PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

UUh

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProj/Acc&Pres/ClRC/DaieDue,indd

 

ASSESSMENT OF OPTIONS FOR THE INTEGRATION OF FOOD AND FUEL
PRODUCTION IN CELLULOSIC ETHANOL REFINING

By

Bryan Bals

A DISSERTATION
Submitted to
Michigan State University
In partial fulﬁllment of the requirements
For the degree of
DOCTOR OF PHILOSOPHY
Chemical Engineering

2010

ABSTRACT

ASSESSMENT OF OPTIONS FOR THE INTEGRATION OF FOOD AND FUEL
PRODUCTION IN CELLULOSIC ETHANOL REFINING

By

Bryan Bals

Reducing petroleum consumption is one of the primary challenges of the United
States and the world in the let century. Biofuels are seen as a primary
alternative, yet concerns of decreasing food production due to increased demand
for land remain. This research proposes two technologies to integrate biofuel
production with animal feeds and lessen this demand: using AF EX treated
biomass as a ﬁber source for ruminants and extracting leaf protein as a protein
source. The purpose of this study is to investigate the viability of these two

technologies for both economic value and increasing the productivity of land.

Experimental results indicate that an early harvest of switchgrass, which would
be required for protein production, requires milder pretreatment condition and
has higher yields than late harvest biomass. Ammonia-based extraction was
successful in removing approximately 40% of the protein from switchgrass.
However, AFEX did not increase extraction yields, and resulting sugar yields
decreased after extraction. After hydrolysis, nearly all of the protein was soluble,
but ultraﬁltration could only concentrate 30-45% of the protein. AF EX increased
the digestibility of ﬁber in multiple feedstocks and increased the crude protein

content to levels comparable to common forages. The digestibility of pretreated

late harvest switchgrass is comparable to high quality forages, while the energy

in corn stover is approximately 85% of the value of corn grain.

From these experimental results, two models were created to determine the
potential of these technologies. The ﬁrst, an economic and material model,
suggests that animal feed integration with ethanol production can displace the
equivalent of 2900-4800 L gasoline per ha of land removed from feed use
compared to 1600 L/ha if no feed integration is performed. Likewise, the
proﬁtability of the land increases to $150-$380/ha for integrated animal feed
scenarios compared to $35/ha for ethanol production. The second model
considers the total land use in the United States and estimates the potential
market for biofuels, AFEX-treated feeds, and protein extracts. These two
technologies increase the amount of biofuel that can be produced by 42 GL of
ethanol on cropland currently used for animal feed or ethanol production without
decreasing animal feed production. Approximately 85 T9 of AF EX-treated feeds

and 52 Tg of biomass for protein extraction are consumed in this scenario.

Thus, this study suggests further research into integrating animal feed production
with biofuel production should be pursued. Emphasis should be applied to
readying AFEX-treated feeds for commercialization, primarily through animal
feeding trials. For protein extraction, future research should be focused primarily

on using the remaining ﬁber for ethanol production.

DEDICATION

To my parents, Carl and Barbara Bals, for all of their love and support

iv

ACKNOWLEDGEMENTS
I would like to thank my advisor, Dr. Bruce Dale, for his guidance and support
throughout the project. I would also like to thank my guidance committee: Dr.

Michael Allen, Dr. Dennis Miller, and Dr. Mark Worden.

I am deeply appreciative of all members of the Biomass Conversion Research
Laboratory for their support, helpful advice, and pertinent questions, especially
Dr. Venkatesh Balan, Elizabeth Sendich, and Rebecca Garlock. In addition, I
thank my undergraduate assistants - Laura Teachworth, Hannah Murnen, and
Chad Rogers — as well as the technicians in the BCRL - Christa Gunawan,
James Humpula, and Charles Donald, Jr. — for their assistance with the

experimental work.

I would like to thank Dr. David Bransby, Dr. Bruce Dien, Dr. A. Daniel Jones, Dr.
Joseph Leykam, Dr. Johan Sanders, Ben Brehmer, David Main, Dr. Mark Laser,

and Dr. Rod McDonald for their contributions.

Financial support for this project was provided by the Midwest Consortium for
Biobased Fuels and Chemicals under the US Department of Energy Contract
DE-FG36-04GO14220, the US Department of Agriculture grant 68—3A75-6-505,
the Michigan State University Research Foundation, the Great Lakes Bioenergy
Research Center under the US Department of Energy Grant DE-FCOZ-

07ER64494, and the Michigan State University Distinguished Fellowship.

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................... viii
LIST OF FIGURES ............................................................................................... xi
KEY TO SYMBOLS AND ABBREVIATIONS .................................................... xvii
CHAPTER 1 : INTRODUCTION ............................................................................ 1
1.1 Justiﬁcation .................................................................................................. 7
1.2 Project Description ...................................................................................... 9
1.3 Objectives .................................................................................................. 10
1.4 Limitations ................................................................................................. 11
CHAPTER 2 : LITERATURE REVIEW ................................................................ 12
2.1 Ethanol Production .................................................................................... 12
2.2 Food vs Fuel .............................................................................................. 16
2.3 Ammoniation of Feeds ............................................................................... 19
2.4 Leaf Protein Concentrate ........................................................................... 22
CHAPTER 3 : ENZYMATIC HYDROLYSIS OF SWITCHGRASS ....................... 27
3.1 Introduction ................................................................................................ 27
3.2 Materials and Methods .............................................................................. 29
3.3 Results ....................................................................................................... 34
3.4 Discussion ................................................................................................. 49
3.5 Conclusions ............................................................................................... 51
CHAPTER 4 : PROTEIN EXTRACTION ............................................................. 53
4.1 Introduction ................................................................................................ 53
4.2 Materials and Methods .............................................................................. 54
4.3 Results ....................................................................................................... 59
4.4 Discussion ................................................................................................. 76
4.5 Conclusions ............................................................................................... 80
CHAPTER 5 : FIBER RUMEN DIGESTIBILITY .................................................. 82
5.1 Introduction ................................................................................................ 82
5.2 Materials and Methods .............................................................................. 83
5.3 Results ....................................................................................................... 89
5.4 Discussion ................................................................................................. 98
5.5 Conclusions ............................................................................................. 104
CHAPTER 6 : SUMMARY OF SWITCHGRASS TREATMENT MODULES ...... 106
6.1 Introduction .............................................................................................. 106
6.2 AFEX Pretreatment and Ethanol Production ........................................... 107

vi

6.3 Aqueous Protein Extraction ..................................................................... 122

6.4 Mechanical Pressing ............................................................................... 127
CHAPTER 7 : DIRECT LAND USE MODEL ..................................................... 136
7.1 Introduction .............................................................................................. 136
7.2 Method .................................................................................................... 137
7.3 Results and Discussion ........................................................................... 144
7.4 Conclusions ............................................................................................. 162
CHAPTER 8 : AGGREGRATE LAND USE MODEL ......................................... 164
8.1 Introduction .............................................................................................. 164
8.2 Method .................................................................................................... 166
8.3 Results and Discussion ........................................................................... 172
8.4 Conclusions ............................................................................................. 178
CHAPTER 9 : CONCLUSIONS AND RECOMMENDATIONS .......................... 180
9.1 Conclusions ............................................................................................. 180
. 9.2 Recommendations for Further Study ....................................................... 182
APPENDIX A: DISTILLER’S GRAIN HYDROLYSIS ......................................... 187
A1 Introduction ............................................................................................. 187
A2 Materials and Methods ............................................................................ 190
A3 Results and Discussion ........................................................................... 192
A4 Conclusions ............................................................................................. 205
APPENDIX B: DISTILLER’S GRAIN PROTEIN FRACTIONATION .................. 207
B1 Introduction ............................................................................................. 207
B2 Materials and Methods ............................................................................ 209
3.3 Results and Discussion ........................................................................... 216
B4 Conclusions ............................................................................................. 235
REFERENCES ......................... 4 ......................................................................... 239

vii

LIST OF TABLES

Table 2.1 : Five leading pretreatment technologies and their operating conditions
for treating corn stover. All information was obtained from Wyman et al. [22]
except for steam explosion, which was obtained from Bura et al. [23]. ............... 14

Table 2.2 : Literature values for potential biofuel production within the United
States .................................................................................................................. 17

Table 2.3 : Total farmland in the United States and their primary usage. [9] Land
in the conservation reserve program may fall into other categories, particularly
idle land. .............................................................................................................. 19

Table 2.4 : Improvements in ﬁber digestibility (NDF) and crude protein (CP)
content due to ammonia treatment for ﬁve types of biomass. ............................. 21

Table 2.5 : Yields of leaf protein concentrates from solid/liquid extraction
obtained from different studies ............................................................................ 25

Table 3.1 : Composition analysis for the July and October harvests of Cave-in-

Rock (CIR) and Alamo switchgrass ..................................................................... 35
Table 3.2 : Reduced linear model for modeling monomeric sugar release with

respect to four AFEX pretreatment parameters .................................................. 37
Table 3.3 : AFEX conditions within the parameters tested in this study .............. 40

Table 3.4 : Reduced linear model for modeling monomeric sugar release with
respect to enzyme addition ................................................................................. 43

Table 3.5 : Optimal enzyme loadings for enzymatic hydrolysis of four harvests of
switchgrass ......................................................................................................... 44

Table 4.1 : Composition of switchgrass used in this study. All values given as

g/1009 dry matter. ............................................................................................... 60
Table 4.2 : Amount of monomeric glucose, monomeric xylose, and protein

solubilized during each operation in potential processes .................................... 67
Table 5.1 :AFEX treatment conditions for eleven potential forages ................... 89
Table 5.2 : NDF and nitrogen content of treated and untreated forages ............. 91

Table 5.3 : NDF digested after 48 hours of in vitro fermentation for eleven
forages ................................................................................................................ 94

viii

Table 5.4 : Comparison of the rate of ﬁber removal for corn stover and Alamo
switchgrass ......................................................................................................... 96

Table 5.5 : Total Digestible Nutrients (TDN), Net Energy available for Lactation at
3X maintenance (NEL), and crude protein (CP) content in ﬁve traditional feeds
and two AF EX feeds ......................................................................................... 102

Table 6.1 : Levels used within the general full factorial model of AF EX
pretreatment conditions ..................................................................................... 109

Table 6.2 : Impact of ammonia loading, water loading, and temperature on
equipment cost within the bioreﬁnery model. Only individual equipment with a
coefﬁcient of variation > 5% and average capital cost > $100,000 are shown. .111
Table 6.3 : Impact of ammonia loading, water loading, and temperature on heat
and electricity streams within the bioreﬁnery model. A residence time of 10
minutes was used for this analysis. Only individual equipment with a coefﬁcient
of variation > 5% are shown .............................................................................. 114

Table 6.4 : Impact of ammonia loading, water loading, and temperature on
variable operating costs within the bioreﬁnery model. ....................................... 114

Table 6.5 : Reduced linear model for the makeup ammonia required. The MESP
model does not include the cost of makeup ammonia. ..................................... 117

Table 6.6 : List of variables used in the bioreﬁnery portion of the direct land use
model ................................................................................................................ 1 19

Table 6.7 : List of constants in the bioreﬁnery portion of the land use model.
Units are such that the ﬁnal equation will result in $/Mg BM. ............................ 120

Table 6.8 : Example material balance for aqueous protein extraction. Stream
labels are shown in Figure 6.2 .......................................................................... 124

Table 6.9 : Total variable and operating costs for the aqueous protein extraction
module. ............................................................................................................. 126

Table 6.10 : Example mass balance for the mechanical pressing module of the
bioreﬁnery. Example is for 1 Mg biomass; all values are in kg. ........................ 134

Table 6.11 : Operating and capital costs for the mechanical pressing module. 135

Table 7.1 : Scenarios considered for the direct land use model ....................... 138

ix

Table 7.2 : Nutrient values of all feeds produced in this study and references for
corn, soy, and grass hay. Values are in Mg/ha for the reference feeds (soy, corn,
hay) and Mg/Mg biomass for the feeds produced in the study. ......................... 140

Table 7.3 : Variables considered in the sensitivity analyses. The medium values
are used for the base case scenario. ................................................................ 143

Table 7.4 : Direct land displaced, gasoline displaced, and total energy displaced
per ha of each scenario produced using the base case assumptions. TED -
Total energy displaced. ..................................................................................... 145

Table 7.5 : Net proﬁt for each operation for all scenarios under the base case
assumptions ...................................................................................................... 146

Table 7.6 : Mean and standard deviations of both total energy density and proﬁt
for the Monte Carlo analysis for each scenario. ................................................ 150

Table 7.7 : Effect of each variable on the proﬁtability for each scenario. Low and
high values are as shown in Table 7.3 .............................................................. 159

Table 8.1 : Cropland and yields used for this study [9] ..................................... 165
Table 8.2 : Livestock population and feed requirements in the United States ..167

Table 8.3 : Yields and nutritional value for all feeds produced in the aggregate
land use model. ................................................................................................. 171

Table 8.4 : Land use (in hectares) of the ﬁve crops under the scenarios studied.
.......................................................................................................................... 176

Table A.1: Composition of four different distiller’s grain samples used in this
study. Sample 1 was obtained from Big River Resources and is used in all
experiments in this chapter. Sample 4 was obtained from the Woodbury, MI
plant of US Bioenergy and used for all experiments in Appendix B ................ 193

Table A.2: Differences in glucose yields released after enzymatic hydrolysis of
two batches of DDGS obtained from the same dry grind facility .................... 202

Table 8.1: Composition of distiller’s grains used in this study. The ﬁrst column
represents the original wet distiller’s grain and solubles while the second column
represents the insoluble residue after enzymatic hydrolysis of ﬁber ............... 229

LIST OF FIGURES

Figure 1.1 : United States primary energy consumption by source (left) and
petroleum use by sector (right) in 2007. All numbers are in quadrillion BTUs. [1] .2

Figure 1.2 : United States ethanol production, 1980-2008. [5] ............................. 4

Figure 1.3 : Renewable fuel standard requirements in the United States as
implemented in the Energy Independence and Security Act of 2007. Current
technology is dominated by corn-based ethanol, and has a limit of 15 billion
gallons per year (bgpy). By 2022, the single largest source of renewable fuel is
cellulosic ethanol (16 bgpy). .................................................................................. 5

Figure 3.1 : Glucose and xylose yields at varying pretreatment conditions. Each
individual data point represents yields after 72 hours of enzymatic hydrolysis for
a speciﬁc AF EX pretreatment. Pretreatment conditions were varied as stated in
the text. For all data points, hydrolysis was performed at 3% solid loading, 50°C,
200rpm rotation, and 5mg Accelerase/g dry biomass. ........................................ 39

Figure 3.2 : Glucose and xylose yields at varying enzyme loadings. All data
points within the same harvest were performed at the same AF EX pretreatment
conditions (listed in Table 3), and each data point represents a separate
combination of enzymes. Hydrolysis was performed at 3% solid loading, 50°C,
and 200rpm rotation, with samples collected after 72 hours. .............................. 45

Figure 3.3 : Rate of hydrolysis. Glucose (right) and xylose (left) released during
enzymatic hydrolysis between 3 and 168 hours of residence time. Pretreatments
were performed at the conditions listed in Table 3 and enzyme addition as listed
in Table 5. Hydrolysis was performed at 3% solid loading, 50°C, and 200rpm
rotation. ............................................................................................................... 48

Figure 4.1 : Effect of extraction temperature on protein yields. All extractions
were performed in duplicate with 3% ammonia at pH = 10.5. The results are
combined after two separate extractions using 11:1 liquid/solid ratio and 3 minute
residence time under 1500 psi pressure. ............................................................ 61

Figure 4.2 : Effect of ammonia concentration on protein yields. All extractions
were performed in duplicate at 50°C and at pH = 10.5 except for 0%
concentration, which was performed using distilled water at 50°C. The results
are combined after two separate extractions using 11:1 liquid/solid ratio and 3
minute residence time under 1500 psi pressure .................................................. 62

Figure 4.3 : Effect of extraction pH on protein yields. All extractions were

performed in duplicate with 3% ammonia and at 50°C. The results were
combined after two separate extractions using 11:1 liquid/solid ratio and 3 minute
residence time under 1500 psi pressure. ............................................................ 63

xi

Figure 4.4 : Effect of reducing agents on protein yields for untreated and AFEX
treated samples. All extractions were performed in duplicate with 3% ammonia,
25°C, and at pH = 10.5. The results are combined after two separate extractions
using 11:1 liquid/solid ratio and 30 minute residence time at atmospheric
pressure. Both the ionic sodium dodecyl sulfate (SDS) and the nonionic Tween
80 (Tw80) surfactants were tested, both with and without the addition of B-
mercaptoethanol. ................................................................................................ 64

Figure 4.5 : Amino acid proﬁles for untreated protein extract, AFEX treated
protein extract, and the native switchgrass protein. Asx and Glx are a
combination of aspartic acid and asparagine and glutamic acid and glutamine,
respectively. Cysteine and tryptophan were not detected due to their instability
during acid hydrolysis .......................................................................................... 66

Figure 4.6 : Mass balance for switchgrass extraction. Water is not included in
the balance, as the biomass was washed and allowed to dry between each step
to insure complete solid/liquid separation. .......................................................... 69

Figure 4.7 : Concentration proﬁle for ultraﬁltration of 100 mL switchgrass extract
using a 50 cm2 10 kda molecular weight cutoff membrane. Both runs were
performed at a constant 30 mUmin crossﬂow rate and the transmembrane
pressure (TMP) allowed to increase as necessary .............................................. 71

Figure 4.8 : Permeate ﬂux for ultraﬁltration of switchgrass extract using a 10 kDa
molecular weight cutoff membrane at two different transmembrane pressures
(TMP). Permeate rate is determined based on the average rate for 10 minutes,
either immediately after starting ﬁltration or after 3 hours of running. The
permeate and retentate were recombined with the feed to keep a constant
protein concentration in the feed. ........................................................................ 72

Figure 4.9 : Protein recovery using ultraﬁltration for four separate protein
solutions: 1 — extraction on AFEX treated switchgrass, 2 — hydrolysate, 3 —
extraction on post hydrolysis solid residue, 4 — extraction on untreated
switchgrass. ........................................................................................................ 73

Figure 4.10 : Effect of ethanol or acid precipitation at pH 3.5 and 5 on protein
concentration. Untreated switchgrass extracted at 50C using 0.1 M NaOH was
used as the extract in this study. PMSF was added at 0.1% concentration during
the extraction for half of the experiments. Error bars represent the high and low
values of duplicate trials. ..................................................................................... 75

Figure 4.11 : Amount of polyphenolic compounds (in equivalent moles of gallic

acid) remaining in the extract after protein precipitation on untreated switchgrass
extract. ................................................................................................................ 76

xii

Figure 5.1: Effect of conventional ammoniation and AFEX treatment compared
to untreated corn stover (upper) and switchgrass (lower) on NDF remaining over
time during NDF digestion. Error bars represent the high and low values for
duplicate samples. .............................................................................................. 95

Figure 5.2 : Relationship between monomeric sugars released during enzymatic
hydrolysis with commercial cellulases (x-axis) and in vitro NDF digestion at 48h
(y-axis) for multiple AFEX conditions of late harvest switchgrass. The line
represents a quadratic curve obtained from an ordinary least squares regression.
AFEX conditions ranged from 0.4-2.0 9 water / g dry biomass, 0.4-2.0 g ammonia

/g dry biomass, 5-30 minute residence time, and 80-150°C ............................... 97
Figure 6.1 : Simpliﬁed process ﬂow diagram of AFEX pretreatment and ammonia
recovery system. ............................................................................................... 108
Figure 6.2 : Process ﬂow diagram for aqueous protein extraction .................... 123

Figure 6.3 : Yield of solubles obtained at different starting water contents (9
water / g wet biomass) for a 1, 2, or 3 press system. ........................................ 129

Figure 6.4 : Utility cost (steam for heat coagulation and electricity for the screw
press) as a function of solubles yield for the 1, 2, or 3 press system. ............... 130

Figure 6.5 : Amount of water required for the process for a 1, 2, or 3 press
system ............................................................................................................... 131

Figure 6.6 : Process ﬂow diagram for mechanical pressing module of the

bioreﬁnery. The scenario where the ﬁber is used for hydrolysis is shown here. If
the ﬁber is used for feed, the remaining whey that would be used for hydrolysis is
evaporated and condensed. .............................................................................. 133

Figure 7.1 : Cumulative distribution functions for proﬁt for all scenarios ........... 152

Figure 7.2 : Cumulative distribution functions for the total energy density (TED)
of each scenario for the Monte Carlo analysis. ................................................. 155

Figure 7.3 : Tornado plot of low and high values for several variables on the
economic value of Scenario J. .......................................................................... 157

Figure 7.4 : Impact on relative crop yields for a one or two harvest system on the
economics of each scenario. The combined crop yield of the two harvest system
divided by the yield of the single harvest is the crop yield ratio, while the net proﬁt
of scenario D, H, or L divided by the net proﬁt of scenario B is the proﬁt ratio. .162

Figure 8.1 : Ethanol production on selected cropland for the base case and 13
scenarios tested. ............................................................................................... 173

xiii

Figure 8.2 : Amount of biomass used for advanced feed technologies. Both
AF EX—feeds and forage for leaf protein production are shown .......................... 175

Figure 8.3 : Amount of digestible nutrients in ruminant diets as a percentage of
dry matter intake. .............................................................................................. 178

Figure A.1: Effect of temperature on the glucose yield of AFEX treated DDGS.
During AFEX, water loading was held constant at 0.13 g/g biomass, and
ammonia loading at 1.0 g/g biomass. Enzymatic hydrolysis was performed at a
cellulose loading of 1%. All runs were done in duplicate and error bars represent
the maximum and minimum values. ....................................................... 194

Figure A.2: Effect of ammonia loading on the glucose yield of AF EX treated
DDGS. Moisture content and temperature during AFEX were held constant at
0.13 g/g biomass and 70°C, respectively. Enzymatic hydrolysis was performed
at a cellulose loading of 1%. All runs were done in duplicate and error bars
represent the maximum and minimum values. ......................................... 195

Figure A.3: The effect of water content during AFEX on glucose yields from
DDGS. AFEX temperature was held constant at 70C and ammonia loading held
constant at 1 glg biomass. Enzymatic hydrolysis was performed at a cellulose
loading of 1%. All runs were done in duplicate and error bars represent the
maximum and minimum values. ............................................................ 197

Figure A.4: The effect of AFEX pretreatment temperature on glucose yields for
both WDG and wetted DDGS. The water content was 1.5 g/g biomass for both
samples, and the ammonia loading was 1.0 g/g biomass. Enzymatic hydrolysis
was performed at a cellulose loading of 1%. All runs were done in duplicate and
error bars represent the maximum and minimum values ............................. 199

Figure A.5: The effect of AF EX pretreatment ammonia loading on glucose yields
for both wet DG and wetted DDGS. The water content was 1.5 glg biomass for
both samples, and the temperature was 80°C. Enzymatic hydrolysis was
performed at a cellulose loading of 1%. All runs were done in duplicate and error
bars represent the maximum and minimum values. ................................... 200

Figure A.6: Effect of amylase on glucose yields for both AFEX treated and
untreated DDGS. Cellulase loading was 16.5 FPU/g glucan, while amylase
addition was added at 14mg enzyme/g glucan. B-glucosidase was added at 56
pNPGU/g glucan in all cases ................................................................. 203

Figure A.7: The effect of solid loading during enzymatic hydrolysis on glucose

yields from AF EX treated DDGS. Two separate batches of DDGS were used. In
all cases, the optimal AFEX treatment of 70°C, 0.8 g ammonia/g biomass, and no
additional moisture was used. Samples were collected after 120 hours ......... 205

xiv

Figure 3.1: Effect of extraction temperature and hydroxide concentration on
solubility of protein, lysine, and biomass. All extractions were performed at 0.1 M
NaOH for the top ﬁgure and 60°C for the bottom ﬁgure on untreated DGS.
Extractions were performed at 20:1 liquid: biomass ratio for 1 h and rotated at
200 RPM .......................................................................................... 217

Figure 3.2: Effect of adding a surfactant (sodium dodecyl sulfate) and reducing
agent (b-mercaptoethanol) during alkaline extraction on protein and lysine yields.
All extractions were performed at 70°C using 0.1M NaOH, a 20:1 liquid: biomass
ratio for 1 hour, and rotated at 200 RPM on untreated DGS ......................... 219

Figure 3.3: Comparison of protein, lysine, and biomass solubility using different
extraction conditions for untreated and AFEX-treated wet DGS. Either 0.1 M or
1.0M sodium hydroxide was used as the solvent, with or without the addition of
sodium dodecyl sulfate at 0.2% loading. Extractions were performed at 70°C,
20:1 liquid: biomass ratio for 1 hour, rotated at 200 RPM ............................ 221

Figure 8.4: Protein (top) and glucose (bottom) yields for integrated extraction
and hydrolysis. Alkaline extractions were performed at 70°C, 10:1 liquid:
biomass ratio for 1 hour and rotated at 200 RPM prior to AFEX pretreatment and
enzymatic hydrolysis of ﬁber. Three extraction conditions — 0.1M NaOH with and
without SDS addition and 0.5M NaOH with no SDS addition - were tested. For
each condition, one set of samples (labeled as Washed) was rinsed at 10:1
liquid: biomass ratio following extraction to remove residual hydroxide and
surfactant. The amount of protein or glucose recovered in the extract, rinse, and
hydrolysate are shown separately. Glucose released in the extract and
hydrolysate was determined 72h after the addition of cellulase enzymes to the
media. The enzymatic hydrolysate from AFEX-treated DGS without a previous
alkaline extraction is also shown as a control ............................................ 225

Figure 3.5: Protein recovery as a function of temperature for a hydrophilic (Ethyl
Lactate, EL) and a hydrophobic (D-Limonene, DL) solvents. Protein recovery was
calculated based on the initial protein content in DGS, i.e., 30 % (dry basis)...226

Figure 3.6: Effect of the enzyme amylase on the enzymatic sacchariﬁcation of
cellulosic materials of DGS present in different solutions. The solvents used are
D—Limonene (DL), distilled methyl esters (DME) and buffer (blank). Time = 120 h,
temperature = 50 °C, AFEX-DGS : buffer: biosolvent = 1 : 2 : 1. The data points
represent the mean experimental value and the error bars represent standard
deviation ........................................................................................... 227

Figure 8.7: Rate of protein removal during protease digestion for both low (0.1%
w/w) and high (1.0% w/w) loading of protease. Lines represent best logarithmic
ﬁt. The amount at 0h represents the initial soluble protein within the whole
stillage. Extractions were performed at 50°C, a pH of 7.5, and 90rpm ........... 232

XV

Figure 8.8: Effect of extractions prior to protease addition. The control is
protease digestion without extraction. Extractions were performed at two
different pH levels at 70°C for 1h. The alkaline extract was neutralized prior to
protease addition. Here, time at 0h represents the beginning of protease
digestion ........................................................................................... 233

Figure 8.9: The effect of different proteases on protein removal. The enzymes
used are Protex 6L, Protex 14L, and Protex 51 P. Digestions were performed for
24h at 50°C and 90rpm. Protease was loaded at 0.1% (w/w) and the pH was
kept constant at 7.5 for Protex 14L and 51 P and 10 for Protex 6L ................. 234

Figure 8.10: Amino acid proﬁles for proteins released by Protex 6L, soluble

proteins in wheat stillage, and total proteins in wheat stillage. Met, Cys, and Trp
were not detected due to being destroyed during acid hydrolysis. ................ 235

xvi

KEY TO SYMBOLS AND ABBREVIATIONS

AF EX — Ammonia Fiber Expansion

CDF - Cumulative Distribution Function

CIR - Cave-in-Rock

CP — Crude Protein

DDGS - Dried Distillers Grains and Solubles

DG — Distillers Grains

DGS - Distillers Grains and Solubles

DMI — Dry Matter Intake

F PU -— Filter Paper Unit

HPLC — High Performance Liquid Chromatography
LPC — Leaf Protein Concentrate

MWCO — Molecular Weight Cutoff

NDF - Neutral Detergent Fiber

NEL - Net Energy Available for Lactation at 3X Maintenance
NOx - Nitrous Oxides

NPN - Non Protein Nitrogen

NRC - National Resource Council

NREL - National Renewable Energy Laboratory
PMSF — Phenylmethanesulphonyl Fluoride
pNPGU — p-nitrophenyl-B-D-galactopyranoside Unit
RBPC - Regional Biomass Processing Center

SDS — Sodium Dodecyl Sulfate

xvii

SEM — Standard Error of the Mean
TDN — Total Digestible Nutrients
TED — Total Energy Density

TMP - Transmembrane Pressure
UF - Ultraﬁltration

VFA — Volatile Fatty Acids

xviii

CHAPTER 1 : INTRODUCTION

One of the primary challenges facing the United States and the world in the 21St
century is the transformation from fossil fuel to renewable energy. Of primary
interest is reducing petroleum consumption. Approximately 40% of the United
States’ total energy consumption is petroleum, the single largest source of
energy consumed (see Figure 1.1). Of this energy, 70% is used for
transportation, primarily gasoline (45%) and diesel (20%) fuel. Furthermore,
transportation is the least diversified end use of energy in the United States, with
over 95% of transportation energy supplied by petroleum [1]. Thus, for
petroleum displacement with renewable energy, careful attention should be

focused on displacing these transportation fuels.

Numerous reasons — including political, economic, and environmental — have
been cited by the media, scientists, private industry, and politicians to move away
from petroleum based transportation fuels. The four primary arguments against

petroleum fuels are summarized below:

0 As a fossil fuel, petroleum is a major pollutant. Besides producing
greenhouse gases, gasoline and diesel engines produce pollutants such
as particulate matter, NOx, and ozone, thereby concentrating pollutants
within cities [2]. In addition, the risk remains of severe ecological disasters

due to oil spills such as the Exxon-Valdez spill of 1989.

Nuclear Power, Transportation,
8.4 Renewable, 6.8 2745

       

Natural Gas,
23.6 Petroleum, .
398 lmustnal, 9.55
, /~-/‘“” Residential and
” \ Commercial,
Coal, 22.8 l 1.99
Electric Power,

0.80

Figure 1.1 : United States primary energy consumption by source (left) and
petroleum use by sector (right) in 2007. All numbers are in quadrillion BTUs. [1]
o The price of oil is highly volatile, increasing from $61/bbl in April 2007 to
$137/bbl in July 2008, before rapidly falling to 640/be by December 2008
[1]. These price shocks can have a dramatic effect on the economy, and

thus should be avoided.

o The difficulty of avoiding these price shocks is increased due to the
political instability of several oil-producing nations. The Middle East
produces 30% of the world’s oil, and has been a source of political unrest

for decades. In addition, the United States’ need to import oil has often

been cited as a confounding factor in its international relations, hampering

the US’ foreign policy goals [3].

c As a nonrenewable resource, it is uncertain how much recoverable
petroleum remains in the world. In 1956, M. King Hubbert correctly
predicted the decline of United States oil production beginning around
1970 based on dwindling supplies. Similar assessments of world supply
are difficult, but some analysts claim we are nearing the peak of world oil
production [4]. Even if large quantities of petroleum still exist, it is in areas
that are harder to extract from (such as shale oil), thereby leading to

increased costs and likely increased environmental impact.

Despite the clear need to move away from petroleum based transportation fuels,
two factors are limiting. The cost of alternatives tends to be higher than gasoline
or diesel fuel, thus limiting their growth. Also of importance is the “chicken and
egg” conundrum. It is difficult to provide the infrastructure for alternative fuels
before there are consumers, and it is difficult to find consumers before there is an
infrastructure. One solution to this problem is to use an alternative fuel that is
compatible with existing infrastructure. Biofuels such as ethanol are a prime
example of this approach. Ethanol can be blended in small quantities with
gasoline and used in all cars. In addition, “flex fuel” cars have been designed to
run on both gasoline and ethanol at little extra cost, allowing the potential

demand for ethanol to grow before the infrastructure is in place. Because of this

innate advantage, ethanol is the leading renewable source of transportation fuel

in the United States, with 9.2 billion gallons produced in 2008 [5].

 

1O

 

 

 

Billion gallons per year

4 IN

0 _
1980 1984 1988 1992 1996 2000 2004 2008
Year

 

 

Figure 1.2 : United States ethanol production, 1980-2008. [5]

The vast majority of ethanol produced in the United States is from corn starch.
However, there is a limit to the amount of ethanol that can be produced from corn
due to limited farmland and the high demand of that same farmland for feed
purposes. While this limit is a source of debate, one valuable estimate is
approximately 15 billion gallons per year (approximately 10% of US gasoline
demand). This value is the maximum starch based ethanol mandated in the
2007 Energy Independence and Security Act, as shown in Figure 1.3 [6].
Cellulosic ethanol is seen as a long-term solution. By obtaining ethanol from

fibrous material rather than starch, several additional feedstocks become

available for bioenergy, including agricultural wastes, municipal solid waste,
forests and wood based residues, and grasslands. Making cellulosic ethanol

commercially viable has been a major focus of research, both academic and

industrial, over the past 5 years.

 

I Other advanced biofuels

 

El Advanced biodiesel

 

I Cellulosic ethanol

 

25 ‘ Current technology

 

 

 

 

 

Billion gallons per year
N
O

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_ . ... Z“; :7 L "_ . :11}:-
‘P‘ ‘55-; 1'7 'L --'L ""31 P: 4'51 ‘1-3 5: ' 5" '
1 0 -:. -‘. E'ux .~‘- :17}. :-'- f'écrs *—~ 5% ‘rl”. 3'5 r— I “
. '1 3' '_~"' "41-" 1‘7 '31:" ' rii ' in}: is"??? 1'3?
. ' .is' 21:: f. " . :f - .‘f‘ -‘f'~. r 5Q '35 '31
: . h . .~ ;.. ..~ . -- " , ... HF": ' :7L w... é a 12“.:
‘- I’ AI.- 1’12? “51'; l‘ v
a! 2.- ~ :3:- zasl E -’
F . a". 3?; :5:
I :2:- 1.-
;4’ - .. ‘ ,_ L;
.441“? . I. “2.. , 2: _.- -'- _ .‘.' . L; __I '_ ‘. . .'. I.‘ ...... ‘ _ q . ‘-
:- 4,” '3. 35-: ,-,-_-..- ::r. "1: -=.-:. r. :: : -,_-, L3; -. .: ‘-
..
I I ’ I I I I I I I I I

 

 

 

 

 

 

 

2006 2008 2010 2012 2014 2016 2018 2020 2022
Year

Figure 1.3 : Renewable fuel standard requirements in the United States as
implemented in the Energy Independence and Security Act of 2007. Current
technology is dominated by corn-based ethanol, yet the mandate limits its
production to 15 billion gallons per year (bgpy). By 2022, the single largest
source of renewable fuel required is cellulosic ethanol (16 bgpy).

While cellulosic ethanol can enhance the amount of biofuels available for use
compared to corn grain, there is no consensus on how much total biofuel energy
is available. Agricultural residues, such as corn stover, rice straw, and wheat

straw, are particularly useful, as they require no additional land for use.

However, low yields per acre, competing uses such as animal bedding, and the

need to leave some material on the farm to prevent erosion and improve soil
quality means these sources are limited. Graham et al. [7] claim that nearly 200
million Mg of corn stover is produced in the United States, but only 58 million Mg
is harvestable for biomass use. Gallagher et al. [8] give a higher number for
collectable stover at 98 million Mg. Accepting this higher number as valid and
assuming 80 gallons per Mg,'com stover alone can only increase the total
ethanol produced in the US by approximately 50%. Other residues can increase
this number, but it is clear that residues alone cannot account for all liquid fuel

demand.

Thus, dedicated energy crops will likely be required. These crops are defined as
those produced solely for bioenergy, thereby eliminating this land from all other
uses. Most dedicated energy crops such as switchgrass, miscanthus, and poplar
are perennial, achieve rapid growth with little water or fertilizer input, achieve
high yields per acre, and require little pesticide or herbicide application. These
crops can often be grown on marginal land unsuitable for traditional crops,
although yields are generally lower than in prime soil. In addition, switchgrass
and other grasses can be grown in land enrolled in the Conservation Reserve
Program, which in 2008 was over 13 million hectares [9]. Again, yields would

likely be lower than in prime farmland.

While there are benefits to large-scale cellulosic ethanol production, concern

regarding a large-scale bioenergy economy remains due to the effect it would

have on current agricultural practices. Walsh et al. [10] studied the effect of
biofuel production on overall agricultural economics. They concluded that if 171
million tons of perennial crops could be produced, traditional crop prices would
rise by 9-14%. These price increases occur naturally due to the increased
demand for land in a bioenergy future. Other researchers believe further impacts
could be caused by indirect land use change. Searchinger et al. [11], for

example, reported a 50% increase in greenhouse gas emissions for cellulosic

 

biofuel compared to gasoline. This is due to land use change; diverting fallow
land to cropland results in the loss of the benefits of these fallow lands. If other
croplands are diverted to biofuels, then it is expected that other fallow lands or
forests would be cleared for food/feed use. If these analyses are correct, then
this further limits the amount of land available to biofuels, particularly if the
benefits of fallow land are perceived as outweighing the benefits of reduced

petroleum consumption.

1.1 Justification

While cellulosic ethanol offers greater variety in terms of potential sources for
biofuels, the question of its viability still remains. Can cellulosic ethanol replace a
significant portion of petroleum use without impacting food production? Due to
the encroachment of cellulosic feedstocks on farmland, would large-scale
production of ethanol be a positive impact on the world? As stated previously,
several analyses claim that large-scale bioenergy production is unsustainable

due to this reason, but such studies are based on the current agricultural

landscape. Rather than accepting this premise, this study envisions a future

where cellulosic feedstocks and food production are compatible with each other.

Such a future would require integrating food production with biofuel production.
This is already performed in the corn ethanol industry, as distiller’s grains, the
byproduct of corn processing, are sold as animal feed, partially offsetting the loss
of farmland for feed purposes. Similar integration may be required for cellulosic
facilities as well. In fact, integrating animal feed operations with cellulosic
ethanol may help to reduce or eliminate several of the potential hurdles for

cellulosic biofuel commercialization.

At 714 g starch per kilogram corn grain [12] and 151 bushels of corn per acre
[13], approximately 6.77 metric tons (Mg) of digestible carbohydrates per hectare
can be produced from com. While dedicated energy crops are a new concept
and their potential is currently unknown, it is expected that they would be able to
produce 10-20 Mg of dry biomass per ha [10; 14]. This relates to the equivalent
of 5.5-11 Mg of carbohydrates per hectare if all cell wall carbohydrates are
available for energy consumption [15]. By the same token, switchgrass
harvested in early summer may have up to 10-15% protein content, although the
biomass yields during spring or summer are approximately 33-50% of the total
harvest [16; 17]. Thus, it is possible to produce 0.5-1.5 Mg of protein per hectare
with switchgrass, which compares favorably with soybean at approximately 1.1

Mg of protein per hectare [13; 15].

Thus, it is conceivable that a hectare of switchgrass could produce as much
animal feed as a hectare of corn or soy while simultaneously providing biofuel.
Thus, land for biofuels can partially displace land for feed with little or no impact
on food production. This increases the amount of land available for cellulosic
ethanol, increasing the potential amount of petroleum that can be displaced. In
addition, these animal feeds would be a large source of revenue for a refinery,
and potentially quite profitable as well. This can help to reduce the economic risk
of early refineries, as there is a second revenue stream, and reduce the impact of
volatile market prices of both inputs (feedstock) and output (ethanol, which would
compete with volatile gasoline prices). Carolan et al. [18] estimate that animal
feed coproducts can reduce the cost of ethanol by approximately 9-20 cents per
gallon ethanol for a fibrous energy feed, while Greene [19] estimates 11 cents

per gallon for a protein feed.

1.2 Project Description

This project seeks to build a foundation upon which animal feed and biofuel
integration can be pursued. As there is currently little interest in such research,
this foundation needs to be built in order to determine which approaches have
the greatest potential. Of primary interest is using dedicated feedstocks such as

switchgrass in mature refineries. Other feedstocks are also tested as needed.

Different options are available to integrate food and fuel production. Two primary
approaches are considered — treating the fibrous matter with ammonia fiber
expansion (AFEX) pretreatment as a feed for ruminants, and separating the
protein from the fiber to displace soybean meal. Different methods and
feedstocks are available for these options, and both must be integrated with

biofuel production.

As stated previously, these options are meant to build a foundation for future
study. Multiple integration options are compared using two different models in
order to assess their value in reducing competition between food and fuel. Thus,
the most promising options can be more fully studied and implemented. The

main objectives and limitations of this research are summarized below.

1.3 Objectives

0 Identify key opportunities for integration of animal feed with cellulosic
ethanol

. Analyze integration options to determine optimal conditions and approach

0 Obtain mass balances around all components for each option on a
consistent basis

0 Compare all options using current models to predict economic and
material impact

0 Identify key knowledge gaps for future research

10

1.4 Limitations

0 Limited to feedstocks available at the time of study. Due to the high
degree of variation among feedstocks, results may change with different
compositions. Sensitivity analyses within the models mitigate this
limitation.

. Limited in scale of experiment. No animal feeding trials were possible. In
vitro rumen digestion experiments were used instead for AFEX-treated
feeds.

. Limited in depth of experimentation. The goal is to identify key areas for
further study, and so not all avenues of research were explored due to
time and material constraints.

0 Limited to current models and assumptions. Due to lack of information,
cost estimates can vary greatly for biofuel or animal feed production.
Sensitivity analyses will mitigate this to some extent.

0 Limited to economic and material outputs rather than environmental

impacts. This is related directly to the justification of the project.

11

CHAPTER 2 : LITERATURE REVIEW

2.1 Ethanol Production

Bioethanol can be produced from various carbohydrates, including simple
sugars, starch, and cellulose. While both simple sugars and starch based
ethanol are currently commercially viable, cellulosic ethanol is still an unproven
technology. There are currently two different approaches being considered for
commercialization: the biochemical and the thermochemical platform [20]. In the
thermochemical platform, the plant material is burned with limited oxygen,
producing syngas (carbon monoxide and hydrogen), which is then converted into
ethanol using either catalysts or microbial fermentation. Alternatively, the
biomass can be burned in the absence of oxygen, called pyrolysis, which
produces a liquid product that can be upgraded to various biofuels. Since the
thermochemical platform uses all materials within the biomass — including
cellulose, lignin, lipids, acids, and protein — there is little opportunity for
coproducing animal feeds. Thus, this research focuses solely on the biochemical
platform. In this approach, the complex carbohydrates are broken down into
component sugars using enzymes, and then these sugars are fermented into

ethanol.

The primary drawback of the biochemical platform is the recalcitrance of cell wall
materials. Cell walls are composed of a dense lignocellulosic structure

containing cellulose nanofibers and hemicellulose linkages, and surrounded by

12

hydrophobic lignin. Because of this dense structure, cellulases and
hemicellulases are unable to effectively break down the carbohydrates, resulting
in poor sugar yields [21]. Thus, a pretreatment step is necessary prior to
enzymatic hydrolysis. This process, generally performed at elevated
temperatures and with either an acid or base catalyst, reduces or eliminates the

barriers to enzymatic hydrolysis and dramatically improves the sugar yields.

A summary of leading pretreatments is shown in Table 2.1. Dilute acid
hydrolyzes hemicellulose, leaving a highly digestible solid residue composed
primarily of cellulose and lignin. However, this process also produces harmful
sugar degradation compounds such as furfural, which can inhibit hydrolysis and
fermentation. Thus, a potentially costly detoxification step is necessary. Hot
water pretreatment reduces the inhibitory compounds formed by controlling the
pH, but does not solubilize as much hemicellulose as dilute acid pretreatment.
Steam explosion reduces the water use and may optionally be performed with
acid catalysts. Due to the low water use, there is no liquid separation after
pretreatment, and so both cellulose and hemicellulose must be hydrolyzed in the
same reactor. Alkaline pretreatments using strong hydroxides or ammonia have
also been used. These pretreatments tend to reduce the inhibitors formed and

sugars degraded, but do not solubilize as much hemicellulose.

Of these, ammonia fiber expansion (AFEX) is the focus of this study. Highly

concentrated ammonia is added to biomass at a moderate temperature and

13

pressure, and allowed to reside for 5-30 minutes. The pressure is then released,
rapidly evaporating ~90% of the ammonia. This process results in partial
solubilization of hemicellulose, removal of lignin to the surface of the biomass,
and partial decrystallization of cellulose [22]. AFEX, like steam explosion, is a
dry-to-dry process, and so all compounds remain in the solid form with no
separate liquid phase. Little sugar degradation occurs relative to acidic
treatments, leading to the potential for high sugar recovery. Various byproducts
are also formed primarily from the reactions between ammonia and various
linkages between the hemicellulose and lignin, primarily acetamide. These
products may inhibit enzymatic hydrolysis and fermentation, or alternatively may
benefit downstream processes by providing a valuable nitrogen source.

Table 2.1 : Five leading pretreatment technologies and their operating conditions

for treating corn stover. All information was obtained from Wyman et al. [23]
except for steam explosion, which was obtained from Bura et al. [24].

 

 

Pretreatment Catalyst Temp (C) L:S ratio Time (min)
DIIUIG ACId H2804 160 4:1 20

Steam Explosion $02 190 Not given 5
Controlled pH None 190 6:1 15

AFEX NH3 90 0.6:1 5

Lime Ca(OH)2 55 5:1 4 weeks

 

AFEX is a strong pretreatment candidate for feed co-products for several
reasons. The moderate temperatures during the AFEX reaction are less likely to
degrade valuable protein. The lack of inhibitory compounds, which may also be
toxic for animal feed purposes, is also an asset. As AFEX is a dry-to-dry
process, only one process stream is present after pretreatment, allowing for a

separate fractionation process dedicated to co-product recovery. Finally, AFEX

14

can be performed at a relatively low capital and operating expense, allowing for

pretreatment and co-product facilities separate from a biorefinery [25].

This latter idea, hereafter named regional biomass processing centers (RBPCs),
is a novel approach to cellulosic refining [18]. These centers would have smaller
capacities than a full-scale biorefinery (between 100-1000 tons per day), and so
several RBPCs would service one refinery. By separating the pretreatment and
co-product formation from the rest of biomass processing, the capital costs per
ton of material of these operations will increase due to lower economies of scale.
However, there are several advantages to RBPCs that may outweigh the

additional cost:

0 Transportation and logistics of biomass collection is streamlined, as the
refinery need not contract with hundreds of farmers and arrange for
transport of material. Instead, the refinery will contract with a manageable
number of RBPCs, who in turn will contract with a smaller number of

farmers.

0 Feed co-products remain in the rural setting, potentially decreasing the
transportation time required. Furthermore, as large feeding operations
may already be centrally located around farms, the RBPC may be co-
located with the animal operation to eliminate the transportation, storage,

and drying requirements.

15

o The risk associated with ethanol production decreases for multiple
reasons. The direct capital cost to the biorefinery decreases, causing a
lower barrier to entry. In addition, RBPCs could be built and begin
producing animal feeds prior to construction of a biorefinery, thus
spreading the overall capital costs required for the entire system over a

longer period of time.

o RBPCs incentivize farmers to produce cellulosic materials by increasing
the potential markets for the biomass. In addition, these RBPCs may be
partly owned by the farmers, thereby retaining more wealth from ethanol

and co-product production in the rural community.

2.2 Food vs Fuel

The primary concern regarding large-scale cellulosic ethanol is the amount of
land required to produce the material. Due to the high use of petroleum fuels in
the United States, over 2 billion Mg of biomass would be required to displace
gasoline use. Few researchers suggest that this amount of material is available.
One prominent study from the USDA and DOE proposed that 1.2 billion Mg of
biomass could be sustainably grown and removed every year within the United
States. This includes corn grain (79 Tg), agricultural residue (388 Tg), forest
resources (334 Tg), and dedicated energy crops (342 Tg), as well as 96 Tg of
other assorted materials such as animal manure [26]. Somewhat optimistic

scenarios were applied to obtain these numbers, including high rates of removal

16

for agricultural residues, large increases in crop productivity, and land use
change. For the land use change, 60 million acres were assumed to be available
for dedicated energy crops, of which 25 million acres would be obtained from
active cropland. Although the study claims that this approach is sustainable, no
mention of the effect on food production is given. Presumably, the increase in
grain and bean yields would offset this loss of land.

Table 2.2 : Literature values for potential biofuel production within the United
States

 

 

Reference EtOHa Notes
(bgpy)
Lynd, 1996 [27] 51 Crop residues, CRP land, mature technology
McLaughlin, 2002 [28] 14 Switchgrass only
Kadam, 2003 [29] 9 Corn stover and wheat straw only
Kim, 2004 [30] 17 Waste crop and crop residues only
Greene, 2004 [19] 140 Land use change, mature technology

USDA/ DOE, 2005 [26] 109 Forest and cropland, land use change
Campbell, 2008 [31] 24 Abandoned crop and pasture land only
Lal, 2008 [32] 65 100% of crop residue, manure, and MSW

a bgpy - billion gallons per year. If total biomass is given in the reference
instead, this value assumes 80 gallons per dry ton of biomass.

 

Most other studies have tended to look at near term possibilities instead, which
provide far less potential bioenergy. A summary of potential ethanol production
from these studies is listed in Table 2.2. In most of these studies, bioenergy can
only produce a small fraction of the transportation fuel currently in use. These
studies do not look at land use change and are thus limited in scope. The USDA
study also included no land use change as a sensitivity as well as little
improvement in yields, and obtained results comparable to the other studies

presented. Those studies that allow for significant land use change show the

17

potential for over 100 billion gallons of ethanol per year, representing over 50%

of current US gasoline demand.

The concern over land use change, however, is significant. Searchinger et al.
[11] proposed that changes in land use would cause biofuels to increase
greenhouse gas emissions relative to gasoline use. This study uses economic
models to predict the effect biofuels would have on the global agricultural market.
In these models, additional land must be converted to agriculture in order to meet
the new demand for food and fiber, and the carbon cost of converting excess
land to cropland offsets the benefits of renewable fuel. In addition to the
environmental concerns, predictions that food costs will rise due to biofuels are
also common. An economic study by Oak Ridge National Laboratory on a
county-by-county basis throughout the United States suggests that food prices
would rise by 9-14% if 55 million Mg of biomass is sold for biofuels [10]. A
considerable expansion beyond 55 Tg could increase prices even further. If
these studies are correct, it is clear that considerable effort must be made to

reduce the impact on land use change if biofuels are to be a large-scale industry.

Current land use is dominated by the production of animal feeds, particularly in
the United States. Table 2.3 displays the major allocation of rangeland and
cropland within the United States. The two primary crops on cropland are com
and soy, which represent 21% and 16% of total US cropland, respectively. Corn

is primarily used as an energy feed due to its high starch content, while soy acts

18

as a protein supplement. These two crops are the primary components of
nonruminant diets, particularly swine and poultry. Ruminants, primarily beef and
dairy cattle, also consume a great deal of fibrous forages in addition to corn and
soy, which are produced on both rangeland and cropland. A large portion of corn
in the United States is devoted to ethanol production as well.

Table 2.3 : Total farmland in the United States and their primary usage. [9] Land

in the Conservation Reserve Program may fall into other categories, particularly
idle land.

 

 

Land MM acres % total Land MM acres % total
Cropland 406.4 44.1 Rangeland 408.8 44.3
Grazing 35.8 3.9 Wood pasture 28.7 3.1
Idle/failed 61.0 6.6 CRP land* 38.5 4.2
Corn 86.2 9.3 Other 78.2 8.5
Soy 63.9 6.9
Wheat 50.9 5.5
Other 108.6 11.8 Total 922.1

 

Thus, in order to reduce the conversion of native grasslands or forests to
croplands in a biofuel economy, it will be necessary to adapt current cropland or
rangeland to biofuel use while maintaining current agricultural output. This
research focuses on increasing land use efficiency, or the amount of feed or

biofuel production produced on an acre of land.

2.3 Ammoniation of Feeds

Forages, defined as feedstufts high in fiber such as hay and silage, are a
common source of energy for ruminant animals such as dairy and beef cattle.
These forages are digested in the rumen, where microbes ferment the fiber into

volatile fatty acids (VFA), namely acetate, propionate, and butyrate. The acids

19

are then absorbed through the wall of the rumen and metabolized by the cattle.
While forages are generally less expensive than grain, ranging from $50-$150
per dry ton depending on the quality, their use has declined in modem livestock
taming due to relatively slow growth and poor utilization of fiber compared to
starch. Despite this, a minimal amount of forage is generally included in the diet.
For example, the NRC recommends at least 25% of a lactating dairy cow’s diet
should be composed of NDF (neutral detergent fiber, or the sum of cellulose,

hemicellulose, and lignin) [33].

The digestibility of forage material is one of its most important qualities, and thus
a major factor in its price. Digestible forages include alfalfa and orchardgrass
hay. These feedstocks tend to have a high leaf to stem ratio, low NDF
concentration, and very little lignin. These characteristics are common for
immature material, although harvesting early in the growing season tends to
lower overall yields. Thus, chemical treatment, particularly alkaline treatment, of
poor digestible forages has been developed to increase the range of acceptable
feeds in livestock production. These treatments, as with AFEX, tend to cleave
ester bonds between cellulose and lignin, allowing for improved rates of digestion

while leaving most of the fiber intact.

Particular interest has been on ammonia or urea treatments, as these nitrogen

containing compounds can increase the crude protein content in the biomass as

well as improving fiber digestibility [34]. In general, anhydrous ammonia is

20

pumped under a tarp containing bales of hay, at a rate of 3-4 g ammonia per 100
g dry biomass and allowed to stand at ambient temperatures for 6-8 weeks. This
process tends to increase NDF digestibility by 100-150 g/kg NDF and crude
protein by 50-100 g/kg biomass, as seen in Table 2.4. Ammoniation also
improves ruminant uptake and helps to eliminate spoilage.

Table 2.4 : Improvements in fiber digestibility (NDF) and crude protein (CP)
content due to ammonia treatment for five types of biomass.

 

 

 

NDF Unt. NDF Treat. CP Unt. CP Treat Ref.

(g/kg NDF) rg/kg NDF) (g/kg BM) (9/_k9 BM)
Wheat Straw 384 509 46 1 10 [35]
Bennudagrass 407 584 66 171 [36]
Cotton 394 489 78 1 60 [37]
Sorghum 528 674 58 106 [38]
Rice Straw 419 540 n/a n/a [34]

 

While ammoniation does have commercial value, it Is limited due to the modest
increase in digestibility as well as the relatively high costs associated with the
process. The ammonia used in this process is not recoverable, requiring 30-40
g/kg forage. In addition, the tarp or plastic covering may also be included in the
costs. Likewise, the increase in digestibility is not likely to make the treated
forages comparable to high quality feeds. Orchardgrass silage, for example, can
have fiber digestibility of 720 g/kg, higher than all treated values shown in Table
2.4 [39]. Thus, ammoniation is used primarily during the winter, when fresh

forages are not available, or in countries such as China where land is limited.

Using more extreme ammonia treatments such as AFEX to further improve

digestibility has been considered but is currently not in practice. One animal

21

feeding trial has been performed using AFEX treated rice straw at 7% of the total
diet in lactating dairy cows [40]. A greater proportion of the total diet was NDF,
with AFEX treated rice straw displacing high protein alfalfa. Here, total dry
matter intake did not significantly change, but milk production increased by 3% (p
= 0.02). Because the level of inclusion of treated rice straw was low, it is difficult
to draw concrete conclusions on the effectiveness of AFEX treatment relative to
the conventional ammonia treatments. Perdok and Leng fed rice straw treated at
elevated temperatures and relatively long residence time (24-72h) at higher
concentrations of total diet, but this led to hyperexcitability in the cows [41]. It is
likely that this is due to the production of 4-methylimidazole from reducing sugars
reacting at elevated temperatures and residence time. Weiss et al., however, did
not see signs of hyperexcitability in sheep when directly fed 4-methylimidazole at
high levels [42]. The exact mechanism of hyperexcitability due to ammonia

treated feeds is currently unknown.

2.4 Leaf Protein Concentrate

The concept of using leaf proteins as a protein source has been investigated
since the 19403 [43]. The amino acid profile of leaf protein is fairly consistent
across a wide variety of species, and compares favorably to other sources of
protein such as corn and soy. Leaf protein concentrates (LPC) also tend to be
highly digestible, and their overall biological value is superior to soy proteins.
The major setbacks to leaf protein concentrates are poor color and flavor and the

possibility of Maillard reactions and lipid oxidation [44].

22

Due to the presence of fiber, which is indigestible for nonruminants such as
swine and poultry, it is necessary to first remove these proteins from the leaf.
One such method is mechanically pressing a protein-rich juice from fresh
forages, while another is to extract the proteins using water or a solvent. This
second option can be performed on either fresh or dried forages, allowing more
versatility, but may require higher downstream costs due to a potentially less

concentrated protein stream.

Investigation into mechanical pressing of leaf proteins has been ongoing for
several decades [44; 45]. In brief, the plant is first macerated before undergoing
a dewatering step, most likely using a screw press. Emphasis is focused on the
percentage of cells ruptured and insuring maximum solubility and filtration of
proteins [46]. Protein from the extracted juice is precipitated and dried, either
through steam injection or acid precipitation. The remaining whey can be either
recycled to the incoming biomass or evaporated and combined with the fiber-rich

press cake.

Attempts have been made to commercialize leaf protein production using
mechanical pressing. One of these attempts was an alfalfa dehydration facility
modified to produce a protein product known as Pro-Xan [47]. This commercial
scale demonstration of leaf protein technology created a 56% protein product,

which accounted for 35% of the protein from alfalfa. An economic analysis of the

23

process indicated that profitability could be obtained depending upon the length
of operation of the facility, which depended heavily upon the growing season of
alfalfa, as well as the value of the deproteinated fiber. Furthermore, it was found
that adding ammonia prior to pressing the fiber improved the quality of the final
protein product, and that recycling a portion of the deproteinated juice improved

overall protein yields.

Commercial leaf protein production was also attempted in New Zealand, which
also used alfalfa as a feedstock [48]. This approach also used mechanical
pressing to release the protein. However, after the first pressing, a disc mill was
used to further disrupt the fiber, and a second pressing was performed. This
resulted in higher protein extraction yields, obtaining up to 70% of the available
protein. However, this production facility ceased operations due to difficulties in
accessing the export market [49]. A third commercial facility, Desialis, is
currently operational In France, and specializes in aquaculture feed and other

specialty protein sources.

In addition to mechanical pressing, solid-liquid extraction has been considered
for the production of LPC. There have been no reported studies of extracting
proteins from switchgrass, although several other types of biomass have been
considered for production of protein concentrates. Dilute solutions of a strong
alkali such as sodium hydroxide are generally used, with the pH between 8 and

12. Extractions generally range from 30 to 60 minutes at 10:1 or higher liquid to

24

solid ratio. Protein yields varied considerably depending upon the types of
biomass, generally resulting in high yields of protein from grains and moderate to
low yields from leaf proteins. Betschart and Kinsella [50] were able to extract
35% of protein from soybean leaves, for example, while Fernandez et al. [51]
obtained 41% from Atriplex leaves. In general, it appears that simple extractions
are not sufficient to obtain complete protein recovery from leafy biomass.

Table 2.5 : Yields of leaf protein concentrates from solid/liquid extraction
obtained from different studies

 

 

Biomass Solvent L/S Ratio Extracted Ref.
Sunflower seed 0.5M KOH 20:1 66% [52]
Rice Bran Water 10:1 12% [53]
Tobacco pH=7 buffer 10:1 1.6%3 [54]
Soybean NaOH pH=8 30:1 71% [55]
Atriplex Leaf pH=10 buffer 5:1 41% [51]
SoLLeaf Tris pH=7.4 13:1 35% [50]

 

3 Given as percent total biomass (protein content of leaves not given)

Few of these studies speculate on the economic attractiveness of this method.
Cost estimates of the commercial Pro-Xan facility indicated that this process was
economical, but the economics depended greatly on the length of the operating
season and the yield obtained. Due to the need to extract material shortly after
harvesting in the mechanical pressing method, such considerations are
extremely important. A diversity of crops, such as cover crops and alfalfa, can
extend the operating season significantly, as can extracting protein from dried
hay. Also of interest is the value of the remaining fiber. The Pro-Xan report
speculates a selling price of $82/ton (in 1977 dollars), compared to $51/ton for

untreated alfalfa [47]. However, much of alfalfa’s value lies in its protein, and so

25

it is uncertain whether this value is correct. Other studies suggest biofuel

production as a viable secondary market [56], yet this too has not been proven.

While leaf protein extraction is a known technology, very little research has been
performed on integrating extraction with fiber processing. De la Rosa [57] and
Urribarri [58] found increases in protein yields from coastal berrnudagrass and
dwarf elephant grass, respectively, when undergoing ammonia pretreatment prior
to extraction. By disrupting the lignocellulosic structure of the biomass, proteins
appear to more easily diffuse out of the biomass and into solution. Over 50% of
the protein in AFEX treated dwarf elephant grass, for example, was extracted
using calcium hydroxide compared to 11% for untreated samples [58]. The feed
value of these protein concentrates, relative to untreated samples, was not tested

in either publication.

26

CHAPTER 3 : ENZYMATIC HYDROLYSIS OF SWITCHGRASS

3.1 Introduction

Switchgrass (Panicum vergatum) is perhaps the most commonly cited dedicated
feedstock for cellulosic ethanol production. As one of the leading feedstocks,
several different treatments have been investigated for its use in pretreating
switchgrass [59]. However, due to the variability in switchgrass type and harvest
dates, these results are not comparable to each other. Few of these publications
give details of harvest time and switchgrass type. Of particular interest is a
previous study using AFEX pretreatment [60]. Here, approximately 90% of the
glucose and 70% of the xylose were released using relatively modest
pretreatment conditions. An optimum was found for both ammonia and water
loadings in terms of maximum sugar yields, while xylose yields continued to
increase with rising temperature. While the harvest date of switchgrass is not

mentioned, it is likely that the material was immature.

If protein extraction for animal feed purposes is to be considered for a biorefinery,
multiple harvest dates must be used. Protein concentration decreases rapidly
between the summer and autumn, with generally less than 3% protein by
October. It is thus unlikely that protein extraction would be practiced on this
mature material. Switchgrass harvested early in the season will also have lower

fiber concentration, thus limiting the potential sugar production. However, early

27

harvest material also has less lignin, and thus the fiber is likely more susceptible

to enzymatic attack.

In addition to the amount of recoverable sugar and protein within the switchgrass,
overall yield per acre also is affected by harvest date. In general, there will be
less switchgrass harvested in the summer compared to one harvest later in the
season, although multiple harvests may produce as much material as a single
harvest [17; 61; 62]. Multiple harvests also remove more minerals and nitrogen
from the field, and thus require more fertilization. These considerations are also
important if protein recovery is included in a biorefinery, and so must be included

in any analysis.

Several different varieties of switchgrass have been created for use as
bioenergy. These varieties can generally be categorized into two classes:
lowland and upland switchgrass. Lowland varieties are more acclimated to
wetter environments, and generally have coarser stems and grow taller than
upland varieties. Different varieties may also react to pretreatment in different

manners as well.

Thus, further research is necessary on the pretreatment and hydrolysis of
switchgrass before any conclusions can be made regarding a protein recovery
system as part of the biorefinery concept. A comparative study between two

harvest dates - July and October — and two switchgrass varieties grown in

28

 

different locations — Cave-in-Rock (upland) grown in East Lansing, MI, and
Alamo (lowland) grown in Auburn, AL — has been undertaken. This study aims to
address the differences between pretreatments at different harvest dates and
locations and the resulting sugar yields. Differences in pretreatment conditions
and enzyme requirements are considered as these factors are important in

evaluating the overall costs of ethanol production.

The objectives of this chapter are as follows:
0 Obtain optimization parameters for AFEX and hydrolysis for the different
Varieties and han/est dates of switchgrass
0 Determine the sugar yields obtainable for early and late harvest
switchgrass
. Analyze the differences in AFEX and hydrolysis conditions between

different varieties and harvest dates of switchgrass

3.2 Materials and Methods

3.2.1 Feedstock

Two varieties of switchgrass were used for this study. Alamo (a lowland variety)
switchgrass, grown at Aubum University (Auburn, AL), was harvested in mid-July
and mid-October of 2005. Cave-in-Rock (CIR) switchgrass (an upland variety)
was grown at Michigan State University (East Lansing, MI) and harvested in early
July and mid-October of 2008. All varieties were milled using a Fitzpatrick JT-6

Homoloid hammer mill (Continental Process Systems, Westmont, IL) to a mesh

29

size of 2mm. Samples were dried to less than 10% moisture and stored at 2°C

unﬁluse.

3.2.2 AFEX Conditions Experiment

For screening AFEX conditions, pretreatment was performed in a 22mL reactor.
Switchgrass was premixed with water at the desired loading and 39 dry weight
was added to the reactor before being sealed shut. Air was removed from the
reactor using a vacuum. Anhydrous ammonia was preheated to a desired
pressure in the ammonia loading vessel, and the biomass preheated to the
desired temperature. Both the ammonia pressure and biomass temperature
were chosen in order to reach a specified temperature once the ammonia was
added to the biomass. The heat of mixing between ammonia and water raises
the temperature beyond the preheated values, and a precise final temperature is
therefore difficult to obtain. Instead, a range of preheated values was used to
obtain a range of temperatures, and the final temperature of the biomass was
recorded and used. Pressure was released at the end of the desired residence
time by turning a ball valve. After the reaction, the biomass was removed and
allowed to dry in a fume hood overnight. Based on previous studies [63], it was

assumed that no net change in the biomass weight occurred during pretreatment.

AFEX conditions ranged from 04-2 g ammonia / g dry biomass, 0.4-2 9 water/ g

dry biomass, and 5-30 minute residence time. In addition, the temperature range

was generally between 80-200°C. At least 45 total AFEX conditions were

30

 

 

chosen for each type of switchgrass tested. The “corner points” of ammonia and

water, representing the maximum and minimum values in the scope of this
experiment, were specifically chosen at a moderate temperature and pressure.
In addition, a near center point (1 g ammonia/g dry biomass, 1 g water/g dry
biomass, 15 minute residence time, moderate temperature) was replicated
multiple times. For the remaining data points, the experimental conditions were

randomly assigned.

Enzymatic hydrolysis was performed at 3% dry biomass loading and 15 mL total
volume. A 0.05 M citrate buffer was used to keep the pH constant at ~5.0.
Tetracycline and cycloheximide were added to prevent microbial and fungal
growth. Accelerase 1000 (Genencor, batch# 1600844643) was used as the
cellulase and loaded at 5mg Accelerase/g dry biomass (equivalent to 3.2 FPU/g
biomass). Hydrolysates were incubated at 50°C and rotated at 200rpm for 72
hours. After the incubation period, enzymes were deactivated by heating
samples to 99°C. Monomeric glucose and xylose concentration was determined
through HPLC using a Bio-Rad (Richmond, CA) Aminex HPX-87P carbohydrate
analysis column. Degassed HPLC water with a flow rate of 0.6 mL/min was used
as the mobile phase, while the temperature in the column was kept constant at

85°C.

A reduced linear model based on the total monomeric glucose and xylose

released during enzymatic hydrolysis was used to analyze the results using

31

~—

 

Minitab 15 as the statistical software package. Ammonia loading, water loading,
residence time, and the final reaction temperature were used as parameters as
well as all interaction effects. Any parameter or interaction term that did not have
a significant effect (p<0.05) was eliminated. The final model was used to analyze
the response of total sugar yield to each pretreatment parameter as well as to

estimate the optimal AFEX conditions for each switchgrass sample.

3.2.3 Enzyme Addition Experiment

For all subsequent experiments in this paper, AFEX was performed at the
conditions determined from the above experiments. These estimated optimal
conditions are listed in Table 3.3. The same treatment method was used except
that AFEX was performed in a 1.5 L reactor rather than a 22 mL reactor.
Between 80-150 g dry switchgrass was used for each batch. The amount of
switchgrass depended on the ammonia loading, as a practical limitation of the
ammonia loading vessel was 160 9. Multiple batches of AFEX treatment were
performed, and no significant differences (p<0.05) were observed in sugar
released through enzymatic hydrolysis between batches. All batches were then

combined before proceeding with further experiments.

Four commercial enzymatic mixtures were used in these experiments:
Accelerase 1000, the B-glucosidase Novozyme 188 (Novozymes,
Batch#058K1144), Multifect Xylanase (Genencor, Batch#4900805391), and

Multifect Pectinase (Genencor, Batch#4010833580). Enzyme loading varied

32

between 5-20 mg/g biomass for Accelerase and 0-10 mg/g biomass for the other
enzyme mixtures. Enzyme loadings were based off previous experiments, in
which more cellobiohydralase (found primarily in Accelerase) is required than
xylanases in corn stover hydrolysis. Likewise, very low yields were obtained
when no cellobiohydrolase was included, which was not the case for the

xylanases [64].

A total of 48 hydrolysis experiments were run for each type of switchgrass,
representing 25 different enzyme combinations determined using the Box-
Behnken method [65]. Hydrolysis was performed in the manner stated above.
Results were analyzed with Minitab 15 using response surface methodology to

determine the importance of each type of enzyme in releasing sugars.

3.2.4 Rate Determination

For the rate experiments, the enzyme loading for each biomass was used as
determined from the previous experiment and listed in Table 3.5. Hydrolysis was
performed in duplicate in 250 mL flasks with a working volume of 100 mL. All
other conditions were as stated previously. Samples of 1 mL were taken at 0, 3,
6, 10, 24, 72, and 168 hours. Cut pipette tips were used in order to sample both
solids and liquid from the flasks, thereby preventing bias in later hydrolysis time

periods due to changing the solid loading.

33

3.2.5 Composition Analysis

Switchgrass cell wall composition was determined based upon the standard
method described by NREL [66]. For total carbohydrate analysis, unextracted
switchgrass was hydrolyzed in 72% sulfuric acid at 30°C for 1h, followed by 1h
hydrolysis in 4% sulfuric acid at 121°C. The resulting hydrolysate was filtered,
and the remaining solids were gravimetrically analyzed to determine acid-
insoluble lignin. Total sugars released within the hydrolysate were analyzed
using a Biorad Aminex 87H column with a constant flow rate of 0.6mL/min using
5mM sulfuric acid and a temperature of 65°C. Ash content was gravimetrically
determined by combusting at 575°C for 16 h. Total extractives were determined
using an accelerated solvent extractor with water followed by ethanol as the
solvent at 1500 psi. A portion of the water extract was analyzed via HPLC for

soluble sugars.

3.3 Results

3. 3. 1 Compositional Analysis

Table 3.1 shows the composition of the four harvests of switchgrass. In general,
later harvest materials contain more glucan, xylan, and lignin than early harvest.
Of particular importance is the lignin, which increases from 16.9% to 22.6%
between the July and October harvest of CIR switchgrass. The amount of lignin
present has been linked to poor hydrolysis yields, particularly as AFEX does not
completely remove lignin. However, increased glucan and xylan content shows

the potential for later harvests to achieve higher overall sugar yields. Both xylan

34

and lignin showed a general trend of increasing with increasing maturity, as the
July harvest of the Michigan CIR switchgrass contained the least amount of both
compounds between all harvests, while the October harvest of CIR contained the
most of each. The CIR switchgrass, grown in Michigan, experienced a shorter
growing season than the Alamo harvests, and thus the July harvest of CIR is less
mature than Alamo, while the October harvest of CIR is more mature than
Alamo, as the northern variety tends to senesce earlier in the year.

Table 3.1 : Composition analysis for the July and October harvests of Cave-in-
Rock (CIR) and Alamo switchgrass

 

 

CIR — July CIR — October Alamo — July Alamo - Oct.
% sema % sem % sem % sem
Glucanb 30.4 1.6 30.5 0.3 29.7 2.2 33.2 0.5
Xylan 15.8 0.7 19.5 0.2 17 1.1 18.2 0.1
Arabinan 2.6 0.2 2.4 0.1 2.2 0.2 2.3 0.1
Ligninc 16.9 0.4 22.6 0.7 18.9 0.2 21.4 0.3
Solublesd 26.0 0.6 15.8 0.1 18.1 0.2 15.0 0.2
Ash 5.0 0.06 3.9 0.02 2.5 0.03 2.0 0.01

 

Closure 96.7 1.9 94.7 0.8 88.4 2.5 92.1 0.6

a Standard error of the mean

9 Includes both glucan and soluble monomeric and oligomeric papers.

0 Acid insoluble (Klason) lignin

9 Includes both water and ethanol solubles

As expected, solubles and ash decreased for later harvests, as the grass
senesces at the end of the growing season and mobilizes those compounds for
storage in the root system. CIR switchgrass also tended to contain more
solubles and ash than the Alamo switchgrass. Overall mass balance closure
ranged from 88% to 97%. Remaining material may include acid-soluble lignin,

galactan and other minor sugars, and insoluble protein. Acid soluble lignin in

particular can range from 2-6% of the total composition [67; 68]. AFEX does not

35

remove any material from the biomass, so this acid soluble lignin may still impact

hydrolysis yields.

3.3.2 AFEX Conditions

The reduced linear model for each of the four switchgrass samples is seen in
Table 3.2. In all cases, the models gave a reasonable approximation of the
results, with R2 values above 75%. Each model showed different main and
interaction effects, indicating that both the harvest date and either the ecotype or
location are important in determining pretreatment parameters. Based on these
models, conditions that obtained maximum sugar yield for each type of
switchgrass are shown in Table 3.2. As the optimal point is often outside the
range of conditions tested, the maximum of each condition is used instead. The
ranges of conditions studied are considered to be the practical limits of AFEX

treatment of forages.

The range of glucose and xylose yields is shown in Figure 3.1. Xylose yields are
within the same range for all harvests of switchgrass, with most samples
between 40 to 100 g/kg switchgrass depending upon the conditions used. The
maximum yields for the October harvests were slightly higher than the July
harvests for xylose. The CIR July harvest had the least variability among xylose
yields, as most AFEX conditions produced xylose yields between 70 and 100
g/kg switchgrass. The amount of glucose released tended to be higher for CIR

switchgrass harvested in July relative to all other harvests, and tended to be

36

lower for the October CIR harvest. Furthermore, different AFEX conditions did
not greatly affect glucose yields for this harvest, with most samples between 120
and 170 g/kg biomass. In comparison, Alamo October harvest had samples
between 130-220 g/kg switchgrass. The July harvest of CIR also behaved
dramatically different in terms of glucose yields between various AFEX
conditions. Both harvests of Alamo switchgrass behaved similarly and showed
similar yields. However, since the October harvest of Alamo switchgrass has
significantly higher glucan and xylan content than the July harvest, the July
harvest has a greater conversion of fiber than the October harvest. Thus, like the
CIR switchgrass, the earlier harvest was more digestible, although not to the

same extent as the CIR material.

Both xylan and glucan hydrolysis were affected in a similar manner for most
AFEX conditions. AFEX conditions that produced high glucose yields also
provided strong xylose yields. This correlation was particularly high in the
October harvests, with a correlation coefficient of 0.88 and 0.94 for CIR and
Alamo harvests, respectively. However, there was more variability in the July
harvests. In particular, two different conditions for the Alamo harvest and one for
the CIR harvest resulted in high glucose yields, but low xylose yields. Both of
these points were at high treatment temperatures, suggesting sugar degradation
or competitive reactions as a possible explanation. This may be general xylan
degradation or the formation of other inhibitors that specifically reduce

hemicellulase activity.

37

Table 3.2 : Reduced linear model for modeling monomeric sugar release with
respect to four AFEX pretreatment parameters

 

Model Parametersa

 

 

CIR — July CIR — Oct Alamo - July Alamo - Oct

Constant 489.6 -376.1 602.6 -177.0
Amb - 88.90 262.4 140.5
Wac -52.95 110.6 -194.0 34.55
Red - 9.237 7.282 12.07
Tee -1.572 6.168 9.086 1.624
Am*Am -27.44 - -40.13 -17.35
Am*Wa - -43.22 - -16.01
Am*Re - - - -1.466
Am*Te 0.6324 - -0.8142 -
Wa*Wa - - - -
Wa*Re - -3.377 - -
Wa*Te 0.5298 - 1 .423 -
Re*Re -0.0612 -0.1 178 - -0.0796
Re*Te 0.0302 - -0.041 7 -0.0356
Te*Te -0.0033 -0.0234 -0.0289 -

R2 80.0% 79.9% 83.6% 94.1 %

 

a Model predicts total monomeric sugars (glucose + xylose) in g/kg biomass.
Terms with a p value less than 0.05 were dropped and the model reanalyzed.
b Ammonia loading measured in g/g biomass

0 Water loading measured in g/g biomass

9 Residence time measured in minutes

9 Temperature measured in C

38

as m Ea 52.92 59 com .008
5 823m 8 not? 962, wcozﬁcoo EmEﬁmEmL
mEmS 9:39am. ESQ San 8:229: comm

.9550. 3.9... exam “a pocketed was 99:93...

a .EoEEmEoE xmu< 050QO m 8.. £3.93: 0

.mcoEncoo EmEﬁmEma 9.3.9 B $63 $0.

.mmeoB >6 9396.603.
.958 £8 :5 an. .5me 9:
:mE>~cm Go 9:0... mm 6cm
>x new 3820 H ﬁn 2:9".

 

 

 

 

 

 

Aconcagua?» 9:3 cocoa—0
com com com 09 09 cm
_ _ _ _ ON
#00 OEm_< I X
22. 88?. x on.
80 m6 . . mx . l o:
x
22. m_o+ + x 30.x . Mix
xx x . m
X le+ o. r 00 \3’
+ m . g x W
+ xXxx - a. w
+ rTT + ++ X X I, m.
+ V“MA... xxvi o o ox . om m.
.1. #+ + x Mm? o. .w.
+ awe x». . i... s . a
I O
+ l I ”MA X I on .83
. II‘
++ +++ . ..Pl ....«. -2:
I 0v“. 0

 

 

 

ON?

39

In general, the October harvests required more severe conditions (higher
temperature, more ammonia, and/or longer residence times) than the July
harvests, as seen in Table 3.3. In particular, the maximum ammonia loading
required for both Cave-in-Rock and Alamo switchgrass was higher for October
harvests than July. For CIR switchgrass, the differences are also pronounced in
terms of temperature and residence time, with the later harvest requiring a higher
temperature and longer residence time. Both harvests of Alamo switchgrass had
similar pretreatment conditions other than ammonia loading. This further
suggests that biomass maturity, which is a function of both harvest date and

location, provides the greatest impact on ethanol production.

Table 3.3 : Optimal AFEX conditions within the parameters tested in this study
CIR — July CIR — Oct Alamo — July Alamo - Oct

 

 

Ammonia (g/g BM) 0.9 2.0 1.6 2.0
Water (g/g BM) 0.4 0.4 2.0 2.0
Res. Time (min) 20 30 30 25
Temperature (°C) 80 130 160 150

 

The CIR July harvest was the switchgrass sample most responsive to
pretreatment and hydrolysis in this study, producing over 350 9 sugar/ kg
biomass at optimal conditions. Here, increasing temperature had an adverse
effect on sugar yields, although high ammonia, water, and residence times
reduced this negative effect. The high temperatures may be degrading soluble
sugars and creating inhibitory products, which are offset by improved digestibility
at more severe conditions. Ammonia requirements were moderate, with yields
peaking at 0.9 g ammonia / g biomass. There was also minimal water

requirement. However, more water was needed at temperatures greater than

40

100°C. At low temperatures, residence time does not play a major role in total
sugars released, but increased residence time greatly improves sugar yields at

high temperature.

For CIR switchgrass harvested in October, few interaction terms were present,
with only water and ammonia as well as water and residence time having
interactive effects. In both cases, the interaction reduced total sugar yields. Due
to the strong positive effect of residence time on sugar yield, increased ammonia
and residence time are offset with decreased water content. [High ammonia
loading and residence time are desired compared to a low water content. There
is no interaction term with temperature. However, optimal sugar yields are

obtained at 130°C, a relatively low temperature for recalcitrant biomass.

The July harvest of Alamo switchgrass, like CIR, was more digestible than the
October harvest, with total sugars released approaching 300 g/kg biomass.

Here, temperature plays a dominant role, with interaction effects with ammonia,
water, and residence time. A negative interaction effect is seen for both
ammonia and residence time interacting with temperature. However, the
coefficients for these negative interactions are relatively small, and so a
temperature of 160°C achieves the highest yields. At this temperature, residence
time does not make a large impact on yields, although the highest residence time
is desired. For both water and ammonia content, the maximum amount gives the

greatest yields at T=160°C.

41

 

The highest correlation was seen with the Alamo switchgrass harvested in
October, as over 90% of the data can be explained by the model. Here,
interaction effects are present for both ammonia/water, ammonia/residence time,
and residence time/temperature. Increasing water tends to increase sugar
yields, but this effect is very small at high ammonia loadings. Likewise,
increasing residence time has a smaller effect at high ammonia loadings and
residence times. At high temperatures and ammonia loadings, a moderate

residence time provides the optimal conditions.

3.3.3 Enzyme Conditions

The response of different enzyme loadings is seen in Table 3.4. The model fit
the data well for three harvests, although the correlation coefficient for the
October Alamo harvest was fairly low (65%). In general, higher enzyme loadings
led to greater sugar production, as expected. However, due to the high costs of
enzymes, high enzyme loadings are unlikely to provide maximum economic
benefit. As the actual cost of enzymes is unknown, the economic optimal
enzyme loading is currently unknown and will likely change with future research
into enzyme combinations and production. Instead, the maximum sugar yields
produced using at most 15 mg enzyme / kg biomass was used to determine
optimal enzyme loadings. The relative amounts of each enzyme that provide

these maximum sugar yields are listed in Table 3.5.

42

Table 3.4 : Reduced linear model for modeling monomeric sugar release with
respect to enzyme addition

 

Model Parametersa

 

 

CIR — July CIR — Oct Alamo — Alamo -
July Oct

Constant 392.69 315.60 344.48 450.49
Acb 3.986 2.606 3.727 -0.634
NoC 6.299 -3.721 -2.990 6.809
Xyd 13.108 20.302 5.013 27.018
Pee 15.435 10.184 4.154 11.199
Ac*Ac - - - -
Ac*No - - - -
Ac*Xy -0.706 - - -1 .458
Ac*Pe - - - -
No*No - - - -
No*Xy - 1.350 1.414 -
No*Pe -0.697 - - -1.369
Xy*Xy - -2.446 - -
Xy*Pe - - - -2.221
Pe*Pe -0.599 -0.561 - -

R2 81 .34% 76.47% 81 .50% 64.87%

 

a Model predicts total monomeric sugars (glucose + xylose) measured in g/kg dry
switchgrass. Terms with a p value less than 0.05 were dropped and the model
reanalyzed.

b Accelerase 1000 loading, measured in mg/g switchgrass

C Novozyme 188, measured in mg/g switchgrass

d Multifect Xylanase, measured in mg/g switchgrass

e Multifect Pectinase, measured in mg/g switchgrass

Despite different conditions providing varying amounts of cellulase and
hemicellulase, there was also a reasonably strong correlation between glucose
and xylose yields for all harvests studied, as seen in Figure 3.2. Correlation
coefficients ranged from 0.65 for the October harvest of Alamo switchgrass to
0.76 for the July harvest of CIR switchgrass. This suggests a degree of synergy
between glucan and xylan hydrolysis regardless of the enzyme used. Given how

closely glucan and xylan polymers are intertwined with each other in the cell wall,

this is not an unexpected result. Furthermore, each of the enzyme complexes

43

contain activities on several different compounds [69], and so the cellulase
mixtures also contain hemicellulase activity and vice versa. Increased
breakdown of cellulose in cellulase-rich enzyme conditions may be increasing
accessibility to the xylan, and vice versa in hemicellulase-rich enzyme conditions,
thus explaining the strong correlation.

Table 3.5 : Optimal enzyme loadings for enzymatic hydrolysis of four harvests of
switchgrass

 

 

 

CIR — July CIR - Oct Alamo — Alamo -

July Oct

Accelerasea 5.0 6.4 5.0 5.0

Novozymeal O 0 0 0

M Xylanasea 5.0 3.6 5.0 5.0

M Pectinasea 5.0 5.0 5.0 5.0

Pred. Sugarsb 523 411 409 557

Act. SugarsC 521 410 410 445

 

aAlI values are in mg enzyme per g dry switchgrass

bMonomeric sugars released (g/kg switchgrass) after 72h of hydrolysis as
predicted by the model in Table 4.

CActual monomeric sugars released (g/kg switchgrass) after 72h of hydrolysis.
The harvest time and cultivar/location also affected the ratio of glucose and
xylose released. The CIR switchgrass harvested in July produced significantly
more glucose than the other three harvests, yet released similar amounts of
xylose. In contrast, the Alamo October switchgrass produced substantially more
xylose than either the CIR October or Alamo July switchgrass, despite a similar

range of glucose released.

44

 

 

.920... an Loam 886.60

$3.5m 5.2, £0.88. EB com new .98 .9550. Bow can 3 accreted mm; $9.66.»: .moEﬁcm .o cozchEoo
2933 6 9:39am. ESQ Emu some new .8 29...? c. 85... 82:28 EoEﬁgma xm“.< 6.53 m... .w 8&6th
9m; 60ch mEmm 6... £53.. Egon Emu __< .3530. oE>Nco 9.3.9 6 £3.63 omo.>x new $820 ”Nd 0.59".

 

 

 

 

 

 

$355325 9:9 3003.0
omm mmm oom mum omm mmm oom mi 05
F H P . . . . OOF
BO 0:52 u
22. oEm_<x
0

~00 E0 0 ++ X r mNF
22. m_o+ . x
X o o .M.
0
WOW? 0 OX 7 omF m
X X 0 I O O W

X 00¢“ O O
+++ mxxxnx . w . A . .xx -. :5 w
+ H + .x x . . x s . . . m.
+ + + xx .x 5 m.
+ I
fawn” 3.4+ x xv.A ..x... . . I - 09.... w
....#+ ++.I...uw - "an. .. In I II- . lav
.4 X I
+ ..—r + I. a I I r mNN
I - ‘II
D

 

 

0mm

45

Sugar yields exceeding SOOg/kg biomass were seen at relatively low (<20 g/kg
biomass) enzyme dosage for the Alamo October and CIR July harvests,
suggesting strong potential as cellulosic feedstocks. Some synergistic effects
appear to be occurring, as glucose yields increase significantly in the presence of
pectinase and xylanase when Accelerase is held constant (data not shown).
Thus, while Accelerase is only a third of the total enzyme loading, both glucose
and xylose are being effectively released. While yields increased for the CIR
October harvest compared to no additional enzymes, the overall conversion
remains low. The Alamo July harvest was also lower than expected. During
AFEX pretreatment for this material, the temperature rapidly rose to the desired
set point, yet decreased to 90-110°C throughout the reaction in all runs. This
large drop in temperature was not seen in the other switchgrass harvests, and
thus the low conversions seen here is likely due to the non-optimal AFEX

conditions.

Harvest date and cultivar/Iocation do not appear to affect the composition of the
enzyme complex required to break down the biomass, despite the different
compositions of the biomass. With the exception of the October harvest of CIR
switchgrass, all switchgrass harvests obtained optimal yields using an equal
blend of Accelerase, Multifect Pectinase, and Multifect Xylanase. For the CIR
October harvest, slightly more pectinase is needed than xylanase. Strong
responses to the hemicellulases suggest that the AFEX pretreatment is not

completely separating the cellulose from the surrounding cell wall material in any

46

 

substrate. Although small amounts of cellobiose were seen in hydrolysates
without Novozyme 188, its presence did not have a positive impact on sugar
yields. Multifect Pectinase also contains strong B-glucosidase activity [69], likely
eliminating the need for Novozyme 188. Increasing Accelerase beyond 5g/kg
produced only modest increases in sugar yields, which are unlikely to be

economically competitive.

3.3.4 Rate of Hydrolysis

All harvests of switchgrass showed a rapid response to enzyme addition, as seen
in Figure 3.3. As expected, both glucose and xylose released during hydrolysis
rose rapidly within the first 24 hours, with a slow release aftenNards.
Interestingly, xylose was released faster than glucose for all samples except the
July harvest of CIR switchgrass. The initial rate (defined as sugar release within
the first 3 hours) was between 35-45 g/kg/h for xylose compared to 25-30 g/kg/h
for glucose. In addition, glucose released after 168 hours was 25% higher than
24 hours for all harvests except for the CIR July harvest (which was 16% higher).
Xylose at 168 hours compared to 24 hours was only 8-14% higher for all
harvests. The xylan appears to be readily accessible to enzymes after AFEX
pretreatment relative to cellulose and responds rapidly to enzymatic attack with

high xylanase addition.

47

 

 

 

    

 

    

 

 

 

 

 

400

5 a:

“300 . c»

E a

8 E

w 3

g 200 8

G) _ a I i a ,,

f, + Alamo July I 5‘; 100 ‘ + Alamo July

g 100 + Alamo October, g + Alamo October

5 |+C|R July I g 50 l +CIR July

0 l + C'R October . 0 + CIR October
0 24 48 72 o 24 48 72

Time (h) Tlme (h)

Figure 3.3 : Rate of hydrolysis. Glucose (left) and xylose (right) released during
enzymatic hydrolysis between 3 and 168 hours of residence time. Pretreatments
were performed at the conditions listed in Table 3.3 and enzyme addition as
listed in Table 3.5. Hydrolysis was performed at 3% solid loading, 50°C, and
200rpm rotation. Results are the average of duplicate samples with error bars
representing the high and low values.

With the exception of the CIR July harvest, the trends for glucose released were
similar in all harvests tested. These three harvests saw a similar initial rate, with
the primary differences appearing between 3-24 hours of hydrolysis. In
comparison, the CIR July harvest showed a very rapid initial release of glucose,
with nearly all glucose released within 24 hours. It is clear that AFEX
pretreatment is very effective in opening up cellulose to enzymatic attack in this
harvest relative to the other three harvests, likely due to its being the most

immature sample. Further research is needed to determine speciﬁcally what

factors inﬂuence this immediate glucose release.

48

The harvest location or type appears to have a greater effect on xylose released
than harvest date. The Alamo harvests showed a significantly higher initial rate
(42-45 g/kg/h) of xylose release than the CIR harvests (34-35 g/kg/h). In

addition, a greater proportion of the total xylose release was seen within the first

24 hours for the Alamo harvests compared to CIR harvests.

3.4 Discussion

1 .

As expected, substantial differences are present between both the harvest date
and either the location or ecotype of switchgrass. These differences are seen in
the response to both pretreatment and hydrolysis conditions. In general, early
harvest switchgrass requires milder pretreatment conditions and lower enzyme
requirements. The ClR switchgrass has greater variability in total sugars
released between the two harvests than the Alamo switchgrass. Optimization of
both pretreatment and enzyme conditions greatly improved both glucose and
xylose yields. All harvests showed a rapid initial response to enzymatic
hydrolysis and, with the exception of the CIR July harvest, acceptable

fermentation of both glucose and xylose.

AFEX pretreatment is a complex process involving several physical and chemical
interactions with the biomass. Thus, it is not surprising that the different
compositions of each type of switchgrass react differently to changing AFEX
conditions. In general, a balance is needed between disrupting the

lignocellulosic structure and reducing or eliminating competing reactions that

49

form degradation products or inhibitory compounds. Early harvest materials
have more soluble sugars, which are more likely to degrade then polymeric
compounds. Furthermore, an important reason for the effectiveness of AFEX
appears to be that large amounts of lignin and hemicellulose oligomers are
dissolved and brought to the surface, thereby creating a network of pores in the
treated biomass (unpublished data). The July harvest of CIR switchgrass
contained the least hemicellulose and lignin, and requiring fewer reactions to

[solubilize and remove this material.

In particular, high temperatures are generally not desired due to the increase in
competing reactions. In all four harvests, the highest temperatures did not give
the highest yields, suggesting that higher temperatures do not provide additional
access to cell wall polymers. High temperatures primarily affect xylose yields,

‘ particularly for the October harvest of Cave-in-Rock switchgrass. Above 160°C,
xylose yields drop from between 65-105 g/kg biomass to 35-65 g/kg biomass,

depending on the other conditions.

The different responses to harvest date and location/ecotype have large
implications for biomass refining. Harvesting early in the season provides lower
costs for pretreatment and higher potential ethanol yields, which may help offset
the costs of a second harvest. With the potential additional revenue from co-
products such as proteins, the benefit for early harvests appears strong.

However, any pretreatment facility will likely be designed to satisfy both harvests,

50

 

. I!

and so reductions in capital costs may not actually occur. In addition, if yields
from the stand decrease over time due to multiple harvests [62], a multiple

harvest scenario may not be an economically viable for agronomic reasons.

Also of concern is the fact that different harvest locations or different types vary
in response to pretreatment and hydrolysis. The Cave-in-Rock switchgrass,
grown in Michigan, begins growing later in the season and senesces earlier than
the Alamo switchgrass grown in Alabama. As such, while the harvest dates were
similar, the relative maturities of the two varieties were different. As the CIR
switchgrass was more digestible in July and less digestible in October than the
Alamo material, the different relative maturities appear to have more effect on
digestibility than the types of cultivars. It remains to be seen if different cultivars
harvested in the same region and same season react differently to pretreatment
and hydrolysis parameters. Thus, harvest practices in the northern latitudes may

be adjusted to avoid the highly recalcitrant late harvest material.

3.5 Conclusions

Ferrnentable sugars are effectively released from switchgrass using AFEX
pretreatment and enzymatic hydrolysis. Relatively mild pretreatment conditions
result in high ethanol yields for early harvest material, while more severe
conditions are necessary for later harvests. A mixture of enzymes, including
xylanase and pectinase, are required to release the greatest amounts of sugar.

Xylan was digested faster than glucan for all types of biomass. The upland

51

Cave-in-Rock switchgrass from Michigan harvested in July showed the greatest
response to pretreatment and hydrolysis, while harvesting in October showed the
least response. All hydrolysates were fennentable, although the July CIR

harvested material showed poor xylose utilization.

Harvest date and location/ecotype have a substantial impact on AFEX
pretreatment conditions and sugar released, although this is not a major factor in
enzyme requirements. Early harvest switchgrass was generally more digestible
and required less severe AFEX conditions than the later harvests. For the
northern CIR switchgrass used in this study, large differences were seen in
pretreatment conditions and response between the two harvests, while this
difference was muted in the southern Alamo switchgrass. The known effect of
maturity suggests that the harvest location is the dominant response, as the
northern climate has a shorter growing season than the southern climate.
Further research is required to separate the effects of harvest location and

ecotype.

These findings have major implications for the possibilities of co-producing
protein from early harvest grasses. It is clear that pretreatment operations must
be flexible to both the needs of early and late harvest material. This requires
flexibility in sizing the equipment, electricity and other operating costs, and
ammonia requirements. The impacts of pretreatment conditions on the

economics of the process are discussed in further detail in Chapter 7.

52

CHAPTER 4 : PROTEIN EXTRACTION

4.1 Introduction

As described in Chapter 2, there are two primary methods to extract protein from
leafy materials: an aqueous extraction using an alkaline solution and mechanical
pressing of wet material. Both options are considered here, although little
research could be performed on the mechanical pressing option. There were two
primary limitations to mechanical pressing research. Since it requires fresh
grass, it could only be performed at the proper time in the growing season. More
importantly, a proper mill and screw press was considered impractical for
laboratory scale studies, limiting the control over the rate and input of the

material. Thus, most of the research was conducted on alkaline extraction.

The interaction of AFEX pretreatment and subsequent hydrolysis with protein
extraction is of particular interest. Due to the presence of ammonia in AFEX
pretreatment, it was chosen as the alkaline medium for protein extraction. If a
bleed stream is required in the ammonia recycle process, adding this bleed
stream to a post-AFEX extraction would be an efficient use of materials. If the
extraction is prior to AFEX, then any residual ammonia remaining on the biomass
would offset the amount required during AFEX pretreatment. The effect of

extraction on sugar yields is also important.

The goals of this chapter are as follows:

0 Optimize alkaline extraction on dry switchgrass using aqueous ammonia

53

0 Investigate options for concentrating the protein following extraction

0 Integrate AFEX pretreatment and subsequent hydrolysis with protein
extraction and determine its impacts

0 Determine key gaps in commercializing alkaline extraction and

concentration relative to traditional methods

4.2 Materials and Methods

4.2.1 Feedstock

The feedstock used in this experiment was Alamo switchgrass obtained from
Auburn University and harvested on May 22, 2005, unless othenNise specified.
The moisture content of the material was approximately 9%. All other grasses
used in this study were obtained from Michigan State University farms and dried
at 50°C prior to use. All material was ground to less than 2 mm prior to
experiments. Composition of the biomass was determined as described in

Chapter 3.

4.2.2 Pretreatment

The AFEX pretreatment was performed in a 300 mL stainless steel pressure
vessel. Water was mixed with the switchgrass to increase the moisture content
to 80% dry weight basis. Glass spheres were added to minimize void space,
thereby reducing the amount of ammonia in the gaseous state. The lid was

bolted shut, and a sample cylinder loaded with 1 (+/-0.04) g NH3 per g dry

biomass, allowing the ammonia to be charged into the vessel. The reactor was

54

heated using a 400W Parr heating mantle, and allowed to stand at 100°C (+/-
1°C) for five minutes. The pressure was explosively released by rapidly turning
the exhaust valve. The treated samples were removed and were placed in a

fume hood overnight to remove residual ammonia.

4.2.3 Hydrolysis

The enzymatic hydrolysis procedure used is described in Chapter 3. The
enzyme loading used for this chapter is 15 FPU Spezyme CP per g glucan and

64 pNPGU Novozyme 188 per g glucan. A solid loading of 10% was used for all

experiments in this chapter.

4.2.4 Protein Extractions

Screening for optimal protein extraction conditions was done using a Dionex
(Sunnyvale, CA) ASE 200 Accelerated Solvent Extractor. Extractions were
performed at 1500 psi, which reduces the required residence time from 30 to 3
minutes. Extractions were done using 11:1 (w/w) liquid/solid ratio and two
separate extractions per sample. For experiments involving varying the pH,
hydrochloric acid was used to reduce the pH. The pH of the solution was
measured after the extraction was complete. Once the optimal extraction
conditions were obtained, all further extractions were performed in flasks for 30
minutes with a 10:1 liquid/solid ratio while continuously stirred. After the

extraction was complete, the solids and liquids were separated using filtration.

55

kwir

“9‘

I

.i
,,
.1
.
t."
9
f.’
3.:
.é
.7-
1.;
”3
'5!
)2
vi
3
H
:9!
4.:
G
‘al
1".
t4,
i
I

4
‘I'ﬁ"

'Ov

.- . . ' "'{ 'IV ,. . . v.7. . ‘i' -7 ‘ _ ' ﬂ _ V"; 1". .1. "‘1-
" ‘ gum "iraMWWﬂWa-nmza .mw. -. ~ . 9'

 

Solid cakes were washed with distilled water at approximately 10:1 liquid/solid

ratio to insure that all soluble material is removed.

Due to the presence of ammonia nitrogen, both during the AFEX pretreatment
and subsequent extractions, it is impossible to use standard nitrogen analysis
methods (the Kjehldahl or Dumas methods) to measure total protein content.
Instead, protein concentration was measured using a Pierce (Rockford, IL)
bicinchoninic acid colorimetric assay kit using bovine serum albumin (BSA) as a
standard. To reduce the effects of interfering agents such as ammonium salts,
lignin components, and glucose, the proteins were first precipitated and
resolubilized [70]. A 100 uL 0.15% sodium deoxycholate was added to 100 uL
protein solution and allowed to sit for 15 minutes. 200 uL of 15% trichloroacetic
acid solution was added, and allowed to sit at 2°C overnight. The mixture was
centrifuged at 13000 RPM for 10 minutes, and the resulting pellet washed with
acetone. The pellet was resolubilized in a buffer solution containing 0.1 M Tris,
2.5M urea, and 4% sodium dodecyl sulfate (SDS). Known concentrations of

protein extracts were used to calibrate the protein recovery of this method.

In order to obtain more accurate numbers for the mass balance shown in Figure
4.6, a full amino acid analysis of the extracts and solid residues was undertaken.
Samples were hydrolyzed in concentrated acid to their component amino acids

and measured using HPLC analysis. All derivitization and HPLC analysis was

56

performed at the Macromolecular Structure Facility in the Department of

Biochemistry at Michigan State University.

In the experiments involving multiple feedstocks and precipitation studies, 0.1 M
sodium hydroxide was used as the extraction medium on untreated samples.
Because no ammonia was used for treatment, nitrogen analysis was used to
analyze the protein removed. A Skalar Primacs SN Total Nitrogen Analyzer was
used to obtain the nitrogen concentration in all samples as described in Chapter

5.

4.2.5 Protein concentration

Ultrafiltration was performed using a Sartorius Sartoflow 200 crossflow
Ultrafiltration system equipped with a Tandem 1081 peristaltic pump. This
system allows can create a constant crossflow rate or a constant transmembrane
pressure (T MP) by varying the speed of the pump. Additional control could be
created by restricting the flow between the pump and the filter. Transmembrane
pressure was measured as the difference between atmospheric pressure (the
permeate) and the average pressure of the inlet and retentate. The permeate
flux was determined gravimetrically as an average over 10 minutes. A 50 cm2,
10 kDa molecular weight cutoff (MWCO), polyethersulfone Pellicon Biomax 10
cartridge was used as the filter for all experiments. The concentration factor was
calculated using the following equation:

MI

CF=—————
MI'MP

(4.1)

57

where M. is the mass of the initial extract and Mp is the mass of the permeate.

Precipitation was performed in 50 mL centrifuge tubes using either 6 M H2804 or

100% ethanol. Acid was added dropwise until the desired pH was obtained with
a tolerance of pH 0.1. Ethanol was added until the final concentration of ethanol
was 70%. After precipitation, samples were centrifuged at 4400 RPM for 10
minutes. The pellet was washed in acetone and recentrifuged before being
allowed to dry. Protein determination was performed using total nitrogen
analysis on the solids while polyphenolic concentration was determined using the

remaining liquid.

4.2.6 Polyphenolics Concentration

Total polyphenolics was measured using the modified Prussian blue method [71].
For each sample, 100 uL 0.016 M potassium ferricyanide and 100 uL 0.02 M iron
chloride was added to 10 (LL extract and 100 uL distilled water and immediately
vortexed. After 15 minutes, 500 uL solution of 17% phosphoric acid and 0.2%
gum arabic is added to stabilize the color. The absorbance was read at 700 nm.
Gallic acid was used as the standard. Because of the potential variety of
polyphenolics present in solutions and the uncertain composition of these
compounds, a precise measurement of polyphenolics is not possible. Thus, all

values are measured and reported as uM of gallic acid equivalent.

58

4.3 Results

4.3.1 Composition Analysis

The composition of the switchgrass used in this study is shown in Table 4.1. Not
included in this composition is minor sugars and acid soluble lignin. The amount
of protein present was lower than reported in literature for other strains of
switchgrass [17]. Switchgrass grown as a biomass energy crop and harvested
early in the growing season would likely have protein contents near 10%, and
thus might be more suitable for integrated protein and sugar processing. The
amount of fiber present is lower than switchgrass harvested at a later date (see
Chapter 3), which seems to suggest lower sugar yields would also result from
using an earlier out. However, early out switchgrass is less recalcitrant than that
harvested in the fall as shown in Chapter 3, and thus the lower cellulose and
hemicellulose content may not be a significant factor. The low amount of lignin is
a promising sign, as this implies less interference with hydrolysis as well as fewer
harmful degradation products that could inhibit sugar production or otherwise be

present in the protein product.

59

Table 4.1 : Composition of switchgrass used in this study. All values given as

g/100g dry matter.

 

 

 

Component % Value
Glucan 29.8
Soluble glucan 3.4
Xylan 16.4
Arabinan 3.5
Protein 7.3
Al Lignina 10.8
Lipids 7.3
Water solubles 21.5
Ash 4.8
Totalb 101.4

 

3 Al - Acid insoluble (klason) lignin
b A portion of the ash and protein may be double-counted in the water solubles.
Acid soluble lignin and minor sugars such as galactan are not counted.

4.3.2 Extraction optimization

Figure 4.1 shows the effect of the temperature of the extraction on the overall

protein and mass yields. Protein yields increased significantly from 25°C to

40°C, but further increases in temperature did not result in major improvements

in protein yield. It is likely that most if not all of the proteins present in the

switchgrass are in their natural state, and so the mild temperatures should not

unfold the proteins or significantly affect their solubility.

6O

 

75 _ I Protein solubilized
I Mass solubilized

 

 

 

 

60

  

 

 

% Solubilized

  

 

 

 

 

 

 

   

25 40 50
Temperature (°C)

Figure 4.1 : Effect of extraction temperature on protein yields. All extractions
were performed in duplicate with 3% ammonia at pH = 10.5. The results are
combined after two separate extractions using 11:1 liquid/solid ratio and 3 minute
residence time under 1500 psi pressure.

The effect of ammonia concentration on extraction yields is seen in Figure 4.2. It
is clear that the presence of hydroxide is necessary to achieve strong yields, as

the yield without ammonia was approximately 65% of those that were performed

in ammonia. Protein yield remains constant from 1-3% NH4+, but then begins to

drop off. This is most likely due to “salting out” the protein, as the increase in salt
concentration decreases the amount of water available to solubilize the protein.

It is also possible that greater hydrolysis of the peptide chains is occurring at
higher salt concentrations, as these smaller peptide chains would then be lost
during the TCA precipitation. However, lower salt concentrations are still more
desirable, due to both the lower cost of solvent as well as the relative ease of
concentration larger protein fragments compared to small peptide chains. There

does not appear to be any salting in effect, likely because 1% salt solution is

61

already a sufficient concentration to solubilize the protein. The total mass

solubilized was unaffected by salt concentration, as expected.

75 AiliProtein saubilEed fr 7
0 ‘ Cl Mass solubilized ‘

% Solubilized

 

0 1 8 10

3 5
Concentration (% NH3)

Figure 4.2 : Effect of ammonia concentration on protein yields. All extractions
were performed in duplicate at 50°C and at pH = 10.5 except for 0%
concentration, which was performed using distilled water at 50°C. The results
are combined after two separate extractions using 11:1 liquid/solid ratio and 3
minute residence time under 1500 psi pressure.

The most significant factor in determining protein yields is the pH of the system,
as seen in Figure 4.3. The amount of protein extracted increased dramatically
from a pH of 8 to 10.5 before leveling off. Similar trends have been seen in other
types of biomass [45; 50; 52]. Most proteins have an acidic isoelectric point, the
pH at which the protein will have no net charge and therefore be the least soluble
in a polar medium. Thus, increasing the pH should increase protein solubility, as
demonstrated here. The most alkaline solution also produced a significant drop

in the total mass solubilized, a potentially useful characteristic. If there is less

biomass in solution, it should be easier to purify the proteins. In addition, the

62

biomass lost during extraction likely includes hemicellulose that could be
hydrolyzed into sugars for ethanol production. Further increases in pH would
require a stronger base than ammonia and might degrade the protein, and thus

were not considered.

 

 

\1
01
l

+ Protein solubilized V

 

—I— Mass Solubilized

 

O)
O
l

 

 

 

% Solubilized
00 A
O 01

 

_L
01

 

 

 

0 w r

7 8 9 pH 10 11 12

Figure 4.3 : Effect of extraction pH on protein yields. All extractions were
performed in duplicate with 3% ammonia and at 50°C. The results were
combined after two separate extractions using 11:1 liquid/solid ratio and 3 minute
residence time under 1500 psi pressure.

As seen in Figure 4.4, attempts were made to improve yields by the addition of
the nonionic surfactant Tween 80, the ionic surfactant SDS, and [3-
mercaptoethanol, a reducing agent. No significant improvements were found by
the addition of either surfactant or reducing agent for the untreated switchgrass.
However, adding B-mercaptoethanol and Tween 80 to AFEX treated grass did
increase protein removal. This would seem to suggest that the AFEX process
affects the proteins in some manner. This effect might be through the creation of

sulfur-sulfur bonds, which would then be cleaved by B-mercaptoethanol, or by

63

proteins unfolding and exposing hydrophobic sites, which can be resolubilized
with surfactants. The total mass solubilized also increased with the addition of
surfactants, most likely due to interactions between the surfactants and

hydrophobic portions of the biomass.

 

40

(A)
O
l

N
O
l

   

_L
O

 
 

liltlo reducing agent“

El Mercaptoethanol 1

 

 

Protein Solubilized (%)

O
. l

None ‘ sos Tw80 None 808 ] Tw80
Untreated AFEX ]
Figure 4.4 : Effect of reducing agents on protein yields for untreated and AFEX
treated samples. All extractions were performed in duplicate with 3% ammonia,
25°C, and at pH = 10.5. The results are combined after two separate extractions
using 11:1 liquid/solid ratio and 30 minute residence time at atmospheric
pressure. Both the ionic sodium dodecyl sulfate (SDS) and the nonionic Tween
80 (Tw80) surfactants were tested, both with and without the addition of [3-
mercaptoethanol.
Thus, optimal extraction conditions for switchgrass are approximately 3%
aqueous ammonia at a pH of 10 and temperature of 40-50°C. These conditions
are in line with those seen for protein extraction of other types of biomass, and

are the conditions used for all subsequent experiments reported here. Total

protein yields are approximately 40%.

64

4.3.3 Integration

To determine whether AFEX pretreatment affects the types of proteins
recovered, the composition of the individual amino acids was determined, as
seen in Figure 4.5. Both the untreated and AFEX treated samples were
extracted at the optimal ammonia conditions without adding surfactant or
reducing agent. Although the amino acid profile for the proteins solubilized
during extraction compared to the total protein from switchgrass is quite different,
there is very little difference between extractions from untreated and AFEX
treated grass. Although AFEX does disrupt the cellular structure of the biomass,
it does not appear to release any other proteins to be available for extraction.
Therefore, it appears that the primary hindrance to protein extraction does not
appear to be the disruption of cell walls, but rather the solubility of the proteins

themselves.

Thus, it appears that AFEX has little impact on the amount of protein removed
from the biomass. However, there may be an impact on both subsequent
processes such as enzymatic hydrolysis and fermentation as well as the quality
of the protein. AFEX acts to remove lignin to the surface of the biomass, and is
also partially broken down into smaller molecular weights. As lignin is partially
soluble in alkaline media, some of these polyphenolic lignin compounds may also
be removed during protein extraction. In addition, these compounds may be
bound to the protein, potentially rendering them indigestible. A Prussian blue

analysis of the extract revealed that polyphenolic compounds are twice as high in

65

the AFEX extract as in untreated switchgrass extract, confirming the potential

problems of this AFEX treatment prior to extraction.

 

 

 

 

0.2 I Untreated
l AF EX
0-15 ‘ a Switchgrass

 

 

I:
1*

£-
‘3'
=5
.
3
.3
.
I

Figure 4.5 : Amino acid profiles for untreated protein extract, AFEX treated
protein extract, and the native switchgrass protein. Asx and Glx are a
combination of aspartic acid and asparagine and glutamic acid and glutamine,
respectively. Cysteine and tryptophan were not detected due to their instability
during acid hydrolysis.

 

 

     

Amino Acid (g/g protein)

.6

17
s “
E
: ~
5 3’
Z 3

Q3

  

 

 

 

 

     

 

 

 

 

g. a g e

JEQQCD

_rlr

$9 2.2
<<2001

lle
Le

Several potential scenarios for integrated sugar and protein recovery were
studied, with the overall sugar released presented in Table 3.2. Extracting
protein prior to AFEX decreased sugar yields significantly, likely due in part to
removing soluble sugars and oligomeric hemicellulose from the biomass.
Removing these compounds may also impact the effectiveness of pretreatment
and may require different pretreatment conditions. However, this was not
studied. An initial extract may also be changing the amount of type of inhibitors
produced during AFEX. If a second extraction is performed after AFEX and prior

to hydrolysis, an additional 40 g/kg glucose is produced during hydrolysis,

66

although the xylose yield remains low. After the proteins are concentrated, the
remaining liquid can be used as a hydrolysate media, thus recovering the soluble
sugars. In a separate experiment, an additional 20 g/kg glucose is seen in the
hydrolysate when the protein extract prior to AFEX is used as the hydrolysate
medium. Extracting protein after AFEX only showed a slight decrease in
hydrolysis yields. Again, this decrease is likely due to removing soluble
carbohydrates, although this is partially offset by removing hydrolysis inhibitors.

Table 4.2 : Amount of monomeric glucose, monomeric xylose, and protein
solubilized during each operation in potential processes

 

Amount released in each step (g/kg dry biomass)
Extraction

 

 

 

 

 

Processa Component Hydrolysate Extraction 1 2b Total
GlucoseC 165 1 8 165
EAH XyloseC 82.7 1 6.1 82.7
Protein 19.4 1 6.2 22.4 1 3.7 n/a 41.8
Glucose 223 1 8 223
AEH Xylose 104 1 3 104
Protein 22.9 1 6.5 20.2 1 5.3 n/a 43.1
Glucose 254 1 10 254
AHE Xylose 113 1 6 113
Protein 33.9 1 4.0 18.9 1 3.8 n/a 42.8
Glucose 208 1 20 208
EAEH Xylose 79.7 1 7.8 79.7
Protein 24.1 1 0.6 22.4 1 3.7 10.6 1 1.2 57.1
Glucose 165 1 8 165
EAHE Xylose 82.8 1 6.1 82.8
Protein 19.4 1 6.2 22.4 1 3.7 1.4 1 1.0 43.4

 

3 Sequence of operations performed for each process: A — AFEX treatment, E —
alkaline extraction, H — enzymatic hydrolysis. For example, Process EAH is
protein extraction followed by AFEX treatment and then hydrolysis.

b For process EAEH and EAHE, this column represents the protein recovered in
the second alkaline extraction operation.

C Monomeric glucose and xylose were only measured in the hydrolysate.

67

The overall mass balance for integrated sugar and protein with extraction prior to
hydrolysis is seen in Figure 4.6. Final yields were 240 g glucose, 85.4 g xylose,
and 80.7 g protein per kg dry biomass. Sugar recovery was approximately 74%
of theoretical values, indicating further improvements in sugar recovery can be
made. Approximately 40% of the protein was found in the extract and 60% in the
hydrolysate, demonstrating that protein must be recovered from both streams in
order to be economical. The insoluble biomass was washed after hydrolysis to
insure all soluble components were recovered, and thus this may have acted as
a second extraction to remove any remaining proteins bound to insoluble
portions of the biomass. Total protein yield is approximately 87% of the total,
taking into account both the switchgrass protein and the enzymes used in
hydrolysis. However, very little insoluble protein remained in the biomass, thus
suggesting that the remaining protein was broken down and lost at some point

during the process.

Approximately 40% of the biomass is solubilized during the initial protein
extraction step. It may be possible to utilize this soluble fraction of the biomass
after the proteins have been removed. The protein might be concentrated and
removed through Ultrafiltration or heat precipitation, while the remaining solution

undergoes further processing.

68

1000 g Biomass

t

——800 9 NH AFEX 782 g Ammonia—>

 

l
1000 g Biomass
7309mem

 

 

 

 

47.5 g Ash
v
__9000 9 Water . 31.7 g Protein
3% NH3 EXtraCtm” 32.9 9 Ash
l
597 g Biomass
433ngmm
14.6 9 Ash
v
240 g Glucose
_5400 9 Water . 85.4 g Xylose
15 9 Enzyme Hydrolysrs 49.0 g Protein
9.1 9 Ash
l
171 g Biomass
0.2 g Protein
5.5 9 Ash
v

Figure 4.6 : Mass balance for switchgrass extraction. Water is not included in
the balance, as the biomass was washed and allowed to dry between each step
to insure complete solid/liquid separation. In the hydrolysis stage, the enzyme
entering is included as protein exiting.

Most of the ash was removed from the biomass during the first extraction step. It

is important to remove this ash, as the final insoluble residue would likely be

burned to provide heat and power for the refinery. The ash content in

switchgrass, particularly sodium and potassium, has been shown to cause

problems with slagging in coal/biomass co-firing power plants [72]. The

69

remaining biomass contains only 3% ash, and thus should reduce this risk in
heat and power generation. It remains to be seen if the ash in the extraction step
can be separated and returned to the land. The fact that most of the ash is
removed during one unit operation should help keep the costs of any ash

processing step low, as only one stream needs to be treated.

Approximately 17% of the biomass remains insoluble throughout this process.
There is virtually no protein or ash still present in this residue, which is mostly
composed of unhydrolyzed fiber and insoluble lignin. This material would likely
be burned for heat and power generation in the refinery, thus reducing natural

gas or coal requirements.

The amount of insoluble material remaining is less than that of the previous
scenario, indicating that less heat and power can be produced. Although less
ash is present, there is still a great deal of protein remaining. Thus, due primarily
to the higher protein yields, an extraction prior to hydrolysis is likely to be the best

option despite the slightly lower sugar yields.

4.3.4 Ultra filtration

The first method of protein concentration tested was Ultrafiltration. Here, a 10
kDa membrane was used for protein recovery. A minimum transmembrane
pressure (TMP) of 15 psi is required to effectively concentrate the protein, as

seen in Figure 4.7. Such a pressure differential is not uncommon in Ultrafiltration

70

systems, as a system to recover proteins from cheese whey uses a TMP of 22
psi [73]. The permeate flux was fairly low, however, peaking at 5 g/s/m2 at a
TMP of 30 psi, as seen in Figure 4.8. In comparison, cheese whey UF had a flux
of 14 g/s/m2. A high crossflow rate (50 mUmin) is needed to achieve this value.
This likely speaks to significant fouling of the membrane, as a higher crossflow

rate tends to remove materials deposited on the surface.

 

4o

 

 

 

 

 

 

 

 

 

 

—— Initial TMP = 7
5, — Initial TMP = 15
g 30
LL
C I
.9
E 20
E
m
0
8 10
O
O i k . 1
0:00 0:43 1 :26 2:09 2:52

Time (h)
Figure 4.7 : Concentration profile for Ultrafiltration of 100 mL switchgrass extract

using a 50 cm2 10 kda molecular weight cutoff membrane. Both runs were
performed at a constant 30 mL/min crossflow rate and the transmembrane
pressure (TMP) allowed to increase as necessary.

After 3 hours, no difference is seen in the permeate flux between 30 and 50
mL/min. This suggests that the fouling occurring during filtration is not limited to
the surface of the membrane, as higher flow rates are unable to remove the
fouling material. Deposits within the pores of the membranes are not as easily

removed and decreases the permeate flux beyond the level seen for surface

deposition. The decrease in flux is fairly rapid followed by a steady flux,

71

suggesting that fouling will not be a significant issue in protein recovery.
Washing the membranes in 0.1 M sodium hydroxide was effective at removing all
fouling material, and multiple experiments after washing were all consistent,

suggesting a reasonably long life time for industrial Ultrafiltration membranes.

 

 

 

 

  
 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6 I TMP = 20 psi
<2: 5 — TMP = so psi
3 4
(D
a 3*
a)
{a 2 ' .1
a) 1 7 T
0- i".

0 -

20 30 4O 50
Initial 3 hours

 

Flow rate (mUmin/50 cm/\2)

Figure 4.8 : Permeate flux for Ultrafiltration of switchgrass extract using a 10 kDa
molecular weight cutoff membrane at two different transmembrane pressures
(TMP). Permeate rate is determined based on the average rate for 10 minutes,
either immediately after starting filtration or after 3 hours of running. The
permeate and retentate were recombined with the feed to keep a constant
protein concentration in the feed.

Unfortunately, protein recovery was fairly low, as seen in Figure 4.9. This is
especially true for the protein released in solution following enzymatic hydrolysis.
An SDS-PAGE analysis of the initial solutions and permeate streams confirm that
no protein larger than 10 kDa is present in the permeate for any media, indicating

that the low recovery is not due to larger than expected pores or damage to the

membrane. The primary enzyme mixture used in these experiments (Spezyme

72

CP) contains few proteases [74], although the long residence time (168 hours)
may be sufficient to break down proteins in solution. In addition, the pH of the
hydrolysis is ~5, which is near the isoelectric point for many proteins. In other
words, larger proteins may remain insoluble at these conditions, leaving only the
peptides released when the cell walls were broken down. High protein retention

is seen in the protein recovered after a second extraction.

 

 

 

 

 

Protein Recovery (%)

 

 

 

2 _ 3
Solution

Figure 4.9 : Protein recovery using Ultrafiltration for four separate protein
solutions: 1 — extraction on AFEX treated switchgrass, 2 — hydrolysate, 3 —
extraction on post hydrolysis solid residue, 4 — extraction on untreated
switchgrass.

Interestingly, AFEX improved the recovery of proteins compared to untreated
extracts. This may indicate that protein is being bound to polyphenolics, which
would increase their size and therefore retention. As stated previously, AFEX
increases the amount of polyphenolics that is solubilized during extraction.
During Ultrafiltration, the permeate stream is a light yellow color while the

retentate remains dark, suggesting that these phenolics do not pass through the

membrane. The large amount of protein or peptides passing through the

73

membrane suggests that much of the protein within the switchgrass is already
degraded prior to extraction. The moderate alkaline and temperature during
extraction should not significantly break down the protein, indicating that the
degradation happened between cutting the switchgrass and milling it. If the
protein is degraded due to cutting and drying the biomass on the field, allowing
the grass to release proteases which reduce overall yield, it is not clear whether

aqueous extraction is a viable approach.

4.3.5 Precipitation

Attempts were first made to precipitate the protein through heat coagulation, the
method used in attempts at commercial leaf protein production in both Colorado
and New Zealand. However, very poor precipitation yields were observed. It is
possible that the rate of heat increase was too low or the protein too dilute to
effectively coagulate the protein. Commercial operations used direct steam
injection, a system unavailable for this project, and so the extract had to be

heated in an oven. Instead, acid precipitation using 6 M H2804 was used. In

addition, ethanol precipitation was also considered due to the possibility of

integrating protein production with an ethanol facility.

Acid precipitation at pH = 3.5 was highly successful at concentrating protein, as
seen in Figure 4.10. Between 65-70% of the protein is recoverable using this
protein precipitation technique, greater than values for heat coagulation in the

literature [47; 75]. By contrast, ethanol precipitation only recovered 45% of the

74

protein. Furthermore, non-protein materials were also precipitated in greater
quantities for this product, leading to a less pure protein product. Ethanol
precipitates were only 22% protein, compared to 30-40% protein for acid
precipitation. Farmers have an incentive to feed protein products in as high a
concentration as possible (i.e., g protein / kg protein supplement), as more non-
nutritious material in the feed decreases intake and therefore reduces growth.
Interestingly, the addition of phenylmethanesulphonyl fluoride (PMSF), a
protease inhibitor, appears to increase protein concentration in the precipitates
for acid pretreatment, although this trend is not significant (p=0.10). A slight
increase is also seen in the amount of protein precipitated, although this too is
not significant. The addition of PMSF had no impact on the amount of protein

extracted, as both extracts solubilized 43% of the total protein in the switchgrass.

 

80%

 

 

60% -

40% ~

   
  

     

 

 

I Protein Precipitated (%);

I Protein in Meal (%)
pH 3 1 pH 5 I EtOH
PMSF

20% ~

 

 

 

0% —

 

 

 

 

NoPMSF

Figure 4.10 : Effect of ethanol or acid precipitation at pH 3.5 and 5 on protein
concentration. Untreated switchgrass extracted at SOC using 0.1 M NaOH was
used as the extract in this study. PMSF was added at 0.1% concentration during
the extraction for half of the experiments. Error bars represent the high and low
values of duplicate trials.

75

Acid precipitation does also precipitate some polyphenolics, as seen in Figure
4.11. No change in polyphenolic concentration is seen using acid precipitation at
pH = 5, although as seen in Figure 4.10 above this results in lower precipitation
yields. At pH = 3.5, however, approximately 23% of the phenolics is removed

with the protein for untreated material.

 

 

 

 

 

 

 

40
. o
8
.‘2 35
8 0 ’
g 30 . 9
:‘g . o o
E 25 A :
a O . . .
20 i i
0 2 4 6 8
pH of precipitation

Figure 4.11 : Amount of polyphenolic compounds (in equivalent moles of gallic
acid) remaining in the extract after protein precipitation on untreated switchgrass
extract.

4.4 Discussion
The results presented in this chapter show both the promise and difficulty of
extracting proteins using an aqueous alkaline extraction. Ammonia has been
shown to be an effective solvent for removing proteins from the biomass, thus

’ opening up possibilities of integrating with AFEX pretreatment. Up to 40% of the
protein could be extracted using dilute solutions of ammonia, consistent with

literature values for other species of leaf protein.

76

The largest issue to be considered is determining whether extraction should be
performed before or after AFEX pretreatment. AFEX does not significantly affect
protein extraction yields, but does decrease resulting sugar yields and increases
the polyphenolics removed in the extract. Although further research is needed to
confirm this assumption, it is likely that extraction prior to AFEX is the preferred
option. While there is concern regarding sugar loss, no effort was made to
improve AFEX treatment on extracted switchgrass. Several options are available
here for increasing yields, including AFEX pretreatment conditions, not drying the
biomass between extraction and pretreatment, and using the extract as the
hydrolysate media to recover soluble sugars. Thus, it is likely that sugar yields
when protein is extracted can be increased to levels comparable with

switchgrass in which no extraction is performed.

If extraction is performed after AFEX treatment, the polyphenolic concentration
may be an issue. If the protein is precipitated, only a portion of these compounds
remain with the biomass, yet most appear to remain with the protein if
Ultrafiltration is used. Further studies on the digestibility of these protein products
are needed prior to determining its true impact. One other concern with protein
extraction following AFEX treatment is ammonia reactions with soluble sugars.
Protein extraction is likely to occur on early-harvest grasses due to higher protein
contents, which contain significant quantities of reducing sugars such as sucrose.
These sugars can react with ammonia at elevated temperatures to produce

degradation compounds such as 4-methylimidizole, which has been linked to

77

nervous disorders in cattle. If the post-AFEX fiber is to be used as an animal
feed, extraction prior to AFEX will insure this compound is not formed during

AFEX by removing the soluble sugars.

In addition to a dedicated protein extraction, protein appears to naturally
solubilize after enzymatic hydrolysis of the lignocellulosic material. This suggests
that protein recovery after hydrolysis and fermentation is a feasible approach as
well. If simultaneous saccharification and fermentation is performed, then
removing the protein must be performed after fermentation or distillation. In this
case, there may be concern with the protein being consumed by the fermentation
organism. One method may be to lyse the cells prior to separating the lignin

residue from the fermentation media, allowing the protein to return to solution.

Despite the promise of this approach, the low recovery from Ultrafiltration remains
a concern. Likewise, the dark color of the concentrated protein also suggests
high phenolic content. It appears as though the small protein size associated
with poor Ultrafiltration recovery is due both to the switchgrass itself and the
protein breaking down during enzymatic hydrolysis. Inhibiting proteases during
hydrolysis may improve the latter result, yet protein yields would still be low.
Either smaller molecular weight cutoffs are required for Ultrafiltration or the
switchgrass must be processed in a manner that does not degrade the protein.
Using smaller MWCO membranes would increase the energy requirement of the

process and increase the amount of nonprotein compounds in the protein

78

product, both undesirable results. Further research is clearly required in this

area if Ultrafiltration is to be an alternative to steam injection.

Aqueous alkaline protein extraction would be an alternative to the mechanical
pressing that has been attempted commercially, as discussed in Chapter 2.
These two operations have substantial differences that must be taken into
account if protein concentrates are to be developed commercially. Mechanical
pressing would tend to use less water, leading to reduced downstream costs.
However, the power requirement for presses can remain high. Furthermore,
decisions must be made between different methods of concentration, including
acid precipitation, steam injection, and Ultrafiltration. While attempts were made
to obtain experimental results from pressing the biomass, no results are
presented here. This is due to poor process control with the equipment, as the
feed rate of biomass and pressure were highly variable and uncontrollable with

the lab-scale equipment used.

Despite the promise of leaf protein extraction, the low protein concentration in the
switchgrass used for these experiments may be problematic economically. The
protein content in the switchgrass samples used in this study ranged from 5-7%
of the total dry weight. Assuming protein recovery of 40%, this amounts to only
20-30 kg protein per Mg of switchgrass. Even an optimistic selling price for the
protein product would not recover the cost of the biomass itself, leading to

concerns about the economics of the process. This issue is discussed in more

79

detail in Chapter 7. In addition, alfalfa and other legumes are potential sources
of protein as well. An intriguing possibility is to double crop corn and soy
rotations, harvesting these grasses or legumes in the spring for their protein

product before planting corn or soy. This possibility is considered in Chapter 8.

4.5 Conclusions

These experimental results show that the integrated recovery of sugar and
protein from grasses appears to be a feasible approach to a cellulosic.
biorefinery. Ammonia has been shown to be an effective solvent for removing
proteins from the biomass, thus opening up possibilities of integrating with AFEX
pretreatment or providing a nitrogen source during fermentation. Additional
protein is released during enzymatic hydrolysis; however, sugar yields decrease
if an extraction is performed prior to AFEX pretreatment. Multiple methods of
concentrating the protein were also investigated. Of these, acid precipitation
provides the highest recovery, although precipitation of polyphenolic compounds

is also occurring.

The major obstacles for commercializing aqueous protein extraction appear to be
in concentrating the protein using ultrafiltration following extraction and the
difficulties with integrating the process with AFEX and subsequent hydrolysis and
fermentation. Protein degradation, both prior to extraction and during
lignocellulosic hydrolysis, is a major concern, lowering protein recovery to 45%

following extraction and 30% following hydrolysis. In addition, the tradeoff

8O

between sugar and protein yields for extraction prior to AFEX is a cause for
concern. Although using the protein extract as a hydrolysate media replaces
some sugar loss, the overall result is declining ethanol yields. Finally, the quality
of the protein product, particularly after AFEX treatment, may be an issue. It is
clear that polyphenolic compounds are included in the extract and concentrated
with the protein. It remains to be seen if these compounds interfere with protein

digestibility.

Thus, further research is necessary before alkaline aqueous extraction can be
developed commercially. Particular emphasis should be placed on pretreating
the fiber remaining after protein extraction. Identifying the reasons behind the
poor sugar yields and adapting to these reasons are likely to be vital to the
economics of this process, as seen in Chapter 7. If necessary, other
pretreatments should be pursued, although AFEX is preferable due to the use of
alkaline materials for extraction. The second area of research to be pursued
should be the quality of the protein and the impact of AFEX treatment on this
quality. This may require animal testing, although experiments with pepsin and
trypsin digestibility can be performed as well. Improving ultrafiltration recovery is
also important. However, because of the low protein recovered in the untreated
extract, other concentration techniques, including steam injection, should be

studied as well.

81

CHAPTER 5 : FIBER RUMEN DIGESTIBILITY

5.1 Introduction

As stated in Chapter 2, ammonia-based treatments have been used to enhance
the fiber digestibility of forages for decades. However, the conventional method
of ammoniation - low pressure, room temperature, and a residence time of
several weeks — does not sufficiently break down the structure of highly
indigestible forages, and so even ammonia treated straws are not as attractive as
feedstuffs compared to traditional forages such as alfalfa. A more severe
treatment such as AFEX may be necessary to allow low quality forages to
compete economically. If so, then agricultural residues or dedicated crops that
have higher yields than traditional forages may displace traditional forages and

possibly some corn grain in cattle diets, thereby allowing more land for biofuels.

Pretreating feedstocks with AFEX technology for animal feed purposes also
aligns with the RBPC concept. By removing pretreatment from the biorefinery
and placing it at a smaller scale, it can be closer to animal feed operations and
therefore reducing transportation costs. Furthermore, it allows two potential
revenue streams from the RBPC: pretreated feed and pretreated biomass for

ethanol, thus limiting the risk for these centers.

If AFEX-treated feeds are to be used for ruminant feeding, there are several

important questions that need to be answered. While animal feeding trials are

82

required to fully understand their use in feeding operations, it is currently
impractical to prepare enough material for such trials due to the size of the AFEX
reactors available. Thus, in vitro analysis is used to estimate the true fiber

digestibility. The objectives of this chapter are as follows:

0 Determine which feedstocks are suitable as AFEX-treated animal feed.
This includes traditional energy crops such as switchgrass, traditional
forages such as alfalfa, and agricultural residues traditionally associated
with ammonia treatments. Because research into AFEX-treated animal
feeds is so limited, it is important to expand the range of feedstocks

beyond switchgrass in this study.

0 Determine the effect of pretreatment on both fiber digestion and nitrogen
addition. For fiber digestion, several different factors must be considered.
These include the rate and extent of digestion, the breakdown of fiber due
to pretreatment vs rumen enzymes, and the relative amount of fiber

breakdown.

5.2 Materials and Methods

5.2.1 Feedstocks
Eleven different feedstocks were used during this experiment, and are listed in
Table 5.1. Three of these forages — corn silage, orchardgrass hay, and alfalfa

hay — are defined as traditional forages in this paper, as they are commonly used

83

for ruminant feeding without treating the forage. The remaining materials are
defined as nontraditional forages. Two of these — corn stover harvested in
October 2007 from Michigan State University (East Lansing, MI) and Alamo
switchgrass harvested in October 2005 from Auburn University (Auburn, AL) —
were chosen for further experiments as being representative feedstocks of

agricultural residues and dedicated energy crops, respectively.

5.2.2 Treatments

AFEX pretreatment was performed in a 1.5 L stainless steel reactor vessel in a
manner described in Chapter 3. As the purpose of this paper is to evaluate the
potential for several AFEX-treated forages rather than to compare the forages to
each other, different treatment conditions were used for each forage. These
conditions were chosen to be as close to optimal values as possible as reported
in the literature while accounting for differences in AFEX procedures. These
optimal values provide the greatest disruption of cell wall material, allowing for
maximum yields of sugars after hydrolysis using commercial cellulases. If no
reference was available, conditions were determined based on other forages with
fiber that is expected to be similar in recalcitrance to enzymatic breakdown. In
general, digestible forages required mild AFEX conditions (lower temperatures,
residence times, and ammonia loadings). For these forages, harsher conditions
generally do not significantly improve accessibility to enzymatic attack (see

Chapter 3). For more indigestible material, relatively harsh AFEX conditions are

84

required to effectively break down the cell wall structure. AFEX conditions for

each biomass are shown in Table 5.1.

To compare AFEX with other treatment methods, corn stover and Alamo
switchgrass were ammoniated at room temperature according to Solaiman et al.
[35]. Forages were ammoniated by placing 30 g dry weight of samples into
plastic bags. Concentrated (30%) ammonia was added at 40 g/kg dry matter,
and the total moisture content was adjusted to 300 g/kg dry biomass. Samples
were thoroughly mixed and left sealed for 30 days at room temperature. After 30
days, samples were dried to remove residual ammonia. For the purpose of this

paper, this treatment is defined as conventional ammoniation.

5.2.3 In Vitro Rumen Digestibility

Neutral detergent fiber (NDF) digestibility was based upon the standard method
of Goering and Van Soest [76]. A rumen buffer solution was prepared including
peptone, macrominerals, and microminerals, and 40 mL were added to 0.5 9
corn stover (or other feedstock). The flasks were placed in a 40°C water bath
and 2 mL of a reducing solution containing cysteine, sodium hydroxide, and

sodium sulfide was added. Flasks were placed under C02 and allowed to

reduce prior to inoculating with rumen fluid. Rumen fluid was collected from a
fistulated dairy cow at Michigan State University’s Dairy Farm approximately 2 h
after feeding. The fluid and partially digested fiber from the cow were blended

before filtering the fiber from the fluid, and the fluid was kept under C02 at all

85

times to minimize bacterial death. After removing the fiber, 10 mL of rumen fluid
was injected into each flask. Samples were kept at 40°C for 48 h, although other
time periods were studied. After the desired residence time, 10 mL of neutral

detergent solution were added to each vial to stop fiber digestion.

NDF, both before and after in vitro digestion, was determined as described by
Mertens [77]. Amylase and sodium sulfite were both added to the samples prior
to neutral detergent digestion. Samples were boiled for 1h in neutral detergent
solution before cool being filtered through crucibles. The remaining fiber was
rinsed with water and acetone, and allowed to dry to determine its dry weight.
Samples were then ashed, and the remaining ash subtracted from the weight of

the sample.

5.2.4 Commercial Enzyme Hydrolysis

Fiber was hydrolyzed using a mixture of commercial enzyme cocktails to
simulate cellulosic ethanol production. These experiments used the pretreatment
parameters and hydrolysis conditions from the October harvest of Alamo
switchgrass as described in Chapter 3. As material was limited for both
experiments, not all AFEX conditions originally tested in Chapter 3 were used for

ruminant testing.

5.2.5 Nitrogen Analysis

86

Nitrogen content within the biomass was determined using a Skalar Primacs SN
Total Nitrogen Analyzer (Breda, The Netherlands), which uses the Dumas
method of combusting all nitrogen to NOx [78]. Approximately 100-300 mg of
biomass, depending on the estimated nitrogen content is placed in a crucible and

combusted at 1100°C for 6 minutes. The resulting gas is reduced to N2 and

measured through thermal conductivity. A standard curve using EDTA from 0.5-

10 mg nitrogen was used to calibrate results.

5.2.6 Statistical Analysis

Only differences between treated and untreated samples were considered in this
study, not differences between different forages. Two experiments, the rate of
digestion for corn stover and switchgrass and the impact of varying AFEX
treatment conditions for switchgrass, were performed in duplicate for each
sample or time point. All other nitrogen and NDF values were determined in
triplicate. A two-tailed student t-test was used to determine if differences
between untreated and treated samples, and all statements of significance were

based on a probability level of 0.05.

The rate of digestion for corn stover and late harvest switchgrass was
determined for each treatment using the following first order degradation model
[79]:

NDFzA-e‘k’ +U (5.1)

87

where NDF is the amount of undigested NDF remaining, A is the amount of
digestible NDF in the forage (g/kg dry matter), k is the rate constant, t is time (h), ‘
and U is the amount of indigestible NDF in the forage (g/kg dry matter).

Constants were determined using the Gauss-Newton method to minimize the

sums of squares for the error.

An analysis of variance was performed for the rate of digestion model. A sum of
squares reduction test was used to compare each pair of treatments [80]. In this
case, the comparison is between the full model and a reduced model with the

following constraints:

Ai = AJ- (5.2)
ki = kj (5.3)
u, = u, (5.4)

where i and j are either no treatment, AFEX treatment, or ammonia treatment.

The test statistic is calculated as

(SSRreduced — SSR full )/(df full — dfrea’uced

)
F = (5.5)
)b
( s MSEfull

 

using an F distribution with the numerator degrees of freedom as the difference
in degrees of freedom between the full and reduced models and the denominator
degrees of freedom as the residual degrees of freedom in the full model. In
addition, each pair of parameters was also compared for significance using the
same method. No differences between switchgrass and corn stover were tested.

All statistical analysis was performed using SAS 9.1 software (Cary, NC).

88

Table 5.1 : AFEX treatment conditions for eleven potential forages

 

 

Name NH3 Water Time Temp Referencea

g/g BM g/g BM min °C
Corn silage 1.0 2.0 15 130 Unpublished data
Alfalfa hay 1.0 0.8 15 130
Orchardgrass hay 1.0 0.8 15 130
Rice straw 1.0 0.8 15 140 Balan 2008 [81]
July Cave-in-rock 1 0 0.8 15 130 Alizadeh 2005 [60]
switchgrass
Forage sorghum 2.0 1.2 15 140 Unpublished data
Corn stover 1.0 0.6 15 130 Teymouri 2005 [63]
October Alamo 1 .5 1 .0 30 150
switchgrass
Wheat straw 1 .0 0.8 15 140
Sugarcane 1 .5 1 .0 30 150
bagasse
Miscanthus 2.0 1.5 30 150 Mumen 2007 [82]

 

aWhen available, optimal AFEX conditions were chosen from previous
references, with time and temperature data adjusted for rapid heating of material.

5.3 Results

5.3.1 Characteristics of AFEX-treated forages

Ammoniation darkened the color of all forages, and the color was darker for
AFEX treatment compared to traditional ammoniation for corn stover and Alamo
switchgrass. Nitrogen and NDF values for all samples are shown in Table 5.2.
AFEX treatment decreased NDF concentration in all samples, ranging from 48-
195 g/kg dry matter, with an average of 110 g/kg dry matter. A slight linear trend
was observed between untreated initial NDF and NDF loss for AFEX-treated
samples (R2 = 0.348, p = 0.052). This trend can partially be explained as
forages with a higher NDF concentration have more reactive sites for the AFEX
treatment to disrupt. The different cell wall compositions among the various

forages likely account for these differences as well. Ammoniation of biomass,

89

particularly AFEX treatment, serves primarily to cleave the ester linkages within
lignin and hemicellulose as well as the linkages between lignin and polymeric
carbohydrates [83]. Furthermore, research in enzymatic digestion of AFEX-
treated corn stover has shown that washing the biomass prior to digestion
reduces pentose yields yet increases glucose yields, strongly suggesting that a
portion of the hemicellulose is broken down into oligomeric sugars during
pretreatment [84]. It should be noted that, as washing is not necessary following
AFEX treatment, these oligomeric sugars should still be present in the final

product and therefore available to the cattle for digestible energy.

This NDF loss appears to be higher than what can be expected for conventional
ammoniation. For example, two references for ammoniation of wheat straw give
NDF losses of 63 and 78 g/kg forage, respectively [35; 85], while 82 g/kg was
hydrolyzed during AFEX treatment. In addition, NDF loss for Alamo switchgrass
was significantly (p<0.05) higher for AFEX than conventional ammoniation (136
vs 38 g/kg for switchgrass) in this study, although the greater NDF loss was not
significant for corn stover (89 vs 58 g/kg). This data suggests a greater
disruption of cell wall material for AFEX-treatment compared to conventional
ammoniation, likely due to the higher temperature and ammonia loading during

treatment.

90

.8898 098:8 :8: 888:8

5:85:08 883 9858: 8:8 __< .8898 098: 0:8 098:8 8828: 30.080 8:85:08 0: 8880:
888: XmE< 88 9858: 850 __< .9858: 8818998 88585000

8.098 098:8 08 8.8: 8828: 888.50:
8.098 858::38

 

 

0.0? >00 0.0m 00.0 md 00? Yam 0: es 000 85802.2
v.0? «v.0 v.0: :0 0.0 mi. 0.3 ~00 0.5 08 8808
888000
0.3 NN.0 0.9 00.0 7v mm 0.0 00> 50 E0 288 88:3
0.3 00.0 v.0? N00 0.0 mm 00 SK 0890:0538
oEm_<
50: 00.0 m.mm «0.0 0.0 09 ﬁn 000 f: 05 8908228
98?. 8880
ms 30 0.0.. 5.0 0.0? :00 N? 03 08>on :80
50.. 050 v.5. 5.0 0.0? 00 «.0 tn 0.? 000 888 :80
0.0 0v.0 00> 3.0 ﬁn >0 0.0 :00 v5 3K 90:98 8088:
v.3 0N0 N.mm 3.0 0.: 0: N0 80 F0 00> 8908128
89-5880 >30
0.: 00.0 0.0m 00.0 0.0 0v 0.? 000 0d PK 2,98 83
3: 8.0 toe 9.: 8 mo: ox 8 9: 98 >2 899558
0.NN ~00 N0: 0N0 Now >0 0.0 owe. n. 5 mmv >8: 8:92
mm? 00.0 0.0m 9.0 NE N0 >0 000 0.0? 5v 808.8 :80
9:5 ._2.m_.w 89... ._2.m_.w m.::: 9.50 ._2.m_.m 89:. ._>..w.w m.:::
$8908 >8 9:9 809:2 88:88 >8 9:00 n52

 

8099 098:8 08 888: 8 585:8 809:: 08 n52 ” «.0 8.8...

91

Ammonia treatment increased the nitrogen concentration for all forages, ranging
from 8.9 g/kg dry matter increase for sorghum to 22.0 g/kg increase in alfalfa.
Untreated nontraditional forages all had low nitrogen concentration. With the
ammonia addition, these samples are more comparable to the untreated
traditional forages. Furthermore, nitrogen addition was significantly (p<0.05)
greater for AFEX treatment compared to ammoniation for both switchgrass and
corn stover, with AFEX increasing nitrogen over conventional ammoniation by
4.9 g/kg for switchgrass and 9.1 g/kg for corn stover. Increases in crude protein
for AFEX compared to ammoniation were approximately the same as literature
values for rice straw [86], with 10.8 g/kg nitrogen increase vs 10.5 g/kg cited. For
wheat straw, literature values for ammoniation [35; 85] were slightly higher than
AFEX treatment (9.5 and 11 g/kg, respectively, vs 9.3 g/kg reported here). It is

believed that most of the additional nitrogen is in the form of acetamide [87].

5.3.2 In vitro digestibility of AFEX-treated feeds

AFEX improved the 48 h in vitro NDF digestibility of most of the forages tested,
as seen in Table 5.3. However, AFEX treatment did not increase in vitro
digestibility of NDF for four forages: the three conventional forages and the early
harvest switchgrass. These forages are already highly digestible, and over 40%
of the fiber was digested for the untreated forages. AFEX treatment of Alamo
switchgrass and wheat straw increased rumen digestion by 223 and 197 g NDF/
kg dry matter respectively, the largest increases among all forages. AFEX

treatment increased NDF digestibility of several nontraditional forages to levels

92

that are greater than untreated traditional forages. AFEX-treated corn stover had
the greatest amount of fiber digested (564 ‘g/kg dry matter) of any treated or
untreated forage in this study. AFEX treatment of forage sorghum and
miscanthus improved NDF digestibility, but the overall amount of fiber digested
was still poor compared to other untreated traditional forages. When taking into
account both loss due to treatment and microbial digestion, AFEX increases the

total fiber loss for all forages.

Also of note is the percentage of fiber digested for AFEX treated samples.
Several forages showed an increase of 25-35 percentage points compared to
untreated samples. Nearly 80% of the NDF in AFEX-treated corn stover was
digested, as well as nearly 70% for wheat straw and 60% for Alamo switchgrass.
NDF digestibility of untreated traditional forages ranged from 40-50%, with early
harvest switchgrass being the only highly digestible untreated forage. Thus,
these treated samples have both more digestible NDF per ton of NDF as well as

per ton of dry matter compared to untreated traditional forages.

93

898:-xm.._< 9m 9858: 8:8 __< .9858: 8.8990 88.9880:

8.098 8899: 9o: 998:8 >_98:_:0.m

99s 9858: 8:8 __< .8898 898: 88 88899: 88.28 888:8 60.080 98:908 o: 8888

8.098 8899: 88 8.098 898: c. 888:. 9889

8.098 898:8 88 8888: 882:8 8:98:58

8.098 8899::

889 >8 0x \ n52 0 8 820 00.89898: 5.: E 091.8 89808 9:09:

9: 88 9858: 8.8998 8 xmu< 9 8:8 8899 9898 85 :8: 0:828. .8898 898: 8.. 88899 82 88.5
9858: 9: :. “.02 .59 8 808988: 8 8 89808 n52:

8099 >8 0v. \ n52 0 8 :8>.0 .>_:o 8:808 899 098: 89808 ”.02:

 

 

New 9m 9 Rm ed: 2: 8 mm: 5 mm «2 e S 95:81.2
m2 8 om 8.. mm 4.9 0mm 8 m3 5 88 mm NE @888
888000
9 9m 2 4% 8..” 0% 8m 8 8: 3 NE E m 5 age 692,
9 S 9 EN 8 8m em 8 9 new £89928
8m 9m 9 m9 8m N9 3m 19 mmm 0 Sn 9 st .80 e820
890:9.30
m 2 mm 8.0, ,3 eat. 9- 8... om mmm {9% 58
9 new 0 98 8 :9 8.. mm .9 e 88 5 OR 99.0. 58
8 mm: 9 e9, 1.: 3m 0.9 mm 8 2 RN 0 at 22.98
em 8: : e8 4.: e. 5 8 N- 87 e 9.. 3 New 32. :8.
88:08:30
8 wow om 98 SN 09 84 9. 09 9 Be 4 us. 391... 8p.
5 o: E 9.. .8 Se m. B F e. 2 8m 2 sem 899.920
mm 9 9 8m tn 8.. as. n- 2:- S n9 9 8 >9 £92.
9 a: 2 8m New :8 v.2 mm 98 9 9m 9 wow 8% Eeo
2: .8 2.0.0 as: .8 .8: .90 ass 88 .200 8: .s..m.m use:
0.98. 8588 .93 802 .8 8890 .198. .8890

 

8099 88.8 :9 8.89898: 53 :_ 8 98: we 88 89808 n52 H 0.0 28:.

94

Switchgrass

 

 

   
 

  
  

 

 

 

 

 

 

 

Com Stover

1000 [ . AFEX l 1000 +AFEX . —I
,. l —- _ Ammoniated 1 A — I —Ammontated _
g’ 800 l ——-A-— Untreated 3 800 -+-- Untreated J
9 9 x.
,‘g” 600 — X E 600 -
C ‘ .5
'8 a
5400* L\\\:\ I [E4008
LL ‘ \1 LL
0 -
z 200 \N g 200 (I

l
0 8 T r r . AT . I T _]
0 48 95 144 192 o 48 96 144 192
Tlme (h) Time (h)

Figure 5.1: Effect of conventional ammoniation and AFEX treatment compared
to untreated corn stover (left) and switchgrass (right) on NDF remaining over time
during NDF digestion. Error bars represent the high and low values for duplicate
samples.

AFEX treatment appears to have a greater effect on both the rate and extent of
fiber digestion over both untreated and conventional ammonia-treated samples,
as seen in Figure 5.1. The amount of NDF remaining after 168 hours was more
than twice as high for untreated switchgrass vs AFEX treated (574 g/kg vs 267
g/kg, p < 0.01), and conventional ammoniation (420 g/kg) was slightly higher than
AFEX treated grass (p = 0.06). AFEX treatment of corn stover resulted in even
greater digestion compared to both untreated (145 g/kg vs 365 g/kg, p = 0.02)
and ammoniated (266 g/kg, p = 0.04) samples. A first order degradation model
was determined for each treatment, and the parameters are shown in Table 5.4.

For corn stover, there was no significant difference between ammonia treatment

and no treatment (p=0.10). However, AFEX treatment significantly improved the

95

degradation of NDF compared to either ammonia treatment or untreated
samples. For switchgrass, all three treatments were signiﬁcantly different from

each other.

Table 5.4 : Comparison of the rate of ﬁber removal for corn stover and Alamo

 

 

switchgrass
Parametersa
A SEb k SE C SE
Corn Stover AFEXC 593C 44 0.029C 0.005 136C 37
Ammoniad 512C 48 0.020C 0.005 2460'd 46
Untreatedd 471C 45 0.023C 0.006 339d 41
Alamo

Switchgrass AF EXc 427c 26 0.026C 0.004 257° 23
(late harvest) Ammoniad 405C 29 0.0180.d 0.004 398d 29
Untreatede 31 QC 103 0.008d 0.005 490d 110

 

aParameters obtained using a least squares nonlinear regression on the rate of
NDF removal (Data from Figure 5.1). The regression equation is NDF = Ae'kt +
U, where NDF is the NDF remaining in the biomass (g/kg dry forage), t is time
after inoculation (h), A is the amount of digestible NDF, k is the rate constant,
and U is the amount of indigestible NDF.

t>Approximate Standard Error

CidveDifferent letters denote signiﬁcant differences (p<0.05) among the
parameters and treatments using the sum of squares reduction test. Signiﬁcant
differences are tested only between treatments and not between forages.

The primary impact of AF EX treatment is to decrease the amount of indigestible
NDF. For both corn stover and switchgrass, AFEX has signiﬁcantly less
indigestible ﬁber than untreated samples, and less indigestible ﬁber than
ammonia treated samples for switchgrass. In addition, AFEX treatment

signiﬁcantly improves the rate of digestion for switchgrass compared to no

treatment.

96

O)
01
O

 

 

i v
%: 600 l 9
l
v 550 ~ t
1:
£3
8 500 -
.9
0 y = -0.0042x"2 + 3.0865x + 57.302
:5 450 “ RA2 = 0.8652
2 l p < 0.001
400 T T l l l
150 200 250 300 350
Sugar Released (glkg BM)

Figure 5.2 : Relationship between monomeric sugars released during enzymatic
hydrolysis with commercial cellulases (x-axis) and in vitro NDF digestion at 48h
(y-axis) for multiple AFEX conditions of late harvest switchgrass. The line
represents a quadratic curve obtained from an ordinary least squares regression.
AFEX conditions ranged from 0.4-2.0 9 water/ g dry biomass, 0.4-2.0 g ammonia
/ g dry biomass, 5-30 minute residence time, and 80-150°C.

As previously stated, the effectiveness of AFEX pretreatment depends on the
conditions present during the reaction. Much literature has been published for
optimizing AFEX conditions based upon theoretical ethanol production; these
conditions are likely to be optimal for ruminant feed as well. As expected, there is
a significant correlation between enzymatic digestion using commercial
cellulases and in vitro rumen digestion, as seen in Figure 5.2. AFEX increases
the accessibility of cellulose and hemicellulose to enzymatic attack, which should
affect fiber digestibility using both methods. The amount of NDF digested during
in vitro studies is approximately twice as high as sugar released by commercial

cellulases. This is due partly to different conditions, as a low enzyme loading

was used and without the complete array of enzymes present in rumen microbes,

97

and partly to a difference in analyses, as the sugar analysis does not include

oligomers or solubilized lignin.

5.4 Discussion

Based upon our results, it appears that AFEX pretreatment is an effective
treatment for improving the digestibility of some forages for ruminant feeding.
For in vitro digestibility studies, we see a clear improvement in NDF digestibility
for both corn stover and late harvest switchgrass compared to ammonia
treatment. An increase in nitrogen content was also seen, providing additional

non protein nitrogen and therefore additional value as a feed.

Many types of biomass appear to be viable candidates for AFEX-treated animal
feeds. In general, early-harvest feedstocks and forages commonly used as
feeds are not appropriate for AFEX treatment, as only modest improvements in
digestibility are seen. In addition, NDF digestibility of miscanthus remained low
(18.5% of NDF) after AFEX treatment, and thus further research is required in
order to make it viable for animal feed. Corn stover and late harvest switchgrass
are of particular interest due to being commonly cited sources of cellulosic
ethanol while simultaneously offering large improvements in crude protein and

fiber digestibility [88; 89].

98

5.4.1 Non Protein Nitrogen

The increase in nitrogen due to ammoniation is also important due to the high
protein content of many untreated forages. Balancing protein is an important part
of feeding operations, and so similar crude protein values between untreated
forages and AFEX-treated feedstocks allow for the displacement of these
untreated forages without adding potentially expensive protein supplements. It is
believed that most of the nitrogen addition is in the form of acetamide. Several
studies have reported that acetamide is digestible to rumen microorganisms,
although less so than common non-protein nitrogen supplements such as urea
[90]. Furthermore, improved digestibility of fiber requires greater N uptake in

order to allow for increased microbial production.

While there are advantages to non-protein nitrogen, primarily the price compared
to protein supplements such as soy, there have been concerns regarding non-
protein nitrogen (NPN) addition. Rapid urea intake has caused rumen toxicity
due to spiking ammonia levels within the rumen. However, a proper transition
period to non-protein nitrogen can allow the cattle to have relatively high intake of
urea without toxicity problems. Furthermore, studies strongly suggest that, while
beef and dairy cattle can be raised solely on NPN, the highest milk producers
require true protein for full milk production [91]. However, these studies were
performed with urea as the NPN source. Acetamide likely breaks down at a

slower rate than urea within the rumen, lessening the potential for ammonia

99

toxicity. Also, there remains some crude protein in the biomass, as well as any

protein in other supplements to the cattle diet.

While it remains to be seen whether the NPN addition is sufficient for cattle diets,
it does not seem likely that it will have a detrimental effect. As stated previously,
the increased digestibility of the fiber necessitates increased ammonia use, and
the likely slow breakdown of acetamide should reduce the risk of rumen toxicity.
Furthermore, as seen in chapter 8, it is possible to adjust NPN addition
significantly by changing the conditions present during AFEX pretreatment.
While this will also likely affect digestibility, conditions may be arrived at to

provide optimal nitrogen addition and fiber digestibility improvement.

5.4.2 NDF Digestibility

The greater total NDF in traditional bioenergy feedstocks leads to the potential
for higher total energy per ton of feed. Com stover, for example, contains more
digestible fiber than the total fiber in alfalfa or corn silage. Furthermore, all
potential feedstocks suggested here have more digestible fiber per kg biomass
than any of the three traditional forages tested. Because of the higher amounts
of digestible fiber, the diet requires less tonnage of fibrous material. Thus, the
value of the fiber in AFEX-treated feedstocks should be higher than that of
untreated forages, potentially increasing their selling price. In addition, this
raises the possibility of adding more digestible energy into the diet, improving

weight gain or milk production.

100

The improvement in the percentage of NDF digested, rather than the total NDF
digested, is also worth noting. The rate of NDF digestion directly corresponds to
the rate of passage through the rumen. As NDF breaks down, it becomes less
aerated due to less microbial action. This increases the density of the material,
allowing it to sink in the rumen and be removed. Increased passage through the
rumen allows for increased dietary intake, leading to greater growth rates for beef
cattle or milk production for dairy cattle. While this will increase overall feed
costs, the cost per kg weight gain or per gal milk will decrease, as the energy for

maintenance for the cattle will be the same.

It seems likely that the increased digestibility of AFEX-treated feeds can lead to
greater intake of feed, thereby increasing production. A second possibility may
be to displace a portion of the grain in cattle diets with AFEX feed. Such an
approach has the potential to reduce overall land use, as bioenergy feedstocks
either have greater yields per acre than corn or, for agricultural residues,
increase the productive biomass per acre for grains. Table 5.5 shows the total
digestible nutrients (T DN), net energy available for lactation at 3X maintenance
(NEL), and crude protein (CP) for AFEX treated corn stover and switchgrass vs
corn grain and several common forages. Whereas switchgrass has numbers
comparable to the forages, corn stover is more digestible, although not as
digestible as corn grain. These numbers are based on calculations obtained

from the composition of the materials and fiber digestibility, and thus only an

101

approximation. The true values will also depend on currently unknown factors,
including the digestibility of the ﬁber in vivo, as well as the digestibility of protein
and its binding to lignin or other polyphenolics, as stated in Chapter 4.

Table 5.5 : Total Digestible Nutrients (TDN), Net Energy available for Lactation at

3X maintenance (NEL), and crude protein (CP) content in ﬁve traditional feeds
and two AFEX feeds

 

TDN NEL CP
% DM Mcal/kg % DM

 

Corn graina 88.7 2.01 9.4
Soybean hulls 67.3 1.46 13.9
Corn silagea 68.8 1.45 8.8
Orchardgrass haya 63.1 1.37 18.1
Alfalfa haya 58.9 1.27 20.2
AFEX Corn Stover 75.6 1.74 17.2
AFEX Switchgrass 63 1.48 14.6

 

a Values obtained from NRC 2001 [33]

5.4.3 Displacing Corn Grain

AF EX treated corn stover is highly digestible at approximately 85% of the value
of corn grain measured using either TDN or NEL. While discussion has
previously focused on replacing traditional forages with AFEX treated feeds, it
may be possible to replace grains in the diet with corn stover as well. Several
recent studies suggest that soybean hulls, a ﬁbrous and highly digestible
feedstock, can replace corn or other grains in beef cattle with little detrimental
effect [92-94]. In general, these studies report dry matter intake (DMI) of feed as
well as dry matter digestibility to be similar between corn grain diets and soybean

hull diets, as seen in Table 5.5. Furthermore, soybean hull diets tended to

102

improve the NDF digestibility of other forages within the diet. This may be due to
the easily digestible fiber stimulating the enzymatic response of ruminant
microbes, increasing the overall activity of fiber breakdown. These studies also
suggest that hulls may improve protein and overall energy utilization compared to

corn.

While soybean hulls are more digestible than AFEX treated com stover, these
results suggest corn stover may also be used in a similar manner. AFEX treated
corn stover would need to be fed at greater levels than corn grain, although the
potential for improved digestibility and energy utilization may reduce this need.
This would effectively increase the productivity per acre of corn by nearly 60%,

assuming 70% of the stover can be harvested.

While this may be a solution for beef cattle, replacing all of the corn grain in dairy
cattle is unlikely to be effective. During ruminant digestion, carbohydrates are
converted to volatile fatty acids, which are absorbed by the cattle and converted
for use. While multiple VFAs are produced, fiber digestion is high in acetate
relative to propionate or butyrate. Starch, meanwhile, produces a greater
amount of propionate as well as lactate. These latter two acids are
gluconeogenic, and thus can be used by the cow to produce the sugars in milk.
Thus, some amount of corn grain will remain in dairy diets in order to optimize

milk production.

103

5.5 Conclusions

Ammonia Fiber Expansion (AFEX) pretreatment caused an improvement in
neutral fiber digestibility of multiple feedstocks during in vitro studies. The
greatest improvements were observed for moderately indigestible material not
commonly used as cattle feed. Of particular interest is com stover and late-
harvest switchgrass, which saw 53% and 128% improvement in 48h digestibility
over untreated material and 74% and 70% improvement over ammonia treated
samples, respectively. AFEX treatment improved corn stover’s total digestible
nutrients to a level comparable with highly digestible fiber sources such as
soybean hulls, while improving switchgrass nutrient value to levels comparable to

traditional forages.

Improvements were seen in both the rate and extent of fiber digestion. Although
NDF is lost during the pretreatment, it is expected that much of that fiber is
converted to oligomeric sugars, which can still be of nutritional value for the
ruminants. Crude protein content also increased, with treated samples being

comparable to common ruminant feeds.

This study strongly suggests that treated agricultural residues such as corn
stover or dedicated energy crops such as switchgrass can compete with common
forages such as alfalfa or orchardgrass for ruminant diets. In addition, corn
stover may be competitive with energy crops such as corn grain. Such

competition will depend upon the cost of AFEX pretreatment and the true value

104

of the final feed. Initial information suggests these feeds can be produced at a
cost comparable to traditional forages while potentially giving added benefits

such as improved intake and milk production.

105

CHAPTER 6 : SUMMARY OF SWITCHGRASS TREATMENT MODULES

6.1 Introduction

In the previous chapters, several options were considered for integrating food
and fuel production cellulosic ethanol refineries. While the general advantages
and disadvantages of each option were discussed, they were not comparable on
an established metric. In addition, the aqueous alkaline protein extraction
studied experimentally could not be compared to the mechanical pressing
extractions performed commercially. Such comparisons are required in order to
narrow these options to the ones with the greatest potential for economic

development and maximum productivity.

The technology behind both potential new technologies of feeds as well as
mechanical pressing of protein is fairly well established. Solid/liquid extraction
can be easily modeled and has been included in an AFEX biorefinery model [95].
In addition, extensive literature references to the mechanical pressing of green
juice and subsequent heat coagulation are also available. For AFEX treated
feeds, the only additional processing needed is drying the material. Thus, the
economic viability of each approach can be considered when analyzing each

opﬁon.

These two protein extraction models - aqueous protein extraction and
mechanical pressing — are combined with AFEX pretreatment and

hydrolysis/fennentation within a biorefinery to obtain the relevant costs and

106

revenues. These models can be adapted for given scenarios. For example, the
fiber produced after protein extraction must be dried and the whey evaporated if
the fiber product is to be used for feed, but this operation not necessary if it is
used for ethanol production. Each module —- extraction, AFEX treatment, and the
remaining refinery — can be combined in different ways to produce all possible
combinations of feeds and fuel from one specific source of biomass. For each
scenario, the parameters in the models must be adjusted to reflect the conditions

required.

The objective of this chapter is to introduce the three major components of the
cellulosic refining models — aqueous extraction, mechanical pressing, and AFEX
and subsequent hydrolysis and fermentation — and determine the effect of

individual adjustments for different processing conditions.

6.2 AFEX Pretreatment and Ethanol Production

The AFEX pretreatment module is based on an integrated biorefinery model
produced by NREL and later adapted by Dr. Mark Laser at Dartmouth University.
This model provides material and energy balances for a cellulosic ethanol
biorefinery with AFEX pretreatment using Aspen Plus 2006 software
(AspenTech, Burlington MA). After the simulation is successfully completed, an
economic analysis tool created in Microsoft Excel uses the material, heat, and
work streams to determine the capital and material costs. Details of this model

can be found in Aden et al. [96], Sendich et al. [25], and Laser et al. [95]. A

107

process flow diagram for the major components in the pretreatment block of the

simulation model is seen in Figure 6.1.

 

 

 

 

 

 

 

 

 

 

 

Chilled Water Cooling Water Water
Condenser Condenser
m g g NH3 Vapor
‘—
r:>
Steam Heater NH3
( W Column
' AFEX :=
B Reactor J St “JP t t d
iomass eam re rea e
V Slurry

 

 

Figure 6.1 : Simplified process flow diagram of AFEX pretreatment and ammonia
recovery system.

6.2.1 Full Refinery Analysis

Due to the time consuming nature of the Aspen model, it is impractical to run it
for multiple simulations. Thus, a simplified model of the biorefinery was created
by determining the trends in costs associated with the refinery at different
pretreatment conditions. A general full factorial model was attempted, with three
levels for each factor, as seen in Table 6.1. However, the Aspen model did not
successfully converge at most individual runs at the high value for water loading
and temperature. Thus, these conditions were dropped, and only two levels
were used for water and temperature. Thus, a total of twelve Aspen simulations
were performed with varying ammonia, water, and temperature conditions, and

each simulation was considered at 3 different residence times.

108

Table 6.1 : Levels used within the general full factorial model of AFEX
pretreatment conditions

 

 

Units Low Medium High
Ammonia Loading g/g dry BM 0.5 1.25 2.0
Water Loading g/g dry BM 0.5 1.25 (2.0)81
Temperature °C 80 140 (200)3
Residence Timeb minutes 10 20 30

 

3 Levels dropped from factorial design due to errors associated with the Aspen
model

b Factor dropped from equipment, energy, and material cost analysis due to only
affecting one piece of equipment

For details regarding how varying pretreatment conditions impact equipment,
material, and energy costs, only the data points at the low residence time were
used. The Aspen simulation model is not affected by residence time within the
AFEX pretreatment reactor, although residence time is a variable during
pretreatment. In order to estimate the effect of residence time, the size of the
pretreatment reactor was varied within the accompanying Excel spreadsheet in
order to achieve the same flow rate at different residence times. Thus, it was
assumed that changing the residence time would have no impact on material

streams, energy costs, or other equipment costs. As the reactor accounts for

~10% of the overall equipment costs, this change can be quite significant.

For the purposes of this study, the scale of the biorefinery was assumed to be
850 tons per day. This value is equivalent to 23 million gallons ethanol produced
per year, and was chosen as a near term scenario consistent with the size of

many current commercial cellulosic ethanol projects. The electricity selling prices

109

were assumed to be $0.04/kWh. The feedstock was assumed to be com stover
sold at $40/ton. Overall ethanol yield was assumed to be 78 gal/ton biomass.
Ethanol selling price is assumed to be $1.70/gal. However, all of these variables

are separated out in the final model and can be varied separately.

6.2.2 Effect of Pretreatment Parameters

For AFEX pretreatment conditions, the primary driver of capital costs within a
biorefinery is the ammonia loading, as seen in Table 6.2. The water loading
during pretreatment is the second most important factor, while temperature only
has a mild impact. Overall, changing these three pretreatment conditions can
increase or decrease capital costs by several million dollars, indicating its
importance. For the biological conversion area, the only piece of equipment with
a high variation due to pretreatment conditions is the cooler immediately
following ammonia recovery and prior to the fermentation vessel. However, this
is because the model assumes a constant amount of water added to the
pretreated material, changing the solids loading during hydrolysis and

fermentation. This is an unlikely scenario, and so this variation can be discarded.

Most of the variation in capital costs occurred within the pretreatment and utilities
areas. For the pretreatment area, the amount of ammonia present has the
largest impact on the cost of most pieces of equipment. This is due to the fact
that the ammonia recycle process is the dominant process within the

pretreatment area. Water content also increases the size of the recovery

110

system, as more water is vaporized and included in the recycle stream during
ammonia stripping. Reaction temperature only has a strong impact on the initial
condenser. The two most important pieces of equipment in terms of their share

of cost are the AFEX reactor itself and the ammonia stripping column.

Table 6.2 : Impact of ammonia loading, water loading, and temperature on
equipment cost within the biorefinery model. Only individual equipment with a
coefficient of variation > 5% and average capital cost > $100,000 are shown.

 

 

 

Average C.V.a Impact of factorsb
A W T
Feedstock Handling Area $4,157,610 0.00% 0 0 0
Pretreatment Area $8,845,306 20.7% +++ ++ +
Cooling water condenser $300,959 58.3% + +++ +++
Chilled water condenser $405,930 49.7% +++ - -
AFEX Reactor $5,034,677 19.6% +++ ++ +
Ammonia stripping column $2,204,306 16.7% +++ ++ +
Ammonia Day Tank $613,906 32.8% +++ ++ +
Biological Conversion Area $1,695,628 1.3% + +++ --
Fermentation feed cooler $26,776 23.2% + +++
Product Recovery Area $7,921,171 0.4% + ++ --
Wastewater Treatment Area $17,538,525 0.2% + +++ --
Storage Area $722,567 0.2% + ++ --
Residue Processing Area $19,731,042 0.0% 0 0 0
Utilities Area $3,425,759 15.3% +++ + +
Cooling Tower System $455,953 22.3% + +++ +++
Cooling Water Pump $382,232 22.6% + +++ +++
Chilled Water Package $1,424,040 32.9% +++ - -
Total Cost $64,037,608 3.7% +++ ++ +

 

a Coefficient of Variation

9 Qualitative assessment of the importance of ammonia loading (A), water
loading (W), and Temperature (T) on the costs of individual pieces of equipment
or area. A + represents an increase in price with an increase in level, a -
represents a decrease of price with increase in level, and 0 represents no
change, with the number of + or - representing its relative importance.

111

The remaining variation is in the utilities area, or specifically within the cooling
tower system. This is due to changing the duties on cooling water throughout the
ammonia recovery system as well as pretreatment. Here, the primary impacts
are due to reaction temperature and water loading. More heat is added to the
system at higher temperatures, which leads to greater heat removal in the
recovery system. Likewise, excess water has a high heat capacity, which also
exacerbates the heating duty required. Because the utilities are required for
AFEX treatment, the capital costs in this area must be considered for both the full

refinery and for AFEX-treated feeds.

The impact of AFEX conditions on heat and work streams is shown in Table 6.3.
The most important stream is the steam used for the ammonia stripping column.
As expected, the temperature of the pretreatment reactor is the most important
variable, as less additional heat is required to remove the ammonia at high
temperatures. Interestingly, ammonia loading has only a slight impact, as much
of it is removed at the initial flash. Water loading plays a greater role, due both to
its high heat capacity as well as increasing the amount of ammonia that can
remain soluble in it. The only other steam used during pretreatment is to heat
the ammonia before adding it to the biomass. However, this stream is five orders
of magnitude lower than the steam used to strip ammonia, and so is not
significant to the overall cost of the refinery. The remaining differences in heat
streams are due to the cooling duty around the pretreatment and ammonia

recovery systems. The changes in cooling duty affects the amount of heat

112

removed in the cooling tower system, which impacts costs due to the required

size of the cooling system, as seen in Table 6.2. As stated previously, water

loading and temperature has the greatest effect on cooling duties.

Table 6.3 : Impact of ammonia loading, water loading, and temperature on heat
and electricity streams within the biorefinery model. A residence time of 10
minutes was used for this analysis. Only individual equipment with a coefficient

of variation > 5% are shown

 

 

Average C.V.a Impact of factorsb

Heat streams Mcal/Mg A W T
Heat removed from feed 100.51 53.8% + +++ ---
Chilled water condenser 360.89 64.2% ++ - -
Cooling water condenser 238.17 78.4% + ++ +++
AFEX Reactor heat removal 313.58 115.8% ++ - ---
Cooling Tower System 675.68 28.6% + +++ +++
Steam for ammonia strippingC 1469.82 69.6% + ++ ---
Electricity Streams kWh/Mg A W T
Total Electricity 144.00 25.1% +++ - -
Cooling Tower System 4.37 28.6% + +++ +++
Chilled Water System 53.40 64.2% +++ - -
Makeup water pump 0.02 79.9% - +++ -
Recycle NH3 pump 1.48 41.9% +++ ++ +
Hydrolysate Feed Pump 11.76 9.1% + ++ --
Cooling Water Pump 8.26 28.6% + +++ +++
Water Circulation Pump 1.38 7.9% + ++ +++

 

a Coefficient of Variation

b Qualitative assessment of the importance of ammonia loading (A), water
loading (W), and Temperature (T) on the costs of individual pieces of equipment
or area. A + represents an increase in price with an increase in level, a -
represents a decrease of price with increase in level, and the number of + or —
representing its relative importance.
C Measured as the enthalpy of the steam entering the column

While electricity required for the plant is assumed to be produced from the

insoluble residue, it reduces the amount that can be sold to the grid for profit.

Several pieces of equipment are affected by changing pretreatment conditions.

Temperature and water loadings both have a large role in the cooling tower

113

electricity requirements, for reasons stated previously. The greatest electricity
usage is associated with the chilled water system. This system is required to
condense the recycled ammonia to a liquid, preventing the need for a costly
compression system. As the ammonia exiting the chilled water condenser must
be at a very low temperature, this requires a specialized refrigeration system for
the chilled water as opposed to the cooling tower. This system accounts for
approximately 50% of the total electricity in the refinery at high ammonia
loadings. Because of this, ammonia loading is also the dominant factor for
overall electricity costs.

Table 6.4 : Impact of ammonia loading, water loading, and temperature on
variable operating costs within the biorefinery model.

 

Average C.V.€=l Impact of factorsb

 

cents/gal A W T
Makeup Water 0.49 32.1% ++ + +++
Cooling Tower Chemicals 0.04 28.6% ++ +++ +++
Electricity Credit 0 14.81 11.3% + +
Ammonia 2.92 2.74% - ++ --
Diammonium phosphate 0.35 3.37% - ++ --
Biomass 51 .13 0.04% - ++ --
Enzymes 1 6.06 0.04% - ++ --
Total Operating Costs 60.32 2.8% +++ - +

 

a Coefficient of Variation

'3 Qualitative assessment of the importance of ammonia loading (A), water
loading (W), and Temperature (T) on the costs of individual pieces of equipment
or area. A + represents an increase in price with an increase in level, a -
represents a decrease of price with increase in level, and the number of + or —
representing its relative importance.

C Credit due to selling excess on-site electricity produced to the grid. The
qualitative assessment is based off the magnitude of this credit; i.e., high
ammonia loadings decrease the size of the credit, which increases the total
operating costs.

114

Changes in electricity usage is the primary effect of pretreatment conditions on
the total variable operating costs of the biorefinery, as seen in Table 6.4. In this
model, fixed operating costs such as salaries, etc. are determined as a function
of the capital cost and so are not considered here. Virtually no change is seen in
either the biomass feed or the enzyme requirement, which are the two largest
sources of raw material cost. Together, these components represent over
$70/Mg biomass of operating cost. Makeup ammonia represents the third largest
raw material cost, but pretreatment conditions in this model do not show a large
effect on ammonia costs. This issue is explored in further detail below. Makeup
water, which is affected primarily by the temperature of the AFEX reactor, does
vary significantly, as does the amount of chemicals required in the cooling tower.

However, these two components are small relative to the total operating costs.

Thus, the magnitude of the electricity credit provides the largest impact on
variable operating costs. Pretreatment conditions have virtually no effect on the
amount of insoluble residue remaining after fermentation, and so the changes
seen are due to changes in heat and power requirements as seen in Table 6.3.
Ammonia has the largest impact in total electricity requirements due to the chilled
water system. This is represented in the electricity credit, as an increase in
ammonia loading decreases the amount of electricity available to sell back to the
grid. As the electricity credit offsets approximately 20% of the total variable
operating costs, ammonia loading also has the largest impact on the total

variable operating cost.

115

6.2.3 Makeup Ammonia

As stated in Chapter 5, it is known that some ammonia reacts with the biomass
to produce nitrogen-based side products such as acetamide. The Aspen model
assumes a constant 10.8 g ammonia reacting per kg biomass regardless of
pretreatment conditions. Experiments were carried out to determine the impact
of pretreatment conditions on the total ammonia lost due to reactions during
pretreatment. The pretreated samples of October harvest CIR switchgrass
described in Chapter 3 were used for this analysis. Nitrogen analysis was
performed using a Skalar Primacs SN Total Nitrogen Analyzer as described in
Chapter 5. It was assumed that the difference in mass of the switchgrass before
and after pretreatment was negligible [63]. The amount of ammonia lost due to
competing reactions was therefore calculated based on the difference in nitrogen

on AFEX treated switchgrass against untreated grass.

All four pretreatment parameters impact the nitrogen increase in AFEX treated
biomass, as seen in Table 6.5. In general, increased water content decreases
nitrogen addition, as water can also react with acetyl or other groups, thus
competing with ammonia. Interestingly, at low temperatures (<100°C), greater
ammonia loadings decrease acetamide formation, although the effect is small.

At higher temperatures, ammonia loading has a positive impact on ammonia
reactivity, as expected. Residence time also increased ammonia reactivity at mid

and high levels of water loading. Temperature has the greatest positive impact

116

at high temperatures, as the two largest nitrogen increases were seen when
AFEX was performed at 200°C. Within the range of conditions studied in the

Aspen simulation, makeup ammonia can vary between 15 and 25 g / kg biomass.

Table 6.5 : Reduced linear model for the makeup ammonia required. The MESP
model does not include the cost of makeup ammonia.

Predictora N addition (g/kg) Predictor N addition (glkg)

 

 

 

Constant 44.81 132 R*W 0.49561
R -0.34281 T*T 0.00153
T -0.34917 T*A 0.05169
A -5.00470
W -9.66895
R-Sq 91 .30%

 

a Predictors are R - residence time, T — temperature, A — ammonia loading, W —
water loading. Units are as shown in Table 6.1.

In the pretreatment model, all reacted ammonia is assumed to be in the form of
acetamide. Changing the amount of acetamide produced during AFEX
pretreatment within the model does affect the capital and energy costs of the
biorefinery, although these costs are insignificant compared to the changing cost
of makeup ammonia. For example, decreasing the acetamide formation by 50%
increases the capital costs in the pretreatment area by less than 0.1%. This
increase is due to the slight increase in ammonia that needs to be recovered.
However, the additional processing cost due to more ammonia recycled is an
order of magnitude lower than the savings in makeup ammonia when acetamide
formation is decreased. Further research on the interactions of ammonia and
biomass during AFEX pretreatment is required in order to build an accurate

model of acetamide formation within the Aspen simulation. A reasonable

117

approximation of its impact can be made by accounting for the change in makeup

ammonia costs without considering changes in the capital or energy costs.

6.2.4 Impact of Biorefinery Size

In the Aspen model, the size of the biorefinery does not impact the variable
operating costs nor the overall material balance. As expected, the capital costs
per ton of biomass are affected, as economies of scale allow a larger refinery to
reduce the cost per ton of biomass. Since the fixed operating costs are
determined as a function of capital costs, these too are affected. Thus, for all
future purposes, direct capital costs and fixed operating costs will be considered
simultaneously and labeled as fixed costs. These costs also include all relevant
financial assumptions as well. These fixed costs were plotted against biorefinery
size and fitted using a power regression. When plotted against annual Mg
biomass, the capital cost was determined to follow a power law with exponent -

0.5012.

6.2.5 Simplified Economic Model of the Biorefinery
Using the information above, a simplified economic model of a biorefinery can be
produced. For a consistent basis, the revenue is determined as $/Mg biomass.

A list of equations used in this model is shown below:
P = REtOH + RElec — CFeed _ CEnzyme — CNH 3 - COIher — CFixed (6'1)

-M +Y

glu xyl ' Mxyl) ' 0'3348 (6'2)

REIOH : EtOH '(Yqu

118

 

R =P . ECons+EA°a+ET't (63)
£1“ E,“ +EAW-a-w+EAT-a-I+EWT~w-t .
(TENS-Vine : PEnzyme ' Eloaded (6'4)
CNH3 = NH3'NAdded (6'5)

NCons+NR'r+NT't+N/t'a
CNHX =PNH3' 2 (5-5)
‘ ‘ +NW-w+NRw~r-w+NTr-t' +NTA~t-a
FC(ms+FR-r+FA-a+FRA-r-a 41.5013
2 5 ’

+FAW-a-w+FAT-a-t

Table 6.6 : List of variables used in the biorefinery portion of the direct land use

 

 

yield

119

yield

model

Explanation Unit Explanation Unit

P Profit $/Mg BM P NH3 Ammonia Price $/Mg

REtOH Ethanol $/Mg BM PEleC Electricity Price $/kWh
Revenue

REIeC Electricity $/Mg BM PEnZyme Enzyme Price $/kg
Revenue

CFeed Feed Cost $/Mg BM NAdded Ammonia on BM g/g BM

CEnZyme Enzyme Cost $/Mg BM a Ammonia g/g BM

loading

CNH3 Ammonia Cost $/Mg BM w Water loading g/g BM

COther Other material $/Mg BM r Residence time min
cost

CFixed Fixed Costs $/Mg BM t Temperature C

PEtOH Ethanol price $/gal S Biorefinery size Mg/day

YGlu Glucose kg/Mg ny. Xylose kg/Mg
hydrolysis yield BM hydrolysis yield BM

MG,U Glucose kg/kg Mxyl Xylose kg/kg
fermentation glucose fermentation xylose

A list of variables, their units, and their explanations is shown in Table 6.6.
Constants for eq 6.6 are seen in Table 6.5. Constants for eq 6.3 and 6.7 are
shown in Table 6.7. These equations were obtained by combining the
information obtained in the above sections. In eq 6.2, the extra constant (0.3348)
is the conversion of gallons of ethanol to kg, as the price of ethanol is more
intuitive as gallons than kg. Two equations, (6.5 and 6.5) are given for the cost
of nitrogen. If the amount of nitrogen addition is known, then eq 6.5 should be
used. If it is unknown, it can be estimated with eq 6.6, which uses the study
based on October Cave-in-Rock switchgrass as seen above. The constants
used to calculate fixed costs were determined at 850 short tons per day, which is

771.11 Mg/day.

Table 6.7 : List of constants in the biorefinery portion of the land use model.
Units are such that the final equation will result in $/Mg BM.

 

 

Electricity Fixed Costs
Constant Value Constant Value
ECO”S -0.1833 FCOnS 44.41
E A 0.0260 FR 0.0755
ET 9.63E-5 FA 1.481
EAW 3.66E-3 FRA .0571
E AT -3.00E-5 FRW 0.0343
EWT -8.20E-5 FRT 2.72E-4

F AA -0.3092
FAW 0.0591
F AT 6.25E-3

 

6.2.6 Separated AFEX Economics
For AFEX-treated animal feed, the cost of AFEX must be separated from the rest

of the refinery. Fortunately, as AFEX conditions do not significantly impact any

120

aspect of the reﬁnery other than the utilities, only minor adjustments in the
bioreﬁnery model need to be made. Feedstock handling, pretreatment, chilled
water package, and the cooling tower were included in the capital cost; the
remaining equipment is not required. This remaining equipment was determined
to be a constant $36.52/Mg in ﬁxed costs, and so this amount is eliminated from
the ﬁxed costs determined in eq 6.7. Likewise, the electricity model was
changed to eliminate all electricity produced from the lignin and consumed during

the remaining portion of the reﬁnery.

In addition, additional capital and energy costs are required to dry the AFEX-
treated feed after processing. For the early harvest material, these costs are
accounted for in the protein extraction models shown in Sections 6.3 and 6.4
below. For the late harvest, the cost of the dryer is assumed to be $1 million for
a 1000 Mg/day facility, which is approximately 20% of the combined
dryer/evaporator cost shown in Section 6.3 and 6.4. This reﬂects both the fact
that no evaporator is required as well as the fact that there is less water to
remove in this scenario. Due to the uncertainties in how AFEX treated feed will
be fed to the cattle, no post-processing of the AFEX material is included other
than drying to 15% moisture. However, additional transportation cost is included
to ship the AFEX-treated material back to the farms. This cost is assumed to be

identical to the transport cost to the reﬁnery.

121

6.3 Aqueous Protein Extraction

A separate version of the bioreﬁnery model includes a protein extraction option.
This model has two extractions, one immediately before and immediately after
AFEX pretreatment. However, such a conﬁguration is not likely due to the
information presented in Chapter 4. Thus, the process must be redesigned.
Rather than use the Aspen model directly, a separate process ﬂow diagram for
the proposed setup was created with economic assumptions based on the Aspen

model.

The process ﬂow diagram for pretreatment and protein extraction is shown in
Figure 6.2. A crossﬂow extraction column is used to remove protein at a low
liquid/solid ratio prior to performing AFEX. It is assumed that the remaining ﬁber
exits at 30% solids loading, which is fed into the AFEX reactor as is. Since
ammonia is used during the extraction process, the ammonia fed into the AFEX
reactor is reduced by the amount present in the biomass. The protein is
removed from the extract via ultraﬁltration, and is then dried and sold. A
concentration factor of 30 is used for the ultraﬁltration step, as seen from Chapter
4. The remaining liquid enters a stripping column to remove the residual
ammonia, which is then recycled. The remaining water is used as a diluent for
hydrolysis and fermentation in order to recapture the soluble sugars removed
during extraction. If desired, the fermentation broth can undergo a separate
ﬁltration after distilling the ethanol to recover protein from the fermentation media.

A material balance for this process is shown in Table 6.8.

122

/—_ﬂ

Ammonia— Concentration <1,

 
 
 
    
 
 

  

    
   

 

 

 

 

Water .
p—Ammonia———1
‘ Ammonia
. Extraction Dryer ' ' Hydrolysrs/
Biomass K . Fermentation
Extract ___.__T____._—_Ammonia
Filtration >1KStripperj1 Water
Protein

 

      
 

, P

Filtration neumapress<1

To Distillation Lignin

1 v 1

Figure 6.2 : Process ﬂow diagram for aqueous protein extraction

The primary economic drivers of this operation are the capital cost and the
operating cost of the ultraﬁltration system. For capital costs, the extraction
column, ﬁlter, dryer, and stripping column must be included. In addition, minor
capital costs such as pumps, etc will also be present. No major changes are
made to the AFEX process, although pretreatment conditions must be adapted to
the high water loading. The ammonia dryer is not affected by the extraction
process, although the remaining ammonia recovery system must increase in size
to account for the additional ammonia from the stripping column. Operating

costs for ultraﬁltration modules were assumed to be $0.56/Mg water [97].

123

Table 6.8 : Example material balance for aqueous protein extraction. Stream
labels are shown in Figure 6.2

 

 

 

 

To To
Inlet BM Ext water To AFEX To UF Dryer stripper
Ammonia 0.0 52.0 18.9 33.1 1.1 32.0
Water 110.0 5150.0 1870.0 3280.0 109.0 3171.0
Biomass 1000.0 0.0 872.6 127.4 22.9 104.5
Protein 100.0 0.0 80.9 19.1 11.5 7.6
To To AFEX Hydrolysis/ To
hydrolysis recovery Ferment. To UF 2 Dryer 2
Ammonia 1.6 30.4 1.6 0 0
Water 3105.0 66.0 5129.6 5129.6 331.3
Biomass 104.5 0.0 977.1 n/a 82.8
Protein 7.6 0.0 88.5 103.5 41.4

 

Other assumptions must be made regarding the costs of the system. Heating
costs are relatively low, requiring only steam for the stripping column and the
dryer. Drying costs for distillers grains are 0.06 Mg natural gas/ Mg water
removed, or 3.2 GJ/Mg water [98]. Although lignin would be used the energy
source for the dryer, this would decrease the amount of electricity generation in
the refinery. It is assumed that electricity generation is 30% efficient, and at
$0.05/kWh this amounts to $4.17/GJ of reduced electricity production. For the
purposes of this model, this is counted as a cost in the protein model, with the
refinery model continuing to produce all electricity. For the stripping column,
estimates using Aspen suggest that 0.29 GJ/Mg water is sufficient to remove
95% of the ammonia. Because the water requirement for AFEX is flexible, most
of the water from the extraction process can remain in the biomass. Only modest
pressure is needed to increase the solids content to 30%. Electricity costs are
low, as pumping costs are only a fraction of the total process electricity required

in the plant. As a first approximation, it is assumed that total process electricity

124

costs increase 5% due to the addition of a protein recovery process. Water loss
is primarily through drying the protein product. It is assumed that this water is not
recovered. Drying the fiber, if necessary, was assumed to require the same
amount of steam per Mg water evaporated as the protein dryer. Evaporating the
solubles was assumed to be performed using a waste heat evaporator , which
would require no additional thermal energy but require $0.75/Mg water

evaporated in electrical costs [47].

Protein recovery is also included after hydrolysis and fermentation assuming the
AFEX—treated fiber is not used for animal feed. The only additional cost of this
process is the added filtration unit and extra dryers. Filtration and drying must be
separated if two separate protein products are to be marketed due to the
potential for differing quality. As specifications for the capital cost of drying

required multiple dryers, this is not explicitly taken into account.

Thus, overall capital and operating costs are listed in Table 6.9. Virtually all of
the capital and operating costs are associated with the amount of liquid in the
process rather than the solid portion. Not included are the added costs to the
ammonia recovery system; these costs are included in the AFEX treatment
process. The capital cost for fiber drying and evaporation was estimated from
the cost of distiller’s grains [98]. All other capital costs were obtained from the
integrated biorefinery Aspen model. Minor equipment such as pumps, etc., were

assumed to be 10% of the overall capital costs. Overall, the cost for producing

125

protein in this manner is $22/Mg initial biomass if the fiber is used for ethanol

production using the base case assumptions and $48/Mg if the fiber is used for

animal feed. The primary difference in costs is due to the high energy cost of

drying the fiber after AFEX, which is not required if it is to be used for ethanol

production.

Table 6.9 : Variable and operating costs for the aqueous extraction module.

 

 

 

 

Variable Costsa Cost Unit Cost ($/Mg BM)b
Filtration 1 0.56 $/Mg water $1.84
NH3 Stripping 1.21 $/Mg water $3.84
Filtration 2 0.56 $/Mg water $2.87
Drying 1 - electricity 0.55 $/Mg water $0.06
Drying 1 - heat 13.3 $/Mg water $1.42
Drying 2 - electricity 0.55 $/Mg water $0.13
Drying 2 - heat 13.3 $/Mg water $3.18
Other electricity 0.46 $/Mg biomass $0.46
Lost ammonia 350 $/Mg ammonia $0.56
Fiber drying 13.3 $/Mg water $23.13
Evaporator Electricity 0.75 $/Mg water $3.73
Capital Costsal Cost TPl Depreciationb

$MM $MM $/Mg BM
Extraction column 0.795 2.709 $1.27
Ultrafiltration 1 0.132 0.448 $0.21
Protein Dryer 1 1.949 6.643 $3.11
Ultrafiltration 2 0.142 0.486 $0.23
Protein Dryer 2 2.111 7.195 $3.37
Fiber Drying and evaporator 5.223 17.800 $8.34
Minor Equipment 1 0.288 0.980 $0.46
Minor Equipment 2 0.225 0.768 $0.36

 

a List of operations and equipment. The number 1 on some operations refers to
the initial extraction prior to AFEX; the number 2 refers to the filtration of the

fermentation media.

9 Example of costs for the base case scenario, using the material balance shown

in Table 6.8.

126

6.4 Mechanical Pressing

6.4.1 Model choice
Two separate processes for mechanical pressing are established in the literature.
The first approach as demonstrated in the Pro-Xan process, uses two twin screw
presses to remove the juice following a hammer mill [47]. The second approach,
as demonstrated in New Zealand, uses only one screw after a hammer mill, but

then passes the fiber residue through a disk mill [48]. After the second milling, a
second and third screw press removes the remainder of the juice. In both cases,

the protein is removed from the juice through heat coagulation.

The primary difference between these two approaches is a tradeoff between
yield and mechanical energy for the mills and presses. The second approach,
with the disk mill, has resulted in protein yields greater than 70% compared to
50-60%% for the first approach. However, the disk mill requires approximately
twice as much electricity as the hammer mill. As the focus of this study is to

maximize production of feed or fuel, the second approach will be used.

Using information from these two sources, an economic and material model can
be created for mechanical pressing. There are three primary operations
performed in this process: milling the biomass, extracting the juice, and
precipitating the protein from the juice. For the purposes of this model, heat
coagulation will be used to precipitate the protein, being the most widely

accepted practice. A value of 75 g steam/kg water was used for precipitation,

127

obtained from Enochian et al. [47]. In addition, two assumptions are made
regarding the mass balance around the coagulation step. It is assumed that 70%
of the soluble protein can be precipitated, and it is assumed that the precipitants
are 50% protein. The second assumption agrees with multiple sources [48; 99],

while the first assumption is obtained from the New Zealand model [48].

The role of the screw press is to dewater the biomass. Both literature studies
considered multiple presses in order to maximize the amount of protein
recovered, although both two and three screw presses have been considered. In
addition, both studies supply excess water to the biomass. Since presses cannot
remove all water, the excess water serves to increase the proportion of water

removed during the press, and therefore increase the soluble protein content.

In order to optimize this procedure, a simple model was constructed to determine
the number of sequential presses to use and the water content prior to extracting.
In this model, it was assumed that the water content in the biomass was reduced
to 65% (total weight basis). Likewise, it was assumed that all solubles were well
mixed, and so the proportion of solubles removed is equal to the proportion of
water removed. The water content of the biomass was varied between 70-90%
(total weight basis) and one, two, or three presses were considered. For multiple
press scenarios, the water content was increased back to the initial value for

subsequent presses. The biomass was assumed to be composed of 20% water

128

solubles and 80% insolubles based on the composition data obtained in

 

 

 

 

 

 

 

Chapters 3 and 4.
3‘; 0. ''''''''''
m .
33 0.
2
2 0.4 «
3' . —----1 Press
g 0.2 1 —2 Press
8 0 1 r 1 _——T—3 Press
0.7 0.75 0.8 0.85 0.9 0.95

Water Content (g/total 9)
Figure 6.3 : Yield of solubles obtained at different starting water contents (9
water/ 9 wet biomass) for a 1, 2, or 3 press system.
The yield of solubles removed by the screw press is shown in Figure 6.3. This
does not take into account loss of yield due to incomplete cell disruption or the
protein yield after heat coagulation. As expected, yield increases with increasing
water content as well as increasing the number of presses. A one-press system
does not appear to be economical, as the soluble yield was only 92% when the
initial water content of the biomass was an unreasonably high 95% water. In
comparison, both of the multipress systems obtained yields in excess of 99%.
As yield is a primary driver of the economics of the process, it is unlikely that a

one-press system would be desirable.

129

 

 

101

 

 

 

 

 

 

 

A -°---1Pl‘eSS‘

E 8 ~ —— 2 Press

in —— 3 Press 1’
Q

175

O

0

g

3

O . i 1 1
0.4 0.5 0.6 0.7 0.8 0.9 1

Solubles Yield (g/g init)

Figure 6.4 : Utility cost (steam for heat coagulation and electricity for the screw
press) as a function of solubles yield for the 1, 2, or 3 press System.

In order to determine these tradeoffs, the utility cost of both options was
considered. It was assumed that the electricity cost for the screw press is
equivalent to 25 kW / dry ton of insoluble residue. One reference fixes a value of
25 kW / dry ton for the last press, where most of the solubles are removed [48].
In addition, a second reference gives a value of 15 kW / initial dry ton [47]. Thus,
this value should be reasonable. Steam price was assumed to be $2.40/ 1000
lb. This steam is used to coagulate the juice after pressing. The cost to cool the
whey after protein precipitation is not considered, nor is the capital cost of

additional presses.

The utility cost relative to solubles release is shown in Figure 6.4. For each
press, costs increase as the yield increases due to increased water content,
which increases the cost of coagulation. The price trends appear to rise

exponentially due to diminishing retums at high water contents, returning only

130

slight improvements in solubles yield despite adding a large amount of water that
must be heated during coagulation. The single press system is the cheapest
option for utility costs unless a yield in excess of 82% is desired. However, since
high yields are expected and the utility cost is relatively low, this option is
unlikely. The three-press system is the cheapest at yields above 95%. If one
assumes that the maximum protein product yield is 60% of the solubles (a
reasonable value based on McDonald et al. ) and a selling price of $360/Mg, a

yield of 95% is also the economic optimum.

 

1’0
0

-----1Pressi

—2 Press /

_L
01
1

 

 

01
1

 

Water Use (g/g water)
E3

 

 

 

0 . . e ,
0.7 0.75 0.8 0.85 0.9 0.95 1
Solubles Yield (g/g init)

Figure 6.5 : Amount of water required for the process for a 1, 2, or 3 press
system.

Since yield and utility cost are the same at this point, other considerations must
be taken into account. A three-press system will require the additional capital
cost of an extra press, while the two-press system requires more water use. At
95% soluble yield, this amounts to approximately 12 Mg water/ Mg biomass for
the two press system, while a third press reduces the water requirement to ~95

Mg / Mg. In both cases, the water requirement is much greater than that required

131

for hydrolysis and fermentation, which is likely to be performed between 15-20%
solid loading. The excess water is recycled through the process, which
increases the concentration of uncoagulated solubles in the protein extraction
process. Alternatively, the solubles can be condensed through evaporation.
Further information on the impact of concentrated recycles must be obtained
before the economic optimum can be determined. For the purposes of this
study, however, the two-press option is used in order to save on the additional
capital costs. Any adverse effect of increased solubles concentration is not

considered.

6.4.2 Mechanical Press model

Thus, a process flow diagram of the mechanical protein extraction process is
shown in Figure 6.6. The process consists of both a hammer mill and disk mill to
increase cell disruption. The model does not consider the amount of cell
disruption from the individual mills, but instead uses the total degree of cell
disruption. These two mills are thus treated as one in the process flow diagram.
Two presses are used in sequence to remove the juice, which is then combined
for direct steam injection. After centrifugation, the protein is dried while the whey
is recycled to either the presses or the hydrolysate. An example material

balance is shown in Table 6.10.

132

 

 

Whey

Disc Mill Press 1 Press 2 AFEX
~—————Extract Extract
. Hammer Mill
Biomass 1 1

 

 

 

 

 
 

 

 
 

Coagulation Centrifuge Hydrolysis

Steam Protein

I

Figure 6.6 : Process flow diagram for mechanical pressing module of the
biorefinery. The scenario where the fiber is used for hydrolysis is shown here. If
the fiber is used for feed, the remaining whey that would be used for hydrolysis is
evaporated and condensed.

 

For this model, it was assumed that the final moisture content would be 63%
after pressing [100]. The solubles are included in the dry biomass, which means
the amount of whey recycled impacts the water loading at each press. Because
of the presence of the feedback loop, an iterative process was used to obtain the
final mass balance in the model. The water content prior to each press was fixed
at 85%. The amount of whey added to the hydrolysate was determined by
insuring the final ethanol concentration after fermentation would be 6% assuming
90% of the fiber is hydrolyzed. Under these conditions, evaporating a portion of
the whey in order to condense solubles was not deemed necessary. If the fiber
is to be used as AFEX-treated feed, then the whey that would be used for
hydrolysis is evaporated and the solubles returned to the fiber. The biomass was
assumed to be reasonably fresh and entered at a moisture content of 50%. If

this number increases, it decreases the amount of makeup water required.

133

Table 6.10 : Example mass balance for the mechanical pressing module of the
biorefinery. Example is for 1 Mg biomass; all values are in kg.

 

 

 

 

Cake Juice
Makeup Whey to after from Whey to

Biomass water press 1 press 1 press 1 press 2
Water 1000.0 1937.4 4452.4 1281.8 6108.1 5073.7
lnsolublesa 664.0 0.0 0.0 664.0 0.0 0.0
Proteinb 56.0 0.0 0.0 9.7 46.3 0.0
Solubles 280.0 0.0 304.1 101.3 482.8 346.5

Juice Cake Whey to

from after hydro-

press 2 press 2 Steam Protein Whey lysis
Water 5102.0 1253.5 763.2 432.7 11540.5 2014.4
Insolubles 0.0 664.0 0.0 0.0 0.0 0.0
Protein 7.8 1.9 0.0 54.1 0.0 0.0
Solubles 359.5 88.3 0.0 54.1 788.2 137.6

 

3 Includes solubles that were not released due to incomplete cell disruption

9 Does not include non-coagulating protein

The cost of the process is shown in Table 6.11. The cost of the hammer mill was
determined based on the value obtained in Mani et al. [101]. The disk mill was
assumed to be twice the cost. Biomass drying and whey evaporation (if not used
for ethanol production) were estimated based on distillers grains as stated in
Section 6.3. Protein dryers were calculated by the same method used in Section
6.3. The cost of steam was calculated using the heat of vaporization of water at
1 atm (2.257 MJ/kg). This energy was assumed to displace electricity, assuming
the efficiency of electricity conversion is 30%. All other capital costs were
estimated using Peters et al. [102]. Utility costs were obtained from the two
models in the literature [47; 48]. The chemicals mentioned include stabilizers,
antifoam agents, and other compounds required to obtain a valuable protein

product.

134

Table 6.11 : Operating and capital costs for the mechanical pressing module.

 

 

 

 

Operating costs Required Unit Cost ($/Mg BM)a
Steam 14.35 kWh $0.72
Makeup water 1.94 Mg $0.00
Hammer mill energy 45.00 kWh $2.25
Disk mill energy 145.00 kWh $7.25
Press energy 33.20 kWh $1.66
Centrifuge energy 50.30 kWh $2.52
Other electricity 10.00 kWh $0.50
Chemicals $2.80
Cost ($MM) TPI ($MM) Depreciation ($/Mg)al
Screw Press 0.720 2.045 $0.96
Centrifuge 1 .000 2.840 $1.33
Dryer 3.028 8.598 $4.03
Disk mill 0.160 0.454 $0.21
Hammer mill 0.080 0.227 $0.11
Steam injector 0.100 0.284 $0.13
Extruder 0.320 0.909 $0.43
Minor Equipment 0.54 1.536 $0.72
Dryer/ Evaporator 5.223 14.833 $6.95

 

a Values shown are for the base case scenario with values as shown in Table

6.10.

135

CHAPTER 7 : DIRECT LAND USE MODEL

7.1 Introduction

In Chapter 6, several models were proposed to calculate the costs associated
with AFEX feed and LPG production. Each aspect of the process: pretreatment
conditions, pretreatment and feedstock costs, refining costs, and various
methods to produce LPCs, were considered separately and adapted from
existing models or data. These models must be combined in order to obtain the

overall impact of each of these technologies.

Two primary issues are addressed in this chapter. The first is to maximize the
economic potential of switchgrass land by considering various technologies.
AFEX feeds, cellulosic ethanol, and LPCs are all possible technologies to
pursue; however, the economic viability of these technologies is unknown. The
second consideration is the productivity of these technologies. Given the
concern regarding biofuels overcrowding the food supply, an economically
attractive option that requires more acreage for the same amount of food and fuel
relative to current technology may not be publicly feasible. Thus, the

technologies are considered for the maximum production on the land as well.

Furthermore, given the general lack of research in this field and the immaturity of
the cellulosic ethanol industry, the costs and products associated with these

technologies are still unknown. While the experimental chapters are intended to

136

provide a greater understanding of these costs and the material balances,
variations between switchgrass types and harvest location are known to produce
differences in the response to AFEX treatment. In addition, different LPC
technologies are available, as well as differences in various AFEX configurations
possible. Naturally, petroleum prices also fluctuate, which affects many of the
costs and selling prices used in these models. Because of these differences,
sensitivity analyses are vital to fully understanding the potential of these

technologies.

Thus, the goals of this chapter are as follows:

. Combine the models from the previous chapter in order to determine the
land productivity and economic benefit of different feed and biofuel
technologies

. Determine the robustness of each technology through multiple sensitivity
analyses

0 Identify the most promising technologies for further research

7.2 Method

The economic benefit and land productivity of an acre of switchgrass are
determined using the economic models described in Chapter 6. Several options
are considered for producing ethanol and/or feed from an acre of switchgrass,
and are listed in Table 7.1. These options include all possible scenarios that

involve either LPC technology or AFEX treated feeds as well as appropriate

137

controls. Two potential switchgrass harvests are used: an early July harvest and

a late October harvest.

For the economic model, the cost and revenue of each option are determined
from the models described in Chapter 6. Agricultural inputs and costs are
determined based on yields as well as the number of switchgrass harvests used
during the process. Ethanol yields and costs are based on experimental data
and the AFEX conditions required for each harvest. Protein yields are based on
experimental data and literature, while costs are based on the two types of
extraction processes described previously. AFEX feed costs are based on
pretreatment costs only, and the selling price of these feeds is determined as the
average of the two models seen in Chapter 5. The revenues from ethanol,
protein, and AFEX feeds are balanced against the costs, and the overall profit

per hectare is determined.

Table 7.1 : Scenarios considered for the direct land use model

 

 

First (July) Harvest Second (OctobeQ Harvest

A None Ethanol production
B None AFEX Feed

C Feed (no technology) Ethanol production
D Feed (no technology) AFEX Feed

E Aqueous protein extract Ethanol production
F with fiber as feed AFEX Feed

G Mechanical protein extract Ethanol production
H with fiber as feed AFEX Feed

| Aqueous protein extract Ethanol production
J with fiber as ethanol AFEX Feed

K Mechanical protein extract Ethanol production
L

with fiber as ethanol AFEX Feed

 

138

For agricultural production costs, three separate costs are calculated. The first is
the price per Mg of biomass produced. This accounts for all of the harvesting
costs, which will increase as yield increases due to the need for extra passes,
larger equipment, etc. The second is the price per hectare, which accounts for
land rent and weed/insect control. The last is the price per harvest, which deals
with fertilizing costs. Because an early harvest removes several nutrients from
the field, it is assumed that greater fertilizer inputs are required for a two harvest

system.

To determine land productivity, each product is compared to the amount of land
or petroleum it displaces. For ethanol production, the control is gasoline at a
ratio of 0.67 L gasoline/L ethanol based on the energy density of each fuel. For
protein feeds, the comparison is the hectares of soy displaced assuming 1.02 Mg
protein/ha. For AFEX feeds, an equivalent amount of corn, soy, and grassland is
displaced based upon the available TDN, protein, and fiber. For fiber, only
feedstocks with long fibers — grass hay, untreated switchgrass, and late harvest
October switchgrass - are included. Only long fibers, which comprise about 75%
of a ruminants’ diet, are balanced in this model. Early harvest switchgrass that
has undergone extraction and pretreatment is not included, as the extraction
process will reduce the length of these fibers. The values of these three nutrients

for each crop are shown in Table 7.2.

139

In order to compare all scenarios on a constant metric, a value defined as Total

Energy Density (TED) is used. This value is defined as the total gasoline

potentially displaced by removing one hectare of farmland from animal feeding

purposes. For each scenario, the amount of gasoline directly displaced by the

hectare of land is added to the potential gasoline displaced from all other animal

feed land displaced by the original hectare. It is assumed that all animal feed

land displaced is replaced with a single harvest of switchgrass for ethanol. Thus,

the equation for total energy displacement is as follows:

(7.1)

where TED is the total energy displaced for scenario i, G] is the gasoline directly

displaced (L/ha) for scenario i, L; is the land (ha/ha) from animal feed displaced

by scenario i, and Ga is the gasoline displaced (L/ha) for scenario A.

Table 7.2 : Nutrient values of all feeds produced in this study and references for
corn, soy, and grass hay. Values are in Mg/ha for the reference feeds (soy, corn,

hay) and Mg/Mg biomass for the feeds produced in the study.

 

 

TDN<’=l Protein Fiber
Soybean meal 1.73 1.02
Corn grain 8.92 0.945
Grass hay 2.38 0.45 2.72
July untreated SGb 0.563 0.1 0.739
October AFEX SG 0.63 0.146 0.819
July AFEX SG — MPC 0.809 0.056
July AFEX SG — AQd 0.809 0.097
Proteierroduct 0.814 0.5

 

a Total digestible nutrients (see Chapter 8 for a full description)

b SG — Switchgrass

C AFEX treated fiber from the July harvest after mechanical pressing
d AFEX treated fiber from the July harvest after aqueous extraction

140

In addition, two sensitivity analyses were performed. In the first case, a Monte
Carlo simulation was performed by randomly varying several unknown or poorly
defined factors. Three levels of each factor are considered and listed in Table
7.3. Protein quality refers to its value in comparison to soybean meal at an
equivalent protein concentration. This affects both the selling price of the protein
as well as land use. Ammonia, harvesting, and fertilizer costs are varied
simultaneously, as all are assumed to be linked to the price of fossil energy.
Corn and soy prices are likewise varied simultaneously. Biorefinery size and
transport costs are also linked, as the distance to the refinery decreases as the
refinery size decreases. All of the switchgrass yields are also varied

simultaneously.

The selling price of AFEX-treated feeds is currently unknown. Based upon their
composition, digestibility, and cost of corn and soybean meal, an estimate of
$140/Mg was obtained. This is similar to the selling price of good quality hay. If
the fiber can compete with this hay, then the price should be reasonable [103].
Soybean hulls sell for approximately $80/Mg, and thus represent the low end of
the price range. Likewise, leaf protein concentrate value is unknown, and so in
this model is tied to soybean meal. The price is determined by the selling price
of soybean meal and the quality of the LPC. It is not clear if the quality would be
above or below the value of soybean meal. If there’s a large amount of
indigestible polyphenolics in the concentrate, particularly if bound to the protein,

then the quality and therefore selling price will likely be lower than soybean meal.

141

However, LPCs may contain valuable pigments such as xanthophyll, which can
greatly increase the value of the meal [47]. Thus, a wide range of values for the

quality of LPCs is used in the sensitivity model.

The costs of producing and harvesting the switchgrass were determined by the
only farm-scale study on switchgrass farming economics available [104]. The
average cost across all sites studied was considered. To better adapt to the
needs of this study, the cost was broken down into three segments. Costs that
were constant (cost per hectare) included land rent, seeding, and weed control.
It is expected that an early harvest would require greater fertilizer costs due to
the first harvest removing nutrients from the soil. Thus, the cost of fertilizer (cost
per harvest) is doubled in the two-harvest scenario compared to the single
harvest. Finally, it was assumed that harvesting costs were directly comparable

to the yield produced, and so make up the final portion (cost per Mg).

The Monte Carlo simulation is used to determine the robustness of each
technology using stochastic dominance analysis. This analysis is important in
decision making and considers the variability within each process. Both the
expected outcome and risk associated with this outcome are taken into account.
The variation in land productivity is also considered, although the variation is not
as great due to fewer factors impacting material changes. The mean and

standard deviation for each approach is shown as well.

142

Table 7.3 : Variables considered in the sensitivity analyses. The medium values
are used for the base case scenario.

 

 

Low Med High Unit Variable Ref
250 350 500 $/Mg Ammonia price [105]
80 160 240 $/Mg Corn selling price [106]
180 360 540 $/Mg Soybean Meal selling price [106]
1.2 1.7 2.2 $/gal Ethanol selling price [107]
80 140 200 $/Mg AFEX Feed selling price a
50 100 150 % of soy LPC Quality
33 67 100 % of LPC Post hydrolysis LPC quality
12 16 20 $/Mg Agricultural cost per Mg [104]
110 210 310 $/ha Agricultural cost per hectare [104]
27 37 47 $/ha Agricultural cost per harvest [104]
5 10 15 $/Mg Transport cost
1000 2000 4000 Mg/day Biorefinery size
7 9.66 13.71 Mg/ha Fall only SG yield [17]
2.23 4.46 6.69 Mg/ha July yield [17]
1.67 3.34 5.01 Mg/ha Fall second cut yield [17]
50 100 150 g/kg Protein in July [17]
65 80 95 % Degree of Cell Disruption [48]
30 40 50 % Aqueous protein yield b
40 60 80 % Extraction filtration yield b
20 40 60 % Fermentation filtration yield b
75 100 125 % Relative protein capex costsC

 

3 Estimated based on data in Chapter 5 and feeding value calculator obtained

from the University of Wisconsin [108]

9 Based on results from Chapter 4.
C Reflects the uncertainty present in the capital cost of both mechanical pressing
and aqueous extraction.

In addition to the Monte Carlo simulation, the impact of each of these variables

on the fourteen options is considered. The variables are changed to determine

what, if any, conditions are necessary to cause certain options to be superior to

others. This is especially important for the multiple protein operations available.

143

7.3 Results and Discussion

7.3.1 Base Case Simulation

A single October harvest used solely for forage feed (scenario B) displaces more
land than any other scenario, as seen in Table 7.4. This is due in part to the
higher yield of a single harvest system and the high displacement value of AFEX
treated forages. AFEX-treated October switchgrass would be used
predominantly as a fiber source, and grass hays have relatively low yields (2.72
Mg/ha of fiber) in this study. While this fiber product also changed the amount of
corn and soy displaced, the displacement of grass accounted for 97% of the total
land displaced. Because of the high land displacement, this scenario also has
the highest potential for total gasoline displacement. The effect of taking one ha
out of feed production is equivalent to producing ethanol from switchgrass on 3
ha of land, thus enabling nearly 5000 L of gasoline to be displaced. Likewise,
this scenario also produced the largest profit per ha, producing $376/ha, which

would likely be split between the farmer and the AFEX processing facility.

In contrast, a single harvest for ethanol production creates a very low profit per
ha. The high cost and the relatively small revenue of ethanol production is the
primary reason for the low value of this scenario, making it an unlikely option.
Likewise, total energy density was also fairly low compared to other scenarios.
Because no animal feed is produced on this land, there is no land displacement.
This difference between feed and fuel production from the October harvest

extends to the multiple harvest scenarios as well. Regardless of how the July

144

harvest is used, the land displacement is 1.11 ha higher, the total energy density
is 1100 L/ha higher, and the profitability is nearly $120/ha higher if the October
harvest is used for feed instead of fuel.

Table 7.4 : Direct land displaced, gasoline displaced, and total energy displaced

per ha of each scenario produced using the base case assumptions. TED -
Total energy displaced.

 

Protein July Fiber October Gasoline Land disp TED

(ha) (ha) Fiber (ha) disp. (L) (ha) (L/ha)
A 0.00 0.00 0.00 1608 0.00 1608
B 0.00 0.00 3.00 0 3.00 4825
C 0.00 1.11 0.00 554 1.11 2343
D 0.00 1.11 1.03 0 2.15 3452
E 0.06 0.45 0.00 554 0.51 1375
F 0.06 0.45 1.03 0 1.54 2484
G 0.24 0.28 0.00 554 0.51 1377
H 0.24 0.28 1.03 0 1.55 2486
l 0.15 0.00 0.00 1453 0.15 1687
J 0.15 0.00 1.03 898 1.18 2796
K 0.24 0.00 0.00 1453 0.24 1833
L 0.24 0.00 1 .03 898 1 .27 2942

 

Economic and TED values for the multiple harvest scenarios tend to be between
these two scenarios. Scenarios that include using the July harvest as AFEX-
treated feed (E-H) provide poor land displacement. This is due to the
assumption that the fiber, despite its high digestibility, is not valuable as a
displacement for hay. Due to the high yields of corn and soy for energy and
protein, respectively, AFEX-treated July switchgrass cannot displace more land
than it uses. Combined with protein extraction, only 0.51 ha are displaced per ha
of July harvest switchgrass. For aqueous protein extraction, most of this land
displacement is in the fiber, as it has a higher protein content due to the lower

yields of this process. For mechanical pressing, the land displacement is similar.

145

In comparison, simply using the July harvest switchgrass as feed without

processing the material displaces slightly more than 1 ha of land, as its yield is

slightly higher than the expected value for grass hay.

Table 7.5 : Net profit for each operation for all scenarios under the base case
assumptions

 

 

Refinery October Total

gate Protein July Fiber Fiber Ethanol Profit

($/ha) ($/ha) ($/ha) ($/ha) ($/ha) ($/ha)
A -498.16 0.00 0.00 0.00 533.36 35.20
B -498.16 0.00 0.00 874.31 0.00 376.15
C -486.54 0.00 401.40 0.00 183.86 98.72
D -486.54 0.00 401.40 301.39 0.00 216.25
E -486.54 -163.63 508.46 0.00 183.86 42.14
F -486.54 -163.63 508.46 301.39 0.00 159.68
G -486.54 -61.78 510.19 0.00 183.86 145.72
H -486.54 -61.78 510.19 301.39 0.00 263.26
I -486.54 21.22 0.00 0.00 570.92 105.60
J -486.54 21.22 0.00 301.39 387.06 223.13
K -486.54 59.31 0.00 0.00 570.92 143.69
L -486.54 59.31 0.00 301.39 387.06 261.23

 

Despite the poorer land displacement potential of treating July harvest
switchgrass for feed relative to feeding directly to cattle, the economics of these
scenarios tend to be better than direct feeding. When mechanical protein
pressing is performed, the total profit per ha is $47 greater than if the July
harvest is fed directly as feed, although it is $56 less if aqueous protein extraction
is performed. This is despite the fact that protein processing is unprofitable in
both sets of scenarios. The lack of profitability is due to the high cost of
evaporating the solubles and drying the biomass after processing. In addition,
the low yields obtained during aqueous extraction also contribute to the high

costs. AFEX treatment of the remaining fiber, however, provided additional value

146

over the untreated fiber, thus leading to improved economics overall. While
these results suggest that a scenario in which no extraction process is performed
prior to AFEX treatment would further improve the economics, concerns
regarding sugar degradation creating neurotoxins (see Chapter 2) may prevent

such an approach.

If the July harvest is to be processed, then the scenarios involving using the fiber
for ethanol (l-L) have higher energy densities than using the harvest for feed (E-
H). Less feed land is displaced, as only 0.17 ha are displaced from aqueous
protein refining and 0.24 ha for mechanical refining. However, the July harvest
has high ethanol yields, displacing nearly 900 L gasoline per ha. Because of this
difference, the overall energy density is higher for July harvests, as the amount of
gasoline displaced per ha removed from feed ranges from 1700 — 2900 L/ha
compared to 1400 - 2500 L/ha for scenarios with AFEX-feeds in July. In
addition, whereas the energy density for aqueous protein extraction and
mechanical pressing were nearly identical when the fiber was fed as feed,
mechanical pressing displaces an additional 110 Uha compared to aqueous
extraction when the fiber is used as ethanol. Despite the high yields obtained
from aqueous extraction, the low quality of the post-fermentation protein limits its
value as land displacement, which accounts for the difference in total energy

density.

147

Despite the improvement in total energy density, using the July harvest for
ethanol production does not necessarily improve the economics of the process.
For mechanical protein extraction, the profit produced per ha is $2 less than if the
fiber portion is used for feed. However, if aqueous protein extraction is
performed, more profit is produced per acre ($64) for ethanol production than
animal feed. In both cases, protein extraction is profitable, as there is no need to
evaporate the whey or dry the fiber. However, ethanol production from the July
fiber is less profitable than feed production. If ethanol production is performed,
there is less difference in the economics between mechanical and aqueous
protein extraction, as the post-fermentation protein recovery adds little in cost but

greatly increases the revenue of protein recovery.

While the total profit is positive in all scenarios, the opportunity cost in
comparison to traditional crops may not be valuable in all cases. Profits for crops
can vary widely across seasons and locations. For example, the farmer’s profit
for traditional crops in North Dakota (one of the states that the switchgrass
production costs used in this model is based on) ranged from $180 to $230 per
hectare in 2008 [109], but only $50 to $120 per hectare in 2009 [110]. The total
profit per ha in this model must be split among both the farmers and the refinery,
and thus the farmer sees only a portion of the values shown in Table 7.5. Thus,
only a few scenarios would be competitive with traditional crops in 2008, while
several could compete in 2009. Farmers would be forced to undergo risk

assessment; since switchgrass is perennial, a field of switchgrass would be

148

planned for several years. Alternatively, switchgrass may only be grown on
marginal land rather than prime cropland. In this case, the opportunity cost of
growing switchgrass may be less. One approximation of this cost is the
Conservation Reserve Program credit, given for land taken out of production for
the purpose of growing grasses for environmental purposes. These payments
will be $125 per ha in 2010 [111], and thus comparable with the profits produced

in this model for both LPC production and AFEX treated feeds.

7.3.2 Monte Carlo Analysis

The mean and standard deviation of the Monte Carlo analysis for economic and
material analysis of each condition are shown in Table 7.6. As with the base
case simulation, the single harvest animal feed production scenario is the most
profitable scenario per hectare of land, and is 34% higher than the second most
profitable scenario. Likewise, for all multiple harvest scenarios, using the second
harvest for feed production is, on average, more profitable than any scenario
where the second harvest is used for fuel production. For scenarios where the
October harvest is used for ethanol, mechanically extracting the protein from the
July harvest and using the remaining fiber for feed tends is on average the most
profitable scenario, while allowing the July switchgrass to be used for feed
without any treatment or the fiber to be used for ethanol after protein extraction
tends to be less profitable. Scenario E, in which aqueous protein extraction is

performed on the July harvest and ethanol is produced from the October harvest,

149

is the least profitable, showing on average virtually no value. All scenarios had a

large standard deviation however, ranging from $203/ha to $596/ha.

Likewise, the average total energy density follows the same patterns as the base
case simulation. Again, the single October harvest for ethanol displaces the
most gasoline, averaging nearly 5700 L of gasoline per ha of AFEX-treated feed
produced. If a July harvest is to be used, either using the untreated material as
feed or removing the protein and producing ethanol from the remaining biomass
have strong potential for overall gasoline displacement.

Table 7.6 : Mean and standard deviations of both total energy density and profit
for the Monte Carlo analysis for each scenario.

 

 

Total Energy Density (L/ha) Profit ($/ha)

Scenario Mean St. Dev. Mean St. Dev.

A 1684 466 18.09 304.58
B 5697 3037 341 .94 596.04
C 2649 1622 74.01 203.02
D 4032 261 1 179.24 302.44
E 1 524 940 2.22 376.75
F 2907 1905 107.45 443.26
G 1526 942 1 10.52 454.76
H 2910 1907 215.76 515.90
| 1713 782 72.03 319.23
J 3096 1776 177.27 343.30
K 1907 927 119.16 352.17
L 3290 1913 224.39 374.06

 

For both economic and land displacement analysis, high variability is seen in this
analysis. For a closer inspection of this variation, the cumulative distribution
function (CDF) of each option is shown in Figure 7.1. As seen from these
figures, a single harvest for ethanol production is a very high-risk endeavor, with

nearly a 50% chance of being unprofitable. The single harvest for ethanol has a

150

higher risk associated with it than any other scenario except scenario E, as it has
both a higher chance of being unprofitable and greater losses than other
scenarios. A two harvest system, with no treatment of the first harvest,
decreases this risk, but at the cost of lower profits at the more desirable market
conditions. The risk associated with this scenario is also lower than any two
harvest scenario with protein extraction. However, scenario K has much greater
potential for profitability. If the latter harvest is to be used for ethanol production,
these two scenarios, C and K, are likely to be the two strongest depending upon

the propensity for risk taking among the producers.

Mechanical protein processing tends to be superior to aqueous extraction across
nearly the entire CDF regardless of how the July fibrous fraction is used. If the
July harvest is used for ethanol production, aqueous extraction is better in the
worst case scenarios (less than 1% probability), although this difference is only
~$10/ha. The benefit of mechanical pressing is greater when the July fiber is
used for ethanol, particularly at the more advantageous market conditions. For
the top third of the CDF, mechanical pressing on average is $200/ha more
profitable than aqueous extraction when the fiber is used as feed, while only
$80/ha when the fiber is used for ethanol. This is due to the greater protein
recovery for aqueous extraction when integrated with biofuels. Thus, according
to this model, it is unlikely that aqueous protein extraction would be pursued.
For both aqueous and mechanical protein production, there is less risk

associated with using the fiber for ethanol as opposed to feed. However, there is

151

greater potential for profit with fiber feeding. The choice of these two options is

dependent upon the amount of risk that a developer is willing to commit to.

 

 

 

 

    

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

   

 

 

 

 

 

 

 

1
0.75 0.75
0.5 0.5
0.25 0.25 — C
-1000 0 1000 2000 -1000 0 1000 2000
Profit ($/ha) Profit ($/ha)
1 J 1
0.75 0.75
0.5 ,. 0.5
0.25 4 —E, 0.25 —Gl
./ ' _
0 . 1 — F 0 ? H
-1000 0 1000 2000 -1000 0 1000 2000
Profit ($/ha) Profit ($/ha)
1 1
0.75 0.75
0.5 0.5
0.25 0.25
0 i 0 I
-1000 0 1000 2000 -1000 0 1000 2000
Profit ($/ha) Profit ($/ha)

Figure 7.1 : Cumulative distribution functions for profit for all scenarios.

152

The greatest potential for profit, however, is from the single harvest used as
animal feed. This approach has the potential to produce $2000/ha according to
this model, and has only a 33% chance of being unprofitable. This is likely due
to either low biomass yield or poor selling price of feed. Despite this, it does not
stochastically dominate all other scenarios, as several scenarios do not lose as
much money at the low end of the spectrum. It should be noted, however, that
no new technology is introduced between this scenario and scenario D, where
the early harvest is used as untreated feed. In other words, a farmer could
potentially choose between a one or two harvest system dependent upon the
economic conditions of the year without regard to the capabilities of a refinery or
RBPC. Thus, both scenarios are interchangeable on a year-to-year basis given
the economic conditions present, further improving the value of these two options

relative to all others.

Since economics are unimportant for the land use function, the probability
distribution is more discrete and primarily based on farm yields as well as protein
extraction yields. Despite this, the preference for a single harvest for feed is
also seen in the land use functions, as seen in Figure 7.2. Likewise, Scenario D
has stochastic dominance over all other scenarios as well except scenario B. It
is clear that, in this model, producing fibrous feed is the strongest way to
increase land displacement and therefore increase biofuel production. These are
also the only two scenarios that show land displacement greater than 1 ha/ha in

the lower 33% probability. This probability range is associated with low

153

switchgrass yield, and so high yields are vital if co-production of food and fuel is
to be performed. However, it should be noted that switchgrass yield is often a
factor of location, weather, and soil quality. Thus, a hectare of switchgrass
planted in poor cropland would not be competing with the average yield of corn

or soy, but rather with lower yields of these crops.

For land use distribution, differences were seen between mechanical and
aqueous protein processing when the July fiber was used for ethanol production.
Despite the fact that the aqueous production is more profitable, the mechanical
press tends to displace more land than the aqueous extraction. This is due
primarily to the lower quality of the post-AFEX protein, which decreases its ability
to displace soy despite higher yields. When the fiber is used as animal feed,
however, there is virtually no difference in land displacement. Any protein that is

not extracted remains with the fiber, and thus eventually is used as animal feed.

154

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.75 — 0.75 -
0.5 ~ 0.5 ~
0.25 — l—A 0.25 1 —C
;———Bl —
O i F1 11 0 I l D
0 4000 8000 12000 0 4000 8000 12000
TED (Uha) TED (L/ha)
1 1
0.75 ~ 0.75 «
0.5 — 0.5 ~
0.25 — —E 0.25 - —G
—F _
O 1 l O T I H l
0 4000 8000 12000 0 4000 8000 12000
TED (L/ha) TED (L/ha)
1 1
0.75 — 0.75 —
0.5 — 0.5 ~
0.25 ~ —I 0.25 - —K
0 1 . _J 0 . . —L
0 4000 8000 12000 0 4000 8000 12000
TED (L/ha) TED (Uha)

Figure 7.2 : Cumulative distribution functions for the total energy density (TED)
of each scenario for the Monte Carlo analysis.

155

7.3.3 Other Sensitivity Analyses

The impact of each variable on the economics of scenario J is shown in Figure
7.3. This scenario was chosen as representative of all scenarios, as it is the only
one in which each variable impacted the economics. All values for each scenario
is shown in Table 7.7. The selling prices of the products and co-products, crop
yield, and harvest costs, are the dominant influences on the economics of all
scenarios. The selling price of either ethanol or AFEX treated fibrous feed has a
larger impact on the single harvest scenarios than the multiple harvest, as the
single harvest scenarios rely solely on one product. If the selling price of AFEX-
treated feed drops from $140/Mg to $80/Mg, for example, the profit from scenario
B drops $580/ha, while only dropping $200/ha for the multiple harvest scenarios.
Alternatively, crop yields had the largest impact on the multiple harvest
scenarios, ranging from $192-$328/ha difference between low and medium crop
yields compared to $79-$222/ha for the single harvest system. For the multiple
product scenarios such as Scenario J, the selling price of both feed and ethanol

are important, as both are produced.

Protein selling price, which depends on the crop price and quality of protein
products, is not as important to the overall economics due to the lower overall
yields of a protein product. This is true for all scenarios with protein production,
although the impact is larger with mechanical pressing due to its higher yields.
Despite the uncertainty in the capital cost of the protein recovery options, it has

little impact on the overall economics, with less than $10/ha difference for all

156

scenarios. Biorefinery size and transportation costs have modest impacts on all
scenarios. Interestingly, since biorefinery size and transportation cost were
varied simultaneously, both the low and high refinery sizes produce less profit
than when the size is 2000 Mg/day. This only occurs when AFEX-treated feeds
are sold, as they must be returned to the farms. This suggests that, as refinery

sizes grow, the RBPC concept is critical if animal feeds are produced as well.

Scenario J

 

 

J Crop Yield

J AFEX Feed Price

Ethanol Price

Energy Price

LPC Quality

Crop Prices

Protein Content

AFEX LPC Quality

Filtration Weld
Biorefinery Size

I Low value I Extraction yield

El High value 1] Protein Capex

 

 

 

 

 

 

 

 

 

 

 

 

-300 -200 -100 0 100 200 300
Price difference ($/ha)

Figure 7.3 : Tornado plot of low and high values for several variables on the
economic value of Scenario J. All variable values are seen in Table 7.3. Energy
price refers to ammonia price, electricity price, and price per harvest. Crop
prices refers to both transportation cost and biorefinery size.

Of particular interest is the impact of the amount of protein recoverable during the
filtration process in aqueous protein extraction. For scenario J, the impact of

filtration on economics was $29/ha between the low and medium values,

compared to $9/ha for protein extraction yields. If the fiber is not used for ethanol

157

production (scenarios E and F), the impact of filtration drops to only $17/ha due
to the very low yields involved with only one extraction. Given the small
difference between aqueous extraction and mechanical protein when the fiber is
used for ethanol production, improvements in filtration technology or decreasing
the breakdown of protein are critical to implementing aqueous extraction.
Likewise, the impact of the quality of the post fermentation protein also has a
significant impact, changing the economics by $33/ha. If both improved filtration
recovery and a high quality of post-AFEX protein can be obtained, then it is
possible that aqueous protein extraction can compete with mechanical pressing

with improved technology.

This model assumes that the amount of switchgrass harvested in a two-harvest
scenario is approximately 20% less than the yield in a single harvest. To test the
impact of this assumption, the ratio between the combined yield of the two-
harvest system and the single harvest was allowed to vary. The ratio between
the yield of the July harvest and October harvest for the two harvest scenario
was kept constant. The overall profit per hectare for scenario D, H, and L was
compared to scenario B. These three scenarios are the most profitable multiple

harvest scenarios.

158

 

 

8 8 .8 .8 28 28 2 2 8 8 8 8 880
£9on

8 8 88$ 88$ 8 8 82$ 8$ 8 8 8 8 2.28;
888.8

88$ 88$ 8$ 8$ 88$ 88$ 82$ 82$ 8 8 8 8 8.283
cozomzxw

88$ 88$ 88$ 82$ 88$ 88$ 88$ 82$ 88$ 82$ 82$ 88$ 8.28; 86
88$ 88$ 88$ 88$ 88$ 88$ 88$ 88$ 8 8 8 8 2980
£995

82$ 88$ 82$ 88$ 88 8$ 8.8 8$ E8 88 88 88$ 888
8288.65

8 8 88$ 88$ 8 8 8 8 8 8 8 8 8830
9: x22

88$ 88$ 88$ 88$ 88$ 88$ 88$ 88$ 8 8 8 8 88:0 08..
88$ 8 88$ 8 88$ 8 88$ 8 88$ 8 88$ 8 8cm.
88 me<

82$ 88$ 82$ 88$ 8 82$ 8 82$ 8 82$ 8 88$ 8:0.
885m

88$ 88$ 88$ 88$ 88$ 82$ 88$ 88$ 8 8 8 8 885 85
8828 8828 8828 88F8 28F8 88_8 4828 88F8 2828 8828 8828 88_8 888:;
>985

._ v. 8 n. o o m 829, 25..

 

ms Bow... c_ 53058 mm 28 82:85 :9: new 26.. .2358 comm .2 3:32:05 2: :0 838:? zomo ho wootm H 8.8. 035...

159

060E co=2c0E.2 05 B 2.0 20:00.80 520.0 20 c0305: 000205 .

20:00.20 0:00:80 .2 8:00.80 28 85000.8 822.2802. .0. 2.038206 :00 .0 00.800 000205 0
0E20>0 20022.. 0.5.. 2.0 #0022. 06:8 0 £00 .0. 0203 80.0 008305 0

2000 .8852. 2.0 050 80.50.05 £00 00.0205 0

000:8 85:00 >00 2.0 Eoo 500 000205 a

.0022. .00 00:0 8.23030 2.0 02.0 £03020 .00_.n_ mEoEEm 008205 a

 

8$ 8$ 8$ 8$ 82$ 82$ 8$ 8$ 8 8 8 8 .880
£991

8 8 88 88 8 8 28 28 8 8 8 88 22>
88:...

N8 N8 8 8 m8 88 28 28 8 8 8 88 22>
cozowbxm

28 .28 88 028 E8 28 88 828 88 N28 88 28 22> 8.0
m8 88 88 88 88 88 8 8 8 8 8 8 E980
E991

80$ 8$ 80$ 8$ 82$ 80$ 82$ 88$ 82$ 80$ 82$ 82$ 80
805.085

8 8 m8 m8 8 8 8 8 8 8 8 8 8:80
on: x22

88 88 88 88 88 88 88 98 8 8 8 8 88:0 0.:
88 8 88 8 88 8 88 8 88 8 88 8 8:8
88". x22

28 88 828 88 8 828 8 828 8 828 8 28 02...
885m.

88 88 88 88 88 88 88 2.8 8 8 8 8 82.8 8.0
82$ 82$ 82$ 82$ 88$ 88$ 88$ 88$ 82$ 82$ 88$ 82$ 02.8 8.2.0

 

00205 :91

1 60

The ratio between these scenarios and scenario B compared to the ratio
between two-harvest yields and a single harvest is shown in Figure 7.4. If
protein extraction is to be performed, then this option becomes more profitable
than a single harvest when the combined harvest is 98% of the yield of a single
harvest, regardless of whether the July fiber is used for feed or fuel. Thus,
despite the potential for a single harvest devoted to animal feed, more
information is required regarding the yields of switchgrass under multi-harvest
scenarios. If two harvests produce more biomass than a single harvest, then
protein extraction becomes a viable choice under the base case scenario.
However, simply allowing the July material to be used as animal feed without
treatment would require the multiple harvests to produce approximately 7% more
biomass than the single harvest in order to be profitable. Despite the changes in
profitability, the total energy displacement still tends to be superior for scenario B.
Scenario D required a harvest ratio of 1.13 to obtain higher energy displacement
than the single harvest, while scenarios H and L required ratios of 1.58 and 1.34,
respectively. Thus, while it may be possible for a two harvest system to be more
profitable than the single harvest under the base case scenario, it is unlikely that
this approach will be able to displace more petroleum per ha removed from

farmland.

161

 

 

 

 

 

 

 

 

 

1.2 , ,
.e ,I-
E , ,.
o: /./'
,5
(I -----o
———H
0.4 i , T “'-
07 0.8 0.9 1 1.1 1.2

Crop Yield Ratio

Figure 7.4 : Impact on relative crop yields for a one or two harvest system on the
economics of each scenario. The combined crop yield of the two harvest system
divided by the yield of the single harvest is the crop yield ratio, while the net profit
of scenario D, H, or L divided by the net profit of scenario B is the profit ratio.

7.4 Conclusions

The direct land use model presented in this chapter strongly suggests that a
single harvest used solely for animal feed is the optimal use of a hectare of
switchgrass land. Under the base case assumptions, this scenario displaced
more land and produced a greater profit than any other scenario. Multiple
sensitivity analyses confirmed the robustness of this option. While the single
harvest for feed scenario showed great variability in response to changing
assumptions, it had the lowest possibility of being unprofitable and stochastically
dominated all other scenarios except for a two harvest system with no treatment
for the July harvest. The yield of switchgrass and the selling price of AFEX-
treated fiber were the most important variables in determining the economics of

this scenario.

162

In contrast, a single harvest for ethanol production gave the poorest economic
return of all scenarios tested, although it displaced more land than some
scenarios. According to the Monte Carlo analysis, there is nearly a 50% chance
of this scenario being unprofitable using current information. This result,
combined with the benefits of a single harvest for feed as mentioned previously,
clearly illustrate the need for dedicated energy crops to diversify beyond

bioenergy alone.

For protein extraction, the simulation suggests focusing on mechanical pressing
as opposed to aqueous extraction. If the fiber after protein removal is to be used
as feed, mechanical pressing is more profitable than aqueous extraction due to
obtaining much higher yields. Both protein operations, however, are highly
unprofitable due to the need to evaporate the whey after protein concentration. If
the fiber is to be used for ethanol production in which no evaporation is required,
then there is only a small difference in the profitability of the two protein options.
If aqueous extraction is to compete with mechanical protein extraction, research
should be focused on improving the quality and recovery of the protein extract

rather than the initial protein yield.

163

CHAPTER 8 : AGGREGRATE LAND USE MODEL

8.1 Introduction

In Chapter 7, the farm scale economics and material benefit were considered.
However, this level of analysis does not take into account the potential changes
to the overall farm landscape due to changes in animal diets. Likewise, the
model does not consider the amount of AFEX-treated feeds or leaf protein
concentrate required in the feed marketplace. Thus, it is not clear what the
maximum amount of biofuel that can be produced on cropland due to improving

feed efficiency. This chapter will explore this impact.

The goal of this model is to determine the maximum ethanol production on
cropland harvested for animal feed provided that the land also produces enough
animal feed to satisfy current requirements. In addition to animal feed land,
current US corn land for ethanol production is also considered, as this land is
already producing biofuels as well as animal feed. The model does not take into
account the economics of any process involved, nor is geographical location
taken into account. Instead, average yields are used for all farmland, and it is
assumed that all crops are available for processing or for animal feed regardless
of where they are grown. In the large scale, the prices of these commodities will
fluctuate if significant portions of the feed market are displaced with new types of
feed based on the laws of supply and demand. Because of this, economic

modeling would drastically increase the complexity of the model beyond the

164

scope required here. Instead, it is assumed that the economics will support both

AFEX treated feeds and LPC production.

This model serves several primary purposes. Most importantly, it provides a
theoretical upper limit to the amount of ethanol produced on current cropland
used for feed. This theoretical upper limit is only on the cropland studied, and is
thus not a theoretical upper limit on US biofuel production. However, this model
can be added to estimates on biofuel production from forest land, idle land,
changes in exports, or municipal solid waste. Furthermore, this model is a
general estimate of indirect land use change. The second purpose is to
determine the size and scope of the two feed technologies used. These two
purposes combined can determine what impact the two feed technologies have
on the volume of ethanol produced. Third, the model can showcase what the
crop distribution would be in a future with the two new feed technologies

incorporated.

Table 8.1 : Cropland and yields used for this study [9]

 

 

 

Size Yield Yield
Crop (MM acres) (Mg/acre) (Tg)
Corn for feed 40.5 4.07 164.6
Soy 37.8 1.07 40.4
Alfalfa 20.2 3.19 64.6
Other Haylage 41.2 1.71 70.5
Cropland for pasture 35.8
Corn for ethanol 20.0 4.07 81.6
Total 195.5 421 .7

 

165

8.2 Method

The land use considered in this model is shown in Table 8.1. This land amounts
to 48% of US cropland. Specifically excluded from this land are crops grown for
human consumption (primarily wheat), crops grown for export, and idle land.
Current animal feed requirements are shown in Table 8.2. These do not include
minor livestock such as goats. Table 8.1 does not consider rangeland pasture,
specialty protein products such as synthetic amino acids or bone meal, or non-
protein nitrogen for ruminants. For the model, the feed requirements were used,
with values increased by 10% to account for potential losses. These losses can

be due to spoilage, overfeeding (particularly proteins), and other causes.

Due to the scope of the model, various assumptions are made in order to simplify
the model for use. Thus, only material balancing is used here, and it is assumed
that the economic vitality of the animal feed production is sufficient to produce
these changes. The time required to convert farms to these crops and build the
refineries and treatment centers is also not considered. Therefore, no attempt is
made to predict yields or feed requirements for a specific year in the future,
although both are varied as a sensitivity analysis. Finally, location is not
considered. It is assumed that all farmland is available for any type of crop and
use. Again, sensitivity analyses are used to determine the impact of this

assumption.

166

Table 8.2 : Livestock population and feed requirements in the United States

 

 

 

 

 

 

 

 

 

 

Herd Protein TDN
sizea Requirementb Requirementb
Me Animal Class 1000 g/d Tg/y kg/d Tg/y
Dairy Lactating cow 7997 2750 8.03 11.07 32.31
Dry cow 1333 1310 0.64 5.14 2.50
Heifer 4410 1070 1.72 4.59 7.39
Calves 1877 421 0.29 1.59 1.09
Total 15617 10.68 43.29
Beef Lactating cow 15850 1440 8.33 6.00 34.71
Dry cow 15850 880 5.09 4.41 25.51
Heifer 5530 900 1.82 4.50 9.08
Finishing 16800 859 5.27 5.05 30.94.
"Other" 9650 900 3.17 4.50 15.85
Calves 13023 656 3.12 2.27 10.80
Bulls 2180 572 0.46 5.30 4.21
Total 78883 27.25 131.1 1
Swine Gestating sow 5127 233 0.44 1.41 2.64
lactating sow 840 936 0.29 3.98 1.22
Hogs < 27 kg 21673 194 1.53 0.59 4.67
Hogs 27-54 kg 15008 350 1.92 1.50 8.22
Hogs 54-82 kg 12584 401 1.84 1.93 8.87
Hogs > 82 kg 10847 406 1.61 2.18 8.64
Total 66079 7.62 34.26
Eggs and
Poultry Layers 340,000 16.8 2.08 0.07 8.18
Broilers 9075261 1117 10.14 4.00 36.30
Turkeys 271425 6517 1 .77 24.09 6.54
Total 9686686 13.99 51 .02
Nonruminant 9752765 21 .62 85.28
Ruminant 94500 37.92 174.39
Total 59.54 259.67

 

a Values obtained from the USDA [9]
b Values obtained from Dale et al. [112]

Two primary feed requirements are considered: energy and protein. Energy is

measured as total digestible nutrients (TDN), which is defined as 4.4 kcal/g TDN.

The energy requirement used is digestible energy. Specific constraints are

added to the model to insure ruminant and non-ruminant needs are met. It is

assumed that any feedstocks with large amounts of fiber present are unavailable

167

for non-ruminants, including distiller’s grains. Although distiller’s grains have

been included in limited quantities in non-ruminant diets, this would require

separate TDN values compared to ruminants, as they cannot digest the fiber.

Since soybean oil is currently used for industrial applications and also grown on

this land, oil was also balanced based on the amount produced from soybeans in

2007 [9]. Here, the options available for oil were canola, soy, or, in one

sensitivity scenario, corn.

Other major assumptions are as follows:

Ammonia-based nitrogen is limited to 27% of ruminant protein
requirements. This is based on recommended values of 10% for dairy
cattle and 33% for beef cattle [113]. While more can be fed, it lowers the
performance of the animals. In addition, even if excess true protein is
added, excess nitrogen increases the environmental impact of ruminant
production, and thus should be avoided.

Rough fiber is required at 20% of ruminant TDN feed. This is based on
the general determination that approximately 25-28% of a ruminant diet
should be fiber, of which 75% should be rough fiber [113]. Only alfalfa,
cover crops, and AFEX-treated switchgrass are available for rough fiber.
AFEX-treated corn stover and DDGS fiber are not considered rough fiber,
as they are rapidly degradable and unlikely to be in the form of long fibers.
At least 60% of non-ruminant protein must be from soy or LPCs in order to

satisfy the lysine requirement (approximately 5% of total amino acids)

168

[113]. Lysine is often the limiting amino acid in non-ruminant diets, and

thus is the only amino acid considered.

In addition, several assumptions are made on the farmland available. It is
assumed that only 1/3 of the cropland is suitable for growing a cover crop. This
applies to all non-forage crops. For corn land without a cover crop, only 40% of
stover was allowed to be removed vs 70% for land with a cover crop [29; 114]. In
addition, non-forage crops are constrained to the 98.28 million acres, or the
amount of corn and soy land currently used for ethanol and feed [9]. Crop

rotation was limited to 3 years of corn for every year of either corn or canola.

The model assumes the following crops. The yield and composition of each crop

are listed in Table 8.3.

0 Corn grain and stover collection: The corn grain can be fed directly to
cattle or can be used for ethanol and DG production. The com stover is
AFEX treated and used for either ethanol production or a ruminant feed.

0 Alfalfa: Alfalfa can be either fed to cattle as is or used for LPC production.
If LPC is produced, the remaining fiber is converted to ethanol. The fiber
is not considered for animal feed based on the low digestibility

improvements seen for alfalfa in Chapter 7.

169

Soybean: The possibility remains that soy is the optimal source for protein
and oil production. In addition, soy is needed to insure that com is not
grown continuously on the same land.

Canola: If soybeans are not produced, this represents a loss of oil for
industrial purposes. Canola can produce more oil per acre than soy.
Nothing else is produced from canola land other than vegetable oil.
Switchgrass: Switchgrass can be used for two purposes. The first is as an
AFEX treated feed for ruminants, while the second is solely ethanol
production. Because of the low protein content present, no early harvest
is considered here.

Cover crops: No additional land is dedicated to cover crops. Instead, 1/3
of corn, soy, and canola land are also used to produce the cover crop. As
with alfalfa, the crops can either be fed as is or used for LPC and ethanol

production.

In addition to the base case scenario, several alternative scenarios were

considered as a sensitivity analysis.

0 - Base scenario

A - All crop yields increase by 25%

B - Switchgrass yields increase by 25%
C - No cover crop is produced

D - Animal feed needs increase 25%

E - A maximum of 15 billion gallons of corn grain ethanol is produced

\

170

F - Corn oil is produced in addition to ethanol + DDGS

G - Protein extraction yields increase 25%

H - Protein extraction yields decrease 50%

l - A three year rotation is used rather than four years

J - Non-protein Nitrogen constraint lifted

K - Non-protein Nitrogen constraint decreases 50%

L - No LPC is produced; cover crops are used for feed alone
M - No AFEX treated feeds are produced

N - Neither AFEX treated feeds nor LPC are produced

Table 8.3 : Yields and nutritional value for all feeds produced in the aggregate

land use model.

 

 

Yield TDN Protein NPNa Fiber Oil Ref
Mg/ha glg g/g glg 9/9 9/9
Corn 10.05 0.887 0.094 [33]
Corn Stoverb 6.10 0.756 0.042 0.130 Ch 5
Alfalfa 7.90 0.589 0.202 0.396 [33]
Alfalfa LPCC 0.214 0.132
Canola 3.36 0.510 [115]
Switchgrassb 13.71 0.630 0.014 0.132 0.819 Ch 5
Cover Crop 5.61 0.631 0.180 0.496 [33]
Cover LPCC 0.190 0.117
Soybean 2.64 0.654 0.386 0.196 [33]
DDGSd 0.256 0.096

 

3 Crude protein from ammonia addition (non protein nitrogen)
b AFEX-treated feeds
0 Yield assumes 65% protein recovery and TDN comparable to soybean meal.
Values listed are per Mg alfalfa or cover crop, not per Mg LPC.

d DDGS — Nutritional values are calculated per Mg corn grain

These scenarios reflect the uncertainty in a future, intensely managed farm

scenario. In addition, they reflect the possible limitations of intensive farming.

171

Cover crops may not be economical to harvest or even available in all areas.
Yields for crops can improve with time and technology. This is particularly true
for switchgrass, which does not have a history of farming associated with it.
Concerns regarding the environmental performance of corn, particularly corn
ethanol, also influence the creation of these scenarios. The final three scenarios

reflect the importance of the technologies considered in this study.

8.3 Results and Discussion

As stated in the previous chapter, the purpose of this model is to determine the
material impact on biofuel production in the United States due to more intensive
farming and feed production practices. In addition to the base-case scenario,

thirteen additional scenarios were also considered.

The total ethanol produced per year in each of these scenarios is seen in Figure
8.1. In the control scenario, 271 billion liters (GL) per year of ethanol or its
equivalent are produced, which is equivalent to approximately 35% of total US
gasoline consumption. This value is not the total amount of biofuel that can be
produced in the United States, as stated previously. An average of over 3400
liters of ethanol are produced per hectare on land currently devoted to animal
feed or ethanol land, which compares favorably to 4200 liters per hectare on
current corn ethanol land. This includes 64 GL of starch-based ethanol, slightly
higher than the cap given in the EISA 2007 bill as seen in Chapter 1. While this

scenario does create more com land for ethanol production, the presence of a

172

cover crop means that this land is also producing protein or forage. Furthermore,
over 43 million hectares are removed from animal feed production and converted
to either switchgrass or corn for ethanol. In other words, the amount of land

required to feed animals decreases under this intensive scenario, thereby freeing

more land for ethanol production.

 

400

300 "

200 '

Ethanol production (GLIy)

 

 

 

0ABCDEFGHIJKLMN
Scenario

Figure 8.1 : Ethanol production on selected cropland for the base case and 13
scenarios tested.

The first four scenarios produced the largest changes in ethanol production. As
expected, increasing crop yields improved ethanol production, in this case to
nearly 400 GL per year. This is larger than a 25% change, as animal feed needs
remain constant. Eliminating cover crop production and increasing animal feed
requirements both have similar impacts, driving the ethanol production down to
210-220 GlJyr. Of the remaining scenarios, lifting the non-protein nitrogen
constraint increased production the most, as an extra 13 GL/yr of ethanol are

produced. Because AFEX treated switchgrass produces more fiber per ha than

173

alfalfa or a cover crop, lifting this constraint frees additional land for ethanol
production. Both the decrease in protein extraction yields and a three year crop
rotation decreased ethanol yields, as both scenarios forced more land into soy

acreage.

Of particular emphasis are the last three scenarios. Interestingly, the addition of
these two technologies did not cause a great increase in ethanol production, as
ethanol production only decreased by 42 GL/yr for scenario N. In other words,
adding these two technologies to the market can only increase the potential
ethanol produced from farmland by ~18%. Instead, most of the ethanol seen in
the base case scenario comes from more intensive farming, namely the addition
of cover crops and high producing alfalfa. AFEX-treated forages have a larger
impact on improving ethanol yields than LPC production. Compared to scenario
N, allowing AFEX treated feeds increases ethanol production by nearly 28 GL/yr

compared to only 14 GL/yr for LPC production.

Despite only modest improvements in ethanol production, the two animal feed
technologies considered in this study are a large industry in these scenarios, as
seen in Figure 8.2. Approximately 85 Tg of AFEX treated forages are produced
in nearly all studies. In all cases, the non-protein nitrogen constraint is limiting for
this forage. Thus, only scenario J and D (where the constraint is increased 25%
due to increased animal feed requirements) increase the size of the industry,

while only scenario K (where the constraint is decreased 50%). The amount of

174

LPC produced varies greatly, however. The base scenario produced
approximately 13 Tg of LPC from 52 Tg forage. When animal feed requirements
increase, LPC production is more important in order to meet these requirements
while still maintaining high ethanol production. Alternatively, LPCs become less
important as crop yields increase, or as corn land decreases due to crop rotation
limitations. Relaxing the NPN constraint also increases LPC production, mostly
due to decreasing the amount of alfalfa and cover crop dedicated to cattle feed.
Also of note is scenario H. When low LPC yields are considered, this technology
is almost entirely eliminated, indicating the necessity of reaching very high yields
associated with a large degree of cell disruption. For the base case scenario,
these two industries represent a $17 billion per year industry, assuming an

average selling price of $150/Mg AFEX feed and $350/Mg for LPC feed.

 

I AFEX Forage
150 - 5 Leaf Forage

 

 

 

 

 

 

 

 

 

Scenario

Figure 8.2 : Amount of biomass used for advanced feed technologies. Both
AFEX-feeds and forage for leaf protein production are shown.

175

The makeup of cropland for each scenario is shown in Table 8.4. Corn land is
maximized in most scenarios, due to its high overall productivity between grain,
stover, and cover crops. It is only when LPC production is eliminated (thus
requiring more soy land) or corn grain ethanol is limited (thus eliminating much of
the land’s value) that com land is reduced below the maximum constraint. Also
of interest is the amount of land present in alfalfa. Alfalfa land tends to be low in
most scenarios, and is eliminated at high crop yields due to the amount of cover
crops present as well as when the NPN constraint is lifted, as switchgrass
produces all fiber needs for ruminants in this case. Eliminating LPC production
also nearly eliminates alfalfa, as the cover crop is used solely as cattle feed in
these two scenarios. Alfalfa production is high when no cover crop is produced,

animal feed requirements increase, or no AFEX feed is used.

Table 8.4 : Land use (in hectares) of the five crops under the scenarios studied.

 

 

Corn Soy Canola Alfalfa Switchgrass

0 29.8 7.7 2.3 3.9 35.5
A 29.8 9.0 1.0 0.0 39.4
B 29.8 7.7 2.3 3.9 35.5
C 29.8 7.7 2.3 12.4 26.9
D 29.8 7.7 2.3 13.0 26.3
E 28.6 9.5 1.7 2.9 36.5
F 29.8 9.9 0.0 1.7 37.7
G 29.8 7.7 2.3 2.6 36.7
H 25.4 14.0 0.4 0.1 39.2
I 26.5 12.4 0.9 1.1 38.3
J 29.8 7.7 2.3 0.0 39.4
K 29.8 7.7 2.3 7.4 32.0
L 25.4 14.0 0.4 0.1 39.2
M 29.8 7.7 2.3 12.1 27.3
N 25.4 14.0 0.4 8.5 30.8

 

176

One interesting metric in these scenarios is the percentage of digestible nutrients
in ruminant diets, as seen in Figure 8.3. Here, the base scenario had the most
digestible diet, matched only by the high switchgrass yield scenario (this scenario
only affected ethanol production and not animal feed makeup) and the high LPC
yield scenario. The lowest TDN diet is the scenario in which no advanced feed
technologies are produced, as the digestible nutrients in nonruminant diets
decrease by five percentage points compared to the base case. This mirrors the
discussion in Chapter 5, and indicates that hidden benefits in the efficiency of
meat or dairy production may be present in new feeding technologies. If, as this
model states, ruminant diets overall are more digestible, then the cattle
performance (growth or milk production) would not be as limited by the amount
they are able to eat per day. Again, AFEX treated feeds show greater
improvements in %TDN compared to LPC production relative to scenario N,
which is unsurprising given LPCs tend to displace soy meal, which has the same

TDN value as LPC.

177

80% ”g 7" ' ﬂ'—' -—
78°/o f '7 ’78;— —
76% '
74%
72%

Ruminant TDN Diet

             

70% - * :

 

 

 

 

 

 

 

 

 

 

 

 

68%

 

0ABCDEFGHIJKLMN
Scenario

Figure 8.3 : Amount of digestible nutrients in ruminant diets as a percentage of
dry matter intake.

8.4 Conclusions

In summary, nearly half of the cropland currently devoted to animal feeding can
be diverted to cellulosic ethanol production without impacting feed production.
Most of this land is freed by adding and harvesting cover crops to corn, soy, and
canola land. However, an additional 43 million hectares can be devoted to
ethanol production with the addition of LPCs and AFEX feeds. While these two
technologies only result in an 18% increase in potential ethanol production, they
may be vital to the economics present. Growing and harvesting a cover crop
may not be economical without LPC production, which dramatically reduces the
potential for ethanol as seen in Scenario C. Likewise, the potential for high
profits from AFEX treated feeds increases the possibility of collecting

lignocellulosic material for both feed and ethanol production. Thus, scenarios K-

178

M do not supply the full picture for the impact of these technologies on ethanol

production due to the limitations of this model.

These results clearly indicate the potential for biofuel production within the United
States without harming feed production. Nearly 35% of current US gasoline
consumption can be displaced on cropland alone while still maintaining the total
meat and dairy requirements. Additional biofuels can be produced on rangeland,
forest land, cropland used for exports or human use, and idle land. The
improvements are primarily due to intensive farming rather than new
technologies; however, these technologies may enable the intensive farming to

be profitable.

179

CHAPTER 9 : CONCLUSIONS AND RECOMMENDATIONS

9.1 Conclusions

Based on the experimental results and models presented, AFEX-treated fibrous
feeds have strong potential as a method to integrate feed and fuel production.
Preliminary experimental results were promising for a wide variety of feedstocks,
ranging from a 50 g NDF/kg BM increase in digested fiber for alfalfa to 350 g/kg
for late harvest switchgrass. According to the direct land use model, AFEX-
treated switchgrass can displace large amounts of animal feed land and is highly
profitable relative to biofuel production. These advantages are robust, as the
value of this approach tends to remain strong among a wide variation in unknown
variables. There is also a large potential market for AFEX-treated feeds, as up to
85 Tg can be annually produced within the United States according to the
aggregate land use model compared to only 52 Tg of harvested biomass for

protein production.

The nature of AFEX feeds also allows for flexibility in commercializing and
marketing the technology. As stated in Chapter 5, certain feedstocks such as
early harvest switchgrass or corn stover may be competitive with corn grain,
while other feedstocks such as late harvest switchgrass can be competitive with
high value forages. Because AFEX treatment improved digestibility in multiple
feedstocks, there is potential for commercialization throughout the United States

and the world. Furthermore, standalone AFEX treatment facilities can market

180

their product to ethanol facilities as well, opening multiple markets to

commercialization.

Despite the strong potential for commercialization, there is a large knowledge
gap for this technology before it is viable. Although the in vitro studies show

large improvements in digestibility, the results must be replicated in vivo. Full
animal feed trials must be completed in order to determine the actual value of

AFEX treated feeds.

Leaf protein concentrates also appear to be an economically viable integration
technique, although the amount of land saved due to this technology is less than
that of the AFEX-treated fibrous feeds. Using the remaining fiber for ethanol
production rather than animal feed eliminates the cost of drying the fiber and
evaporating the whey. Furthermore, the productivity of the land increases if the
fiber is used as ethanol, as early harvest extracted switchgrass would displace
highly productive corn land. Aqueous protein extraction combined with
enzymatic hydrolysis of cellulose was able to solubilize virtually all of the protein
in switchgrass. However, only 35% of this protein could be concentrated using
ultrafiltration. According to the direct land use model, a hectare of switchgrass
could conceivably displace 0.24 ha of soybean while still producing a biofuel

such as ethanol.

181

Both yield and economic value were not appreciably higher with alkaline
extraction and ultrafiltration than literature values for mechanical pressing and
steam injection. This latter method of producing leaf proteins can be
commercialized in conjunction with biofuels, although potential for improving the
technology of LPC production is still possible. Likewise, it is unlikely that

integration with AFEX or other pretreatments will improve protein yield or quality.

9.2 Recommendations for Further Study

Based on the results of this study, further research into animal feed integration
with biofuel production is warranted. For AFEX treated feeds, research into
potential ammonia-based toxins is required before feeding trials can be
performed. Determining the amount of 4-methylimidazole formed during AFEX is
necessary to insure the concentration is below levels mentioned in the literature
that caused toxicity in cattle. If too much imidazole is produced, then research
into reducing these levels, either through performing AFEX at milder conditions
or preprocessing the biomass, is required. Likewise, a complete mass balance
of all ammonia based compounds produced during AFEX is preferable in case
other toxins are present. This research can coincide with fundamental research
into the kinetics of ammonia-biomass interactions, and would additionally provide

more information into the nutritional value of non-protein nitrogen within the feed.

Another possibility for future research is to determine the fermentation profile for

AFEX treated feeds. The relative ratios of volatile fatty acids produced can help

182

determine the value of these feeds, particularly for lactating dairy cows and
fattening steers. In addition, the amount of methane produced can be quantified
in these fermentations. This can be particularly important given the potential
environmental regulations surrounding feed and fuel production. The quality of
the fiber can also be assessed to determine if the AFEX treated feed can

produce a fibrous mat required for a healthy rumen.

Finally, feeding trials are required to insure that AFEX treated feeds are healthy
and effective, and would also help assess how such feeds could be produced
commercially. Important considerations are the particle size of the biomass
being used, whether or not pelletization after AFEX is required or desirable, and
if the treated fiber needs to be dried for storage. The length of the fibers may be
a tradeoff between the ease of transferring biomass within the treatment process
and the value to the cattle. AFEX treatment is likely to improve the stability of the
fiber, which could reduce or eliminate a costly drying process. Additionally, these
trials would improve the economic modeling, as they would provide a better
estimate of the costs of the process as well as the potential selling price of the

fiber.

For leaf protein concentrates, further research into pretreatment and hydrolysis of
extracted biomass is needed to improve sugar yields. Likewise, this research
can be expanded to include feedstocks likely to be used for protein production

such as alfalfa and viable winter cover crops. Potential fiber processing research

183

includes re-optimizing pretreatment conditions for the fiber product and
determining if glucan yield can be increased to levels comparable with non-
extracted biomass. The impact of drying the material after extraction in
comparison to pretreating the material at its native pretreatment may also be
considered. Separate research into changes in inhibitor formation during AFEX

due to a pre-extraction would also be useful for advancing biofuel production.

Further research into aqueous alkaline extraction is required in order to become
economically competitive with mechanical protein extraction. The impact of lignin
or ammonia-based degradation products on the quality of protein concentrates
obtained from a post-AFEX extraction or from the hydrolysate or fermentation
media is currently unknown. Simple pepsin and trypsin digestibility tests may
provide some information, although small scale animal feed trials may be
performed as well. In addition, a comparison between protein recovered after
hydrolysis and after fermentation can be performed to determine the effect of
microbes on the recovered protein. This approach does not necessarily require
an initial protein extraction, and can be used on late harvest materials as well if
the microbes can produce enough protein from residual ammonia compounds

following AFEX pretreatment.

Finally, research into reducing protein degradation in order to improve

ultrafiltration yields may be performed. This may require research into fresh

harvests to determine if the proteins are degraded prior to initially drying the

184

biomass. Likewise, protein degradation during hydrolysis and fermentation must
be closely monitored and controlled if possible due to the poor yields in
concentrating the hydrolysate media. If it is impossible to keep the protein size
above 10 kDa, then research into smaller filtration cartridges may be necessary.
Concern here is that the protein content of the final product may be too low. If
so, some pre-processing may be required to hydrolyze any polyphenolics or

other compounds that were concentrated.

With improved experimental research, the models can be updated to improve
their reliability. Likewise, the models themselves can also be refined. Because
no pilot scale facility of AFEX pretreatment and particularly ammonia recovery
system has been constructed, the costs associated with pretreatment are only
estimates. The ammonia recovery system modeled here is a novel approach
that does not currently have experimental validation. Replacing this model with a
more conventional ammonia recovery system may be more applicable in the
near term. Furthermore, improved information on handling for AFEX treated
materials and designs for their use as animal feeds can be included as well. This
can result in one or multiple models of a regional biomass processing center,
which may or may not include protein production. An RBPC model could expand
the scope of the land use model, allowing for multiple scales of operation to be
considered at once. Likewise, including geographical considerations into both

the direct land use model and aggregate model would allow for more accuracy.

185

These considerations include changing biomass yields and the feasibility of

certain operations in different climates.

186

APPENDIX A: DISTILLER’S GRAIN HYDROLYSIS

A.1 Introduction

While the primary aim of this research was focused on switchgrass and other
cellulosic feedstocks, an interesting opportunity for co-producing food and fuels is
through distiller’s grain fractionation. The dominant process for producing grain
ethanol in the United States is the dry-grind process. During this process,
distiller’s dry grain and solubles (DDGS) is created as a coproduct generally sold
as an animal feed. The various forms of distiller’s grain have been cited as being
ideal for early cellulosic ethanol production due to the ability to integrate into
starch-based facilities and the low lignin content of DG [116; 117]. Such a
process may also lower the risk of saturating the market for DGs, which would
their selling price and thus affecting the viability of the industry. In addition,
removing the fiber from DG increases the possibility of using the remaining

material as feed in the swine and particularly poultry markets.

Despite the potential of this resource, little effort has been placed in studying
enzymatic hydrolysis and fermentation of DG before this study. Furthermore,
due to the high quantity of protein present, it is prudent to use mild pretreatments
on the grains in order to prevent degradation of this protein. Thus, AFEX is a
strong contender for such a first generation refinery. The goal of this chapter,

then, is to assess the potential of AFEX for DG hydrolysis.

187

There are several concerns specific to DG pretreatment and hydrolysis that
needs to be addressed. First, the variability of DO, both in types and across
various refineries, may result in differences in both yield and conditions. Belyea
et al. found significant variations in DDGS composition from one ethanol facility
across five years of operation, including both the fiber and starch components
[12]. While research into DDGS is common due to its ability to be stored
indefinitely and ease of use in handling, a first generation ethanol facility would
likely use wet DG or DGS as a feed source in order to eliminate the costly drying
operation. The high moisture content of these materials, as well as the different
composition of DG vs DGS, may require changes in AFEX conditions and result

in different yields.

All forms of DG also have relatively low fiber content compared to traditional
lignocellulosic feedstocks. In general, 30—40% of DG consists of carbohydrates,
compared to 50-70% for grasses or woody materials [12]. Ethanol production
requires high ethanol titers in order to make distillation economically viable.
However, since DG fiber content is so low, a higher initial DG solid loading is
required in order to achieve high ethanol concentration. This may create mixing

issues as well as the traditional problems associated with end product inhibition.

Because DG is a unique Iignocellulose source, it may also require unique

enzyme mixtures to effectively break down the structure. The hemicellulose in

corn grain is a complex arabinoxylan structure, consisting of a xylan backbone

188

with several branching and crosslinked chains [118]. The xylanases found
naturally in enzyme cocktails such as Spezyme GP or Multifect Xylanase may not
be sufficient. Research performed cocurrently with this project at the USDA
National Center for Agricultural Utilization Research has shown that a
combination of pectinase and ferulic esterase are required to effectly break down
hemicellulose [69]. In addition, a significant portion of the glucan in DG is starch,
which requires amylases to remove. However, these enzymes are known to be

inhibitory to cellulase enzymes.

Finally, throughout the pretreatment and hydrolysis, careful attention must be
given to the protein content present in the DG. The revenue associated with
increased ethanol production per ton of corn due to DG processing will be less
than the revenue lost from DG sales. Thus, it is vital that the remaining material
in DG, particularly the valuable protein, retain its value throughout the process.

This scrutiny will be further discussed in Appendix B.

Thus, the objectives of this chapter are as follows:
0 Optimize AFEX conditions for both DDGS and WDG
. Assess the variability of DDGS, both from different vendors and different
batches of material
0 Determine the effect of high solid loadings

0 Investigate cellulase and amylase addition

189

A.2 Materials and Methods

A.2.1 Material Source

Several sources of DG were used throughout these experiments. Unless
otherwise indicated, the DDGS or WDG from Big River Resources, LLC. Two
separate batches of DDGS were obtained from Big River, one in 2005 and the
second in 2006. Unless otherwise indicated, the 2005 batch was used for these

experiments. The WDG used was obtained in 2005 as well.

A.2.2 Biomass Pretreatment

For early experiments focusing on changing AFEX conditions, the AFEX
pretreatment process was performed in a 300 mL stainless steel pressure vessel.
When not performed at the native moisture content, water was added to DDGS
and evenly mixed before loading it into the vessel. Glass spheres were added to
minimize void space, thereby reducing the amount of ammonia in the vapor
phase within the reactor. The lid was bolted shut, and a separate cylinder loaded
with the proper amount of liquid anhydrous ammonia was connected, allowing
the ammonia to be charged into the vessel. The reactor was heated using a
400W PARR heating mantle, and allowed to stand at the desired temperature
(+/- 1°C) for five minutes. The pressure was explosively released by rapidly
turning the exhaust valve. The treated samples were removed and were placed

in a fume hood overnight to remove any residual ammonia.

190

With improved AFEX process designs, high solid loading experiments were
performed using rapid heating of biomass. Both the biomass and ammonia were
heated prior to the introduction of ammonia into the reaction vessel. In addition,
the temperature of external heating mantle is limited to 20°C above the desired
temperature in order to reduce burning or charring. After ammonia addition, the
biomass rapidly heats to the desired temperature within one minute. The total

residence time before the ammonia is released is 15 minutes.

A.2.3 Enzymatic Hydrolysis

The enzymatic hydrolysis procedure is based on that described in Chapter 3. An
amount of biomass equal to 0.15 g cellulose was placed in a vial and brought to
a total volume of 15 mL with autoclaved water. The solution was buffered to pH
4.8 by 0.75 mL 1M citrate buffer. Spezyme CP (Genencor, Palo Alto, CA)
cellulase was loaded at 16.5 FPU/g glucan (31 mg protein/g glucan), and B-
glucosidase (Novozyme 188, Bagsvaerd, Denmark) at 56 pNPGU/g glucan. One
experiment involved the amylase Stargen (Genencor) added along with the
cellulases. All samples were incubated at 50°C with 75 rpm rotation. Samples

were collected at various time points after hydrolysis, generally 24 and 72 hours.

A.2.4 Analytical Methods
Sugar analysis was done using a Waters High Performance Liquid
Chromatograph (HPLC) system equipped with a Bio-Rad (Richmond, CA)

Aminex HPX-87P carbohydrate analysis column. Degassed HPLC water with a

191

flow rate of 0.6 mL/min was used as the mobile phase, while the temperature in
the column was kept constant at 85°C. A Waters 410 Differential Refractometer

was used to measure the peaks obtained.

A.3 Results and Discussion

A.3.1 DG Composition

A complete composition analysis on the distiller’s grains and wet cake of four
different refineries is seen in Table A.1. Mass closure for the wet DG samples
were near 100% except for the third sample. In general, there were not large
differences in protein, glucan, or xylan content (<10% difference) between the
first three refineries, indicating a reasonable consistency among separate
feedstocks. The fourth sample shows significant differences, both in the dry
matter content as well as the overall composition. This sample was taken after
the solubles were added, making it wet DGS rather than wet DG. This may be
an acceptable alternative to separating wet cake and solubles, and thus is also a
potential feedstock. All four samples of DDGS were also consistent, although
sample 1 has a lower protein and higher glucan content than the other samples.
Mass closures greater than 100% are due to the presence of protein and ash in
the water extracts. As seen in Appendix B, only a small portion of protein is
water soluble, and the mass closures would be near 100% if 20-25% protein is

soluble, which is in the range reported by various studies.

192

Table A.1: Composition of four different distiller’s grain samples used in this
study. Sample 1 was obtained from Big River Resources and is used in all
experiments in this chapter. Sample 4 was obtained from the Woodbury, Ml
plant of US Bioenergy and used for all experiments in Appendix B.

 

DDGS
1 2 3 4
Water Extractives 23.1 23.0 24.6 21.7
Ether Extractives 11.4 11.9 11.3 10.0

 

 

 

 

 

Crude Protein 27.3 31.1 31.7 32.0
Glucan (total) 21.3 19.3 19.4 19.4
Xylan and Arabinan 18.9 17.6 16.7 18.5
Xylan 12.5 11.6 11.1 12.1
Arabinan 6.4 6.0 5.6 6.3
Ash 4.5 4.4 4.3 4.4
Mass Closure 106.5% 107.4% 108.0% 105.9%
Wet DG
1 2 3 4
Water Extractives 8.1 10.4 3.6 20.2
Ether Extractives 8.1 8.2 7.0 11.4
Crude Protein 33.7 36.4 36.0 29.6
Glucan (total) 22.1 20.3 21.1 18.1
Xylan and Arabinan 26.9 23.9 24.2 15.8
Xylan 15.8 14.7 15.3 10.3
Arabinan 11.1 9.1 9.0 5.5
Ash 2.2 2.1 1.9 5.5
Mass Closure 101.1 101.2 93.9 100.5

 

A.3.2 Optimizing AFEX Conditions for DDGS

The effect of pretreatment temperature on glucose released after enzymatic
hydrolysis is seen in Figure A.1. Glucose yields increase sharply from 60°C to
70°C, while further temperature increase has a statistically insignificant (p<0.05)
effect at 72h. Further temperature increases do appear to slightly increase the
rate, as the glucose released after 24 hours was higher at 80°C than 70°C. At

100°C, the grain appeared to be burnt, although this did not affect glucose yields.

193

In all cases from 70°C to 100°C, greater than 100% of the theoretical glucose
yield from cellulose can be obtained after 72 hours of hydrolysis. Thus, there are
two possible explanations for obtaining yields in excess of 100%: either the
measured cellulose composition is lower than its actual value or there is some
starch hydrolysis occurring. It was found that the B-glucosidase complex used in
this study, Novozyme 188, has some amylase activity, making it likely that at
least a portion of the residual starch in the DDGS is hydrolyzed as well.
200
175 -
150
125 -
100 -
75 ,
50 7~
25 7 ~

 

Glucose Yield (glkg DBM)

 

 

 

 

 

 

 

 

 

50 60 70 80 90 1 00 unt
Temperature (°C)

Figure A.1: Effect of temperature on the glucose yield of AFEX treated DDGS.
During AFEX, water loading was held constant at 0.13 g/g biomass, and
ammonia loading at 1.0 g/g biomass. Enzymatic hydrolysis was performed at a
cellulose loading of 1%. All runs were done in duplicate and error bars represent
the maximum and minimum values. Untreated (unt) is shown as a control.

194

 

 

200 *
175 0 '
150 -
125 *
100 *

U1 \1
0 U1
1 I

 

Glucose Yield (glkg DBM)

M
01

 

 

 

 

 

 

 

O

0.6 0.8 1 1 .2 1 .4 unt
Ammonia Loading (g/g DBM)

Figure A.2: Effect of ammonia loading on the glucose yield of AFEX treated
DDGS. Moisture content and temperature during AFEX were held constant at
0.13 g/g biomass and 70°C, respectively. Enzymatic hydrolysis was performed
at a cellulose loading of 1%. All runs were done in duplicate and error bars
represent the maximum and minimum values. Untreated (unt) is shown as a
control.

Relatively low ammonia loadings are also needed for DDGS hydrolysis, as seen
in Figure A.2. There is a slight increase in glucose yield from 0.6 to 0.8 g NH3/g
biomass, and a slight decrease from 0.8 to 1 g NH3 / g biomass. Yields
dramatically decline at higher ammonia loadings. It is not immediately clear why
high ammonia loadings reduce glucose yields, although it may be due to an
increase in competing reactions such as those between ammonia and sugar.
These trends are similar to other types of easily digestible biomass, including

early harvest switchgrass (see Chapter 3). Approximately 190 9 glucose was

released after 72 hours at the optimal conditions, representing 81% of the total

195

glucan (starch + cellulose) conversion. When hydrolysis was extended to 168 h,

the increase in glucose released compared to 72h was not statistically significant.

At 70°C and 1:1 ammonia to biomass ratio, the amount of water added had little
effect on glucose yields, as seen in Figure A.3. Here, the total water added to
the biomass varied between no additional water and up to 60% water on a total
weight basis (1.5 g water per g biomass). This high value of total water addition
is approximately the moisture content prior to drying the material. No significant
effect (p<0.05) of water addition is seen for either 24 or 72 hour glucose yields.
With little lignin present in DDGS, the primary effect of AFEX on cellulose
conversion is likely the swelling and decrystallization of cellulose. The presence
of water appears to be irrelevant to these reactions. Despite the similar glucose
yields, AFEX treated DDGS at high moisture appeared much darker than low
moisture samples. As this did not effect glucose conversion, the darker color is
likely due to increased reactions with other components in the biomass, including

possibly protein.

While AFEX improved yields over untreated material, high glucose yields were
obtained from untreated DDGS as well, at nearly 150 g/kg biomass. This is most
likely due to the relatively low lignin and cellulose content as well as the effect of
the dry grind ethanol process. The presence of lignin can create inhibitory
byproducts during pretreatment such as coumeric acid. In addition, lignin that is

not removed to the surface of the biomass prevents enzymes from reaching the

196

cellulose, and enzymes may also bind to lignin and deactivate, lowering the
overall enzymatic activity of the hydrolysis. During the dry grind process, the
corn fiber is naturally pretreated, being subjected to milling and elevated
temperatures during the steeping, distillation, and drying processes. These act
as mild pretreatments, improving sugar yields without undergoing AFEX or other

processes required for lignocellulosic biomass.

200
175 0* *
150 *8
125
100 * r
75 --
50 'ﬁ
25

 

 

Glucose Yield (glkg DBM)

 

   

 

 

 

 

 

 

0.13 0.36 0.58 0.80 1.50 unt
Water Content (g/g DBM)

Figure A3: The effect of water content during AFEX on glucose yields from
DDGS. AFEX temperature was held constant at 700 and ammonia loading held
constant at 1 g/g biomass. Enzymatic hydrolysis was performed at a cellulose
loading of 1%. All runs were done in duplicate and error bars represent the
maximum and minimum values.

Optimal AFEX conditions for DDGS were determined to be 70°C, no additional
water, and 0.821 kg/kg ammonia loading. These are fairly mild compared to

optimal conditions for other types of biomass. As untreated material is already

digestible, extreme conditions are not needed. Under these conditions,

197

approximately 1909 glucose/kg dry biomass is released after 72 hours. Both the

rate and extent of digestion is improved over untreated material.

A.3.3 WDG Optimization

Wet distiller’s grain was obtained from the same source as DDGS, and separate
experiments were conducted to determine the optimal AFEX conditions. As this
material was already at high moisture content, only temperature and ammonia
loading were varied. Furthermore, in order to more accurately compare DDGS
with wet DG, the effect of reaction temperature and ammonia loading was
determined for DDGS at high (60% total weight, or 1.5 9 water/ 9 biomass)

moisture content.

Increasing temperature increased glucose yields in wet DG up to 80°C, as seen
in Figure A.4. More glucose was released at 72 hours at 90°C than either 80°C
or 100°C, but no significant difference was seen at either 24 or 168 hours. As
with the dry DDGS, the glucose yield for high moisture DDGS increased between
60°C to 70°C and remained constant at higher temperatures. No reduction in
glucose yields at elevated temperatures was seen for either sample in the range
of temperatures tested. Lower temperatures are generally required to reduce
inhibitor formation or reduce the risk of damaging protein, and so no high

temperatures were tested.

198

200 . , ,

 

_L
\l
01

150 ~
125
100

01V
001

 

Glucose Yield (glkg DBM)

    

E] W00.
I Wetted DDGS]

N
01

 

 

 

 

 

O

 

60 70 80 90 untreated
Temperature (°C)

Figure A.4: The effect of AFEX pretreatment temperature on glucose yields for
both WDG and wetted DDGS. The water content was 1.5 glg biomass for both
samples, and the ammonia loading was 1.0 g/g biomass. Enzymatic hydrolysis
was performed at a cellulose loading of 1%. All runs were done in duplicate and
error bars represent the maximum and minimum values.

Interestingly, the amount of ammonia present did not appear to significantly
affect glucose yields for either wet DG or high moisture DDGS, as seen in Figure
A.4. For DDGS, glucose yields also remained constant through the range shown
in the figure, although lower yields were seen at high (1.6 g/g) ammonia loadings.
For wet DG, no decline in glucose yields was seen up to 1.2 9/9, although no
conditions were tested at higher ammonia loadings. Research into
understanding the interactions between ammonia and water on lignocellulosic
material is ongoing, so it is not immediately clear why the addition of water allows

higher ammonia loading to be used relative to the low water DDGS. This may be

due to lowering the concentration of ammonia ions or the pH of the system.

199

 

75 ,
5o ,--

 

    
 

Glucose Yield (glkg DBM)
3
0

III WDG 1
l Wetted DDGSl

N
01

 

 

 

 

 

 

0.6 0.8 1 1 .2 untreated
Ammonia (glg biomass)

Figure A.5: The effect of AFEX pretreatment ammonia loading on glucose yields
for both wet DG and wetted DDGS. The water content was 1.5 glg biomass for
both samples, and the temperature was 80°C. Enzymatic hydrolysis was
performed at a cellulose loading of 1%. All runs were done in duplicate and error
bars represent the maximum and minimum values.

While there was no significant difference in wet DG glucose yields between 70°C
and 80°C at 1:1 g/g ammonia loading, and no difference between 0.8:1 and 1:1
g/g ammonia loading at 80°C, the glucose yield at 70°C and 0.8:1 g/g ammonia
loading was lower than these values (145 g/kg biomass). Thus, some interaction
between ammonia and temperature is occurring at low severity conditions. Such
interactions were also seen for highly digestible switchgrass. As the increased
ammonia addition is likely to be more expensive than the increased temperature

(see chapter 6), 0.8 g/g ammonia and 80°C was determined to be the optimal

conditions for wet DG.

200

A.3.4 Variability within DGs

Both wet DG and high moisture DDGS performed similarly in terms of glucose
released at various AFEX conditions. There was no significant difference in the
amount of glucose released at optimal conditions, and both temperature and
ammonia loading affected yields in similar manner. Thus, it does not appear that
drying DGs or the composition of wet DG vs DDGS affects the response of
glucan to AFEX treatment. While WDG does not contain the solubles portion,
the glucan content is similar (see Table A.1) to DDGS. Thus, the oligomeric
sugars and fine cellulose particles found in the solubles do not appear to be more

reactive to enzymatic breakdown than the grain portion.

There are differences in the response to AFEX treatment between wet DG and
dry DDGS, however. As the differences are more pronounced between high
moisture DDGS and no added moisture DDGS than high moisture DDGS and
wet DG, it appears that these differences are due primarily to the additional water
present during pretreatment. The water appears to reduce the adverse effect of
high ammonia loadings. This may be due to reducing the concentration of
ammonia contacting with biomass or to reduce the pH of the system. Yields
were slightly lower for wet material than dry, although this difference was not

significant.

A comparison was made between the two separate batches of DDGS obtained in

different years from Big River Resources. Table A.2 shows the results of

201

hydrolysis on untreated DDGS. At both 48 and 120 hours, the glucose yields for
the new batch of DDGS were about 80% of the values for the old batch of DDGS.
Differences within duplicates were in line with previous experiments. No
significant differences were seen in the composition of the two batches.
However, the second batch was a darker color than the first, which may mean it
was subjected to higher temperatures during the dry-grind process, as the darker
color is likely due to Maillard reactions with the protein. Such differences in
enzymatic response may reduce the likelihood of DGS as a cellulosic ethanol

source, as this increases the economic risks of the process.

Table A.2: Differences in glucose yields released after enzymatic hydrolysis of
two batches of DDGS obtained from the same dry grind facility.

 

 

48 hour 120 hour
2005 DDGS 138.1 153.1
2006 DDGS 108.0 122.1

 

A.3.5 Enzyme Addition

Although glucose yields were high throughout all previous experiments, xylose
yields were negligible. Multifect Xylanase was added at up to 50% of the
cellulase loading. However, no xylose was detected at any level of xylanase
loading after 168 hours, meaning less than 20% of the xylan was hydrolyzed.
Separate experiments performed at USDA NCAUR (Peoria, IL) determined that
Multifect Pectinase and Ferulic Esterase could convert 63% of xylan and 98% of

arabinan into monomeric sugars [69].

202

 

200 '

—L
01
O

Glucose Yield (glkg DBM)
0 S
o o

 

 

 

 

 

O

i ‘ l 1
Cellulase Amylase Both Cellulase,Amylase Both

 

AFEX Untreated

Figure A.6: Effect of amylase on glucose yields for both AFEX treated and
untreated DDGS. Cellulase loading was 16.5 FPU/g glucan, while amylase
addition was added at 14mg enzyme/g glucan. B-glucosidase was added at 56
pNPGU/g glucan in all cases.

As the dry grind process did not obtain complete conversion to ethanol, DDGS
contains some residual starch. Thus, attempts were also made to increase the
glucose yields by the addition of amylase. However, as seen in Figure A.6,
adding Stargen amylase at 14 mg enzyme per gram of glucan to the cellulase
cocktail did not significantly (p<0.05) affect the overall glucose yield. As stated
previously, a portion of the starch may already be being broken down by the
cellulase enzymes, thus reducing the need for additional amylase. Novozyme
188 contains some amylase activity, producing monomeric glucose from starch.
Moreover, the amylase is likely inhibiting the cellulase by binding to cellulose
sites, thus decreasing cellulose conversion and offsetting any increase in starch

hydrolysis.

203

A.3.6 High Solid Loading

Increased solid loading did not affect the glucose yields 2006 batch of DDGS, as
seen in Figure A.7. However, yields did decline for the 2005 batch. The primary
cause of decreasing yields for cellulosic ethanol is end product inhibition, or the
amount of glucose and cellobiose released [119]. Thus, the lower yields present
in the 2006 batch do not cause as much inhibition. If the end-product inhibition
issues are not resolved, then this may displace the issue of DG variability. Other
factors that may be associated with lower yields, including byproduct inhibitors,
competitive binding on lignin, and poor mixing do not appear to decrease yields.
This is consistent with the mild pretreatment conditions required for DDGS and
its low lignin content. In addition, DDGS solubilizes rapidly even at high solid
loadings after enzymes are added. High solid loading hydrolysis of wet DG
showed similar results to the 2005 batch of DDGS. At 15% solid loading, yields

decreased from 180 g/kg wet DG to 150 g/kg.

At the same solids loading the AFEX treated wet DG gives 34 g/L glucose and 27
g/L xylose as the maximum theoretical concentrations. Enzyme digestion of the
AFEX treated WDG at 15% solids loading with the addition of the auxiliary
enzymes showed a slight increase (from 68% to 72%) in the glucose yield and
approximately 4 times increase in the xylose yield (from 12% to 45%) as
compared to the case without addition the xylanase and feruloyl esterase

enzymes. The rise in glucan conversion as well as xylan is likely due to

204

synergistic effects between the different enzymes. As more hemicellulose is
hydrolyzed, this likely increases the glucan susceptibility for attack, thereby

slightly improving glucose yield as well.

 

75

01
O
1

 

Glucose Yield (glkg DBM)

25

 

 

 

Solids Loading (%)

Figure A.7: The effect of solid loading during enzymatic hydrolysis on glucose
yields from AFEX treated DDGS. Two separate batches of DDGS were used. In
all cases, the optimal treatment of 70°C, 0.8 g ammonia/g biomass, and no
additional moisture was used. Samples were collected after 120 hours.

A.4 Conclusions

Our experimental results show that the AFEX process is an effective
pretreatment for the enzymatic hydrolysis of distiller grains. High yields of
glucose were obtained from relatively mild AFEX conditions for both wet DG and
DDGS. Xylose yields were negligible, and require further enzymes to release
monomeric pentoses. Sugar yields did vary across different types of distiller

grains as well as different batches of grain.

205

These results suggest that dry grind refineries can improve their overall ethanol
yields from corn by hydrolyzing the fiber as well as the starch. The primary
impediments to this approach appear to be the low pentose yields and difficulty in
achieving high sugar concentrations. Both of these problems can be solved by
co-fermenting the DG hydrolysate with the starch hydrolysate. The pentoses will
not be fermented, although the lost revenue will be offset by removing the costs
of additional enzymes or the difficulty in co-fermenting glucose and xylose [120].
The low sugar concentration from the DG hydrolysate will be offset by the high
concentration during starch hydrolysis. Conversely, the fiber may be hydrolyzed
simultaneously with the starch, an option not considered here. The third primary
issue, variability within DG, is something that must be confronted by individual
refineries. It remains to be seen whether the use of DG as a feedstock for

ethanol will be an economically attractive option for all or some refineries.

206

APPENDIX B: DISTILLER’S GRAIN PROTEIN FRACTIONATION

8.1 Introduction

As seen in the previous chapter, distiller’s grains show strong potential as a fiber
source in first generation cellulosic ethanol facilities. Despite this potential, the
remaining protein must retain its value as a protein source. Ethanol from corn
fiber and residual starch can increase the overall ethanol yields of an integrated
starch-fiber facility by 11%, but the revenue from DG sales approaches 15-25%
of the total value of current refineries [98; 121]. One option, discussed in the
previous chapter, is simply to sell the remaining fiber, either with or without the

solubles added, as a hydrolyzed distiller grain feed.

However, the possibility of fractionating DG for biofuels allows one to consider
other methods of adding value to the material, particularly the protein. The
protein in DG is limited in value as an animal feed by its low lysine content, and
thus it may be possible to fractionate the protein into low and high lysine
products. The high lysine material can be used as a feed while the low lysine
used for other purposes. If the low lysine product is easily separated from the
rest of the grain, it can be used for value added products such as bioplastic
precursors. If it is not easily separated, it and the remaining material may be

burnt for energy.

207

Likewise, the high lysine product’s value will be dependent upon the other
material obtained with it. High amounts of lipid or phosphorous can reduce the
value of DG, particularly if their concentrations increase with reduced fiber
content. Thus, obtaining a high lysine protein that is relatively pure will likely be
of the highest value. As the low lysine protein is primarily zein, which is
hydrophobic, aqueous extraction of protein should produce a solubilized protein
higher in lysine. This protein would then be concentrated and be relatively free of

fiber, lipids, and high levels of ash.

Another method of separating protein is through enzymatic extraction using

proteases. As the solubilized protein will be broken down into peptides or amino
acids, these peptides are not useful as animal feed, and instead should be used
as precursors for value-added products [122]. Thus, in this situation, the protein

susceptible to enzymatic attack should be low in lysine.

The primary objective of this study, then, is to determine if fractionation of protein
is feasible and, if so, what methods are preferable. The specific objectives are
as follows:
0 Optimize protein extraction using aqueous or alkaline solvents with
particular emphasis on the amount of lysine extracted
. Integrate protein extraction using solvents with AFEX
0 Determine the effectiveness of protease extraction on distiller’s grain

0 Determine the recovery of lysine in all methods tested

208

B.2 Materials and Methods

8.2.1 Feedstock

For aqueous extraction, fresh wet distiller’s grain and solubles (DGS) was
generously donated by the US Bioenergy ethanol plant in Woodbury, MI on
October 4, 2007. Samples were kept frozen at -20°C until use. The composition
of DGS is summarized in Table A.1 in the previous experiment. For protease
experiments, the feedstock tested is wheat heavy stillage, obtained from
Wheyfeed Ltd. (Nottingham, UK). This stillage contains the residue, both solid
and soluble, in a dry grind process immediately after distilling off the ethanol.
The dry weight of the stillage was 14.4% of the total weight, with the insoluble
portion of the stillage at 10.4% of the total weight while the solubles contributed
4%. Nitrogen analysis gave a total protein content of 37.8 mg protein per g

stillage (270 mg protein per g dry weight). The pH of the stillage was 3.5.

32.2 Pretreatment

The AFEX pretreatment was performed in a 1.5 L stainless steel pressure vessel
preheated to 130°C. Approximately 150 g (dry weight) wet cake was added to
the vessel at its native moisture content. The lid was bolted shut and the air
removed using a vacuum pump. A cylinder loaded with 0.6 (+/—0.04) g NH3 per g
dry biomass was connected and the ammonia charged into the vessel. The
temperature of the wet cake reached 120°C in 5 minutes and decreased to
114°C after 15 minutes. The final pressure inside the reactor was 1.45 MPa.

After 15 minutes, the pressure was explosively released. The treated samples

209

were immediately placed in a sealed bag to minimize exposure to oxygen and

thus potentially harmful oxidation reactions.

3.2.3 Aqueous Protein Extractions

Protein extractions were performed in Erlenmeyer flasks. Solutions of sodium
hydroxide at the desired concentrations were allowed to preheat to the desired
temperature before adding to the flask. An amount of wet cake equivalent to 2.0
g dry biomass was added to each flask, and the preheated sodium hydroxide
solutions were added to bring the total weight to 40 g. When B-mercaptoethanol
was used, it was added immediately prior to the extraction. Extractions were
performed in shake flask incubators set at the desired temperature for 1 hour and
shaken at 200 rpm. After extraction, the samples were centrifuged and the
supernatant removed. Samples of the supernatant were removed for protein
analysis. Pellets were weighed and allowed to dry to determine the total water
remaining in the pellet. It was assumed that the concentration of soluble proteins
within the water remaining on the pellet was the same as the concentration in the
supernatant. The additional soluble protein remaining with the water in the
pellets were not included in total protein extracted when integrating with
hydrolysis, as this would lead to double-counting the protein with the subsequent

operation.

210

8.2.4 Enzymatic Protein Extraction

All protease extractions and hydrolyses were performed in 50mL Erlenmeyer
flasks placed in a shake-flask water bath set at 90 rotations per minute. Unless
otherwise stated, the temperature was held constant at 50°C. An amount of
stillage equal to 2.0 g dry weight was added, and the pH brought up to 7.5 by the
addition of 3.6 mL of 1.0M NaOH. Water was added to bring the mixture to 209
total weight. Although higher concentrations may be used in industry to
decrease energy use downstream, this 10% solid loading was deemed to be an
acceptably high solid loading while remaining easy to handle and remove
samples. Protease was added based on weight, as the operational energy costs
in the model were based on weight rather than activity. Unless otherwise stated,
protease was added at 0.1% w/w loading, or 1g protease per kg stillage protein.

Samples for analysis were taken at 24 hours after the addition of protease.

In certain treatments, the stillage was allowed to incubate for one hour at 70°C
prior to the addition of protease in an attempt to extract extra proteins from the
insoluble matrix. The stillage was brought to the desired pH to either 7.5 or 12
using 1.0M NaOH and placed in a shake flask incubator. After 1h, the flasks
were cooled and the alkaline flasks neutralized with HCI prior to the addition of
the protease. Residence time was measured as beginning from the time of
protease addition. Samples were then taken immediately after neutralization as

well as 24 h after protease addition.

211

8.2.5 Protein Analysis

Due to the presence of ammonia nitrogen resulting from AFEX pretreatment, it is
impossible to use standard nitrogen analysis methods, such as the Kjeldahl or
Dumas methods, to measure total protein content in all samples. For three
experiments, the complete amino acid profile was determined at the
Macromolecular Structure Facility at Michigan State University. Samples were
hydrolyzed in either concentrated acid or base and quantified through high
pressure liquid chromatography. Both the insoluble residues after processing as
well as the liquid extracts were quantified. The amino acids were then summed

together to obtain the protein content.

Due to the cost and time associated with this method, it was impractical to
perform for all trials. Instead, individual amino acids were measured based on an
LC/MS/MS procedure developed at the Mass Spectrometry Facility at Michigan
State University [123]. Protein samples were first hydrolyzed into their
component amino acids using 6 M HCI at 110°C overnight. Deuterated valine, D-

valine-d8 (Sigma) was added as an internal standard. The resulting hydrolysates

were diluted and filtered through a 0.22 pm filter. Samples passed through a
Waters Symmetry C18 LC column before entering a Waters Quattro Micro mass
spectrometer and quantified using QuanLynx software. The flow rate was 0.3
mUmin and the mobile phase was a gradient of 1 mM perfluoroheptanoic acid
and acetonitrile. Multiple reaction monitoring was performed using electrospray

ionization in positive mode.

212

This method allowed the detection and quantitation of eleven amino acids. The
nine amino acids not detected were ala, asn, asp, cys, gly, met, ser, thr, and trp.
From this data, ratios of individual amino acids were determined among different
treatments to determine their variance. It was found that only hydroxide
concentration had a large effect on this variance. For example, in the experiment
determining the effect on temperature at low concentration (Figure B.1a), the
average ratio of proline to glutamic acid in all samples and replicates was 0.518
with a standard deviation of 0.024 and no significant interaction with temperature.
In comparison, this average ratio was 0.717 for high (20.5M) concentrations of
hydroxide and significantly different than the low composition. Thus, a complete
amino acid profile was obtained at the Macromolecular Structure Facility as
stated previously for one representative experiment at both low and high
concentration hydroxide, as well as hydrolysate medium. The ratio of the eleven
detected amino acids to the complete amino acid profile was then used to

estimate the total protein content in all samples.

For enzymatic digestion, the primary method of protein quantification used was
by amino acid analysis [124]. All peptides that were solubilized were assumed to
be available for bioplastic precursors. Samples were centrifuged at 13000 RPM
for five minutes in order to remove suspended solids, and the liquid portion
hydrolyzed into individual amino acids by heating at 110°C in GM HCI overnight.

The acid was removed by evaporating under a vacuum at room temperature and

213

the samples resolubilized in 0.1 N HCI. The amino acids were then derivatized
using 6-aminoquinolyI-N-hydroxysuccinimidyl carbamate (AQC) and separated
using a Waters Nova-Pak C18 amino acid column. The amino acid peaks were
measured and summed together to obtain the total protein. The protein value
obtained was adjusted upward using an amino acid profile averaged from
multiple milling companies to estimate the added weight of the undetected amino
acids (tryptophan, cystein, methionine). The weight of the protease added was
then subtracted from the final amount of protein, as it was assumed that all of the

protease added would be present in the liquid phase.

3.2.6 Fiber Hydrolysis

The enzymatic hydrolysis procedure was based upon the procedure described in
Chapter 3. Samples were hydrolyzed in Erlenmeyer flasks at 10% solid loading
buffered to pH 5.0 by 1M citrate buffer. Spezyme CP (Genencor) cellulase was
loaded at 75 FPU/g DGS (13.6 FPU/g glucan), and B-glucosidase (Novozyme
188) at 320 pNPGU/g DGS. All samples were incubated at 50°C with 200 rpm
rotation. Sugar concentration after 72 hours was determined using a Waters
High Performance Liquid Chromatograph (HPLC) system equipped with a Bio-
Rad (Richmond, CA) Aminex HPX-87P carbohydrate analysis column.
Degassed HPLC water with a flow rate of 0.6 mUmin was used as the mobile

phase, while the temperature in the column was kept constant at 85°C.

214

In certain treatments, samples were hydrolyzed using aqueous extraction after
protein extraction. Solids were separated from the liquid extract through
centrifugation and the mass loss determined using the method described in the
following paragraph. The remaining solid residue was hydrolyzed at 10% solid
loading at 75 FPU/g residue. The extract solution was brought to a pH of 5.0
using 1M citric acid and enzymes loaded at 75 FPU/g DGS solubilized and 320
pNPGU/g DGS solubilized in order to hydrolyze solubilized oligomeric sugars. In
addition, half of the samples were rinsed with approximately 10g water per g

biomass after extraction and prior to hydrolysis.

8.2. 7 Mass Balances

Results are listed either in percent yield or in g/kg biomass. In all cases, the
reference value is the original, untreated dry biomass. Thus, the percent yield for
protein is the amount of protein released divided by 296 g protein/kg dry
untreated biomass. In cases where multiple operations are performed, a mass
balance was used to insure all values were compared to untreated rather than
treated biomass. After each operation, solid/liquid separation was performed
using centrifugation at 4000 rpm for 20 minutes or 8000 rpm for 15 minutes. The
solids were weighed after centrifugation and portions dried overnight at 105°C to
determine moisture content. This information was used to calculate the mass
solubilized at each step, and used to convert protein and sugar values released

from later operations to g substrate/kg dry, untreated biomass.

215

8.2.8 Error Analysis

The extractions testing temperature and alkaline concentration were performed in
duplicate, while all other extractions were performed in triplicate. Error bars
represent the range for duplicates and the standard deviation for triplicates. For
all significance tests, a student t-test was used requiring a probability p<0.05 to

be significant.

8.3 Results and Discussion

8.3.1 Extraction Optimization

The effect of extraction temperature on protein yields is shown in Figure B.1a.
Protein and lysine yields as well as total biomass solubilized increased steadily
as temperature increased before leveling off at 70°C. However, protein yields
were quite small, as no more than 15% of the proteins were solubilized, a result
comparable to literature values. This protein is disproportionately high in lysine,
as 31% of the total lysine was extracted at 70°C compared to 13% of the bulk
protein. It is likely that zein protein, which is deficient in lysine, is not as soluble
as other proteins in DGS, as has been suggested in previous research with com
[125]. Although the solubilized protein is relatively high in lysine, nearly 70% of

the total lysine remains within the insoluble residue.

216

50 F meters”

E Lysine ]
-- qBiomass,

 

  
  

 

 

 

 

Percent Solubilized (%)

 

 

 

 

 

 

30 45 60 70 80
Temperature (°C)

 

70

{I Protein
60“ Lysine ‘77 [-
50 1 ”210,055? 58 '0‘

 

 

 

Percent Solubilized (%)

 

 

 

0 0.01 0.05 0.1 0.5 1
NaOH concentration (M)

Figure 8.1: Effect of extraction temperature and hydroxide concentration on
solubility of protein, lysine, and biomass. All extractions were performed at 0.1 M
NaOH for the top figure and 60°C for the bottom figure on untreated DGS.
Extractions were performed at 20:1 liquid: biomass ratio for 1 h and rotated at
200 RPM.

A much greater effect was seen with the concentration of NaOH used, as given
in Figure 8.1 b. Here, protein yields increased significantly between 0.05 and 0.5

M NaOH, increasing from less than 4% to 22% of the total protein. Further

increases in hydroxide concentration did not result in further improvements in
solubility. The increased alkalinity of the solution was effective in solubilizing
proteins from the previously insoluble residue, as evidenced from the increased
yields compared to water extraction as well as changes in the relative amino acid
ratios of the extracted protein. The stronger alkaline solutions are likely more
effective in breaking down the structure of the distiller’s grain, thereby freeing

more proteins. Lysine solubility also increases in a similar manner.

As hydroxide concentration increased it became more difficult to separate the
solids and liquids after extraction. A film of insoluble biomass less dense than
the liquid appeared at 0.5 and 1.0 M NaOH. This film of biomass was not
considered part of the pellet, and thus partially explains the large amount of
biomass solubilized during extraction. However, nearly 30% of the total biomass
is soluble even at low hydroxide levels. This is consistent with the large amount
of water solubles seen in Table A.1, but this may make downstream processing
of the protein stream more difficult. If the remaining solubles are not separated
from the soluble protein, the protein content within the final product may be too

low to provide maximum value as an animal feed.

218

01
O

 

 

A
o
"I

(D
O

 

 

   
 

Percent Solubilized (%)

 

 

 

204
10~ Ip'rotein
0J0 Ilysine
0.05 0.1 o.2|o.3 0.05 0.1 0.2 0.3
OmM 2mM

SDS concentration (%), BME concentration (mM)

Figure 8.2: Effect of adding a surfactant (sodium dodecyl sulfate) and reducing
agent (b-mercaptoethanol) during alkaline extraction on protein and lysine yields.
All extractions were performed at 70°C using 0.1 M NaOH, a 20:1 liquid: biomass
ratio for 1 hour, and rotated at 200 RPM on untreated DGS.

The effect of adding a surfactant (sodium dodecyl sulfate, or SDS) and a
reducing agent (B-mercaptoethanol) is seen in Figure 3.2. The reducing agent,
which can cleave sulfur-sulfur bonds, did not significantly affect protein solubility
except at an SDS concentration of 0.05%. The concentration of SDS, however,
did have a significant role in protein yield, peaking at approximately 0.2% loading
(w/v). The extent of the increase, however, was to increase protein solubility to
approximately 20%, doubling the amount of protein released compared to no
surfactant. The increase in lysine solubility showed similar trends. It is likely that
as the corn was processed in the dry grind facility, hydrophobic portions buried
within the protein were exposed to the surface, thus limiting their solubility.

Furthermore, the SDS may also be increasing oil emulsification and potentially

displacing protein from the oil-water interface [126]. Boatright and Hettiarachchy

219

[127] found that the presence of oils in soy protein isolates reduced their

solubility in aqueous solutions.

However, the addition of a surfactant did not affect protein solubility at high
hydroxide concentrations and temperatures, as seen in Figure B.3. One possible
explanation is that the surfactant is only aiding in solubilizing the protein fragment
that is already soluble in strong basic solutions while having no effect on the
remaining insoluble protein. Another possibility is that saponification is occurring
between the oils present in DGS and the NaOH at high concentrations. A soap-
like substance has been observed in highly alkaline extractions when no SDS
was added to the extract. This resulting soap may then act as a surfactant, thus

eliminating the need for further SDS addition.

Approximately 40% of the protein and 50% of the lysine was extracted at 70°C
and 1M NaOH, a significant improvement over other conditions tested. The
percentage of protein that is lysine is 4.5% in this extract, compared to 3.5% in
the original DGS. Saponification is also likely to be occurring during AFEX,
although the addition of SDS still improved protein yields in extraction of AFEX
treated grain, as protein solubility increased from 21% to 32% with the addition of
the surfactant. Performing AFEX pretreatment prior to extraction improved
protein and lysine yields when 0.1 M NaOH was the solvent, but not at higher

concentrations.

220

 

\1
O

   

    

           
   

 

 

 

 

 

 

E 60

E 50* ~

3 30 E E

E 20 , E 2

0 g: g:

8 10 E E3 ILysine

“ 04 .1..- % alotamags.
0.1M 0.1M, 1M 1M, ‘0.1M 0.1M,1 1M 1M,
1 1 SDS 1 SDS SDS 1 SDS ‘

Untreated AFEX ]

Figure 8.3: Comparison of protein, lysine, and biomass solubility using different
extraction conditions for untreated and AFEX-treated wet DGS. Either 0.1 M or
1.0M sodium hydroxide was used as the solvent, with or without the addition of
sodium dodecyl sulfate at 0.2% loading. Extractions were performed at 70°C,
20:1 liquid: biomass ratio for 1 hour, rotated at 200 RPM.

The total mass solubilized also increased as protein solubility increased. The
addition of SDS increased the total mass solubilized at 0.1 M NaOH, as an
additional 6% of the biomass went into solution in both the untreated and AFEX
treated samples. Much of this additional mass is the increase in protein
solubility. However, performing AFEX causes an additional 12% solubilization,
indicating that compounds other than protein are solubilized due to AFEX
pretreatment. At high hydroxide concentrations, the total biomass solubilized is
approximately 65%, regardless of whether the grain is treated or a surfactant is

added. At these conditions, only 20% of the soluble material is protein. The

amount of non-protein biomass solubilized clearly indicates that at least a portion

221

of the oil and carbohydrate fractions are among those extracted. If the extract is
to be used as an animal feed, these extra fractions may affect the quality and
therefore the selling price of the feed. In addition, the loss of carbohydrates may
adversely affect resulting ethanol yields, thereby reducing the value of such an

operation.

8.3.2 Integration of Protein Extraction with Hydrolysis

Overall protein solubility is improved when enzymatic hydrolysis is performed
following alkaline extraction, as seen in Figure B.4a. Protein removed during the
extracted step was slightly lower than in previous experiments. This is likely due
the higher solid loading used (10:1 liquidzsolid ratio vs 20:1 for previous
experiments), as more water and thus soluble protein is trapped within the
biomass. These proteins are later released in either the wash phase or
hydrolysis. Approximately 30% of proteins were solubilized for 0.1 M NaOH
samples, only a slight improvement over the amount of protein released during
hydrolysis alone. This is due to a decrease in the protein present within the
hydrolysate, indicating that alkaline extraction at low hydroxide concentrations
works primarily on proteins that are already soluble once the fiber matrix is '
destroyed. In comparison, 0.5 M NaOH was able to solubilize 46% of the total
protein, a significant improvement over 24% protein solubilized without
extraction. Including a rinse step in between extraction and hydrolysis slightly
decreased yields in the three scenarios studied. This may be due to the

presence of added salts formed as the alkaline is neutralized, which may be

222

beneficial for protein solubility during hydrolysis. Furthermore, a noticeable
amount of protein is found in the rinse water, which is unlikely to be economically
recoverable. Although the addition of SDS improved protein solubilized during
extraction at low hydroxide concentrations, the total protein solubilized was
approximately equal. Relative amino acid ratios of the total protein solubilized
were similar for samples with and without SDS addition (data not shown). Thus,
adding a surfactant does not appear to affect total protein solubility when fiber

hydrolysis is performed.

In addition, extracting the biomass also decreased glucose yields in the
hydrolysate, as seen in Figure B.4b. Due to the large amounts of glucose
present in the extract, it is clear that much of the glucan, most likely starch and
oligomers, is soluble under alkaline conditions [128]. This is especially true at
higher concentrations of hydroxide, perhaps due to the greater disruption of the
structure of DGS or to increased solubility of starch. At 0.5 M hydroxide, 42% of
the glucan and 47% of the xylan was removed during the protein extraction
according to acid hydrolysis of the resulting pellet. Thus, while protein removal
improved the percent glucan conversion over unextracted samples (68% vs 55%
of available glucan present), glucose yields were significantly lower. Adding
cellulase enzymes to the extract media released additional glucose, particularly
for the 0.5 M hydroxide extract. Despite this additional glucose, overall yields

were significantly lower than unextracted material for 0.5 M hydroxide, while only

223

one case at 0.1 M hydroxide saw slightly improved glucose yields. Thus, it is

unlikely that extraction is opening up more glucan to enzymatic attack.

From these data, it appears that extraction prior to enzymatic hydrolysis is not an
effective method of DGS fractionation. The protein extracted is separated into
two different streams, as is the sugar released. This could potentially lead to
additional downstream costs due to lower final ethanol concentrations. Although
it is possible that the extract could be used as hydrolysate media after removing
the protein in order to recombine the two sugar streams, this would require
neutralizing the hydroxide. Furthermore, such a process would be unlikely for
extracts with surfactants due to their adverse effects on the enzymes. In
addition, a tradeoff is seen between sugar and proteins solubilized, as stronger
alkaline conditions lead to greater protein content yet lower sugar yields.
Furthermore, protein yields remain below 50% despite their improved lysine

content, thus making any additional value achieved from this process unlikely.

224

 

 

01
O

 

 

 

   

 

 

 

 

E 40 s
'o
o
E 4
3 30 7%
(g 20 4 {it A"... :23; 23:1,: ,
.3 10 ”K '- EJHydrolysate
E; El Rinse
I
O - ‘ Extract
0.1M Rinse 0.1M Rinse 0.5M Rinse
SDS

150 ,___,E_,, s s s z s s -" s s _‘,__~___j
’5 l
E 120 , , , ,, , , , , E ,
33 1
'0 l
a 90 - a a a 4
a l
2 ;
9 .
o 60 - ’ , , , ‘
m l
8 i
g 30 7 IExtract

0 Cl Hydrolysate

 

 

0.1M Rinse 0.1M Rinse 0.5M Rinse No
SDS extract

Figure 3.4: Protein (top) and glucose (bottom) yields for integrated extraction
and hydrolysis. Alkaline extractions were performed at 70°C, 10:1 liquid:
biomass ratio for 1 hour and rotated at 200 RPM prior to AFEX pretreatment and
enzymatic hydrolysis of fiber. Three extraction conditions — 0.1 M NaOH with and
without SDS addition and 0.5M NaOH with no SDS addition — were tested. For
each condition, one set of samples (labeled as Washed) was rinsed at 10:1
liquid: biomass ratio following extraction to remove residual hydroxide and
surfactant. Glucose released in the extract and hydrolysate was determined 72h
after the addition of cellulase enzymes to the media. The enzymatic hydrolysate
from AFEX-treated DGS without a previous alkaline extraction is also shown as a
control.

225

B. 3.3 Integration with Biosolvent Extraction

Attempts were made to use various biosolvents to extract protein simultaneously
with enzymatic hydrolysis of cellulose. Hydrophobic solvents were effective at
solubilizing proteins at 50°C, as seen in Figure 8.5. D-Limonene, a hydrophobic
solvent, removed 45% of the protein in DGS, while the hydrophilic solvent ethyl
lactate solubilized only 30%. The protein removed was primarily zein protein and

therefore low in lysine, unlike the alkaline extractions.

 

IDL
Cl EL

Protein Recovery, %

 

 

 

 

 

 

 

 

30 50

Temperature of Extraction, 0C
Figure 8.5: Protein recovery as a function of temperature for a hydrophilic (Ethyl

Lactate, EL) and a hydrophobic (D-Limonene, DL) solvents. Protein recovery was
calculated based on the initial protein content in DGS, i.e., 30 % (dry basis)

Performing hydrophobic solvent extraction in conjunction with enzymatic

hydrolysis did not harm hydrolysis yields, as seen in Figure 8.6. Yields were

226

low

GSt

imp

tor

loa

U.L: >Vr‘_r‘.- < h"‘-.‘ §E\<:<<II-: Ii

Fi

tel
re
de

me

lower than the control (no solvent added) for limonene, but similar for the methyl
esters. In addition, when only cellulase was added, methyl ester addition
improved sugar yields. Overall glucose yield was low, however, potentially due
to end product inhibition. As the solvents formed a separate layer, the DGS

loading was 33%.

 

 

100
Cellulase and Amylase

m D Cellulase
8 80 s
>2:
hi
kl
‘1 60 -
2‘
c
E
g0 40 -
U)
8
E.
3 20 —
E

0 _

 

 

 

 

 

 

 

DL DME Blank

Figure 3.6: Effect of the enzyme amylase on the enzymatic saccharification of
cellulosic materials of DGS present in different solutions. The solvents used are
D-Limonene (DL), distilled methyl esters (DME) and buffer (blank). Time = 120 h,
temperature = 50 0C, AFEX-DGS : buffer : biosolvent = 1 : 2 : 1. The data points
represent the mean experimental value and the error bars represent standard
deviation.

Protein solubilization was low, with negligible protein in the solvent phase for the

methyl ester. It is likely that mass transfer issues are preventing the protein from

227

reaching the solvent. As the enzymes break down the cellulosic structure, large
portions of the DGS are solubilized leading to a separate aqueous layer. This
layer prevents the hydrophobic proteins from interacting with the solvent, limiting
its solubility. The insoluble DGS, meanwhile, is denser than water, further
insuring no interaction with the less dense methyl ester. While the hydrolysis is
performed under vigorous rotation, no emulsification was seen during the

reaction.

Thus, integrating biosolvents with enzymatic hydrolysis does not appear to be a
viable option. However, protein extraction prior to hydrolysis provides 45%
protein yield while not decreasing sugar yields. The removed protein is low in
lysine, and thus may not be valuable as an animal feed. However, it could be
used for other high value processes. The remaining insoluble residue would be
high in lysine, and could be used as an animal feed rather than for heat and

power.

3.3.4 Composition of Insoluble Residue

One potential option to fractionate the protein is simply to separate the solubles
from insoluble material after hydrolysis and fermentation of the fiber. A complete
composition analysis of the insoluble fraction of DGS after fiber hydrolysis
compared to the initial composition is shown in Table 8.1. Here, the protein
fraction increases dramatically, to 56% of the total product. The lysine content

also increased, although roughly at the same level as the protein increase. Thus,

228

the relative lysine content is the same as before. Due to the high concentration
of protein and no loss of lysine, this insoluble residue may be a more valuable
animal feed than extracted solubles. Further research on the digestibility of both

protein and specifically lysine within hydrolyzed DG is ongoing.

Table 8.1: Composition of distiller’s grains used in this study. The first column
represents the original wet distiller’s grain and solubles while the second column
represents the insoluble residue after enzymatic hydrolysis of fiber.

 

Component (%) Untreated Hydrolyzed

 

 

Protein 29.6 56.2
Lysine 1.05 1.80
Glucan 18.1 ‘ 7.0
Xylan 10.3 5.0
Arabinan 5.5 2.1
Water Extracts 20.2 8.1
Ether Extracts 11.4 13.6
Ash 5.5 1.6
Phosphorous 1.09 0.23
Total 100.5 93.6

 

Carbohydrate content decreased due to hydrolysis, although the total glucose
yields (120 g/kg biomass) are well below expected (Kim et al., 2008b) for the loss
of fiber. This is likely due to incomplete hydrolysis of the fiber creating oligomers
in the resulting hydrolysate. In addition, natural variance between DG at
separate refineries may account for the lower response to enzymatic hydrolysis.
Mass closure on the insoluble residue did not reach 100%, and the remaining
material is possibly fiber not broken down during acid hydrolysis. Xylan and
arabinan in particular do not appear in the amounts expected due to their low
enymatic digestibility. A more complete set of enzymes may be able to further

break down these oligomers, thereby increasing ethanol yields. The lack of fiber,

229

while decreasing the value as an energy source in ruminants, may serve to

improve ileal amino acid digestibility in swine [129].

Ash content in the insoluble residue also significantly decreased, including
phosphorous, as most of the ash was soluble during hydrolysis. It is unclear
whether a lower composition of ash and especially phosphorus will improve or
detract from the value of the DGS in nonruminant diets. Phosphorus in DG is
highly digestible compared to that in cereal grains, and thus is of great value
[130]. However, high phosphorus levels may lead to toxicity. Furthermore,
phosphorus and calcium requirements are closely linked, even though D6 are
relatively deficient in calcium. Carefully managing the amount of solubles that
are returned to the insoluble residue prior to feeding may be necessary to

balance the mineral requirements for livestock.

Although an oil layer appeared in the hydrolysate, a significant amount of lipids
remain in the insoluble material. This may cause problems with swine diets, as
high fat DDGS has been linked to soft bellies in swine, generally an undesirable
property [131]. Extraction of corn oil, either prior to starch conversion to ethanol
or prior to cellulose conversion, may be necessary in order to keep lipid content

low as well as providing a second co-product.

230

8.3.5 Protease Extraction

Figure 8.7 shows the rate of protein solubility for both high (1.0%) and low (0.1%)
protease loadings up through 168 hours of protein digestion. Data points were
fitted to a logarithmic curve, and high protease loading had a significant (p<0.05)
increase in the rate of protein solubility. Low protease loading approximately
doubled the amount of protein in solution after 24 hours, but very little additional
protein was released after this point. Higher protease loadings showed both an
improved rate and extent of protein solubility, continuing to release protein
throughout the experiment. The reduced rate of protein solubility, particularly for
the low protease loading, is likely due to protease degradation over time, either

due to protein unfolding or the protease attacking itself.

Gel electrophoresis suggests the latter explanation, as no band is seen for the
protease after 24 hours (data not shown). Approximately 18% of the total protein
is available as soluble protein prior to protease addition, demonstrating the
importance of using whole stillage as the substrate rather than solely the
insoluble material. Furthermore, more protein is initially soluble in the stillage for
wheat compared to com [132], indicating that wheat based dry mills may be

better suited to this technology than com.

231

 

01
O

 

 

 

 

 

E
‘3 40 ~ i
_ , i
E i
.0 30 ~ g ..
m
c
E 20 g
2
a
1 a
O o 0.1% Loading
0 I 1% Loading

 

 

 

 

O 24 48 72 96 120 144 168
Time (h)

Figure 8.7: Rate of protein removal during protease digestion for both low (0.1%
w/w) and high (1.0% w/w) loading of protease. Lines represent best logarithmic
fit. The amount at Oh represents the initial soluble protein within the whole
stillage. Extractions were performed at 50°C, a pH of 7.5, and 90 rpm.

Attempts to extract the proteins prior to protease addition are seen in Figure 3.8.
The high temperatures used were necessary to obtain a significant extract. As
expected, the initial amount of protein in solution increased when a prior
extraction was performed, although little difference is seen between the neutral
and basic extracts. This led to a subsequent increase in the final protein
concentration in solution after protease addition. Despite the increase in protein
concentration, virtually all of it is still broken down into smaller peptides (data not
shown). Thus, it is likely that the number of active sites within the insoluble

matrix is the limiting factor in the release and breakdown of the proteins, rather

than the activity of the proteases themselves.

232

It is interesting to note that, despite the lack of initial increase in protein removal
in alkaline extraction, the final result after the addition of proteases did continue
to increase the protein removal. Thus, the alkaline conditions may also be

opening up more sites for the protease to attack.

 

 

01
O

A
O

 

 

Percent solubilized (%)
ro co
0 O

  

—L
O
l

 

Mass Solubilized

   

 

 

O
1

 

pH12 I control pH7.5 I pH12
0h 1 24h

 

 

Figure 8.8: Effect of extractions prior to protease addition. The control is
protease digestion without extraction. Extractions were performed at two
different pH levels at 70°C for 1h. The alkaline extract was neutralized prior to
protease addition. Here, time at on represents the beginning of protease
digestion.

Due to the slight improvement seen with alkaline extraction, the alkaline protease
Protex 6L was also tested. In addition, Protex 51 P was tested due to its
exopeptidase activity, as well as a combination of 51 P and 14L. Protex 6L
digestions were performed at pH 10, the optimal level suggested by the

manufacturer, whereas 14L and 51 P digestions were performed at pH 7.5. As

seen in Figure 8.9, Protex 6L showed a significant improvement over the other

233

two proteases. This could be due to the increased solubility at alkaline
conditions. The exopeptidase enzymes released fewer proteins than either of
the other two enzymes tested. This is most likely due to a decrease in active
sites, as the ends of the proteins are embedded within the insoluble matrix or
have been modified during the original dry grind process. The combination of
14L and 51 P were only slightly lower than 14L alone, despite the decrease in 14L
loading. This indicates that there may be some synergistic activity between the
two types of protease, indicating other protein combinations may also improve

protein removal with lower overall loadings.

60 ,_ ,,,_-,_E,, ssss ,_L_,-_.
50 ~
40

30

   

 

Percent solubilized (%)

 
 

I Pr'ot'ein Solubilized 1
1:: Mass Solubilized E

 

 

 

 

 

6L 14L 51 P 14IJ51 P
Figure 3.9: The effect of different proteases on protein removal. The enzymes
used are Protex 6L, Protex 14L, and Protex 51 P. Digestions were performed for
24h at 50°C and 90rpm. Protease was loaded at 0.1% (w/w) and the pH was
kept constant at 7.5 for Protex 14L and 51 P and 10 for Protex 6L.

Figure 8.10 shows the effect of Protex 6L protease digestion on the amino acid

profiles of the stillage. Both the profile of the whole stillage as well as the soluble

234

portion are shown for comparison. In general, the protease digestion served to
moderate the amino acid profile, increasing the relative ratios of amino acids that
were not originally soluble in stillage. However, relative amounts of both aspartic
acid/asparagine and lysine increased more than expected, while glycine and
histidine decreased more than expected. Proteases attack at specific sites within
a protein, and thus are not likely to attack all proteins equally. Thus, a protease
cocktail may help digest specific proteins or peptide sequences that Protex 6L

alone is unable to attack, potentially increasing the yield of soluble amino acids.

 

1 8%
l Protex 6L

15% - I untreated - soluble
El untreated

 

 

 

 

 

1 2%

 

 

 

9% _

 

 

6%

% of Total Protein

  

 

3%

 

    

 

 

 

 

 

 

 

 

 

mm 111 IIIII

0% .
ALA ARG ASX GLY HIS ILE LEU LYS PHE PRO THR WR VAL

Figure 3.10: Amino acid profiles for proteins released by Protex 6L, soluble
proteins in wheat stillage, and total proteins in wheat stillage. Met, Cys, and Trp
were not detected due to being destroyed during acid hydrolysis.

8.4 Conclusions

Extraction of proteins from wet distiller’s grains using alkaline solvents in
combination with pretreatment and enzymatic hydrolysis of cellulose does not

appear to be an effective method of enhancing the protein quality of distiller’s

235

 

grains. Only 32% of the proteinsand 43% of the lysine was extracted using 0.1
M NaOH. In addition, this separated the glucose into two separate streams,
adding to downstream costs. In comparison, if enzymatic hydrolysis is performed
without alkaline extraction, 45% of protein is solubilized. Extraction of hydrolyzed
material resulted in an additional 30% of protein extracted, to bring the total

amount of soluble protein to 62%.

Protease extraction of wheat stillage using 0.1% Protex 6L yielded a maximum of
57% of protein in soluble form. However, approximately 32% of these proteins
were already soluble prior to protease addition, indicating the importance of using
the whole stillage as a feedstock rather than solely the solid distiller’s grains.
These soluble proteins as well as those extracted were effectively broken down
into smaller peptides as well. With 0.1% protease loading, digestion was
essentially complete after 24 hours. Improvements in protein solubilization were
also seen with an alkaline extraction prior to protease addition for proteases
active in neutral conditions. Increased protease addition primarily improves the
extent of hydrolysis, rather than the rate. In addition, a cocktail of different
alkaline proteases may further improve yields and therefore reduce carbon

emissions.
Simultaneous aqueous phase enzymatic saccharification of cellulosic materials

and hydrophobic biosolvent phase extraction of proteins was not successful.

This is likely due to mass transfer issues, as little contact was present between

236

 

the hydrophobic solvent and the proteins. Thus, bio-solvents were unable to
recover additional proteins from DGS at the current conditions. A separate
extraction and hydrolysis, however, was successful in removing proteins and

hydrolyzing the fiber, although the protein was low in lysine.

Thus, it remains necessary to recover proteins in the insoluble, solid portion of
the biomass, as an enhanced distiller’s grain product. Added value is obtained
from removing fat, fiber, low-lysine protein, and water solubles from the product
rather than removing the high—lysine protein. This product is low in ash and fiber
and high in protein, potentially increasing its value as an animal feed to
nonruminants. Corn oil may be removed prior to the dry grind process and sold
as a separate product or alternatively removed using a hydrophobic solvent.
Glucan can be removed easily by enzymatic hydrolysis, although hemicellulose
requires a more complete set of enzymes if using a mild pretreatment such as
AFEX. While a cocktail of enzymes has been shown to be effective in releasing
pentoses, it remains to be seen whether the added cost of the enzymes would
make this approach economically viable. The protein remaining in solution after
hydrolysis and fermentation can then be dried and returned to the insoluble

portion, similar to what is currently performed in dry grind ethanol facilities.
It the soluble portion of the hydrolysate is not recombined with the insoluble

portion, it may still be possible for the remaining proteins to have some value. It

has been suggested that protein in distiller’s grains be used as a feedstock for

237

bioplastics [133]. This would require breaking the proteins down into their
component amino acids, probably using proteases. However, these proteases
are costly and cannot completely solubilize all protein in DGS. Thus, one option
would be to simply break down those proteins in solution after hydrolysis. Less
protease would be needed, thereby reducing costs. In addition, it may also be
possible to separate out the lysine after protease hydrolysis and return it to the

DGS to improve its value as an animal feed.

238

REFERENCES

[1] Energy Information Administration. 2009. Annual Energy Review 2008. US
Department of Energy, Washington, DC.

[2] Yli-Tuomi T, Aarnio P, Pirjola L, Makela T, Hillamo R, Jantunen M. 2005.
Emissions of fine particles, NOx, and CO from on-road vehicles in Finland.
Atmospheric Environment 39(35):6696-6706.

[3] Lugar RG, Woolsey RJ. 1999. The New Petroleum. Foreign Affairs 78:88-102.

[4] Bentley RW. 2002. Global oil & gas depletion: an overview. Energy Policy
30:189-205.

[5] Dinneen B. 2009. 2009 Ethanol Industry Outlook. Renewable Fuels
Association, Washington, DC.

[6] Energy Independence and Security Act of 2007. Public Law # 110-140.

[7] Graham RL, Nelson R, Sheehan J, Perlack RD, Wright LL. 2007. Current and
Potential U.S. Com Stover Supplies. Agronomy Journal 99:1 -1 1.

[8] Gallagher P, Dikeman M, Fritz J, Wailes E, Gauther W, Shapouri H. 2003.
Biomass from crop residues: cost and supply estimates. Washington DC: United
States Department of Agriculture, Office of the Chief Economist, Office of Energy
Policy and New Uses. Report nr 819. 30 p.

[9] US Department of Agriculture. 2009. USDA Agricultural Projections to 2018.
Office of the Chief Economist, Long-term Projections Report OCE-2009-1.

[10] Walsh ME, Ugarte DG, Shapouri H, Slinsky SP. 2003. Bioenergy crop
production in the United States: Potential Quantities, Land Use Changes, and
Economic Impacts on the Agricultural Sector. Environmental and Resource
Economics 24:313-333.

[11] Searchinger T, Heimlich R, Houghton RA, Dong F, Elobeid A, Fabiosa J,
Tokgoz S, Hayes D, Yu T-H. 2008. Use of US Croplands for Biofuels Increases
Greenhouse Gases Through Emissions from Land-Use Change. Science
319:1238-1240.

[12] Belyea RL, Rausch KD, Tumbleson ME. 2004. Composition of corn and
distillers dried grains with solubles from dry grind ethanol processing.
Bioresource Technology 94:293-298.

[13] Holcomb R, Petrone R. 2008. Agricultural Statistics 2008. 1-20 p.

239

[14] Heaton EA, Long SP, Voigt TB, Jones MB, Clifton-Brown J. 2004.
Miscanthus for Renewable Energy Generation: European Union Experience and
Projections for Illinois. Mitigation and Adaptation Strategies for Global Change
9:433-451.

[15] Adler P, Sanderson MA, Boateng A, Weimer PJ, Jung H-JG. 2006. Biomass
yield and biofuel quality of switchgrass harvested in fall or spring. Agronomy
Journal 98:1 518-1 525.

[16] Vogel KP, Brejda JJ, Walters DT, Buxton DR. 2002. Switchgrass biomass
production in the Midwest USA: Harvest and nitrogen management. Agronomy
Journal 94:413-420.

[17] Madakadze IC, Stewart KA, Peterson PR, Coulman BE, Smith DL. 1999.
Cutting frequency and nitrogen fertilization effects on yield and nitrogen
concentration of switchgrass in a short season area. Crop Science 39:552-557.

[18] Carolan JE, Joshi SV, Dale BE. 2007. Technical and Financial Feasibility
Analysis of Distributed Bioprocessing Using Regional Biomass Pre-Processing
Centers. Journal of Agricultural & Food Industrial Organization 5:10.

[19] Greene N. 2004. Growing energy: How biofuels can help end America's oil
dependence. National Resource Defense Council.

[20] Wright MM, Brown RC. 2007. Comparative economics of biorefineries based
on the biochemical and thermochemical platforms. Biofuels, Bioproducts and
Biorefining 1:49-56.

[21] Akin D. 2007. Grass lignocellulose. Applied Biochemistry and Biotechnology
137-140:3-15.

[22] Mosier N, Wyman C, Dale BE, Elander R, Lee W, Holtzapple M, Ladisch M.
2005. Features of promising technologies for pretreatment of lignocellulosic
biomass. Bioresource Technology 96:673-686.

[23] Wyman CE, Dale BE, Elander RT, Holtzapple M, Ladisch MR, Lee W. 2005.
Comparative sugar recovery data from laboratory scale application of leading
pretreatment technologies to com stover. Bioresource Technology 96220262032.

[24] Bura R, Chandra R, Saddler J. 2009. Influence of xylan on the enzymatic
hydrolysis of steam-pretreated com stover and hybrid poplar. Biotechnology
Progress 25:315-322.

[25] Sendich E, Laser M, Kim S, Alizadeh H, Laureano-Perez L, Dale BE, Lynd
LR. 2008. Recent process improvements for the ammonia fiber expansion
(AFEX) process and resulting reductions in minimum ethanol selling price.
Bioresource Technology 99:8429-8435.

240

[26] Perlack RD, Wright LL, Turhollow AF, Graham RL, Stokes BJ, Erbach DC.
2005. Biomass as feedstock for a bioenergy and bioproducts industry: the
technical feasibility of a billion-ton annual supp/y. US Dept of Energy/ US Dept of
Agriculture Report DOE/GO-102995-2135.

[27] Lynd LR. 1996. Overview and evaluation of fuel ethanol from cellulosic
biomass: Technology, Economics, the Environment, and Policy. Annual Review
of Energy and Environment 21 :403-465.

[28] McLaughlin SB, Ugarte DG, Garten CT, Lynd LR, Sanderson MA, Tolbert
VR, Wolf DD. 2002. High-value renewable energy from prairie grasses. p
Environmental Science and Technology 36:2122-2129. :

[29] Kadam KL, McMillan JD. 2003. Availability of com stover as a sustainable .
feedstock for bioethanol production. Bioresource Technology 99:17-25.

 

[30] Kim S, Dale BE. 2004. Global potential bioethanol production from wasted ‘
crops and crop residues. Biomass and Bioenergy 26:361-375. ’

[31] Campbell JE, Lobell DB, Genova RC, Field CB. 2008. The Global Potential
of Bioenergy on Abandoned Agriculture Lands. Environmental Science and
Technology 42:5791-5794.

[32] Lal R. 2008. Crop residues as soil amendments and feedstock for bioethanol
production. Waste Management 28:747-758.

[33] NRC. 2001. Nutrient Requirements of Dairy Cattle: Seventh Revised Edition,
2001. Washington, DC: National Academy Press.

[34] Van Soest PJ. 2006. Rice straw, the role of silica and treatments to improve
quality. Animal Feed Science and Technology 130: 1 37-1 71.

[35] Solaiman SG, Horn GW, Owens FN. 1979. Ammonium hydroxide treatment
on wheat straw. Journal of Animal Science 49:802-809.

[36] Dean DB, Adesogan AT, Krueger NA, Littell RC. 2008. Effects of treatment
with ammonia or fibrolytic enzymes on chemical composition and ruminal
degradability of hays produced from tropical grasses. Animal Feed Science and
Technology 145:68-83.

[37] Reddy GVN, Reddy MR. 1985. Effect of ammonia treatment and processing
of whole cotton plants as sole source of roughage in complete feeds for growing
cross-bred calves. Animal Feed Science and Technology 13:93-102.

[38] Ben Salem A, Nefzaoui A, Rokbani N. 1994. Upgrading of sorghum stover
with anhydrous ammonia or urea treatments. Animal Feed Science and
Technology 48:1 5-26.

241

[39] Weiss WP, Shockey WL. 1991. Value of Orchardgrass and Alfalfa Silages
Fed with Varying Amounts of Concentrates to Dairy Cows. Journal of Dairy
Science 74:1933-1943.

[40] Weimer PJ, Mertens DR, Ponnampalam E, Severin BF, Dale BE. 2003.
FlBEX-treated rice straw as a feed ingredient for lactating dairy cows. Animal
Feed Science and Technology 103241-50.

[41] Perdok HB, Leng RA. 1987. Hyperexcitability in Cattle Fed Ammoniated
Roughages. Animal Feed Science and Technology 17:121-143.

[42] Weiss WP, Conrad HR, Martin CM, Cross RF, Shockey WL. 1986. Etiology
of Ammoniated Hay Toxicosis. Journal ofAnimaI Science 63:525-532.

[43] Pirie NW, editor. 1978. Leaf Protein and Other Aspects of Fodder
Fractionation. Cambridge: Cambridge University Press.

[44] Kohler GO, Bickoff EM. 1970. Leaf Protein. Proceedings of the 3rd
International Congress of Food Science and Technology. Washington, DC:290-
295.

[45] Fiorentini R, Galoppini C. 1981. Pilot plant production of an edible alfalfa
protein concentrate. Journal of Food Science 46:1514-1520.

[46] Carroad PA, Anaya-Serrano H, Edwards RH, Kohler GO. 1981. Optimization
of cell disruption for alfalfa leaf protein concentration (Pro-Xan) production.
Journal of Food Science 46:383-386.

[47] Enochian RV, Kohler GO, Edwards RH, Kuzmicky DD, Vosloh CJ. 1980.
Producing Pro-Xan (leaf protein concentrate) from alfalfa: Economics of an
emerging technology. Agricultural Economic Report No. 445, US Dept. of
Agriculture.

[48] McDonald RM, Ritchie JM, Donnely PE, Bushnell PG, Mace MJ. 1981.
Economics of the leaf-protein extraction process under New Zealand conditions.

[49] Sinclair S. 2009. Protein extraction from pasture: The plant fractionation bio-
process and adaptability to farming systems. Ministry of Agriculture and
Forestry, New Zealand, SFF No. 008/001.

[50] Betschart A, Kinsella JE. 2002. Extractability and solubility of leaf protein.
Journal of Agricultural and Food Chemistry 21:60-65.

[51] Fernandez SS, Padilla AP, Mucciarelli S. 1999. Protein extraction from
Atriplex lampa leaves: Potential use as forage for animals used for human diets.
Plant Foods for Human Nutrition 54:251-259.

242

 

[52] Ordonez C, Asenjo M, Benitez C, Gonzalez J. 2001. Obtaining a protein
concentrate from integral defatted sunflower flour. Bioresource Technology
78:187-190.

[53] Tang S, Hettiarachchy NS, Shellhammer TH. 2002. Protein Extraction from
Heat-Stabilized Defatted Rice Bran. 1. Physical Processing and Enzyme
Treatments. Journal of Agricultural and Food Chemistry 50:7444-7448.

[54] Balasubramaniam D, Wilkinson C, Van Cott K, Zhang C. 2003. Tobacco
protein separation by aqueous two-phase extraction. Journal of Chromatography
A 989(1):119-129.

[55] Lawton JW; 1986 Nov 25, 1986. Process for recovery of protein from
agricultural commodities prior to alcohol production. US Patent #4,624,805.

[56] Dale BE. 1983. Biomass refining: protein and ethanol from alfalfa. Industrial
& Engineering Chemistry Product Research and Development 22:466-472.

[57] De La Rosa L, Reshamwala S, Latimer V, Shawky B, Dale B, Stuart E. 1994.
Integrated production of ethanol fuel and protein from Coastal Berrnudagrass.
Applied Biochemistry and Biotechnology 45-462483-497.

[58] Urribarri L, Ferrer A, Colina A. 2005. Leaf protein from ammonia-treated
dwarf elephant grass (Pennisetum purpureum Schum cv. Mott). Applied
Biochemistry and Biotechnology 121 -124:721-730.

[59] Keshwani DR, Cheng JJ. 2009. Switchgrass for bioethanol and other value-
added applications: a review. Bioresource Technology 100:1515-1523.

[60] Alizadeh H, Teymouri F, Gilbert Tl, Dale BE. 2005. Pretreatment of
Switchgrass by Ammonia Fiber Explosion (AFEX). Applied Biochemistry and
Biotechnology 1 21 -1 24:1 1 33-1 142.

[61] Fike JH, Parrish DJ, Wolf DD, Balasko JA, Green JT, Rasnake M, Reynolds
JH. 2006. Switchgrass production for the upper southeastern USA: Influence of
cultivar and cutting frequency on biomass yields. Biomass and Bioenergy
30:207-213.

[62] Monti A, Bezzi G, Pritoni G, Venturi G. 2008. Long-term productivity of
lowland and upland switchgrass cytotypes as affected by cutting frequency.
Bioresource Technology 99:7425-7432.

[63] Teymouri F, Laureano-Perez L, Alizadeh H, Dale BE. 2005. Optimization of
the ammonia fiber explosion (AFEX) treatment parameters for enzymatic
hydrolysis of corn stover. Bioresource Technology 96:2014-2018.

243

[64] Gao D, Chundawat SPS, Krishnan C, Balan V, Dale B. 2009. Identification
and optimization of six core glycosyl hydrolases for the hydrolysis of ammonia
fiber expansion (AFEX) pretreated com stover. Bioresource Technology In press.

[65] Box G, Behnken D. 1960. Some new three level designs for the study of
quantitative variables. Technometrics 2:455-475.

[66] Sluiter A, Hames B, Ruiz R, Scarlata C, Sluiter J, Templeton D, Crocker D.
2008. Determination of structural carbohydrates and lignin in biomass. US Dept.
of Energy Technical Report NREL/TP-510-42618.

[67] Isci A, Himmelsbacj JN, Pometto AL, Raman DR, Anex RP. 2008. Aqueous
Ammonia Soaking of Switchgrass Followed by Simultaneous Saccharification
and Fermentation. Applied Biochemistry and Biotechnology 144:69-77.

[68] Suryawati L, Wilkins MR, Bellmer DD, Huhnke RL, Maness NO, Banat IM.
2009. Effect of hydrotherrnolysis process conditions on pretreated switchgrass
composition and ethanol yield by SSF with Kluyveromyces marxianus IMB4.
Process Biochemistry 44:540-545.

[69] Dien BS, Ximenes EA, O'Bryan PJ, Moniruzzaman M, Li X-L, Balan V, Dale
BE, Cotta MA. 2008. Enzyme characterization for hydrolysis of AFEX and liquid
hot-water pretreated distillers’ grains and their conversion to ethanol. Bioresource
Technology 99:521 6-5225.

[70] Lovrien R, Matulis D. 1995. Current Protocols in Protein Science. Section
341-3.4.24.

[71] Graham H. 1992. Stabilization of the Prussian blue color in the determination
of polyphenols. Journal of Agricultural and Food Chemistry 40:801-805.

[72] Sami M, Annamalai K, Wooldridge M. 2001. Co-firing of coal and biomass
fuel blends. Progress in Energy and Combustion Science 27:171-214.

[73] Breslau BR, Kilcullen BM. 1977. Hollow Fiber Ultrafiltration of Cottage
Cheese Whey: Performance Study. Journal of Dairy Science 60:1379-1386.

[74] Nagendran S, Hallen-Adams HE, Paper JM, Aslam N, Walton JD. 2009.
Reduced genomic potential for secreted plant cell-waII-degrading enzymes in the
ectomycorrhizal fungus Amanita bisporigera, based on the secretome of
Trichoderma reesei. Fungal Genetics and Biology 46:427-435.

[75] Koschuh W, Povoden G, Thang V, Kromus S, Kulbe K, Novalin S,
Krotscheck C. 2004. Production of leaf protein concentrate from ryegrass (Lolium
perenne x multiflorum) and alfalfa (Medicago sauva subsp. sativa). Comparison
between heat coaguIation/centrifugation and ultrafiltration. Desalination 1632253-
259.

244

[76] Goering HK, Van Soest PJ. 1970. Forage fiber analyses (apparatus,
reagents, procedures, and some applications). US Agricultural Research Service.

[77] Mertens D. 2002. Gravimetric determination of amylase-treated neutral
detergent fiber in feeds with refluxing in beakers or crucibles: collaborative study.
Journal of AOAC lntemational 85: 1 21 7-1240.

[78] Wiles P, Gray I, Kissling R. 1998. Routine analysis of proteins by Kjeldahl
and Dumas methods: Review and interlaboratory study using dairy products.
Journal of AOAC International 81 :620-632.

[79] Robinson P, Fadel J, Tamminga S. 1986. Evaluation of mathematical
models to describe neutral detergent residue in terms of its susceptibility to
degradation in the rumen. Animal Feed Science and Technology 15: 249-271.

[80] Seber GAF, Wild CJ. 1989. Nonlinear regression. New York: Wiley & Sons.

[81] Balan V, Sousa LC, Chundawat SPS, Vismeh R, Jones AD, Dale BE. 2008.
Mushroom spent straw: a potential substrate for an ethanol-based biorefinery.
Journal of Industrial Biotechnology 35:293-301.

[82] Mumen HK, Balan V, Chundawat SPS, Bals B, Sousa LdC, Dale BE. 2007.
Optimization of Ammonia Fiber Expansion (AFEX) Pretreatment and Enzymatic
Hydrolysis of Miscanthus x giganteus to Ferrnentable Sugars. Biotechnology
Progress 23:846-850.

[83] Buranov AU, Mazza G. 2008. Lignin in straw of herbaceous crops. Industrial
Crops and Products 28:237-259.

[84] Chundawat SPS, Balan V, Dale BE. 2007. Effect of Particle Size Based
Separation of Milled Corn Stover on AFEX Pretreatment and Enzymatic
Digestibility. Biotechnology and Bioengineering 96:219-231.

[85] Birkelo CP, Johnson DE, Ward GM. 1986. Net energy value of ammoniated
wheat straw. Journal of Animal Science 63:2044-2052.

[86] Broudiscou LP, Agbagla-Dobnani A, Papon Y, Comu A, Grenet E,
Broudiscou AF. 2003. Rice straw degradation and biomass synthesis by rumen
micro-organisms in continuous culture in response to ammonia treatment and
legume extract supplementation. Animal Feed Science and Technology 105295-
108.

[87] Weimer PJ, Chou Y, Weston W, Chase D. 1986. Effect of supercritical
ammonia on the physical and chemical structure of ground wood. Biotechnology
and Bioengineering Symposium 17:5-18.

245

 

[88] Wyman C, Dale B, Elander R, Holtzapple M, Ladisch M, Lee Y. 2005.
Coordinated development of leading biomass pretreatment technologies.
Bioresource Technology 96:1 959-1966.

[89] Parrish D, Fike J. 2005. The biology and agronomy of switchgrass for
biofuels. Critical Reviews in Plant Science 24:423-459.

[90] Belasco U. 1954. New Nitrogen Feed Compounds for Ruminants--A
Laboratory Evaluation. Journal of Animal Science 13:601-610.

[91] Huber JT, Kung L. 1981. Protein and Nonprotein Nitrogen Utilization in Dairy
Cattle. Journal of Dairy Science 64:1 170-1 195.

[92] Anderson SJ, Merrill JK, Klopfenstein TJ. 1988. Soybean hulls as an energy
supplement for the grazing ruminant. Journal of Animal Science 66:2959-2964.

[93] Ludden PA, Cecava MJ, Hendrix KS. 1995. The value of soybean hulls as a
replacement for corn in beef cattle diets formulated with or without added fat.
Journal of Animal Science 73:2706-2711.

[94] Galloway DL, Goetsch AL, Forster LA, Patil AR, Sun W, Johnson 28. 1993.
Feed Intake and Digestibility by Cattle Consuming Berrnudagrass or
Orchardgrass Hay Supplemented with Soybean Hulls and(or) Com. Journal of
Animal Science 71:3087-3095.

[95] Laser M, Jin H, Jayawardhana K, Dale BE, Lynd LR. 2009. Projected mature
technology scenarios for conversion of cellulosic biomass to ethanol with
coproduction, thermochemical fuels, power, and/or animal feed protein. Biofuels,
Bioproducts and Biorefining 3:231 -246.

[96] Aden A, Ruth M, Ibsen KN, Jechura J, Neeves K, Sheehan J, Wallace B,
Montague L, Slayton A, Lukas J. 2002. Lignocellulosic Biomass to Ethanol
Process Design and Economics Utilizing Co-Current Dilute Acid Prehydrolysis
and Enzymatic Hydrolysis for Corn Stover. US Dept. of Energy Technical Report
NREL/TP-51 0-32438.

[97] Afonso MD, Ferrer J, Borquez R. 2004. An economic assessment of proteins
recovery from fish meal effluents by ultrafiltration. Trends in Food Science and
Technology 1 5:506-512.

[98] Kwiatkowski JR, McAloon AJ, Taylor F, Johnston DB. 2006. Modeling the
process and costs of fuel ethanol production by the corn dry-grind process.
Industrial Crops and Products 23:288-296.

[99] Saunders RM, Connor MA, Booth AN, Bickoff EM, Kohler GO. 1973.
Measurement of Digestibility of Alfalfa Protein Concentrates by in vivo and in
vitro Methods. Journal of Nutrition 103:530-535.

246

 

[100] Edwards RH, de Fremery D, Mackey BE, Kohler GO. 1978. Factors
affecting juice extraction and yield of leaf protein concentrated from ground
alfalfa. Transactions of the American Society of Agricultural Engineers 21 :55-62.

[101] Mani S, Sokhansanj S, Bi X, Turhollow AF. 2006. Economics of producing
fuel pellets from biomass. Applied Engineering in Agriculture 22:421-426.

[1 O2] Peters MS, Timmerhaus KD, West RE. 2003. Plant Design and Economics
for Chemical Engineers, 5th ed. Boston: McGraw-Hill.

[103] Anonymous. 2009. Hay Market News.
http:/Iwww.extension.iastate.edu/adair/news/haymarketnews.htm: Iowa State
University.

[104] Perrin R, vogel KP, Schmer M, Mitchell RB. 2008. Farm-Scale Production
Cost of Switchgrass for Biomass. Bioenergy Research 1:91 -97.

[105] 2009. ICIS Pricing 10th April 2009. Elsevier. p
httpz/Iwww.icispricing.com/iI_shared/Samples/SubPage149.asp.

[106] US Dept. of Agriculture. 2009. In: Service U-NAS, editor. Agricultural
Prices: Kansas Agricultural Statistics. Released Aug 3, 2009.

[107] Anonymous. 2009. Market trends. Biofuels, Bioproducts and Biorefining
3:424.

[108] Howard WT, Shaver RD. 1997. FEEDVAL: Comparative values calculated
from crude protein, TDN, Ca and P In: Wisconsin-Madison Uo, editor.

[109] Swenson A. 2009. North Dakota farm profits remained strong in 2008. In:
Service NE, editor. http://www.ag.ndsu.edu/news/newsreleases/2009/june-22-
2009/north-dakota-farm-profits-remained-strong-in-2008.

[110] Swenson A. 2009. Crop profit projections down for 2009. In: Service NE,
editor. http://www.ag.ndsu.edu/news/newsreleases/2009/jan-8-2009/crop-profit-
projections-down-for-2009/.

[111] Groveman J. 2009. Agriculture Secretary Vilsack announces $1.7 billion in
Conservation Reserve Program rental payments. In: Agency UFS, editor.
http://www.fsa.usda.gov/FSA/newsReleases?area=newsroom&subject=landing&t
opic=ner&newstype=newsrel&type=detail&item=nr__20091 007_rel__0497.html.

[112] Dale BE, Allen MS, Laser M, Lynd LR. 2009. Protein feeds coproduction in
biomass conversion to fuels and chemicals. Biofuels, Bioproducts and Biorefining
3219-230.

[113] Ensminger ME, Olentine CG, editors. 1978. Feeds and Nutrition. Clovis,
CA: Ensminger Publishing.

247

 

[114] Kim S, Dale BE. 2005. Life cycle assessment of various cropping systems
utilized for producing biofuels: Bioethanol and biodiesel. Biomass and Bioenergy
29:426-439.

[115] Gibson L, Cheema M, Patrick G. 2007. Spring canola variety performance
in Iowa: 2007 Final Report. Ames, IA: Iowa State University, Dept. of Agronomy.

[116] Ladisch M, Dale B. 2008. Distillers grains: On the pathway to cellulose
conversion. Bioresource Technology 99:5155-5156.

[117] Tucker MP, Nagle NJ, Jennings EW, Ibsen KN, Aden A, Nguyen QA, Kim
KH, Noll SL. 2004. Conversion of distiller’s grain into fuel alcohol and a higher-
value animal feed by dilute-acid pretreatment. Applied Biochemistry and
Biotechnology 1 13-1 16:1 139-1 159.

[118] Koukiekolo R, Cho H-Y, Kosugi A, lnui M, Yukawa H, Doi RH. 2005.
Degradation of Com Fiber by Clostridium cellulovorans Cellulases and
Hemicellulases and Contribution of Scaffolding Protein Cpr. Applied and
Environmental Microbiology 71 :3504-351 1 .

[119] Kristensen JB, Felby C, Jorgensen H. 2009. Yield-determining factors In
high-solids enzymatic hydrolysis of lignocellulose. Biotechnology for Biofuels
2:11.

[120] Lau MW, Dale BE. 2009. Cellulosic ethanol production from AFEX-treated
corn stover using Saccharomyces cerevisiae 424A(LNH-ST). Proceedings of the
National Academy of Sciences 106:1368-1373.

[121] Perkis D, Tyner W, Dale R. 2008. Economic analysis of a modified dry
grind ethanol process with recycle of pretreated and enzymatically hydrolyzed
distillers’ grains. Bioresource Technology 99:5243-5249.

[122] Scott E, Peter F, Sanders J. 2007. Biomass in the manufacture of industrial
products—the use of proteins and amino acids. Applied Microbiology and
Biotechnology 752751 -762.

[123] Gu L, Jones AD, Last RL. 2007. LC-MS/MS Assay for Protein Amino Acids
and Metabolically Related Compounds for Large-Scale Screening of Metabolic
Phenotypes. Analytical Chemistry 79:8067-8075.

[124] Heidelbaugh ND, Huber CS, Bednarczyk JF, Smith MC, Rambaut PC,
Wheeler HO. 2002. Comparison of three methods for calculating protein content
of foods. Journal of Agricultural and Food Chemistry 23:61 1-613.

[125] Wu YV, Sexson KR. 1976. Protein concentrate from normal and high-lysine
corns by alkaline extraction: composition and properties. Journal of Food Science
41 :512-515.

248

 

[126] Dickinson E, Euston SR, Woskett CM. 1990. Competitive adsorption of
food macromolecules and surfactants at the oil-water interface. Progress in
Colloid Polymer Science 82:65-75.

[127] Boatright WL, Hettiarachchy NS. 1995. Effect of lipids on soy protein isolate
solubility. Journal of the American Oil Chemistry Society 72:1739-1444.

[128] Okahisa Y, Yoshimura T, Imamura Y. 2005. An application of the alkaline
extraction - glucoamylase hydrolysis method to analyze starch sugar contents of
bamboo. Journal of Wood Science 51 :542-545.

[129] Lenis NP, Bikker P, van der Meulen J, van Diepen JT, Bakker JG,
Jongbloed AW. 1996. Effect of dietary neutral detergent fiber on ileal digestibility
and portal flux of nitrogen and amino acids and on nitrogen utilization in growing
pigs. Journal of Animal Science 74:2687-2699.

[130] Pedersen C, Boersma M, Stein H. 2007. Digestibility of energy and
phosphorus in 10 samples of distillers dried grain with solubles fed to growing
pigs. Journal of Animal Science .

[131] Whitney MH, Shurson GC, Johnston LJ, Wulf DM, Shanks BC. 2006.
Growth performance and carcass characteristics of grower-finisher pigs fed high-
quality corn distillers dried grain with solubles originating from a modern
Midwestern ethanol plant. Journal of Animal Science 84:3356-3363.

[132] Wilkie A, Riedesel K, Owens J. 2000. Stillage characterization and
anaerobic treatment of ethanol stillage from conventional and cellulosic
feedstocks. Biomass and Bioenergy 19:63-102.

[133] Brehmer B, Bals B, Sanders J, Dale B. 2008. Improving the com-ethanol
industry: Studying protein separation techniques to obtain higher value-added
product Options for distillers grains. Biotechnology and Bioengineering 101.

249

H