incubation time during enzymatic hydrolysis. For this purpose, a statistical design of
experiments will be implemented followed by regression using a full quadratic equation. The
performance of EA-CS during enzymatic hydrolysis will be compared with the same feedstock
pretreated using a leading ammonia-based pretreatment technology designated by AFEX
TM

AFEX

TM

.

technology is currently on the path for commercialization by Michigan Biotechnology

International (MBI) and is known to be highly effective on grasses, such as corn stover,
achieving glucose conversions that range 75-80 % during high solid loadings enzymatic
hydrolysis [78, 171]. From the empirical regression models developed for EA and AFEX

TM

pretreated corn stover, it is possible to compare the kinetics of enzymatic hydrolysis for various
solid loadings and enzyme loadings. The improvements offered by EA pretreatment technology
during high solids loading enzymatic hydrolysis will be analyzed in detail as a function of
enzyme loading and solids loading.

5.3. Experimental
5.3.1. Untreated corn stover
Untreated corn stover (UT-CS), a Pioneer 36H56 hybrid, was harvested in September 2009 in
Wisconsin (USA) and oven dried at 60 ºC for approximately 2 weeks. The biomass was further
passed through a 5 mm screen installed in a Christy hammer mill (Christison Scientific LTD,
England) and stored at 4 ºC in heat sealed bags prior to utilization. The moisture content of the
dried and milled corn stover was approximately 6 % on a wet weight basis. The biomass
131

composition analysis was performed using NREL protocols NREL/TP-510-42618 and
NREL/TP-510-42620. On a dry weight basis, the untreated corn stover contained approximately
38 % glucan, 23 % xylan, 1 % galactan, 3 % arabinan, 14 % Klason lignin, 2 % acid soluble
lignin, 5 % ash and 15 % extractives (i.e. ethanol and water soluble compounds).
5.3.2. Extractive Ammonia (EA) pretreatment reactor
For this work, a new EA reaction system was constructed based on the design presented on
Section 4.3.2. This version is composed of 3 reactors of ~600 mL each, which are capable of
working in parallel and extracting biomass components to a single extraction collector. This new
system is capable of producing larger amounts of pretreated biomass, which can be used in high
solid loading enzymatic hydrolysis and fermentation studies. Each reactor is made of 316
stainless steel sanitary tubing of 2” x 12” (OD x H) and can pretreat ~40 g of corn stover per
batch, per reactor. Sanitary tubes are connected to inlet and outlet fittings through high-pressure
bolted clamps, sealed with PTFE gaskets, which can withstand a maximum pressure of 1200 psi.
The bottom fitting clamped to the reactor was customized to fit a stainless steel filter of 80 μm
pore size, sealed with a custom made PTFE gasket. This connection is linked to a valve that
separates the reactors from the extractives collector. All three reactors are equipped with
individual heating mantles, connected to a single PID controller (Chemglass Life Sciences,
Vineland, NJ, USA) that can be programmed to control temperature of each reactor
independently.
The ammonia delivery system is composed of a tubular high-pressure vessel made of 316
stainless steel, containing liquid ammonia at room temperature. This vessel is secured on top of a
balance (A&D, Inc., San Jose, CA, USA) and connected to the high-pressure reactors through
132

flexible tubing made of stainless steel and lined with PTFE. Ammonia delivery is measured by
weight displacement from the vessel located on top of the balance to the pretreatment reactors.
For accurate delivery measurements all lines from the ammonia vessel to the reactors must be
primed with liquid ammonia.

Figure 5-1 Picture of scaled-up EA pretreatment system, composed by 3 high-pressure reactors
of 600 mL each, ammonia delivery system and extractives collector.
The extractives collector is also made with stainless steel 316 sanitary tubing of 2” x 12” (OD x
H), connected to a sanitary tee tube on the top. A deep tube was installed from the top fitting of
the sanitary tee tube to be able to drain the extractives from the reactor to the bottom of the
collector. A valve was installed on the side fitting of the tee tube, which can be used to control
the ammonia evaporation rate during extractives condensation. Heating tape was installed around
133

the outer surface of the extractives collector to better control ammonia evaporation, as it absorbs
energy from the surroundings, resulting in a significant temperature drop. Figure 5-1 shows a
picture of the reactor system fully assembled.
5.3.3. Pretreatment of corn stover
EA pretreatment was conducted in the tubular reactors described above (Figure 5-1). The
moisture content of corn stover was raised to the desired level by spraying distilled water
homogeneously throughout the untreated biomass (40 g in dry weight basis) before transferring it
to the reactors. The desired amount of ammonia was loaded from a high pressure vessel located
on the top of a balance, which measured the displaced weight of ammonia transferred to the
reactors. Immediately after loading ammonia, the reactors were heated to the desired
temperature. Set point temperatures were achieved in < 7 min. The top fitting of the reactor was
connected to a nitrogen line and the bottom fitting was connected to the extractives collector.
The extractives collector was pressurized with nitrogen to equalize the pressure developed in the
reactor during the course of pretreatment. After reaching the desired residence time, the valve
between the reactor and collector was opened and the extractives were filtered and drained to the
collector. Nitrogen overpressure of 1200 psi was applied to the reactor to maintain ammonia in
the liquid state during biomass extraction. The valve in the top of the reactor was kept opened to
allow a pressure of approximately 1200 psi to be maintained in the system for ~10 min. During
this time, nitrogen was allowed to slowly flow through the system by slightly opening the
ammonia exhaust valve on the top of the collector. After 10 min, the nitrogen inlet valve was
closed and the system was depressurized slowly (~5 min). The pretreated biomass was
transferred from the reactor to a tray, which was placed under the hood overnight to remove any
134

residual ammonia. The biomass was fairly dry after EA pretreatment and this drying procedure
was performed to stabilize the biomass with respect to moisture and residual ammonia contents.
After collecting the pretreated biomass, the reactor was reconnected to the extractives collector
and all the system lines were cleaned with 70 % ethanol and 90 % acetone (both in water) to
remove residual extractives from the lines, which were drained to the extractives collector. After
drying, the total weight and moisture content of the EA pretreated corn stover (EA-CS) was
measured with an analytical balance and moisture analyzer A&D MX-50 (A&D Engineering,
Inc., San Jose, CA, USA), respectively. The samples were stored at 4 C in sealed bags prior to
compositional analysis, following NREL protocols NREL/TP-510-42618 and NREL/TP-51042620, and enzymatic hydrolysis. EA pretreatment conditions used in this work were designed to
evaluate the effect of moisture content, temperature, ammonia loading and residence time on
enzymatic hydrolysis and lignin extraction performance in this new reactor system, and to
compare them with results obtained in the smaller scale reactors in CHAPTER 4.
Ammonia Fiber Expansion (AFEX

TM

TM

) pretreated corn stover (AFEX

-CS) was provided by

Michigan Biotechnology International (MBI). AFEX pretreatment conditions were optimized by
MBI to maximize monomeric sugar conversion during enzymatic hydrolysis.
5.3.4. X-ray powder diffraction (XRD)
XRD was performed on an X-ray powder diffractometer with beam parallelized by a Gobel
mirror (D8 Advance with Lynxeye detector; Bruker, Bruker AXS Inc., Madison, WI, USA).
CuKα radiation (wavelength = 1.5418 Å) was generated at 40 kV and 40 mA. The detector slit
was set to 2.000 mm. Sample was analyzed using a coupled 2θ/θ scan type with a continuous
PSD fast scan mode; 2θ started at 8.000° and ended at 30.0277° with increments of 0.02151°,
135

while θ started at 4.0000° and ended at 15.0138° with increments of 0.01075°. Step time was
1.000 sec (i.e., 1025 total steps, effective total time 1157 sec per run). Cellulose samples
(approximately 0.5 g) were placed in a specimen holder ring made of PMMA with 25 mm
diameter and 8.5 mm height, rotating at 5 degrees per minute during analysis.
5.3.5. Enzymatic hydrolysis for EA pretreatment optimization
Enzymatic hydrolysis was performed at 1 % glucan loading, using 15 mg of enzyme per gram of
glucan in 15 mL vials, incubated at 50 ºC, with pH 4.8 for 24 h in an orbital shaking incubator
(New Brunswick, USA). The enzymes used in this work were Cellic® CTec2 (138 mg
protein/ml, batch No.VCNI0001), Cellic® HTec2 (157 mg protein/ml, batch No.VHN00001),
generously provided by Novozymes (Franklinton, NC, USA), and Multifect Pectinase (MP) (72
mg/ml, batch No. 4861295753) a gift from Genencor/Dupont (Palo Alto, CA, USA). The protein
concentration of the enzymes was determined using the Kjeldahl nitrogen analysis method
(AOAC Method 2001.11, Dairy One Cooperative Inc., Ithaca, NY, USA). The enzyme ratios
utilized in this work were 50 % Cellic® CTec2, 25 % Cellic® HTec2 and 25 % MP in a dry
protein weight basis. These ratios were previously optimized to maximize total sugar conversion
on EA pretreated corn stover. After enzymatic hydrolysis, samples of the hydrolyzate were
prepared for glucose and xylose analysis using HPLC equipped with a Bio-Rad Aminex HPX87H column (Bio-Rad, Hercules, CA, USA) as previously described [88].
5.3.6. Experimental design for high solid loading enzymatic hydrolysis
The impact of variables such as residence time, solids loading and enzyme loading on glucan and
xylan conversion was determined using a Box Behnken design of experiments (DOE), for
enzymatic hydrolysis on both on EA and AFEX
136

TM

pretreated corn stover. The DOE was

performed using Minitab software (Minitab Inc., State College, PA, USA), where 30
experimental points were generated, including duplicates and six center points. The low and high
values that defined the boundaries of our design are shown in Table 5-1 .

Table 5-1 High and low limits defined for the experimental points using Box Benhken
design.
Factor
Low
High
Time (h)
24
120
Solid Loading (%)
15
25
Enzyme Loading (mg/g glucan)
5
30

The experimental design used in this work is presented in Table 5-2. Glucan and xylan
conversions (monomeric and monomeric + oligomeric) were determined for the various
enzymatic hydrolysis conditions and used to fit a full quadratic equation containing the factors
described in Table 5-1. All interaction effects between factors were considered in this analysis
and parameters were chosen according to their P value and influence on the model prediction
2

(R -predicted and model P values). The regression equations were used to predict the responses
of the various effects as a function of enzymatic hydrolysis conditions within the boundaries set
by the experimental design. Contour plots were used to provide a graphical interpretation of the
regression model and compare predicted enzymatic hydrolysis results between EA and AFEX
pretreated corn stover.

137

TM

Table 5-2 Enzymatic hydrolysis conditions defined by a Box Benhken DOE.
Time
Solid Loading Enzyme Loading
RunOrder
(h)
(wt%)
(mg/g glucan)
1
24
15
18.75
2
120
15
18.75
3
24
25
18.75
4
120
25
18.75
5
24
20
7.5
6
120
20
7.5
7
24
20
30
8
120
20
30
9
72
15
7.5
10
72
25
7.5
11
72
15
30
12
72
25
30
13
72
20
18.75
14
72
20
18.75
15
72
20
18.75
16
24
15
18.75
17
120
15
18.75
18
24
25
18.75
19
120
25
18.75
20
24
20
7.5
21
120
20
7.5
22
24
20
30
23
120
20
30
24
72
15
7.5
25
72
25
7.5
26
72
15
30
27
72
25
30
28
72
20
18.75
29
72
20
18.75
30
72
20
18.75

138

5.3.7. High solids loading enzymatic hydrolysis
Enzymatic hydrolysis was performed in 250 mL Erlenmeyer flasks with 100 mL reaction
volume, incubated at 50 ºC, with pH adjusted to 4.8, in an orbital shaking incubator at 250 RPM
(New Brunswick, USA). The enzymes used in these experiments were Cellic® CTec2 (138 mg
protein/ml, batch No.VCNI0001), Cellic® HTec2 (157 mg protein/ml, batch No.VHN00001),
generously provided by Novozymes (Franklinton, NC, USA), and Multifect Pectinase (MP) (72
mg/ml, batch No. 4861295753) a gift from Genencor/Dupont (Palo Alto, CA, USA). The protein
concentration for the enzymes was determined using the Kjeldahl nitrogen analysis method
(AOAC Method 2001.11, Dairy One Cooperative Inc., Ithaca, NY, USA). The enzyme ratios
utilized in this work were 50 % Cellic® CTec2, 25 % Cellic® HTec2 and 25 % MP on a dry
protein weight basis for both EA and AFEX

TM

pretreated corn stover. This relative amounts of

the various commercial enzymes was previously optimized and found to maximize total sugar
conversion of both feedstocks. The other enzymatic hydrolysis conditions used in this work are
defined in Table 5-2. After enzymatic hydrolysis, samples of the hydrolyzate were analyzed for
glucose and xylose content using HPLC equipped with a Bio-Rad Aminex HPX-87H column
(Bio-Rad, Hercules, CA, USA) as previously described [88].
5.4. Results and Discussion
5.4.1. Evaluation of EA pretreatment on corn stover
The EA pretreatment reactor used in this work was developed for larger scale production of EACS in order to supply feedstock for experiments that require significant amounts of material,
notably high solid loading enzymatic hydrolysis and fermentation. In this new reactor system,
139

pretreatment conditions already known to maximize sugar yields and lignin extraction (from
CHAPTER 4) were evaluated in parallel with other unexplored conditions. The major goals
were to verify if extended residence times, larger than 30 min, would further improve EA
performance and if the results obtained in previous studies could be replicated in the larger EA
pretreatment setup. This evaluation was important to define the best EA conditions before
expanding our work to high solid loading enzymatic hydrolysis. Figure 5-2 i) shows the impact
of biomass moisture content on glucose and xylose conversions as well as lignin extraction.
From these results is apparent the negative influence of biomass moisture content on glucose
conversion. In these experiments, highly digestible cellulose III (CIII) was formed in all cases as
was confirmed by XRD analysis [data not shown], however it is possible that the presence of
additional water may impact the amount of converted CIII in the sample, as water is known to
facilitate reversion of CIII to cellulose I (CI) [80]. XRD of the pretreated material is not
sufficient to confirm this hypothesis, because these samples contain different amounts of
amorphous material (e.g., hemicelluloses and lignin) and it is impossible to deconvolute the
spectra using the technique presented in CHAPTER 1. Further work is required to develop
methodologies to extract lignin and hemicelluloses from the pretreated biomass without reverting
CIII to CI, thereby allowing proper CIII quantification to be performed through XRD or FTRaman analysis [163, 164]. Xylose conversion seems to be positively affected by the presence of
moisture, even though the effects are not large. In the case of lignin extraction, water did not play
a major role at this range of conditions, as the highest level of lignin extraction was 36.3 % (10
% moisture) and the lowest was 31.6 % (40 % moisture). From these results, it is possible to
conclude that lower moisture contents seem to benefit lignin extraction as well as total sugar
conversion and therefore, 10 % moisture was used in the following experiments. Figure 5-2 ii)
140

shows the effect of temperature on EA pretreatment performance. These experiments were
performed using biomass with 10 % moisture, 3:1 ammonia:biomass ratio and 60 min residence
time. From these results it is clear that temperature positively impacts 24 h glucose conversion in
the range between 90 °C and 110 °C, where conversions reached 81.6 % and 93.8 %,
respectively. From 110 °C to 120 °C, glucose conversions were maintained practically constant.
In the case of xylose conversions, temperature has little impact in the range of conditions tested
herein. The results show a slight increase of xylose conversion from 90 °C to 110 °C, reaching
71.1 % and 76.3 %, respectively. This increase was followed by a small decrease to 72.8 % at
120 °C. In the other hand, temperature plays an important role on lignin extraction. From Figure
5-2 ii) is clear that % lignin extraction remained practically constant at ~30 % for pretreatment
temperatures ranging from 90 °C to 110 °C. However, a sudden increase was observed for EA
pretreatment performed at 120 °C, where 36.3 % lignin extraction was achieved. Ammoniabased extraction of lignin is commonly known to be positively affected by temperature [157,
160], however further work is required to evaluate the reasons behind this unexpected sudden
increment. One possible hypothesis can be related to the fact that a great portion of the extracted
lignin has a Tg (glass transition temperature) of approximately 117 °C (as demonstrated in
Appendix from CHAPTER 6). Once the Tg is reached, lignin becomes “softer”, and can
perhaps be delocalized more easily from the cell wall ultra-structure and solubilized by liquid
ammonia.

141

ii)
80%

60%

60%

40%

40%

20%

20%

30%

60%

60%

40%

40%

20%

20%

0%

40%

60%
40%

80%

60%
60%

40%

40%

20%

20%
0%

20%

100%
Ammonia:Biomass Ratio (60 min.)

4:1

5:1

100%

0%

0%
3:1

6:1
10%

120

iv)
Lignin % Extraction

Conversion

80%

110 °C

Temperature

100%

80%

100 °C

Moisture

100%
Ammonia:Biomass Ratio (30 min.)
100%

0%
90 °C

80%

80%

Conversion

20%

Biomass Moisture Content
Glucose
Xylose
Lignin Extraction

Conversion

80%

0%
10%

iii)

80%

60%
60%

40%

40%

20%

20%
0%

20%

Ammonia:Biomass Ratio

30%

100%

Lignin % Extraction

0%

100%

Lignin % Extraction

80%

Temperature
100%

80%
60%
40%
20%
0%

0%
3:1
40%

4:1

Lignin % Extraction

100%

Lignin % Extraction

Conversion

Moisture
100%

Conversion

i)

5:1

6:1

Ammonia:Biomass Ratio

Biomass Moisture Content
Glucose
Xylose
Lignin Extraction

Figure 5-2 i) Effect of moisture content on enzymatic hydrolysis and lignin extraction
performances. Experiments were performed at 110 °C for 60 min residence time and 3:1
ammonia loading; ii) Effect of temperature on enzymatic hydrolysis and lignin extraction
performances. Experiments were performed with 10 % moisture content, 3:1 ammonia loading
and 60 min residence time; iii) Effect of ammonia:biomass ratio at 30 min residence time on
enzymatic hydrolysis and lignin performances. Experiments were performed with 10 % moisture
content at 120 °C; iv) Effect of ammonia:biomass ratio at 60 min residence time on enzymatic
hydrolysis and lignin extraction performances. Experiments were performed with 10 % moisture
content at 120 °C. Enzymatic hydrolysis in all cases was performed with an enzyme loading of
15 mg/g glucan for 24 h at 50 °C and 250 RPM.

The effect of the ammonia:biomass ratio on EA performance is shown in Figure 5-2 iii), for a 30
min pretreatment time. The results indicate that glucose conversion does not vary significantly at
this range of conditions, where 87-90 % conversion was achieved after 24 h enzymatic
hydrolysis at 1% glucan loading. In the other hand, xylan conversion was slightly improved from
142

69 % to 76 % when ammonia:biomass ratio increased from 5:1 to 6:1. No major differences in
xylan conversion were observed from 3:1 to 5:1, where values remained between ~69 and ~70 %
conversion. Lignin extraction was significantly improved with increasing ammonia:biomass
ratio. This result is in agreement with our early observations that showed the importance of
ammonia loading for lignin extraction CHAPTER 4. Comparing results from the model
generated in CHAPTER 4 with experimental data presented in Figure 5-2 iii), it is possible to
validate that the model has good ability to predict new observations (Figure 5-3), even using a
larger scale experimental setup. Figure 5-3 shows that lignin extraction could be predicted well
for ammonia:biomass ratios from 4:1 to 6:1, showing residuals ranging from 0.60 % to 1.95 %.
At 3:1 ammonia:biomass ratio the residual was largest in absolute terms, as the model
underpredicted lignin extraction by 3.81 % compared to the experimental value. From the same
figure, it is also possible to compare glucose and xylose conversions with model predictions. The
residuals obtained for glucose conversion varied from -3.17 % (3:1) to 6.35 % (5:1), while
xylose conversion residuals varied from 0.47 % (6:1) to 5.59 % (5:1). Therefore, experimental
results show good agreement with model predictions, even though the data obtained to generate
the model was produced with a different experimental setup. Additionally, the enzyme
combination used in this study is different from the one used to produce the model, with
inclusion of 25 % MP by replacement of 25 % of HTec 2, although both enzyme combinations
generate very comparable results at 1 % glucan loading with respect to glucose and xylose
conversions [data not shown].
Increasing the pretreatment residence time from 30 min to 60 min did not result in significant
improvements both for glucose and xylose conversions (Figure 5-2 iv)).

However, lignin

extraction improved significantly with pretreatment time, especially at 4:1 ammonia:biomass
143

ratio, where lignin extraction increased from about 35 % to 48 %. Lignin extraction remained
constant between 5:1 to 6:1 ammonia:biomass ratio, where about 50% of the lignin from corn
stover was extracted. The mathematical model generated in CHAPTER 4 does not predict
observations for pretreatment times above 30 min. Therefore, the comparison between predicted
and experimental results could not be performed for these conditions.

100%

50%

80%

40%

60%

30%

40%

20%

20%

10%

0%

Lignin % Extraction

Conversion

Pretreatment Model vs. Experimental Data

0%
3:1

4:1

5:1

6:1

Ammonia:Biomass Ratio
Glucose

Model Glucose

Xylose

Model Xylose

Lignin Extraction

Lignin Extraction Model

Figure 5-3 Comparison between experimental observations and predicted observations by the
model presented in CHAPTER 4 for glucose, xylose and lignin extraction performances.
Pretreatment conditions were performed with a fixed biomass moisture of 10 %, 120 °C and 30
min residence time, varying ammonia:biomass ratio from 3:1 to 6:1.

Using the results presented herein, we defined the EA condition for producing larger quantities
of material in order to perform high solid loading enzymatic hydrolysis. The preferred condition
144

allowed sugar and lignin extraction yields to be close to maximum with minimal energy inputs.
Therefore, we chose to use 10 % moisture in biomass, 6:1 ammonia:biomass ratio at 120 °C for
30 min. This choice of pretreatment time had to do with the small improvements obtained by
extending the pretreatment to 60 min, as can be observed in Figure 5-2 iii) & iv). The process
mass balance with the preferred EA pretreatment condition is shown in Figure 5-4. In this work,
glucose and xylose conversions (Xglucose and Xxylose, respectively) will be modeled as a
function of enzymatic hydrolysis parameters such as incubation time, solids loading and enzyme
loading. The empirical model, based on statistical design of experiments, will provide a useful
tool to optimize conditions and perform future techno-economic evaluation studies. Furthermore,
it helps us to analyze the effect of the various parameters on enzymatic hydrolysis performance.

145

Figure 5-4 Process mass balance for the EA pretreatment condition chosen for this study.
Enzymatic hydrolysis conversion (%X) is a function of variables such as time, % solids loading
and enzyme loading during enzymatic hydrolysis. Empirical modeling was performed to
determine these relationships in this study and complete the material balance around
pretreatment and enzymatic hydrolysis.

5.4.2. High solid loading enzymatic hydrolysis of EA corn stover
5.4.2.1. Definition of enzymatic hydrolysis conditions: fixed variables
High solid loading enzymatic hydrolysis was studied using EA-CS pretreated under the
conditions determined above. The performance of EA-CS was compared with AFEX

TM

-CS, as

this is an established pretreatment technology that is currently on the path for commercialization.
Sugar yields during enzymatic hydrolysis are known to be dependent on various factors that
146

include, solids loading, enzyme loading and incubation time [170, 172, 173]. Other parameters
are also revealed to be important, notably the mixing regime, pH, incubation temperature and
enzyme mixture, which is dependent on the lignocellulosic substrate [172, 173]. For this study,
incubation temperature was set constant at 50 °C, as enzymes show maximum activity around
this temperature according to the literature provided by the enzyme suppliers. Preliminary
experiments varying pH from 4.2 to 5.2 revealed a pH between 4.8 and 5.2 maximize sugar
conversion of AFEX

TM

-CS and therefore, a pH of 4.8 was chosen for this study. Other

published studies have successfully used this pH, as it is also recommended by enzyme
manufacturers [4, 171]. Enzyme cocktail optimization was performed in conical shake flask
experiments using 20 % solids and 18.75 mg/g glucan enzyme loading for 72 h incubation time,
TM

both for AFEX

- and EA-CS. Based on our knowledge and experience optimizing enzyme

cocktails [22], a set of 6 commercial enzyme combinations known to maximize sugar yields
were chosen to determine the optimal mixture. From these experiments the combination of 50 %
Ctec2, 25 % Htec2 and 25 % MP maximized total sugar conversion for both AFEX

TM

and EA

pretreated corn stover (in Appendix) at 20% solids. Preliminary experiments were also
performed to optimize mixing speed, where 20 % solid loading enzymatic hydrolysis was
performed in a shaker incubator for 72 h at optimal pH and temperature. A range of mixing
speeds was tested in this preliminary study, from 100 to 300 rpm, demonstrating that 250 rpm is
sufficient to maximize sugar conversion [data not shown]. By fixing enzyme mixture, pH,
incubation temperature and mixing regime, we will be able to focus our attention on the effects
of 1) solid loading, 2) enzyme loading and 3) incubation time on glucose and xylose conversion
TM

for EA- and AFEX

-CS.
147

5.4.2.2. Empirical modeling of high solid loading enzymatic hydrolysis: AFEX

TM

-CS

For this work, a statistical design of experiments was developed using Minitab software. The
Box Benhken design was composed by a total of 30 experiments, including duplicates and six
middle points. The experimental results, notably glucose and xylose conversions, were used to
develop a regression model composed of a quadratic equation containing the various factors
(solid loading, enzyme loading and incubation time) and the respective interaction terms. The
coefficients from the terms were considered based on their P value (< 0.05), goodness of the
regression fit and their effect on the model’s ability to predict unknown observations, analyzing
2

statistical parameters such as predicted R . These statistical analyses were performed both for
TM

enzymatic hydrolysis on AFEX

and EA pretreated corn stover. From Figure 5-5 A, it is

possible to observe residual plots from the regression of monomeric glucose conversion as a
function of enzymatic hydrolysis parameters for AFEX

TM

-CS. Figure 5-5 B shows the

coefficients considered for the regression model and respective P values. The residual plots show
that the residuals are evenly spread through the fitted values, even though there are more
experimental points above 50 % conversion (0.5). The residuals varied from -4 % (-0.04) to 4 %
(0.04) glucose conversion, however the histogram shows that most residuals waere close to zero.
From the normal probability plot, it is also possible to observe that the residuals conform
reasonably well to a normal distribution.

148

A

Residual Plots for Monomeric Glucose Conversion (AFEX)

B

Regression Coefficients for Monomeric Glucose Conversion

Term
Constant
Time
Solid Loading
Enzyme Loading
Time*Time
Solid Loading*Solid Loading
Enzyme Load.*Enzyme Load.
Time*Solid Loading
Time*Enzyme Loading
Solid Loading*Enzyme Loading

S = 0.0229884
R-Sq = 98 %

Coef
-0.034311
0.006288
0.016447
0.03727
-0.000026
-0.000686
-0.000578
-0.000039
-

PRESS = 0.0231912
R-Sq (pred) = 96.01%

SE Coef
0.139759
0.000611
0.013584
0.002778
0.000004
0.000338
0.000067
0.000015
-

T
-0.246
10.289
1.211
13.416
-7.039
-2.028
-8.644
-2.595
-

P
0.808
0.000
0.239
0.000
0.000
0.055
0.000
0.017
-

R-Sq (adj) = 97.36%
TM

Figure 5-5 (A) Residual plots for the regression of glucose conversion on AFEX -CS as a
function of enzymatic hydrolysis variables (time, solid loading and enzyme loading) and
respective interaction terms. Residuals and fitted values are expressed in (glucose conversion).
(B) Estimation of regression coefficients and P values for the terms affecting glucose conversion.
Parameters that determine the goodness of the fit are also shown.
149

From Figure 5-5 B, it is possible to observe that the interaction parameters Time*Solid Loading
and Solid Loading*Enzyme Loading were not considered for the model equation. These two
factors had a P value higher than 0.05 and had negative impacts on the predictive ability of the
2

model by generating lower predicted R values. Other parameters also have P values above 0.05
and were included in the model, as their absence negatively impacted the goodness of the fit and
2

the ability of the model to predict unknown observations, notably by reducing the predicted R .
2

2

2

The final model presented an R value of 98 %, a predicted R of 96.01 % and an adjusted R of
97.36 %. These three parameters are similar to each other and demonstrate the goodness of the fit
and predictability of the model, also suggesting the model is not overfitting the experimental
data.
The final equation predicting glucose conversion for AFEX

TM

-CS as a function of enzymatic

hydrolysis parameters was determined as follows:

Glucose Conversion = (-0.034311) + (0.006288 X1) + (0.016447 X2) + (0.03727 X3) -

5-1

(0.000026 X1X1) - (0.000686 X2X2) - (0.000578 X3X3) - (0.000039 X1X3)
Where,
X1 = Time (h) [24 to 120]; X2 = Solid Loading (%) [15 to 25]; X3 = Enzyme Loading (mg/g
glucan) [7.5 to 30].

150

A

Residual Plots for Monomeric Xylose Conversion (AFEX)

B

Regression Coefficients for Monomeric Xylose Conversion

Term
Constant
Time
Solid Loading
Enzyme Loading
Time*Time
Solid Loading*Solid Loading
Enzyme Load.*Enzyme Load.
Time*Solid Loading
Time*Enzyme Loading
Solid Loading*Enzyme Loading

S = 0.0188255
R-Sq = 95.48%

Coef
0.431136
0.003516
-0.0086
0.016491
-1.5E-05
-0.00023
-1.6E-05
-

PRESS = 0.0151929
R-Sq (pred) = 91.58%

SE Coef
0.034083
0.000499
0.000941
0.00227
0.000003
0.000055
0.000012
-

T
12.65
7.041
-9.139
7.266
-4.985
-4.285
-1.333
-

P
0.000
0.000
0.000
0.000
0.000
0.000
0.196
-

R-Sq (adj) = 94.3%
TM

Figure 5-6 (A) Residual plots for the regression of xylose conversion on AFEX -CS as a
function of enzymatic hydrolysis variables (time, solid loading and enzyme loading) and
respective interaction terms. Residuals and fitted values are expressed in (xylose conversion). (B)
Estimation of regression coefficients and P values for the terms affecting xylose conversion.
Parameters that determine the goodness of the fit are also shown.
151

Figure 5-6 shows the residual plots and regression parameters for xylose conversion of
TM

AFEX

-CS as a function of enzymatic hydrolysis parameters. In Figure 5-6 A, the residuals

for this model are randomly distributed throughout the residual plot (versus fit), showing that the
model is balanced and does not present systematic lack-of-fit. The histogram and normal
distribution plots show that most of the residuals are close to zero, following an approximate bell
shape curve ranging from -3 % (0.03) to 4 % (0.04). This observation suggests that the model fits
2

experimental data reasonably well, which can be confirmed by the R value of 95.48 % (Figure
5-6 B). The regression coefficients and terms considered for this model are shown in Figure 5-6
B. Most of the terms had P values below 0.05, showing their significance in explaining the
variability of the experimental observations. One exception was the term Time*Enzyme Loading,
which had a P value of 0.196, however, it was included in the final model equation because it
2

positively affected the predictive ability of the model, by increasing the predicted R . Since the
2

2

2

R , predicted R and adjusted R are similar to each other and relatively high (> 90 %), it
suggests that the model has good predictive ability, fitting the experimental data reasonably well
without overfitting experimental values.
The final equation predicting xylose conversion for AFEX
hydrolysis parameters was determined as follows:

152

TM

-CS as a function of enzymatic

Xylose Conversion = (0.431136) + (0.003516 X1) - (0.0086 X2) + (0.016491 X3) -

5-2

(1.5 E -5 X1X1) - (0.00023 X3X3) - (1.6 E -5 X1X3)
Where,
X1 = Time (h) [24 to 120]; X2 = Solid Loading (%) [15 to 25]; X3 = Enzyme Loading (mg/g
glucan) [7.5 to 30].

5.4.2.3. Empirical modeling of high solid loading enzymatic hydrolysis: EA-CS
A similar analysis was performed for enzymatic hydrolysis of EA-CS, resulting in regression
equations for glucose and xylose conversions as a function of incubation time, solid loading and
enzyme loading, including their respective interaction terms. Figure 5-7 A shows residual plots
and histograms for the glucose conversion model. In this model, there is a tendency for the
residuals to be negative in the lower and higher extremities of glucose conversion (versus fits
plot). The residuals in those regions can go up to -0.07 (-7% conversion). In the intermediate
glucose conversions the residuals appear to be randomly distributed and do not show visible
evidence of systematic lack-of-fit. It is also important to mention that the experimental
observations did not cover a region between ~0.62 and ~0.72 glucose conversion, which can
negatively impact the predictive ability of the model. The histogram and normal probability plots
show good distribution of the residuals, ranging from -0.07 (-7 %) to 0.04 (4 %), with the
majority of the fits having residuals near zero, as preferred. As a consequence of the minor
2

2

systematic lack-of-fit, the predicted R value for the best fit was 88.73 %, although the R value
2

was 94.16 %. These statistical indicators and the adjusted R (92.3 %) are similar to each other,
153

which indicate the model does not overfit experimental data. The coefficients considered for the
best fit are shown in Figure 5-7 B and included all the terms with P value below 0.05.
As a result of this analysis, the final equation predicting glucose conversion for EA-CS as a
function of enzymatic hydrolysis parameters was determined as follows:

Glucose Conversion = (0.695671) + (0.004804 X1) - (0.02252 X2) + (0.02295 X3) -

5-3

(1.2 E -5 X1X1) - (0.00051 X3X3) - (9.9 E -5 X1X2) - (0.000582 X2X3)
Where,
X1 = Time (h) [24 to 120]; X2 = Solid Loading (%) [15 to 25]; X3 = Enzyme Loading (mg/g
glucan) [7.5 to 30].

The residual plots and histograms for the xylose conversion model of EA-CS are shown in
Figure 5-8 A. The plot “versus fits” shows that the residuals are randomly distributed throughout
the space of the model. The residual amplitude is also relatively small, ranging from -0.003 (3
%) to 0.024 (2.4 %). The histogram and normal probability plots show that the residuals follow a
normal distribution relatively well, with the highest number of residuals being close to zero.
Thus the model fits the experimental observations reasonably well and suggests the absence of a
systematic lack-of-fit within the space of the model. Figure 5-8 B shows that the model has a
2

good predictive ability, with a predicted R value of 93.02 % and fits experimental data
2

2

reasonably well with an R value of 95.87 %. As the adjusted R value (94.79 %) is similar to

154

2

the formerly mentioned R values, there is a strong indication that the model does not overfit
experimental data and can be used for predict new observations with good probability of success.
The model parameters used for predicting xylose conversion on EA-CS are also shown in Figure
5-8 B. The terms of this model were chosen based on their P value and their influence on the
predictive ability of the model. Terms with P values above 0.05 were initially rejected, however
the model was negatively impacted by the absence of the terms Solid Loading and Solid
2

Loading*Solid Loading, due to decrease of predicted R value and increase of P value from the
F test. Therefore, those two terms were included in the model. As a result, the final equation
predicting xylose conversion for EA-CS as a function of enzymatic hydrolysis parameters was
determined as follows:

Xylose Conversion = (0.503478) + (0.001913 X1) + (0.00947 X2) + (0.01392 X3) -

5-4

(7 E -6 X1X1) - (0.00048 X2X2) - (0.00023 X3X3)
Where,
X1 = Time (h) [24 to 120]; X2 = Solid Loading (%) [15 to 25]; X3 = Enzyme Loading (mg/g
glucan) [7.5 to 30].

155

A

Residual Plots for Monomeric Glucose Conversion (EA)

B

Regression Coefficients for Monomeric Glucose Conversion

Term
Constant
Time
Solid Loading
Enzyme Loading
Time*Time
Solid Loading*Solid Loading
Enzyme Load.*Enzyme Load.
Time*Solid Loading
Time*Enzyme Loading
Solid Loading*Enzyme Loading

S = 0.0299862
R-Sq = 94.16%

Coef
0.695671
0.004804
-0.02252
0.02295
-1.2E-05
-0.00051
-9.9E-05
0.000582

PRESS = 0.0381324
R-Sq (pred) = 88.73%

SE Coef
0.089122
0.000795
0.003839
0.005223
0.000005
0.000087
0.00002
0.000188

T
7.806
6.04
-5.865
4.394
-2.53
-5.818
-5.06
3.09

P
0.000
0.000
0.000
0.000
0.019
0.000
0.000
0.005

R-Sq (adj) = 92.3%

Figure 5-7 (A) Residual plots for the regression of glucose conversion on EA-CS as a function
of enzymatic hydrolysis variables (time, solid loading and enzyme loading) and respective
interaction terms. Residuals and fitted values are expressed in (glucose conversion). (B)
Estimation of regression coefficients and P values for the terms affecting glucose conversion.
Parameters that determine the goodness of the fit are also shown.
156

A

Residual Plots for Monomeric Xylose Conversion (EA)

B

Regression Coefficients for Monomeric Xylose Conversion

Term
Constant
Time
Solid Loading
Enzyme Loading
Time*Time
Solid Loading*Solid Loading
Enzyme Load.*Enzyme Load.
Time*Solid Loading
Time*Enzyme Loading
Solid Loading*Enzyme Loading

S = 0.016133
R-Sq = 95.87%

Coef
0.503478
0.001913
0.00947
0.013927
-7E-06
-0.00048
-0.00023
-

PRESS = 0.0101034
R-Sq (pred) = 93.02%

SE Coef
0.097039
0.00038
0.009533
0.001795
0.000003
0.000237
0.000047
-

T
5.188
5.028
0.993
7.758
-2.618
-2.02
-4.812
-

P
0.000
0.000
0.331
0.000
0.015
0.055
0.000
-

R-Sq (adj) = 94.79%

Figure 5-8 (A) Residual plots for the regression of xylose conversion on EA-CS as a function of
enzymatic hydrolysis variables (time, solid loading and enzyme loading) and respective
interaction terms. Residuals and fitted values are expressed in (xylose conversion). (B)
Estimation of regression coefficients and P values for the terms affecting xylose conversion.
Parameters that determine the goodness of the fit are also shown.
157

TM

5.4.2.4. Performance of high solid loading enzymatic hydrolysis on EA- and AFEX

-CS

To better evaluate our empirical modeling work, we compared experimental and predicted
observations side-by-side, as shown in Figure 5-9 . The predicted values match reasonably well
TM

average glucose (Figure 5-9 I) and xylose (Figure 5-9) conversions for EA- and AFEX

-CS

at the various conditions given by the Box Benhken design (Figure 5-9 III). In this figure it is
also clear that EA-CS is more digestible than AFEX

TM

-CS for all enzymatic hydrolysis

conditions, as reflected by superior glucose and xylose conversions.
To better visualize of the effect of the various enzymatic hydrolysis variables on glucose and
xylose conversions, a series of contour plots are presented in Figure 5-10 and Figure 5-11.
Figure 5-10 shows contour plots for glucose A) and xylose B) conversions on AFEX

TM

-CS

using “middle point” experimental conditions as fixed variables, i.e., 72 h time, 18.75 mg/g
glucan enzyme loading and 20 % solid loading. From the contour plots, it is possible to observe
the negative effect of increasing solid loading on both glucan and xylan conversion (plots I and
II). This effect, also known as the “solids effect” in the literature, is commonly observed when
performing high solid loading enzymatic hydrolysis studies, independently of the pretreatment
technology and lignocellulosic substrate applied [170]. As expected, increasing enzyme loading
and residence time have a positive effect on enzymatic hydrolysis yields from AFEX

TM

-CS. The

effect of time and enzyme loading (within the boundaries of this study) seems to be more
impactful to improve glucose conversion than xylose conversion from this substrate as is evident
from the range of conversions that can be achieved when controlling these two variables. Using
TM

AFEX

-CS, the maximum glucose conversion for “middle point” conditions range from 85 %
158

to 92 %, achieved using low solid loadings, relatively high enzyme loadings and residence times.
The scenario for xylose conversion is significantly different, as maximum xylose conversions
range 70 % to 75 % for “middle point” conditions. These significantly lower yields for xylose
conversion agree with previous reports in the literature that studied enzymatic hydrolysis on
TM

AFEX

-CS [171]; however the reasons behind these observations are still under investigation.

If we compare these results for AFEX

TM

-CS with EA-CS (Figure 5-11), it is possible to observe

significant differences and some similarities. The performance profile and the dependence of the
enzymatic hydrolysis variables are similar between the two pretreated substrates. The “solids
effect” is still present in EA-CS, especially at lower enzyme loadings and residence times. At
higher enzyme loadings the “solids effect” seems to be reduced for glucose conversion as can be
TM

observed on Figure 5-11 A-I. The major differences between EA-CS and AFEX

-CS is that

EA pretreatment improves glucose and xylose conversions significantly. Maximum glucose
conversions of > 92 % and xylose conversions of > 85 % can be achieved for “middle point”
conditions on EA-CS in contrast with AFEX

TM

-CS. Much of this improvement is due to

cellulose III conversion [105, 117] and also to lignin removal [157, 165] as discussed in
CHAPTER 4.

159

Glucose Conversion Results

i)

Glucose Conversion (%)

Experimental (EA)

Model (EA)

Model (AFEX)

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
1

2

3

4

5

6

7

8

9

10

11

12

13

Xylose Conversion Results

ii)
Experimental (EA)
Glucose Conversion (%)

Experimental (AFEX)

Model (EA)

Experimental (AFEX)

Model (AFEX)

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
1

2

iii)
Condition #
1
2
3
4
5
6
7
8
9
10
11
12
13

Time (h)
24
24
24
24
72
72
72
72
72
120
120
120
120

3

4

5

6

7

8

9

10

11

12

13

Enzymatic Hydrolysis Conditions
Solid Loading (%)
Enzyme Loading (mg/g glucan)
15
18.75
20
7.5
20
30
25
18.75
15
7.5
15
30
20
18.75
25
7.5
25
30
15
18.75
20
7.5
20
30
25
18.75

Figure 5-9 Comparison of average experimental observations and model predictions for i)
TM
glucose conversion and ii) xylose conversion for EA- and AFEX -CS. iii) Experimental
conditions formulated by the Box Benhken design, which were used for model formulation.
160

To better observe the benefits of EA pretreatment during high solids loading enzymatic
hydrolysis, a few kinetic curves were extracted from the model for two solid loadings (15 % and
20 %) and three enzyme loadings (7.5, 10 and 20 mg/g glucan), which are shown in Figure 5-12.
As mentioned during the analysis of the contour plots, both residence time and enzyme loading
seem of little benefit to xylose conversion. These curves clearly show that in the period between
24 and 120 h, predicted xylose conversions do not improve more than ~10 % for all the enzyme
and solid loadings presented. This observation is also independent of the pretreatment type
investigated here, as it was observed for both EA and AFEX
enzymatic hydrolysis are greater for EA-CS than for AFEX

TM

TM

-CS. However, the initial rates of

-CS and that seems to be the key

factor for achieving higher xylose conversions. Also, a relatively small improvement (~10 %) in
the final xylose conversion is predicted when increasing enzyme loading from 7.5 to 20 mg/g
glucan (i.e., 2.7 fold increase enzyme loading). Likewise, increasing solids loading from 15 % to
20 % did not produce considerable effects on the final xylose conversions for the range of
enzyme loadings evaluated in Figure 5-12.

161

A

Time = 72 h

Enzyme Loading = 18.75 mg/g glucan

Solid Loading = 20 %

20
15

22.5

20.0

17.5

Hold Values
Time 72

15.0

17.5

20.0

22.5

15.0
24

25.0

40

60

80

100

120

Time

Solid Loading
Solid
Loading (%)

25

25

20

20

15

Time (h)

monomeric
glucose
conv ersion
< 0.55
0.55 – 0.60
0.60 – 0.65
0.65 – 0.70
0.70 – 0.75
0.75 – 0.80
0.80 – 0.85
0.85 – 0.92
> 0.92

glucose conv ersion
< 0.55
0.55 – 0.60
0.60 – 0.65
0.65 – 0.70
0.70 – 0.75
0.75 – 0.80
0.80 – 0.85
0.85 – 0.92
> 0.92

Hold Values
Enzy me Loading 18.75

15

10

Glucan
monomeric
glucose
Conversion
% monomeric
conv ersion

III

30

conv ersion
< 0.55
0.55 – 0.60
0.60 – 0.65
0.65 – 0.70
0.70 – 0.75
0.75 – 0.80
0.80 – 0.85
0.85 – 0.92
> 0.92

Enzyme Loading

Solid Loading

25

Contour Plot 30
of % monomeric
glucose conversion vs Enzyme Loading,%Time
%

%

IImonomeric
glucose

Enzyme Load. (mg/g glucan)

25.0

I

Enzyme Loading

30

Solid Loading (%)

Enzyme Loading

Enzyme Load. (mg/g glucan)

Contour Plot of % monomeric gluc vs Enzyme Contour
Loading,Plot
Solid
of Loading
% monomeric glucose conversion
vsPlot
Solid
Time glucose conversion vs Enzyme Loading, Time
Contour
ofLoading,
% monomeric

10

10

40

24

40

60

80

Time 100
80
Time
(h)

60

100

120

< 0.55
0.55 – 0.60
0.60 – 0.65
0.65 – 0.70
0.70 – 0.75
Values
0.75SolidHold
– Loading
0.80
20
0.80 – 0.85
0.85 – 0.92
> 0.92
Hold Values
Solid Loading 20

120

Time

B

Enzyme Loading = 18.75 mg/g glucan
25.0

I
Solid Loading

25
20
15

22.5

20.0

17.5

10
15.0

17.5

20.0
Solid Loading

22.5

Solid Loading (%)

25.0

40

60

80
Time

Time (h)

100

Xylan
monomeric
xy lose
Conversion
conv ersion
% monomeric

III

30

conv ersion
< 0.60
0.60 – 0.65
0.65 – 0.70
0.70 – 0.75
0.75 – 0.80
0.80 – 0.85
> 0.85
Hold Values
Time 72

15.0
24

Solid Loading = 20 %

Contour Plot30
of % monomeric
xylose conversion vs Enzyme Loading,%Time
%

%

IImonomeric
xy lose

Enzyme Load. (mg/g glucan)

30

Enzyme Loading
Enzyme Loading

Time = 72 h

Solid Loading (%)

Enzyme Load. (mg/g glucan)

Enzyme Loading

Contour Plot of % monomeric xylo vs Enzyme Loading, Solid Loading
Contour Plot of % monomeric xylose conversion vs Solid Loading, Time
Contour Plot of % monomeric xylose conversion vs Enzyme Loading, Time

25

25

20

20

monomeric
xy lose
conv ersion
< 0.60
0.60 – 0.65
0.65 – 0.70
0.70 – 0.75
0.75 – 0.80
0.80 – 0.85
> 0.85

< 0.60
xy lose conv ersion
0.60 – 0.65
0.65 – 0.70
0.70 – 0.75
0.75 – 0.80
0.80 – 0.85
> 0.85

< 0.55
0.55 – 0.60
0.60 – 0.65
0.65 Hold
– 0.70
Values
0.70Solid– Loading
0.75 20
0.75 – 0.80
0.80 – 0.85
> 0.85

Hold Values

15Enzy me Loading

18.75

15

10

120 10

24

24
40

40

60
60

80

Time
Time
80 (h)
100

100

120
Hold Values

120

Time

Figure 5-10 Contour plots with model predictions for A) glucose and B) xylose conversion for
TM
AFEX -CS, as a function of time, enzyme loading and solids loading during enzymatic
hydrolysis. Fixed variables were chosen to be at the “middle point” conditions.

162

A

Time = 72 h

Enzyme Loading = 18.75 mg/g glucan

Solid Loading = 20 %

20
15

22.5

20.0

17.5

conv ersion
< 0.55
– 0.60
– 0.65
– 0.70
– 0.75
– 0.80
– 0.85
– 0.92
> 0.92

25

0.55
0.60
0.65
0.70
0.75
0.80
0.85

Hold Values
Time 72

Enzyme Loading

Solid Loading

25

30

%

IImonomeric
glucose

Enzyme Load. (mg/g glucan)

25.0

I

Enzyme Loading

30

Solid Loading (%)

Enzyme Loading

Enzyme Load. (mg/g glucan)

Contour
% monomeric
Contour Plot of % monomeric gluc vs EnzymeContour
Loading,
Plot
Solid
of %
Loading
monomeric glucose conversion
vs Plot
SolidofLoading,
Time glucose conversion vs Enzyme Loading, Time
Contour Plot of % monomeric glucose conversion vs Enzyme Loading, Time

10

20
15
10

15.0

17.5

20.0

22.5

15.0
25.0
24

40

Solid Loading

80

100

120

%
Glucan
% monomeric
monomeric
glucose conv
ersion
glucose
Conversion
ersion
<conv0.55

%
monomeric
glucose
conv ersion
< 0.55
0.55 – 0.60
0.60 – 0.65
0.65 – 0.70
0.70 – 0.75
0.75 – 0.80
0.80 – 0.85
0.85 – 0.92
> 0.92

III

25
20
15

< 0.55

– 0.60
0.55 –0.550.60
0.60 – 0.65
– 0.70
0.60 –0.650.65
0.70 – 0.75
0.65 –0.750.70
– 0.80
– 0.85
0.70 –0.80
0.75
0.85 – 0.92
> 0.92
0.75 – 0.80
Values
0.80SolidHold
– Loading
0.85
20
0.85 – 0.92
> 0.92

Hold Values
Enzy me Loading 18.75

10

24

24
40
60
80
100
120
40
60
80
100
120

Time
Time
(h)

Solid Loading (%)

B

60

30

Time

Hold Values
Solid Loading 20

Time

Time (h)

Enzyme Loading = 18.75 mg/g glucan

Solid Loading

25
20
15

%

IImonomeric
xy lose
22.5

20.0

17.5

25

conv ersion
< 0.55
– 0.60
– 0.65
– 0.70
– 0.75
– 0.80
– 0.85
> 0.85

0.55
0.60
0.65
0.70
0.75
0.80

Hold Values
Time 72

15.0

17.5

20.0

22.5

SolidLoading
Loading
Solid
(%)

20
15
10

10
15.0
25.0
24

Enzyme Load. (mg/g glucan)

25.0

I

Enzyme Loading

30

30

Enzyme Loading

Time = 72 h

Solid Loading (%)

Enzyme Load. (mg/g glucan)

Enzyme Loading

Contour Plot of % monomeric xylose conversion vs Enzyme Loading, Time
Contour Plot of % monomeric xylo vs Enzyme Contour
Loading,Plot
Solid
of Loading
% monomeric xylose conversion
vs Solid
Loading,
Time xylose conversion vs Enzyme Loading, Time
Contour
Plot of
% monomeric

40

60

80

Time
Time
(h)

100

120

Solid Loading = 20 %

30

% monomeric

%
xy lose conv ersion
Xylan
<monomeric
0.55
xy
lose
Conversion
ersion
0.55 –conv0.60

%
monomeric
xy lose
conv ersion
< 0.55
0.55 – 0.60
0.60 – 0.65
0.65 – 0.70
0.70 – 0.75
0.75 – 0.80
0.80 – 0.85
> 0.85

III

25
20

< 0.55

– 0.60
–0.550.65
0.60 – 0.65
–0.650.70
– 0.70
– 0.75
–0.70
0.75
0.75 – 0.80
– 0.85
–0.800.80
> 0.85
– 0.85
Hold Values
Solid
>Loading
0.85 20

0.60
0.65
0.70
0.75
0.80

15

Hold Values
Enzy me Loading 18.75

10

24

Hold Values

40
60
80
100
120
24
40
60
80
100
120
Time

Time (h)
Time

Figure 5-11 Contour plots with model predictions for A) glucose and B) xylose conversion for
EA-CS, as a function of time, enzyme loading and solids loading during enzymatic hydrolysis.
Fixed variables were chosen to be at the “middle point” conditions.

163

i a)

ii a)

15 % Solids and 20 mg Enzyme/g glucan

20 % Solids and 20 mg Enzyme/g glucan

80%

80%

Conversion

100%

Conversion

100%
60%

40%
20%

60%

40%
20%

0%

0%
0

20

40

60

80

100

120

0

20

40

Time (h)

i b)

ii b)

15 % Solids and 10 mg Enzyme/g glucan

Conversion

80%

Conversion

100%

80%
60%

100

120

40%
20%

60%
40%
20%

0%

0%

0

20

40

60

80

100

120

0

20

40

Time (h)

60

80

100

120

Time (h)

ii c)

15 % Solids and 7.5 mg Enzyme /g glucan

20 % Solids and 7.5 mg Enzyme/g glucan

100%

100%

80%

80%

Conversion

Conversion

80

20 % Solids and 10 mg Enzyme/g glucan

100%

i c)

60
Time (h)

60%
40%
20%

0%

60%
40%
20%

0%
0

40
60 at 15
80% Solids
100 and
120 7.5 mg/g glucan
0 Enzyme
20
40
60
Kinetics
Loading

20

Time (h)

Glucose (EA)

80

100

120

Time (h)

Glucose (AFEX)

Xylose (EA)

Xylose (AFEX)

100%

Figure 5-12 Kinetics of enzymatic hydrolysis performed at i) 15 % solids and ii) 20 % solids, for
enzyme loadings
of a) 20 mg/ g glucan, b) 10 mg/ g glucan and c) 7.5 mg/ g glucan. The kinetics
90%
TM
was performed 80%
on EA- and AFEX -CS and accounts for both glucose and xylose conversion.

Conversion

70%
60%
50%
40%
30%
20%

164

10%
0%
0

20

40

60
Time (h)

80

100

120

Incubation time and enzyme loading seem to be the variables that impact glucose conversion in
more than they do xylan conversion. Even though the first 24 h of enzymatic hydrolysis is when
most of the glucose is generated from pretreated biomass, the model predicts considerable
improvements of glucose conversion in the period from 24 to 120 h incubation time. An
TM

additional 15-25 % glucose conversion is predicted in this period for both EA- and AFEX

-CS

and therefore, extended residence times benefit glucose conversions more than xylose
conversion. The effect of residence time is very similar for both pretreatment technologies tested
in this study and this variable seems to be even more important when operating at lower enzyme
loadings. More time is required to achieve terminal glucose conversions at lower enzyme
loadings. A major difference between the performance of EA- and AFEX
TM

of conditions is that glucose conversion of AFEX

TM

-CS under these set

-CS seems to be greatly impacted by

decreasing the enzyme loading from 20 to 7.5 mg/g glucan. The drop in glucose conversion by
decreasing enzyme loading from 20 to 7.5 mg/glucan is approximately 6 % at 15 % solids and
TM

~11 % at 20 % solids for EA-CS, while for AFEX

-CS this drop can be as much as ~20 % at

both 15 % and 20 % solids (Figure 5-13 i)). Comparing EA and AFEX

TM

pretreatments side-

by-side, it is possible to evaluate the absolute improvement in terminal glucose and xylose
conversion at 120 h enzymatic hydrolysis. From Figure 5-13 ii), it is clear that the absolute
improvement (AI) in terminal glucose conversion is larger as the enzyme loading drops from 20
mg/g glucan to 7.5 mg/g glucan. Also, this difference decreases with increasing solids loading
from 10 % to 20 %. This observation suggests that EA pretreatment is more susceptible to the
“solids effect”, especially at lower enzyme loadings, as the absolute difference drops
165

significantly with increasing solids loading. However, it is important to notice that for the same
solids loading, EA-CS contains higher levels of carbohydrates, which will potentiate higher
concentrations of sugars for similar sugar conversion levels. This fact will negatively impact
enzymatic activity due to sugar inhibition at lower sugar conversions and therefore, the “solids
effect” between these two pretreated feedstocks cannot be compared under the same terms.
Interestingly, the AI value for xylose conversion was not affected by enzyme loading or solid
loading, which was maintained constant at approximately 15 % conversion. In a different type of
analysis, it is possible to determine the relative improvement (RI) with respect to AFEX

TM

conversion (Figure 5-13 iii)). From this perspective, it is also possible to have a better idea of
the predicted impact of EA pretreatment as a function of enzymatic hydrolysis conditions. For
example, glucose conversion improvements at high enzyme loading (i.e., 20 mg/g glucan)
represent ~10 % of the final conversion for AFEX

TM

-CS. This RI value is similar for both at 15

and 20 % solids loading enzymatic hydrolysis. However, if we decrease the enzyme loading by
half (i.e., to 10 mg/g glucan), we can practically increase the RI for glucose conversion up to ~3
fold for 15 % solids loading and ~2.5 fold for 20 % solids enzymatic hydrolysis. This RI value
can increase up to ~35 % for 15 % solids and ~29 % for 20 % solids as we reduce the enzyme
loading down to 7.5 mg/g glucan. A similar trend is observed for xylose conversion, however a
major difference is that increasing solids loading from 10 to 15 % benefits the xylose RI value in
contrast to the glucose conversion RI. In summary, this analysis helped us to determine the
impact of EA pretreatment with respect to a leading ammonia-based pretreatment technology
such as AFEX

TM

. From these results, it is clear that EA pretreatment is more beneficial for

enzymatic hydrolysis performed at lower enzyme loadings and lower solid loadings. These
166

results were expected, since at high enzyme loading the final sugar conversion is very close to
100 % for AFEX

TM

-CS, leaving little room for terminal conversion improvements. However,

the initial rate of enzymatic hydrolysis at high enzyme loading was still considerably higher for
EA-CS, which means that the benefit of using EA pretreatment and high enzyme loading may be
better observed for residence times below 24 h. At higher solids loading, the impact of EA
pretreatment seems to be slightly decreased, since effects such as sugar inhibition [174] start
playing an important role on terminal enzymatic hydrolysis conversion.

167

i)

120 h Sugar Conversion

EA
AFEX
G - Glucose
X - Xylose

Conversion

100%
80%
60%
40%

20%
0%

Enzyme Loading (mg/g glucan)

GX

GX

GX

GX

GX

GX

20

10

7.5

20

10

7.5

Solid Loading (%)

ii)

15

20

Absolute Improvement in 120 h Sugar Conversion
(AI = %XEA - %XAFEX)

Glucose
Xylose

Conversion

25%
20%
15%
10%
5%
0%
Enzyme Loading (mg/g glucan)

20

Solid Loading (%)

15

10
15
15

7.5

20

15

20

10
20
20

7.5
20

Relative Improvement in 120 h Sugar Conversion
(RI = [%XEA - %XAFEX]/ %XAFEX )

iii)

Glucose

Xylose

Conversion

40%
30%

20%
10%
0%

Enzyme Loading (mg/g glucan)

20

Solid Loading (%)

15

10
15
15

7.5

20

15

20

10
20
20

7.5
20

Figure 5-13 i) Terminal (120 h) glucose and xylose conversion as a function of solids loading
and enzyme loading; ii) Absolute improvement (AI) in glucose and xylose conversion observed
TM
for EA-CS with respect to AFEX -CS; iii) Relative improvement (RI)

168

5.5. Conclusion
A larger scale EA pretreatment system, capable of pretreating 3 x 40 g of biomass per run, was
developed to produce enough feedstock for studying high solid loadings enzymatic hydrolysis.
The experimental values obtained for this larger EA system are in agreement with predicted
values from the model developed in CHAPTER 4 for glucose and xylose conversions, as well as
lignin extraction. As this model was developed using experimental data obtained from the small
scale high-throughput EA system, these observations suggest that this pretreatment technology is
scalable without having to perform new optimization protocols. The effects of moisture, and
residence time were evaluated beyond the workspace of the earlier model (see CHAPTER 4).
By increasing the moisture content of the biomass up to 40 %, it was possible to observe a
gradual decrease in glucan conversion and a very marginal increase in xylan conversion. As
moisture reduces cellulose III (CIII) formation, this observation may be a result of lower CIII
content of biomass samples treated under higher moisture conditions. Although XRD analysis
revealed CIII formation under every condition, this methodology revealed limitations to quantify
CIII conversion in lignocellulosic biomass and therefore, further work is required to develop
methods that allow determination of ratios between CIII, CI and amorphous cellulose in EA-CS.
Extending residence time from 30 to 60 min during pretreatment impacted lignin extraction more
than sugar conversion, especially at lower ammonia loadings. At high ammonia loading, i.e., 6:1
ammonia:biomass ratio, the increase in residence time had marginal effects on EA pretreatment
performance, and therefore the EA condition used for enzymatic hydrolysis studies was 6:1
ammonia:biomass ratio, 10 % moisture, 30 min residence time and 120 °C. Under these
conditions, glucan and xylan extraction during EA pretreatment reached 4.7 % and 5.2 %
respectively, while lignin extraction reached 45.3 %.
169

The performance of EA pretreatment was compared with AFEX

TM

during high solid loading

enzymatic hydrolysis, where the effects of 1) enzyme loading, 2) solids loading and 3) residence
time were evaluated. For this purpose, a statistical design of experiments was implemented,
allowing the formulation of mathematical models for predicting glucose and xylose conversions
as a function of the variables mentioned above, for both pretreated substrates. The resulting
models were statistically robust with good predictive ability within their workspace, reflected by
2

predicted R values higher than 88.73 %. Using model predictions, it was possible to conclude
that the initial rates of enzymatic hydrolysis are greater for EA-CS than for AFEX

TM

-CS. Initial

rates seem to be a determining factor for maximizing sugar yields at 120 h incubation time. For
example, little improvement in xylose conversion was observed after 24 h enzymatic hydrolysis,
independently of the enzyme loading and solid loading, showing the importance of the initial
rates of hydrolysis to xylose conversion. Glucose conversion required longer times to reach high
levels; however, most of the terminal conversion was obtained within the first 24 h of hydrolysis.
More importantly, EA pretreatment offers greater benefits to enzymatic hydrolysis at lower
enzyme loading when compared with AFEX

TM

. At high enzyme loadings the benefits of CIII

conversion and lignin extraction are diluted by the fact that enzymes are non-limiting, even
TM

though initial and final conversions on EA-CS are still higher compared to AFEX

-CS. At

high solid loadings, the impact of EA pretreatment on glucose conversion seems to be reduced
with respect to AFEX

TM

pretreatment. Perhaps factors such as sugar inhibition start playing

important roles on the performance of cellulases, especially because sugar conversions can reach
very high values within the first 24 h of enzymatic hydrolysis for EA-CS. Also, for the same
170

solids loading, EA-CS presents higher carbohydrate content due to lignin removal during the
pretreatment process. Therefore, the solids loading effect for pretreated feedstocks with distinct
compositions cannot be compared on the same terms. Although glucose conversion
improvements decrease with increasing solid loadings, EA-CS is still able to achieve
approximately 10 to 30 % higher glucose conversion with respect to AFEX

TM

at 20 % solids. In

the other hand, the beneficial effect of EA pretreatment on xylose conversion improved with
higher solid loadings, in contrast with what was observed for glucose conversion. Understanding
these phenomena requires detailed studies beyond the scope of this work. In that regard, it is
important to determine the contributions of sugar inhibition, lignin inhibition and CIII reversion
during enzymatic hydrolysis for individual cellulases and xylanases. Those studies will help
unveil factors that should be improved for enzyme technology and pretreatment technology in
order to reduce enzyme loading during enzymatic hydrolysis operations at high solid loadings.
From the analysis presented herein, EA pretreatment decreases required enzyme loading by up to
2.7 fold (from 20 to 7.5 mg/g glucan) compared to AFEX

TM

, at both 15 % and 20 % solids,

without reducing performance in sugar conversion. This reduction of enzyme loading will
certainly impact process economics. However, as EA pretreatment operates at higher pressures
and is energy intensive, process economic studies are required to determine whether this enzyme
loading reduction (along with lignin valorization) can offset the increase in pretreatment capital
cost and operating cost.

171

APPENDIX

172

Table 5-3 Enzyme mixtures used during enzyme cocktail optimization for EA and
TM
AFEX -CS
Ctec 2 Htec 2
MP
100
0
0
Mix 1
70
15
15
Mix 2
50
25
25
Mix 3
25
50
25
Mix 4
25
25
50
Mix 5
33.33
33.33
33.33
Mix 6

Table 5-4 Enzyme optimization results for EA and AFEX-CS using enzyme mixtures 1 to 6.
Glucose (Glc), xylose, (Xyl) and arabinose (Ara) concentrations are presented here as a
measure of each enzyme cocktail performance after 72 h enzymatic hydrolysis. Highlighted
in bold letters is the enzyme mixture used for this study.
EA
AFEX
Glc (g/L) Xyl (g/L) Ara (g/L) Glc (g/L) Xyl (g/L) Ara (g/L)
Mix 1
87.17
39.75
4.89
63.86
29.03
3.38
Mix 2
89.53
43.97
6.17
66.17
32.76
4.77
Mix 3
89.86
45.60
6.28
68.05
36.12
5.22
Mix 4
89.46
45.38
6.32
66.38
35.01
5.04
Mix 5
88.82
47.30
6.53
62.56
35.60
5.19
Mix 6
89.25
45.83
6.34
67.36
34.56
5.02

173

CHAPTER 6 - ISOLATION AND CHARACTERIZATION OF LIGNIN DERIVED
FROM THE NOVEL EXTRACTIVE AMMONIA (EA) PRETREATMENT.

6.1. Abstract
With the increasing interest in sustainable lignocellulosic biofuels, and the demand for an
economic conversion process, several authors have begun to look at lignin as a potential revenue
source for the biorefinery. Until recently, ethanol was seen as the only revenue source for the
biorefinery, where lignin-rich residues were envisioned as combustion fuel to generate electricity
and heat to the process. However, these energy forms can be supplied by other kinds of
alternative sources and lignin could be potentially used for further upgrade to liquid fuels or
valuable chemicals. In this context, Extractive Ammonia (EA) is a pretreatment technology of
lignocellulosic biomass that uses anhydrous ammonia in the absence of water to extract lignin
and convert native cellulose I to highly digestible cellulose III by cellulolytic enzymes. Due to
the higher ammonia loadings used in this process, would be important to use the lignin-rich
streams to improve the economics of the process. One of the challenges for lignin utilization is
related with the level of impurities and structural modifications that usually occur during
pretreatment processes. Such factors limit the potential of the lignin as an economic material to
produce bio-based fuels and chemicals. For this reason, a detailed characterization of the various
lignin streams generated by the EA-based biorefinery was performed and discussed herein.
Process mass balances, compositional analysis, structural determination by NMR, elemental
analysis by ICP, molecular weight determination by GPC) and thermogravimetric analysis
(TGA) were performed and discussed in this work to provide a comprehensive understanding of
the potential of EA lignins for commercial applications.
174

6.2. Introduction
Recent advances on ammonia-based pretreatment technologies have demonstrated the
importance of transforming the naturally occurring crystalline allomorph of cellulose I to IIII
[105, 117]. This rearrangement is responsible for improved enzymatic hydrolysis rates up to 2folds in pure cellulose, thereby allowing a reduction of the enzyme loading to achieve equivalent
glucose release compared to native cellulose [105]. Efficient conversion to this crystalline
structure is allowed by contacting liquid anhydrous ammonia at elevated liquid-to-solid ratios (>
3:1, w/w) for relatively short residence times (> 10 min). The utilization of such conditions
during pretreatment of lignocellulosic biomass also enables partial solubilization of plant cell
wall components, which include lignin, monomeric and oligomeric carbohydrates, proteins,
lipids, minerals, etc. The presence of lignin has been considered one of the most detrimental
factors for achieving efficient biomass conversion, due to its role on inhibition of both
cellulolytic enzymes and microorganisms [29, 30]. Therefore, the possibility of efficient lignin
solubilization and subsequent extraction from the carbohydrate polymers is desirable to enhance
enzymatic hydrolysis and fermentation performances downstream. In this context, we developed
a novel Extractive Ammonia (EA) pretreatment process that is able to selectively remove lignin
from corn stover (up to ~50 %), with minimal carbohydrate loss, being simultaneously able to
convert cellulose Iβ to the highly digestible cellulose IIII (CHAPTER 4). While most of the
reports found in the literature describe the utilization of aqueous ammonia to improve biomass
delignification, EA pretreatment process utilizes concentrated anhydrous ammonia at mild
temperatures (< 120 ºC) and relatively short residence times (15-60 min). Variations on this
pretreatment method can also be implemented with the utilization of organic solvent systems in

175

combination with ammonia, allowing the reduction of operation pressure while extracting lignin
and converting cellulose Iβ to IIII (CHAPTER 3 & CHAPTER 4).
Special attention has been recently dedicated to lignin as a possible bio-based precursor of
aromatic building blocks that are currently obtained through petroleum refining (e.g. BTX
chemicals and phenol) [26]. Lignin represents about 10-40 % of the lignocellulosic biomass
content, depending on the type of plant and variety within the same species, as well as on the
growth conditions of the plant [175]. Several methods have been developed to depolymerize
lignin and further convert it to simpler monomeric aromatic compounds. These methods
comprise catalytic cracking and hydrolysis, catalytic and electrocatalytic reductions and
oxidations, among other methods [26]. In most of these processes, the preservation of lignin
functionalities, such as β-O-4 aryl ether linkages, are desirable for efficient conversion to value
added products [26]. Lignin isolation methods often affect these functionalities, possibly
resulting in condensation reactions by the formation of chemical resistant C-C bonds that will
require extra energy and controlled chemistry for selective cleavage [176, 177]. Therefore,
utilization of lignin isolation methods that minimize lignin decomposition, providing lignin
streams closer to its “native” form, are preferable to facilitate depolymerisation chemistry with
high selectivity at acceptable yields. In this work, we present a methodology to fractionate and
isolate lignin from EA pretreatment of corn stover. Further characterization of EA lignin
fractions was performed by composition analysis, NMR, GPC, TGA, DSC and ICP analysis.

176

6.3. Experimental
6.3.1. Untreated corn stover
Corn stover (Pioneer 36H56) was harvested in September 2009 in Wisconsin (USA) and oven
dried at 60 ºC for approximately 2 weeks. The biomass was further passed through a 5 mm
screen installed in a Christy hammer mill (Christison Scientific LTD, England) and stored at 4 ºC
in heat sealed bags prior to utilization. The moisture content of the dried and milled corn stover
was approximately 6 % on a wet weight basis. The biomass composition analysis was performed
using NREL protocols NREL/TP-510-42618 and NREL/TP-510-42620. On a dry weight basis,
the untreated corn stover contained approximately 38 % glucan, 23 % xylan, 1 % galactan, 3 %
arabinan, 14 % Klason lignin, 2 % acid soluble lignin, 5 % ash and 15 % extractives (i.e. ethanol
and water soluble compounds).
6.3.2. Extractive Ammonia (EA) pretreatment of corn stover
EA pretreatment was conducted in a high pressure, stainless steel tubular reactors, equipped with
individual heating mantles, temperature and pressure gauges. The reactors were connected to a
temperature controlled, high pressure flash tank to allow separation of ammonia from the
extractives by evaporation (Figure 6-1). In each reactor, 40 grams of corn stover (dry weight
basis), containing 10 % moisture (wet weight basis) reacted with 240 g of ammonia for 30 min at
120 ºC. Set point temperature was achieved in the first 7 min of reaction, allowing the pressure to
reach 83 bar. The pressure in the flash tank was equilibrated with nitrogen to the same pressure
observed in the reactor during pretreatment. Upon reaction completion, the bottom valve in the
reactor was open, allowing liquid ammonia and soluble extractives to pass through a stainless
steel filter of 50 μm pore size. The ammonia gas was then allowed to be released from the top of
177

the flash tank while pressure was maintained in the reactor at 83 bar using nitrogen for 10 min.
This procedure was designed to maintain ammonia in the liquid state during extraction and allow
residual ammonia to be efficiently released from the biomass. After extraction, the reactor and
flash tank pressures were equalized to ambient pressure by releasing the ammonia and nitrogen
gas from the top of the flash tank. During evaporation, the liquid extractives precipitated in the
bottom of the flash tank.

Figure 6-1 Schematic illustration of the EA pretreatment apparatus and process streams.

The biomass was then removed from the tubular reactors and allowed to dry overnight in the
hood. The system lines were further cleaned with 70 % ethanol and 90 % acetone (both in
water) to remove residual extractives from the lines, which were collected in the flash tank. All
the EA extractives were then drained from the flash tank, collected in a 1 L round bottom flask
and further concentrated in vacuum using a rotary evaporator (BUCHI Labortechnik AG,
178

Switzerland) set at 70 ºC. The EA extractives were then dried using a freeze drier (Labconco,
Kansas City, MO, USA). The dry weight of the extractives and EA pretreated biomass was
further recorded for mass balance purposes. The dried samples were stored at 4 ºC in sealed
containers to avoid major moisture exposure.
6.3.3. EA extractives fractionation and production of lignin rich streams from EA process
Freeze dried EA extractives were solubilized in 100 % ethanol using 1:20 (w/v) extractives-tosolvent ratio for 30 min at continuous mixing conditions. The ethanol insoluble fraction was
filtered using a fiberglass filter installed in a Millipore vacuum filter holder (EMD Millipore,
Billerica, MA, USA). The filtrate was further washed with fresh 100 % ethanol to remove
residual ethanol-soluble components adsorbed to the solid fraction. The solid fraction was air
dried for 2 h in the hood to evaporate residual ethanol in the sample. The dried ethanol insoluble
sample was weighed and placed in a beaker containing distilled water in a 1:30 (w/v) extractivesto-water ratio and stirred for 30 min. The resulting suspension was vacuum filtered using a
fiberglass filter and the filtrate washed with water to remove residual water soluble components.
The water insoluble fraction resulting from this separation was transferred to a pre-weighed
container and dried using a freeze dried (Labconco, Kansas City, MO, USA). The dried sample
was weighed for mass balance purposes and labeled as Fraction 1 (F1). The water soluble
fraction was collected in a round bottom flask and concentrated using a rotary evaporator
(BUCHI, Labortechnik AG, Switzerland) under vacuum, while not allowing it to reach dryness.
The sample was then transferred to a pre-weighed container and freeze dried. The dried sample
was weighed for mass balance purposes and labeled as Fraction 2 (F2).

179

The ethanol soluble fraction was transferred to a pre-weighed round bottom flask and dried
using a rotary evaporator at 60 ºC under vacuum. Distilled water in a 1:30 (w/v) extractives-towater ratio was added to the dried ethanol soluble fraction and mixed for 30 min to solubilize
water soluble extractives. The suspension was vacuum-filtered and washed with distilled water
to remove additional water soluble components adsorbed to the water insoluble fraction. The
water insoluble fraction was transferred to a pre-weighed container and freeze dried. The dry
weight of the sample was recorded for mass balance purposes and labeled as Fraction 3 (F3). The
filtered water soluble fraction was further concentrated using a rotary evaporator under vacuum
at 80 ºC, while not allowing it to reach dryness. The concentrated fraction was transferred to a
pre-weighed container and freeze dried. The dried sample was weighed for mass balance
purposes and labeled as Fraction 4 (F4). All freeze dried samples were stored in the fridge at 4 ºC
in sealed plastic containers prior to utilization to avoid moisture exposure.
6.3.4. Enzymatic hydrolysis (EH) of EA pretreated corn stover
Enzymatic hydrolysis (EH) was performed at 6 % glucan loading, using 15 mg of enzyme per
gram of glucan in a 5 L bioreactor set to control mixing speed at 120 RPM, temperature of 50 ºC
and pH 4.8 for 72 h. The enzymes utilized in this work were Cellic® CTec2 (138 mg protein/ml,
batch No.VCNI0001) and Cellic® HTec2 (157 mg protein/ml, batch No.VHN00001),
generously provided by Novozymes (Franklinton, NC, USA). The enzymatic cocktail was also
supplemented with Multifect Pectinase (MP) (72 mg protein/ml, batch No. 4861295753), a gift
from Genencor (Pala Alto, CA, USA). The protein concentration for the enzymes was
determined using the Kjeldahl nitrogen analysis method (AOAC Method 2001.11, Dairy One

180

Cooperative Inc., Ithaca, NY, USA). The enzyme ratios utilized in this work was 50 % Cellic®
CTec2, 25 % Cellic® HTec2 and 25 % MP in a dry protein weight basis.
After EH, the resulting suspension was centrifuged at 8,000 RPM for 30 min in a Beckman
Coulter Avanti J-26XP centrifuge, equipped with a rotor model JLA 8.1000 (Beckman Coulter,
Inc., Brea, CA, USA), to separate the unhydrolyzed solids (UHS) from the liquid hydolyzate.
The liquid hydrolyzate was decanted to a volumetric cylinder and the volume was recorded for
mass balance purposes. The UHS were washed twice with the distilled water. In each washing
step, the volume of distilled water used was equal to the volume of hydrolyzate generated during
EH. This washing step was performed by sequential re-suspension of the solids, centrifugation
and decantation. The solution resulting from the washing steps was transferred to a volumetric
cylinder and the volumes were recorded for mass balance purposes. Samples of the hydrolyzate
and water washing solutions were prepared for glucose and xylose analysis using HPLC
equipped with a Bio-Rad Aminex HPX-87H column (Bio-Rad, Hercules, CA, USA) as
previously described [88].
6.4. Lignin mass balance
Mass balance on lignin was performed around the EA pretreatment, EA extractives fractionation
and enzymatic hydrolysis. The dry weight loss observed during pretreatment of corn stover was
calculated by measuring the weight of the biomass before and after pretreatment, along with the
moisture content using a moisture analyzer A&D MX-50 (A&D Engineering, Inc., San Jose, CA,
USA). Composition analysis was performed on corn stover before and after EA pretreatment
using the standard NREL protocols NREL/TP-510-42618 and NREL/TP-510-42620. Nitrogen
analysis was performed using a nitrogen analyzer (Skalar PrimacsSNC, Breda, The Netherlands).
181

Lignin extraction yield was calculated by the difference in total lignin weight before and after
pretreatment, divided by the total lignin weight of the untreated sample. The percent recovery of
each fraction was normalized with respect to the total lignin present in corn stover and total
extracted lignin during EA pretreatment.
6.4.1. Characterization of EA extractives fractions
6.4.1.1. Lignin characterization by NMR
6.4.1.1.1. 2D-HSQC NMR analysis
Nuclear magnetic resonance (NMR) spectra of samples in DMSO-d6/pyridine-d5 (4:1, v/v) were
acquired in University of Wisconsin, Madison, using a Bruker Biospin (Billerica, MA, USA)
AVANCE 500 (500 MHz) spectrometer fitted with a cryogenically-cooled 5 mm TCI gradient
probe with inverse geometry (proton coils closest to the sample) and spectral processing used
Bruker’s Topspin 3.1 (Mac) software. The central DMSO solvent peaks were used as internal
reference (δH/δC: 2.50/39.51).
13

6.4.1.1.2. C-NMR and

31

P-NMR analysis

Lignin was isolated from corn stover and UHS by ball milling and organic-solvent extraction.
Ball milled lignin (BML) samples of corn stover and UHS were prepared according to the
procedures described by Guerra et al. [178] and Holtman et al. [179]. Lignin isolation of EA
extractive fractions was performed according to the same procedures as corn stover and UHS,
excluding ball milling.

182

Isolated lignin samples (~100 mg) were dissolved in DMSO-d6 (500 mg) and analyzed by
quantitative

13

C NMR using a Bruker Avance-400 MHz spectrometer at a frequency of 100.59

MHz with an inverse gated decoupling pulse sequence using a 12 s pulse delay and 10K scans.
Quantitative

31

P NMR analysis of BML (~25 mg) was accomplished by using a pyridine/CDCl3

(1.6:1, v/v) solvent, cyclohexanol as an internal standard and 2-chloro-4,4,5,5-tetramethyl-1,3,2dioxaphospholane (TMDP) as the derivatization agent following literature methods [180]. The
31

P NMR spectra were acquired using an inverse gated decoupling pulse sequence with a 25 s

pulse delay and 128 scans.
6.4.1.2. Gel permeation chromatography (GPC)
The isolated lignin samples (100.00 mg) were treated with a mixture of pyridine and acetic
anhydride (1:1, v/v, 4.00 mL) with stirring at room temperature for 24-36 h. The reaction
mixture was diluted with ethanol (30.00 mL) and stirred for 30 min and then concentrated under
lower pressure. The acetylated lignin samples were dissolved in chloroform (2.00 mL) and added
dropwise into diethyl ether (100.00 mL) to precipitate the sample followed by centrifugation.
The precipitate was washed with diethyl ether and centrifuged three times. After air drying, the
acetylated samples were dried for 24 h in a vacuum oven at 40 °C prior to GPC analysis.
Molecular weight determination was conducted using a Polymer Standards Service (PSS) GPC
Security 1200 system equipped with four Waters Styragel columns (HR0.5, HR2, HR4, HR6) at
30 ºC, Agilent isocratic pump, Agilent auto-sampler, Agilent degasser, Agilent refractive index
(RI) detector and Agilent UV detector (270 nm) using THF as the mobile phase (1.0 mL/min)
with injection volumes of 20 μL.
183

The weight average molecular weight (Mw) values of the derivatized lignin samples were
acquired by using an relative calibration curve and this relative calibration curve was created by
fitting a third order polynomial equation to the retention volumes obtained from a series of
narrow molecular weight distribution polystyrene standards (1.36 x 106, 5.38 x 105, 3.14 x 104,
7.21 x 103, 4.43 x 103, 5.80 x 102 g/mol). The curve fit had an R2 value of 0.9984.
6.4.1.3. Elemental analysis
The inorganic elements in the samples were determined by Inductively Coupled Plasma
Emission Spectroscopy (ICP) according to methodology previously employed by Allison et al.
[181]. Nitrogen, Carbon, Hydrogen and Oxygen analysis in fraction F3 were performed by
Galbraith Laboratories, Inc., Noxville, TN, USA.
6.4.1.4. Thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC)
analysis
Thermal gravimetric analysis (TGA) data were obtained using PerkinElmer Simultaneous
Thermal Analyzer (STA) 6000. Samples were put into a ceramic crucible with a lid. The ramp
rate was 10 ºC/min from 25 to 600 ºC. When the temperature reached 600 ºC, it was held under
isothermal condition for 2 minutes. Nitrogen was used as the flushing gas set at a flow rate of 50
mL/min throughout the test.
Differential scanning calorimeter (DSC) data were obtained using a Q20 TA Instruments V24.7
Build 119. Samples were put into an aluminum pan with a lid and underwent a heat-cool-heat
cycle at a ramp rate of 10 ºC/min, including 25 to 400 ºC, 400 to 100 ºC and 100 to 400 ºC. As a
reference, an identical empty aluminum pan with a lid was used. Nitrogen was used as the
flushing gas set at 50 mL/min.
184

6.5. Results and Discussion
6.5.1. EA pretreatment process and lignin mass balance
EA pretreatment conditions that produce high lignin removal, while allowing simultaneous
conversion of cellulose Iβ to IIII, provided the pressure limit for operation of the EA reactors
(~1200 psi). Enzymatic hydrolysis results at 6 % glucan loading demonstrates a superior
digestibility of EA pretreated corn stover (EA CS) when compared to AFEX
TM

stover (AFEX

TM

pretreated corn

CS), which is a leading pretreatment technology, based on ammonia, that

performs particularly well on grasses [182]. From Figure 6-2, it is possible to observe a
maximum glucan and xylan conversions of ~90 % for EA pretreated corn stover, obtained during
the first 24 hours of incubation, while for AFEX

TM

CS was only possible to observe similar

conversions after 72 hours of incubation. These experimental results confirm the point that EA
pretreatment, while removing lignin and converting cellulose I to III (CHAPTER 4) is able to
significantly decrease biomass recalcitrance to enzymatic digestion.
The ammonia-soluble EA fraction, which was recovered after extraction and evaporation of
ammonia, is a black viscous fluid that contains approximately 44 % of the total lignin from
untreated corn stover (F0 in Figure 6-3).

This black liquor was further fractionated by

sequential precipitation using 100% ethanol and water, as demonstrated in Figure 6-3. The
choice of solvents was based on their cost and availability in the ethanol biorefinery, keeping in
mind the goal to pursue the most economic and sustainable fractionation process. In this context,
ethanol will certainly be one of the cheapest solvents available once the lignocellulosic
bioeconomy becomes a reality and will be readily available in the biorefinery.
185

Glucose

Xylose

Glucose

Xylose

Figure 6-2 Sugar conversion based on monomeric + oligomeric glucose and xylose during
TM
enzymatic hydrolysis of AFEX
and EA pretreated corn stover.

Using this methodology, it was possible to fractionate EA lignin extractives into four different
streams (F1-F4 in Figure 6-3). The ethanol insoluble/water insoluble fraction (F1) represents
approximately 1.5 % (dry weight) of the untreated biomass and contains approximately 1% of
the lignin present in untreated corn stover. This fraction contains approximately 69 wt% of lignin
and relatively low levels of carbohydrates and ash, as demonstrated in Table 6-1. Fraction 2 (F2)
is ethanol insoluble and water soluble, containing approximately 3 % of the initial lignin present
in untreated corn stover. Composition analysis results show that this stream is abundant in water
soluble carbohydrates (~28 %), ash (~4 %) and water soluble lignin (~14 %). The ethanol
soluble fractions (F3 & F4) contain most of the lignin extracted during EA pretreatment (F0),
representing approximately 40% of the lignin present in untreated corn stover. After adding
186

water to the ethanol soluble extracts, it was possible to observe the creation of lignin aggregates,
beige in color, while the darker brown extracts were solubilized by the added water. Mass
balance and composition analysis showed that about 32 % of the lignin present in untreated corn
stover is present in water insoluble fraction F3 while the remaining 8% is water soluble (F4).
More importantly, this methodology allowed the isolation of a fraction that contained 92.35 % of
lignin in its composition (i.e., F3).

Figure 6-3 Illustration of EA lignin processing and mass balance. Fraction F6 is the enzymatic
hydrolyzate, rich in glucose and xylose, which is usually fermented to ethanol in a
lignocellulosic ethanol biorefinery. Fractions F1-5 may be utilized for value addition within the
biorefinery, with special attention to fraction F3 and F5 that contain 32% and 51% of the lignin
originally present in untreated corn stover, respectively.
187

From the EA CS that was hydrolyzed to fermentable sugars, it was possible to obtain a solid (F5)
and a liquid (F6) stream. The liquid stream (F6) contains mostly the sugars that will be
fermented to ethanol, while F5 is in great part composed by lignin (42.90 %) and recalcitrant
carbohydrates (25.78 % total) that could not be solubilized in water.

Table 6-1 Compositional analysis of fractions F1 to F5.
F1

F2

F3

F4

F5

Glucan

0.45% 13.29% 0.04% 4.43% 18.97%

Xylan

0.16% 11.72% 0.01% 1.28%

5.54%

Arabinan 0.03%

3.21%

0.00% 0.53%

Acetyl

0.39%

1.24%

0.35% 27.28% 0.36%

Lignin

69.12% 14.44% 92.35% 13.49% 42.90%

Ash

1.43%

4.01%

0.18% 2.35%

1.27%

4.05%

Further work will be required to identify the remaining unknown components of fractions F1-F5.
However, fractions F1, F2 and F4 do not show great potential from the perspective of lignin
utilization, since they only carry 1%, 3% and 8%, respectively, of the total lignin in untreated
corn stover, respectively. In this regard, fractions F3 and F5 are highly enriched in lignin and
therefore, these fractions may provide a good source of aromatic precursors for advanced fuels
and chemicals in an EA-based biorefinery. To evaluate the potential of these lignin streams, a
detailed characterization was performed and will be further discussed herein.

188

6.5.2. NMR characterization of EA extractives fractions from corn stover

6.5.2.1.

13

C NMR and 2D NMR analysis of lignin derived from EA pretreated corn stover

Quantitative

13

C NMR was performed in isolated lignin from fractions F1, F3 and F5 derived

from EA pretreatment process. These are the samples that had the highest content of lignin and,
therefore, are the most representative to perform this analysis. Due to the complexity of these
samples, quantification requires a number of conditions to be fulfilled, which include having a
sample as free of contaminants as possible (e.g. proteins, fatty acids, carbohydrates, etc.) and
having a highly concentrated sample to maximize signal-to-noise ratio during analysis.
Therefore, NMR was performed on isolated lignins by the methods described above. Lignin,
usually defined as a complex polymer composed of p-coumaryl alcohol (H), coniferyl alcohol
(G), and sinapyl alcohol (S), also intercalates p-coumaric acid and ferulic acid units in grasses
(e.g. maize) [152, 153, 183]. These two monomers are typically ester and ether-linked to the
lignin backbone and ferulic acid can also participate in lignin-carbohydrate complex (LCC)
cross-links with arabinoxylans [96, 152, 153]. In Table 6-2, it is possible to observe some ligninspecific NMR assignments and the relative signal intensities with respect to the aromatic
contents (Ar) of fractions F1, F3 and F5. The Ar content was determined by integrating the
signal intensity between 162.0 and 103 ppm and subtracting the integration value for the two
vinyl carbons of ferulate and p-coumarate.This table shows the assignments for conjugated
COOR and CONH2, corresponding to the Cγ carbon signal from both p-coumaric and ferulic
acids and the respective amides (169-166 ppm). Also, it is possible to identify the C4 carbon of
p-coumaric acid from NMR signals at 163-157 ppm. These signals are very similar between the
189

ammonia extracted samples (F1 and F3) and F5. By subtracting the 163-157 ppm integration
from the 169-166 ppm, it is possible to obtain the ferulic content in these fractions (both amide
and acid forms). Using this methodology we can observe that the F5 fraction has the highest
ferulic (0.08 per Ar) content, but the lowest coumaric (0.19 per Ar) content. However, no major
differences can be found between the integration values of all the fractions analyzed here
concerning both coumaric and ferulic contents. The release of ferulic acid and amides during
ammonia pretreatment of corn stover has been reported in literature [23, 97]. As mentioned
before, some of these phenolic residues are often present in the plant cell wall of C4 plants
connected by ester linkages. These linkages are easily disrupted in aqueous ammonia through
hydrolysis or ammonolysis reactions [23, 95, 99], releasing the respective phenolic acids or
amides, which were further extracted by the liquid phase during EA pretreatment. Further
evidence of the formation of amides is revealed by 2D NMR (e.g. pCM – p-coumaroyl amide)
from F1-F4, as can be observed in Figure 6-4. As suggested earlier, it is possible that most of the
ferulic and coumaric content of the biomass was extracted during EA pretreatment, under the
conditions applied in this study (CHAPTER 4). However, both these phenolic acids and amides
are water and ethanol soluble and therefore, they should be present in greater levels in sample F4
after ammonia extraction and fractionation.
The amount of guaiacyl (G) units for the lignin from F1, F3 and F5 can be calculated from the
integration values of the peaks between 123 and 117 ppm, subtracting the values calculated for
ferulic acid (Table 6-2). This methodology suggests that the G content of F1, F3 and F5 is 0.3,
0.23 and 0.3 units per aromatic ring, respectively. Concerning the syringyl content (S), it can be
calculated from half of the integration value of the peak in the 109-103 ppm interval. The values
obtained from this calculation were 0.55, 0.40 and 0.51 S units per Ar, for F1, F3 and F5,
190

respectively. From the S and G content, it is possible to perform S/G ratio calculations, which
resulted in 1.83, 1.77 and 1.72 for F1, F3 and F5, respectively. The S/G ratio can vary
substantially in maize depending on the type of hybrid, cell type and location of the internode
(from bottom to top) of the plant. However, in the literature is possible to observe variations
from 0.69 to 1.47 measured by thioacidolysis of various hybrids [184, 185]. Therefore, the
numbers calculated in this study for the various lignin fractions are at the high end of what has
been normally reported in the literature for maize lignin.
One of the most important linkages in lignin is the -O-4 ether bond, which is usually cleaved by
most alkaline or acidic pretreatments, often leading to further condensation reactions and the
formation of C-C bonds [177]. In the context of lignin valorization, it is important to preserve
these -aryl ether bonds, which could potentially be oxidized or reduced during catalytic
reactions to valuable products [26].
From Figure 6-4, it is possible to observe that the -O-4 linkages (in light blue) are still present
in fractions F1-F4, which resulted from ammonia extraction of corn stover. Quantification of
these linkages can be performed by

13

C NMR integrating the peak at 61-57 ppm, correspondent

the Cγ in -O-4 without α carbonyl structures. The results obtained from this measurement
indicate that these linkages are more abundant in F1 (0.26/Ar), followed by F3 (0.17/Ar) and F5
(0.11/Ar). Therefore, these results suggest that these lignins could offer good potential for
conversion to fuels, chemicals or polymers, with special interest in the highly abundant F3, that
is ethanol soluble and F5 that contains high molecular weight and could potentially be of interest
to the polymer industry.

191

Table 6-2 Assignments and integration value of quantitative
fractionated lignin from EA pretreated corn stover [186].

13

C NMR spectra of
Amount/Ar

Chemical Shift
Assignments
(ppm)
184-180
C=O in spirodienone unit

F3

F5

0.08 0.04 0.07

176-169

Aliphatic COOR

0.40 0.42 1.03

169-166

Conjugated COOR

0.30 0.27 0.27

163-157

C4 of p-coumaric acid

0.25 0.21 0.19

156-142

C3/C’3 in 5-5’ biphenyl and C3/C4 in G units, C3/C5 in S
units

1.79 1.60 1.98

142-124

C1 G units, C5/C5’ in etherified 5-5 units

1.97 2.42 2.07

123-117

C6 in G units

0.34 0.28 0.38

117-113

C5 in G units

0.96 1.01 0.72

113-109

C2 in G units

0.15 0.20 0.17

109-103

C2/C6 in S units

1.11 0.80 1.02

90-78

Cβ in β-O-4, Cα in β-5 and β-β

0.84 0.50 0.51

65-62

Cγ in β-5 and β-O-4 with α-carbonyl structures (G and S 1.28 1.31 1.23
units)

61-57

Cγ in β-O-4 without α carbonyl structures (G and S units)

0.26 0.17 0.11

58-55

Methoxy CH3

0.66 0.36 0.56

53-51

Cβ in β-β and β-5 structures

0.02 0.04 0.04

22-20

Acetyl CH3

0.00 0.36 1.03

a

S/G ratio (I 108-103/2)/I114-108 [187]
a

F1

b

1.83 1.77 1.72

b

I – Integration value, Ar – Aromatic ring

192

F1

F2

F4

F5

F3

Figure 6-4 2D-HSQC NMR of fractions derived from EA pretreated corn stover.
193

31

6.5.2.2. P NMR analysis of lignin derived from EA pretreatment of corn stover
To evaluate the relative distribution of the hydroxyl groups of the various fractions F1-F5,
quantitative

31

P NMR was performed. This technique uses TMDP derivatization, which attacks

all of the exposed hydroxyl groups present in lignin, allowing the analysis by

31

P NMR. Table

6-3 summarizes the results based on the integration of the peaks associated with the different
assignments. The relative abundance of each type of hydroxyl group was calculated dividing the
peak area of each assignment by the sum of the areas of all the peak assignments. From these
results, it is possible to rapidly observe that the aliphatic hydroxyl groups are the most abundant
peaks in all the lignins analyzed in this study. This result is often observed in the literature as it
has been reported by several authors for biomass materials that include wheat straw [188],
switchgrass [189] and poplar [190]. One other important observation is that the most hydrophilic
lignins, i.e., F2 and F4, have more hydroxyl groups, with a total of 9.39 mmol/g and 11.60
mmol/g, respectively. The water insoluble lignins contain very similar amounts of hydroxyl
groups, ranging between 4.87 and 5.08 mmol/g. Also, the relative level of hydroxyl groups from
carboxylic acids is higher for ethanol soluble fractions F3 and F4, with values of 18.59% and
36.59%, respectively. The differences in hydroxyl content and in the percentage of the various
types of hydroxyl groups can provide a great insight about the chemical diversity among the
lignins that were fractionated from corn stover by EA pretreatment. This data, along with

13

C

NMR and 2D NMR analysis presented here, can provide good chemical information to
researchers that are interested of using these streams, especially the abundant F3 and F5, to
convert to fuels and chemicals.
194

Table 6-3 Abundance of assigned hydroxyl groups of fractionated EA extractives from corn
stover, determined by 31P NMR [189].
Percentage of Total -OH
Assignments

F1

F2

F3

F4

F5

Aliphatic -OH

61.89%

84.28%

47.28%

51.53%

73.79%

C5 + C3 Substituted -OH

6.82%

2.99%

8.17%

1.37%

6.85%

C3 Substituted, p-Coumaryl and
p-Hydroxyphenyl -OH

22.39%

4.38%

25.96%

10.51%

11.57%

Carboxylic acid -OH

8.90%

8.34%

18.59%

36.59%

7.79%

5.08

9.39

4.94

11.60

4.87

Total OH (mmol/g of lignin)

6.5.3.Gel Permeation Chromatography (GPC)
The weight average molecular weights (Mw) of the acetylated lignins, isolated from the various
fractions (F1-F5), were measured by GPC (Figure 6-5). These values were calculated in
reference to polystyrene standards, which are not representative of lignin in its shape or
chemistry. Therefore, the Mw values presented in this work cannot be interpreted as being
absolute, but can help us have a relative comparison between the average sizes of the lignin
molecules present in the various fractions and also compare with other literature values derived
by this technique [186].
The chromatograms show that the lignin molecular weights decrease between F1 to F4, as the
Mw value decreases from 3314 g/mol to 416 g/mol. This result was expected, as the ethanol
insoluble fractions should contain higher molecular weight polymers in comparison to the
ethanol soluble fractions. It is also possible to observe that from the ethanol insoluble fractions,
F2 has a greater range of molecular weights compared to F1, as it shows a larger number of
195

peaks in the lower molecular weight region of the chromatogram. These lower molecular weight
components could be associated with highly polar compounds that are not ethanol soluble, but
can be solubilized by water. From the

31

P NMR results discussed above, it is clear that F2 lignin

is highly rich in aliphatic OH groups, which provide hydrophilic properties to the lignin
molecules.
Similarly to F1, the ethanol soluble lignins from F3 and F4 show relatively narrow molecular
weight distributions compared to F2. However, the most important fraction derived from the EA
extractives, i.e. F3, presents an Mw value of 1470 g/mol, which is relatively low compared to F1
and F2 (Figure 6-5) and significantly lower than the Mw from dioxane-extracted lignin from
corn stover (3400-3900 g/mol) [191].The polydispersity of the lignin from F3 was calculated to
be approximately 2. The lowest molecular weight lignin is found in F4, however it is important
to mention that the highest peak shown in the chromatogram of F4 represents the sum of the low
molecular weight molecules, which could not be resolved by the column during analysis.
Therefore, due to this limitation, the Mw value presented in Figure 6-5 (F4) is overestimated.

196

Figure 6-5 Gel permeation chromatograms of fractions F1-F5. Weight average molecular weight
(Mw) was calculated based on polystyrene standards.
The highest molecular weight lignin from the fractions obtained in this study is present in F5.
This lignin was not solubilized by ammonia during EA pretreatment and remained water
insoluble after enzymatic hydrolysis. Due to its abundance, this lignin stream can be potentially
used in chemical conversion, especially in processes that prefer higher molecular weight lignin
polymers. The Mw value obtained by GPC was approximately 4140 g/mol and is slightly higher
than literature values from lignin isolated directly from corn stover (3400-3900 g/mol) by
dioxane or acidic dioxane extraction [191]. This result is expected, as the fractionation process
197

eliminated part of the lower molecular weight lignin from F5, resulting in a higher Mw value.
The polydispersity of the lignin from F5 was also calculated and determined as approximately
4.38, which is considerably higher than the values that are usually observed in dioxane extracted
lignin (i.e. 2.2 to 2.5) [191]. From this fact we may infer that the lignin might have been
subjected to some structural modifications during pretreatment and enzymatic hydrolysis that
made the lignin polymer more heterogeneous in size, even though the average degree of
polymerization was increased by the nature of the fractionation process. However, more studies
are required to evaluate the nature of such structural modifications and the chemistry involved.
6.5.4. Elemental analysis
Elemental analysis of the fractions F1-F5 was performed by ICP and the results are found in
Table 6-4. If we compare the minerals present in F5 with the remaining fractions, we can depict
that the most abundant minerals extracted from corn stover by ammonia include: calcium (Ca),
copper (Cu), iron (Fe), potassium (K), magnesium (Mg), phosphorous (P), sulfur (S) and silicon
(Si). However, in F5 we can still observe relatively high levels of aluminum (Al), Ca, Fe, K, Mg,
sodium (Na), P, S, Si and zinc (Zn) that remained insoluble after pretreatment and enzymatic
hydrolysis.
Most of the ammonia-extracted minerals (F1-F4) were precipitated by the addition of ethanol
and are present in the fractions F1 and F2. As a result, these fractions contain relatively higher
content of mineral elements compared to the ethanol soluble fractions (F3 and F4), notably the
content of K that was exceeded by two orders of magnitude. The addition of water to the ethanol
insoluble extractives was able to enrich mostly K, P and S in F2.

198

From the ethanol soluble minerals, which are present in F3 and F4, P and S are the most
abundant. From these results, it is visible that the early ethanol extraction could deplete these two
fractions from most of the mineral elements. One of the important reasons for this fractionation
process, apart from enriching the EA-extracted lignin, was to remove possible mineral elements
that could potentially interfere with catalytic processes during lignin valorization to fuels and
chemicals. With this respect, this strategy could remove most of the minerals from F3 (the most
important lignin rich fraction), except P (209.7 mg/Kg) and S (1310.0 mg/Kg) that are present in
reasonable amounts. Other strategies could be studied to further remove these residual levels of
inorganic elements, if processing limitations require their total absence. In the other hand,
process streams that are rich in minerals could be potentially used in soil amendment, to improve
the sustainability of the bioeconomy.
In the perspective of lignin processing, the most relevant fraction (F3) was further analyzed by
CHNO analysis. The results show that this sample contains approximately 67 % of carbon (C),
8.2 % of hydrogen (H), 2.3 % of nitrogen (N) and 22.5 % of oxygen (O). Apart from the
presence of relatively high N content, this element composition is very similar to the ones
typically observed in commercial lignins [192]. Nitrogen values are typically below 1% for most
commercial lignins, however F3 lignin was subjected to ammonia treatment which may likely
have functionalized some of the lignin compounds with nitrogen-based groups, such as amides or
amines. Also, is possible that some lipids may be part of the remaining 7.65% that compose the
F3 sample, which contains approximately 92.35% of lignin. The ester linkages present in
triglycerides can react with ammonia during ammonolysis reactions, potentially forming amides,
which may be present in F3 [193]. Moreover, it is also possible that some proteins derived from
corn stover are still present in F3, contributing to the nitrogen content in this sample. The high
199

levels of nitrogen are not only present in F3, but also in the remaining samples from F1 to F4. As
mentioned above, in F4 there is high abundance of coumarates and ferulates that were cleaved
from the plant cell wall structure by hydrolysis and ammonolysis reactions during EA
pretreatment. Ammonolysis reactions formed coumaroyl amide and feruloyl amide that will
highly contribute to the nitrogen abundance of this sample (~13.24 %).

Table 6-4 Mineral inorganic elements and CHNO analysis of fractions derived from the EA
process.
Composition (mg/kg of Fraction)
F1
F2
F3
F4
F5
Element
12.5
11.7
3.0
2.1
390.1
Al
0.4
11.4
< 0.3
< 0.3
3.2
B
2.5
3.3
1.3
1.0
29.0
Ba
< 0.0
< 0.0
< 0.0
< 0.0
< 0.0
Be
361.7
288.6
27.6
24.5
2714.0
Ca
< 0.5
< 0.5
< 0.5
< 0.4
< 0.5
Cd
< 0.7
< 0.7
0.7
1.0
< 0.7
Co
4.8
2.7
0.9
4.6
14.6
Cr
134.0
36.6
33.2
51.7
59.2
Cu
248.5
63.6
11.5
17.5
750.3
Fe
3830.0
26643.4
131.2
75.6
845.9
K
1120.0
496.8
11.2
7.5
345.3
Mg
34.6
20.9
2.1
1.0
28.7
Mn
< 1.6
4.3
1.6
3.2
3.0
Mo
40.3
388.5
39.6
12.3
187.9
Na
3.7
2.5
1.5
10.2
9.8
Ni
898.9
2573.3
209.7
1033.1
754.9
P
1340.0
2305.0
1310.0
911.6
1860.0
S
< 5.1
< 5.1
< 5.1
< 4.6
< 5.1
Sb
< 11.3
< 11.3
< 11.3
< 10.1
< 11.4
Se
100.9
242.9
53.5
100.0
162.6
Si
6.3
8.3
7.4
6.5
4.7
Sn
1.2
1.0
0.1
0.1
7.4
Sr

200

Table 6-4 (Cont’d)
Ti
V
Zn

0.8
< 0.2
62.5
a

ND
C
ND
H
3.67 %
N
ND
O
a
ND – Not determined

1.4
< 0.2
59.7

0.5
< 0.2
30.1

0.1
< 0.2
16.5

13.2
1.6
99.1

ND
ND
2.74 %
ND

67.0 %
8.2 %
2.3 %
22.5 %

ND
ND
13.24 %
ND

ND
ND
ND
ND

6.5.5.Thermogravimetric analysis (TGA) of the EA fractions
The thermal stability of the various fractions obtained in this study (i.e., F1-F5) was measured by
TGA as shown in Figure 6-6. Thermal decompositions were measured in a temperature range
between 100 C and 600 C at a heating ramp of 10 C/min. This figure shows very distinct
curve profiles associated with the thermal decomposition of the different fractions. For all the
samples, a substantial mass loss was recorded at a temperature range between 100 and 450 C,
after which no major decomposition was observed. The water soluble fractions (F2 and F4)
showed significant thermal decomposition at temperatures ranging 100 to 200 C, which was not
observed in the remaining water insoluble fractions (F1, F3 and F5). This observation may be
associated to the fact that these samples are rich in water soluble carbohydrates, most likely
monomeric and small oligomeric sugars that can be thermally decomposed at these temperature
ranges [194, 195]. For example, literature reports show that glucose can be decomposed at onset
temperatures ranging 146-158 C, xylose at 153-156 C and fructose can be decomposed at onset
temperatures ranging from 102-132 C [195, 196]. Also, some of the initial mass loss may be
derived from water evaporation, as these water soluble samples are highly hygroscopic. One
other consideration is that samples F1 and especially F4 contain high levels of acetic acid as can
201

be observed in Table 6-1, which can be evaporated at temperatures of 118 C. All these factors
have contributed to some extent to the weight loss due to thermal decomposition of these two
samples, which resulted in approximately 70 % weight loss for F4 at temperatures below 300 C
and 50 % weight loss for F2 at temperatures below 350 C.

Figure 6-6 Thermogravimetric curves for the various biomass fractions (F1-F5) derived from
the EA pretreatment of corn stover. The heating ramps for all the samples were set at 10 C/min.

Water insoluble fractions, F1 and F3, show a significantly higher thermal stability compared to
the water soluble samples. Sample F1 is slightly decomposed (~5 % weight loss) at temperatures
below 250 C followed by an abrupt decrease in weight (~40 %) before reaching temperatures
around 375 C. This fraction (F1) produced higher levels of solid products of decomposition than
the water soluble fractions, as it was only possible to evaporate about 60 % of its total weight.
202

On the other hand, F3 showed an unusual decomposition pattern for a sample that contains
approximately 92 % lignin. The initial thermal stability in relation to the other samples is
explained by the absence of major carbohydrate content as well as the low presence of volatile
organic compounds. The onset temperature of ~220 C was followed by a rapid decrease in mass
loss before reaching ~420 C, resulting in a mass loss of approximately 70% in F3. For example,
at 10 C/min, organosolv lignin behaves quite differently, as it is possible to observe an onset
temperature around 350 C (~130 C higher than F3) followed by a very slow decrease in mass
weight [197]. At temperatures around 420 C, only about 30 wt% of the organosolv lignin was
lost in TGA (as opposed to 70% in F3) [197]. It is also evident that sample F3 leaves behind very
little solid intermediates after pyrolysis, as only 26 wt% of the sample remained in the TGA tray
at 600 C. At these temperatures, only about 50 % of the organosolv lignin was pyrolyzed to
gases while the remaining 50% required higher temperatures to continue thermal decomposition
[197]. DSC analysis on F3 (in Appendix) also showed that the glass transition temperature (Tg)
value for this sample is approximately 117 C, which helps explain the reason why EA lignin
extraction yields can increase significantly after reaching temperatures ranging 115 to 120 C
(CHAPTER 4).
Sample F5, in the other hand contains high levels of recalcitrant carbohydrates. Cellulose and
hemicellulose decomposition occurs at 315-400 C and 220-315 C, respectively, while lignin
decomposes at a temperature range from 130-1000 C, depending on the lignin origin. Therefore
the thermal decomposition occurring between 220 C and 375C should be mostly associated
with cellulose and hemicellulose present in F5. Lignin decomposition should occur in great part
after 375 C, where is possible to see a very small weight loss with increasing temperature. A
203

similar TGA profile is possible to observe in maplewood, as described by Joungmo Cho et al.
(2012) [197].
6.6. Conclusion
A comprehensive characterization of various lignin-rich streams generated during the processing
of corn stover in an EA-based biorefinery was performed and reported here. By treating corn
stover exclusively with anhydrous ammonia, using 6:1 ammonia to biomass ratio and 120 C for
30 min, it was possible to extract approximately 44% of the lignin from corn stover, which was
further fractionated into four streams (F1-F4) by sequential ethanol and water precipitation. In
one of these fractions (i.e., F3), it was possible to recover approximately 32 % of the lignin
content in untreated corn stover. F3 contained about 92% lignin and was practically carbohydrate
free. Another important lignin rich stream was a solid residue that resulted from enzymatic
hydrolysis of EA pretreated biomass. This fraction (F5) contained about 51 % of the lignin
originally present in the untreated corn stover, however it was composed of only about 43%
lignin.
NMR characterization of the various fractions showed that these lignins contain a high S/G ratio,
ranging from 1.72 to 1.83 and still retain a good level of -O-4 linkages, which were not cleaved
by the ammonia treatment. The preservation of these linkages is of great importance to some
chemical processes that target ether bonds during conversion of lignin aromatics to valuable
products.

31

P NMR also showed that the water soluble fractions of EA lignin (F2 and F4)

contain a considerably higher amount of hydroxyl groups compared to the water insoluble
lignins (F1, F3 and F5), which makes them more hydrophilic, allowing the water fractionation
process to be possible. The weight average molecular weight of these lignins range from 4140
204

g/mol to <416 g/mol according to GPC, against polystyrene standards. The highest molecular
weight was observed in F5, which is comparable to the lignins from various grasses (e.g.
switchgrass) [186] and the lowest was observed in F4. The weight average molecular weight of
F3 was considerably lower than what is usually observed for grasses, with a value of 1470 g/mol.
Therefore, we can conclude that we have produced of a very unique lignin stream in F3, one
which is highly abundant, with low molecular weight, ethanol soluble, with 92% lignin content,
and that maintained most of the native chemistry practically intact. One other feature of this
lignin stream was highlighted by the TGA results, where F3 showed a relatively low
decomposition temperature for lignin, with the formation of very low levels of solid
decomposition products (i.e., ~30 wt%). Furthermore, F3 contains reduced levels of inorganic
elements, which can potentially benefit catalytic processes by avoiding catalyst poisoning. The
lignin from F5, also abundant in the EA-based biorefinery, shows good chemical properties with
the preservation of functionalities such as -aryl ether bonds and high molecular weight.
However this lignin stream requires some further processing to eliminate the carbohydrate
content and thereby improve its quality for future usage.
One of the important goals of this process is also to provide good quality lignin, so it can be used
as a starting material for its valorization and help to reduce biofuel price, contributing to the
economy of the EA pretreatment. With this regard, is expected that the F3 lignin, as well as F5,
can provide good starting materials for lignin-based conversion to fuels and chemicals. Within
the scope of this work, was possible to provide a summary of the most important properties of
these lignin streams, so they can be used in research related to lignin valorization in the future.

205

APPENDIX

206

Figure 6-7 DSC of F3 lignin showing the glass transition temperature at approximately 117°C. Black and red curves (top and
bottom) are sample replicates, while the green curve (middle) is the average of the two replicates.

207

CHAPTER 7 – CONVERSION OF EXTRACTIVE AMMONIA PRETREATED CORN
STOVER TO ETHANOL BY RaBIT PROCESS.

7.1. Abstract
Rapid Bioconversion with Integrated Enzyme Recycling Technology (RaBIT) has been shown as
a very promising process on ammonia fiber expansion (AFEX

TM

) pretreated corn stover (CS) in

terms of enzyme loading reduction and ethanol productivity enhancement. The present study
applied RaBIT on EA (Extractive Ammonia) pretreated corn stover, enzyme loadings were
optimized and the results were compared to conventional SHF process. The RaBIT on EA
pretreated corn stover applied the enzyme loading as low as 8.4 mg protein per g glucan,
however, reached a glucan conversion and a xylan conversion as high as 97.7% and 98.0%,
respectively. The overall ethanol yield was 227 g ethanol/ kg EA-CS (191 g ethanol/ kg
untreated CS). Compared to conventional SHF, enzyme loading was 30% less and ethanol
productivity was two times higher with greater sugar conversions.
7.2. Introduction
The advances in ammonia pretreatment of lignocellulosic biomass and the understanding of how
the plant cell wall chemistry can be modified to facilitate enzyme action towards improved
sugars conversion enabled us to introduce the novel Extractive Ammonia (EA) pretreatment
technology. This technology allies the disruption of lignin-carbohydrate complex (LCC) bonds to
the high selectivity of ammonia for extracting lignin and phenolic compounds, thereby allowing
the removal of these components from the plant cell wall, which are inhibitory to microbes and
enzymes, leaving the desired carbohydrate polymers behind. Additionally, EA pretreatment uses
208

high ammonia-to-biomass ratios (from 3:1 to 6:1) without addition of water, which enables a
rewiring of the hydrogen bonding from native cellulose, yielding cellulose IIII (CHAPTER 4)
This non-native cellulose crystal structure has been identified as being more vulnerable to
cellulase degradation [105, 117]. From these results, initial rates of enzymatic hydrolysis can be
enhanced more than two fold as compared to native cellulose I, which can drive a significant
reduction of enzyme loading for processes that promote this cellulose crystal modification. Our
work has already confirmed the processing advantages of cellulose III conversion allied with
removal of lignin from the plant cell wall. However, the performance of EA-CS hydrolyzates has
not yet been evaluated with current fermentation technology.
New advances in the bioprocessing methodology to convert lignocellulosic biomass to ethanol
have been also investigated with great success. For example, adding to the traditional Separate
Hydrolysis and Fermentation (SHF) and Simultaneous Saccharification and Fermentation (SSF)
a new approach was developed by Mingjie Jin and co-authors [4], denominated by Rapid
Bioconversion with Enzyme Recycling Tehcnology (RaBIT). These processes take advantage of
the fact that the least recalcitrant portion of the biomass consists of nearly 80 % of the total
sugars of the plant cell wall and can be hydrolyzed in the first 24 h of enzymatic hydrolysis. The
resulting hydrolysate is fermented to fuels while the unhydrolyzed solids remaining after the first
24 h are transferred to another vessel where fresh biomass is added along with make-up enzymes
to start another fermentation cycle. This process can be carried on for multiple cycles with
consistent process efficiency. A major advantage of this process is that cellulases and
hemicellulases can be recycled from cycle to cycle, as they are adsorbed to the surface of the
recalcitrant biomass that is transferred to the multiple cycles. This feature allowed 30% enzyme
209

reduction using AFEX

TM

CS as the substrate, when compared to the traditional SHF and SSF

[4]. Additionally, this process could improve ethanol productivity up to approximately four fold
from 0.13 g/L/h to 0.5 g/L/h [4] . To further reduce enzyme usage in the biorefinery process, we
evaluated the integration of EA pretreatment with the RaBIT process. Some of the benefits and
limitations of these two state-of-the-art technologies developed in BCRL, at Michigan State
University, integrated in a full biorefinery concept will be presented and discussed here and
TM

compared with the traditional AFEX

and SHF processes.

7.3. Experimental
7.3.1. EA pretreated corn stover
EA pretreatment method was performed on GLBRC 2009 corn stover as described CHAPTER
5. Pretreatment conditions applied were ammonia loading: 6.0 g/g dry biomass, 10% moisture
content, 120 °C, 1300 psi, and pretreated for 1 h. The composition of EA-CS includes glucan
content: 43.3%, xylan content: 26.7%, arabinan content: 4.0%, acid insoluble lignin content:
9.0% and acetyl content: 1.1%.
7.3.2. Microorganisms and seed culture preparation
Saccharomyces cerevisiae 424A(LNH-ST) [198] was obtained from Prof. Nancy W. Y. Ho,
Purdue University and used for fermenting both glucose and xylose in the EA-CS hydrolysate.
The seed culture was prepared in a 250 ml Erlenmeyer flask with 100 ml YEP medium (10 g/L
yeast extract, 20 g/L peptone, 75 g/L glucose and 20 g/L xylose). A frozen glycerol stock stored
at -80 C was used for seed culture inoculation with a final optical density (OD600) of 0.1. Seed
210

cultures were cultivated at 30 °C and 150 rpm under micro-aerobic conditions for 24 h. Then the
seed culture was centrifuged at 4400 rpm for 6 min and the cell pellets were used for
fermentation inoculation.
7.3.3. Conventional separate enzymatic hydrolysis (EH) and fermentation (SHF)
Conventional separate hydrolysis and fermentation was carried out using 96 h enzymatic
hydrolysis followed by 96 h fermentation. Enzymatic hydrolysis was performed using a
commercial enzyme mixture with a total protein loading of 12 mg protein/g glucan, which
included 66.7% Ctec2 (Novozymes, USA), 16.7% Htec2 (Novozymes, USA) and 16.7%
Multifect pectinase (Dupont, former Genencor Inc., USA). Pretreated biomass solids loading
used in this study was 7.5% (w/w) glucan loading, which corresponding to 17.3% (w/w) total
solids loading. The Enzymatic hydrolysis reaction was carried out in a 250 ml baffled flask with
a total mixture weight of 100 g at 50 °C, 250 rpm and pH around 5.0. After 96 h, the hydrolysate
was centrifuged at 10,000 rpm for 30 min. The supernatant was used for fermentation.
Fermentation was performed at 150 rpm, 30 °C and pH 5.5 with an initial OD600 of 2.0.
A mass balance was performed around the enzymatic hydrolysis step as previously described in
the literature [4, 78]. Total ethanol yield was calculated based on the sugar yield during
enzymatic hydrolysis and the sugar consumption and ethanol metabolic yield during
fermentation. Ethanol metabolic yield (%) was calculated based on the consumed sugars and the
maximum ethanol metabolic yield of 0.51 g ethanol per gram of glucose/xylose consumed.

211

7.3.4. Rapid Bioconversion with Integrated Enzyme Recycling Technology (RaBIT)
The RaBIT process is shown in Figure 7-1, adapted from the work of Jin, Mingjie et al., 2012
[4]. Enzymatic hydrolysis was conducted under the same conditions as the conventional
enzymatic hydrolysis except that it only proceeded for 24 h. The same amount of pretreated
biomass and water were added for enzymatic hydrolysis of each cycle. After 24 h hydrolysis, the
liquid hydrolysate was separated from the unhydrolyzed biomass solids (solid residue) by
centrifugation in 50 ml falcon tubes at 5300 rpm for 20 min. The solid residue was then recycled
and fed to the next cycle of enzymatic hydrolysis. The liquid hydrolysate was fermented with a
high initial yeast cell density (an initial OD600 of 20) at pH 5.5 and 32 °C for 24 h. After
fermentation, the broth was centrifuged and the yeast cells were reused for the next cycle
fermentation. Since the unhydrolyzed solid biomass was recycled with some enzymes absorbed
on it, the enzyme loadings for cycles other than the first cycle were reduced and optimized. We
performed five cycles in total for this study. The last step was carried out with the addition of
water but no fresh biomass input. However, the enzyme loading for the last step was calculated
based on the assumption that the same amount of fresh biomass was added as previous cycles.
After 24 h enzymatic hydrolysis of the last step, the hydrolysate pH was adjusted to 5.5 and yeast
cells were directly added without solids-liquid separation. The last fermentation step also lasted
for 24 h.
Concentrations of glucose, xylose and ethanol were analyzed using an HPLC equipped with a
Biorad Aminex HPX-87H column as described previously [4].

212

Figure 7-1 Diagram of the RaBIT process (adapted from Jin, M. et al. (2012) [4])

7.4. Results and Discussion
7.4.1. Conventional SHF
Conventional enzymatic hydrolysis of EA-CS resulted in approximately 75% total glucan
conversion and 91% total xylan conversion in the first 24 h (Figure 7-2a). However, it took
another 72 h to reach 91% glucan conversion. Rapid hydrolysis at the beginning of enzymatic
hydrolysis and then a reduced rater was typically observed during high solids loading hydrolysis
[199]. The rate reduction is due to several factors including sugar inhibition [174], enzyme deactivation [200] and reduction of substrate availability [172, 173]. Nevertheless, high sugar
conversions (91% for both glucan and xylan) were achieved for EA-CS after 96 h hydrolysis
with a low enzyme loading of 12 mg/g glucan.

213

Figure 7-2 Enzymatic hydrolysis (a) and fermentation (b) performance of conventional SHF on
EA pretreated corn stover.
Hydrolysate fermentation began with glucose and xylose concentrations of 77.4 and 38.5 g/L,
respectively. There were also 10.1 and 17.2 g/L oligomeric glucose and oligomeric xylose,
respectively in the hydrolysate (Figure 7-2a), which cannot be fermented by Saccharomyces
cerevisiae 424A(LNH-ST) . Glucose was rapidly converted to ethanol and yeast cell biomass as
expected (Figure 7-2b). Xylose was consumed to a concentration of 5.9 and 4.2 g/L after 72 h
and 96 h fermentation, respectively. The ethanol concentration was 46.8 g/L for both 72 and 96
h. Therefore, 72 h fermentation results were used for further mass balance and ethanol
productivity calculations. The ethanol metabolic yield in the fermentation of EA corn stover
hydrolysate was approximately 84%, which is lower compared to AFEX

TM

hydrolysate

fermentations that can typically achieve ~90% metabolic yield [78]. Currently, the reasons for
this result are still unclear; however, it is possible that removal of cell growth inhibitors during
pretreatment favor carbon flux toward cell growth, resulting in lower ethanol production. This
214

hypothesis is supported by the fact that the OD600 for EA hydrolyzate is slightly higher that for
TM

AFEX

hydrolyzate (15 compared to 11, respectively) [78]. Another hypothesis is that some of

the nutrients present in corn stover are removed from the biomass during EA pretreatment,
TM

resulting in lower ethanol yields comparing with AFEX

hydrolyzate.

7.4.2. Optimization of enzyme loadings for the cycle 2 and the last step of RaBIT
Enzyme loading for cycle 1 was performed based on the work described in CHAPTER 5.
However, after enzyme recycling, the make-up enzyme must be optimized to obtain constant
sugar yields in all the cycles. For this purpose, an enzyme optimization experiment was
performed as described in Table 7-1.

Table 7-1 Design for the optimization of enzyme loadings (EL) for the Cycle 2 and the last
step.
Total EL
(mg protein/g glucan)
Cycle 1

% Ctec2

% Htec2

% Pectinase

EL1
EL2
EL3
EL4
EL5

12
66.7%
16.7%
Optimization of enzyme loadings for the Cycle 2
0
0
0
66.7%
16.7%
7.5
50%
25%
7.5
80%
12%
7.5
66.7%
16.7%
6

ELa
ELb
ELc

Optimization of enzyme loadings for the last step
80%
10%
3.75
66.7%
16.7%
4.5
66.7%
16.7%
6

a

16.7%
0
16.7%
25%
8%
16.7%

a

10%
16.7%
16.7%

In the experiment of optimizing the enzyme loadings for the last step, the enzyme loadings for
the Cycle 2 was 7.5 mg/g protein and was 6 mg/g protein for the Cycle 3-5. The enzyme
combinations were all 66.7% Ctec2, 16.7% Htec2 and 16.7% Multifect Pectinase.
215

TM

Jin, Mingjie et al., (2012) showed that by implementing the RaBIT process on AFEX

pretreated corn stover, it is possible to reduce external enzyme addition from cycle 2 to 5 by 4050% of the enzyme added in cycle 1, while achieving the same sugar concentrations [4]. EA
TM

pretreatment converts cellulose I to cellulose III (unlike AFEX

), which has a lower binding

affinity to cellulases [118]. Therefore, when the unhydrolyzed solid biomass is recycled, less
number of cellulase enzymes will be adsorbed to it which could negatively affect enzyme
recycling. In other words, enzyme loading reduction for cycles 2 to 5 on EA CS was expected to
TM

be lower than for AFEX

CS. However, also EA removes around 44% of the lignin, which

reduces the non-productive binding of the enzymes to lignin and thereby improves the quality of
the recycled enzymes. Our results on EA CS (Figure 7-3a) showed that with no freshly added
enzymes in the cycle 2, it was possible to produce 30 g/L glucose and 18 g/L xylose after 24 h
TM

hydrolysis, which was lower than the results obtained with AFEX

CS (around 40 g/L glucose

and 20 g/L xylose) [4]. Nevertheless, with an enzyme loading of 7.5 mg protein/ g glucan in
cycle 2 (62.5% of the cycle 1 enzyme loading), it was possible to achieve almost the same sugar
concentrations as in cycle 1 (around 60 g/L glucose and 30 g/L xylose). Thus, under these
TM

conditions, the enzyme loading reduction on EA CS for cycle 2 was similar to that for AFEX
CS.

216

70

a)

Mono-Glc.
Oligo-Glc.

Mono-Xyl.
Oligo-Xyl.

60
50
40
30
20
10
0

Cycle1 Cycle2 EL1
24hEH 0h

80
Concentration (g/L)

Concentration (g/L)

80

70

b)

Mono-Glc.
Oligo-Glc.

Mono-Xyl.
Oligo-Xyl.

60
50
40
30
20
10

EL2 EL3 EL4
Cycle 2 24hEH

EL5

0

ELa
24h

ELa
48h

ELb
24h

ELb
48h

ELc
24h

ELc
48h

Figure 7-3 Optimization of the enzyme loadings (EL) for the Cycle 2 (a) and the last step (b)
during RaBIT process on EA CS. Experiment design was shown in Table 7-1.
TM

Our previous study of RaBIT on AFEX

CS indicated that less xylanases were recycled with

solids compared to cellulases [4]. One possible explanation for this fact is that hemicellulose was
hydrolyzed faster compared to cellulose and hence, less substrate remained for binding xylanases
after 24 h hydrolysis, although considerable oligomeric xylose was generated. This same trend is
observed during RaBIT on EA CS (Figure 7-2a & Figure 7-3a), where hemicellulose is mostly
solubilized in the first 24 h of enzymatic hydrolysis. However, since cellulases present in
commercial cocktails have lower binding affinity to cellulose III from EA CS, it may lead to less
number of cellulases recycled with solids compared to AFEX

TM

CS as well. Under this

hypothesis we investigated various combinations of cellulases and hemicellulases to minimize
enzyme loading while maximizing enzyme performance (Figure 7-3a, EL2-4). Surprisingly, the
results obtained show that increasing the hemicellulases ratio (EL 3) or increasing the cellulases
ratio (EL 4) does not affect significantly the enzymatic hydrolysis performance, as the sugar
concentrations were almost unchanged. The reasons for this observation remain unclear;
217

however, is possible that by increasing the amount of hemicellulases, cellulases could be more
efficient performing hydrolysis of cellulose and vice-versa. From this result we concluded that
the original enzyme combination (66.7% Ctec2+16.7% Htec2+16.7% Multifect pectinase) did
not need to be readjusted and therefore, it was also used for Cycles 2-5 of enzymatic hydrolysis.
Also similarly to Cycle 1, the enzyme loading of Cycle 2 remained as 7.5 mg protein /g glucan,
however as the number of enzyme transfer cycles increased we observed more potential to
reduce make-up enzymes and therefore, an enzyme loading of 6.0 mg protein /g glucan was used
for Cycles 3-5. An enzyme loading of 6.0 mg protein /g glucan was also tested in Cycle 2
(Figure 7-3a, EL5), but sugar concentrations were slightly lower compared to the first cycle.
However, this lower make-up enzyme loading was adequate for Cycles 3-5, as demonstrated in
Figure 7-4a.
The enzyme loading of the last step was also optimized (Figure 7-3b). Three enzyme loadings
and two time points (24 & 48 h) were tested. In this step there was no addition of fresh biomass
and therefore, this step deals with the hydrolysis of the accumulated biomass from the previous
cycles. This results in the presence of a highly recalcitrant substrate in the last step, and because
of that, the sugar concentrations were relatively low without the addition of make-up enzymes
[201-203]. With the addition of an enzyme loading of 4.5 mg protein /g glucan (ELb) (Figure
7-3b), the glucose and xylose concentrations reached 57 and 22 g/L, respectively, after 24 h.
Since most of the xylan was hydrolyzed in previous cycles and the leftover xylan was
recalcitrant, the resulting xylose concentration was lower than in Cycles 1-5. Increasing the
enzyme loading to 6 mg protein/g glucan (same as Cycle 3-5) did not result in the expected
improvement in sugar yield. Further attempts were made to reduce the enzyme loading by
keeping the same levels of Ctec 2 loading (3 mg protein /g glucan) as ELb but reduce the Htec 2
218

and Multifect Pectinase usage, since the available xylan in the last step was lower than in
previous cycles (ELa). However, this enzyme addition reduced the sugar concentration in
comparison to the other conditions evaluated here. It may be that that the recalcitrant biomass
used in the last step contains more branched polymers which could possibly require accessory
enzymes present in hemicellulase cocktails, allowing a greater synergy with cellulases during
enzymatic hydrolysis. However, further work is required to understand and explain this
phenomenon at the molecular level. We also evaluated the effect of extending hydrolysis time to
48 h. However, the increase of hydrolysis time did not show significant improvement of sugar
concentrations (solely 3-4 g/L). Therefore, ELb (4.5 mg protein /g glucan) and 24 h enzymatic
hydrolysis time was used for the last step of the study that will now be discussed.
7.4.3. Performance of RaBIT
Based on the results and discussion of enzyme loading for Cycle 2 and the last step, the
following enzyme loadings were applied for the process: 12 mg protein/g glucan for Cycle 1, 7.5
mg protein/g glucan for the Cycle 2, 6 mg protein/g glucan for Cycles 3-5, and 4.5 mg protein /g
glucan for the last step. The average enzyme loading of the process was 8.4 mg protein/ g glucan.
Cycle 1-5 enzymatic hydrolysis yielded similar glucose and xylose concentrations (around 60
g/L and 30 g/L, respectively), which are enough for producing >40 g/L ethanol if the ethanol
metabolic yield is above 90%. The sugar concentrations were slightly lower after the last cycle of
enzymatic hydrolysis (approximately 55 g/L glucose and 20 g/L xylose). The last cycle of
fermentation is performed without solid-liquid separation, using simultaneous saccharification
and co-fermentation (SSCF) mode, where enzymes are still breaking down the unhydrolyzed
glucan and xylan during fermentation. Therefore, it is possible that the last step fermentation also
219

generates around 40 g/L ethanol, which is a common threshold for economic distillation [166].
The large amount of oligomeric sugars (10-25 g/L oligomeric glucose and 10-30 g/L oligomeric
xylose) generated during enzymatic hydrolysis was the major source of sugar loss since the S.
cerevisiae does not consume those sugars.

Concentration (g/L)

80

Mono-Glc.
Oligo-Glc.

a)

70

Mono-Xyl.
Oligo-Xyl.

60
50
40
30
20
10
1

2

4
3
Cycle

Mono-Glc.
Oligo-Glc.
EtOH

60 b)
50

5

last step

Mono-Xyl.
Oligo-Xyl.
OD600

40
30
20
10
0

1

80 c)
70

2

4
3
Cycle

Oligo-Glc.(%)
Oligo-Glc. (g/L)

5

last step

Oligo-Xyl. (%)
Oligo-Xyl.(g/L)

80
70

60

60

50

50

40

40

30

30

20

20

10

10

0

1

2

4
3
Cycle

5

0
last step

Consumed oligo-sugars (g/L)

Consumed oligo-sugar
percentage (%)

Concentration (g/L) & OD600

0

Figure 7-4 Enzymatic hydrolysis performance (a), fermentation performance (b) and oligomeric
sugars hydrolyzed and consumed during fermentation (c) using the RaBIT process.
220

Fermentation for each cycle also proceeded for only 24 h with a high initial OD600 of 20. Most
of the monomeric glucose and xylose was consumed leaving only 3-7 g/L xylose in the
fermentation broth (Figure 7-4b). It should be noted that the HPLC column used cannot resolve
xylose and arabinose peaks. Since arabinose cannot be utilized by Saccharomyces cerevisiae
424A(LNH-ST), is possible that great part of the leftover sugar (here designated as xylose) is in
reality unfermented arabinose. Therefore, the real leftover xylose concentration is lower than
what is shown in Figure 7-4. The oligomeric sugar concentrations were also reduced after each
fermentation cycle. The enzymes that did not adsorb on unhydrolyzed solids, notably βglucosidase and β-xylosidase, stayed in the liquid hydrolysate, which was recovered from
enzymatic hydrolysis reactor and used for fermentation. Due to short period of enzymatic
hydrolysis (24 h), these enzymes were still active during fermentation of the liquid hydrolysate
and hence still could hydrolyze soluble oligomeric sugars to fermentable glucose and xylose.
This activity on soluble oligomeric sugars could be further facilitated by the fact that
fermentation removed monomeric sugars rapidly, thereby decreasing sugar inhibition of those
enzymes. Around 55%-70% of the oligomeric glucose and 22%-35% of the oligomeric xylose in
the liquid hydrolysate were hydrolyzed to monomeric glucose and xylose, which were further
consumed by the yeast during fermentation (Figure 7-4c). The observed lower reduction of
oligomeric xylose during fermentation in comparison to oligomeric glucose was partly due to the
fermentation profile of this yeast strain. Xylose is typically consumed after glucose is depleted
from the media, which allowed the oligomeric glucose to be efficiently hydrolyzed first.
However, when xylose starts being consumed by the yeast, the ethanol concentration has already
reached inhibitory levels for xylanases, even though xylose inhibition is reduced during
221

fermentation. Consequently, 10-20 g/L of oligomeric xylose were still unutilized from each
cycle.
Using this processing setup, approximately 40 g/L of ethanol were consistently produced in
Cycles 1-5, while in the last step only 31 g/L of ethanol could be generated. The ethanol
metabolic yield for each cycle varied from 74%-80%, which was significantly lower than
TM

AFEX

hydrolysate fermentation (>90%) and also slightly lower than conventional SHF

fermentation (~84%). Lower ethanol metabolic yield of RaBIT compared to SHF was not
observed on AFEX

TM

pretreated corn stover. Zymomonas mobilis, which utilizes Enter-

Doudoroff pathway for fermentation, may be a better choice for enhancing ethanol metabolic
yield in the EA process platform [204]. The cell density (OD600) increased from cycle to cycle,
which guaranteed rapid fermentation of sugars. Recently, a native yeast strain Spathaspora
passalidarum was found able to efficiently ferment glucose and xylose to ethanol [205]. If this
strain is applied for fermentation, it could generate another revenue source from hydrolysate
fermentation because the wild-type yeast residues could be used as animal feed.
7.4.4. Comparison of RaBIT and conventional SHF
The RaBIT process works on the concept that the less recalcitrant portion of the cell wall can be
rapidly hydrolyzed in the first 24h, generating high concentration of sugars that can be readily
fermented to ethanol. Also, this process provides more time for the recalcitrant biomass to be
hydrolyzed by recycling the unhydrolyzed solids to the following cycle of enzymatic hydrolysis.
Using RaBIT process, 97.7% total glucan and 98.0% total xylan conversions during enzymatic
hydrolysis were achieved, which are higher compared to the ones reached by conventional SHF
222

(90.7% and 91%, respectively, Figure 7-5a). However, the monomeric glucose and monomeric
xylose conversions were slightly lower for RaBIT compared to SHF. The major reason for this
result is that in RaBIT, enzymatic hydrolysis was only performed in 24 h while SHF enzymatic
hydrolysis was carried out for 96h. A significant amount of oligomeric sugars was generated
during the early stages of enzymatic hydrolysis and then were slowly degraded to monomeric
sugars in the later stages (Figure 7-2a). Nevertheless, most of the oligomeric sugars were still
converted to monomeric sugars during RaBIT fermentation, which was not observed during
conventional SHF fermentation, most likely because the enzyme activities were very low after 96
h enzymatic hydrolysis. Therefore, higher glucan-to-ethanol conversion and higher total sugarto-ethanol conversions were observed for RaBIT process when compared to SHF (Figure 7-5b).
The xylan-to-ethanol conversion followed a similar trend and was much lower than glucan-toethanol conversion. Higher concentrations of oligomeric xylose are present in the hydrolysate
(Figure 7-2a & Figure 7-4b), and cannot be utilized by this fermentation yeast. This is the major
cause for low xylan-to-ethanol conversion. A similar observation was elaborated in previous
TM

study on AFEX

pretreated corn stover [109].

223

RaBIT

SHF

100

RaBiT

90

b)

Glucan to ethanol
Xylan to ethanol
Total sugar to ethanol

80
70

20

SHF

30

SHF

40

RaBIT

50

RaBIT

60
SHF

Xyl.
Oligo-Xyl.

RaBIT

Glc.
Oligo-Glc.

Sugar-to-ethanol conversion (%)

Sugar conversion (%)

120
a)
110
100
90
80
70
60
50
40
30
20
10
0

10
0

SHF

Figure 7-5 Sugar conversions during enzymatic hydrolysis (a) and sugar-to-ethanol conversions
(b) for the processes of RaBIT and conventional SHF.
Even though RaBIT showed higher sugar-to-ethanol conversion, the ethanol yield was the same
as the conventional SHF process (227 g ethanol per kg EA CS) because of a lower ethanol
metabolic yield (Figure 7-6a). However, the process ethanol productivity was around 2 times
higher than the conventional SHF (0.58 vs. 0.28 g·L-1·h-1, Figure 7-6b) and the enzyme loading
was 30% lower (8.4 mg protein/ g glucan vs. 12 mg protein/g glucan). Higher productivity was
mostly due to less time needed for the RaBIT process (Figure 7-6c).

224

Ethanol Productivity (g/L/h)

Ethanol yield (g/kg E-AFEX-CS)

0.6

250 a)
200
150
100
50
0

RaBIT

0.4
0.3
0.2
0.1
0.0

Conventional
SHF

b)

0.5

RaBIT

Conventional
SHF

c)
SHF

96 h EH

24 h
EH

RaBIT
0

72 h Ferm.

24 h
Ferm.

24

48

72

96

120

144

168

Time (h)

Figure 7-6 Comparison of ethanol yield (a), ethanol productivity (b) and timelines (c) for RaBIT
and conventional SHF processes. Ethanol productivity was calculated based on enzymatic
hydrolysis plus fermentation.
Based on 1 kg of untreated corn stover, 840 g of EA pretreated corn stover were obtained after
pretreatment, which contained 401.3 g of glucose equivalent glucan and 254.8 g of xylose
equivalent xylan (Figure 7-7). The maximum ethanol yield from this amount of sugar is
approximately 334.6 g. Both the RaBIT process and the conventional SHF yielded around 190 g
of ethanol from 1 Kg of untreated corn stover. However, the enzyme addition per kilogram of
untreated corn stover was 3.0 g for the RaBIT process versus 4.3 g for the conventional SHF
process. In the RaBIT process, around 2% and 23.3% of the total sugars were lost in the solid
unhydrolyzed biomass and in the liquid fermentation broth, respectively, which corresponded to
225

a maximum loss of 85 g ethanol. This result means that 74.5% of the total sugars (488.8 g) could
be converted to ethanol, yielding 249.3 g of ethanol considering for 100% metabolic yield. This
amount is considerably higher than what has been reported in the literature [4]. However, the
ethanol metabolic yield of the fermentation was low (around 76.6% of the theoretical maximum).
This problem can probably be solved with some stain development and adaptation, to direct the
carbon flux towards ethanol production instead of cell mass production. Nevertheless, we still
TM

achieved the same ethanol yield (191 g ethanol/kg CS) as reported for AFEX

CS [78] due to

higher sugar-to-ethanol conversion. However, this result was obtained with much lower enzyme
loading and higher ethanol productivity. If the ethanol metabolic yield issue can be solved in the
future by achieving values near 90%, it will be possible to generate 224 g of ethanol from 1 kg
CS combining the RaBIT process and EA pretreatment technology.
As a side product, 29.3 g and 25.8 g glycerol per Kg untreated corn stover was produced from
RaBIT and SHF processes, respectively (Figure 7-7). Glycerol production could alleviate the
redox imbalance caused during xylose metabolism through xylose reductase and xylitol
dehydrogenase. Another side product is the unhydrolyzed solids, which compositions are showed
in the Table 7-2. Currently, the unhydrolyzed solids are considered to supply heat/electricity for
the biorefinery.

226

RaBIT

RaBIT

Figure 7-7 Mass balance comparison of RaBIT and conventional SHF processes.

Table 7-2 Compositions of unhydrolyzed solids from RaBIT and conventional SHF
processes.
Process
RaBIT
SHF

% Glucan

% Xylan

% Arabinan

% Klason Lignin

% Ash

11.6%
±0.8%
18.1%
± 2.4%

6.1%
±0.3%
7.2%
± 1.1%

0.9%
±0.1%
1.4%
± 0.1%

47.4%
±0.3%
43.7%
± 0.4%

11.4%
±0.1%
9.3%
± 0.8%

7.5. Conclusion
The integration of the novel EA pretreatment with state-of–the-art bioprocessing technologies,
such as RaBIT, further reduced the enzyme loading by 30% (from 12 mg/g to 7.5 mg/g glucan)
compared with the conventional SHF process. It was demonstrated that processing engineering
innovations allied with novel pretreatment strategies can decrease enzyme usage in the
biorefinery and reduce processing time and capital costs. It is also important to mention that
227

conventional AFEX

TM

technology uses approximately 20 mg/g glucan for optimal results in a

traditional SHF process, as reported in CHAPTER 5. Therefore, this technology advancement
and, in particular, the integration of EA pretreatment with the RaBIT process, allowed a very
substantial decrease in enzyme loading. Furthermore, RaBIT process enhanced ethanol
productivity by 2 fold and allowed higher sugar conversions compared to SHF process on EA
CS.
While continuing to develop incremental improvements to this novel biomass processing
strategy, it is also important to generate appropriate enzymes and microorganisms that are
adapted to the properties of these new substrates. For example, developing novel strains that can
generate higher ethanol metabolic yields is important to continue decreasing enzyme loading and
improve process yields. Also, advances in enzyme technology can provide novel engineered
cellulases with higher binding affinity to cellulose III, which will not only allow improvements
in glucan conversion rates but also more efficient cellulase recycling, which is currently not
possible with wild-type cellulases.

228

CHAPTER 8 –

EXTRACTIVE

AMMONIA

(EA)

PRETREATMENT

PROCESS

DESIGN, PRELIMINARY ECONOMICS AND FINAL CONSIDERATIONS.

8.1. Abstract
The novel Extractive Ammonia (EA) pretreatment of lignocellulosic biomass has been studied
on various fronts that include the three most important units of the biorefinery: pretreatment,
enzymatic hydrolysis and fermentation. Even though this pretreatment technology presents good
downstream processing performance, the most important aspect to consider is the overall
economics of the EA-based biorefinery. Therefore, this study aimes to evaluate EA pretreatment
economics, by determining capital and operating costs. These were adapted to the existing
downstream operations described by the NREL economic model for lignocellulosic ethanol
production. By coupling the economic model with enzymatic hydrolysis and fermentation
performance results, a techno-economic evaluation of the EA biorefinery was performed. For the
base case pretreatment conditions (6:1 ammonia to biomass ratio, 30 min residence time and 120

°C), the minimum ethanol selling price (MESP) at an internal rate of return (IRR) of 10 % was
TM

$2.36 per gallon of ethanol. Compared with an AFEX

biorefinery, this value is slightly higher

(about 8 cents per gallon of ethanol); however, there is great room for improvements as this
result represents a preliminary level of optimization on the pretreatment conditions and process
design. From sensitivity analysis is clear that the heat duty during ammonia-water separation can
highly impact the economy of the EA biorefinery. Also, the lignin selling price is a very
important factor that can dictate the viability of the EA process. These and other considerations
will be evaluated in detail in this work.
229

8.2. Introduction
The development and study of Extractive Ammonia (EA) pretreatment has provided
understanding on how plant cell wall modifications can improve enzymatic hydrolysis yields and
potentiate reduce enzyme loadings during biorefinery operations. However, besides the scientific
knowledge that has been presented in the previous chapters of this dissertation, is also important
to provide techno-economic information and thereby determine the most important areas for
scientific and technological development. In 2011, the National Renewable Energy Research
Laboratory (NREL) published a new technical report and respective model tools to simulate
biorefinery conditions based on dilute acid (DA) pretreatment. This report provides the most
updated and detailed economics information on the literature [206]. From the report, a minimum
ethanol selling price (MESP) of approximately $2.15 per gallon ethanol ($3.27/gal gasoline
energy equivalent) was determined for a DA-based biorefinery. Based on this study, this work
will focus on determining the impact of EA pretreatment on the overall economics of the
biorefinery. The base case pretreatment conditions for EA will be considered as 6:1 ammonia to
biomass ratio, 30 min residence time and 120

°C.

These conditions are not optimized for

achieving the lowest ethanol price, but will serve as a benchmark for future improvements of this
technology. For this base case, EA process design will be simulated to generate material and
energy balances for processing 2000 ton/day of biomass, which will be further used for sizing of
equipment and determination of energy requirements. From this information, capital and
operational costs of EA pretreatment will be determined and incorporated into the NREL
economic model [206], replacing the existing DA pretreatment. Downstream conditions will be
optimized using the enzymatic hydrolysis performance models generated for corn stover in
230

CHAPTER 5, which will generate the most profitable biorefinery operation for the EA base
case scenario. The MESP will be determined for the base case EA biorefinery and compared
TM

with an AFEX

biorefinery. Further sensitivity analysis will be used to provide technical

recommendations for research and development of new processing conditions of EA
pretreatment, always aiming to improve the profitability of the process.
8.3. Experimental
8.3.1. EA pretreatment process flow diagram (PFD)
The EA process was scaled based on the batch design presented in CHAPTER 4 to pretreat
2000 tons of dry biomass per day. The process flow diagram is presented in Figure 8-5, and
consists of two reactors that operate intermittently in what we called a “reactor system”. After
pretreatment, the ammonia from one reactor is always recycled to the other reactor belonging to
the same system. The pretreatment unit operation for 2000 ton per day is composed of five
reactor systems with two pretreatment reactors each, in a total of ten reactors. For this operation,
two ammonia recycling systems comprised of one distillation column, a compressor, a pump and
a flash evaporator, operate continuously throughout the production of EA pretreated biomass,
following the schedule in Figure 8-6. For this analysis, approximately 13.9 tons of dry biomass
were added to each batch reactor (streams 1), along with 1.39 tons of water and 83.4 tons of
ammonia (6:1 ammonia to biomass ratio) in our base case (streams 2). The reactor operates at
1200 psi of pressure and 120

°C

for 30 min residence time. The liquid ammonia is filtered

through the bottom of the reactor and flows (streams 7) to a distillation column that operates
under 200 psi pressure. The top stream of the distillation column (stream 9) is fully condensed to
231

liquid and is pumped to a new reactor that has been freshly loaded with biomass. The bottoms of
the distillation column contain lignin rich extractives, which can be used as a valuable byproduct for the biorefinery (stream 8). The EA pretreated biomass is further transferred to a flash
evaporator column (stream 3) where heat is applied to recover the ammonia adsorbed on the
biomass surface and inside the pores. The evaporated ammonia (stream 5) is further compressed
to the ammonia tank (stream 6) and reused in the pretreatment system. For all the ten reactors
that compose the pretreatment operation, two evaporators and two compressors are used
continuously during ammonia recycling.
8.3.2. Material and energy balances around EA pretreatment
Material and energy balance calculations around ammonia-water system were performed using
Aspen Plus software (Aspen Tech, Burlington, MA, USA). Biomass energy balance calculations
were added based on heat capacity and enthalpy values, which were determined through
regression of experimental numbers obtained for lignin [207] and cellulose [208].
Thermodynamic parameters for hemicellulose and the remaining unknown components of the
biomass were assumed to be the same as those for amorphous cellulose [208].
The reactor was modeled based on flash calculations at constant pressure (1200 psi) and
temperature (120

°C)

for the ammonia-water system, entering at 25

°C

and 150 psi. All

thermodynamic calculations were determined using the Non-Random Two-Liquid (NRTL)
electrolyte model with default parameters provided by Aspen Plus software. The heat required to
increase biomass temperature from 25 to 120 °C during reaction was added to the heat duty
calculations. Heat of mixing of ammonia with water and heat of adsorption of ammonia/water to
the biomass were not accounted to simplify these calculations. This assumption was made based
232

on the fact that ammonia loadings during EA pretreatment are about sixty times higher than
water loadings and also because the amounts of ammonia/water adsorbed on the biomass is
considerably smaller than the free ammonia in the bulk phase. Mass balance on the reactor was
performed experimentally as described in CHAPTER 4. In this study, ammonia reactions with
biomass were considered to result in 0.5 wt% ammonia loss (from the total ammonia added to
pretreatment) [95], where ammonia is chemically bound to the biomass. This assumption is
based on information provided by Michigan Biotechnology International (MBI), who is
performing ammonia recycling tests during AFEX pretreatment. The chemistry of AFEX is
similar to EA pretreatment and therefore, we assumed the ammonia loss per amount of pretreated
biomass to be the same.
The energy balance for the evaporator was simulated based on flash calculations for ammonia,
which was allowed to reach a vapor fraction of 1.0 at ambient pressure, assuming an input of
ammonia in liquid state (pressure of 150 psi) at ambient temperature. The enthalpy of desorption
of ammonia from the biomass (33 kJ/mol adsorbed ammonia) [209] and the enthalpy associated
with biomass temperature variations was accounted for during heat duty calculations for the
evaporator. Mass and energy balances for distillation were modeled based on ammonia-water
mixture with RadFrac simulation tool from Aspen software. Extractives composition and mass
balances were not accounted for during distillation calculations. However, the heat duty required
to increase the extractives temperature in the reboiler was accounted for the heat requirements of
the distillation process. The extractives were assumed to be recovered on the bottom of the
distillation column with most of the water entering in the front of the process.

233

8.3.3. Equipment sizing and cost
All equipment, with the exception of the reactor and ammonia evaporator, were sized and cost
estimated using Aspen Capital Cost Estimator software (Aspen Tech, Burlington, MA, USA).
The reactor volume was sized based on the experimentally measured capacity of 1L for every
100 g of dry biomass input. The ammonia evaporator was sized assuming that the biomass
volume occupies 1/3 of the evaporator volume. Capital cost estimations for these two pieces of
equipment were performed according to tabulated values for high and low pressure vessels in the
literature [210-213].
8.3.4. Techno-economic analysis
A biorefinery designed to process 2000 ton biomass/day was evaluated for 30 years of operation
based on the NREL economic model for biochemical conversion of lignocellulosic biomass to
ethanol [206]. The capital and utility costs for EA pretreatment were calculated in this work and
added to the NREL economic model in replacement of the dilute acid pretreatment. The base
case conditions for EA pretreatment was defined as 6:1 ammonia to biomass ratio, 120 °C and
30 min residence time. On Table 8-12, the capital cost for EA pretreatment and utility costs
assumed for economic analysis are given. The cost of delivered biomass at the biorefinery was
assumed to be $60 per dry ton (similarly to the NREL model) and ammonia price was assumed
to be $600 per ton [214] in the base case. Revenues from lignin were also accounted with a
selling price of 10 cents/lb [215]. Economic evaluation was performed without in-house power
production from unhydrolyzed solids that are released after enzymatic hydrolysis and
fermentation because they are assumed to be used as feedstock for lignin valorization. Instead,
electricity is bought from the grid at cost of 5.72 cents/kWh.
234

The MESP was calculated based on an internal rate of return of 10% for different processing
conditions of enzymatic hydrolysis using a cellulase cost of $3600/ton and hemicellulase cost of
$4500/ton, as the base case scenario, which results in a similar average value to the NREL model
[206]. These conditions were the same as used on the Box-Benhken design from CHAPTER 5,
with variable solid loadings, enzyme loadings and incubation times. Experimental values of
enzymatic hydrolysis were obtained from the work presented in CHAPTER 5. Fermentation
metabolic yields were fixed constant at 90% of theoretical. This fermentation performance was
TM

assumed based on the yields obtained for AFEX

treated corn stover (AFEX

TM

-CS)

hydrolysates [78]. The calculated MESPs for each enzymatic hydrolysis condition of EA-CS
were used to produce an empirical, full quadratic regression model that relates MESP ($ gallon
EtOH) with processing variables during enzymatic hydrolysis. This model, produced with
Minitab software (Minitab Inc., State College, PA, USA), was used to further optimize the
enzymatic hydrolysis conditions that generate the lowest MESP. This lowest MESP value for the
TM

EA biorefinery was further compared with the lowest MESP obtained for an AFEX

-based

biorefinery, also using the same methodology as described for EA-CS, with experimental values
from CHAPTER 5. For the AFEX
in the biorefinery. Instead, AFEX

TM

TM

-based economic model, the pretreatment was not included

-CS pellets were assumed to be delivered at the door of the

biorefinery from regional biomass processing depots (RBPDs) [216]. The cost of AFEX

TM

-CS

pellets was assumed to be $120 per ton (dry weight), as estimated by Michigan Biotechnology
Institute (MBI). Sensitivity analysis on the MESP was performed for enzyme cost, biomass cost,
lignin price, utility cost and capital cost of EA pretreatment.
235

8.4. Results and Discussion
8.4.1. Material and energy balances for the EA pretreatment model
The preliminary EA pretreatment configuration presented herein (Appendix Figure 8-5) is
composed of a set of five reactor systems, with two reactors each, and two ammonia recycling
systems containing by a pump, a distillation column, an evaporator and a compressor each.
While the pretreatment reactors are operating in batch mode, ammonia recycling is operated
continuously to fulfill the demand of the ten reactors that compose this unit operation (see
schedule on Appendix Figure 8-6). Mass and energy balance calculations were performed
assuming the base case EA conditions of 6:1 ammonia-to-biomass ratio, 120 °C and 30 min
residence time. From the material balances around EA pretreatment, presented in Appendix
Table 8-4, it is possible to predict loss of ammonia from the bottom of the distillation column.
This loss represents 0.12% of the ammonia input during pretreatment. The major ammonia loss
represents 0.5% of the ammonia loaded during pretreatment, which is chemically bound to the
biomass. This assumption was based on previously quantified ammonolysis products and Millard
reaction products formed during AFEX pretreatment [23, 95, 97], which was assumed to be
similar for EA pretreatment. More detailed ammonia mass balance should be performed
experimentally to verify the validity of this assumption for a more detailed economic evaluation.
The mass balance around the biomass was determined experimentally from CHAPTER 4 and
CHAPTER 5 where about 45 % of the lignin present in the biomass was extracted during the
process. Energy balances on the biomass streams were evaluated based on heat capacities from
literature associated with crystalline and amorphous cellulose and lignin [207, 208].
Hemicellulose heat capacities and enthalpies were assumed to be the same as amorphous
236

cellulose. From those literature values it was possible to account for changes in composition
during energy balance calculations for every stream, which are presented in Appendix Table
8-4. Biomass composition changes for the various pretreatment streams were determined
experimentally and are presented in CHAPTER 5. Regarding the water mass balance, some
water cannot be efficiently removed from the system, causing water build up. The optimum
moisture content during EA pretreatment is 10 % in a dry weight basis. To avoid water build up
in the system and maintain the pretreatment operating at steady state, the biomass should contain
about 7% moisture when added to the EA pretreatment system, because the remaining 3%
moisture is already present in the ammonia recycling system.
The energy requirements of each unit operation involved on EA pretreatment are shown in
Appendix Table 8-5. From this table, it is possible to observe that EA process is energy
intensive, especially during ammonia recovery. Simulation values for total heat requirements of
EA pretreatment are 9.08 GJ/ ton biomass and electricity requirements are 89.59 kWh/ ton
biomass. The energy requirement for the EA pretreatment operation represents about 50 % of the
HHV of the corn stover input. Analyzing the distribution of these energy requirements by the
various equipments is possible to observe that the distillation is the most energy demanding
operation, covering about 73 % of the total heat duty of EA pretreatment, followed by the
reactors (20.59 %) and evaporator (6.21 %) (Figure 8-1). The electricity duty is shared by the
pump and compressor, where the later represents about 92.43 % of the total electricity
consumption and the former represents 7.57 % (Figure 8-2).

237

Heat
consumption
pretreatment
Heat
consumption
for on
EAEA
pretreatment
6.21%
20.59%
Reactors

Distillation
Evaporator

73.20%

Figure 8-1 Pie chart describing the heat consumption of the various unit operations during EA
pretreatment. The total heat duty is 9.08 GJ/ ton biomass.

Electricityconsumption
consumptionfor
on EA pretreatment
Electricity
pretreatment
7.57%

Compressor
Pump

92.43%
Figure 8-2 Pie chart describing the heat consumption of the various unit operations during EA
pretreatment. The total electricity consumption is 89.59 kWh/ ton biomass.
238

8.4.2. Capital cost of EA pretreatment
EA pretreatment equipment was sized based on Aspen Plus simulations and on experimental
information. In Appendix Table 8-6 to Table 8-11, the various pieces of equipment used in the
EA pretreatment system are described. The cost estimation was performed with Aspen Capital
Cost Estimator for most of the equipment except the reactor vessel and the evaporator, which
were estimated based on literature information [211-213]. Table 8-1 summarizes the equipment
used for a 2000 ton/day of biomass processing capacity with respective unit cost, number of units
and total cost. Analyzing this information, is clear that the pretreatment reactors represent the
highest expenditure on the EA capital cost ($17,447,100.00), followed by the distillation
columns ($12,789,984.00) and the compressors ($7,298,000.00). The other unit operations are
relatively cheaper and do not impact capital cost significantly. The total capital cost estimated for
the EA pretreatment system was $51,907,471.20 for the base case scenario. Compared with the
DA pretreatment operation for the same size biorefinery, EA pretreatment is more capital
intensive, as DA pretreatment unit was estimated to be purchased under MM$30 [206]. Thus EA
pretreatment would benefit more from being implemented at a very large scale in centralized
biorefineries.

Table 8-1 Summary of equipment and capital cost for EA pretreatment.
Name
Unit Cost ($) # Units Total cost ($)
Reactor
1,744,710.00
10 17,447,100.00
Distillation Column
6,394,992.00
2 12,789,984.00
Evaporator
360,910.00
2
721,820.00
Compressor
3,649,000.00
2 7,298,000.00
Pump
354,775.00
2
709,550.00
Ammonia Tank
320,790.00
3
962,370.00
1.3
Installation factor
Total installed cost ($) 51,907,471.20
239

8.4.3. Material balance for enzymatic hydrolysis and fermentation and MESP calculation
To evaluate enzymatic hydrolysis conditions that produce the optimal MESP, a Box Benhken
design of experiments was implemented with the same enzymatic hydrolysis parameters as
described in CHAPTER 5. The sugar yields obtained for EA-CS as a function of pretreatment
conditions were presented in CHAPTER 5 and were used in this work for determination of the
economic optimal conditions. For this purpose, enzymatic hydrolysis conditions and respective
sugar yields were added to the modified NREL economic model. This model was built based on
the NREL economic model for Dilute Acid (DA) pretreatment, where EA pretreatment data
(both fixed and variable costs) were introduced in replacement of DA pretreatment parameters.
The MESP was calculated for the various enzymatic hydrolysis conditions for a 30 year plant
operation, with 2000 ton/ day of dry biomass capacity, using an internal rate of return of 10%.
Fermentation performance was assumed to be constant at 90 % of the theoretical metabolic yield
from the sugar produced during enzymatic hydrolysis. The costs associated with pretreatment are
defined in Appendix Table 8-12. For the economic model, both the EA extracted lignin and the
lignin from unhydrolysed solids were considered as revenue for the biorefinery, with an average
selling price of $220 per ton of lignin. The lignin market is currently occupied by kraft lignin,
ligno-sulphonates or organosolv lignin. The Chinese market for lignin seems to be the most
competitive, with prices ranging 200 to 600 US dollars per metric ton (FOB), depending on the
type and quality of the lignin [215].
The calculated MESP values after running the economic model were further processed using
Minitab software to fit a full quadratic equation as a function of enzymatic hydrolysis parameters
(i.e., incubation time, solids loading and enzyme loading). The residuals and coefficients
obtained for this model are presented in Appendix Figure 8-7 and Appendix Table 8-13,
240

respectively. From the residual plots, it is possible to observe that the residual values are
reasonably dispersed throughout the MESP range and do not denote any systematic deviation.
The histogram also presents a reasonably normal distribution with the highest number of
residuals being similar to zero, as expected. From the coefficient table, it is possible to observe a
P value above 0.05 for the terms Enzyme Loading and Solid Loading*Solid Loading. These
terms were considered in the model equation because their absence impacted negatively the
2

predictive ability of the model (predicted R ) and the goodness of the fit. Using this model along
with the optimization tool from Minitab software, the minimum MESP was determined along
with the enzymatic hydrolysis conditions (time, enzyme loading and solids loading) that provide
this ethanol price. Appendix Figure 8-8 shows the optimization plots obtained in this study,
with a predicted MESP of $2.37 per gallon of ethanol. The enzymatic hydrolysis conditions and
predicted glucose and xylose yields are also shown. The optimal enzymatic hydrolysis condition
that projects the lowest MESP for EA-CS is 93.7 h incubation time, 21.44% solid loading and
7.5 mg enzyme/g glucan. At these conditions, the regression model predicts a monomeric
glucose conversion of 72.46 % and monomeric xylose conversion of 69.77 %. When introducing
these predicted enzymatic hydrolysis results and processing conditions into the modified NREL
economic model for EA pretreatment, it was possible to recalculate the MESP and compare with
the regression model prediction. The value given by the modified NREL economic model for
these enzymatic hydrolysis conditions was $2.36 per gallon of ethanol, which means the MESP
prediction from the regression model was $0.01 higher than the value calculated by the modified
NREL model for that processing condition.

241

Table 8-2 shows the ethanol yields on a mass and volume basis, per ton of biomass and per year,
using the optimal conditions of enzymatic hydrolysis for the base case EA pretreatment
condition. From this table, it is possible to observe that for each ton of untreated biomass
entering the biorefinery about 61.8 gallons of ethanol are produced, which results in an annual
production of 47.74 million gallons of ethanol. The ethanol titer after fermentation was projected
to be 6 %, based on a fermentation yield of 90 % of theoretical maximum. Extracted lignin from
the EA process is about 63 Kg lignin/ton of biomass, while about 77 Kg lignin/ton of biomass
are recovered from the unhydrolyzed solids after enzymatic hydrolysis and fermentation.
For the overall EA biorefinery economics in the base case scenario, the total capital investment
was approximately MM$292.10 (Table 8-3). The MESP of $2.36 per gallon of ethanol was
calculated based on an IRR of 10 %. For this process, the enzyme cost per gallon of produced
ethanol is approximately $0.16, which is considerably lower than the one obtained by the AFEXbased biorefinery ($0.43). This result is a direct consequence of the lower enzyme loading used
in the EA biorefinery during enzymatic hydrolysis, due to the high digestibility of EA-CS.

Table 8-2 Ethanol and lignin yields for the optimized base case scenario of an EA
biorefinery.
242.20
Ethanol yield (kg/ton BM)
61.80
Ethanol yield (gal/ton BM)
47.74
Ethanol volume (MMgal/yr)
0.06
Ethanol titer (w/w)%
63.00
Extracted lignin (kg/ton BM)
77.00
Lignin from solids (Kg/ton BM)

242

Table 8-3 Economic considerations for the base case EA biorefinery (2007 US$).
292.10
Fixed Capital Investment (MM$)
6.46
TCI/Annual gallon Ethanol ($/gal EtOH)
2.36
0.10
0.16
0.19
0.53

Ethanol Selling Price ($/gal EtOH)
IRR
Enzyme Cost ($/gal EtOH)
Fixed Costs ($/gal EtOH)
Other Operation Costs ($/gal EtOH)

8.4.4. Performance of EA-based biorefinery
A comparative study was also performed to evaluate the performance of the EA biorefinery and
the base case scenario presented here. Figure 8-3 shows the MESP values for biorefineries based
TM

on two different ammonia pretreatment technologies, notably AFEX
TM

AFEX

and EA (base case). The

biorefinery was also modeled based on NREL economic model for all unit operations

with the exception of pretreatment. The scenario evaluated for AFEX
system, with a delivered cost of pretreated AFEX

TM

TM

was based on the depot

-CS pellets to the biorefinery of $120/ton on

dry weight basis. Enzyme cost was assumed to be the same as for the EA biorefinery. The
enzymatic hydrolysis data were generated from CHAPTER 5 and used in this work to project
TM

the AFEX

biorefinery MESP. The AFEX

TM

–based biorefinery was evaluated considering

internal power generation by combustion of the unhydrolysed solids. Excess electricity
production was assumed to be revenue to the biorefinery. For the EA biorefinery, lignin was
assumed to be a valuable raw material with good potential for future valorization, according to

243

the physico-chemical characterization presented in CHAPTER 6. Therefore, lignin was assumed
to be a source of revenue to offset the high energy demand of EA pretreatment.

Biorefinery Economy Based on Different Pretreatments

EA ( Base Base)

AFEX (MBI)

1.00

1.20

1.40

1.60

1.80

2.00

2.20

2.40

2.60

MESP ($/gallon EtOH)
Figure 8-3 MESP values for biorefineries based on different pretreatment technologies.

From Figure 8-3, it is possible to conclude that the base case EA biorefinery shows a higher
TM

MESP compared to an AFEX

biorefinery, with a value of $2.36 per gallon of ethanol

compared with $2.28. However, the base case scenario represents the starting point for the
development of this technology. Besides being important to evolve on developing technologies
to better separate lignin from plant biomass, generating important revenue to the sustainability of
an EA biorefinery, is also imperative to find processing conditions that minimize energy input to
the EA pretreatment. From Figure 8-4, is clear that heat usage in EA pretreatment is the factor
that most affects the economy of the biorefinery, followed by biomass cost. As mentioned above
244

and observed in Figure 8-1, the distillation unit used for separating ammonia and water is the
most energy demanding operation in the EA pretreatment process.The temperature requirement
for the distillation is 290 °C for the reboiler and 110 °C for the condenser. It is highly probable
that the 6:1 ammonia-to-biomass ratio condition used here as a base case is not the most
economical for the EA biorefinery. Analyzing the sensitivity chart in Figure 8-4, it is clear that
reduction in the ammonia loading and heat duty by 30% could have significant impact on the
MESP of EA biorefinery. Besides heat integration and process optimization, this is possible to
achieve by reducing the ammonia loading between 3:1 and 4:1, or possibly by directly reusing
the liquid ammonia with the additional water for one more pretreatment cycle before separating
the ammonia from the water by distillation. The former strategy would not have significant
impact on the enzymatic hydrolysis yields but would impact lignin extraction, unless the
pretreatment time is extended to 60 min (Figure 5-2 iv). This extension of residence time would
practically duplicate the capital investment of EA pretreatment, which does not significantly
impact the projected MESP from the sensitivity chart. Therefore, doing pretreatment at 4:1
ammonia-to-biomass ratio would possibly benefit the economics of the EA biorefinery. In the
case of liquid ammonia reuse before distillation, the major downfall is that it will cause some
temporary water build up in the system, which may have some negative impact on the sugar
yields, although it should not be very significant by analyzing the results shown in Figure 5-2 i).
However, the impact in lignin extraction should be determined experimentally using this method.

245

Figure 8-4 Sensitivity analysis chart for measuring the impact of variations in utility and capital
cost of EA pretreatment on the MESP. The impact of enzyme cost and lignin selling price on the
MESP are also presented.

Another important aspect to note about the EA technology is that it reduced the impact of
enzyme cost on the overall biorefinery economics, as shown in the sensitivity analysis chart. This
effect is a direct consequence of the reduced enzyme loading (7.5mg/g glucan) during efficient
enzymatic hydrolysis of EA-CS. Even though we could achieve this reduction on enzyme
loading effect on the economics of the biorefinery, the EA process is highly dependent of the
lignin value generated as a by-product from the biorefinery. Therefore, creating value addition to
EA-derived lignin is a major area that will determine the success of this technology. Moving
forward on the development and improvement of the EA biorefinery, it is critical to expand our
research on developing technologies for converting EA lignin to value added products.
246

A final consideration is the operating pressure of EA pretreatment in the base case scenario
(1200 psi). For industrial processes, high pressure has significant impact on capital cost, but also
on process safety and social acceptance of the industrial activity. Therefore, decreasing operating
pressure is also an important area that deserves attention moving forward during development of
the EA technology. A possible strategy is to use co-solvents of higher boiling point that can
interact with ammonia and further reduce operating pressure at higher temperatures.
Preferentially, solvents that can be produced within the biorefinery, such as ethanol, which
present good properties for selectively dissolving lignin and that do not impact cellulose III
conversion (CHAPTER 1). The impact of these solvent systems on EA pretreatment
effectiveness and process economics should be evaluated in future research.
8.5. Conclusion
EA pretreatment process optimization was performed in this study and compared with a leading
TM

ammonia pretreatment technology, i.e., AFEX

. From optimization studies, it was possible to

observe that EA pretreatment minimizes enzyme utilization (7.5 mg/g glucan) during enzymatic
hydrolysis at high solids loading (21.44 % solids), while being able to achieve about 72.5 % and
69.8 % glucose and xylose conversions, respectively. Oligomeric glucan and xylan predicted for
these conditions were 8.2 % and 25.5 %, respectively. These projected results allow the
biorefinery to produce an ethanol titer of 6 wt% and a process yield of 242.20 Kg ethanol/ton
untreated biomass, assuming 90 % of theoretical maximum ethanol yield during fermentation.
Under these conditions, the predicted MESP was $2.36 per gallon of ethanol ($3.59/gal gasoline
equivalent), assuming an IRR of 10%. This value was obtained by assuming that the lignin
produced during EA is a source of revenue to the biorefinery (10 cents/lb lignin) and electricity
247

was externally purchased. Using these base case assumptions, the EA economic performance is
TM

slightly worse than AFEX

($2.28/gal ethanol). However, this technology is still taing its first

steps and offers great room for improvement from the base case scenario, as pretreatment
conditions were not optimized for producing the lowest possible MESP.
From the sensitivity analysis results, the heat consumption during EA pretreatment has the
highest impact on the overall economics of the biorefinery. Also, most of the heat consumption
is associated with distillation activities during ammonia recycling. Therefore, it is important to
focus attention on the development of less heat intensive conditions for EA, possibly by better
heat integration and process optimization, as well as by decreasing ammonia loading or the
number of cycles of distillation. Capital cost and enzyme cost seem to have significantly lower
impact on the economics of the EA biorefinery, and therefore, the ~M$52 capital cost predicted
for the EA unit operation did not affect the MESP to a large extent. In the other hand, the lignin
price had a great impact on the MESP and can significantly affect the economies of the EA
biorefinery. Therefore, it is important to create a large market for lignin in the future in order to
make this process economically viable. From this perspective, future research on EA technology
should also be focused on creating value addition to EA lignin streams and provide diverse
market opportunities for lignin in the future.
Future developments should also include the decrease of operating pressure during EA
pretreatment. Although pretreatment capital costs do not significantly affect the economics of the
biorefinery, handling substantial amounts of ammonia at 1200 psi can constitute a liability issue
and therefore, social acceptance of the process may be jeopardized. This goal can be achieved by
various strategies that include the utilization of co-solvents along with ammonia that could
248

potentially improve lignin extraction without compromising cellulose III conversion. For
example, ethanol is used for efficient extraction of lignin in other processes and its usage along
with ammonia would decrease the operating pressure of EA pretreatment, as well as ammonia
usage. In CHAPTER 3 we showed that it is possible to convert cellulose I to cellulose III while
operating at a under 2.4:1 ammonia to cellulose ratio if we use a solution of 60 wt% ammonia in
ethanol. This could potentially decrease the pressure below 800 psi. Other solvents of higher
boiling point could be considered in the future research. However, is important to focus attention
on inexpensive solvents, preferably produced within the biorefinery system.

249

APPENDIX

250

Ammonia
6

Ammonia
Reservoir

10

Compressor

PUMP
2

11
EA 1

EA 2
2a

2b

3a

3b

1a

1b
5

Untreated
Biomass

3

Untreated
Biomass

Flash Evaporator

7a

7b

EA
Biomass

4

9

8
Distillation

Extractives

7

Ammonia
Recycling
Figure 8-5 EA pretreatment process flow diagram. EA1 and EA2 work intermittently one after
the other, sharing the equipment used for ammonia recovery (highlighted in blue). This system of
two reactors (EA1 and EA2) is replicated five times to compose the full biorefinery, scaled for
processing 2000 tons/day of untreated corn stover. The ammonia recycling system duplicated to
be shared among the 10 reactors that are synchronized during operation.
251

Time (hr)
Time (min) 10
System 1
System 2
System 3
System 4
System 5

1
20

Load Reactor 1

30

2
40

50

Reactor 1 Pretreatment
Load Reactor 2

Load Reactor 1

60

Ammonia Rec
Ammonia Rec

Reactor 1 Pretreatment
Load Reactor 2

Load Reactor 1

70

3
100

110

130

150

4
160

180

200

210

220

230

240

UL Dryer Load Reactor 1
Reactor 2 Pretreatment

Ammonia Rec
Ammonia Rec

Reactor 1 Pretreatment
UL Dryer Load Reactor 2

Ammonia Rec
Ammonia Rec

190

UL Dryer Load Reactor 1
Reactor 2 Pretreatment

Ammonia Rec
Ammonia Rec

Reactor 1 Pretreatment
UL Dryer Load Reactor 2

Ammonia Rec
Ammonia Rec

UL Dryer Load Reactor 1
Reactor 2 Pretreatment

170

Ammonia Rec
Ammonia Rec

Reactor 1 Pretreatment
UL Dryer Load Reactor 2

Ammonia Rec
Ammonia Rec

UL Dryer Load Reactor 1
Reactor 2 Pretreatment

Ammonia Rec
Ammonia Rec

140

Reactor 1 Pretreatment
UL Dryer Load Reactor 2

Ammonia Rec
Ammonia Rec

UL Dryer Load Reactor 1
Reactor 2 Pretreatment

Ammonia Rec
Ammonia Rec

Reactor 1 Pretreatment
Load Reactor 2

120

Ammonia Rec
Ammonia Rec

UL Dryer Load Reactor 1
Reactor 2 Pretreatment

Ammonia Rec
Ammonia Rec

Reactor 1 Pretreatment
Load Reactor 2

Load Reactor 1

90

UL Dryer Load Reactor 1
Reactor 2 Pretreatment

Ammonia Rec
Ammonia Rec

Reactor 1 Pretreatment
Load Reactor 2

Load Reactor 1

80

UL Dryer Load Reactor 1
Reactor 2 Pretreatment

Ammonia Rec
Ammonia Rec

Reactor 1 Pretreatment
UL Dryer Load Reactor 2

UL Dryer Load Reactor 1
Reactor 2 Pretreatment

Ammonia Rec
Ammonia Rec

UL Dryer Load Reactor 1
Reactor 2 Pretreatment

Figure 8-6 Schedule for EA pretreatment operation in the biorefinery. The pretreatment unit is composed by five systems of two
reactors that load and unload (UL) intermittently. While pretreatment reactors requires the utilization of five systems of two reactors,
ammonia recycling can be operated with two systems, each of them composed by a distillation column, an ammonia evaporator, a
compressor and a pump.

252

Table 8-4 Stream table for EA pretreatment process.
Stream Temperature (°C) Temperature (K) Pressure (Psi)
1a/1b
2/2a/2b
3a/3b/3
4
5
6
7a/7b/7
8
9
10
11

25.00
99.60
25.00
-33.40
-33.40
195.70
120.00
278.00
105.10
99.20
25.00

298.15
372.75
298.15
239.75
239.75
468.85
393.15
551.15
378.25
372.35
298.15

14.70
1200.00
150.00
14.70
14.70
150.00
1200.00
200.00
200.00
1000.00
150.00

Flow Rate (kg/min)
Biomass Ammonia Water
1388.89
0.00
78.76
0.00
8333.40
60.12
1166.67
1679.54
0.27
1166.67
41.67
0.27
0.00
839.77
0.00
0.00
839.77
0.00
222.22
7493.56 138.614
222.22
10.35
78.492
0.00
7483.21
60.122
0.00
8333.40
60.122
0.00
850.19
0.00

Table 8-5 Total energy duty of EA pretreatment.
Heat duty
Electricity
Heat duty
Name
(GJ/h)
(kW)
(GJ/Ton BM)
Reactor
155.76
1.87
Distillation Column
553.84
6.65
Evaporator
46.97
0.56
Compressor
6901.04
Pump
565.08
Ammonia Tank
Total Energy Spending
9.08
253

H (GJ/hr)

Phase

-47.04
-1627.30
-373.74
-0.66
-210.83
-117.17
-1297.83
-43.28
-1145.10
-1629.56
-196.97

Solid
Liquid
Solid/Vapor
Solid/Liquid
Vapor
Liquid
Liquid
Liquid
Vapor
Liquid
Liquid

Electricity
(kWh/Ton BM)
82.81
6.78
89.59

Table 8-6 EA reactor specifications and cost.
Reactor
Volume (gallon)
31,530.68
Material of construction
Stainless steel
Design temperature (°C)
147.78
Design pressure (psig)
1318.05
Heat duty (GJ/h)
155.76
1,744,710.00
Unit cost ($)
10.00
Number of systems
17,447,100.00
Total cost ($)

Table 8-7 Ammonia storage tank specifications and cost.
Ammonia tank
Volume (gallon)
82,187
Material of construction
Stainless steel
Design temperature (°C)
70
Design pressure (psig)
160.30
320,790.00
Unit cost ($)
3
Number of systems
962,370
Total cost ($)

Table 8-8 Ammonia evaporation column specifications and cost for ammonia recycling
system.
Evaporation column
Volume (gallon)
94,591.98
Material of construction
Stainless steel
Design temperature (°C)
70
Design pressure (psig)
160.30
Heat duty (GJ/h)
23.49
360,910.00
Unit cost ($)
2
Number of systems
721,820.00
Total cost ($)

254

Table 8-9 Distillation equipment specifications and cost for ammonia recycling system of
EA pretreatment.
Distillation
Tower
Column diameter (ft)
14
Number of stages
13
Feed stage
3
Distillate to feed ratio
1
Reflux ratio
1
Tray type
Sieve
Tray spacing (ft)
2
Design Pressure (psig)
1,035.30
Design Temperature (°C)
306.00
Condenser
Heat transfer area (sqft)
4,304.79
Tube design temperature
132.87
(°C)
Operating Temperature
110.86
(°C)
Tube design pressure (psig)
1,035.30
Reflux pump efficiency
0.70
Heat duty (GJ/h)
-161.45
Reboiler
Heat transfer area (sqft)
31,917.26
Tube design temperature
343.33
(°C)
Operating Temperature
290.03
(°C)
Tube design pressure (psig)
1,035.00
Shell design temperature
343.00
(°C)
Shell design pressure (psig)
685.30
Heat duty (GJ/h)
114.82
6,394,992.00
Unit cost ($)
2.00
Number of systems
12,789,984.00
Total cost ($)

255

Table 8-10 Pump specifications and cost for ammonia recycling system of EA
pretreatment.
Pump
Pump type
centrifugal
Flow rate (gpm)
2,643.66
Fluid head (ft)
1,008.73
Design pressure (psig)
1,245.30
Design temperature (°C)
127.38
Pump efficiency
0.74
Power (kW)
282.54
354,775.00
Unit cost ($)
2.00
Number of systems
709,550.00
Total cost ($)

Table 8-11 Compressor specifications and cost for ammonia recycling system of EA
pretreatment.
Compressor
Actual gas flow rate (CFM)
14,691.30
Design pressure inlet (psig)
0.30
Design pressure outlet (psig)
160.30
Compressibility factor inlet
0.99
Compressibility factor outlet
0.98
Driver power (kW)
2,998.63
3,649,000.00
Unit cost ($)
2.00
Number of systems
7,298,000.00
Total cost ($)

256

Table 8-12 Utility costs and pretreatment capital cost assumed for economic model of
a
biorefinery
Pretreatment installed cost
51.907 Million $
Thermal energy use
9.08 GJ/ton biomass
Electricity use
89.59 kWh/ton biomass
Ammonia use
0.0375 kg/kg biomass
Ammonia price
600 $/ton
Biomass price
60 $/ton
Electricity price
5.72 Cents/kWh
Lignin price
10 Cents/lb
Cellulase price
3600 $/ton
Hemicellulase price
4500 $/ton
a

Other variable costs were assumed to be as defined by the economic model from NREL [206].

Residual Plots for MESP
Normal Probability Plot

Versus Fits

99
0.1

Residual

Percent

90
50
10
1
-0.2

-0.1

0.0
Residual

0.0
-0.1
-0.2

0.1

2.4

2.5
2.6
Fitted Value

Histogram

Versus Order
0.1

7.5

Residual

Frequency

10.0

5.0
2.5
0.0

2.7

-0.15 -0.10 -0.05 0.00 0.05
Residual

0.10

0.0
-0.1
-0.2

0.15

2

4

6

8 10 12 14 16 18 20 22 24 26 28 30

Observation Order

Figure 8-7 Residual plots and histograms for MESP prediction as a function of enzymatic
hydrolysis conditions (enzyme loading, solid loading and incubation time) for base case EA
pretreated corn stover.

257

Table 8-13 Regression coefficients for MESP prediction as a function of enzymatic
hydrolysis conditions for base case EA pretreated corn stover.
Term
Coef
SE Coef
T
P
Constant
3.60571
0.382689
9.422
0
Time
-0.00373
0.0015
-2.485
0.021
Solid Loading
-0.09476
0.037595
-2.521
0.019
Enzyme Loading
-0.00788
0.00708
-1.113
0.277
Time*Time
0.00002
0.00001
1.94
0.065
Solid Loading*Solid Loading
0.00218
0.000937
2.325
0.029
Enzyme Loading*Enzyme Loading
0.00048
0.000185
2.618
0.015
S =0.063623
R-Sq = 77.63%

PRESS = 0.161458
R-Sq(pred) = 61.20%

258

R-Sq(adj) = 71.79%

Figure 8-8 Optimization plots for defining enzymatic hydrolysis conditions that minimize MESP
($/gallon ethanol). % sugar conversions are also presented in the left side on the plots in blue
numbering. These values were calculated for base case EA pretreated corn stover.

259

REFERENCES

260

REFERENCES

1.

Sousa, L.D., et al., 'Cradle-to-grave' assessment of existing lignocellulose pretreatment
technologies. Current Opinion in Biotechnology, 2009. 20(3): p. 339-347.

2.

Roberts, K.M., et al., The effects of water interactions in cellulose suspensions on mass
transfer and saccharification efficiency at high solids loadings. Cellulose, 2011. 18(3): p.
759-773.

3.

Kristensen, J.B., C. Felby, and H. Jorgensen, Yield-determining factors in high-solids
enzymatic hydrolysis of lignocellulose. Biotechnology for Biofuels, 2009. 2.

4.

Jin, M., et al., A novel integrated biological process for cellulosic ethanol production
featuring high ethanol productivity, enzyme recycling and yeast cells reuse. Energy &
Environmental Science, 2012. 5(5): p. 7168-7175.

5.

Monti, A., N. Di Virgilio, and G. Venturi, Mineral composition and ash content of six
major energy crops. Biomass and Bioenergy, 2008. 32(3): p. 216-223.

6.

Pettersen Roger, C., The Chemical Composition of Wood, in The Chemistry of Solid
Wood. 1984, American Chemical Society. p. 57-126.

7.

Sun, Y. and J. Cheng, Hydrolysis of lignocellulosic materials for ethanol production: a
review. Bioresource Technology, 2002. 83(1): p. 1-11.

8.

G. W. Huber, B.E.D., Grassoline at the Pump. Scientific American Magazine, 2009.
301(1): p. 8.

9.

Wyman, C.E., et al., Coordinated development of leading biomass pretreatment
technologies. Bioresource Technology, 2005. 96(18): p. 1959-1966.

10.

Cosgrove, D.J., Growth of the plant cell wall. Nat Rev Mol Cell Biol, 2005. 6(11): p.
850-861.

11.

Fengel D, W.W., Wood - Chemistry, Ultrastructure, Reactions. 1989, Berlin, Germany:
Walter de Gruyter. 613.

12.

Wada, M., T. Okano, and J. Sugiyama, Allomorphs of native crystalline cellulose I
evaluated by two equatoriald-spacings. Journal of Wood Science, 2001. 47(2): p. 124128.

13.

Goro Honjo, M.W., Examination of cellulose fiber by the low-temperature specimen
method of electron diffraction and electron microscopy Nature 1958. 181: p. 326.
261

14.

Jeffries, T., Biodegradation of lignin and hemicelluloses, in Biochemistry of microbial
degradation, C. Ratledge, Editor. 1994, Springer Netherlands. p. 233-277.

15.

Chundawat, S.P.S., Ultrastructural and Physicochemical Modifications Within Ammonia
Treated Lignocellulosic Cell Walls and Their Influence on Enzymatic Digestibility. 2009:
UMI.

16.

Laine, C., Structures of Hemicelluloses and Pectins in wood and Pulp, in KCL
communications 2005, Helsinki University of Technology.

17.

Sederoff, R.R., et al., Unexpected variation in lignin. Current Opinion in Plant Biology,
1999. 2(2): p. 145-152.

18.

Besle, J.-M., A. Cornu, and J.-P. Jouany, Roles of structural phenylpropanoids in forage
cell wall digestion. Journal of the Science of Food and Agriculture, 1994. 64(2): p. 171190.

19.

Himmel, M., et al., Biomass recalcitrance: Engineering plants and enzymes for biofuels
production. Science, 2007. 315: p. 804 - 807.

20.

Aden, A., et al., National renewable energy laboratory. Golden, Colorado, 2002.

21.

Wooley, R., et al., Process Design and Costing of Bioethanol Technology: A Tool for
Determining the Status and Direction of Research and Development. Biotechnology
Progress, 1999. 15(5): p. 794-803.

22.

Gao, D., et al., Mixture optimization of six core glycosyl hydrolases for maximizing
saccharification of ammonia fiber expansion (AFEX) pretreated corn stover. Bioresource
Technology, 2010. 101(8): p. 2770-2781.

23.

Chundawat, S.P., Ultrastructural and physicochemical modifications within ammonia
treated lignocellulosic cell walls and their influence on enzymatic digestibility, in
Chemical Engineering and Materials Science. 2009, Michigan State University: East
Lansing.

24.

Wu, Z. and Y.Y. Lee, Inhibition of the enzymatic hydrolysis of cellulose by ethanol.
Biotechnology Letters, 1997. 19(10): p. 977-979.

25.

Zyl, W.H.v., R.d. Haan, and D.C.l. Grange, Developing Organisms for Consolidated
Bioprocessing of Biomass to Ethanol, in Biofuel Production-Recent Developments and
Prospects, D.M.A.D.S. Bernardes, Editor. 2011, InTech.

26.

Zakzeski, J., et al., The Catalytic Valorization of Lignin for the Production of Renewable
Chemicals. Chemical Reviews, 2010. 110(6): p. 3552-3599.
262

27.

Lora, J.H. and W.G. Glasser, Recent industrial applications of lignin: A sustainable
alternative to nonrenewable materials. Journal of Polymers and the Environment, 2002.
10(1-2): p. 39-48.

28.

Kadla, J.F., et al., Lignin-based carbon fibers for composite fiber applications. Carbon,
2002. 40(15): p. 2913-2920.

29.

Nishikawa, N., R. Sutcliffe, and J. Saddler, The influence of lignin degradation products
on xylose fermentation by Klebsiella pneumoniae. Applied Microbiology and
Biotechnology, 1988. 27(5-6): p. 549-552.

30.

Pan, X., Role of functional groups in lignin inhibition of enzymatic hydrolysis of cellulose
to glucose. J Biobased Mater Bioenergy, 2008. 2: p. 25 - 32.

31.

Weiss, N., et al., Enzymatic lignocellulose hydrolysis: Improved cellulase productivity by
insoluble solids recycling. Biotechnology for Biofuels C7 - 5, 2013. 6(1): p. 1-14.

32.

Lee, D., A.H.C. Yu, and J.N. Saddler, Evaluation of cellulase recycling strategies for the
hydrolysis of lignocellulosic substrates. Biotechnology and Bioengineering, 1995. 45(4):
p. 328-336.

33.

Cantu, D., et al., Strangers in the matrix: plant cell walls and pathogen susceptibility.
Trends in plant science, 2008. 13(11): p. 610-617.

34.

Ishizawa, C.I., et al., Porosity and Its Effect on the Digestibility of Dilute Sulfuric Acid
Pretreated Corn Stover. Journal of Agricultural and Food Chemistry, 2007. 55(7): p.
2575-2581.

35.

Zadrazil, F. and A.K. Puniya, Studies on the effect of particle size on solid-state
fermentation of sugarcane bagasse into animal feed using white-rot fungi. Bioresource
Technology, 1995. 54(1): p. 85-87.

36.

Fan, L.T., Y.H. Lee, and D.R. Beardmore, The influence of major structural features of
cellulose on rate of enzymatic hydrolysis. Biotechnology and Bioengineering, 1981.
23(2): p. 419-424.

37.

Chang, V. and M. Holtzapple, Fundamental factors affecting biomass enzymatic
reactivity. Applied Biochemistry and Biotechnology, 2000. 84-86(1): p. 5-37.

38.

Laureano-Perez, L., et al., Understanding Factors that Limit Enzymatic Hydrolysis of
Biomass. , in Twenty-Sixth Symposium on Biotechnology for Fuels and Chemicals. 2005,
Humana Press. p. 1081-1099.

263

39.

Sidiras, D.K. and E.G. Koukios, Acid saccharification of ball-milled straw. Biomass,
1989. 19(4): p. 289-306.

40.

Tassinari, T., et al., Energy requirements and process design considerations in
compression-milling pretreatment of cellulosic wastes for enzymatic hydrolysis.
Biotechnology and Bioengineering, 1980. 22(8): p. 1689-1705.

41.

Alvo, P. and K. Belkacemi, Enzymatic saccharification of milled timothy (Phleum
pratense L.) and alfalfa (Medicago sativa L.). Bioresource Technology, 1997. 61(3): p.
185-198.

42.

Heinze, T. and A. Koschella, Solvents applied in the field of cellulose chemistry: a mini
review. Polímeros, 2005. 15: p. 84-90.

43.

Pan, X., et al., Bioconversion of hybrid poplar to ethanol and co-products using an
organosolv fractionation process: Optimization of process yields. Biotechnology and
Bioengineering, 2006. 94(5): p. 851-861.

44.

Arato, C., E.K. Pye, and G. Gjennestad, The Lignol Approach to Biorefining of Woody
Biomass to Produce Ethanol and Chemicals, in Twenty-Sixth Symposium on
Biotechnology for Fuels and Chemicals, B. Davison, et al., Editors. 2005, Humana Press.
p. 871-882.

45.

Zhang, Y.-H.P., et al., Fractionating recalcitrant lignocellulose at modest reaction
conditions. Biotechnology and Bioengineering, 2007. 97(2): p. 214-223.

46.

Moxley, G., Z. Zhu, and Y.H.P. Zhang, Efficient Sugar Release by the Cellulose SolventBased Lignocellulose Fractionation Technology and Enzymatic Cellulose Hydrolysis.
Journal of Agricultural and Food Chemistry, 2008. 56(17): p. 7885-7890.

47.

Swatloski, R.P., et al., Dissolution of Cellose with Ionic Liquids. Journal of the American
Chemical Society, 2002. 124(18): p. 4974-4975.

48.

Turner, M.B., et al., Ionic liquid salt-induced inactivation and unfolding of cellulase from
Trichoderma reesei. Green Chemistry, 2003. 5(4): p. 443-447.

49.

Joglekar, H.G., I. Rahman, and B.D. Kulkarni, The Path Ahead for Ionic Liquids.
Chemical Engineering & Technology, 2007. 30(7): p. 819-828.

50.

Lee, S.H., et al., Ionic liquid-mediated selective extraction of lignin from wood leading to
enhanced enzymatic cellulose hydrolysis. Biotechnology and Bioengineering, 2009.
102(5): p. 1368-1376.

51.

Weil, J., et al., Pretreatment of corn fiber by pressure cooking in water. Applied
Biochemistry and Biotechnology, 1998. 73(1): p. 1-17.
264

52.

Weil, J.R., et al., Removal of Fermentation Inhibitors Formed during Pretreatment of
Biomass by Polymeric Adsorbents. Industrial & Engineering Chemistry Research, 2002.
41(24): p. 6132-6138.

53.

Klinke, H.B., A.B. Thomsen, and B.K. Ahring, Inhibition of ethanol-producing yeast and
bacteria by degradation products produced during pre-treatment of biomass. Applied
Microbiology and Biotechnology, 2004. 66(1): p. 10-26.

54.

Lloyd, T.A. and C.E. Wyman, Combined sugar yields for dilute sulfuric acid
pretreatment of corn stover followed by enzymatic hydrolysis of the remaining solids.
Bioresource Technology, 2005. 96(18): p. 1967-1977.

55.

Kim, J., Y.Y. Lee, and R. Torget, Cellulose hydrolysis under extremely low sulfuric acid
and high-temperature conditions. Applied Biochemistry and Biotechnology, 2001. 9193(1-9): p. 331-340.

56.

Buhner, J. and F.A. Agblevor, Effect of detoxification of dilute-acid corn fiber
hydrolysate on xylitol production. Applied Biochemistry and Biotechnology, 2004.
119(1): p. 13-30.

57.

Chen, H., et al., New process of maize stalk amination treatment by steam explosion.
Biomass and Bioenergy, 2005. 28(4): p. 411-417.

58.

Playne, M.J., Increased digestibility of bagasses by pretreatment with alkalis and steam
explosion. Biotechnology and Bioengineering, 1984. 26(5): p. 426-433.

59.

Ewanick, S.M., R. Bura, and J.N. Saddler, Acid-catalyzed steam pretreatment of
lodgepole pine and subsequent enzymatic hydrolysis and fermentation to ethanol.
Biotechnology and Bioengineering, 2007. 98(4): p. 737-746.

60.

Brownell, H.H., E.K.C. Yu, and J.N. Saddler, Steam-explosion pretreatment of wood:
Effect of chip size, acid, moisture content and pressure drop. Biotechnology and
Bioengineering, 1986. 28(6): p. 792-801.

61.

Wyman, C., Handbook on bioethanol: production and utilization. 1996: Taylor & francis.

62.

Mosier, N., et al., Features of promising technologies for pretreatment of lignocellulosic
biomass. Bioresource Technology, 2005. 96(6): p. 673-686.

63.

Mosier, N., et al., Optimization of pH controlled liquid hot water pretreatment of corn
stover. Bioresource Technology, 2005. 96(18): p. 1986-1993.

265

64.

Mosier, N.S., C.M. Ladisch, and M.R. Ladisch, Characterization of acid catalytic
domains for cellulose hydrolysis and glucose degradation. Biotechnology and
Bioengineering, 2002. 79(6): p. 610-618.

65.

Peter van Walsum, G. and H. Shi, Carbonic acid enhancement of hydrolysis in aqueous
pretreatment of corn stover. Bioresource Technology, 2004. 93(3): p. 217-226.

66.

Luo, C., D.L. Brink, and H.W. Blanch, Identification of potential fermentation inhibitors
in conversion of hybrid poplar hydrolyzate to ethanol. Biomass and Bioenergy, 2002.
22(2): p. 125-138.

67.

Öhgren, K., M. Galbe, and G. Zacchi, Optimization of Steam Pretreatment of SO2Impregnated Corn Stover for Fuel Ethanol Production, in Twenty-Sixth Symposium on
Biotechnology for Fuels and Chemicals, B. Davison, et al., Editors. 2005, Humana Press.
p. 1055-1067.

68.

Wyman, C.E., et al., Comparative sugar recovery data from laboratory scale application
of leading pretreatment technologies to corn stover. Bioresource Technology, 2005.
96(18): p. 2026-2032.

69.

Wyman C, K.R., Dale B, Elander R, Holtzapple M, Ladisch M, Lee Y, Moniruzzaman
M, Saddler J, Comparative data for enzymatic digestion of corn stover and poplar wood
after pretreatment by leading technologies., in AIChE Annual Meeting. 2006: San
Francisco, California.

70.

Öhgren, K., et al., Effect of hemicellulose and lignin removal on enzymatic hydrolysis of
steam pretreated corn stover. Bioresource Technology, 2007. 98(13): p. 2503-2510.

71.

Kim, T.H. and Y.Y. Lee, Pretreatment and fractionation of corn stover by ammonia
recycle percolation process. Bioresource Technology, 2005. 96(18): p. 2007-2013.

72.

Laureano-Perez, L., Spectroscopic and chemical characterization of biomass. 2005,
Michigan State University: East Lansing.

73.

Kim, S. and M.T. Holtzapple, Lime pretreatment and enzymatic hydrolysis of corn stover.
Bioresource Technology, 2005. 96(18): p. 1994-2006.

74.

Kim, S. and M.T. Holtzapple, Delignification kinetics of corn stover in lime pretreatment.
Bioresource Technology, 2006. 97(5): p. 778-785.

75.

Chang, V. and M. Holtzapple, Fundamental factors affecting biomass enzymatic
reactivity. Applied Biochemistry and Biotechnology, 2000. 84-86(1-9): p. 5-37.

76.

Laureano-Perez, L., et al., Understanding factors that limit enzymatic hydrolysis of
biomass. Applied Biochemistry and Biotechnology, 2005. 124(1-3): p. 1081-1099.
266

77.

Sendich, E., et al., Recent process improvements for the ammonia fiber expansion
(AFEX) process and resulting reductions in minimum ethanol selling price. Bioresource
Technology, 2008. 99(17): p. 8429-8435.

78.

Lau, M.W. and B.E. Dale, Cellulosic ethanol production from AFEX-treated corn stover
using Saccharomyces cerevisiae 424A(LNH-ST). Proceedings of the National Academy
of Sciences, 2009. 106(5): p. 1368-1373.

79.

Lau, M.W., B.E. Dale, and V. Balan, Ethanolic fermentation of hydrolysates from
ammonia fiber expansion (AFEX) treated corn stover and distillers grain without
detoxification and external nutrient supplementation. Biotechnology and Bioengineering,
2008. 99(3): p. 529-539.

80.

Lewin, M. and L.G. Roldan, The effect of liquid anhydrous ammonia in the structure and
morphology of cotton cellulose. Journal of Polymer Science Part C: Polymer Symposia,
1971. 36(1): p. 213-229.

81.

Weimer, P.J., et al., Effect of supercritical ammonia on the physical and chemical
structure of ground wood. 1986. Medium: X; Size: Pages: 5-18.

82.

Kim, S. and Y. Lee, Fractionation of herbaceous biomass by ammonia-hydrogen
peroxide percolation treatment. Applied Biochemistry and Biotechnology, 1996. 5758(1): p. 147-156.

83.

García-Cubero, M.T., et al., Effect of ozonolysis pretreatment on enzymatic digestibility
of wheat and rye straw. Bioresource Technology, 2009. 100(4): p. 1608-1613.

84.

Martin, C. and A.B. Thomsen, Wet oxidation pretreatment of lignocellulosic residues of
sugarcane, rice, cassava and peanuts for ethanol production. J. Chem. Technol.
Biotechnol., 2007. 82: p. 174-181.

85.

Gould, J.M., Alkaline peroxide delignification of agricultural residues to enhance
enzymatic saccharification. Biotechnology and Bioengineering, 1984. 26(1): p. 46-52.

86.

Lee, J.-W., et al., Evaluation of waste mushroom logs as a potential biomass resource for
the production of bioethanol. Bioresource Technology, 2008. 99(8): p. 2736-2741.

87.

Christian, V., et al., Degradation of xenobiotic compounds by lignin-degrading white-rot
fungi: enzymology and mechanisms involved. Indian journal of experimental biology,
2005. 43(4): p. 301.

88.

Balan, V., et al., Mushroom spent straw: a potential substrate for an ethanol-based
biorefinery. Journal of Industrial Microbiology & Biotechnology, 2008. 35(5): p. 293301.
267

89.

Searchinger, T., et al., Use of U.S. Croplands for Biofuels Increases Greenhouse Gases
Through Emissions from Land-Use Change. Science, 2008. 319(5867): p. 1238-1240.

90.

Yang, B. and C.E. Wyman, Pretreatment: the key to unlocking low-cost cellulosic
ethanol. Biofuels, Bioproducts and Biorefining, 2008. 2(1): p. 26-40.

91.

E, C.J., J.S. V, and D.B. E, Technical and Financial Feasibility Analysis of Distributed
Bioprocessing Using Regional Biomass Pre-Processing Centers. Journal of Agricultural
& Food Industrial Organization, 2007. 5(2): p. 1-29.

92.

Weimer, P.J., et al., FIBEX-treated rice straw as a feed ingredient for lactating dairy
cows. Animal Feed Science and Technology, 2003. 103(1–4): p. 41-50.

93.

Fengel, D. and G. Wengener, Wood. Chemistry, Ultrastructure, reactions. Walter de
Gruyter, Berlin, Germany, 1989: p. 613.

94.

Sjöström, E., Tappi 60, 1977. 9: p. 151-154.

95.

Chundawat, S.P.S., et al., Multifaceted characterization of cell wall decomposition
products formed during ammonia fiber expansion (AFEX) and dilute acid based
pretreatments. Bioresource Technology, 2010. 101(21): p. 8429-8438.

96.

Buanafina, M.M.D., Feruloylation in Grasses: Current and Future Perspectives.
Molecular Plant, 2009. 2(5): p. 861-872.

97.

Vismeh, R., et al., Profiling of diferulates (plant cell wall cross-linkers) using ultrahighperformance liquid chromatography-tandem mass spectrometry. Analyst, 2013. 138(21):
p. 6683-6692.

98.

French, H.E. and G.G. Wrightsman, Action of Ammonia on Esters. Journal of the
American Chemical Society 1938. 60: p. 50-51.

99.

Gordon, M., J.G. Miller, and A.R. Day, Effect of Structure on Reactivity.1 II. Influence of
Solvents on Ammonolysis of Esters. Journal of the American Chemical Society, 1949.
71(4): p. 1245-1250.

100.

Williamson, P.M., R.C. Anderson, and G.W. Watt, Electrolyte Catalysis in the
Ammonolysis of 9-Phenyl-9-chlorofluorene by Liquid Ammonia1. Journal of the
American Chemical Society, 1944. 66(3): p. 376-379.

101.

Fellinger, L.L. and L.F. Audrieth, Acid Catalysis in Liquid Ammonia. II. Ammonolysis of
Ethyl Benzoate. Journal of the American Chemical Society, 1938. 60(3): p. 579-581.

268

102.

Shibamoto, T., et al., A study of pyrazine formation. Journal of Agricultural and Food
Chemistry, 1979. 27(5): p. 1027-1031.

103.

Wada, M., et al., Cellulose IIII Crystal Structure and Hydrogen Bonding by Synchrotron
X-ray and Neutron Fiber Diffraction. Macromolecules, 2004. 37(23): p. 8548-8555.

104.

Wada, M., Y. Nishiyama, and P. Langan, X-ray Structure of Ammonia-Cellulose I: New
Insights into the Conversion of Cellulose I to Cellulose IIII. Macromolecules, 2006.
39(8): p. 2947-2952.

105.

Chundawat, S.P.S., et al., Restructuring the Crystalline Cellulose Hydrogen Bond
Network Enhances Its Depolymerization Rate. Journal of the American Chemical
Society, 2011. 133(29): p. 11163-11174.

106.

Cheng, G., et al., Transition of Cellulose Crystalline Structure and Surface Morphology
of Biomass as a Function of Ionic Liquid Pretreatment and Its Relation to Enzymatic
Hydrolysis. Biomacromolecules, 2011. 12(4): p. 933-941.

107.

Zhang, Y.H.P., et al., A transition from cellulose swelling to cellulose dissolution by ophosphoric acid: Evidence from enzymatic hydrolysis and supramolecular structure.
Biomacromolecules, 2006. 7(2): p. 644-648.

108.

Humpula, J.F., et al., Probing the nature of AFEX-pretreated corn stover derived
decomposition products that inhibit cellulase activity. Bioresource Technology, 2013.
152(0): p. 38-45.

109.

Jin, M., et al., Quantitatively understanding reduced xylose fermentation performance in
AFEXTM treated corn stover hydrolysate using Saccharomyces cerevisiae 424A (LNH-ST)
and Escherichia coli KO11. Bioresource Technology, 2012(available online).

110.

Shao, X., et al., Conversion for Avicel and AFEX pretreated corn stover by Clostridium
thermocellum and simultaneous saccharification and fermentation: Insights into
microbial conversion of pretreated cellulosic biomass. Bioresource Technology, 2011.
102(17): p. 8040-8045.

111.

Mooney, C.A., et al., The effect of initial pore volume and lignin content on the enzymatic
hydrolysis of softwoods. Bioresource Technology, 1998. 64(2): p. 113-119.

112.

Palonen, H., et al., Adsorption of Trichoderma reesei CBH I and EG II and their catalytic
domains on steam pretreated softwood and isolated lignin. Journal of Biotechnology,
2004. 107(1): p. 65-72.

113.

Ximenes, E., et al., Deactivation of cellulases by phenols. Enzyme and Microbial
Technology, 2011. 48(1): p. 54-60.
269

114.

Hall, M., et al., Cellulose crystallinity – a key predictor of the enzymatic hydrolysis rate.
FEBS Journal, 2010. 277(6): p. 1571-1582.

115.

Nishiyama, Y., et al., Crystal Structure and Hydrogen Bonding System in Cellulose Iα
from Synchrotron X-ray and Neutron Fiber Diffraction. Journal of the American
Chemical Society, 2003. 125(47): p. 14300-14306.

116.

Finkenstadt, V.L. and R.P. Millane, Crystal Structure of Valonia Cellulose I.
Macromolecules, 1998. 31(22): p. 7776-7783.

117.

Igarashi, K., M. Wada, and M. Samejima, Activation of crystalline cellulose to
cellulose IIII results in efficient hydrolysis by cellobiohydrolase. FEBS Journal, 2007.
274(7): p. 1785-1792.

118.

Gao, D., et al., Increased enzyme binding to substrate is not necessary for more efficient
cellulose hydrolysis. Proceedings of the National Academy of Sciences, 2013.

119.

Schleicher, H., C. Daniels, and B. Philipp, Changes of cellulose accessibility to reactions
in alkaline medium by activation with ammonia. Journal of Polymer Science: Polymer
Symposia, 1974. 47(1): p. 251-260.

120.

Barry, A.J., F.C. Peterson, and A.J. King, x-Ray Studies of Reactions of Cellulose in NonAqueous Systems. I. Interaction of Cellulose and Liquid Ammonia1. Journal of the
American Chemical Society, 1936. 58(2): p. 333-337.

121.

Davis, W.E., et al., X-Ray Studies of Reactions of Cellulose in Non-Aqueous Systems. II.
Interaction of Cellulose and Primary Amines1. Journal of the American Chemical
Society, 1943. 65(7): p. 1294-1299.

122.

Clark, G.L. and E.A. Parker, An X-ray Diffraction Study of the Action of Liquid Ammonia
on Cellulose and Its Derivatives. The Journal of Physical Chemistry, 1937. 41(6): p. 777786.

123.

Mittal, A., et al., Effects of alkaline or liquid-ammonia treatment on crystalline cellulose:
changes in crystalline structure and effects on enzymatic digestibility. Biotechnology for
Biofuels, 2011. 4(1): p. 41.

124.

Zhang, J., et al., Dissolution of Microcrystalline Cellulose in Phosphoric Acid-Molecular
Changes and Kinetics. Molecules, 2009. 14(12): p. 5027-5041.

125.

Ruland, W., X-ray determination of crystallinity and diffuse disorder scattering. Acta
Cryst, 1961. 14: p. 1180 - 1185.

126.

Park, S., et al., Cellulose crystallinity index: measurement techniques and their impact on
interpreting cellulase performance. Biotechnology for Biofuels, 2010. 3(1): p. 10.
270

127.

Hult, L., T. Iversen, and J. Sugiyama, Characterization of the supramolecular structure
of cellulose in wood pulp fibres. Cellulose, 2003. 10: p. 103 - 110.

128.

Garvey, C., I. Parker, and G. Simon, On the interpretation of X-ray diffraction powder
patterns in terms of the nanostructure of cellulose I fibres. Macromol Chem Phys, 2005.
206: p. 1568 - 1575.

129.

He, J., S. Cui, and S.-Y. Wang, Preparation and crystalline analysis of high-grade
bamboo dissolving pulp for cellulose acetate. J Appl Polym Sci, 2008. 107: p. 1029 1038.

130.

Schuerch, C., Plasticizing wood with liquid ammonia. J Ind Eng Chem, 1963. 55: p. 39.

131.

Thygesen, A., et al., On the determination of crystallinity and cellulose content in plant
fibres. Cellulose, 2005. 12(6): p. 563-576.

132.

Barry, A., F. Peterson, and A. King, X-ray studies of reactions of cellulose in nonaqueous systems. I. Interaction of cellulose and liquid ammonia. J Am Chem Soc, 1936.
58: p. 333 - 337.

133.

Clark, G. and E. Parker, An X-ray diffraction study of the action of liquid ammonia on
cellulose and its derivatives. J Phys Chem, 1937. 41: p. 777 - 786.

134.

Sinitsyn, A., A. Gusakov, and E. Vlasenko, Effect of structural and physico-chemical
features of cellulosic substrates on the efficiency of enzymatic hydrolysis. Appl Biochem
Biotechnol, 1991. 30: p. 43 - 59.

135.

Gharpuray, M.M., Y.-H. Lee, and L.T. Fan, Structural modification of lignocellulosics by
pretreatments to enhance enzymatic hydrolysis. Biotechnology and Bioengineering,
1983. 25(1): p. 157-172.

136.

Murnen, H.K., et al., Optimization of Ammonia Fiber Expansion (AFEX) Pretreatment
and Enzymatic Hydrolysis of Miscanthus x giganteus to Fermentable Sugars.
Biotechnology Progress, 2007. 23(4): p. 846-850.

137.

Garlock, R.J., V. Balan, and B.E. Dale, Optimization of AFEXTM pretreatment conditions
and enzyme mixtures to maximize sugar release from upland and lowland switchgrass.
Bioresource Technology, 2012. 104(0): p. 757-768.

138.

Mielenz, J.R., et al., Lignocellulosic Biomass Pretreatment Using AFEX, in Biofuels.
2009, Humana Press. p. 61-77.

139.

Wada, M., et al., Improved structural data of cellulose IIII prepared in supercritical
ammonia. Macromolecules, 2001. 34: p. 1237 - 1243.
271

140.

Bellesia, G., et al., Probing the early events associated with liquid ammonia pretreatment
of native crystalline cellulose. J Phys Chem B, 2011. 115(32): p. 9782-8.

141.

Saafan, A., S. Kandil, and A. Habib, Liquid-ammonia and caustic mercerization of cotton
fibers using X-ray, infrared, and sorption measurements. Text Res J, 1984. 54: p. 863 867.

142.

Wadehra, I.L., R.S.J. Manley, and D.A.I. Goring, Permanently amorphous cellulose.
Journal of Applied Polymer Science, 1965. 9(7): p. 2634-2636.

143.

Bel'chev, F.V., N.I. Shuikin, and S.S. Novikov, The catalytic amination of alcohols.
Bulletin of the Academy of Sciences of the USSR, Division of chemical science, 1961.
10(4): p. 599-602.

144.

Klinkenberg, J.L. and J.F. Hartwig, Catalytic Organometallic Reactions of Ammonia.
Angewandte Chemie International Edition, 2010. 50(1): p. 86-95.

145.

Chundawat, S.P.S., et al., Multi-scale visualization and characterization of
lignocellulosic plant cell wall deconstruction during thermochemical pretreatment.
Energy & Environmental Science, 2011. 4(3): p. 973-984.

146.

Iyer, P., et al., Ammonia recycled percolation process for pretreatment of herbaceous
biomass. Applied Biochemistry and Biotechnology, 1996. 57-58(1): p. 121-132.

147.

Morrison, W.H., et al., Investigation of the ester- and ether-linked phenolic constituents
of cell wall types of normal and brown midrib pearl millet using chemical isolation,
microspectrophotometry and 13C NMR spectroscopy. Journal of the Science of Food and
Agriculture, 1993. 63(3): p. 329-337.

148.

McMillan, J., et al., Corn Stover Fractions and Bioenergy, in Twenty-Seventh Symposium
on Biotechnology for Fuels and Chemicals. 2006, Humana Press. p. 104-116.

149.

Argillier, O., et al., Genotypic variation in phenolic components of cell-walls in relation
to the digestibility of maize stalks. Agronomie, 1996. 16(2): p. 123-130.

150.

Bergvinson, D.J., J.T. Arnason, and L.N. Pietrzak, Localization and quantification of cell
wall phenolics in European corn borer resistant and susceptible maize inbreds. Canadian
Journal of Botany, 1994. 72(9): p. 1243-1249.

151.

Lu, F. and J. Ralph, Detection and Determination of p-Coumaroylated Units in Lignins.
Journal of Agricultural and Food Chemistry, 1999. 47(5): p. 1988-1992.

272

152.

Ralph, J., J.H. Grabber, and R.D. Hatfield, Lignin-ferulate cross-links in grasses: active
incorporation of ferulate polysaccharide esters into ryegrass lignins. Carbohydrate
Research, 1995. 275(1): p. 167-178.

153.

Grabber, J.H., J. Ralph, and R.D. Hatfield, Cross-Linking of Maize Walls by Ferulate
Dimerization and Incorporation into Lignin. Journal of Agricultural and Food Chemistry,
2000. 48(12): p. 6106-6113.

154.

Riboulet, C., et al., Genetic variation in maize cell wall for lignin content, lignin
structure, p-hydroxycinnamic acid content, and digestibility in set of 19 lines at silage
harvest maturity Maydica, 2008. 53(1): p. 11-19.

155.

Huang, G., C. Zhang, and Z. Chen, Pulping of Wheat Straw with Caustic PotashAmmonia Aqueous Solutions and Its Kinetics. Chinese Journal of Chemical Engineering,
2006. 14(6): p. 729-733.

156.

Bludworth, J. and F. Carl Knopf, Reactive extraction of lignin from wood using
supercritical ammonia-water mixtures. The Journal of Supercritical Fluids, 1993. 6(4): p.
249-254.

157.

Liu, Z., et al., Aqueous-ammonia delignification of miscanthus followed by enzymatic
hydrolysis to sugars. Bioresource Technology, 2012. 135(0): p. 23-29.

158.

Zhang, Y.-j., et al., Effects of supercritical ammonia on bamboo pulping. Forestry Studies
in China, 2011. 13(1): p. 80-84.

159.

Obst, J.R., Kinetics of Alkaline Cleavage of β-Aryl Ether Bonds in Lignin Models:
Significance to Delignification. Holzforschung, 1983. 37(1): p. 23-28.

160.

Huang, G., J.X. Shi, and T.A.G. Langrish, NH4OH-KOH pulping mechanisms and
kinetics of rice straw. Bioresource Technology, 2007. 98(6): p. 1218-1223.

161.

del Rio, J.C., et al., Structural Characterization of Wheat Straw Lignin as Revealed by
Analytical Pyrolysis, 2D-NMR, and Reductive Cleavage Methods. Journal of Agricultural
and Food Chemistry, 2012. 60(23): p. 5922-5935.

162.

Stewart, D., Lignin as a base material for materials applications: Chemistry, application
and economics. Industrial Crops and Products, 2008. 27(2): p. 202-207.

163.

Agarwal, U.P., R.R. Reiner, and S.A. Ralph, Estimation of Cellulose Crystallinity of
Lignocelluloses Using Near-IR FT-Raman Spectroscopy and Comparison of the Raman
and Segal-WAXS Methods. Journal of Agricultural and Food Chemistry, 2013. 61(1): p.
103-113.

273

164.

Agarwal, U., R. Reiner, and S. Ralph, Cellulose I crystallinity determination using FTRaman spectroscopy: univariate and multivariate methods. Cellulose, 2010. 17(4): p.
721-733.

165.

Yu, Z., et al., The effect of delignification of forest biomass on enzymatic hydrolysis.
Bioresource Technology, 2011. 102(19): p. 9083-9089.

166.

Dien, B.S., M.A. Cotta, and T.W. Jeffries, Bacteria engineered for fuel ethanol
production: current status. Applied Microbiology and Biotechnology, 2003. 63(3): p.
258-266.

167.

Jin, M., et al., Quantitatively understanding reduced xylose fermentation performance in
AFEXTM treated corn stover hydrolysate using Saccharomyces cerevisiae 424A (LNHST) and Escherichia coli KO11. Bioresource Technology, 2012. 111(0): p. 294-300.

168.

Hodge, D., et al., Model-Based Fed-Batch for High-Solids Enzymatic Cellulose
Hydrolysis. Appl Biochem Biotechnol, 2009. 152: p. 88 - 107.

169.

Hodge, D., et al., Soluble and insoluble solids contributions to high-solids enzymatic
hydrolysis of lignocellulose. Bioresour Technol, 2008. 99: p. 8940 - 8948.

170.

Kristensen, J., C. Felby, and H. Jorgensen, Yield-determining factors in high-solids
enzymatic hydrolysis of lignocellulose. Biotechnology for Biofuels, 2009. 2(1): p. 11.

171.

Lau, M.W., et al., An integrated paradigm for cellulosic biorefineries: Utilization of
lignocellulosic biomass as self-sufficient feedstocks for fuel, food precursors and
Saccharolytic enzyme production. Energy & Environmental Science, 2012.

172.

Caminal, G., J. López-Santín, and C. Solà, Kinetic modeling of the enzymatic hydrolysis
of pretreated cellulose. Biotechnology and Bioengineering, 1985. 27(9): p. 1282-1290.

173.

Okazaki, M. and M. Moo-Young, Kinetics of enzymatic hydrolysis of cellulose:
Analytical description of a mechanistic model. Biotechnology and Bioengineering, 1978.
20(5): p. 637-663.

174.

Xiao, Z., et al., Effects of sugar inhibition on cellulases and beta-glucosidase during
enzymatic hydrolysis of softwood substrates. Appl Biochem Biotechnol, 2004. 113-16: p.
1115 - 1126.

175.

Studer, M.H., et al., Lignin content in natural Populus variants affects sugar release.
Proceedings of the National Academy of Sciences, 2011. 108(15): p. 6300-6305.

176.

Gellerstedt, G., J. Pranda, and E.L. Lindfors, Structural and Molecular Properties of
Residual Birch Kraft Lignins. Journal of Wood Chemistry and Technology, 1994. 14(4):
p. 467-482.
274

177.

Shimada, K., S. Hosoya, and T. Ikeda, Condensation Reactions of Softwood and
Hardwood Lignin Model Compounds Under Organic Acid Cooking Conditions. Journal
of Wood Chemistry and Technology, 1997. 17(1-2): p. 57-72.

178.

Guerra, A., et al., Toward a Better Understanding of the Lignin Isolation Process from
Wood. Journal of Agricultural and Food Chemistry, 2006. 54(16): p. 5939-5947.

179.

Holtman, K.M., et al., Quantitative 13C NMR Characterization of Milled Wood Lignins
Isolated by Different Milling Techniques. Journal of Wood Chemistry and Technology,
2006. 26(1): p. 21-34.

180.

Granata,
A.
and
D.S.
Argyropoulos,
2-Chloro-4,4,5,5-tetramethyl-1,3,2dioxaphospholane, a Reagent for the Accurate Determination of the Uncondensed and
Condensed Phenolic Moieties in Lignins. Journal of Agricultural and Food Chemistry,
1995. 43(6): p. 1538-1544.

181.

Allison L, R.A., Hsieh J, Metal profiling of southeastern U.S. softwood and hardwood
furnish. Tappi J, 2000. 83(8): p. 97-1.

182.

Balan, V., et al., Overview to ammonia pretreatments for lignocellulosic biorefineries.
Dynamic Biochemistry, Process Biotechnology and Molecular Biology. Vol. 6. 2012:
Global Science Books.

183.

Withers, S., et al., Identification of Grass-specific Enzyme That Acylates Monolignols
with p-Coumarate. Journal of Biological Chemistry, 2012. 287(11): p. 8347-8355.

184.

Fornale, S., et al., Altered Lignin Biosynthesis Improves Cellulosic Bioethanol
Production in Transgenic Maize Plants Down-Regulated for Cinnamyl Alcohol
Dehydrogenase. Molecular Plant, 2011. 5(4): p. 817-830.

185.

Chabbert, B., et al., Biological variability in lignification of maize: Expression of the
brown midrib bm3 mutation in three maize cultivars. Journal of the Science of Food and
Agriculture, 1994. 64(3): p. 349-355.

186.

Hu, Z., M. Foston, and A.J. Ragauskas, Comparative studies on hydrothermal
pretreatment and enzymatic saccharification of leaves and internodes of alamo
switchgrass. Bioresour Technol, 2011. 102(14): p. 7224-8.

187.

Hallac, B.B., et al., Biomass Characterization of Buddleja davidii: A Potential Feedstock
for Biofuel Production. Journal of Agricultural and Food Chemistry, 2009. 57(4): p.
1275-1281.

275

188.

Crestini, C. and D.S. Argyropoulos, Structural Analysis of Wheat Straw Lignin by
Quantitative 31P and 2D NMR Spectroscopy. The Occurrence of Ester Bonds and -O-4
Substructures. Journal of Agricultural and Food Chemistry, 1997. 45(4): p. 1212-1219.

189.

Samuel, R., et al., Structural Characterization and Comparison of Switchgrass Ballmilled Lignin Before and After Dilute Acid Pretreatment. Applied Biochemistry and
Biotechnology, 2010. 162(1): p. 62-74.

190.

Akim Leonid, G., et al., Quantitative 31P NMR Spectroscopy of Lignins from Transgenic
Poplars, in Holzforschung. 2001. p. 386.

191.

Sun, X.-F., et al., Extraction and characterization of lignins from maize stem and
sugarcane bagasse. Journal of Applied Polymer Science, 2011. 120(6): p. 3587-3595.

192.

Jiang, G., D.J. Nowakowski, and A.V. Bridgwater, Effect of the Temperature on the
Composition of Lignin Pyrolysis Products. Energy & Fuels, 2010. 24(8): p. 4470-4475.

193.

Black, L.T., G.F. Spencer, and O.L. Brekke, Reactions of lipids in corn with ammonia.
Journal of the American Oil Chemists' Society, 1978. 55(6): p. 526-529.

194.

Parashar, V., S.K. Pandey, and A.C. Pandey, Low-temperature synthesis of quantum size
gadolinium monosulfide (GdS) nanoparticles and their pathogen capture efficiency.
Chemical Communications, 2010. 46(18): p. 3143-3145.

195.

Hurtta, M., I. Pitkanen, and J. Knuutinen, Melting behaviour of d-sucrose, d-glucose and
d-fructose. Carbohydrate Research, 2004. 339(13): p. 2267-2273.

196.

Räisänen, U., et al., Formation of the main degradation compounds from arabinose,
xylose, mannose and arabinitol during pyrolysis. Journal of Thermal Analysis and
Calorimetry, 2003. 72(2): p. 481-488.

197.

Cho, J., et al., Kinetics and reaction chemistry for slow pyrolysis of enzymatic hydrolysis
lignin and organosolv extracted lignin derived from maplewood. Green Chemistry, 2011.
14(2): p. 428-439.

198.

Ho, N.Y., et al., Successful Design and Development of Genetically Engineered
Saccharomyces Yeasts for Effective Cofermentation of Glucose and Xylose from
Cellulosic Biomass to Fuel Ethanol, in Recent Progress in Bioconversion of
Lignocellulosics. 1999, Springer Berlin Heidelberg. p. 163-192.

199.

Jin, M., et al., Continuous SSCF of AFEX™ pretreated corn stover for enhanced ethanol
productivity using commercial enzymes and Saccharomyces cerevisiae 424A (LNH-ST).
Biotechnology and Bioengineering, 2013. 110(5): p. 1302-1311.

276

200.

Howell, J.A. and M. Mangat, Enzyme deactivation during cellulose hydrolysis.
Biotechnology and Bioengineering, 1978. 20(6): p. 847-863.

201.

Kumar, R. and C.E. Wyman, Does change in accessibility with conversion depend on
both the substrate and pretreatment technology? Bioresource Technology, 2009.
100(18): p. 4193-4202.

202.

Zhou, W., Y. Xu, and H.-B. Schüttler, Cellulose hydrolysis in evolving substrate
morphologies III: Time-scale analysis. Biotechnology and Bioengineering, 2010. 107(2):
p. 224-234.

203.

Zhang, S., D.E. Wolfgang, and D.B. Wilson, Substrate heterogeneity causes the
nonlinear kinetics of insoluble cellulose hydrolysis. Biotechnology and Bioengineering,
1999. 66(1): p. 35-41.

204.

Dawes, E., D. Ribbons, and L. PJ, The route of ethanol formation in Zymomonas mobilis.
Biochem J. , 1966. 98(3): p. 795-803.

205.

Long, T.M., et al., Cofermentation of Glucose, Xylose, and Cellobiose by the BeetleAssociated Yeast Spathaspora passalidarum. Applied and Environmental Microbiology,
2012. 78(16): p. 5492-5500.

206.

Humbird, D., et al., Process Design and Economics for Biochemical Conversion of
Lignocellulosic Biomass to Ethanol: Dilute-Acid Pretreatment and Enzymatic Hydrolysis
of Corn Stover. 2011, NREL.

207.

Voitkevich, O.V., et al., Thermodynamic Properties of Plant Biomass Components. Heat
Capacity, Combustion Energy, and Gasification Equilibria of Lignin. Journal of
Chemical & Engineering Data, 2012. 57(7): p. 1903-1909.

208.

Blokhin, A.V., et al., Thermodynamic Properties of Plant Biomass Components. Heat
Capacity, Combustion Energy, and Gasification Equilibria of Cellulose. Journal of
Chemical & Engineering Data, 2011. 56(9): p. 3523-3531.

209.

Adjei, T., Characterization of a novel biodegradable material to reduce emission of
ammonia, in Biological Systems Engineering. 2007, Virginia Polytechnic Institute and
State Univeristy: Blacksburg, Virginia.

210.

Perry, R.H. and D. Green, Perry's Chemical Engineers' Handbook. 6th ed. 1984, New
York: McGraw-Hill.

211.

Peters, M.S. and K.D. Timmerhaus, Plant Design and Economics for Chemical
Engineers. 1991, New York: McGraw-Hill.

212.

Page, J.S., Conceptual Cost Estimating Manual. 1996, Houston: Gulf Publishing.
277

213.

Matches. http://matche.com. 2013 [cited 2013.

214.

http://www.duncanseddon.com/docs/pdf/c9-the-price-of-ammonia.pdf. 2013 [cited 2013.

215.

http://www.alibaba.com/showroom/lignin.html. 2013 [cited 2013.

216.

Eranki, P.L., B.D. Bals, and B.E. Dale, Advanced Regional Biomass Processing Depots:
a key to the logistical challenges of the cellulosic biofuel industry. Biofuels, Bioproducts
and Biorefining, 2011. 5(6): p. 621-630.

278