IMPROVING BIOMASS PROCESSES VIA STRUCTURE

CHARACTERIZATION AND VALORIZATION OF HUMINS AND PROCESS

INTENSIFICATION

by

Ziwei (Lily) Cheng

A dissertation submitted to the Faculty of the University of Delaware in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Chemical Engineering

Winter 2020

© 2020 Ziwei (Lily) Cheng

All Rights Reserved

IMPROVING BIOMASS PROCESSES VIA STRUCTURE

CHARACTERIZATION AND VALROZATION OF HUMINS AND PROCESS

INTENSIFICATION

by

Ziwei (Lily) Cheng

Approved: ______Eric M. Furst, Ph.D. Chair of the Department of Chemical & Biomolecular Engineering

Approved: ______Levi T. Thompson, Ph.D. Dean of the College of Engineering

Approved: ______Douglas J. Doren, Ph.D. Interim Vice Provost for Graduate and Professional Education and Dean of the Graduate College I certify that I have read this dissertation and that in my opinion it meets the academic and professional standard required by the University as a dissertation for the degree of Doctor of Philosophy.

Signed: ______

Dionisios G. Vlachos, Ph.D. Professor in charge of dissertation

I certify that I have read this dissertation and that in my opinion it meets the academic and professional standard required by the University as a dissertation for the degree of Doctor of Philosophy.

Signed: ______Raul F. Lobo, Ph.D. Member of dissertation committee

I certify that I have read this dissertation and that in my opinion it meets the academic and professional standard required by the University as a dissertation for the degree of Doctor of Philosophy.

Signed: ______Bingjun Xu, Ph.D. Member of dissertation committee

I certify that I have read this dissertation and that in my opinion it meets the academic and professional standard required by the University as a dissertation for the degree of Doctor of Philosophy.

Signed: ______Marat Orazov, Ph.D. Member of dissertation committee I certify that I have read this dissertation and that in my opinion it meets the academic and professional standard required by the University as a dissertation for the degree of Doctor of Philosophy.

Signed: ______Donald Watson, Ph.D. Member of dissertation committee

ACKNOWLEDGMENTS

I would like to thank my advisor Dr. Dion Vlachos, whose diligent passion for research affected me deeply. He forced me out of my limits and has shaped me into an independent and rigorous researcher who’s no longer afraid of exploring the unknown or public speaking. Besides Dion, I greatly appreciate the mentorship from Dr. Basudeb

Saha and Jeffery Everhart, who are both very approachable and helpful with organic chemistry and analytical methods development. Dr. Vladimiros and Dr. George

Tsilomelekis were helpful in my first year and will be dearly missed.

I would then like to acknowledge the instrument managers-Dr. Jan Ilavsky and Dr.

Kuzmenko from Argonne National Lab, and Gerald Poirier and Caroline Golt from the

Advanced Material Characterization Lab at University of Delaware. They were the best at what they do and were always there to answer my questions and offer technical support.

I would like to thank the supportive and encouraging staff at the UD Counseling

Center-Dr. Kimberly Zahm, Dr. Sharon Lee, Dr. Richard Kingsley and Dr. Jeremy

Cohen. They taught me to be self-compassionate and inspired me to understand the minds myself and others. They pulled me out of the most depressing year of my life and their kind and encouraging words meant a lot to me in times of extreme distress.

Most parts of my research are exploratory, and I greatly appreciate the people who are willing to give up their time to explore the unknown together with me -Dr. (now

Prof.!) Konstantinos Goulas, Natalia Rodriguez and Jiayi Fu. Without their company to

v the Argonne National Lab beam times, I could never survive the 48 hours straight experiments. Special thanks also go out to Hannah Nguyen, Pierre Desir and Sibao Liu, who made the late-night work in the lab more fun and less lonely; and to Matt Gilkey- who answered so many of my questions during thesis writing. Also, the RAPID Biomass team members-Pierre Desir, Sebastian Prödinger, Zhaoxing Wang, Ali Mehdad, and Tai- ying Chen are gratefully acknowledged for their wit, great sense of humor and insightful discussions. Working with them as a team has made me happier and more productive than ever. The heartiest thank you belongs to Angela Norton, my crazy dog and cat lady friend. Although growing up in two different cultures, our minds have so much in common and her sweet, tender personality always makes my day (or night).

Lastly, I cannot thank enough my family members-those whom I wouldn’t change for the world. My mother-an incredible woman who truly understands how to raise children. Her wit, curiosity and her big, tender heart has shaped me into someone who’s dedicated to spreading kindness to others. My father, although not good with words, had always supported me financially and would do anything to make me, his little girl happy.

I loved the one-of-a-kind way you taught me division, solving equations and elementary school level English. I missed the adventurous boat watching, fish catching, cat rescuing, bike rides and midnight bus trips we had together. This is the 14th year since I got to know my husband (and the 5th year of our relationship), Chao Wang, and I won’t change a single thing. From views of the world to knowledge and hobbies-We have so much in common like two halves of the same body. His decency as a man, unconditional love, understanding and support is like spring, flowers, and sunshine to my life. Oh, not only

vi did he bring those-he brought our adopted cat Meatball too! She’s the best pet I could ever ask for-a cute, cuddly, vocal, and large furball who loves me wholeheartedly.

vii TABLE OF CONTENTS

LIST OF TABLES ...... xii LIST OF FIGURES ...... xiv LIST OF SCHEMES...... xxi ABSTRACT ...... xxii

Chapter

1 INTRODUCTION ...... 1

1.1 Conversion of Biomass to Chemicals ...... 1 1.2 Humins ...... 6

1.1.1 Literature Review on Humins Formation ...... 7 1.1.2 Literature Review on Applications of Humins ...... 13

1.3 Glucose Dehydration to HMF ...... 14

1.3.1 Significance of Glucose ...... 14 1.3.2 Glucose Dehydration and Mechanism ...... 15 1.3.3 Catalyst Selection for Optimal HMF Production...... 20

1.3.3.1 Homogeneous Catalysts ...... 20 1.3.3.2 Heterogeneous Catalysts ...... 21 1.3.3.3 Lewis/Brønsted Catalyst Ratio ...... 23

1.3.4 Solvents and Phase Modifiers for Suppressing HMF Degradation ...... 25 1.3.5 Process Design ...... 27

1.4 Motivation and Scope of Thesis ...... 30

2 STRUCTURAL ANALYSIS OF HUMINS FORMED IN THE BRØNSTED CATALYZED DEHYDRATION OF FRUCTOSE ...... 33

2.1 Introduction ...... 33 2.2 Experimental ...... 37

2.2.1 Materials ...... 37 2.2.2 Solubility Studies ...... 38 2.2.3 Characterization of Soluble and Insoluble Humins ...... 40

viii 2.3 Results and Discussion ...... 43

2.3.1 Solvent Screening ...... 43 2.3.2 LC-MS and GPC Analysis of Soluble Humins ...... 47 2.3.3 Multistage Dissolution Experiments ...... 51

2.4 Conclusions ...... 60

3 GROWTH KINETICS OF HUMINS STUDIED VIA X-RAY SCATTERING ...... 62

3.1 Introduction ...... 62 3.2 Methods...... 63

3.2.1 Materials ...... 63 3.2.2 Ultra-Small Angle X-ray Scattering ...... 64 3.2.3 Typical Reactivity Experiments ...... 67 3.2.4 Analysis of the Soluble Reaction Products ...... 68

3.3 Results and Discussion ...... 68

3.3.1 Effects of Temperature, Substrate, and Concentration ...... 68 3.3.2 Connection to Nucleation and Growth Theory of Silica ...... 84

3.4 Conclusions ...... 86

4 CATALYTIC HYDROTREATMENT OF HUMINS TO BIO-OIL IN METHANOL OVER SUPPORTED METAL CATALYSTS ...... 87

4.1 Introduction ...... 87 4.2 Methods...... 90

4.2.1 Materials ...... 90 4.2.2 Hydrotreatment Reactions ...... 91 4.2.3 Calculations of Conversion, Yield and Mass Balance ...... 93 4.2.4 Analysis of Reaction Products ...... 94

4.3 Results and Discussion ...... 96

4.3.1 Elemental Composition of Humins-Oil ...... 96 4.3.2 Catalyst Screening ...... 97 4.3.3 Effect of Temperature, Time, Pressure and Catalyst to Humins Ratio ...... 103

4.4 Conclusions ...... 114

5 GLUCOSE CONVERSION TO 5-HYDROXYMETHYL FURFURAL AT SHORT RESIDENCE TIMES IN A MICROREACTOR...... 117

ix 5.1 Introduction ...... 117 5.2 Methods...... 120

5.2.1 Materials ...... 120 5.2.2 Microreactor Setup and Experiments...... 120 5.2.3 Batch Reactor Experiments ...... 122 5.2.4 Product Analysis ...... 123 5.2.5 Characterization of Cr(III) Species ...... 123

5.3 Results and Discussion ...... 124

5.3.1 Comparison Between Batch and Flow Reactor Data ...... 124 5.3.2 Improved HMF Production Rate Using Tandem CrCl3/HCl Catalysts ...... 126 5.3.3 Understanding the Effect of CrCl3 Speciation on Reactivity...... 130

5.4 Conclusions ...... 142

6 GLUCOSE CONVERSION TO 5-HYDROXYMETHYL FURFURAL IN BIPHASIC SOLVENT MIXTURES IN A MICROREACTOR ...... 143

6.1 Introduction ...... 143 6.2 Methods...... 146

6.2.1 Materials ...... 146 6.2.2 Microreactor Setup and Experiments...... 146 6.2.3 Product Analysis ...... 148

6.3 Results and Discussion ...... 149

6.3.1 Effect of Temperature ...... 149 6.3.2 Comparison with Single Phase Data of This Work and Biphasic Systems of Literature ...... 151 6.3.3 Effect of Organic to Aqueous Flow Rate Ratio ...... 154 6.3.4 Comparison between Water/2-Pentanol and Water/MIBK Systems ...... 157

6.4 Conclusions ...... 161

7 CONCLUSIONS, PERSPECTIVES, AND FUTURE RESEARCH DIRECTIONS ...... 163

7.1 Overall Summary and Key Conclusions ...... 163 7.2 Future Research Directions ...... 166

7.2.1 Addressing Challenges in the Valorization of Humins...... 166 7.2.2 Process Intensification for HMF Production and Extraction ...... 170

x REFERENCES ...... 172

Appendix

A SUPPLEMENTARY INFORMATION FOR CHAPTER 2 STRUCTURAL ANALYSIS OF HUMINS FORMED IN THE BRØNSTED- CATALYZED DEHYDRATION OF FRUCTOSE ...... 187

B SUPPLEMENTARY INFORMATION FOR CHAPTER 3 GROWTH KINETICS OF HUMINS STUDIED VIA X-RAY SCATTERING ...... 197

C SUPPLEMENTARY INFORMATION FOR CHAPTER 4 CATALYTIC HYDROTREATMENT OF HUMINS TO BIO-OIL IN METHANOL OVER SUPPORTED METAL CATALYSTS ...... 199

D SUPPLEMENTARY INFORMATION FOR CHAPTER 5 GLUCOSE CONVERSION TO 5-HYDRXYMETHYL FURFURAL AT SHORT RESIDENCE TIMES IN A MICROREACTOR ...... 210

E SUPPLEMENTARY INFORMATION FOR CHAPTER 6 GLUCOSE CONVERSION TO 5-HYDROXYMETHYL FURFURAL IN BIPHASIC SOLVENT MIXTURES IN A MICROREACTOR...... 215

F ADDITIONAL DATA FOR SINGLE, AQUEOUS PHASE GLUCOSE CONVERSION IN A MICROREACTOR ...... 220

G PARTITIONING OF SOLID ACID CATALYST PARTICLES BETWEEN WATER AND 2-PENTANOL PHASES ...... 234

H PERMISSION ...... 237

xi LIST OF TABLES

Table 1.1 Partition coefficient P for HMF in selected organic solvents at 25 °C without any phase modifiers...... 26

Table 1.2 Summary of literature on glucose conversion to HMF in water solvent only...... 28

Table 1.3 Summary of aqueous phase fructose conversion to HMF in flow reactors. . 30

Table 2.1 Mn , Mw and PDI values for solubilized humins determined using GPC.... 50

Table 4.1 Major classes of products and detected compounds in humins oil (obtained from GC-MS shown in Figure C1)...... 98

Table 6.1. Summary of literature on glucose conversion to HMF in biphasic solvents...... 152

Table A. 1 Concentrations of dominant species of solubilized humins detected by LC-MS...... 192

Table A. 2 Concentrations of prevalent, MeOH soluble humins masses of Soxhlet extracted and manually washed humins...... 192

Table A. 3 Mole percentage (%) of each mass in the total amount of monitored species, determined by LC-MS analysis...... 193

Table C. 1 Detailed reaction conditions used for hydrotreatment of humins in methanol...... 199

Table C. 2 Oxyophilicity of metals used as catalysts and the O/C, H/C ratio of the corresponding humins oil...... 200

Table C. 3 Reproducibility results in the yield of detected products from humins hydrotreatment in methanol over Rh/C...... 200

Table C. 4 GC-MS identified compounds in humins-oil...... 203

Table C. 5 Methanol degradation as a function of temperature...... 205

Table C. 6 Polydispersity indices (PDI) of humins oil obtained from hydrotreatment using Rh/C at different reaction conditions...... 206

xii Table C. 7 GC-MS identified compounds derived from IPA in humins oil...... 207

Table C. 8 GC-MS identified compounds in humins oil obtained only from humins during hydrotreatment in IPA...... 208

Table E. 1 Total Flow Rates at Different Residence Times and Organic/Aqueous Flow Rate Ratios (O/A) ...... 215

Table F. 1 Response Factors for Solid Acid Catalyst Particles Suspensions in 2-Pentanol ...... 235

xiii LIST OF FIGURES

Figure 2.1 (a) Pictures of solutions after adding 15 mg humins in 5 mL solvents and stirring the solutions...... 46

Figure 2.2 Concentration of solubilized humins in ACN at stages 1-6. 0.45 g of insoluble humins obtained from the previous stage was mixed with 3mL of ACN and stirred at 25 oC for 24 h at 1000 rpm...... 54

Figure 2.3 Total ion chromatogram (TIC) of ACN solubilized humins for stages 1, 2, and 6 at (a) 10-300 m/z and (b) 301-500 m/z...... 54

Figure 2.4 FTIR spectra of (a) initial humins and insoluble humins remaining after each stage of dissolution...... 57

Figure 3.1 (a) Evolution of I(q) vs. q in a typical experiment...... 71

Figure 3.2 Mean radius of gyration (Rg) of humins particles as a function of time at various temperatures grown in a 10 wt% fructose solution at pH = 0. ... 71

Figure 3.3 Arrhenius plot for humins formation rate from fructose at 80, 85, 90 and 95 ºC...... 72

Figure 3.4 Evolution of polydispersity PD of growing humins over time at 85°C from a 10 wt% fructose solution at pH of 0...... 74

Figure 3.5 Evolution of (a) volume fraction of humins particles, (b) radius of gyration (Rg), and (c) number of particles per volume of solution at 50 wt% and 10 wt% initial fructose concentrations...... 75

xiv Figure 3.6 (a) Scattered intensities of humins originating from 10 wt% fructose and 10 wt% HMF solutions heated at 70 °C and pH of 0, taken at 449 min of reaction time. (b) and (c) Evolution of the corresponding Rg and number of particles as a function of time for the HMF solution...... 79

Figure 3.7 (a) Humins growth rate at different initial HMF concentrations of 2.5 wt%, 5 wt%, 7.5 wt% and 10 wt%...... 80

Figure 3.8 (a). Reaction mixture at 100% HMF conversion after heating at 120 °C for 48 h, before filtration and (b) after filtration using a 0.2 µm syringe filter...... 82

Figure 4.1 van Krevelen diagram of the humins-oil and the starting humins...... 97

Figure 4.2 Comparison of humins conversion and yield of total and GC-detectable humins oil over the Ru/C, Pd/C, Pt/C and Rh/C catalysts...... 102

Figure 4.3 Humins conversion and yields of total and GC-detectable oil at different (a) temperatures, (b) reaction times, (c) hydrogen pressure, and (d) Rh/C catalyst to humins mass ratio...... 104

Figure 4.4 Distribution of GC-detectable products at different (a) temperatures, (b) reaction times, (c) hydrogen pressure, and (d) Rh/C catalyst to humins ratio...... 106

Figure 4.5 Mn and Mw of humin oils obtained at different (a) reaction times, (b) temperatures, (c) hydrogen pressure and (d) Rh/C catalyst to humins ratios...... 108

Figure 4.6 Mass spectrum peaks of representative compounds in 13C isotopic labeling and unlabeled experiments...... 111

Figure 5.1 (a) Glucose conversion and (b) HMF yield as a function of contact (residence) time...... 126

Figure 5.2 Glucose conversion and HMF yield and selectivity at a residence time of 10 min...... 127

Figure 5.3 Glucose conversion and products yield (mannose, fructose, FA, LA and HMF) at temperatures of 200 °C, at short residence times...... 128

Figure 5.4 Glucose conversion and product yields at 200 °C at different pH values. 129

Figure 5.5 Maximum HMF yield per time (productivity) vs. temperature from this work and relevant literature...... 130

xv Figure 5.6 Measured pH of a 3.4 mM chromium chloride solution as a function of contact (residence) time at temperatures indicated...... 133

Figure 5.7 (a) Freshly prepared 3.4 mM CrCl3 solution and (b) a 3.4 mM CrCl3 solution after preheating at 140 °C for 1 h...... 133

Figure 5.8 Glucose conversion and fructose and HMF yields as a function of CrCl3/HCl catalyst preheating time...... 135

Figure 5.9 UV-Vis spectra of a freshly prepared 3.4 mM CrCl3 solution and of the same solution preheated at different times...... 137

Figure 5.10 Reaction rates for glucose dehydration at 160 °C and 180 °C, catalyzed by CrCl3 catalysts preheated for different times indicated...... 140

Figure 6.1. Glucose (GLU) conversion and HMF yield as a function of residence time at temperatures of (a) 160, (b) 180 and (c) 200 °C...... 150

Figure 6.2. (a) Glucose conversion and HMF yield vs. residence time and (b) HMF yield vs. glucose conversion at 200 °C in biphasic and single-phase microreactors...... 152

Figure 6.3. (a) Maximum HMF yield and (b) maximum HMF yield per time (productivity) vs. temperature from this work and relevant literature...... 154

Figure 6.4. (a) Glucose conversion and (b) HMF yield vs. residence time and (c) HMF yield vs. glucose conversion at 200 °C at organic to aqueous flow rate ratios of 1:1, 2:1, and 3:1...... 155

Figure 6.5. (a) Glucose conversion and HMF yield vs. residence time in the water/2-pentanol and water/MIBK systems. (b) HMF, (c) mannose, and (d) humins yield vs. glucose conversion...... 158

Figure A. 1 Solubilized humins concentration in ACN as a function of time...... 187

Figure A. 2 Solubilized humins concentration in ACN as a function of the pore size of the filter...... 188

Figure A. 3 HPLC chromatogram for humins in (a) acetonitrile (ACN) solution and (b) methanol solution...... 188

Figure A. 4 Total ion chromatogram of humins in acetonitrile solution...... 189

Figure A. 5 Total ion chromatogram of humins in methanol solution...... 189

Figure A. 6 Calibration plots of 2-(4-formylphenoxymethyl) furan-3-carboxylic acid (FCA) and HMF (a) 0-0.5 mM and (b) 1-5 mM in acetonitrile...... 189

xvi Figure A. 7 Solubilized humins concentration versus solvent properties...... 190

Figure A. 8 FTIR spectra of humins floating on top and settling to the bottom when placed in water...... 191

Figure A. 9 (a) Concentration of the methanol solubilized humins at stages 1-6...... 191

Figure A. 10 Refractive index detector (RID) intensity as a function of molecular weight for the acetone, ACN, methanol, and THF solubilized humins...... 193 Figure A. 11 Molecular weight distribution for the (a) acetone, (b) acetonitrile, (c) methanol, and (d) THF-soluble portions of humins that were Soxhlet extracted (red) and only manually washed with DI water before dissolution (black)...... 194

Figure A. 12 Zoomed in of the high wavenumber region of Figure 2.4a showing the 2925 cm-1 (C-H stretching) peak for the initial humins and the decreasing intensity of this peak in the insoluble humins from dissolution stages 1~6...... 195

Figure A. 13 FTIR spectra of initial humins, and (a) the methanol solubilized humins at each stage of dissolution, and (b) insoluble humins remaining after each stage of dissolution...... 196

Figure B. 1 Raw and fitted data in a typical experiment...... 197

Figure B. 2 Number of particles per volume of solution as a function of time for (a) 10 wt % initial fructose solution at various temperatures indicated and (b) 50 wt % initial fructose at solution 70 and 80 °C...... 198

Figure B. 3 (a) Fructose and FA, LA, HMF concentrations and (b) calculated fructose conversion and product yields as functions of time...... 198

Figure B. 4 (a) HMF, FA, and LA concentrations and (b) calculated HMF conversion and product yields as a function of time...... 198

Figure C. 1 GC-MS total ion chromatogram of Hunins-oil from the Ru/C catalyzed reaction...... 201

Figure C. 2 GC-MS total ion chromatograms of compounds obtained in blank experiments using (a) Ru/C and (b) Rh/C catalysts without humins...... 202

Figure C. 3 Comparison of GC-detectable oil yield over four commercial catalysts...... 203

Figure C. 4 Molecular weight distribution of species in methanol solubilized humins and humins oil...... 206

xvii

Figure C. 5 (a) Conversion of humins and total oil yield and (b) breakdown GC-detectable class of compounds and their yield in the oil obtained from hydrotreatment in IPA and MeOH without initial H2. .... 209

Figure D. 1 The heat-up profiles for microreactor and batch glass vial reactors heated in oil bath...... 210

Figure D. 2 Concentrations of glucose, fructose, and mannose as a function of time-on-stream...... 211

Figure D. 3 Glucose conversion and product yields vs. residence time...... 212

Figure D. 4. Control experiments: (a) 1 wt% glucose with HCl catalyst at pH=1.25 and 160 °C (no chromium chloride); (b) 1 wt% glucose with no catalyst at 160 °C...... 212

Figure D. 5 UV-Vis spectra of the 3.4 mM of CrCl3 solution heated for 24 h at 140 °C, and the permeate and retentate from ultracentrifugation at (a) 200-350 nm region and (b) 350-800 nm region...... 213

o Figure D. 6 Glucose concentration (CGLU) vs. time (t) at 160 C using a 1.7 mM CrCl3 catalyst which was not preheated...... 214

Figure E. 1 Glucose conversion and selected product (fructose and HMF) yields vs. residence time...... 215

Figure E. 2 Carbon yields of minor products (mannose, fructose, levoglucosan, formic acid, levulinic acid) vs. residence time at temperatures of (a) 160, (b) 180 and (c) 200 °C...... 216

Figure E. 3 Glucose conversion and product yields vs. time in a batch reactor...... 217

Figure E. 4 Yields of minor products (mannose, fructose, levoglucosan, formic acid, levulinic acid) vs. residence time at 200 °C using 2-pentanol/aqueous ratios (O/A) of (a) 2 and (b) 3...... 217

Figure E. 5 Maps of flow pattern vs. the org/aq (v/v) ratio and the total volumetric flow rate using various organic solvents: (a) 2-pentanol and (b)MIBK...... 218

Figure E. 6 Maps of flow pattern at the residence times used in our experiments for (a) 2-pentanol/water solvent system with flow rate ratios of 1, 2 and 3 (b) MIBK/water system at flow rate ratios of 1:1...... 218

Figure E. 7 Carbon yields of minor products (mannose, fructose, levoglucosan, formic acid, levulinic acid) vs. residence time at 200 °C using MIBK/aqueous ratios of 1:1...... 219 xviii Figure F. 1 (a) Glucose conversion and (b) mannose, (c) fructose, and (d) HMF yields vs. residence time at 140 oC at varying catalyst concentrations. .. 221

Figure F. 2 (a) Mannose, (b) fructose, and (c) HMF yields vs. glucose conversion at 140 oC at varying catalyst concentrations...... 222

Figure F. 3 (a) Glucose conversion and (b) mannose, (c) fructose, and (d) HMF yields vs. residence time at 140-200 oC...... 223

Figure F. 4 (a) Mannose conversion and (b) glucose, (c) fructose, and (d) HMF yields vs. residence time at 140-180 oC...... 224

Figure F. 5(a) Fructose conversion and (b) glucose, (c) mannose, and (d) HMF yields vs. residence time at 140-180 oC...... 225

Figure F. 6 (a) HMF conversion, (b) formic acid and (c) levulinic acid yields vs. residence time at 140-180 oC...... 226

Figure F. 7 Glucose conversion and mannose, fructose, HMF, FA, LA and LG carbon yields vs. residence time at (a) 140 °C and (b) 160 °C...... 227

Figure F. 8 Glucose conversion and mannose, fructose, HMF, FA, LA and LG carbon yields vs. residence time at (a) 140 °C, (b) 160 °C, and (c) at 180 oC...... 228

Figure F. 9 Glucose conversion and mannose, fructose, HMF, FA, LA and LG carbon yields vs. residence time at (a) 160 °C and (b) at 180 oC. .... 229

Figure F. 10 Glucose conversion and mannose, fructose, HMF, FA, LA and LG carbon yields vs. residence time at (a) 160 °C and (b)180 oC...... 229

Figure F. 11 Glucose conversion and mannose, fructose, HMF, FA, LA and LG carbon yields vs. residence time at (a) 180 °C and (b) 200 °C...... 229

Figure F. 12 Glucose conversion and mannose, fructose, HMF, FA, and LA carbon yields vs. residence time at (a) 140 °C, (b) 160 °C, (c) 180 °C and (d) 200 °C...... 230

Figure F. 13 Glucose conversion and mannose, fructose, HMF, FA, and LA carbon yields vs. residence time at (a) 160 °C, (b) 180 °C and (c) 200 °C...... 231

Figure F. 14 Glucose conversion and mannose, fructose, HMF, FA, and LA carbon yields vs. residence time at (a) 160 °C, (b) 180 °C and (c) 200 °C...... 232

xix Figure F. 15 Glucose conversion and mannose, fructose, HMF, FA, and LA carbon yields vs. residence time at (a) 160 °C, (b) 180 °C and (c) 200 °C...... 233

Figure G. 1 Turbidity vs. mass concentration for H-BEA, Sn-BEA, Zr-BEA, TiO2, and ZrO2 particles suspended in 2-pentanol...... 235

Figure G. 2 Percent partitioned into the 2-pentanol phase for H-BEA, Sn-BEA, Zr-BEA, TiO2, and ZrO2 particles mixed with water/2-pentanol solvents...... 236

Figure H. 1 Permission for Chapter 2...... 238

Figure H. 2 Permission for Chapter 4...... 238

xx LIST OF SCHEMES

Scheme 1.1. Schematic of cellulose, hemicellulose and lignin, and their locations in a plant cell wall...... 2

Scheme 1.2 Representative structural units of (a) crystalline cellulose, (b) hemicellulose, and (c) lignin...... 4

Scheme 1.3 Formation of DHH intermediate from HMF proposed by Horvat et al. 25.. 8

Scheme 1.4 Humins formation from HMF via DHH proposed by Lund and coworkers...... 8

Scheme 1.5 Pathway of humins formation proposed by Sumerskii et al...... 10

Scheme 1.6 Structures of humins derived from (a) glucose and (b) xylose proposed by van Zandvoort et al. 29 ...... 11

Scheme 1.7 (a) Protonation of different anomeric oxygen atoms (1-5) of glucose and subsequent products, together with their respective calculated ΔG values...... 13

Scheme 1.8 Proposed reaction pathway for the aldose-ketose tautomerization...... 17

Scheme 1.9 Proposed reaction network for the tandem isomerization-dehydration reaction of glucose conversion to HMF by CrCl3 and HCl catalysts in aqueous phase...... 18

Scheme 2.1 Structure of (a) fructose-derived humins proposed by Sumerskii et al.,28 (b) HMF-derived humins proposed by Patil et al.,27 and (c) glucose- derived humins proposed by van Zandvoort et al.29, 57 ...... 36

Scheme 2.2 (a) Structure proposed in the computational work.56 with Mw of 252. .... 55

Scheme 3.1 Refined humins formation and growth network...... 83

Scheme 5.1 Schematic overview of the microchannel flow reactor setup...... 121

Scheme 6.1. Schematic of the microchannel flow reactor setup, and picture of the coiled reactor section...... 148

Scheme 7.1 Proposed model for the molecular structure of alkali-treated, glucose-derived humins by van Zandvoort et al. 57 based on 2D-NMR data...... 169 xxi ABSTRACT

This thesis focuses on some of the grand challenges in biomass conversion to platform chemicals, and specifically, on humins and on converting batch processes to continuous processes. Humins are carbonaceous, polymeric byproducts formed during acid-catalyzed, hydrothermal processing of sugars to bio-based platform molecules, such as 5-hydroxymethylfurfural (HMF). Currently, humins are a low-value byproduct used mainly for combustion. Handling these solid particles increases maintenance costs.

Minimization of humins formation and/or their valorization is essential for improving the process economics of the biorefineries. The lack of thorough understanding of humins’ structure hinders efforts towards their valorization. To bridge this gap, in the first part of the thesis, we developed methods to infer the microstructure of humins through solubility experiments. First, we conducted dissolution experiments in various solvents and correlated the solubility data to molecular properties of solvents to develop suitable descriptors. Next, Liquid Chromatography-Mass Spectrometry (LC-MS) was used to determine the species solubilized from humins in different solvents. Finally, multi-stage dissolution experiments were done to investigate the spatial homogeneity of humins as they get dissolved from the solid. Both Fourier Transform Infrared Spectroscopy (FT-IR) and LC-MS were used to characterize the solubilized and insoluble humin fractions obtained at each stage. Based on our results, an inhomogeneous structure, where macromolecular and molecular components are connected through weak forces, was proposed.

xxii Next, we use for the first time operando ultra-small angle X-ray scattering

(USAXS) to investigate the evolution of size, volume fraction, and number concentration of humins formed from fructose and HMF in acidic solutions. The radius of gyration (Rg) of suspended humins particles grows linearly with respect to time accompanied by an increase in polydispersity (PD), before precipitation occurs. An apparent activation energy of humins growth was found to be 102 kJ/mol. By comparing the growth starting from fructose or HMF, we concluded that humins form mainly from HMF or its derivatives. The time evolution in the number of particles revealed two competing processes, namely formation of new particles characterized by inception of oligomers and humins nanoparticles < 20 nm as well as the aggregation of particles leading to potential precipitation. The results are briefly compared to those of growth of silica, a better- understood colloidal system.

Armed with better understanding of the structure of humins, we then investigated the one-step valorization of humins in methanol using carbon supported catalysts. First, we screened four different noble metal catalysts. Aromatic hydrocarbons, phenols and esters were the main liquid products. Rh/C gave the best GC-detectable oil yield and was chosen for a subsequent set of experiments with varying reaction parameters. Up to 12 wt% light, GC-detectable oil yield was achieved at 75 % humins conversion. The light oil yield was shown to be a strong function of hydrogen pressure and temperature. High pressure and intermediate temperatures, time and catalyst loadings were found beneficial for the light oil yields. As temperature and time increase, the total oil yield decreases as a result of gasification. The product distribution shifts in favor of aromatics and phenols at

xxiii high temperatures and long reaction times and of esters at short reaction times and high catalyst loadings.

Lastly, we investigated continuous flow sugar chemistry. We leveraged a new group flow microchannel reactor with good heat and mass transfer for the effective conversion of glucose to HMF using CrCl3/HCl catalysts. Comparison with batch reactor data acquired at 140 °C demonstrated a two-fold increase in HMF productivity at otherwise identical reaction conditions. Then, we extended the study to 2-pentanol/water biphasic solvent mixtures to continuously extract HMF and improve its yield. We achieved the best HMF productivity of 33.6% at 51.4 % yield in 92 s under 200 °C.

Compared to our single-phase reaction where 30.3% HMF yield is achieved in 120s, the

2-pentanol/water biphasic system shows a two-fold increase in productivity. Changing the 2-pentanol/water volumetric ratio from 1 to 3 did not lead to significant changes in glucose conversion or HMF yields possibly due to the process being kinetics rather than mass transfer limited. The maximum HMF yield of 43.0% was reached at 60s in the

MIBK/water system, and the yields of other products, such as mannose and humins, also showed notable differences. Possible factors contributing to the different reactivity of different solvents are discussed.

xxiv Chapter 1

INTRODUCTION

1.1 Conversion of Biomass to Chemicals

Biomass research and potential industrial applications of bioproducts have drawn an enormous interest in the past decade to mitigate challenges associated with climate change and pollution. Energy-efficient and cost-competitive conversion technologies to transform biomass into bioproducts and energy materials can nurture the growth of a carbon neutral society. The first-generation biofuels, e.g. biodiesel and bioethanol, are produced directly from food crops via transesterification and fermentation, respectively.1, 2 This approach consumes food grains and competes with food driving up the global food grains prices. To solve the “food versus fuel” dilemma, second-generation biofuels, derived from non-edible lignocellulosic biomass, were introduced. Examples of inedible biomass include agricultural residues (corn stover, corn cob, husk, and sugarcane bagasse), forest wood, municipal solid wood waste, and food waste like potato peels. Inedible biomass is attractive for several reasons: first, it does not compete with food sources; second, it is inexpensive and abundant. For example, 1.3 ×109 metric tons of dry biomass are produced in the US each year. 3

1

Scheme 1.1. Schematic of cellulose, hemicellulose and lignin, and their locations in a plant cell wall. Reprinted with permission from Zakzeski et al.4 Copyright (2010) American Chemical Society.

Lignocellulosic biomass consists of three major components: cellulose (38%-

50%), hemicellulose (23%-32%) and lignin (15%-30%). 5 They together make up the major component of the plants cell walls. The minor components include lipids, pigments and wax. Scheme 1.1 illustrates the three major components in a typical cell wall microfibril. Cellulose is a highly crystalline biopolymer composed of monomeric glucose units linked through glycosidic bonds. Hemicellulose is an amorphous heteropolymer made up of C5 and C6 sugar monomers including xylose, arabinose, mannose, and galactose. For example, a typical corn fiber xylan, which is a common form of hemicellulose found in many plants, contains 48–54% xylose, 33–35%

2 arabinose, 5–11% galactose, and 3–6% glucuronic acid.6 While the 3D structured cellulose is strong, recalcitrant and difficult to solubilize and hydrolyze in water and in acid catalysts, amorphous hemicellulose has weaker inter-unit bond strength and can be easily hydrolyzed. Lignin is an amorphous biopolymer of natural aromatic phenolic units connected through various C−O and C−C linkages such as β-O-4, -O-4, β-5, β-

β’, 5,5 and others. It acts as a glue for cell wall, holding cellulose and hemicellulose units together and making lignocellulosic biomass recalcitrant. 5Among these linkages, the most abundant is the β-O-4 type C-O linkage, taking up 49-65% of all the linkages depending on the type of wood.7 A typical soft wood lignin contains 49-51 β-O-4, 6-8

7 -O-4, α-α’, 9-15 β-5, 9.5 5-5, and 2 β-β’ linkages per 100 C9 units. Common phenolic units in lignin are syringyl (S) and guiacyl (G). 4 Lignin, containing up to 40% of biomass carbon, is considered as a promising and natural source for bio-based aromatic compounds. Representative chemical structures of cellulose, hemicellulose and lignin are depicted in Scheme 1.2.

3

Scheme 1.2 Representative structural units of (a) crystalline cellulose, (b) hemicellulose, and (c) lignin. The lignin structure was adapted with permission from Murzin et al. 8 Copyright (2012) Springer Nature.

4 The first step to lignocellulosic biomass processing is its deconstruction into its three major components. This process is also known as pretreatment. Often hemicellulose is also hydrolyzed in the pretreatment step to form lower molecular weight saccharides. The deconstructed cellulose requires an additional step, known as hydrolysis, to be broken down to saccharides. Common hydrolysis methods are enzymatic9 and acidic. Acidic hydrolysis can be done using dilute aqueous acid solutions10 or concentrated acid or acidic ionic liquids3. Milling or mechano-catalytic breakdown of cellulose can weaken or partly cleave its chemical glycosidic bonds,11 thereby improving the yield of lower molecular weight saccharides upon hydrolysis. C6 and C5 monosaccharides, e.g., glucose and xylose, can be upgraded to various bioproducts through different chemical processes. One of the most studied catalytic upgrade processes involve sugar isomerization followed by dehydration to furfural platform chemicals. These platform chemicals can then be used as bio-renewable feedstocks to produce alkane jet fuels12, functional chemicals and monomers for polymers. 13 14

For example, glucose isomerization by Lewis acidic or base catalysts produces fructose, which is conventionally done using enzymes by food industries. Fructose dehydration using Brønsted acid catalysts produces 5-hydroxymethylfurfural (HMF), 5

15 one of the top platform chemicals listed by the U. S. Department of Energy. HMF can be upgraded to levulinic acid (LA) and formic acid (FA) via rehydration16, 2,5- furandicarboxylic acid (FDCA), a potential replacement for petroleum-based terephthalic acid to produce polyester 13, 2,5- dimethyfuran and 5-ethoxylmethylfurfural

5 (potential fuel additives) via selective hydrodeoxygenation and etherification reactions17, 2,6-hexanediol5 via selective furan ring opening and various other bioproducts. Innovative cascade reactions using multifunctional catalysts can potentially improve process economics. LA is an important intermediate feedstock to produce bio-based industrial solvent such as γ-valerolactone (GVL) 18, 19, and fuel additive such as alkyl levulinates19. Detailed reviews on the applications of HMF 5, 20 and LA19 can be found elsewhere. This dissertation focuses on catalytic pathways to produce HMF from glucose as well as the structural understanding and valorization of humins byproducts made in the sugars chemistry. This chapter reviews various catalytic reactions used for biomass valorization.

1.2 Humins

In soil science, humins are defined as the fraction of the humic substances (HS) that is insoluble in an aqueous solution at any pH value. 21 In sugar chemistry, which is the focus of this dissertation, humins are the solid polymeric byproduct derived in the acid-catalyzed dehydration of carbohydrates during production of furfural or HMF.

Humins are currently underutilized and its chemical structure is not well understood.

Humins’ molecular structure is discussed in Chapter 2. In Chapters 2 and 4, the structure of humins and lignin is compared. The structural characterization methods of humins and their upgrade to high value bioproducts are inspired from the rich literature on lignin.4, 22, 23 Currently, humins are combusted for heat, a low-value application.

Typical chemical compositions of humins are 64-67 wt% C, 28-31 wt% O, 24 with the

6 balance being H. This high oxygen content makes humins a poor fuel. Existing literature on the formation mechanism and applications of humins will be discussed here and in more detail in Chapters 2-4.

1.1.1 Literature Review on Humins Formation

Prior reports have described the formation pathway of humins in water from

HMF, although no conclusive agreement was reached due to difficulty in detecting transient reaction intermediates. In 1985, Horvat et al.25 studied humins formation using nuclear magnetic resonance (NMR) spectroscopy. They proposed that HMF first converts to 2,5-dioxo-6-hydroxyhexanal (DHH) intermediate (Scheme 1.4), which then polymerizes to humins. However, the reaction pathways from DHH to humins were not discussed. The authors presumed that DHH was very highly reactive and could not be detected. Following Horvat’s work, Lund and coworkers suggested that humins form via the aldol addition and condensation of DHH with aldehydes and ketones (e.g., HMF and its ring-opened intermediates) that form in the reacting solution.26, 27 An “idealized humins structure proposed is given in Scheme 1.5. They studied humins formation by adding benzaldehyde into the reaction medium containing HMF, HCl (catalyst) and water (solvent) and found evidence of benzene rings in the humins structure using

Fourier Transform Infrared Spectroscopy (FT-IR).

7

Scheme 1.3 Formation of DHH intermediate from HMF proposed by Horvat et al. 25

Scheme 1.4 Humins formation from HMF via DHH proposed by Lund and coworkers. (a) HMF or DHH reacting with another DHH molecule via aldol addition to form a dimer. (b) Idealized humins structure. “R” denotes HMF or DHH groups. Adapted with permission from Reference27. Copyright (2011) of American Chemical Society.

Sumerskii et al. 28 synthesized humins from various pentoses and hexoses using sulfuric acid as catalyst. They proposed several humins formation pathways based on the FT-IR spectroscopic analysis of humins (Scheme 1.6 (A-E)). In Pathway (A), HMF

8 undergoes condensation reactions and forms ether linkages via the -CH2OH group. In

Pathway (B), HMF forms acetal linkages via addition of the -CHO and ring -CH groups of HMF to produce a networked polymer. In Pathway (D), HMF rehydrates and this intermediate enters pathway (A). Humins formation from LA were also proposed in

Pathway (C), where a carbocation is formed at the C4 position and can either undergo addition reactions to C=O or C-OH functional groups of HMF, or electrophilic substitutions in the furan ring of HMF. The humins formation from xylose-derived furfural, Pathway (E), is simpler because furfural lacks the hydroxymethyl group. The carbocation, formed at the carbonyl group, undergoes electrophilic substitution with the furan rings of other furfural molecules.

9

Scheme 1.5 Pathway of humins formation proposed by Sumerskii et al. Pathways (A), (B) and (D) start with HMF, Pathway (C) starts from HMF rehydration product LA, and Pathway (E) starts from furfural derived from the acid catalyzed dehydration of xylose. Reprinted with permission from Sumerskii et al.28 Copyright (2010) of Springer Nature.

10 Scheme 1.6 Structures of humins derived from (a) glucose and (b) xylose proposed by van Zandvoort et al. 29based on FT-IR data. It was suggested that fructose-derived humins were very similar to glucose-derived humins. Reprinted with permission from van Zandvoort et al.29 Copyright (2013) of John Wiley and Sons.

Several works pointed out the difference in the molecular structure of humins derived from different sources, 29, 30 including glucose, fructose, HMF and xylose. Van

Zandvoort et al. suggested that the structure of glucose and fructose derived humins were very similar, while the xylose-derived humins have a more condensed structure based on their FT-IR spectra, as shown in Scheme 1.7. The roles of sugars and HMF- derived reaction intermediates, such as LA, in the formation of humins have also been studied. Summerskii 28 et al. suggested LA forms a carbocation at its C4 position, which can then either adds to the carbonyl or alcoholic functional groups of HMF, or undergoes electrophilic substitution to the furan rings of HMF (Scheme 1.6, Pathway

(C)). Yang et al. 31 published a systematic computational study on the Brønsted acid catalyzed reaction pathways of fructose and glucose. They found that the O2H group at the anomeric carbon was the preferred protonation site for fructose that leads to a series of facile reactions towards forming HMF (Scheme 1.8a). Scheme 1.8b shows that HMF dehydrates and becomes protonated at C5 position, forming a carbocation referred to as

H-1. Then H-1 forms an ether (H-1-H in Scheme 1.8b) with another HMF molecule, similar to the Pathway (A) proposed by Sumerskii et al.28 in Scheme 1.6. H-1 can also react with a fructose molecule in the β-fructofuranose form to form a humins precursor

(H-1-FF in Scheme 1.8b) with a similar free energy change (ΔG). For glucose, protonation on the O-5 atom leads to isomerization to fructose, protonation on the O2-or

11 O-3 atom leads to direct formation of LA, while protonation on the O-1 and O-4 atom leads to the formation of humins and reversion products such as disaccharides including maltose and cellobiose (Scheme 1.8a).31, 32 Gibbs free energy calculations showed that the humins formation is actually favored via fructose and LA formation. This explains the low selectivity of LA and HMF at moderate to low temperatures when glucose is the starting substrate.

12 Scheme 1.7 (a) Protonation of different anomeric oxygen atoms (1-5) of glucose and subsequent products, together with their respective calculated ΔG values. (b) Formation of humins precursors H-1-H and H-1-FF from HMF-derived ether (H-1-H) and the β- fructofuranose form of fructose (β-FF) proposed by Yang et al. 31 with the calculated ΔG values. Reprinted with permission from Yang et al.31 Copyright (2012) of Elsevier.

1.1.2 Literature Review on Applications of Humins

Two major approaches have been developed for utilizing humins. One approach involves their utilization for biomaterial production. For example, De Jong and coworkers 33 34 incorporated humins into a poly furfuryl alcohol (PFA) network to produce composites and characterized the mechanical properties of the composites using rheometry and tensile tests. After comparing this composite with lignin/PFA derived composite, they found that higher interfacial bonding and more efficient stress transfer between the matrix and the fibers is present in the composite obtained from humins/PFA. The higher ductility of the humins-based matrix allows for a two-fold higher tensile strength in the humins/PFA in comparison with the lignin/PFA or neat

PFA tested. 34 The authors envisioned that humins-derived resins for wood adhesion and wood durability could be developed.

The second approach is transformation of humins into low molecular weight high value bioproducts. Gasification and catalytic hydrotreatment have been explored by several groups. Hoang et al. 35, 36 studied the gasification of humins for syngas production using basic catalysts such as Na2CO3. Heeres and coworkers first experimented with the dry pyrolysis of humins.24 The major products formed include

13 furanic compounds (furfural, 2-acetylfuran, 3-methylfurfural and 2-methylbenzofuran), phenolics (phenol and 2-methylphenol) and acids (levulinic and acetic acids). The gas and liquid products summed up to 30 wt% yield based on the humins weight. Later, the same group investigated the catalytic pyrolysis of humins using the ZSM-5 zeolite catalyst. 37 Despite the zeolite catalyst used, the liquid yield was still very low. The yields of aromatic products were 9 wt% and the total liquid yield was 11-14 wt%. More recently, the catalytic hydrotreatment of humins have been studied by the Heeres group using noble metal catalysts loaded on carbon as well as metal oxide supports in isopropanol (IPA) solvent. 38-40 Hydrogen was generated in-situ. Aromatic hydrocarbons and alkylphenolics were the major liquid products formed from humins.

These are valuable bulk chemicals that have a wide range of applications. Aromatics can be used as solvents, and phenolics can be used to produce phenolic resins. IPA was not inert and formed alcohol and ketones, with the majority being methyl isobutyl ketone (MIBK). It can be widely used as an industrial solvent in paint, resin, varnishes, and lacquers. 39 A brief overview of hydrodeoxygenation (hydrotreatment) is discussed in Section 1.4.

1.3 Glucose Dehydration to HMF

1.3.1 Significance of Glucose

Glucose is the most common and cheapest monosaccharide. According to the

ICIS chemical pricing inventory, glucose is nearly 1/3 the cost of fructose, making glucose an ideal feedstock to produce HMF economically. Fructose has been the

14 starting substrate for dehydration to HMF over Brønsted acid catalysts due to its simplicity. Although the Brønsted acids can convert glucose to HMF with low yields, the HMF yield can be increased significantly if glucose is first isomerized to fructose, and fructose is then dehydrated to HMF.41 Many catalysts and solvents have been developed for selective dehydration of glucose to HMF. However, low productivity, which is defined as the yield of HMF per time, is a main challenge when glucose is used at the feedstock.

In this dissertation, we focus on improving HMF production rates from glucose.

First, we summarize and discuss existing literature on the mechanism of glucose conversion to HMF, and efforts to improve HMF yield and selectivity in the batch process. Then, we describe our recent efforts on the development a flow microreactor for HMF productivity improvement. We demonstrate the best HMF productivity compared to existing literature. Next, we conduct time-dependent studies to build a kinetic model for glucose conversion to HMF that extends our group’s existing model42 of lower temperature (below 140 °C) to temperatures higher than 140 °C.

1.3.2 Glucose Dehydration and Mechanism

The most common molecular configuration of glucose in solid form is the α- glucopyranose anomer. Upon dissolution in solvents (e.g., water), some of the α- glucopyranose molecules undergo mutarotation to form the β-anomer, β-glycopyranose.

Lewis acid catalysts such as Sn-beta zeolites and metal salt chlorides (MClx) can catalyze isomerization of glucose to the fructofuranose form of fructose. 43 The

15 mechanism for glucose isomerization involves three main steps: (1) aldose ring- opening, (2) aldose-to-ketose tautomerization, and (3) ketose ring closure.44 45 46, 47

Formally, step 1 involves a hydrogen transfer from the O1 atom to the O5 atom of the aldose and step 3 involves a hydrogen shift from the O5 atom to the O2 atom in closing of the ketose ring. Step 2 involves a hydrogen shift from the O2 atom to the O1 atom and another hydrogen migration from the C2 atom to the C1 atom 44 (Scheme 1.9). This latter hydrogen transfer has been shown to be the rate-limiting step in the tautomerization step through kinetic isotope effect (KIE) experiments. 44 The tautomerization step involves (1) the intramolecular 1,2-hydride shift catalyzed by a

Lewis acid forming a cyclic intermediate, (2) an enediol mechanism activated by abstracting a proton from the C2 position, and (3) a synergistic action of the Lewis acid zeolite lattice sites and a neighboring proton donor, which may be an internal silanol defect or a water molecule adsorbed to the Lewis acid site in the preferred heterogenous

Sn-β zeolite. 45 Leshkov et al. 48 showed for the first time that Sn-beta catalyzed glucose isomerization in water proceeded the tautomerization step through an intramolecular hydride shift instead of a proton transfer. Choudhary et al.44 first demonstrated that the tautomerization mechanism is similar for homogeneous (CrCl3, AlCl3) and heterogeneous (Sn-beta zeolites) Lewis acid catalysts.

16 Scheme 1.8 Proposed reaction pathway for the aldose-ketose tautomerization. “M” refers to the metal center (e.g. Al, Cr, or Sn). Reprinted with permission from Choudhary et al. 44 Copyright (2013) of John Wiley and Sons.

Despite extensive efforts on elucidating glucose to fructose isomerization mechanisms, most studies have focused on understanding the role of Lewis acids in the tautomerization step, rather than the aldose ring-opening and ketose ring-closure steps.

Furthermore, the role of Brønsted acids in the overall isomerization remains unclear.

Recently, Qi et al. 45 theoretically investigated the cooperative roles of Brønsted and

Lewis acids in each of the three steps of isomerization using PBE0/6-311++G(d,p), aug- ccpvtz level theory for an aqueous AlCl3/HCl catalyst system. It was found that

Brønsted acid predominantly catalyzes the aldose-to-ketose tautomerization and the ketose ring closure steps, while the Lewis acid predominantly catalyzes the aldose-to- ketose tautomerization step.

The isomerization process is a reversible reaction limited by thermodynamic equilibrium49, and therefore limits the production of HMF. Generally, two routes have been proposed for the formation of HMF. Antal et al. and Locas et al. suggested that

HMF is formed from dehydration of fructose in its furanose form and occurs through a series of cyclic furan intermediates.50 51 Others and us have suggested HMF is formed through an acyclic pathway proceeding through an 1,2-enediol intermediate. 52, 53

Glucose also has competing reaction pathways that lead to formation of by-products. In one pathway, dehydration forms non-furan cyclic ether intermediates, which usually occurs in alcoholic solvents. In another pathway, C-C bond scission of sugar species

17 occurs through retro-aldol reaction usually at high temperatures or in the presence of metals and metal oxide catalysts.54, 55 However, the formation of solid humins byproduct is a significant challenge. Multiple reaction pathways for humins formation have been proposed. While the dominant pathway start with the condensation of HMF and its ring-opened intermediates, 56 sugars such as glucose, fructose and other reaction intermediates such as LA also participate in humins formation. 57 Swift et al. reported that fructose can directly convert to humins and formic acid by studying the molar ratio of FA and LA products at different fructose conversions.16 A comprehensive reaction network for the tandem isomerization-dehydration reaction of glucose conversion to

HMF by CrCl3 and HCl catalysts was proposed (Scheme 1.10). van Zandvoort et al. characterized the functional groups of humins using FT-IR and 2D-NMR spectroscopy and found evidenced for incorporation of LA units into the humins. 29, 57 A thorough study of the structure and nucleation and growth kinetics of humins particles is presented in Chapters 2 and 3, respectively.

Scheme 1.9 Proposed reaction network for the tandem isomerization-dehydration reaction of glucose conversion to HMF by CrCl3 and HCl catalysts in aqueous phase. Solid lines represent Brϕnsted acid catalyzed steps, and dashed lines represents Lewis acid catalyzed steps. Step 10 is glucose isomerization, and step 11 is glucose

18 epimerization. Steps 1, 6 and 8 are HMF forming reactions, and step 2 is the HMF rehydration reaction. Steps 3, 4, 7, 9, 13, 14, 15, and 16 are humins forming reactions. Reprinted with permission from Swift et al.42 Copyright (2017) of the Royal Society of Chemistry. As mentioned above, homogeneous Lewis acids such as metal salt chlorides

58, 59 (MClx) are effective catalysts for the glucose isomerization to fructose. These salts hydrolyze in water and coordinate to water molecules, forming M(H2O)x (OH)y complexes, and generating Brønsted acidic HCl from released Cl- ions and protons. The generated acid can also partly dehydrate fructose intermediate to HMF. Among several metal salts used, CrCl3 and AlCl3 catalyzed reactions are widely investigated because of

41, 58, 60-63 58 their effectiveness as well as speciation profiles. Choudhary et al. used

Extended X-ray Absorption Fine Structure (EXAFS) to study the CrCl3 catalyzed glucose dehydration and elucidated that the most active catalytic species for glucose

2+ 64 isomerization is the monohydroxy Cr, [Cr(H2O)5(OH)] . Similarly, Norton et al. and

46, 65 others used ESI-MS combined with NMR studies for AlCl3 catalyzed reaction and reported that the active species in aqueous AlCl3 solution is the bishydroxy

+ [Al(H2O)4(OH)2] ion.

A number of kinetic studies66, 67 68 have shown that the isomerization of glucose to fructose is first order at temperatures below 220 °C, which is the temperature range used for most studies. Matsumura et al. 67 conducted kinetic studies at higher temperatures (between 175 and 400 °C) and found that the reaction order decreases to below 1 at above 220 °C. Our work focuses on temperatures from 140 to 200 °C.

A direct pathway from glucose to HMF has also been proposed. 42, 66 The

Brønsted catalyst also catalyzes the rehydration of HMF to LA and FA, as well as the

19 condensation of HMF to humins. These side reactions of HMF can be minimized by simultaneously extracting HMF into an organic phase in a biphasic system as elaborated above. Detailed reviews on this topic can be found in the works by Saha et al.69 and

Romo et al. 70 The biphasic reaction systems including the use of phase modifiers are reviewed below in section 1.3.4.

1.3.3 Catalyst Selection for Optimal HMF Production

The HMF formation and degradation pathways given in Scheme 1.10 indicate that the key to high HMF yield is fast glucose isomerization, fast fructose dehydration to HMF, and slow HMF degradation.42 This section provides a brief review to correlate homogeneous and heterogeneous catalyst performance with their interactions with substrates.

1.3.3.1 Homogeneous Catalysts

In Section 1.1.2, homogeneous metal salt catalysts, CrCl3 and AlCl3, were discussed. To select and design catalysts rationally, it is crucial to identify key descriptors or catalyst properties that can lead to high HMF yield and selectivity. Less selective metal salt catalysts including FeCl3, VCl3 and SnCl4 were found to catalyze the competing pathways-reverse aldol condensation and humins formation. Li et al. conducted in-situ far-infrared (FIR) studies and attributed the superior performance of

59 CrCl3 to its preferential coordination to the glycoaldehyde group of glucose. In another study, Tsilomelekis and coworkers used in-situ Raman spectroscopy and

20 correlated the selectivity to HMF with the ability of the MClx catalyst to stabilize the α- anomer of glucose. It was observed that SnCl4, an active but less selective catalyst, facilitates the glucose mutarotation toward the β-anomer while CrCl3 promotes the formation of the α-anomer of glucose. 71 Not only does the metal type affects the catalytic performance, detailed mechanistic studies also revealed the role of the Cl- ion.

45 Qi et al. computed the ΔG for each step of glucose isomerization using AlCl3 as the catalyst, and proposed that Cl- ion indirectly affects the activity of the Lewis acid by stabilizing the carbocationic sugar intermediates. For the overall glucose-to-fructose

+ - isomerization, the kinetic activity follows the order of H3O ···Cl >

+ - 2+ - [Al(H2O)2(OH)2] ···Cl > [Al(H2O)4(OH)] ···2Cl . But for the Al species not

2+ associated with Cl, the kinetic activity follows the order of [Al(H2O)4(OH)] >

+ [Al(H2O)2(OH)2] .

1.3.3.2 Heterogeneous Catalysts

Heterogeneous catalysts have also drawn considerable attention due to their easier separation and recycling. Moliner et al. 49 first reported that Sn-beta is highly effective for the isomerization of glucose to fructose. In addition, this catalyst was stable in Brønsted acidic environments, which are needed for dehydrating fructose to

HMF. The mechanism for heterogeneous catalyzed glucose conversion to HMF is more complicated compared to homogeneous catalysis, because introduction of a solid phase brings additional steps, namely, species adsorption/desorption. Designing catalysts for high HMF yield and selectivity therefore requires fundamental understanding of the

21 kinetics in complex systems. Partially hydrolyzed Sn sites in zeolite beta (or open Sn sites) with proximal silanol groups were shown to be the active sites for the isomerization of glucose into fructose via a Lewis-acid mediated 1,2 hydride shift mechanism. 72 Kruger et al. 73 showed that a small fraction (1-2 wt%) of many types of zeolite catalyst dissolves in the water as oligomeric aluminosilicate fragments, which catalyzes the fructose dehydration and HMF side reactions homogeneously. However, these homogeneous species do not catalyze the glucose isomerization reaction. Swift et al. 74 conducted a combined modeling and experimental study on glucose conversion to

HMF over zeolite beta catalysts. The relative strengths of HMF and fructose adsorption strongly influence the intrinsic kinetics. It was concluded that reducing the adsorption strength of HMF while strengthening the adsorption of fructose can effectively improve the HMF formation rate. Consequently, H-BEA is not a good catalyst as it is a weaker

Lewis acid catalyst than Sn-BEA, a weaker Brønsted acid than HCl, and a strongly

HMF-adsorbing material which promotes HMF degradation reactions. On the other hand, a combination of Sn-BEA and HCl catalysts can achieve the best HMF yields.

Rajabbeigi et al. 75 also combined experimental and computational studies to model glucose isomerization using Sn-BEA catalysts. It was concluded the kinetic parameters were consistent with a reaction limited model catalyzed by framework Sn sites. Kruger et al. 76 and Otomo et al. 77 both discovered that calcination of zeolite beta at 450-500

°C generated octahedral Al as Lewis acid sites which are in close proximity to the existing Brønsted acid sites. In the glucose conversion to HMF, the calcined catalyst

22 shows a higher selectivity to HMF at similar glucose conversions compared to a

77 physical mixture of acid catalysts with a similar number of acid sites.

Besides zeolite beta, many other solid acid catalysts have been used for the conversion of glucose and fructose to HMF, including ionic exchange resins, 78 oxides, 79 phosphates, 80 and solid acid foams. 81 Ordomsky et al.82 screened alumina, aluminosilicate, zirconium phosphate, zeolite mordenite (MOR) and Amberlyst-15 catalysts for the conversion of fructose to HMF in water at 135 °C. It was concluded that the HMF selectivity correlates with the effectiveness of Brønsted acid sites with the highest selectivity over Amberlyst‐15 and MOR. 82 Lewis acidity was essential for glucose isomerization to fructose, but was also found to be responsible for the decrease in HMF selectivity due to the fast initial condensation of fructose into humins over

Lewis acid sites. 82

1.3.3.3 Lewis/Brønsted Catalyst Ratio

Due to the nature of tandem reactions, the HMF formation rates and yields have been found to change nonlinearly with the ratio of Lewis/Brønsted acid sites (L:B).

Wang et al. 83 synthesized AlNb/SBA-15 catalysts of different Al/Nb ratios between 1:1 and 1:4. An optimal Al/Nb ratio (1:1.5) achieved the best HMF yield (56%) at 94% glucose conversion at 170 °C in 6 h. Evidences from NH3-temperature programmed desorption (TPD) and FT-IR of these catalysts suggested the Al/Nb ratio of 1:1.5 gave the highest total Brønsted and Lewis acidity. 84 Swift et al. 42, 74 studied H-BEA catalyst

23 and showed that when the total number of acid sites are kept constant, as in the case of

H-BEA prepared with varying L:B ratios, the forward HMF production rate first increases with L:B, reaches a maximum at L:B values of 0.3 and then decreases at higher L:B. This phenomenon underscores the interplay between Lewis acid catalysis

(glucose isomerization to fructose) and the subsequent Brønsted acid catalysis (fructose dehydration to HMF) for improving HMF production. When the Lewis and Brønsted acid sites are varied independently, as in the case of varying the relative ratios of Sn-

BEA and HCl catalyst mixtures, the forward HMF production rate first increases with increasing L:B ratio and eventually reaches a plateau. When the Brønsted acid sites increase, both the rate of fructose dehydration (HMF formation) and HMF degradation increase. Several studies have found that HMF yields are maximum at high L:B ratios and long reaction times, although the exact value of the optimal L:B depends on the type of catalyst. 74, 77 Here, the glucose isomerization to fructose is fast, while both

HMF formation and degradation reactions are slow due to limited number of Brønsted sites. Xia et al. 84 synthesized a series of Fe/beta zeolite catalysts with various Fe/Al ratios between 0.06 and 0.18 for the conversion of glucose to HMF in a water/tetrahydrofuran (THF)/NaCl biphasic system. In order to keep the aqueous and

THF phases separate, 35 wt% NaCl was used in the aqueous phase. Under optimal reaction conditions, the Fe/beta-0.06 catalyst afforded 61% HMF yield at glucose conversion of 95% after reacting for 90 min at 120 °C. This Fe/Al ratio corresponds to the L: B of 1.2 while the highest L:B used was 1.3. A volcano behavior was also

24 observed between HMF yield and the L:B ratio, consistent with the work of Swift et al.

74

1.3.4 Solvents and Phase Modifiers for Suppressing HMF Degradation

Sugars and HMF form many side products in Lewis and Brønsted acids catalyzed reactions in water solvent. To suppress these side products and improve HMF yield and selectivity, extensive research has been done on exploiting new organic solvents for biphasic systems and phase modifiers.

Ionic liquids (ILs) have been used instead of water (solvent). ILs gave promising

43 results in the Lewis acid catalyzed reactions. Zhao et al. used a CrCl2 catalyst in 1- ethyl-3-methyl-imidazolium chloride and obtained 68-70% HMF yield from glucose.

Other ILs/Lewis acids combinations were also screened. 52, 85, 86 The major drawbacks of ILs are their high cost and instability in small amounts of water formed in the reaction.

Another approach, as briefly mentioned in Section 1.1.2, is to use a water/organic biphasic solvent system where the organic solvent has a high partitioning coefficient. FT-IR combined with density functional theory (DFT) calculations showed87 enhanced HMF stability in water/DMSO (dimethyl sulfoxide) mixtures and attributed it to the preferential solvation of the C=O group of HMF by DMSO. The solvation of HMF in DMSO increases its LUMO (lower unoccupied molecular orbital) energy and therefore makes it less susceptible to nucleophilic attack towards formation of side products. Organic solvents of moderate to high partition coefficients employed

25 include 2-secbutylphenol41, 88, 2-methyltetrahydrofuran (2-MTHF)89, tetrahydrofuran

(THF)58, 90, DMSO, methyl isobutyl ketone (MIBK) 89, 91 and primary and secondary

C3-C6 linear alcohols89. The partition coefficient (P) is defined as the HMF concentration in the organic phase divided by that in the aqueous phase at equilibrium.

Values of P at 25 °C without any salts or modifiers are summarized in Table 1.1. The separation can be improved using a suitable phase modifier, which is usually a salt.

Nikolla et al.90 used a 35 wt% NaCl aqueous solution and THF with Sn-beta and HCl catalysts and obtained 57% HMF yield at 79% glucose conversion. Binder et al. 92 reported that a water/dimethylacetamide/NaBr mixture and CrCl3 catalyst could reach

HMF yields of 81%, which is as effective as IL systems. Dumesic and coworkers 41 used sec-butylphenol (SBP) solvent which has the highest partition coefficient (Table

1.1). Good glucose conversion of 88% and HMF yield of 62% was achieved using

AlCl3/HCl catalysts and saturated NaCl as modifier at 170 °C. 97% of the HMF generated was extracted to the SBP phase. Leshkov and Dumesic89 screened various solvents including alcohols, ketones and cyclic ethers in combination with different salts, and found THF to be the most effective solvent for HMF extraction. At 150 °C,

88% glucose conversion, 78% HMF yield and 7.3 to 1 HMF in organic/water phase distribution was achieved using NaCl salt at saturation concentration. 89

Table 1.1 Partition coefficient P for HMF in selected organic solvents at 25 °C without any phase modifiers.

26 Solvent Partition Coefficient for HMF (P)

2-sec-butylphenol 10.13

2-MTHF 3.15

MIBK 2.97

1-butanol 2.12

2-butanol 2.04

The use of a suitable additive can have a synergetic effect with the organic solvent. 93 Leshkov et al. 91 developed a process for efficient conversion of concentrated

(30-50 wt%) fructose solutions to HMF using HCl catalyst. DMSO and/or poly(1-vinyl-

2-pyrrolidione) (PVP) was added to the aqueous phase to increase its polarity. 2-butanol was added to the organic MIBK phase to improve the extraction efficiency of HMF.

The optimal result achieved in all three additives was 82% fructose conversion at 83%

HMF selectivity at 180 °C in only 3 minutes of reaction time. Here, the reactor was first pre-heated to 90 °C in an oil bath, and time zero was taken when the reactor was transferred to a 180 °C oil bath.

1.3.5 Process Design

Glucose conversion to HMF has been extensively investigated using batch and continuous reactors. Batch reactors suffer from high capital investments, long start-up and shutdown times, long residence times and slow heat and mass transfer rates. In

27 recent years, glucose processing in flow microreactors is of growing interest because their small inner diameters enable fast heat and mass transfer and higher HMF productivity.94, 95 The microreactors can shorten the reaction times to seconds. Flow operation is also beneficial for high-throughput production of chemicals. 96 However, literature on glucose conversion in microreactors is sparse. Most studies used fructose as starting feed 51, 94, 96-99 and very few used glucose65, 97. The reaction conditions, the optimal HMF yields, productivity and glucose conversions are summarized in Tables

1.2 and 1.3. The yields and productivities of HMF from glucose are much lower compared to those from fructose. The Arai group published studies 100-102 in flow reactors using glucose at 350-450 oC in sub- and supercritical water without catalyst. At these harsh conditions, C-C cleavage of glucose via reverse aldol condensation was a challenge. To date, the highest HMF productivity from glucose was reported by Zhang et al. in a narrow tube with both ends sealed, which is technically a batch reactor. The tube was heated by a fluidized sand bath. A maximum 33% HMF yield was achieved in

3 mins at 180 °C. However, the time it took to reach the set temperature of 180 °C was

2 min, which was comparable to the reaction time of 3 min. This study emphasizes a drawback of conventional designs where the heat-up time is compared to actual reaction time. This makes data at low conversion and short residence times less accurate for fast reactions. Another study achieved 40% yield of HMF in the flow reactor, but at a relatively longer reaction time of 22 min. 97

Table 1.2 Summary of literature on glucose conversion to HMF in water solvent only.

28 Reference Reactor Catalyst T (°C) Other time Max. HMF Conversion Productivity

# conditions (min) Yield (%) (%) (% HMF

yield/min)

1103 Batch None 220 10 Mpa 30 32.5 72 1.08

2104 Batch HCl/AlCl3 180 250mM 3 33 95 11.0

GLU

3105 Batch TiO2 200 556mM 5 20 85 4.0

anatase GLU

4106 Batch ZrO2-P 200 1 wt% GLU 100 47 75 0.470

(55mM)

597 Flow PBS buffer, 180 1 wt% GLU 22 40 100 1.82

pH=2 (55mM)

29 Table 1.3 Summary of aqueous phase fructose conversion to HMF in flow reactors. Reference Catalyst and Temperature Other Residence Max. Conversion Productivity

# pH (°C) Conditions Time (min) HMF (%) (% HMF

Yield (%) yield/min)

198 HCl, pH=1.8 240 4MPa 2 65 86 32.5

2 99 HCl, pH=1.8 300 4MPa 0.083 50 97 600

351 H2SO4, 2mM 250 / 0.5 47 93 94

497 PBS buffer, 180 / 5 60 92 12

pH=2

594 HCl, pH=1 185 / 1 54 71 54

6107 HCl, pH=1 200 / 1 53 95 53

7108 HCl, pH=0.7 200 / 0.0667 54 100 810

To further improve the productivity of HMF from glucose and generate accurate kinetics data, we have conducted glucose dehydration in a specially designed house- built microreactor. This microreactor can reach reaction temperature within one minute without preheating the substrate. Thus, it allows accurate data collection and enables fast mass transfer, eliminating diffusion limitations.

1.4 Motivation and Scope of Thesis

This focus of this dissertation is the structural understanding and valorization of humins byproducts and glucose conversion to HMF. Chapter one described the literature overview. Chapter two presents characterization of the molecular weight and

30 microstructure of humins using analytical methods to enable analysis of molecular, macromolecular and microstructure through dissolution studies in different solvents.

Compared to glucose as starting substrate, humins form faster from fructose and in higher yields. Prior work has shown differences in the structure of humins derived from different carbohydrate sources 29, 30 and reaction conditions109. Using chromatography, mass spectroscopic and FT-IR techniques, we show that humins consist of small, soluble molecules trapped in large, insoluble macromolecular networks and connected through weak interactions.

Chapter three describes the nucleation and growth of humins particles from aqueous, acidified fructose solutions. We employed operando Ultra Small Angle X-ray

Scattering (USAXS) to monitor the evolution of particle size and conformation as a function of time, temperature and reactant concentration.

Chapter four discusses a catalytic hydrotreatment route for humins valorization.

We report conversion of humins to aromatics, esters, phenolics and other low molecular weight products using supported metal catalysts. Effects of process parameters such as temperature, H2 pressure, reaction time and catalyst loading on humins’ conversion and product distribution are discussed. 13C isotopic labeling study was also employed to investigate the role of methanol solvent in coupling with the humins-derived intermediates to form the final products.

Chapter five investigates the kinetics of glucose conversion over CrCl3/HCl in aqueous phase using a flow microreactor. Reaction conditions for higher productivity are identified. Chapter 6 uses the optimal CrCl3 to HCl ratio determined in Chapter 5

31 and extends the study to water/2-pentanol and water/MIBK biphasic systems. The effect of temperature, organic to aqueous volume ratios, and organic solvent type on the HMF yield and productivity is discussed.

Chapter seven summarizes the key findings in this dissertation and provides a perspective on the current state of humins valorization and glucose conversion to HMF.

Potential directions for future research are also discussed.

32 Chapter 2

STRUCTURAL ANALYSIS OF HUMINS FORMED IN THE BRØNSTED CATALYZED DEHYDRATION OF FRUCTOSE

2.1 Introduction

Humins are carbonaceous, polymeric by-products formed during acid-catalyzed dehydration of sugars to platform chemicals, such as furfural and 5- hydroxymethylfurfural (HMF) via reactions of HMF, sugars and other reaction intermediates produced during sugar dehydration and subsequent rehydration of HMF with water.57 Such by-products account for 10-50% carbon loss of the feed, causing unfavorable process economics.35 Furthermore the need to handle slurry poses additional challenges (i.e., plugging of processing equipment, etc.). Currently, humins are used for low-value applications, such as combustion to supply heat in biorefineries33, 34, 36.The process economics of biorefineries can be improved either by inhibiting the formation of humins or by valorizing them. Both approaches require thorough understanding of the humins' formation-process and their molecular structure.

The latter is the focus of this work.

The structure of humins has not been extensively studied27-29, 36, 56, 57, 110 because their complex and recalcitrant nature makes their characterization difficult. Existing literature on humins has mostly been focused on determining its functional groups

(representative proposed structures are shown in Scheme 1). Sumerskii et al.28 first

33 investigated the structure of fructose-derived humins using Fourier Transform Infrared

(FTIR) spectroscopy, nuclear magnetic resonance (NMR) and pyrolysis gas chromatography coupled with mass spectrometry (Pyr GC-MS). They proposed that humins consist of 60% furan rings and 20% aliphatic linkers. Patil et al.27 studied HMF- derived humins using FTIR and proposed that humins originate from aldol condensation of 2,5-dioxo-6-hydroxyhexanal (DHH), the HMF rehydration intermediate, with the carbonyl group of HMF. A conjugated network of C=C and furan rings decorated with aldehyde or ketone functional groups was proposed based on FTIR studies. van

Zandvoort et al.29, 57 synthesized humins using glucose, fructose, and xylose in aqueous sulfuric acid and characterized the resulting humins using FTIR spectroscopy and solid- state NMR. They found that humins consist of furan rings connected via aliphatic linkers, decorated with oxygen-containing functional groups, such as carboxyl, carbonyl, and hydroxyl. The structure of glucose and fructose-derived humins were similar while xylose-derived humins had a more conjugated structure. In a follow-up study, they employed 2-dimensional solid-state NMR and found that the furan rings in glucose-derived humins are connected to each other through short aliphatic chains mainly at alpha carbons.57 Similarly, Reiche et al.109 investigated the effect of pH and oxidative strength of the acid catalyst on the chemical structure of glucose-derived humins using FTIR and Raman spectroscopies and thermogravimetric mass spectrometry (TG-MS). The humins synthesized at lower pH were found to be more highly condensed. The carbonyl content was shown to increase with synthesis pH. For pH values less than 3, using nitric acid as catalyst led to higher humins yield and a

34 higher content of carboxylic functional groups compared to HCl catalyst. All these studies point to the fact that humin structure may depend on synthesis conditions.

Tsilomelekis et al.110 used dynamic light scattering (DLS), scanning electron microscopy (SEM) and FTIR to monitor the size, morphology, and molecular structure of HMF-derived humins as functions of HMF conversion and solvent media. A model, in which humins particles grow via fast diffusion-limited aggregation (DLA) followed by a slow reaction-limited aggregation (RLA) mechanism, was proposed based on DLS data. A multimodal distribution of interconnected spherical particles was observed using SEM. The particle size was around 200 nm at 10% HMF conversion and grew to

2-5 µm at 100% HMF conversion in water. Particles of ~200 nm in diameter were forming throughout the HMF conversion. It was also presented that the size of humins particles formed in the presence of DMSO reaction media (~20 nm) does not grow with

HMF conversion. The changes in the FTIR spectra were consistent with the previously proposed reaction mechanism involving nucleophilic attack of an HMF carbonyl group to the α- or β-position of the furan ring.29 However, other reaction pathways, such as aldol addition and condensation, could not be excluded.

35 Scheme 2.1 Structure of (a) fructose-derived humins proposed by Sumerskii et al.,28 (b) HMF-derived humins proposed by Patil et al.,27 and (c) glucose-derived humins proposed by van Zandvoort et al.29, 57

The current literature provides the types of functional groups present in humins synthesized from different substrates mainly based on FTIR and NMR studies.

However, the molecular weights of the components and their connectivity have not been elucidated for humins generated from any substrate under any reaction condition.

HMF-derived humins have been used as a simplified model for functional group characterization and growth kinetics studies.27, 110 Although condensation between

HMF and HMF-derived intermediates was suggested as the dominant pathway to humins formation,27, 110 the proposed formation mechanism for HMF-derived humins may not apply to humins derived from sugars.16 In addition, humins' solubility in different solvents and the characterization of soluble fragments, which are important for humins' valorization to bio-oils or other high value bio-products, have not been studied in a diverse set of solvents. Although Sumerskii et al.28 reported 20 wt% humins solubility in acetone, thorough dissolution studies correlating humins’ solubility with solvent properties are still lacking.

In this work, we use humins obtained from the Brønsted acid catalyzed dehydration of fructose. Compared to glucose, fructose has been shown to convert to

HMF and humins faster when Brønsted acid is used as the catalyst and is our substrate of choice.42, 111 First we conducted dissolution experiments in various solvents and correlated the solubility data to solvent properties for identifying suitable descriptors.

36 Next, we developed a Liquid Chromatography-Mass Spectrometry (LC-MS) method, in conjunction with gel permeation chromatography (GPC), to determine the Mw of the solubilized humins fragments. Analysis of soluble humins from multistage dissolution experiments using FTIR and LC-MS techniques elucidates the macromolecular structure of humins.

2.2 Experimental

2.2.1 Materials

Fructose (≥99% purity, Sigma Aldrich), hydrochloric acid (37 wt%, Fisher

Scientific), sulfuric acid (5M, Fluka), formic acid (HPLC grade, Sigma-Aldrich), dimethyl sulfoxide (≥99.5% , Sigma Aldrich), γ-valerolactone (99% , Sigma Aldrich), tetrahydrofuran (Optima, Fisher Scientific), acetone (≥99.5%, Fisher Scientific), dichloromethane (≥99.9%, Fisher Scientific), cyclohexane (≥99.0%, Fisher Scientific), methanol (HPLC grade, Fisher Scientific), acetonitrile (HPLC grade, Fisher scientific) and 2-(4-formylphenoxymethyl) furan-3-carboxylic acid (Sigma-Aldrich) were used without further purification. All aqueous solutions were prepared using deionized (DI) water obtained from a Millipore water purification system (model: Direct-Q3 UV R).

Humins were synthesized in the laboratory. In this method, 25.0 g of fructose was dissolved in 250 mL of 0.1 M HCl aqueous solution. The solution was transferred to a

500 mL round bottom flask and heated in an oil bath with continuous stirring using a magnetic bar and refluxed at 120 C for 24 h. Upon cooling down the solution to room

37 temperature, solid humins were separated by vacuum filtration and repeatedly washed with a total of 3 L DI water until the pH of the filtrate became neutral. This ensures the removal of physisorbed acid residue from the solid.27-29, 36, 110 Finally, the washed humins were oven dried at 80 ºC for 12 hours and ground by hand to a fine powder before use. To determine the conversion of fructose and the yields of the HMF, levulinic acid (LA) and formic acid (FA), the filtrate was diluted tenfold and analyzed using a high performance liquid chromatograph (Waters 2695 HPLC) equipped with a refractive index detector (RID, Waters model 2414) and a separation column (Aminex

HPX-87H). A 5 mM aqueous sulfuric acid solution was used as the mobile phase at a flow rate of 0.5 mL/min. The column compartment temperature was maintained at 50

°C and the refractive index detector temperature was 35 C. The characteristic peaks for organic products and reactants were identified from the retention times of the individual primary standards. Each peak was integrated, and the actual concentration of each product was calculated from their respective pre-calibrated plots of peak area vs. concentration. Under the aforementioned experimental conditions, 82% conversion of fructose was achieved. The yields of HPLC-detectable products, HMF, LA and FA sum up to 50 % on a carbon basis and the balance of 32% is assumed to be humins.

2.2.2 Solubility Studies

Screening solvents for humins solubility. Typically, 0.45 g humins were mixed with 3 mL of a solvent in a 5 mL Wheaton glass vial. A magnetic bar was added to the solution and the vial was sealed with a septum cap. It was then placed in an aluminum

38 block oil bath pre-set at 25 C and the solution was stirred at 1000 rpm for 24 h.

Insoluble humins were vacuum filtered using an oven-dried and pre-weighed 11 µm pore size Whatman filter paper, washed with 250 mL of DI water, and oven dried at 80

C for 12 h. The weight of insoluble humins was obtained by subtracting the mass of the filter paper from the total weight of the filter paper and insoluble humins. The concentration of the solubilized humins was calculated in mg/mL by subtracting the weight of insoluble humins from the initial humins weight and dividing it with the volume of the solvent. The results were averaged from three replicates. To verify thermodynamic equilibrium has been reached, a time-dependent study was conducted

(Figure A1). The effect of the filter pore size on the amount of solubilized humins was also investigated. The results in Figure A2 suggest that the concentrations of solubilized humins in ACN are similar for filters of pore sizes ranging between 20 nm and 11 μm.

Multistage solubility experiments. For the first stage, ten replicates were prepared following the procedure outlined above. The 2nd stage experiment was conducted by mixing 0.45 g of the recovered insoluble humins from the 1st stage with 3 mL fresh ACN (or methanol, acetone) and following the same procedure described for the 1st stage. Because of some mass loss during recovery of insoluble humins from the

1st stage experiments, eight replicates were prepared for the 2nd stage experiment. This approach was carried out up to 6 stages, where only three replicates could be prepared.

The solubility of humins was calculated at each stage and the soluble humins at each stage were analyzed using the LC-MS (see below).

39 Soxhlet extraction. Extraction follows a previously described procedure.110 In brief, standard Soxhlet extraction apparatus consisting of a thimble, a distillation flask, a distillation arm, and a condenser, was used. After manual washing with DI water, 2.3 g humins were placed into a thimble contained inside the distillation arm. A 250 mL distillation flask was filled with DI water and heated in a 130 oC oil bath. The water vapor traveled up the distillation arm and reached the condenser. Condensed warm water flowed back to the thimble and accumulated while gradually dissolving humins.

When the thimble volume was almost full, it was emptied back into the distillation flask by the siphon. Each cycle lasted approximately 30 minutes and the extraction was carried out for a total of 24 h. The purified humins were removed from the thimble and oven dried at 80 oC for 12 h.

2.2.3 Characterization of Soluble and Insoluble Humins

Infrared spectroscopy. All humins samples were analyzed using a Nicolet 8700

FTIR spectrometer equipped with a DTG detector and a Golden Gate single reflection diamond attenuated total reflectance (ATR) attachment. All FTIR spectra were acquired at 32 scans with a resolution of 2 cm-1. Each spectrum was normalized with respect to its total area.

LC-MS analysis. The soluble fractions of humins in ACN and methanol were analyzed using an Agilent 1260 Infinity LC equipped with a photodiode array detector and quadrupole mass spectrometer (Agilent model 6120) and an electrospray ionization

(ESI) in positive ion polarity mode. N2 was used in the ionization chamber as the drying

40 gas and kept at 200 ºC, 35 psi and a flow rate of 12 L/min. The applied capillary voltage was 4000 V. An Agilent Zorbax SB-C18 reverse phase column of dimensions 4.6 mm

250 mm 5 µm and pore size of 120 Å was used to separate the soluble humins species. The chromatograms are shown in Figures A3-A5.

The soluble humins samples collected from multistage solubility experiments in

ACN were analyzed using LC-MS to determine the Mw size of soluble humins. The collected solutions were re-filtered with a 0.2 µm syringe filter and 1 μL filtrate was injected for analysis. The LC column temperature was pre-set at 30 °C. A 60/40 (v/v) mixture of ACN and 0.1 vol% of aqueous formic acid was used as the mobile phase at a flow rate of 0.3 mL/min. The total run time was 30 min. The mass spectrometer was set to full scan over a mass range from 10-500 in positive mode. The mass spectra collected were split into two mass ion channels, 10-300 m/z and 301-500 m/z, to improve resolution and signal to noise ratio.

Humins solution in methanol was prepared by mixing 0.45 g humins with 3 mL of methanol and stirring the mixture at 25 C for 24 h. The mixture was filtered with a

0.2 μm syringe filter and 2 μL solution was injected on the LC column that was kept constant at 35 C. A linear gradient flow of methanol and 1 vol% of aqueous formic acid was used as the mobile phase at a total flow rate of 0.3 mL/min. The percentage of methanol in the mobile phase was 70% for the first 2 mins, and linearly decreased to 50

% between 2 and 50 mins, and then brought back to 70 % at 50 mins to re-equilibrate for 5 mins. The percentage of the formic acid solution was similarly adjusted. The total

41 run time was 55 min. Selected soluble humins species were quantified using the selected ion monitoring (SIM) mode of the MS. To account for the effect of molecular weight on the MS response factor, HMF was used as the standard for quantification of humins species of m/z less than 200 by monitoring the 127 mass response. 2-(4- formylphenoxymethyl) furan-3-carboxylic acid (Mw=246) was used as the standard for quantification for humins species of m/z greater than 200 by monitoring the 247 mass response. A 5-point calibration plot was developed for the standard based on the concentration of each standard and the response of the primary mass of 246 (Figure

A6). The response factors of humins fragments were assumed to be equal to that of the

127 and 247 masses of the standards, respectively. Due to this assumption, we refer to this method as semi-quantitative and the results should be considered to provide trends.

Each sample and standard solution was injected two times to ensure reproducibility of the data. The 2 mM check standards had recoveries between 98-102%.

GPC analysis. Gel permeation chromatography (GPC) analysis was conducted using a Waters 2695 HPLC equipped with a refractive index detector ((RID); Waters model 2414) and two Waters Styragel columns (dimensions: 4.6 x 300 mm with packing size of 5 µm) connected in series (Models: HR 3 and HR 4). Tetrahydrofuran

(THF) was used as the mobile phase at a flow rate of 0.3 mL/min. The column compartment and the RID temperature was set at 25 °C. To prepare the samples for analysis, 0.15 g humins were dissolved in 3 mL of a solvent (acetone, ACN, methanol and THF). After stirring for 24 h at 25 ºC, the slurry was filtered with a 0.2 µm syringe filter. Prior to GPC analysis, the THF/humins solution was diluted five times using pure

42 THF. Because an RID was used in our instrument, 0.2 mL of the other humins solutions were vacuum dried at room temperature and 1 mL of THF was added to re-dissolve the humins. 20 μL of each sample was injected. Five polystyrene (PS) standards with peak molecular weights (Mp) of 370-17600 Da were used for calibration. The average molecular weights of samples were calculated based on the retention times of the standards.

2.3 Results and Discussion

2.3.1 Solvent Screening

Following the approach of Sumerskii et al.28 for dissolving humins in acetone, we screened different classes of organic solvents, namely, polar protic, polar aprotic, and nonpolar solvents to identify suitable descriptors for dissolving humins. Under comparable experimental conditions, partial humins dissolution is observed in all solvents tested (Figure 2.1a). The fraction of humins dissolving ranges from very low

(in cyclohexane and water) to high (~20% by weight in THF) despite performing these dissolution experiments at room temperature. Significant fractions of humins dissolve in

DMSO but the viscous slurry nature of the system prevents us from carrying out filtration to determine the concentration of soluble species. In order to reduce the viscosity of the slurry, 0.1 g of humins was used, instead of 0.45 g used in typical experiments. Taking the solubility to vary linearly with humins load up to a saturation level (this behavior was indeed found in experiments in ACN), the concentration of

43 solubilized humins was multiplied by 4.5 times and compared with the other solvents.

The concentration of the solubilized humins in polar solvents was determined and plotted against various solvent properties, such as the Gutmann donor and acceptor numbers and the Hansen solubility parameters (Figure A7). The Hansen solubility parameters were chosen because the contributions of dispersion forces, hydrogen bonding and polarity, are evaluated separately.112 Among these solvent properties, only the donor number, which represents the electron donating ability of a solvent, correlates positively with the solubility of humins (Figure 2.1b). Water, however, is an exception.

As shown in Figure 2.1a, a fraction of hand-ground humins float on top of water and the rest settle to the bottom. The FTIR spectrum of the humins that settle to the bottom of the vial looks alike with that of the humins particles that stay afloat (Figure A8) indicating that both fractions have similar functional groups but different size or density. Mechanical grinding was not used to achieve sample uniformity because the microstructure of humins would possibly be altered, as demonstrated for microcrystalline cellulose.11, 113

A similar correlation between the donor number and solubility has also been reported for bituminous coal, another type of carbonaceous material.114 This correlation was attributed to electron-donor-acceptor (EDA) type interactions between the coal and the solvents (water was not included). It was proposed that the dissolution of coal occurs when EDA adduct formation between the coal and the solvent is more energetically favorable than both the adduct between the solvent molecules and that between the coal molecules. This is essentially a substitution reaction, where an electron

44 acceptor site of the solvent binds to an electron donating site of coal, disrupting the

EDA pairs within the coal molecule. The structural model of bituminous coal proposed by Shinn et al.115 suggests that it contains benzene rings connected through aliphatic linkers and decorated with oxygen-containing functional groups such as hydroxyl, carbonyl, and carboxylic acid groups. Since a similar correlation between the donor number and solubility occurs also for humins, this may indicate the presence of EDA solvent-humins interactions.

Since no clear correlation was found between the concentration of the solubilized humins and the Hansen solubility parameters (Figure A7), other weak interactions are probably not dominant contributors to the dissolution of humins. It is unclear to what extent the solvent can disrupt these interactions within humins, and what causes the insolubility of the remaining humins. But the insolubility is probably due to covalently crosslinked high Mw species present in humins. For example,

Constant et al.116 recently observed that humins produced from fructose dehydration in methanol by HCl can be fully dissolved in DMSO. It was also pointed out that 5- methoxymethylfurfural (MMF) was formed as an intermediate besides HMF. Therefore, the resulting humins possibly contain methoxyl instead of hydroxyl groups as the side chains, which precludes the formation of intermolecular hydrogen bonds.

45

40 35 DMSO 30 THF 25 Acetone 20 15 MeOH 10 DCM ACN

5 Solubilized Humins Solubilized

Concentration (mg/mL) Concentration 0 Cyclohexane Water -5 -5 0 5 10 15 20 25 30 Donor Number

Figure 2.1 (a) Pictures of solutions after adding 15 mg humins in 5 mL solvents and stirring the solutions. The vials are arranged from highest solubility (left) to lowest solubility (right). Numbers 1-8 correspond to (1) dimethyl sulfoxide (DMSO), (2) tetrahydrofuran (THF), (3) γ-valerolactone (GVL), (4) acetone, (5) methanol (MeOH), (6) ACN, (7) dichloromethane (DCM), and (8) cyclohexane. Water (vial 9) is used as a reference. (b) Donor number of the solvent vs. humins solubility for a subset of the solvents (the donor number for GVL is unavailable, and solubility in DMSO was not determined since a very thick slurry was formed, causing difficulty in vacuum filtration). The weight fractions of humins dissolved are 5.8 wt % in DCM, 7.3 wt % in ACN, 10.3 wt % in methanol, 14.2 wt% in acetone, 19.3 wt% in THF, and 23.3 wt% in DMSO.

46 2.3.2 LC-MS and GPC Analysis of Soluble Humins

Even though DMSO dissolved humins the most, it was not introduced into the mobile phase because certain organic solvents, including DMSO, THF, and acetone, decrease the ionization efficiency of the mobile phase.117 Instead, ACN and methanol were selected for LC-MS experiments, as these also dissolve humins effectively (Figure

2.1a). The chromatograms of soluble humins are shown in Figures A3-A5. The mass numbers of the major soluble species are summarized in Table A1. Most of the identified soluble species have relatively low Mw (200-600 Da). Many species of very low concentrations were found in the m/z range of 1000 to 2000 Da. In ACN, the most prevalent mass fragment is 127 that corresponds to HMF, which was not fully removed during manual washing with water. The m/z of 235 and 252 are the next most abundant.

The mass 235 corresponds to the 252 structure with a water molecule lost. In methanol,

HMF is in even higher concentration, and masses 235 and 252 are also relatively in high concentration but not as dominant as in ACN (instead, masses 219, 225, and 109 are more dominant in methanol). Clearly, the dominant masses as well as the total amount depend on the solvent. The 235 and 252 masses detected herein coincide with the hypothesized “humins precursors”.56 Specifically, computations of free energies proposed several “humins precursors” with a mass of 252 with the most stable isomer originating from the aldol condensation between 2,5-dioxo-6-hydroxyhexanal (DHH), an intermediate of HMF rehydration product, and an isomer of HMF (Scheme 2a).56

Further condensation reactions from these precursors could lead to higher Mw compounds. Since the compound (252) shown in Scheme 2a is expected to be a reactive

47 intermediate and thus to be present in solution at low concentrations, assignment of its structure may not unique. Isolating this species for NMR analysis might be helpful in determining its chemical structure.

To investigate whether HMF and low Mw humins can be removed by extensive washing with water, we performed Soxhlet extraction according to the humins purification method reported previously110 (see methods). 0.45 g of Soxhlet extracted humins were dried and subsequently dissolved in 3 mL of methanol following the aforementioned single stage dissolution procedure. The results are shown in Table A2.

The concentration of most soluble species decreased upon Soxhlet extraction. Notably,

HMF is still present in humins after Soxhlet extraction but in significantly lower concentration (0.22 mM compared to that obtained from non-extracted humins of 1.76 mM). This indicates that a small fraction of the HMF formed during fructose dehydration becomes entrapped in humins and is not removed by the Soxhlet extraction; organic solvents of high donor number are more effective than water for this task.

Summing up the amounts of all the major ACN soluble species yields 3.2 mg/mL (Table A1), which is about 29% of the total (12 mg/mL) soluble humins determined with the gravimetric method (Figure 2.2). The corresponding amount in methanol is about 30% of the total soluble humins (Figure A9b). The low recoveries in both solvents could be attributed to the semi-quantitative approach where mass fragments lower than the standard mass response are being under quantified. Another explanation is the selective ion monitoring approach where additional mass fragments may need to be monitored. Finally, the amount of large Mw species present in the

48 sample could be underestimated due to their inability to ionize, known as a

“discrimination effect” commonly observed in the MS analysis of polydisperse materials.118 To compensate for the mass discrimination effects,118, 119 GPC is often used to complement MS analysis of polydisperse polymer solutions and coal liquids.

Next we turn to the GPC results. Here the accuracy of the estimated number average molecular weight (Mn) strongly depends on the resemblance between the sample and the calibration standards used. Mn is accurate when the association between the sample and the column packing are minimal and the retention time is solely a function of molecular size; for unknown samples, such as humins, which are structurally different from the PS standards, association with the column packing could

119 occur, leading to systematic errors in estimating Mn. Although not an absolute value,

Mn can be compared across samples. The results of GPC analysis are shown in Table

2.1 and Figure A10. Here the humins used were not purified by Soxhlet extraction before being dissolved in organic solvents. Observed high polydispersity index (PDI) values are well above 1, implying broad Mw distributions. The peak with a shoulder for all samples, as shown in Figure A10, indicate a bimodal Mw distribution. Our results show that acetone, ACN, methanol and THF dissolve fractions of similar Mn in the range from 1100 to 1300 Da but distinct Mw distributions given the differences in the

PDI values, especially for THF, which appears to be a more selective solvent. For the

ACN solubilized humins, 70% of the fragmented species have masses below 2000 Da.

The LC-MS data show very low concentrations of masses higher than 600 m/z. This difference stems most likely from the inability for the MS to ionize high Mw species

49 and potentially the GPC calibration (see discussion above). However, the low masses determined by both LC-MS and GPC are in qualitative agreement, indicating that a significant fraction of dissolved species consists of fairly small compounds. This may be because the humins used in our study were generated at relatively low temperature and mild pH. In literature, the effect of sugar feedstock, acid catalyst type and pH on the chemical structure of humins has been studied using FTIR, solid-state NMR, Raman spectroscopy and TG-MS. 29, 109. However, the impact of these reaction parameters on the molecular weight and solubility of humins is still poorly understood. Our approach could easily be applied to other humins made from other feedstocks and under various conditions.

To access the effect of Soxhlet extraction, humins were dissolved in acetone,

ACN, methanol and THF following Soxhlet extraction. As shown in Figures A11(a)-

(d), the majority of the soluble humins is still seen after Soxhlet extraction. The molecular weight distribution of the soluble fractions become slightly tighter compared to humins that have not undergone Soxhlet extraction, as indicated by the PDI values in

Table 2.1. The Mn values are similar while the weight average molecular weight (Mw) values are slightly smaller. This is because the molecular weight has greater weighting in Mw than in Mn. Since the Mw decreases only slightly, this shift in molecular weight distribution might be too small to be reflected in Mn. Overall, Soxhlet extraction does not remove all the organic-soluble material and removes preferentially smaller humins fragments.

Table 2.1 Mn , Mw and PDI values for solubilized humins determined using GPC.

50 Solvent Use of Soxhlet Extraction Mn Mw PDI

Acetone No 1142 2517 2.20

Acetone Yes 1109 2127 1.92

ACN No 1102 2632 2.39

ACN Yes 1125 2523 2.24

MeOH No 1170 2857 2.44

MeOH Yes 1057 2333 2.21

THF No 1173 2439 2.08

THF Yes 1289 2269 1.76

2.3.3 Multistage Dissolution Experiments

Structural characterization of hydrothermal carbon (HTC), a similar type of carbonaceous material generated by the hydrothermal treatment of glucose without any acid, has suggested a spherical structure of the material consisting of a dense hydrophobic core and a less dense hydrophilic shell.120, 121 Tsilomelekis et al.110 provided evidence that humins particles grow via a fast DLA mechanism followed by a slow RLA mechanism. RLA tends to form tightly compact structures as compared to

DLA.122

To study the structural and compositional homogeneity of humins, we conducted a six-stage dissolution experiment. We focus on ACN first and discuss methanol in brief later. Figure 2.2 shows that the concentration of the solubilized

51 humins, as determined from the gravimetric method, decreases significantly from stage

1 to stage 2. This observation closely resembles the trend observed in the analytical approach using a LC-MS method. Given errors present in the gravimetric data, the trends in the solubilized humins from stages 2 to 6 follow qualitatively those of the LC-

MS data. More specifically, the LC-MS results show a sharp drop in the total concentration from stage 1 to stage 2 with nearly half of the total concentration in both stages being attributed by the masses of the proposed structures. Furthermore, the mass of HMF is very abundant in stage 1 but observed minimally in stage 2 and can hardly be detected after stage 4. Notably, species with m/z of 252 appearing at 11 minutes of retention time, as depicted in Figure 2.3, which makes up nearly 40% of the total dissolved humins in stage 2, has almost completely vanished after stage 2. In the remaining stages, the concentration values appear to level off, and trends closely resemble the results obtained from the gravimetric approach. Moreover, several trends in the mole fraction of monitored species as a function of dissolution stage are observed

(Table A3). For example, with each dissolution step, the mole fraction of species with

Mw 151, 191, 217, 241, 247, 315 and 343 increases, with Mw 235 decreases, with Mw

109 and 225 does not change significantly, while that of other species passes through a maximum or minimum (e.g., Mw 99, 127 and 252). This observation suggests that each stage is solubilizing different levels of the ion fragments, which are being monitored, and their abundance in subsequent stages is directly impacted by the previous dissolution step. Together with the spatial heterogeneity deduced by Tsilomelekis et al.,110 it appears that humins are spatially and chemically heterogeneous materials.

52 The existence of similar low Mw species in a macromolecular network of

114, 123 bituminous coal has been proposed based on multistep dissolution studies.

Aharoni and Edwards124 suggested that branched low Mw oligomers are structurally inflexible and can be trapped within the macromolecular polymer network. Vahrman et al.125 conducted a four-step dissolution experiment of bituminous coal and proposed that the coal molecules dissolve slowly because they are in restricted pores. Since humins can be thought of as a non-porous material35 with a highly branched chemical structure

(Scheme 1), the gradual dissolution of humins fragments over multiple stages could be explained by their difficulty in diffusing out of the macromolecular network.

Additionally, Painter et al.126 proposed a model to predict the thermodynamics of coal/pyridine solutions; it was proposed that the low Mw and low aromatic fragments are soluble, while the high Mw, highly aromatic coal fragments, upon addition of solvent, would phase separate into a solvent-rich (i.e., a homogeneous solution) phase and a solvent poor (i.e., swollen, gel-like coal) phase. Consequently, they inferred that successive steps result in continued dissolution of the high Mw fragments. Although humins and bituminous coals exhibit similarities regarding the presence of aromatics

(furan rings in the case of humins) and high Mw fragments, the relative amounts of sp2 and sp3 C-C bonds, which could have a substantial effect on the rigidity of the backbones in both materials, are not well understood and analogies may be limited.

53 14 12 10 8 6 4

2 SolubilizedHumins

Concentration (mg/mL) Concentration 0 0 1 2 3 4 5 6 7 Stage

Figure 2.2 Concentration of solubilized humins in ACN at stages 1-6. 0.45 g of insoluble humins obtained from the previous stage was mixed with 3mL of ACN and stirred at 25 oC for 24 h at 1000 rpm. The number of replicates was 10 at the first stage and decreased to 3 at the 6th stage.

8.0x105 1.5x105 (a) Stage 1 (b) Stage 1 Stage 2 Stage 2 Stage 6 6.0x105 Stage 6 1.0x105

4.0x105

5.0x104

MSD Signal MSD MSD Signal MSD 2.0x105 0.0 4 8 12 16 20 24 4 8 12 16 20 24 Time (s) Time (s)

Figure 2.3 Total ion chromatogram (TIC) of ACN solubilized humins for stages 1, 2, and 6 at (a) 10-300 m/z and (b) 301-500 m/z. Stage 3, 4, 5 (not shown) closely resemble the Stage 6 TIC.

54

Scheme 2.2 (a) Structure proposed in the computational work.56 with Mw of 252. The same Mw was detected for the first time in our MeOH and ACN solubilized humins. (b) Proposed structure based on functional group information from FTIR (vide infra) and m/z numbers from LC-MS in ACN solubilized humins with Mw of 342. (c) Proposed structure in methanol solubilized humins with Mw of 340. Because of the protonation in the MS, the actual m/z observed for these species are 253, 343, and 341, respectively.

In order to provide insights into the molecular structure and probe plausible structural differences of dissolved and undissolved humins, FTIR spectra have been acquired at each stage of dissolution and compared with the initial fructose- derived humins (Figure 2.4a). The FTIR spectra of the initial humins are in excellent agreement with those previously reported for sugar-derived humins.29, 30 The highly overlapping bands in the spectra underscore the complexity of their molecular structure.

Comparison of the spectra in the 1450-1800 cm-1 range indicates a consistent decrease of the 1671 cm-1 vibrational band with increasing dissolution stage. The rest of the vibrational bands of the insoluble humins remain the same as those of the parent humins. We have previously ascribed the 1671 cm-1 peak to the carbonyl C=O functionality of HMF-like units that are incorporated in humins at late stages of reaction

(not at full sugar conversion).110 In our previous work, the glucose-derived as well as

55 fructose-derived humins at almost complete sugar conversion was not present the 1671 cm-1. Given that the 82% fructose conversion in the present work, all findings underscore that HMF actively participates in the growth of humins. The ease of dissolving those species indicates the presence of weak forces that hold the “core” of the humins with the HMF-like polymeric units bound on the external surface or that simply HMF-like species are physically adsorbed on the humins particles and cannot be fully removed by Soxhlet extraction with water.

56 (a) 2925 cm-1 1671 cm-1

Stage 6 Stage 5 Stage 4 Stage 3

Stage 2 Absorbance Stage 1 Initial

3000 2800 1800 1600 1400 1200 1000 800 Wavenumber (cm-1)

-1 -1 -1 (b) 1705 cm 1671 cm 1516 cm 2850-2950 cm-1 Stage 6

Stage 5

Stage 4

Stage 3

Stage 2 Absorbance Stage 1

Initial

3000 2800 1800 1600 1400 1200 1000 800 Wavenumber (cm-1)

Figure 2.4 FTIR spectra of (a) initial humins and insoluble humins remaining after each stage of dissolution. The 2925 (C-H stretching) and 1671 cm-1 (C=O stretching) peaks are highlighted for comparison across each stage. (b) initial humins and solubilized humins in ACN at each stage of dissolution. The 2850-2950 (C-H stretching), 1705, 1671 cm-1 (C=O stretching) and 1516 cm-1 (C=C stretching) peaks are highlighted for comparison across each stage.

57 The spectra of the solubilized humins after solvent evaporation and of the initial humins are shown in Figure 2.4b. The 1450-1800 cm-1 spectral range shows significant and consistent changes with increasing dissolution stage, especially in the C=O and furan ring vibrational range. Initially an increase (relative to the rest of the bands) of the

1516 cm-1 and 1671 cm-1 modes occurs. These correspond to the stretching frequencies of the furan ring and carbonyl group of HMF-like units, respectively. Interestingly, comparing the spectra of solubilized humins at stage 1 with those in early stages of

HMF conversion, many similarities are seen (e.g., the 1671 cm-1 mode is always higher in absorbance than the 1705 cm-1 one at low reactant conversion).110 This observation is consistent with Figure 2.4a which shows that HMF-like units dissolve first. With increasing dissolution stage, the 1516 cm-1 and 1671 cm-1 bands consistently decrease and the C=O contribution from aliphatic ketones becomes dominant (1705 cm-1).

Structure (a) in Scheme 2 is an example of such a molecule with a non-conjugated C=O group, which is formed by aldol condensation of DHH with an HMF isomer. This structure agrees with the proposed structure of Ref.56, which presumably forms at early times. Comparing the spectra of the solubilized humins at late stages with those of the initial and early stage humins, it is reasonable to assume that the external surface of the humins contains units which formed in early stages of fructose dehydration via the ring opening mechanism of HMF (acid catalyzed hydrolysis) leading to DHH proposed originally by Horvat25. As components leave the surface of the parent humins and become solubilized, a new surface is exposed with molecular structure similar to the

“core” of the insoluble humins.

58 In the high wavenumber frequency region of the insoluble humins, we observe a decrease in the low intensity vibrational band at 2950 cm-1 (Figure 2.4a), which corresponds to the stretching vibration of sp3 hybridized C-H groups (a blow up of the

3100-2800 cm-1 region is shown in Figure A12). At the same time, in the high wavenumber frequency region of the solubilized humins, we observe a consistent increase in the vibrational bands in the 2750-2950 cm-1 range (Figure 2.4b), which all correspond to the stretching vibration of sp3 hybridized C-H groups. Since these bands increase with increasing dissolution stage, they probably belong to the primary humins polymers formed at very early stages. This indicates that species with Mw 151, 191,

217, 241, 247 and 343, whose mass fraction, as determined by LC-MS, also increases with increasing dissolution stage, are probably rich in these functional groups and actively participate in the initial humins formation. Other soluble species with higher

Mw, which are not detectable by the LC-MS, might also have contributed to the changing spectral pattern observed. More advanced spectroscopic techniques sensitive to C-H bonds, such as 2D-NMR, coupled with isotopic exchange experiments can shed further light towards the accurate elucidation of the complex molecular structure of humins.

To assess the generality of the results, the multistage dissolution experiments were repeated using methanol and acetone as solvents. Figure A9 show that the dissolution pattern in methanol follows the same trend as in ACN. Figure A13 compares the IR spectra of the initial, methanol solubilized, and methanol insoluble humins collected at each stage. The spectra of methanol insoluble humins from stages 1 through

59 6 are very similar to the spectrum of the initial humins. However, the spectra of the methanol solubilized humins shown in Figure A13a show increasing intensities for sp3

C-H (which is not monotonic as opposed to ACN) and non-conjugated carbonyl groups, and decreasing intensities for furanic C=C and conjugated carbonyl groups. These trends are mostly consistent with those in ACN (Figure 2.4). Although the LC-MS and

GPC data indicate that the Mw of methanol- and ACN-solubilized humins are slightly different, these species still have the same types of functional groups and collectively yield similar IR spectra.

2.4 Conclusions

The macromolecular structure of humins derived from the HCl catalyzed dehydration of fructose was investigated using multistage dissolution studies complemented with LC-MS, GPC, and FTIR studies. The donor number of the solvent is correlated with the solubilized weight fraction of humins, suggesting that solvents interact with humins by possibly disrupting the EDA-types of interactions between humins fragments. Solvents of high donor number dissolve significant fractions of humins even at room temperature in a single stage, and this finding may be useful for humins utilization, for example, by selecting a good solvent for catalytic hydrotreatment that improves the contact between fragmented humins and the catalytic sites. This is an important finding given that profound advances in lignocellulosic biomass conversion to valuable products have been achieved, yet carbon loss to humins remains a key barrier to commercialization. Subsequent stages lead to smaller but substantial and

60 approximately similar dissolution. Most of the solubilized humins species have mass numbers ranging from 200 to 600 Da, as detected by LC-MS, and below 2000 as found by GPC. Our results indicate that a significant fraction of humins are agglomerates of oligomeric species held together by weak interactions rather than humins being entirely macromolecules.

Interestingly, the acetonitrile and methanol solubilized species are chemically different but possess similar FTIR spectra indicating similar functional groups. Using

FTIR and LC-MS data, possible structures of dominant species are identified for the first time and indicate that multiple growth mechanisms, proposed in the literature, are at play. The FTIR for the original, soluble, and insoluble fractions indicate that the dissolution of humins at ambient temperature does not involve covalent bond breakage.

Solubility data along with LC-MS and FTIR analysis of the soluble humins species from each stage of the dissolution suggest that humins are structurally and chemically inhomogeneous, as the associated, soluble species dissolving in later stages contain more aliphatic linkers (sp3 C-H) and C=O (non-conjugated with C=C), which likely form via condensation of furanic species. Finally, the methodologies developed in this study could be applied to the characterization of humins generated in industrial processes from various feedstocks and under various conditions.

61 Chapter 3

GROWTH KINETICS OF HUMINS STUDIED VIA X-RAY SCATTERING

3.1 Introduction

Humins are waste by-products formed during acid-catalyzed hydrothermal processing of sugars to platform molecules, such as 5-hydroxymethylfurfural (HMF),57 from uncontrolled cross-polymerization reactions of HMF and reaction intermediates.57

Their formation has a detrimental effect on process economics, as humins are of low- value products that can only be used for combustion127 or syngas production35, 36.

Additionally, they can cause reactor plugging and maintenance issues. Efforts to transform humins to value-added products36, 128 are hindered by limited understanding of the mechanism of their formation. Although the chemical constituents of humins have been elucidated using infrared 110 (IR) and nuclear magnetic resonance (NMR) 27,

57, 116 spectroscopy, certain aspects of their structure remains poorly understood. First, the number and the types of C-C and C-O linkages have not been determined. Second, their microstructure has not been studied in detail. The formation and growth of humins were studied by Tsilomelekis et al.110 using low temperature dynamic light scattering

(DLS) of a 5 wt% aqueous HMF solution. At 70 ºC, pH of 1 and reaction times less than 3 h, the hydrodynamic radius (RH) of humins particles grows at a constant rate of

2.4 nm/min.110 Studies using the DLS instrument were limited to a maximum temperature of ~70 oC. At these temperatures, the conversion of fructose to HMF is too

62 slow, and thus, studies of humins formation from fructose via DLS cannot practically be conducted. Furthermore, the particle size at longer reaction times could not be measured due to the increasing turbidity and light absorption by the sample. Both of these phenomena were associated with the precipitation of larger particles.

In order to overcome the above limitations of DLS, here we use in-situ ultra- small angle X-ray scattering (USAXS) to monitor the size and morphology of fructose- derived humins particles over time. Given the X-ray transparency of the samples, we reasoned that X-ray scattering would not be influenced by the turbidity or the color of the samples. First, we present the growth of humins from fructose, as a function of temperature and time. We then use fructose and HMF as two separate substrates to qualitatively differentiate the two competing pathways of nucleation and growth. Lastly, a refined humins formation and growth scheme is proposed.

3.2 Methods

3.2.1 Materials

D-fructose (≥99% purity, Sigma Aldrich), 5-hydroxymethylfurfural, and hydrochloric acid, (37 wt%, Fisher Scientific) were purchased from Sigma-Aldrich and used without further purification. The hydrochloric acid solutions were prepared by diluting a 37 wt% concentrated stock using deionized water.

63 3.2.2 Ultra-Small Angle X-ray Scattering

Small Angle X-ray Scattering (SAXS) is a well-established tool to obtain information about colloidal samples such as size, shape, volume fraction, and interactions between the scatterers.129 When an X-ray beam shines on a sample, some

X-rays are reflected, most transmit through the sample without changing direction, while some are scattered by the sample. There are two types of scattering: elastic scattering occurs when the kinetic energy of the X-ray is conserved before and after being scattered, and inelastic scattering occurs when the kinetic energy of the X-ray changes. The SAXS instrument works with elastic scattering only. The scattering vector, q, is defined as:

⃗⃗⃗ ⃗⃗⃗ 푞 = 푘푠 − 푘푖

⃗⃗⃗ ⃗⃗⃗ where 푘푠 is the wave vector describing the scattered direction, and 푘푖is the incident wave vector. The scattering angle, θ, is the angle by which the trajectory of the incident

X-ray beam is changed. In elastic scattering, the wavelength of the X-ray, λ, the distance between lattice planes, d, and the angle θ are related by Bragg’s Law:

휆 = 2푑푠푖푛휃 q is in reciprocal space related to the d in the normal space through the relationship:

2휋 푞 = 푑

This is an important relationship that indicates that when we refer to large length scales, we observe them at low q regimes; conversely, we observe small length scales at high q regimes. Combining the above equations, we obtain:

64 4휋푠푖푛휃 푞 = 휆

Ultra-Small Angle X-ray Scattering (USAXS) experiments were conducted at the X-ray Science Division beamline 9-ID-C at the Advanced Photon Source, Argonne

National Laboratory (ANL). It is a versatile instrument that provides a wide range of q values to probe length scales ranging from nanometers to microns. We used photons with X-ray wavelength of 0.59 Å, obtained by beamline monochromator equipped with

Si (111) optics. This USAXS instrument uses Bonse-Hart geometry to reach ultra-small angles of q = 0.0001 Å-1 and can be optionally combined with SAXS and wide-angle X- ray scattering (WAXS) pinhole devices to cover range all the way up to q =6 Å-1.

Details regarding this specific instrument can be found elsewhere.129 For our experiments, however, only the USAXS setup, was used. The beam size was 0.8 × 0.8 mm. The USAXS instrument was operated in slit smeared configuration with slit length of 0.028 Å-1.

In a typical experiment, 0.3 mL of aqueous solution of 10 wt% fructose acidified with HCl to a pH value of 0 was injected into an NMR tube (Fisher Scientific, Catalog

No.16-800-161). The samples were heated in a 10-slot temperature-controlled aluminum capillary holder. A USAXS scan was taken every 3 minutes. For each scan, the fast scan mode was used, covering the q range of 1×10-4 to 0.3 Å-1 in 90 s. For

130 spheroids, the radius of gyration Rg is related to radius of the particle (R) by:

3 푅 = √ 푅 푔 5

The elastic scattering intensity for a monodisperse system is given by:130

65 푁 I(q) = 푉2(∆휌)2푃(푞)푆(푞) 푉 푝 where N is the number of particles, V is the total volume of the system, Δρ is the scattered intensity difference between the suspended particles and the solvent, P(q) is the form factor that includes information on the shape of the particles, and S(q) is the structure factor that accounts for the interaction between particles in concentrated suspensions. When analyzing our dilute system, S(q) was approximated to be equal to 1.

The volume fraction of particles is defined by the total volume of particles over the total solution volume:

푁푉푝 φ = 푉

The number of particles per unit volume (N/V) can be calculated from the fitted

φ and Vp values.

Incorporating the volume fraction into the expression for the scattered intensity

I(q) above, I(q) is rewritten as:

2 I(q) = 휑푉푝(∆휌) 푃(푞)푆(푞)

When the system is polydisperse, I(q) becomes:130

∞ 푁 I(q) = 푉2(∆휌)2 ∫ 푓(푟)[푟2푗 (푞푟)]2푑푟 푉 푝 1 −∞ where f(r) is the distribution function of sizes, and j1 is the Bessel function.

The polydispersity PD was calculated as follows:

Standard deviation PD = 푀푒푎푛 푠푖푧푒

66 The Porod analysis is used for obtaining the shape of particles:

퐼(푞)~푆푞−훼 = 푆푞−(6−푑) where α is the Porod (or power law slope) of the particle, and d, which equals 6-α, is the fractal dimension of the particle. For rods, α equals 1. For mass fractals, α is between 2 and 3. For fractal-like shapes with rough surfaces, α is between 3 and 4. For a smooth sphere, α equals 4.

Data reduction and analysis were done with the Irena software package in IGOR

Pro 7.01 using the procedure described by Ilavsky et al.131 In brief, data was fitted in slit smeared form using the Irena data analysis package.131 A unified fit model for dilute spheroids was used to obtain the power-law slope from the Porod regime.132 A Gaussian size distribution model for dilute spheroids was used to obtain the mean Rg and the standard deviation. An example of the fit is shown in Figure B1. The volume fraction

(φ) can also be given by the software. The total number of particles per volume of solution can be calculated from the volume fraction and the particle size.

3.2.3 Typical Reactivity Experiments

To determine the concentration of HMF as a function of time during humins growth, a typical time-dependent experiment was done in our lab. We used 5 mL batch glass vial reactors loaded with 2 mL of the same reaction mixture. The heating was done using an aluminum block containing multiple slots for oil baths. Each vial was sealed with a crimp cap and put into the oil baths preheated at 80 °C. To be consistent with the reaction conditions at the USAXS beamline at ANL, no stirring was used.

67 After desired reaction times, each vial was taken out and immediately soaked in an ice bath to quench the reaction.

3.2.4 Analysis of the Soluble Reaction Products

To determine the conversion of fructose and the yields of the HMF, levulinic acid (LA) and formic acid (FA), the filtrate was diluted tenfold and analyzed using high-performance liquid chromatography (Waters 2695 HPLC) equipped with a refractive index detector (RID, Waters model 2414) and a separation column (Aminex

HPX-87H). A 5 mM aqueous sulfuric acid solution was used as the mobile phase at a flow rate of 0.5 mL/min. The column compartment temperature was maintained at 50

°C and the refractive index detector temperature was 35 C. The characteristic peaks for the products and reactants were identified from the retention times of the individual primary standards. Each peak was integrated, and the actual concentration of each product was calculated from their respective pre-calibrated plots of peak area vs. concentration.

3.3 Results and Discussion

3.3.1 Effects of Temperature, Substrate, and Concentration

To understand the structural evolution of the suspended humins particles, we recorded USAXS spectra during fructose dehydration at different temperatures. A

68 representative raw scattered intensity I(q) vs. scattering vector (q) plot is given for one set of conditions at various times in Figure 3.1a. Figure 3.1b shows a zoom-in at a specific time. The scattered intensity I(q) increases with time at low q values (0.0002-

0.002 Å-1) but remains low and slow varying at higher q values. As such, we analyze the data of the low q regime that provides quantitative information. The radius of gyration as a function of time and temperature is depicted in Figure 3.2. At short reaction times for all four temperatures, the scattering was too weak to extract meaningful particle size data. At intermediate times and lower temperatures (80 and 85

°C), the Rg increases linearly with time, while at longer times, the growth slows down.

The same trend in particle size vs. time was also reported by Tsilomelekis et al. 110 The linear regime at short times indicates the lack of diffusion limitations (a square root dependence is common for diffusion-limited growth). We attribute the plateau in Rg to the precipitation of the large suspended humins particles (see also below). Apart from precipitation, the scattering pattern collected at longer times suggests the formation of a secondary, smaller size population (Figure 3.1b, between 0.001 and 0.01 Å-1), whose intensity was too weak to be fitted. This small population indicates that new, small size humins form during most of the growth period. At higher temperatures (90 and 95 °C),

Rg continuously increases before exceeding the detection limit of the USAXS instrument and therefore a plateau was not observed. In all cases, the power law slopes of the scattering patterns range between 3 and 4, indicating the suspended particles have fractal-like rough surfaces.132 This finding is consistent with the rough spheres observed in prior work using SEM imaging of humins.27, 35, 110 The lack of a plateau along with

69 the larger Rg values at higher temperatures indicate that the particle size alone is not sufficient to demarcate the existence of a plateau and precipitation as the main physical phenomenon. The particle density at a given size and the growth vs. precipitation rates collectively affect the observed behavior.

(a) 1000 7min 27min 48min 100 67min

) 87min

-1 107min 128min 10 148min

168min I(q)(cm 1

0.1 1E-4 0.001 0.01 0.1 q (Å-1)

(b) 1000

100 )

-1 10

1 I(q)(cm

0.1

0.01 1E-4 0.001 0.01 0.1 -1 q (Å )

70 Figure 3.1 (a) Evolution of I(q) vs. q in a typical experiment. Data shown here were obtained from the reaction of 10 wt% fructose aqueous solution acidified with HCl to a pH of 0 and heated at 80 ºC. (b) I(q) vs. q plot at 62 min of reaction time for 10 wt% fructose at pH of 0 and 85 °C. The two Guinier-Porod regimes at 0.00026-0.001 Å-1 and 0.001-0.01 Å-1 indicate two particle size populations. Due to the weak intensity, scattering from the second population at higher q could not be fitted.

1400 21.8 nm/min 14.2 1200 nm/min 1000 9.7

(nm) nm/min

g 800 R 600 95oC 4.95 90oC nm/min 85oC 400 80oC 20 40 60 80 100 120 140 160 180 Time (min)

Figure 3.2 Mean radius of gyration (Rg) of humins particles as a function of time at various temperatures grown in a 10 wt% fructose solution at pH = 0. The lines guide the eye. Growth rates correspond to the initial growth regime (before the plateau when one is observed).

Taking the derivative dRg/dt of the first regime (shorter times), we constructed an Arrhenius plot to obtain an apparent activation energy Eapp for the humins formation

(Figure 3.3) of 102±0.4 kJ/mol. It is worth noting that this value is determined by directly monitoring the growth of humins. This is in contrast to the common literature

71 method of determining the yields of fructose, FA and LA, and by estimating humins from the carbon balance. Our estimated Eapp for humins formation falls within the 94-

115 kJ/mol range reported in the literature. 27, 133, 134

3.2 3.0 2.8

2.6 /dt) g 2.4 2.2

ln(dR 2.0 1.8 1.6 2.70 2.73 2.76 2.79 2.82 2.85 1000/T (K-1)

Figure 3.3 Arrhenius plot for humins formation rate from fructose at 80, 85, 90 and 95 2 ºC. A linear fit gives an R value of 0.998 and an apparent activation energy Eapp of 102±0.4 kJ/mol.

Figure 3.4 shows the change in the polydispersity index (PD) of humins particles over time. From 40 to 100 minutes of the reaction at 85 ºC, the PD of humins increases from 0.1 to as high as 0.6. The continuous formation of new particles was also been inferred in the work of Tsilomelekis et al.110 using SEM to monitor the morphology of humins at different HMF conversions. The gradual increase in PD is consistent with the Carothers’ equation,135 suggesting that the humins follow a step- growth polymerization model that does not require an initiator. In this model, the

72 monomers react with each other to form dimers, trimers, longer oligomers and eventually polymer networks. Continuous formation of humins, rather than an initial burst in nucleation, is consistent with the increasing polydispersity with time.

Next, we varied the initial fructose concentration and monitored the particle volume fraction and Rg. Figure 3.5a shows a much higher volume fraction of particles at higher fructose concentration. Yet, Figure 3.5b shows that the growth rate is independent of fructose concentration, and identical (within experimental error) to that when HMF is the substrate. These findings strongly suggest that fructose barely participates in humins formation. Combined with prior work, showing new IR peaks in humins when fructose is the reactant,30 the results indicate adsorption of the sugar onto the humins rather than chemical incorporation. Interestingly, the number of particles per solution-volume increases over time for 50 wt % fructose but decreases for 10 wt % fructose (Figure 3.5c). The same dependence is also observed at other temperatures (at

85, 90 and 95 °C for 10 wt % fructose; Figure B2a, and at 70 and 80 °C for 50 wt % fructose; Figure B2b; see Supplementary Information (SI)). We attribute the larger number of particles at higher fructose concentrations to the higher HMF concentration that drives nucleation of new humins. The increase (decrease) of the population arises from competing mechanisms discussed below.

73 0.7 0.6 0.5 0.4 0.3

0.2 Polydispersity 0.1 0.0 40 50 60 70 80 90 100 110 Time (min)

Figure 3.4 Evolution of polydispersity PD of growing humins over time at 85°C from a 10 wt% fructose solution at pH of 0. A Gaussian size distribution was used to obtain the mean and variance of particle sizes needed to calculate the PD.

74 (a) -4 50 wt% Fructose (b) 900 8.0x10 10 wt% Fructose 4.45 nm/min 800 6.0x10-4

700 4.0x10-4

600 4.95 nm/min Rg (nm) Rg -4 Volume Fraction Volume 2.0x10 500 50 wt % Fructose 10 wt % Fructose 0.0 400 60 80 100 120 140 160 180 60 80 100 120 140 160 180 Time (min) Time (min)

(c) 3.0x108 50 wt% Fructose

) 8 10 wt % Fructose 3 2.5x10

2.0x108

1.5x108

1.0x108

5.0x107 # of Particles (1/cm Particles of #

0.0 60 80 100 120 140 160 180 Time (min)

Figure 3.5 Evolution of (a) volume fraction of humins particles, (b) radius of gyration (Rg), and (c) number of particles per volume of solution at 50 wt% and 10 wt% initial fructose concentrations. Reaction at 80 °C and pH=0. In most early kinetics studies,136, 137 fructose was assumed to convert to an unknown intermediate which subsequently reacts with HMF to form humins. To our knowledge, the only Brønsted-catalyzed fructose dehydration kinetics study at low temperature (70-150 °C) which included the direct conversion of fructose to humins was published by Swift et al. 16 The reported reaction rate for the conversion of fructose to humins was faster than that of HMF to humins. The difference from our observations could be due to the model assumption that all four tautomers of fructose directly convert

75 to humins at identical rates or model uncertainty arising from multiple sets of parameters in regression. Very little literature data is available for the degradation of fructose tautomers in water other than β-pyranose, which only accounts for 45-55% of all the fructose over our temperature range as shown by Kimura et al.138 and by Swift et al.42 Swift et al. also used a KCl buffer in their experiments, which is known to decrease the selectivity to humins.139 Although the effect of KCl on the kinetics of each pathway to humins has not been studied in detail, a recent in-situ Raman spectroscopy study by

Ramesh et al.140 showed that the molecular structures of glucose-derived humins formed in the presence of metal chloride catalysts AlCl3 and SnCl4 are quite different from each other. AlCl3 species promote the pathway that involves incorporation of furan rings

(possibly HMF to 2,5-dioxo-6-hydroxyhexanal, or DHH), while SnCl4 species promote probably the fragmentation of glucose and fructose prior to the formation of HMF. It is also unclear what fructose degradation intermediates participate in humins formation.

Moreover, the amount of humins was estimated as the balance between the initial reactant and the known detectable products rather than being directly measured. The existing kinetics studies were carried out with stirring, while in our study the reactants were not stirred in order to avoid disturbance to the X-ray measurement. At early times of the reaction with fructose as the starting material, the number of HMF molecules is low, and the absence of stirring could possibly hamper the diffusion of HMF which in turn slows down the growth rate of humins compared to a stirred system. In summary, both model assumptions and transport effects may have led to the observed differences in the humins formation rates starting from fructose as the substrate.

76 To investigate the effect of substrate, a solution containing 10 wt% HMF and 1 mol/L HCl (pH=0) and another containing 10 wt% fructose and 1 mol/L HCl (pH=0) were heated at 70 ºC. Representative scattering patterns in Figure 3.6a show that the scattered intensity of HMF-derived humins at low q values (0.0001 Å-1) is approximately 10 times greater than that of fructose-derived humins. As was the case with fructose, no useful information could be extracted from the high q regime (>0.01

Å-1) data, suggesting lack of structural features in sizes smaller than 200 Å. Therefore, no useful information can be extracted from the high q data for fructose. By extrapolation of the Arrhenius plot shown in Figure 3.3, we estimate a growth rate of fructose-derived humins at 70 ºC to be 2.32 nm/min, which is slightly faster than the growth rate of 1.92 nm/min for HMF-derived humins (Figure 3.6b). The similar growth rate may indicate a common intermediate for growth, irrespective of the substrate

(fructose vs. HMF). Since the growth rates of fructose and HMF-derived humins are comparable, their particle volumes must also be similar, assuming a similar growth mechanism. Therefore, the significantly higher I(q) of HMF-derived humins at low values of q can be attributed to a greater number of humins particles formed. This is consistent with the conclusions of Lund and coworkers30 and suggests that the direct formation of humins from fructose is slower. Instead, the predominant mechanism of humins formation is condensation of HMF molecules or its derivatives. Since the HMF concentration during the fructose reaction is low (Figure B3a), the humins formation pathway involving the coupled condensation between fructose and HMF molecules could not be assessed from our data.

77 In the HMF experiments, the number of particles per volume increases over time

(Figure 3.6c), similarly to what is seen at higher fructose concentration experiments but opposite from the behavior at low fructose concentrations (Figure 3.5c). The change in the number of particles at reaction times longer than 500 min could not be observed due to the rapid increase of the volume fraction leading to multiple scattering and particle- to-particle interactions under which the model assumptions about the structure factor no longer hold.

104 (b) 1400 (a) 10 wt% HMF 103 10 wt% Fructose 1300 1.92 ) 2 1200

-1 10 nm/min 1100 101 1000

0 Rg (nm) Rg

I(q) (cm I(q) 10 900 -1 10 800 10-2 700 10-4 10-3 10-2 10-1 100 200 300 400 500 -1 q(Å ) Time (min)

(c) 1.2x108

) 3 1.0x108

8.0x107

6.0x107

4.0x107 # of Particles of # (1/cm 2.0x107 100 200 300 400 500 Time (min)

78 Figure 3.6 (a) Scattered intensities of humins originating from 10 wt% fructose and 10 wt% HMF solutions heated at 70 °C and pH of 0, taken at 449 min of reaction time. (b) and (c) Evolution of the corresponding Rg and number of particles as a function of time for the HMF solution. The growth rate value indicated corresponds to the growth (initial) regime.

We also varied the initial concentration of HMF between 2.5 and 10 wt% and estimated the rate of particle growth. The effect of initial HMF concentration was not studied in our prior work.110 The growth rate increases with the initial HMF concentration, as shown in Figure 3.7. A second-order polynomial fit was used. A log- log plot shown in Figure 3.7b gave a slope between 1 and 2, as well as a low coefficient of determination (R2). The nonlinear dependence of the growth rate of humins on HMF concentration underscores the complexity of the growth pathways of humins. We attribute this nonlinearity to the multiple paths contributing to growth, including condensation between HMF molecules, the condensation of HMF with water-soluble humins oligomers, and the addition of HMF to the surface of humins particles. Taking

Figures 3.5-3.7 together further underscores the limited effect fructose plays on the growth of humins compared to the dominant role of HMF.

79 (a) 5

4

3 y=1.367-0.512x+0.082x2 R2=0.998

2 Rate (nm/min) Rate 1

0 0 2 4 6 8 10 12 HMF Concentration (wt %)

(b) 5

4

y=1.44x+0.238

3 R2=0.88 ln (Rate) ln 2

1 0.5 1.0 1.5 2.0 2.5 3.0 ln (HMF Concentration)

Figure 3.7 (a) Humins growth rate at different initial HMF concentrations of 2.5 wt%, 5 wt%, 7.5 wt% and 10 wt%. A second-order polynomial fit is shown here. (b) Log-log plot for humins formation rate vs. initial HMF concentration. Reactions at: 80 °C and pH=0.

Overall, the results can be rationalized by considering the HMF concentration as key in condensation chemistry for inception of new particles and growth of existing ones and the competing mechanisms controlling population size and growth rate.

Specifically, the change in particle population is dictated by (1) the inception of new particles via condensation chemistry, controlled mainly from the level of HMF

80 concentration and happening continuously as the increasing polydispersity indicates

(Figure 3.4), and (2) particle agglomeration and precipitation of larger, denser, or less soluble particles that reduce the number of particles. Rg increases due to monomer condensation and particle aggregation and decreases when particles crash out of solution. Due to difference in molecular weight, the HMF experiments have a higher initial molar concentration of substrate compared to fructose. Furthermore, the HMF concentration is low in low fructose concentration solutions (Figure B3) and much lower than that when HMF is the substrate (Figure B4). Higher concentrations of fructose result in higher HMF concentrations. When the monomer condensation dominates due to a high HMF concentration, the population increases with time irrespective of the substrate; conversely, when the HMF concentration is low, the number of particles decreases with time after a fast, initial growth due to aggregation and/or precipitation dominating over condensation.

81

Figure 3.8 (a). Reaction mixture at 100% HMF conversion after heating at 120 °C for 48 h, before filtration and (b) after filtration using a 0.2 µm syringe filter. This long time results in complete HMF conversion. (c) After filtration, the solution was heated at 120 °C for another 24 h. Large humins particles still form even after all the HMF has been consumed.

To confirm the proposed pathway of oligomer aggregation by Tsilomelekis et al.,110 we ran the reaction in HMF at 120 °C for 48 hours until 100% HMF conversion.

A black solid suspension was formed at the end of the reaction (Figure 3.8a). The absence of HMF in the reaction mixture was confirmed by HPLC analysis, and FA and

LA were the only soluble products detected in 11.1% and 62.1% carbon yields, respectively. The corresponding carbon yield of humins is 26.8%. Then a 0.20 m pore size filter was used to remove the black solid precipitates. The filtrate is shown in

Figure 3.8b. The filtrate containing FA, LA, and humins oligomers smaller than the filter pores was loaded into a clean vial and heated at 120°C for another 24 hours. After that, the color of the suspension became darker, indicating the formation of visible, larger polymers, as shown in Figure 3.7c. This observation is consistent with a condensation polymerization growth model, as oligomers retain reactivity.

82

Scheme 3.1 Refined humins formation and growth network. Black dashed lines show the addition of HMF to other HMF molecules, soluble oligomers, primary particles, or particle aggregates. Black solid lines show the process of particle formation, growth and aggregation. The sizes of the black spheres are not drawn to scale and are for illustration purposes only.

Based on the above observations, we refine the network proposed by

Tsilomelekis et al.110 As shown in earlier work by Lund et al. and van Zandvoort et al.,

HMF undergoes ring opening, condensation56 and nucleophilic attack29 to form soluble oligomers. As the oligomers increase in size by addition of more HMF, a solubility threshold is reached leading to formation of primarily particles. The mechanism shown in Scheme 3.1 involves oligomers and small primary particles <20 nm. These primary particles, referred to above as a secondary population, are smaller than 20 nm and

83 cannot be quantified by USAXS. The oligomers having degrees of polymerization between 3-10 have been identified in slow HMF degradation experiments at room temperature reported by Galkin et al.141 Lund and coworkers also proposed energetically feasible structures of HMF dimers in their computational work.56

However, it is quite possible that there is a continuum of sizes <20 nm. The primarily particles then grow into “larger particles” by addition of HMF units, soluble oligomers, or aggregation with other particles. At short times, the number of particles and polydispersity increase indicating that condensation chemistry to oligomers and aggregation to small primary particles dominate over particle growth leading to larger particles. As the HMF concentration decreases and that of the particles increases, the aggregation between particles becomes eventually dominant and continues even after all the HMF has been consumed.

3.3.2 Connection to Nucleation and Growth Theory of Silica

Although the formation and growth of humins have not been studied as much, the nucleation and growth of particles in general have been extensively studied using silica precipitated from aqueous solutions as the model system.142-145 Here, a base catalyst, usually ammonia, is added to a tetraalkoxysilane in an alcohol solution. The reaction is described by the following equations:

푂퐻− ≡ 푆푖푂푅 + 퐻2푂 → ≡ 푆푖푂퐻 + 푅푂퐻 (Hydrolysis)

푂퐻− 2 ≡ 푆푖푂퐻 → ≡ 푆푖푂푆푖 ≡ +퐻2푂 (Condensation)

84 Solid particles increase in size by molecular addition where soluble species deposit on the solid surface or by aggregation with other solid particles.144 Herein, two extreme cases have been proposed for the growth of silica particles. Matsoukas et al. 145 presented a monomer addition model where the particle nucleation is the result of reaction between two hydrolyzed monomers. The hydrolysis step was found to be rate- limiting. Nucleation happened only in the earliest stage, and growth by monomer addition became dominant and the number of particles stayed constant thereafter. In this addition mechanism, the particle size distribution remains very narrow during growth until the end of the process. 146 Bogush et al. 144 presented an aggregative growth model where particles grow solely by aggregation with other particles. Small nuclei form throughout the entire reaction. Once the aggregates reach a certain size (and thereby a certain colloidal stability because of their surface charge), the growth continues only by aggregation with the small sub-particles and not by collisions with other (larger) particles.147 Subsequently, the size distribution of the particles was rather uniform. The final equilibrium distribution and the size are determined by parameters such as the surface charge.146

Our results clearly indicate that despite silica and humins growth sharing reactive events rather than a classical nucleation mechanism, the signatures of growth are definitely different between the two systems.

85 3.4 Conclusions

We have used in-situ USAXS for the first time to investigate the evolution of size and morphology of growing humins particles from fructose and HMF. We show that the humins form fractal particles that grow at rates of 5-25 nm/min, depending on the reaction temperature. We estimated the apparent activation energy of humins growth from fructose to be 102±0.4 kJ/mol. Based on the nearly equal growth rates of fructose- and HMF-derived humins, combined with the larger particle population of HMF- derived humins, we find that condensation of HMF is the predominant mechanism for humins formation and fructose may simply physisorb onto humins.

For the first time, the inception and growth of humins from HMF have been connected to existing theories developed about the nucleation and growth of silica particles. The rate of humins growth was shown to be quadratic (or a power law of

~1.5) in initial HMF concentration, underscoring the complexity of humins growth pathways. Using USAXS, we were able to demonstrate the competition of humins inception and growth. When the HMF concentration is high, inception and growth by condensation chemistry dominate, the number of particles keeps growing, and the average radius of gyration keeps increasing. In contrast, when the HMF concentration is low, aggregation and precipitation dominate over condensation chemistry and the population decreases giving rise to a constant growth rate.

86 Chapter 4

CATALYTIC HYDROTREATMENT OF HUMINS TO BIO-OIL IN

METHANOL OVER SUPPORTED METAL CATALYSTS

4.1 Introduction

Development of efficient technologies for platform chemicals from sustainable resources has drawn great attention. Lignocellulosic biomass has been identified as a promising source to produce platform chemicals because it is abundant and does not compete with the food chain. An example of a bio-derived platform chemical is 5- hydroxymethylfurfural (HMF), which is a versatile feedstock for polymer precursor, e.g., furan dicarboxylic acid (FDCA), fuel additive, e.g., 2,5-dimethylfuran and other value-added chemicals.148 HMF is produced by the acid-catalyzed dehydration of C6 sugars149 in which humins, a polymeric carbonaceous solid, is formed as a by-product, causing ~10-50%35 carbon loss, unfavorable process economics, and plugging of equipment. Currently, humins are burned to supply heat in biorefineries.33, 34, 36 The process economics of HMF can be improved either by inhibiting the formation of humins or by valorizing them. The latter is the focus of this work.

A number of valorization routes for humins have been proposed, including catalytic gasification,35 pyrolysis,24, 37 and hydrotreatment.38-40 Hoang et al. studied the

35 gasification of glucose-derived humins using a homogeneous Na2CO3 catalyst. At 750

87 °C, almost complete conversion of humins was observed, yielding CO, CO2 and H2 gases as major products. Heeres and coworkers conducted non-catalytic pyrolysis of humins at 500 °C.24 They reported furan derivatives as the major products. Very little aromatic compounds were formed. In a recent study, they conducted H-ZSM-5 catalyzed pyrolysis of humins,37 which produced benzene, toluene, xylene, naphthalene and ethylbenzene (BTXNE) with improved selectivity. The total oil yield varied between 11 and 14 wt% depending on the source of humins. Very high temperatures were used in both gasification and pyrolysis processes, making them energy intensive.

Heeres and coworkers reported also the catalytic hydrotreatment of glucose- derived humins in isopropanol (IPA) solvent using supported noble metal catalysts.40 In this reaction, H2 was generated in-situ from IPA via catalytic transfer hydrogenation

(CTH).150, 151 Major humins-derived products were aliphatic and aromatic hydrocarbons as well as phenolic compounds. Many alcoholic and ketonic by-products, e.g., acetone and methyl isobutyl ketone (MIBK), were formed from the solvent (IPA). Among several catalysts (Ru, Pd, Pt and Rh supported on carbon, ZrO2, TiO2 and CeO2) tested,

Pt/CeO2 gave the highest GC-detectable humins-oil yield (25.9 wt%). Among the carbon-supported catalysts, Pt/C was the best and yielded 9.6% GC-detectable humins- oil. However, little explanation was given on the superior performance of the Pt/CeO2 catalyst or the role of the support. Blank experiments using only CeO2 as the catalyst were not performed. In addition, the yield of humins-oil from the solvent (IPA) itself, if any, was not reported. In a follow-up study, they varied process conditions over the

Pt/C catalyst following a central composite design approach.39 Based on the results,

88 they proposed a reaction scheme involving two parallel pathways. One pathway involves the depolymerization of the furanic network of humins to furanic monomers, and subsequent transformation to phenols, aromatics, and aliphatic hydrocarbons. The other pathway entails the aromatization of humins prior to depolymerization. The process economics was not discussed.

The aforementioned limited number of studies on humins hydrotreatment point out a number of knowledge gaps. The total humins-oil yield, which includes the light,

GC-detectable and the heavy, non-detectable fractions, was neither reported nor shown as a function of reaction parameters. Instead, the overall yield of the liquid was reported. Since the amount of unreacted IPA solvent was included in this calculation, the total liquid yield does not indicate the humins-derived oil yield. Among the GC- detectable products, the yields of other than phenolics products were not reported.

Consequently, the effect of process parameters on the distribution of products remains unclear. In addition, a better understanding of the reaction network is necessary.

Here we study the catalytic hydrotreatment of humins, obtained during the HCl- catalyzed dehydration of fructose,42, 111 over supported metal catalysts in methanol. We previously reported that approximately 20 wt% of humins could be dissolved in methanol at room temperature 152 and this solvent could potentially lead to enhanced mixing of the reactants and the catalyst to achieve a higher yield of humins-oil.

Moreover, methanol forms only gaseous byproducts and water when noble metal catalysts are used, making qualitative and quantitative analysis of the humins-derived liquid products easier. We demonstrate a higher yield of humins-derived product in

89 methanol than in IPA. Through screening experiments, we identify Rh/C as the best catalyst for the highest GC-detectable products. We hypothesize the reaction pathway based on the product distribution and isotopic labeling. To the best of our knowledge, this is likely the first report that demonstrates the coupling of the solvent with humin derived intermediates in the formation of certain products during hydrotreatment.

4.2 Methods

4.2.1 Materials

Fructose (≥99% purity, Sigma Aldrich), hydrochloric acid (37 wt%, Fisher

Scientific), sulfuric acid (5M, Fluka), methanol (reagent grade, Fisher Scientific), 13C- labeled methanol (Sigma-Aldrich), and tetrahydrofuran (Optima, Fisher Scientific) were used without further purification. Nitrogen and hydrogen gases were purchased from

Keen Gas. All catalysts (Ru, Pt, Pd and Rh in 5 wt% loading on carbon) were purchased from Sigma Aldrich and used without pretreatment. All aqueous solutions were prepared using deionized (DI) water obtained using a Millipore water purification system (model: Direct-Q3 UV R). Humins were synthesized in the laboratory following previously reported procedures. 152 In this method, 25 g of fructose was dissolved in 250 mL of 0.1 M HCl aqueous solution. The solution was transferred to a 500 mL round bottom flask and heated in an oil bath with continuous stirring using a magnetic bar and refluxed at 120 C for 24 h. Upon cooling down the solution to room temperature, solid humins were separated by vacuum filtration and repeatedly washed with a total of 3 L

90 DI water until the pH of the filtrate became neutral. This ensures the removal of physisorbed acid residue from the solid.27-29, 36, 110 Finally, the washed humins were oven dried at 80 ºC for 12 h and ground by a mortar-pestle to fine powder before use.

To determine the conversion of fructose and the yields of the HMF, levulinic acid (LA) and formic acid (FA), the filtrate was diluted ten-fold in water and analyzed using a high-performance liquid chromatograph (Waters 2695 HPLC) equipped with a refractive index detector (RID, Waters model 2414) and a separation column (Aminex

HPX-87H). A 5 mM aqueous sulfuric acid solution was used as the mobile phase at a flow rate of 0.5 mL/min. The column compartment temperature was maintained at 50

°C and the detector temperature was 35 C. The characteristic peaks for organic products and reactants were identified from the retention times of the individual primary standards. Each peak was integrated, and the actual concentration of each product was calculated from their respective pre-calibrated plots of peak area vs. concentration.

Under the aforementioned experimental conditions, 82% conversion of fructose was achieved. The yields of HPLC-detectable products, HMF, LA and FA, sum up to 50 % on a carbon basis and the balance of 32% is assumed to be humins.

4.2.2 Hydrotreatment Reactions

All reactions were carried out in a 100 mL batch reactor (Parr Model 4590 of temperature and pressure ratings of 500 ºC and 5000 psi, respectively) equipped with a temperature controller (Parr Model 4848). The reaction conditions are tabulated in

Table C1. In a typical experiment, 0.546 g humins, 0.055 g Rh/C catalyst and 10 mL

91 methanol were added into the reactor. After the reactor was sealed, it was tested for leakage using 50 bar of nitrogen. Subsequently, the pressure was released, and the reactor was flushed three times with nitrogen and then one time with hydrogen. Next, the reactor was pressurized with hydrogen and heated to a desired set temperature at a heating rate of 10 °C/min and with continuous stirring at 1700 rpm. After about 40 min

(for set temperature of 400 ºC), the solution temperature reached the set point, and the reaction was allowed to proceed for a set time. The reactor was cooled down to room temperature by removing the heating mantel. The pressure of the reactor before and after the reaction was recorded for mass balance and gas yield calculations. Two 1 L

Tedlar bags were used to collect the vapor phase for analysis of gaseous products. The remaining vapor in the reactor was released. The liquid phase was transferred into a centrifuge tube and weighed. The volume of the liquid after the reaction was also recorded. The reactor was rinsed five times with 10 mL methanol each time. All the solids and the liquid rinsate were mixed with the collected liquid phase in the centrifuge tube. The collected mixture was centrifuged at 12000 rpm for 10 min to separate the solid and liquid phases. 7 mL of the liquid phase was transferred into a pre-weighed vial and placed in a vacuum oven pre-set at 25 °C and 0.03 bar for removal of solvent and determination of oil yield. The solid phase was dried in the vacuum oven at 40 ºC and

0.03 bar for 12 h and weighed for determination of humins conversion and mass balance. Each reaction was repeated two times to ensure reproducibility of results. The yields of products and the humins conversion are summarized in Table C3.

92 Blank experiments without humins were also conducted. After the reaction at comparable conditions, the reaction mixture was collected and transferred into a graduated cylinder to measure the volume and determine the solvent conversion. The water content in the product mixture was determined by Karl-Fisher titration (see below).

For isotopic labeling experiment, 10 mL 13C labeled methanol was used as the solvent. These reactions were conducted under identical conditions as in unlabeled methanol (400 °C, 3 h, 30 bar H2) and a catalyst to humins mass ratio of 0.1. After the reaction, the liquid phase was filtered and analyzed by GC-MS.

4.2.3 Calculations of Conversion, Yield and Mass Balance

The equations used to calculate humins conversion and the yield of humins oil and gaseous products are shown below. The mass balance for the repeat reactions was close to 90 % (Table C3). Dissolved gaseous products in the solvent may contribute to a small unaccountable part in the mass balance as shown in the results section.

퐇퐮퐦퐢퐧퐬 퐜퐨퐧퐯퐞퐫퐬퐢퐨퐧

퐻푢푚푖푛푠 푖푛푖푡푖푎푙 − (푀푎푠푠 표푓 푠표푙푖푑 푎푓푡푒푟 푡ℎ푒 푟푒푎푐푡푖표푛 − 퐶푎푡푎푙푦푠푡) = 퐻푢푚푖푛푠 푖푛푖푡푖푎푙

× 100 %

푶풊풍 풓풆풔풊풅풖풆 풂풇풕풆풓 풆풗풂풑풐풓풂풕풊풏품 풔풐풍풗풆풏풕 퐓퐨퐭퐚퐥 퐨퐢퐥 퐲퐢퐞퐥퐝 = ( ) × ퟏퟎퟎ% 푯풖풎풊풏풔 풊풏풊풕풊풂풍

93 GC detectable oil yield

푇표푡푎푙 푚푎푠푠 표푓 푝푟표푑푢푐푡푠 푑푒푡푒푐푡푎푏푙푒 푏푦 퐺퐶 푖푛 표푖푙 푟푒푠푖푑푢푒 = × 100% 퐻푢푚푖푛푠 푖푛푖푡푖푎푙

퐌퐚퐬퐬 퐛퐚퐥퐚퐧퐜퐞 퐜퐥퐨퐬퐮퐫퐞

푆푢푚 표푓 푚푎푠푠 표푓 푠표푙푖푑푠, 푙푖푞푢푖푑 푎푛푑 푔푎푠푒푎푠표푢푠 푝푟표푑푢푐푡푠 푎푓푡푒푟 푡ℎ푒 푟푒푎푐푡푖표푛 = 퐻푢푚푖푛푠 푖푛푖푡푖푎푙 + 푐푎푡푎푙푦푠푡 + 푠표푙푣푒푛푡 푖푛푖푡푖푎푙 + 푔푎푠 푖푛푖푡푖푎푙

× 100%

4.2.4 Analysis of Reaction Products

Gas phase. To ensure reproducibility, two 5 µL gas phase samples were injected into an Agilent 7890A GC equipped with a flame ionization detector (FID) and a HP-

PlotQ column (30 m × 0.53 mm  40 µm). Compounds were identified and quantified based on the retention times and calibration plots of known standards. To determine the other components in the gas phase such as hydrogen (H2), carbon monoxide (CO) and carbon dioxide (CO2), the gas bag was connected to an Agilent 490 MicroGC equipped with a thermal conductivity detector (TCD). Two MS5A columns (10 m with 0.32 µm film thickness), one PoraplotU column (10 m × 0.32 µm  0.20 µm) and one Al2O3/KCl column (10 m × 0.32 µm) were used. H2, CO and CH4 were quantified using the first two columns, N2 was quantified using the second column, and CO2 was quantified using the third column. The peaks were identified based on the retention times of known gas standards. The quantification was done using the calibration plots of the standard gases.

94 Liquid phase. GC-Mass Spectrometry (GC-MS) analysis was done using an

Agilent 7890B GC equipped with a DB-5 column (30 m × 0.25 mm  0.25 μm) connected to an Agilent 5977A mass spectrometer. The injector temperature was kept at

250 °C. The oven temperature was kept at 40 °C for 2 min, and brought to 315 °C at a heating rate of 10 °C/min, and held at 315 °C for 2 min. The NIST14 library was used for identification of peaks. C7-C10 linear alkanes, C4-C7 methyl esters, naphthalene, o- cresol, 2-hexanone, 2-hexanol, and 2,5-dimethylfuran were used as representative external standards for quantification. The alkane-type products were quantified using the calibration plots of n-alkanes with the same number of carbons. The response factors of aromatic, phenolic, ester, alcohol, ketone and furanic compounds were assumed to be equal to the response factor of the representative compound with the same functional group.

Gel permeation chromatography (GPC) analysis was conducted using a Waters

2695 HPLC equipped with a RID (Waters model 2414) and two Waters Styragel columns (dimensions: 4.6 mm  300 mm  5 µm) connected in series (Models: HR 3 and HR 4). Tetrahydrofuran (THF) was used as the mobile phase at a flow rate of 0.3 mL/min. The column compartment and the RID temperature were set at 25 °C. To prepare the samples for analysis, 0.1 mL of the humins oil solution was placed in a vacuum oven to remove the methanol solvent using the conditions written above. This dry humins oil was re-dissolved in 0.5 mL of THF and 20 μL of this solution was injected. Five polystyrene (PS) standards with peak molecular weights (Mp) of 370-

95 17600 Da were used for calibration. The average molecular weights of samples were calculated based on the retention times of the standards.

Elemental analysis of the humins oil was conducted on an Elementar Vario EL

Cube CHNS analyzer using 5 ± 0.5 mg sample. Two sulfanilamide samples were used as standards to calibrate the instrument. Since humins oil samples are not expected to contain N or S, the oxygen content was calculated to be the balance of the C and H contents.

Water content in the liquid phase obtained from all hydrotreatment reactions was determined using a Mettler Toledo V20 Karl-Fischer volumetric titrator.

Composite-5 in anhydrous methanol (Hydranal) solvent was used as the titrant.

Approximately 0.1 mL of sample was weighed and injected into the mixing chamber pre-filled with Hydranal.

4.3 Results and Discussion

4.3.1 Elemental Composition of Humins-Oil

The elemental composition of humins oil, obtained by vacuum evaporation of methanol solvent at 20 °C (experiment 9 in Table C1), is shown in the van Krevelen diagram in Figure 4.1 and compared with the starting humins which contain 60.4 wt%

C, 4.2 wt% H and 35.3 wt% O. Compared to the composition of the fructose-derived humins reported in the literature (67.1 wt% C, 4.5 wt% H and 28.4 wt% O),24 our humins have a slightly lower C/O ratio possibly due to the differences in the reaction

96 temperature used to produce humins, catalyst used, and the fructose conversion.

Compared to the starting humins, the oil sample shows a significant decrease in O/C ratio, indicating deoxygenation during hydrotreatment of humins. The recovered solids after reaction were not analyzed due to difficulty in separating out the spent catalyst.

1.0 Fructose

0.8

0.6

Humins O/C 0.4

0.2 Oil 0.0 0.4 0.8 1.2 1.6 2.0 H/C

Figure 4.1 van Krevelen diagram of the humins-oil and the starting humins. Fructose is used as a reference.

4.3.2 Catalyst Screening

Ru, Pt, Pd and Rh on carbon were screened for humins hydrotreatment at

350 °C and 30 bar initial H2 pressure for 3 h. Some of these catalysts have been used for the hydrogenolysis of lignin4, 153 and humins.38, 39 Figure C1 shows representative GC-

MS chromatograms of the oil using 5 wt% Ru/C. Various aromatic and aliphatic hydrocarbons, phenolics, esters, ketones, alcohols, and a small amount of furans were identified; representative compounds are given in Table 4.1. The same types of compounds formed using other metal catalysts. Although humins have polymeric

97 structure containing a high fraction of furanic backbone,29, 57 low molecular weight furans are absent in the detected products. Aromatics compounds, including hydrocarbons and phenols, are present instead. The detected products agree well with those reported in the literature in catalytic gasification36 and hydrotreatment38-40 of humins. A complete list of all identified compounds is given in Table C4. To confirm the origin of these products, blank reactions without humins were conducted under identical conditions using the Rh/C catalyst (Table C5). Approximately 30 wt% conversion of methanol to H2, CO, CO2 and C1-C6 hydrocarbons as gaseous products and water as liquid product (Table C5) was observed. The compounds observed in humins oil are absent in the GC-MS chromatogram of the blank experiments (Figures

C2a and C2b). This indicates that the aforementioned organic compounds in the oil are all derived from humins. Water was also detected in the liquid phase in the blank experiment. Methanol conversion was estimated to be between 30 and 70% depending on reaction temperature (Table C5).

Table 4.1 Major classes of products and detected compounds in humins oil (obtained from GC-MS shown in Figure C1).

98 Compounds Structures of compounds detected

Aromatics

(single and

polycylic)

Phenolics

Esters

Ketones

Alcohols

Alkanes

(Linear and cyclic)

Furans

99

Figure 4.2 compares humins conversions and the yields of oil using different metal catalysts. Ru/C gives the lowest conversion while Pt/C achieves the highest conversion. The total humins-oil yield with Ru/C is much lower than that of the other catalysts. High amount of gaseous products form over the Ru/C catalyst consistent with a prior report.40 Pt/C gives the highest yield of total humins oil, however the GC- detectable oil fraction is higher over the Rh/C catalyst. The Heeres group40 also reported that Pt/C is the most effective catalyst in terms of humins conversion (77%) and GC- detectable fraction of oil yield (9.6 wt%). However, their GC-detectable oil yield from the Rh/C catalyst was only 2.4 wt%. The difference between our GC-detectable oil yield using the Rh/C catalyst and that of Heeres group could be attributed to differences in the conditions used to generate the starting humins and in the solvents used.

The effect of different metals on the performance of hydrodeoxygenation of biomass-derived compounds has been documented to an extent and focused mainly on

C-O bond scission. Resasco and coworkers154 compared the hydrogenation of furfural over silica supported Cu, Pd and Ni catalysts. It was reported that Cu/SiO2 catalyst promotes high selectivity to furfuryl alcohol while Pd/SiO2 mainly converts furfural to furan by decarbonylation. On Ni/SiO2 catalyst, ring opening products (butanal, butanol, and butane) were obtained in significant fraction. The different product distributions were attributed to the strength of interaction of the furan ring with the metal surface and the type of surface intermediates stabilized on each metal. In a similar study155 on

100 phenolic and aromatic ether model compounds representative of lignin, Ru, Pt, and Pd catalysts on carbon were investigated. Pt led to the highest conversion. Time-dependent studies showed that Pd had the lowest hydrogenation activity converting a ketonic intermediate to alcohol products, while Ru and Pt could rapidly convert the ketone to alcohol and eventually to alkane products.155 Metal catalyst performance is typically correlated with the affinity of O for metal sites, i.e., the more oxophilic the metal is, the easier the deoxygenation is. This follows from mechanistic studies of correlations of reaction barriers with thermochemistry descriptors.156-158 In hydrodeoxygenation of furans, it was shown that oxophilic metals like Ru are partially oxidized under working conditions, and a redox, vacancy-based, mechanism is at work. 151, 159, 160

In Figure C3, the yields of each type of products are given. The product distributions were similar over the Pt and Pd catalysts. The same two catalysts were more selective towards phenolics and Rh was more selective towards ester formation.

The Ru/C catalyst, on the other hand, was more selective towards aromatic hydrocarbons. However, the structural complexity of humins resulted in multiple pathways such as furans to aromatics, ring opening of furans, and possibly secondary transformations. Low molecular weight model compounds representative of basic humin units would be necessary to elucidate the correlation between the catalyst properties to the individual reaction pathways.

The O/C molar ratios of the humins oil product by four metal/C catalysts follow a decreasing order of Pt, Ru, Rh and Pd (Table C2) while the H/C ratios are similar except for the Ru/C catalyst, which yields oil with lower H/C ratio. To understand this

101 noticeable change of O/C ratio of the humins oil composition, we compared the oxophilicity of these metals, as determined using density functional theory (DFT).161

The oxophilicity, which showed a decreasing order of Ru, Rh, Pt and Pd, does not follow the trend of the O/C ratios of humin oil. This underscores the structural and chemical complexity of humins as multiple reaction pathways could coexist during the hydrotreatment process40. Correlation of the catalyst properties with humins conversion and products selectivity is challenging.

60 80 50

40 60

30 40 20

20 (%) Conversion

10 Humins Oil Yield (wt%) Yield Oil Humins 0 0 Ru/C Pt/C Pd/C Rh/C Catalyst

■ GC-detectable oil yield ■ Total oil yield ■Conversion

Figure 4.2 Comparison of humins conversion and yield of total and GC-detectable humins oil over the Ru/C, Pd/C, Pt/C and Rh/C catalysts. Reaction conditions: 350 °C, 3 h, 30 bar of initial H2, and a catalyst to humins mass ratio of 1: 10. The dashed lines connecting the black points (conversion) guide the eye.

102 4.3.3 Effect of Temperature, Time, Pressure and Catalyst to Humins Ratio

The Rh/C catalyzed reaction achieved the best yield to GC-detectable oil and to aromatic hydrocarbons and esters (Figure C3). Thus, Rh/C was used to optimize the reaction time, temperature, H2 pressure and Rh/C to humins mass ratio in the range of

0-6 h, 350 - 425 °C, 10 - 45 bar, and 0 - 0.2 (or Rh to humins mass ratio = 0-0.01), respectively (Table C1). The humins-conversion and the total yield of oil as well as of the GC-detectable fraction are compared in Figure 4.3. The conversion increases with an increase in the reaction temperature (Figure 4.3a) while the total oil yield decreases, indicating that humins gasification increases with increasing temperature. The gaseous products from humins hydrotreatment were not determined because of formation of similar gases from the solvent (methanol). The GC-detectable oil yield increases first from 350 to 400 °C and then decreases from 400 to 425 °C. This suggest that higher temperatures are beneficial for the depolymerization of humins and HDO of the depolymerized species, but sufficiently high temperatures cause gasification and decrease the oil yield. The reaction time has a similar effect on the total and GC- detectable oil yields. At 400 °C, the reaction proceeds quickly as the GC-detectable oil yield doubles in the first 1.5 h. Further increase in the reaction time from 1.5 h to 3 h has no significant effect on the yield. However, prolonged time (>3 h) results in a slight decrease of both yields (total and GC-detectable fraction) at comparable conversion, suggesting gasification of the oil with increasing reaction time.

Increased H2 pressure has a positive effect on the conversion and oil yields

(Figure 4.3c). Likewise, a higher catalyst amount results in more HDO in terms of

103 humins conversion and total oil yield (Figure 4.3d). A blank experiment without Rh/C shows some thermal degradation of humins and formation of oil. The conversion increases from 40 to 73 wt% upon increasing the catalyst to humins mass ratio to 1:10 while the total oil yield increases almost three-fold. However, much higher catalyst to humins ratio (1:5) is detrimental to the yield likely because of higher methanol reforming.

50 100 50 100 (a) (b) 40 80 40 80

30 60 30 60

20 40 20 40 Conversion (%) Conversion

10 20 (%) Conversion 10 20

Humins Oil Yield (wt%) Yield Oil Humins Humins Oil Yield (wt%) Yield Oil Humins 0 0 0 0 350 375 400 425 0 1.5 3 6 o Temperature ( C) Time (h)

50 100 50 100 (c) (d) 40 80 40 80

30 60 30 60

20 40 20 40 Conversion (%) Conversion

10 20 (%) Conversion 10 20

Humins Oil Yield (wt%) Yield Oil Humins Humins Oil Yield (wt%) Yield Oil Humins 0 0 0 0 10 20 30 45 No Catalyst 0.025 0.1 0.2 H Pressure (bar) 2 Catalyst/Humins (w/w)

■ GC-detectable oil yield ■ Total oil yield ■Conversion

Figure 4.3 Humins conversion and yields of total and GC-detectable oil at different (a) temperatures, (b) reaction times, (c) hydrogen pressure, and (d) Rh/C catalyst to humins mass ratio. The temperature dependent reactions were carried out at 30 bar H2 and 1:10 catalyst to humins ratio for 3 h. The time dependent reactions were conducted at 400

104 C, 30 bar H2, and 1:10 catalyst to humins ratio. The pressure dependent reactions were run at 400 °C and 1:10 catalyst to humins ratio for 3 h. The catalyst dependent reactions were conducted at 350 °C, 30 bar H2, and 3 h. The dashed lines connecting the black points (conversion) are used to guide the eye.

The distribution of identified products was profoundly affected by the reaction conditions (Figure 4.4). Higher temperatures result in the formation of higher percent of aromatic hydrocarbons and phenols. Our observation of forming aromatic products at elevated temperatures is consistent with earlier literature of pyrolysis of humins,24 glucose and fructose.162 Likewise, slightly higher amounts of aromatic hydrocarbons and phenols form as the reaction time increases from 0 to 1.5 h. Prolonged times show little to no change in the product distribution. Higher H2 pressure shifts the product distribution to non-aromatic compounds such as esters, ketones and alkanes. Addition of catalyst first decreases the aromatic hydrocarbons and phenols and eventually increases the percent of aromatic hydrocarbons, phenolics and esters.

105 (a) (b) 100 100

80 80

60 60

40 40

20 20 Product Distribution (wt%) Distribution Product Product Distribution (wt%) Distribution Product 0 0 350 375 400 425 0 1.5 3 6 Temperature (oC) Time (h)

(c) (d) 100 100

80 80

60 60

40 40

20 20 Product Distribution (wt%) Distribution Product

0 (wt%) Distribution Product 0 10 20 30 45 No Catalyst 0.025 0.1 0.2 H Pressure (bar) Catalyst/Humins (w/w) 2

■ Aromatics ■ Phenols ■ Esters ■ Ketones ■ Alcohols ■ Alkanes ■ Furans

Figure 4.4 Distribution of GC-detectable products at different (a) temperatures, (b) reaction times, (c) hydrogen pressure, and (d) Rh/C catalyst to humins ratio. The temperature dependent reactions were carried out at 30 bar H2, 3 h and 1:10 Rh/C catalyst to humins ratio. The time dependent reactions were run at 400 C, 30 bar H2, and 1:10 catalyst to humins ratio. 0 min corresponds to after heating up to the set point temperature. The pressure dependent reactions were conducted at 400 °C, 3 h and 1:10 catalyst to humins ratio. The catalyst dependent reactions were run at 350 °C, 30 bar H2, and 3 h.

Next, we demonstrate the effect of reaction conditions on the molecular size of products in the oil using GPC. The polydispersity index (PDI) values of all humin oils are similar and range between 1.3 and 1.7 (Table A6). These well-above 1 values

106 indicate broad distributions of molecular weight species. Compared to the humins solubilized species at room temperature prior to the start of the reaction,152 the average molecular weights of the compounds in the oil is greatly reduced (Figure C4). The number average (Mn) and weight average (Mw) molecular weights as a function of temperature, reaction time and Rh/C to humins ratio are given in Figure 4.5. As the temperature and reaction time increase, both Mn and total humins oil yield decrease with a concomitant increase in humins conversion, strengthening our hypothesis that the oil gasifies at high temperatures (Figures 4.3a and 4.5a) and long reaction times

(Figures 4.3b and 4.5b). Figures 4.4d and 4.3d suggest that humins can be depolymerized and hydrodeoxygenated thermally but Rh/C accelerates the reaction and improves the total and GC-detectable oil yields. The Mn and Mw values for the four oils obtained at four H2 pressures are very similar (Figure 4.5c). This indicates that the relative amounts of compounds in the oils are similar although higher H2 pressure results in higher yield of total oil (Figure 4.3c).

107 (a) (b) 800 800

600 600 M Mw w

400 400

200 Mn 200 Mn Average Molecular Average Weight

0 Weight Molecular Average 0 350 375 400 425 0 1.5 3 6 Temperature (oC) Time (h)

(c) (d) 800 800

M 600 w 600

400 400

200 200 Mn

0 Average Molecular Weight Molecular Average 0 Weight Molecular Average No 0.025 0.1 0.2 10 20 30 45 Catalyst H pressure (bar) Catalyst/Humins (w/w) 2

Figure 4.5 Mn and Mw of humin oils obtained at different (a) reaction times, (b) temperatures, (c) hydrogen pressure and (d) Rh/C catalyst to humins ratios. The temperature dependent reactions were run at 30 bar H2, 3 h and 1:10 catalyst to humins ratio. The time dependent reactions were run at 400 C, 30 bar H2, and 1:10 catalyst to humins ratio. The pressure dependent reactions were run at 400 °C, 3 h and 1:10 catalyst to humins ratio. The catalyst dependent reactions were run at 350 °C, 30 bar H2, and 3 h. The dashed lines connecting the points are used to guide the eye.

Humins HDO in IPA solvent was conducted by the Heeres group.38-40 To compare the performance of the Rh/C catalyst in IPA, we conducted two experiments under comparable conditions as in methanol at 400 °C for 3 h using a 1:10 Rh/C to humins mass ratio. Initially, the reactor was not pressurized with H2 to be consistent

38 with the literature experiments. Instead, 30 bar of N2 was used. The IPA-mediated

108 reaction formed several new compounds in the liquid phase, primarily from the solvent as similar compounds formed in a blank experiment using only Rh/C and IPA (Table

C7). Table C8 lists IPA-derived and humins-derived products. Prior studies reported formation of the same classes of products (alkanes, aromatic hydrocarbons, phenols and a very small amount of furans, alcohols and ketones).39, 40 Unlike methanol, IPA mediated reactions form no esters. The conversion, total and GC-detectable oil yields in methanol and IPA are summarized in Figure C5. Humins conversion in both solvents are similar and the total yield of oil is higher in methanol. We found higher humins conversion (70 wt%) in IPA than the reported value (60 wt%) in the same solvent, possibly due to differences in starting humins. The yield of GC-detectable oil is similar in both solvents after accounting for the IPA-derived liquid products; however, the yield of only humins-derived GC-detectable oil (based on detected compounds) is less in IPA than in methanol. As an added benefit, methanol is cheaper than IPA used in Heeres’ work and thus, the overall process economics may be improved. The total GC- detectable humins-oil yield in IPA is 2.1 wt%, in good agreement with the value of 2.4 wt % reported by Wang et al. 40 In IPA, phenolic compounds, mainly isopropyl- substituted, are a larger fraction in the detected aromatic compounds. The results in the two solvents compared in Tables C3, C6, and Figure C5 suggest that solvents play a role during the catalytic hydrotreatment of humins, reminiscent of a prior report on the catalytic depolymerization of lignin.23

109 4.3.4. Effect of Solvent through 13C Isotopic Labeling Study

To investigate the possibility of coupling reactions between solvents and humins or humins-derived intermediates, reactions were conducted in 13C labeled methanol.

Identified products from the isotopic labeling experiment were compared with those obtained in unlabeled methanol. Methyl valerate, m-xylene, 2,6-dimethylphenol, and 4- octanone were chosen as representative compounds for comparison because these are relatively abundant in the oil. The changes in the relative abundance of the fragments for these representative compounds are shown in Figure 4.6.

110 (a) (b) 10000 10000

8000 8000

6000 6000

4000 4000 Counts (a.u.) Counts 2000 (a.u.) Counts 2000 0 0 Unlabeled Labeled Unlabeled Labeled m/z=74 m/z=75 m/z=85 m/z=86 106 107 108 91 92 77 78

(d) (c) 8000 10000

8000 6000 6000 4000

4000 Counts

2000 (a.u.) Counts 2000

0 0 Unlabeled Labeled Unlabeled Labeled

m/z=128 m/z=129 m/z=57 m/z=58 m/z=85 m/z=86 m/z=122 m/z=123 m/z=124 m/z=77 m/z=78

Figure 4.6 Mass spectrum peaks of representative compounds in 13C isotopic labeling and unlabeled experiments. (a) Methyl valerate; (b) m-xylene; (c) 2,6-dimethylphenol; (d) 4-Octanone.

For methyl valerate (Figure 4.6a), the intensity of the molecular ion, m/z=116, was very weak. Thus, the relative abundances of the major fragments (A) of it and their

(A+1) counterparts were used for comparison. In the NIST library mass spectrum for

+ methyl valerate, m/z=74, which corresponds to the [CH2COOCH3] ion, was the most abundant fragment while the intensity of the m/z=75 fragment was very low. However,

13 in the labeled experiment using the CH3OH solvent, m/z=75 became the most abundant fragment while m/z=74 almost disappeared. This indicates that the methanol

111 solvent could take part in the esterification of humins-derived fragments. At the same time, the ratio of fragments corresponding to m/z 85 and 86, which represent the methyl valerate with the -OCH3 group lost, are nearly the same. This strengthens our

13 proposition that the C from the solvent is incorporated into only in the –OCH3 group but not in the aliphatic chain. A similar observation is evidenced for other esters, such as methyl propionate and methyl hexanoate. In prior reports, Huang et al.22 and Ma et al.163 found alkylation and esterification of lignin-derived intermediates with methanol and ethanol. In their work, oxidative esterification was reported despite the reaction taking place in a hydrogen-rich atmosphere. Huang et al. proposed that esters could form from the reaction between the alcohol solvent and the aliphatic -OH groups in lignin or their fragments.163 Ma et al.’s suggestion that the ethanol solvent undergoes oxidative esterification with alcohol products generated from lignin163 may also happen in our reactions. Interestingly, when IPA was used as the solvent, ester products were not observed either in the present work or in prior works.38-40

Next, we investigate the formation of aromatic hydrocarbon and phenolic products. The intensity of the molecular ion (M) for m-xylene, m/z=106, is high (Figure

13 4.6b). The (M+1)/M ratio significantly increases in CH3OH due to formation of the m/z=107 ion. The peak intensity for m/z=108 ion, which was absent in the spectrum in unlabeled methanol, increases in the labeled experiment. Since the intensity of the m/z=106 ion is still dominating over the m/z=107 or 108 ions, these changes together indicate that one or two 13C atoms are incorporated into some, but not all of the m- xylene molecules. The intensity ratio of the m/z=78 (for benzene) and m/z=77 peaks are

112 similar in the unlabeled and labeled experiments, thus ruling out 13C incorporation into the benzene ring. The intensity ratio of the m/z=92 (benzylic ion) and 91 ions increases

13 13 13 sharply in the C labeled methanol, indicating C from CH3OH is incorporated in the methyl group of m-xylene. However, not all the methyl groups in m-xylene are derived from methanol as the m/z=106 peak is enhanced over the m/z=107 and 108 peaks.

Similar comparisons for tri- and tetra-methylated benzene products confirm the partial ring-methylation by methanol.

The alkylation of benzene ring by the methanol solvent was also found to occur in the multi-methylated phenolic products in the labeled experiment. For example, the

(M+1)/M and (M+2)/M intensity ratios for 2,6-dimethylphenol (M=122) ions increase significantly in the isotopic labeling experiment (Figure 4.6c). Since the intensity of the m/z=122 ion is still dominated over the m/z=123 and 124 ions, these changes together indicate that one or two 13C are incorporated into some, but not all the phenolic molecules.

In addition to the methyl group in the aromatic products originating from solvents (methanol or labeled methanol), it is likely that some methyl groups also originate from the C-C cleaved aliphatic linkers of humins at high temperatures, which couple with reactive humins intermediates that form during depolymerization and HDO.

The latter pathway is consistent with the reported catalytic liquefaction of bituminous coal in which C-C bonds cleavage to free-radicals species and self-coupling of the resulting radicals or coupling with hydrogen radicals occur.164 The radical formation and coupling pathway likely follows the C-alkylation mechanism observed in the

113 catalytic liquefaction of lignin in methanol or ethanol. 22 Although isotopic labeling experiments were not conducted in lignin liquefaction,22 the authors performed 2D-

NMR analysis of the reaction mixture and proposed that the alkylation of the ring carbons by the solvents prevented the reactive lignin intermediates from repolymerizing into insoluble products. Given that similar reactive intermediates could form and repolymerize during the hydrotreatment of humins, it is likely that the methanol solvent plays a crucial role in our reaction.

Lastly, we analyze the ketonic products in the labeled and unlabeled experiments (Figure 4.6d). For 4-octanone, the intensity of the molecular ion, m/z=128, is much weaker compared to the molecular ions of the aforementioned phenolic and aromatic products. Thus, the relative abundances of the major fragments (A) and their

(A+1) counterparts were used for comparison. As shown in Figure 4.6d, the intensity ratios of the ions of m/z = 57 and 58 peaks (A/A+1) and that of m/z =85 and 86 peaks

(M/M+1) are similar in both labelled and unlabeled reactions. Therefore, it is unlikely

13 13 that C from CH3OH solvent is incorporated in the 4-octanone molecule. A similar observation for other ketones, such as 2-hexanone, suggests that ketones are solely formed from humins. The ketones might form via pyrolysis of humins as observed by

Rasrendra et al. 24 and Hoang et al.36

4.4 Conclusions

We have studied the catalytic hydrotreatment of fructose-derived humins using noble metal catalysts supported on carbon in methanol and IPA. The major products in

114 the resulted oil are aromatic hydrocarbons, phenols, esters, ketones and alcohols, which are valuable bulk or specialty bio-based chemicals. An increase in H/C ratio and a decrease in O/C ratio in the products is observed upon hydrotreatment. Rh/C gives the highest GC-detectable oil yield. The humins conversion and the yield of oil depend highly on reaction temperature, catalyst to humins ratio and reaction time. Gasification of oil occurs at high temperature (e.g., 425 C) and long reaction times, resulting in loss of oil yield. The product distribution shifts toward aromatics and phenols at high temperatures and long reaction times while esters dominate at short reaction times and high catalyst to humins ratios. External hydrogen improves the yield of total oil as well as of the GC-detectable compounds. Among the GC-detectable products, up to 70 wt% aromatic hydrocarbons and phenols could be achieved. Future improvement of this process could include developing methods to separate the catalyst from the remaining solids. Magnetic catalyst particles or oxide-supported metal catalysts for easier regeneration as well as engineering approaches, such as pumping a humins methanol solution through a tubular reactor with packed catalyst bed165, or using a stirred reactor with a catalyst basket of appropriate mesh size166, 167, could be exploited. A thorough techno-economic analysis should also be done to evaluate the feasibility of this process.

Isotopic labeling experiments provide evidence for the first time of coupling reactions between methanol and humins or its fractionated intermediates and incorporation of labeled carbon in the aromatic and ester products. Kinetics studies using model compounds and correlation of the solvent properties with the catalytic

115 performance would be valuable future directions to further elucidate the reaction pathways of humins hydrotreatment.

116 Chapter 5

GLUCOSE CONVERSION TO 5-HYDROXYMETHYL FURFURAL AT SHORT RESIDENCE TIMES IN A MICROREACTOR

5.1 Introduction

Due to the growing concern about global warming, replacing fossil fuels with molecules derived from renewable lignocellulosic biomass has received attention. As such, converting biomass-derived carbohydrates to platform molecules, including 5- hydroxymethyl furfural (HMF) has been studied extensively. 15, 78, 111, 148A wide variety of catalysts, including both homogeneous and heterogeneous Brønsted and Lewis acid catalysts,42, 47, 49, 74 solvents and solvent mixtures,41, 85, 86, 89, 138, 168 phase modifiers,91 reaction conditions as well as process designs65, 96 have been investigated to optimize the yield of HMF. Recently, the Department of Energy (DOE) has established the Rapid

Advancement of Process Intensification Development (RAPID) institute, through the

American Institute of Chemical Engineers (AIChE),to propel state-of the art research in process intensification. One approach to intensified processes entails using more compact, smaller and therefore more energy-efficient reaction vessels. Micro and milli channel flow reactors are promising in this regard because their narrow channels, between 0.1-3 mm internal diameter96, 169 enable fast heat and mass transfer, which leads to improved productivity and selectivity.94, 95, 170 An advantage of the micro reactors is the ability to study reaction kinetics under differential conversions where

117 short residence times and precise process control that render collection of large amounts of data feasible.171 In contrast, batch reactors heat up slowly and often high conversions are observed when the set point temperature is reached, rendering kinetic studies less accurate.

The application of microreactors to biomass processing is receiving growing attention. In particular, a number of studies 51, 94, 96-99, 107 have focused on the conversion of sugars to HMF using flow reactors. However, most of them have used fructose as the substrate. Specifically, Tierce et al.94 optimized the HMF yield with respect to reaction temperature at a fixed residence time of 1 min and obtained up to

53% HMF yield with improved selectivity compared to batch reactors. They attributed the enhanced selectivity to a steep heating profile and improved mass transfer and mixing characteristics resulting from passive mixers embedded in the microchannel. 108

Muranaka et al.97 used a biphasic solvent mixture of aqueous, acidic phosphate buffered saline (PBS) at pH of 2 and 2-sec-butanol. Up to 81% HMF yield from fructose and

76% HMF yield from glucose was achieved at 180 °C in 12 and 47 min, respectively.

Hansen et al.107 obtained up to 53% HMF yield at 95% fructose conversion using HCl catalyst at 200 oC and a reaction time of 1 min under microwave heating. Using glucose could significantly lower the cost, although achieving high HMF yields in short reaction times remains challenging. The isomerization of glucose to fructose is a reversible, slightly endothermic reaction limited by thermodynamic equilibrium49 with an equilibrium constant of ~1 at 298 K. To make the HMF production more economically viable requires efforts on optimizing catalyst process conditions.

118 Recently, our group built a continuous flow microchannel reactor for ultrafast processing of fructose in water solvent.108 The flow patterns were quantified using laser-induced fluorescence (LIF) and particle image velocimetry (PIV). The mixing times were estimated to range from 0.03 to 4.8 s for residence times of 1 to 120 s, respectively. Moreover, a curved channel geometry induces secondary (Dean) vortices that produce a three-fold increase in mixing. The highest yield of 54% with full fructose conversion was achieved in only 4 s at 200 °C. Since all the fructose was consumed, the need for recycle streams were eliminated. The optimal HMF yield achieved was similar to those obtained by Tierce et al. 94and Hansen et al.,107 but the residence time was much shorter (4 s compared to 1 min). The highest ever HMF productivity (HMF yield per unit time) was achieved. A model was also presented to evaluate the energy savings as a result of optimizing the reactor length; at 200 °C, 58% energy savings could be achieved.

In this work, we improve HMF productivity from glucose by selecting catalyst and optimizing process conditions using the microchannel flow reactor design of Desir et al.108 and benchmarking the performance with batch reactors. We also aim at a deeper understanding of the catalytic pathway by acquiring more accurate kinetics data. We

42, 59 choose CrCl3 as the homogeneous isomerization catalyst in aqueous phase.

Choudhary et al.58 have demonstrated through Extended X-ray Absorption Fine

2+ Structure (EXAFS) studies that the [Cr(H2O)5(OH)] formed during the hydrolysis of

CrCl3 is the active species for glucose isomerization. Furthermore, the HMF yield and

16, 58 selectivity could be improved by adding HCl to CrCl3 at an optimal ratio. The effect

119 of preheating on the speciation and the reactivity of the CrCl3 is also investigated for the first time, using UV-Visible Spectroscopy (UV-Vis) and reactivity studies.

5.2 Methods

5.2.1 Materials

Glucose (≥99% purity, Sigma Aldrich), fructose (≥99% purity, Sigma Aldrich), mannose (≥99% purity, Sigma Aldrich), chromium chloride hexahydrate (≥99% purity,

Sigma Aldrich), hydrochloric acid (37 wt%, Fisher Scientific), sulfuric acid (5M,

Fluka), were used without further purification. All aqueous solutions were prepared using deionized (DI) water obtained using a Millipore water purification system (model:

Direct-Q3 UV R).

5.2.2 Microreactor Setup and Experiments

A microreactor was built in our laboratory as described by Desir et al.108 Figure

D1 shows that the reactant can be heated from room temperature to the set point within

30 s. Due the much shorter heating time comparing to the reaction time, no preheating was used for experiments with residence times longer than or equal to five min. For this reason, a mixture of sugars and catalyst was prepared in one feed reservoir. A Teledyne

SSI MX reciprocating pump with Poly Ether Ether Ketone (PEEK) wetted internal parts was used to deliver the feed into a 1.5 m PEEK tube with 0.02 cm2 cross-sectional area placed in an oven with temperature control (accuracy: ±1°C). The reaction mixture exits

120 the tube and immediately enters a coiled, thin PEEK tube with 0.002 cm2 cross- sectional area placed in an ice-water bath to rapidly quench the chemistry. A stainless- steel 316 pressure gauge in the range of 0-1000 psi was installed after the quenching section to indicate the system pressure. A PEEK back-pressure regulator (BPR) was connected further downstream to pressurize the system and prevent vaporization. The eluents from the BPR were filtered and collected for High Performance Liquid

Chromatography (HPLC) analysis.

Scheme 5.1 Schematic overview of the microchannel flow reactor setup. Reprinted with permission from Desir et al. 108 Copyright (2019) Royal society of chemistry. In this work, glucose was used instead of fructose in all reactions. For the experiments with residence times less than five min, two pumps were used. As shown in Scheme 5.1, one pump (Teledyne SSI-MX) is connected to a reservoir containing a 2 wt% aqueous glucose feed. The other pump (Teledyne SSI-LS)

121 is connected to the catalyst feed where the catalyst concentration is twice that used for reaction. The catalyst feed enters a preheating coil of 1.5 m in length and total volume of 3 mL. The sugars feed is preheated for 30 s in a PEEK tube of 0.02 cm2 cross sectional area and variable length to keep the preheating time constant. Control experiments showed that the conversion of 1 wt% glucose at 200 °C and 1 min of residence time without catalyst is minimal (less than 2 %). Therefore, reactions of glucose in the preheating section can be neglected. This preheated glucose stream is mixed with the preheated catalyst feed at a T-junction. The flow rate ratio of the two streams is 1:1 so that the combined stream contains 1wt% of glucose and the target catalyst concentration. The combined stream enters a coiled PEEK tube of 1 m length and 0.2 mL in total volume.

5.2.3 Batch Reactor Experiments

Thick walled glass vial reactors (Sigma-Aldrich) 5 mL in volume were used for batch reactions. Following the procedure of Swift et al.,42 2 mL of reacting mixture and a stir bar were placed in each vial. Then each vial was sealed with a crimp cap. To prevent water vapor leakage, a stainless-steel shim was inserted between each aluminum cap and the rubber septum. Then the vials were immersed in an aluminum heating block with individual vial slots filled with mineral oil pre-heated to the set point temperature. A thermocouple inserted into a vial filled with mineral oil sitting in the bath was used for monitoring the actual temperature. The stirring rate was 500 rpm.

Time zero was defined at the time when the vial was put into the oil bath. At each

122 desired time point, a vial was taken out of the oil bath and immediately immersed in an ice-water bath to quench the reaction. After the vials were cooled, the cap was opened, and the reaction mixture was filtered for analysis using HPLC.

5.2.4 Product Analysis

A Waters 2695 HPLC equipped with a RID and an Aminex Biorad 87C column heated at 75 °C was used for determination of the glucose, mannose and fructose concentrations. The mobile phase was deionized (DI) water at 0.5mL/min flow rate.

The same samples were also run on a Waters 2695 HPLC equipped with a RID and an

Aminex Biorad 87H column heated at 50 °C for determination of the acids and HMF concentration. The mobile phase was 0.005 mM of aqueous H2SO4 solution at 0.5 mL/min flow rate. External calibration standards were used in both cases.

5.2.5 Characterization of Cr(III) Species

UV-Visible Spectroscopy (UV-Vis) analysis was conducted on a Cary 600 UV-

Visible Spectrometer. All scans were performed in the 200-800 nm range. DI water was used for 100% transmission baseline reference and zero absorbance calibrations. All samples were directly put into a 1 cm pathlength cuvette and inserted into the sample holder for analysis.

123 5.3 Results and Discussion

5.3.1 Comparison Between Batch and Flow Reactor Data

A reaction using 1 wt% glucose and 1.7 mM CrCl3 was conducted at 140 °C.

Samples were taken every 5 min. The sugars and HMF concentrations vs. time on stream are given in Figure D2. Steady state is established two residence times after the set point is reached; the exit species concentrations remain stable afterwards. To ensure reproducibility, two repeat experiments were conducted. The glucose conversion and yields of mannose, fructose, and HMF and the error bars are given in Figure D3. The yields of levulinic acid (LA) and formic acid (FA) are very low (<1%).

Figure 5.1 shows the glucose conversion and HMF yields as a function of contact time for the batch and flow reactors. The error bars indicate good reproducibility in both setups. The flow reactor outperforms the batch reactor in both glucose conversion and HMF yields (Figures 5.1a and 5.1b). When the HMF yields are plotted against glucose conversion (Figure 5.1c), the data from the two reactors overlaps. This overlap indicates the same reaction pathways, while the glucose decomposition and HMF production are faster in the flow reactor. This can be attributed to faster heat and mass transfer of our microreactor: first, the time it took to reach the temperature set point in the microreactor was only 30 s, while the glass vial reactor took more than 5 min, as shown in Figure D1. The initial reaction rate is therefore higher in the microreactor. Second, the high surface-to-volume ratio of the microreactor enables fast diffusive mixing. Diffusion, which is the predominant mixing mechanism compared to convection at low Reynolds numbers,172, 173 is dependent on the interfacial area

124 between species as well as the concentration gradient.173 Even in a homogeneous system, fast diffusive mixing is still important given that our flow reactor operates in the laminar regime. Improving the radial diffusive mixing can bring the performance of the reactor to that of an ideal plug flow reactor (PFR), where each thin cross-section of the reactor has a uniform composition. It has been observed that microreactors often lead to higher yields and selectivity than batch reactors at shorter residence times because of the fast diffusive mixing.108, 170

(a) 60 (b) 10 Flow Flow 50 Batch Batch 8 40 6 30 4

20 HMF Yield (%) Yield HMF

10 2 Glucose Conversion (%) Conversion Glucose 0 0 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Contact Time (min) Contact Time (min)

125 (c) 10 Flow Batch 8

6

4

2 HMF Yield (%) Yield HMF

0

10 20 30 40 50 60

Glucose Conversion (%)

Figure 5.1 (a) Glucose conversion and (b) HMF yield as a function of contact (residence) time. (c) HMF yield versus glucose conversion. Reaction conditions: 1 wt % glucose, 1.7 mM CrCl3 at 140 °C. The error bars were obtained from two repeat experiments.

5.3.2 Improved HMF Production Rate Using Tandem CrCl3/HCl Catalysts

Prior work for glucose conversion to HMF in flow reactors is sparse and exhibits low productivity. To further improve the productivity, we use tandem catalysts

HCl and CrCl3 that have been demonstrated previously to be effective for glucose dehydration in batch reactions 42, 58. Matsumura et al. 67 investigated the decomposition of glucose in water at subcritical pressure of 25 MPa and found that the reaction order decreased with increasing temperature from around unity at 175 °C to 0.7 at 400 °C.

The temperature 140-200 °C was chosen to shorten the reaction time as well as keeping the reaction in the first-order regime. First, we conducted fixed residence time experiments at different temperatures. As shown in Figure 5.2, at ten min of residence

126 time, as the temperature increases from 140 to 200 °C, both the HMF yield and selectivity increase significantly. The increase in HMF selectivity at higher temperatures could be explained by the HMF formation reactions having higher energy barriers than that of the HMF decomposition reactions. 42

100 Glucose Conversion HMF Yield HMF Selectivity 80

60

40 or Selectivity (%) Selectivity or

Conversion, Yield Yield Conversion, 20

0 140 160 180 200 Temperature (oC)

Figure 5.2 Glucose conversion and HMF yield and selectivity at a residence time of 10 min. Inlet composition is 1 wt% glucose and 1.7 mM CrCl3 catalyst. Error bars were obtained from two repeat experiments.

Modeling using the OLI software as well as literature experiments42 have

2+ demonstrated a volcano-like behavior of active species [Cr(H2O)5(OH)] concentration versus solution pH. To investigate optimal reaction conditions, we used a temperature of 200 °C and added HCl to vary the pH between 1 and 2. Given the one-parameter

(residence time, temperature, and pH at fixed CrCl3 concentration) at a time optimization, here, we first chose to vary the residence time. The glucose conversion

127 and product yields as a function of contact (residence) time at 200 °C, 1.7 mM CrCl3 and (HCl) a pH value of 1.25 are given in Figure 5.3. Here the HMF yield increases with increasing residence time. At 2 min, 30.4% yield at 66.2% conversion is achieved.

This corresponds to a 15.2% yield/min productivity and a 46.0% selectivity. Reactions with 2 minutes of residence time at other pH values were also conducted with the results summarized in Figure 5.4. It can be concluded that the HMF yield was the highest at pH of 1.25. Control experiments were also conducted with HCl only and catalyst-free conditions at 160 °C and the results are given in Figure D4. Consistent with prior work, it can be confirmed that CrCl3 and HCl combined give superior performance compared to using the catalysts individually.

80 Glucose Mannose Fructose FA 60 LA HMF

40

20 Conversion or Yield (%) Yield or Conversion 0 0.0 0.5 1.0 1.5 2.0 2.5 Contact Time (min)

Figure 5.3 Glucose conversion and products yield (mannose, fructose, FA, LA and HMF) at temperatures of 200 °C, at short residence times. Data was acquired in the microreactor using 1 wt% glucose, 56.2 mM of HCl (corresponding to pH=1.25) and 1.7 mM CrCl3 in aqueous phase.

128 80 Glucose Mannose Fructose FA 60 LA HMF

40

20 Conversion or Yield (%) Yield or Conversion

0 pH=1 pH=1.25 pH=1.77 pH=2 No HCl

Figure 5.4 Glucose conversion and product yields at 200 °C at different pH values. Data was acquired in the microreactor using 1 wt% glucose, 1.7 mM CrCl3, 2 min residence time, and different concentrations of HCl in aqueous phase.

Here, we report the HMF productivity as a result of combined optimization in

Figure 5.5, and compare with literature studies with glucose as substrate for HMF production in aqueous phase at temperatures equal to or above 180 °C. Studies using glucose as substrate are relatively sparse than the ones with fructose as substrate, and very few used flow reactors. In Figure 5.5, the works of Zhang et al.,104Watanabe et al.,105 Siqueira et al.,106 and Lv et al.,103were done in batch reactors, and the work of

Muranaka et al.,97 was done using a flow reactor. At 200 °C in 2 min, with 1.7 mM

CrCl3 and 56.2 mM HCl (pH=1.25), we achieve the highest HMF productivity of 15.2% yield/min among all studies using flow reactors with aqueous phase and the same initial glucose concentration. Both our fast heat and mass transfer of our microreactor and the

129 effectiveness of the tandem CrCl3/HCl catalysts contributed to the observed high HMF productivity.

20 Batch Flow This Work 15

10 Zhang et al. Watanabe et al.

(% Yield/min) (% 5 Muranaka et al. HMF Productivity HMF Lv et al. Siqueira et al. 0 180 190 200 210 220 o Temperature ( C)

Figure 5.5 Maximum HMF yield per time (productivity) vs. temperature from this work and relevant literature. All literature used 1 wt% glucose solutions at inlet in single (aqueous) phase only. Batch data from our group is not compared as high temperatures require use of Parr reactors and heating to reaction temperature takes considerable time leading to poor productivity.

5.3.3 Understanding the Effect of CrCl3 Speciation on Reactivity

3+ It is well known that the chromium (III) hexaaquo complex [Cr(H2O)6] forms

58 174, 175 176 after CrCl3 is dissolved in water. Choudhary el. and others have observed the further speciation of Cr (III) ions in water as shown below:

3+ 2+ + [퐶푟(퐻2푂)6] + 퐻2푂 ↔ [퐶푟(퐻2푂)5 (푂퐻)] + 퐻3푂

2+ + + [퐶푟(퐻2푂)5 (푂퐻)] + 퐻2푂 ↔ [퐶푟(퐻2푂)4 (푂퐻)2] + 퐻3푂

130 3-y The formation of [Cr(H2O)x(OH)y] ions, water-soluble Cr oligomers, and colloidal Cr(OH)3 particles, accompanied by the release of protons which lowers the solution pH, have been proposed.

However, little is known about the species formed under the reaction conditions of interest and the relevant time scales. The change in Cr(III) speciation and the accompanying change in catalytic activity for sugars dehydration is poorly understood.

Norton et al. investigated the speciation of Al(III) species and solids formation as a function of heating time and HCl addition using 27Al-NMR, Inductively Coupled

Plasma Mass Spectrometry (ICP-MS) and OLI modeling.64 Ultracentrifugation was used to separate the permeate for quantitative ICP-MS analysis to calculate the Al concentration, and the balance was taken as solid particles. It was shown that the soluble Al(III) ions form quickly, while the Al(OH)3 solids form after extended heating, leading to a drop in catalytic activity due to removing the active species from solution.

As a result, there are at least two time scales in the problem: a short one that controls speciation of small ions and the pH and a long time that is controlled by growth of solid particles during which the pH is quasi-equilibrated. At least 20 mM of HCl was needed to suppress solids’ formation. Detailed kinetics for these processes was though difficult to obtain from batch experiments. Similar work for Cr(III) speciation on reactivity remains elusive. To bridge these gaps, we characterize the chromium species formed under reaction conditions as well as perform reactivity measurements to understand their effect on glucose conversion and product yields. These are enabled by the having nearly isothermal conditions in the microreactor and short residence times.

131 First, catalyst treatment experiments were conducted in the absence of a substrate. Specifically, 3.4 mM of CrCl3 solutions, which were the same as the feed concentration used for the optimization experiments discussed in Section 5.3.2, were heated using the flow reactor for various residence times. This catalyst concentration was also chosen to increase the sample absorption for detection by UV-Vis. Prior work by our group on Al(III) speciation suggested that pH measurements done in-situ at reaction temperature do not differ appreciably from those done after quenching the solution to room temperature.64 For this reason, the pH was measured for chilled eluent

(Figure 5.6). At all temperatures, the pH drops rapidly within the first 10 min and stabilizes at longer residence times, pointing to the hydrolysis of Cr3+ ions which releases protons. As expected, the pH stabilizes faster as the temperature increases. At

140 oC, the pH stabilizes within 5 min of heating while at 200 °C, the pH stabilizes in a

1 min. The color of the solution was light green initially and became slightly bluish after 1 h of preheating, as shown in Figure 5.7. After ultracentrifugation at 10,000 rpm for 10 min, the permeate became almost colorless but the retentate kept the green-bluish color. It can be concluded that green colloidal particles had formed upon preheating the

58, 174 CrCl3 solution, consistent with prior literature findings. Dynamic light scattering

(DLS) experiments were attempted aiming at measuring the size of these particles, but due to the sample absorbing visible light at 532 nm, which is the wavelength of the laser of the DLS instrument, the Cr ions gave artifact background signals and unreliable autocorrelation functions.

132 4.0 140 oC 160 oC 3.5 180 oC 200 oC

3.0 pH

2.5

2.0

0 10 20 30 40 50 60 Contact Time (min)

Figure 5.6 Measured pH of a 3.4 mM chromium chloride solution as a function of contact (residence) time at temperatures indicated. The solutions were heated using the flow microreactor.

Figure 5.7 (a) Freshly prepared 3.4 mM CrCl3 solution and (b) a 3.4 mM CrCl3 solution after preheating at 140 °C for 1 h.

Next, we used these preheated CrCl3 solutions as catalysts for glucose conversion and compared their activity at the same reaction time of 45 min. Because of

133 the need to preheat milliliters of CrCl3 catalyst for long hours, batch glass vial reactors were used to perform the preheating and the reactions and compare the trends in glucose conversion and product yields. The glucose conversion data in Figure 5.8a shows that activity of the CrCl3 solutions preheated at 140 °C for more than 1 h continues to drop despite exhibiting a stable pH. This finding indicates that similar to AlCl3, there is a longer time scale than that involved in establishing the proton concentration, most probably associated with formation of solids that remove active species from solution.

Figures 5.8b and 5.8c show that preheating a mixture of HCl and CrCl3 with a pH less than or equal to 2 for various times up to an extended period of 24 h does not lead to an appreciable drop in activity. This can be attributed to the acid suppressing the formation of Cr(OH)3.

134 (a) 60 (b) 70 Glucose Glucose Fructose 50 60 Fructose HMF HMF 40 50 40 30 30 20 20

10 10

Conversion or Yield (%) Yield or Conversion Conversion or Yield (%) Yield or Conversion 0 0 0 1 2 4 12 24 0 1 2 12 24 CrCl Preheating Time (h) CrCl Preheating Time (h) 3 3

(c) 40 (d) 35 Glucose GLU Conversion FRU Yield Fructose 30 HMF HMF Yield 30 HMF Selectivity 25

20 20 15

10

10 or Selectivity (%) Selectivity or

Conversion, Yield Yield Conversion, 5 Conversion or Yield (%) Yield or Conversion 0 0 0 1 2 12 24 Control 0 24 CrCl Preheating Time (h) CrCl Preheating Time (h) 3 3

Figure 5.8 Glucose conversion and fructose and HMF yields as a function of CrCl3/HCl catalyst preheating time. Reaction conditions: 1 wt% glucose with 3.4 mM CrCl3 catalyst and varying HCl at 140 oC in a batch reactor for 45 min. Two repeat experiments were conducted for each set of reaction conditions. (a) No HCl (starting pH=3.76). (b) 0.01 M HCl (pH=2). (c) 0.1 M HCl (pH=1). (d) CrCl3 preheated for 0 or 24 hours before adding HCl to a solution to bring pH to 1. “Control” refers to a batch experiment conducted with 1 wt% glucose and 0.1 M HCl (pH=1) without CrCl3 at 140 °C for 45 min.

To understand the reversibility of forming solids, we preheated a 3.4 mM CrCl3 solution for 24 hours at 140 °C. After cooling down this solution to room temperature, a small amount of 12.1 M of HCl was added to bring the solution pH to 1. Then glucose was added in 1:100 mass ratio with respect to the catalyst solution. The mixture was

135 shaken well and loaded into a reactor vial, capped and transferred to a temperature- controlled oil bath kept at 140 °C. After 45 min of reaction, the vial was taken out and quenched to room temperature. Compared to using a freshly prepared, non-preheated

CrCl3/HCl catalyst mixture, this CrCl3/HCl catalyst gives significantly lower glucose conversion and HMF yield (Figure 5.8d). Compared to a control experiment where only a 0.1 M HCl catalyst is used, this CrCl3/HCl catalyst still gives higher glucose conversion and product yields. When HCl is added at the beginning prior to catalyst heating (Figure 5.8b and c), the reactivity is higher as HCl prevents formation of solids.

Our results indicate that upon CrCl3 preheating, likely soluble oligomers with degree of polymerization varying from 2 to 6175, 177, 178 and solids form by condensation of the

3-y [Cr(H2O)x(OH)y] species. Once formed, dissolution of the solids is slow, i.e., the process appears as irreversible under reaction conditions. Yet, the concentration of active centers is higher compared to using only HCl. Our results indicate that the catalyst pretreatment is important and getting optimal reactivity and reproducible results require recording the catalyst history.

136 (a) (b) 0.10 5 Initial Initial Peak 1 0.6 min 0.6 min Peak 2 Peak 3 5 min 5 min 0.08 4 15 min 15 min 1 h 1 h 4 h 4 h 0.06 3 24 h 24 h

2 0.04

Absorbance Absorbance 1 0.02

0 0.00 200 250 300 350 300 400 500 600 700 800 Wavelength (nm) Wavelength (nm)

Figure 5.9 UV-Vis spectra of a freshly prepared 3.4 mM CrCl3 solution and of the same solution preheated at different times. (a) Low wavelengths (200-350 nm) and (b) high wavelengths (300-800 nm). Preheating for 1 h and below was done in the flow microreactor, and preheating for above 1 h was done in batch glass vials reactors. The observed drop in catalytic activity for the preheated CrCl3 emphasized the need for a standard procedure for preheating. The amount of time or residence time for CrCl3 preheating should be carefully chosen and consistent within each set of time-dependent study.

Next, we employ UV-Visible spectroscopy (UV-Vis) to follow the progress of

Cr(III) speciation in water and connect it with the observed drop in catalyst activity.

The UV-Vis spectrum of the freshly prepared CrCl3 solution shown in Figure 5.9 is in good agreement with the spectra given by Onjia et al. 174 The spectra exhibit three peaks: a broad and intense peak between 200 and 300 nm (Peak 1) which corresponds to at least one pair of negative ligands coordinated in trans positions179, and two much less intense peaks centered at 443 (Peak 2) and 633 nm (Peak 3) which correspond to the d−d transitions between different levels split from the d-orbital set of Cr3+.180 181

- 180 Additionally, the shoulder at 692 nm was assigned to [Cr(H2O)2Cl4] by Elving et al.

After just 0.6 min of heating at 140 °C, a blue shift (i.e., a shift of peak positions to lower wavelengths) was observed for peaks 2 and 3. This is consistent with the color of

137 2- the solution turning from bright green due to [CrCl4] to a more bluish color associated

3+ 182 with [Cr(H2O)6] . At the same time, the intensity of peak 1 reduces significantly.

179 3+ According to Tsuchida et al., the [Cr(H2O)6] does not have any pairs of negatively charged coordinated ligands and therefore should not show an absorption peak in the ultraviolet region. The disappearance of peak 1 and of the shoulder at 692 nm after 0.6 min of preheating is consistent with the complete replacement of coordinated Cl- ions by water molecules. As the preheating time becomes longer, peak 1 regrows in intensity and becomes constant for longer than 4 hours with a different shape than that of the freshly prepared solution. This points to the formation of new electronic transitions

+ arising from most likely an [Cr(H2O)4(OH)2] ion. We were not able to predict changes

2+ in the concentration of the catalytically active [Cr(H2O)5(OH)] species using the UV-

Vis technique, since this ion does not have any pairs of coordinated negative ions, which are essential for exhibiting peak 1. The positions of peaks 2 and 3 for the fresh and preheated CrCl3 samples are summarized in Table D1. The peak positions are overall very similar for preheating times longer than 0.6 min. Peaks 2 and 3 red-shift by

11 nm and 16 nm upon preheating for 0.6 min and 24 h, respectively, pointing to the formation of Cr species coordinated with more OH- ions. According to studies on the effect of anion coordination in the UV-Vis spectra of transition metal complexes179,

- replacing a H2O ligand with a OH could cause a slight red shift of the UV-Vis peak in the 350-800 nm region. Apart from peak-position changes, the intensity of Peaks 2 and

3 decreases from 0 to 0.6 min of preheating and increases at longer heating times. The change in absorbance also implies a change in Cr speciation, but the concentration of

138 each species could not be determined given that their peak positions are very close and hard to deconvolute. Advanced Raman spectroscopic methods may be valuable to

3-y further identify the various [Cr(H2O)x(OH)y] species.

To decouple potential contributions to the overall absorbance spectra from different Cr species, we used ultracentrifugation to separate the particles generated from preheating the CrCl3 for 24 h and washed the retentate with DI water to remove the residual soluble Cr species. The retentate has an intense green color and the permeate is almost colorless. The UV-Vis spectra collected for the unfiltered mixture, the washed and diluted retentate, as well as the permeate are shown in Figure D5. Interestingly, after filtration, the permeate has almost no absorption in the 400-800 nm region (Figure

D5b) but the absorption was still very strong in the 200-350 nm region (Figure D5a).

Because it was not able to measure the volume or the Cr(III) concentration of the retentate, the intensity of the spectrum of the retentate in Figure D5 could not be quantitatively compared with the other spectra. Since the Cr particles absorb considerably, the observed overall spectra of CrCl3 sample preheated at different times

3-y must contain contributions from both soluble [Cr(H2O)x(OH)y] ions and Cr solids.

The observed red shifts of peaks 2 and 3 may result from the formation of soluble

3-y [Cr(H2O)x(OH)y] ions, oligomers, or insoluble Cr(OH)3 species, since all of these

- 3+ contain more OH units compared to the [Cr(H2O)] ion formed right after dissolving

CrCl3 in water.

139 12 160 oC o 10 180 C

8 mol/L-s) -4 6

4

Glucose Reaction Reaction Glucose Rate (10 Rate 2

0 20 40 60 80 100 CrCl Preheating Time (min) 3

Figure 5.10 Reaction rates for glucose dehydration at 160 °C and 180 °C, catalyzed by CrCl3 catalysts preheated for different times indicated. Reaction conditions: 1 wt% glucose, 1.7 mM CrCl3, at 160 and 180 °C. The residence times range from 3-30 s at 160 °C, and 3-20 s for 180 °C. The catalyst was preheated in the flow reactor at the same temperature as that used for the glucose reaction. The dotted lines are guides to the eye.

To model the effect of preheating of preheating on catalyst deactivation, we conducted kinetics studies in our flow reactor and used residence times 3-30 s to keep the glucose conversion below 15% and at the same time minimize further transformation of the catalyst. An example calculation for obtaining the glucose reaction rate is given in Figure D6. Each reaction uses CrCl3 catalyst feeds preheated at different intervals between 0 and 90 min at the desired reaction temperature and cooled to room temperature. Figure 5.10 depicts a non-monotonic variation of glucose rate vs. the CrCl3 preheating time at both temperatures. Preheating the CrCl3 catalyst increases

140 the reaction rate rapidly up to 5 min but decreases it for longer times. These results point towards first an increase in the concentration of the actives species

2+ [Cr(H2O)5(OH)] at short catalyst preheating times followed by a decrease at longer preheating times. This behavior can be explained by the interplay between the

3+ hydrolysis of [Cr(H2O)6] and the formation of oligomers and eventually Cr(OH)3

3+ solids. At short preheating times (< 5 min), [Cr(H2O)6] is rapidly hydrolyzed to form

2+ + [Cr(H2O)5(OH)] releasing H . This is supported by our results in Figure 5.6, where the

2+ pH rapidly drops within the first 5 min of preheating. Since [Cr(H2O)5(OH)] is the active species for glucose dehydration, its increasing concentration is consistent with the increase in the glucose consumption rate seen in Figure 5.10. At longer preheating

3+ times (> 10 min), the [Cr(H2O)6] hydrolysis is either complete and/or the

2+ [Cr(H2O)5(OH)] starts forming oligomers and Cr(OH)3 solids, both of which lower the concentration of the active species and consequently the glucose consumption rate.

Figure 5.10 suggests that to achieve the highest initial glucose reaction rate, a catalyst preheating time of ~1 min should be used for reaction temperatures of 160-180 °C.

Since the residence times used here are from 3 to 30 s, we expect that CrCl3 is

2+ undergoing some speciation change while the active species [Cr(H2O)5(OH)] catalyzes the reaction. The comparable preheating and reaction time scales underscore the need for incorporating the Cr(III) speciation kinetics when modeling the glucose reaction with homogeneous Lewis and Bronsted acid catalysts.

141 5.4 Conclusions

In this work, we achieved fast glucose conversion to HMF using tandem

CrCl3/HCl catalysts in water solvent in a flow microreactor. After optimizing the reaction temperature and HCl concentration, the highest ever HMF productivity of 15.2

% HMF yield/min (a two-fold increase over existing literature) was achieved at 200 °C using 1.7 mM CrCl3 and HCl with a pH of 1.25.

We show for the first time using the flow microreactor that the preheating of

CrCl3 catalyst solution plays a crucial role in the rate of glucose consumption. If the catalyst is not preheated, the concentration of active species for glucose isomerization,

2+ [Cr(H2O)5(OH)] is low, and the glucose reaction rate is relatively low. As the

2+ preheating time increases, more [Cr(H2O)5(OH)] forms due to the hydrolysis of

3+ [Cr(H2O)6] , and the glucose reaction rate increases. However, further increase in preheating time compromises the glucose reaction rate due to the formation of soluble oligomers and Cr(OH)3 particles that reduce the concentration of the active species. For the highest initial glucose reaction rate, catalyst preheating times of ~1 minute is suggested for carrying out the reaction at 160-180 °C. For experiments using tandem

CrCl3/HCl catalysts, enough HCl (>0.01 M) suppresses the formation of these Cr oligomers and solids. In the future, advanced Raman spectroscopy or liquid chromatography techniques could be exploited to resolve the peaks of the different soluble Cr(III) species and determine their concentration. Additionally, particle size measurement techniques not affected by the color of the sample, such as small angle X- ray scattering, may be used to monitor the growth of Cr(OH)3 colloids over time.

142 Chapter 6

GLUCOSE CONVERSION TO 5-HYDROXYMETHYL FURFURAL IN BIPHASIC SOLVENT MIXTURES IN A MICROREACTOR

6.1 Introduction

Due to diminishing fossil fuel resources and the consumers’ growing awareness of global warming, manufacturing fuels and chemicals from renewable lignocellulosic biomass is receiving considerable attention. As such, converting biomass-derived carbohydrates to platform molecules, including 5-hydroxymethyl furfural (HMF), is studied extensively. Since the pioneering work of Roman-Leshkov et al.,89 use of biphasic solvent mixtures has become common to optimize the yield of HMF. In a biphasic system, the organic solvent continuously extracts the HMF formed from the sugar in the aqueous phase. The lower concentration of HMF in the aqueous phase reduces the side reactions and thereby improves HMF yield.69 Organic solvents with low solubility in water and high partition coefficients for HMF are favorable. Sec- butylphenol (SBP),41, 88 methyl isobutyl ketone (MIBK),91, 183 and aliphatic alcohols89,

184, 185 are among the most commonly used organic solvents.

Besides improving the reaction media, the application of microreactors to biomass processing is receiving growing attention. Compared to batch reactors, microreactors achieve orders of magnitude higher heat and mass transfer leading to

143 improved productivity and selectivity.170 Microreactors can also be operated continuously and require less start-up and shut-down times. They also allow precise control of short reaction times.96 Capillary microreactors are particularly promising because the surface-to-volume ratio of the liquid–liquid biphasic system can be increased via generation of a dispersed phase into a continuous phase, e.g., slug or droplet flow.95, 186-188

Several studies65, 97, 189-191 have focused on the conversion of sugars to HMF using biphasic solvent mixtures in flow reactors. Most of them have used fructose because it can be converted to HMF faster and in higher yields. Brasholz et al.191 conducted a HCl catalyzed reaction of 10 wt % fructose using 0.25 M aqueous HCl as the reactive phase and MIBK as the organic phase. The water to MIBK volume ratio was varied. The best results were obtained when the flow rate of MIBK to water was 3.

At 140 °C and 15 min, 74% total HMF yield with 84% fructose conversion was obtained. Shimanouchi et al.189 also used 10 wt% fructose and MIBK as the organic phase. The HCl catalyst concentration was 0.025M. At 180 °C, in only 2 minutes, 93% fructose conversion and 89% of total HMF yield was achieved. Similarly, Lueckgen et al.190 used 10 wt% fructose as substrate and HCl in water and MIBK as the organic phase. Fructose was converted to HMF in less than 40 s with a combined HMF yield in both phase higher than 90% at 150 °C. HMF was obtained in MIBK with a yield of

80%.

144 Compared to fructose, glucose could significantly lower the cost, although achieving high HMF yields in short reaction times still remains challenging. Very few studies have focused on glucose. Muranaka et al.97 used a water/SBP biphasic solvent mixture of aqueous, acidic phosphate buffered saline (PBS) at pH of 2. Up to 81% HMF yield from 1 wt% fructose and 76% HMF yield from 1 wt% glucose was achieved at

180 °C in 12 and 47 min, respectively. Very recently, Guo et al.65 used 1 mol/L (18 wt%) concentrated glucose, AlCl3 catalyst in water and MIBK. 83% glucose conversion and 66% total HMF yield was achieved in 16 minutes. Compared to the work of

Muranaka et al.,97 similar HMF yields were achieved in a shorter time, with a much more concentrated glucose feed.

In this work, we improve HMF productivity from glucose by using water/organic biphasic solvent mixtures using the capillary flow reactor built by Desir et al.108 and benchmark the performance to single phase results. We choose tandem

58 Lewis and Brønsted CrCl3/HCl as the aqueous phase catalysts. Choudhary et al. have demonstrated through Extended X-ray Absorption Fine Structure (EXAFS) studies that

2+ the [Cr(H2O)5(OH)] formed during the hydrolysis of CrCl3 is the active species for glucose isomerization. It has been shown in earlier glucose conversion studies that

42, 59 combination of CrCl3 with HCl in appropriate ratios can maximize the

2+ concentration of the active species [Cr(H2O)5(OH)] which leads to improved HMF yield compared to using either catalyst alone. For the organic solvent, we choose 2- pentanol and MIBK because they are less toxic and cheaper than 2-secbutylphenol.

Specifically, 2-pentanol not only has a good extraction efficiency for HMF but is also a

145 hydrogen donor solvent for the potential hydrodeoxygenation (HDO) of HMF to 2,5- dimethylfuran. First, we find conditions that optimize HMF yield and productivity by varying temperatures and residence times at fixed organic-to aqueous flow rate ratio.

Then, the effects of organic-to aqueous flow rate ratio and different organic solvents on the glucose conversion as well as the yield of HMF and other products are also investigated.

6.2 Methods

6.2.1 Materials

Glucose (≥99% purity, Sigma Aldrich), fructose (≥99% purity, Sigma Aldrich), mannose (≥99% purity, Sigma Aldrich), chromium chloride hexahydrate (≥99% purity,

Sigma Aldrich), hydrochloric acid (37 wt%, Fisher Scientific), sulfuric acid (5M,

Fluka), 2-pentanol (≥99% purity, Sigma Aldrich), and methyl isobutyl ketone (≥99% purity, Sigma Aldrich) were used without further purification. All aqueous solutions were prepared using deionized (DI) water obtained using a Millipore water purification system (model: Direct-Q3 UV R).

6.2.2 Microreactor Setup and Experiments

The microreactor used in this study was built in our laboratory (Scheme 6.1). A total of three pumps were used. Pump 1 (Model: Teledyne SSI-LS010PFT3A) is connected to a 10 wt% aqueous glucose solution. Pump 2 (Model: Teledyne SSI-

146 LS005PFT3A) is connected to the catalyst feed where the catalyst concentration is 1.89 mM CrCl3 and 0.0625 M HCl. Pump 3 (Model: Eldex Optos 1-HM) is connected to a pure organic solvent (2-pentanol or MIBK) feed. The catalyst and organic solvent enter each a preheating coil of 1.5 m in length, 0.16 cm in inner diameter and a total volume of 3 mL. To avoid hydrothermal carbonization, the glucose feed directly enters Tee 1 without being preheated. The preheated catalyst stream is combined with the glucose stream at Tee 1, and the combined aqueous stream immediately enters Tee 2, where it encounters the preheated organic solvent stream. Here, the biphasic mixture enters a reactor coil 500 µm in diameter. The flow rate ratio of the glucose to catalyst streams is kept at 1:9 so that the combined aqueous stream contains 1 wt% of glucose, 1.7 mM of

CrCl3 and 0.056 M HCl (pH=1.25). In our previous glucose dehydration work in single phase, this combination of CrCl3 and HCl catalysts concentrations gave the highest

HMF yield. The volume of the reactor is 0.4 mL, 0.6 mL and 0.8 mL for organic/aqueous flow rate ratios (O/A) of 1:1, 2:1 and 3:1, respectively. The total flow rates at each residence time and O/A value are given in Table E1. To prevent back flow, one in-line check valve (IDEX Health& Science, Catalog No. CV-3345) is used in each of the three feed lines immediately before Tees 1 and 2. A K-type thermocouple was connected to Tee 3 to monitor the temperature at the outlet of the reactor. The eluent from the reactor enters immediately into an ice bath to be quenched. A pressure gauge and a 350 psi back-pressure regulator were used in series to monitor and control the system pressure. After waiting for 10 times as long as the residence time, the eluent from the back-pressure regulator was collected into a clean glass vial and allowed to

147 settle into clear aqueous and organic layers. Then each phase was collected separately for High Performance Liquid Chromatography (HPLC) analysis to quantify the reactant and products.

Scheme 6.1. Schematic of the microchannel flow reactor setup, and picture of the coiled reactor section.

6.2.3 Product Analysis

An Agilent 1260 Infinity HPLC equipped with a refractive index detector (RID), a photodiode array detector (DAD) operated at 254 nm wavelength, and an Aminex

Biorad HPX-87H column (dimensions: 300×7.8 mm with 9 µm packing size) heated at

75 °C was used for determination of the glucose, mannose, fructose, and HMF

148 concentrations in the aqueous sample. The mobile phase was DI water at 0.5 mL/min flow rate. Both the organic and the aqueous phase samples were run on a Waters 2695

HPLC equipped with a RID and an Aminex Biorad HPX-87H column (dimensions:

300×7.8 mm with 9 µm packing size) for determination of the acids and HMF concentration. The column was heated to 50 °C. The mobile phase was 5 mM of aqueous H2SO4 solution at 0.5 mL/min flow rate. External calibration standards were used in both HPLCs.

6.3 Results and Discussion

6.3.1 Effect of Temperature

Two repeat experiments were conducted at 160 °C to ensure reproducibility. The glucose conversion and yields of fructose and HMF with the error bars are given in

Figure E1. The glucose conversion and HMF yields as functions of residence time at three temperatures are given in Figure 6.1. As expected, both the glucose conversion and HMF yield increase with temperature (Figures 6.1a, b and c). For 160 and 180 °C, the HMF yield monotonically increases with increasing residence time; at 200 °C, the

HMF yield reaches a maximum of 51.4% at 92 s and then decreases at 120 s. When the

HMF yields are plotted against glucose conversion (Figure 6.1d), the data from the three temperatures overlap. This overlap points to the same reaction pathways and similar activation energies among paths. The yields of other products, including mannose (MAN), fructose (FRU), levoglucosan (LG), formic acid (FA), and levulinic

149 acid (LA) as functions of residence time and temperature are depicted in Figure E2. The maximum HMF yield of 51.4% is very similar to the maximum HMF yield of 50.8% we obtained using a batch reactor at lower temperature (140 °C) at long reaction times of several hours using water/tetrahydrofuran (THF) biphasic system (Figure E3).

(a) (b) 100 100 GLU GLU HMF HMF 80 80

60 60

40 40

20 20

Conversion or Yield (%) Yield or Conversion (%) Yield or Conversion 0 0 0 50 100 150 200 250 300 0 50 100 150 200 Residence Time (s) Residence Time (s)

(c) (d) 100 50 GLU 160 oC HMF 180 oC 80 40 200 oC

60 30

40 20

20 (%) Yield HMF 10 Conversion or Yield (%) Yield or Conversion 0 0 0 20 40 60 80 100 120 140 0 20 40 60 80 100 Residence Time (s) GLU Conversion (%)

Figure 6.1. Glucose (GLU) conversion and HMF yield as a function of residence time at temperatures of (a) 160, (b) 180 and (c) 200 °C. (d) HMF yield vs. glucose conversion ° at 160, 180, and 200 C. Aqueous feed composition: 1 wt % glucose, 1.7 mM CrCl3 and 0.056 M HCl (pH=1.25). 2-pentanol was used as the organic solvent in 1:1 flow rate ratio to the aqueous phase.

150 6.3.2 Comparison with Single Phase Data of This Work and Biphasic Systems of

Literature

To benchmark the performance of the biphasic microreactor setup, Figure 6.2a compares the glucose conversion and HMF yield vs. residence time in single and biphasic systems at the same temperature and inlet feed composition. As expected, the

HMF yield is higher in the biphasic system compared to the single phase. Surprisingly, the glucose conversion is also significantly higher in the biphasic system. The reason for this trend is not exactly known from literature. A maximum HMF yield of 51.4% occurs at 92 s for the biphasic system, while the HMF yield reaches only 30% in 120 s in the single phase. When the HMF yields are plotted against glucose conversion

(Figure 6.2b), the data for the biphasic and single-phase systems overlap. This overlap again indicates the glucose dehydration follows the same reaction pathway despite the

HMF-extracting solvent present in the biphasic system.

(b) (a) 100 50 Biphasic-GLU Single Phase Single Phase-GLU Biphasic-HMF Biphasic 80 Single Phase-HMF 40

60 30

40 20 HMF yield (%) yield HMF

20 10 Conversion or Yield (%) Yield or Conversion 0 0 0 20 40 60 80 100 120 0 20 40 60 80 100 Residence Time (s) GLU Conversion (%)

151 Figure 6.2. (a) Glucose conversion and HMF yield vs. residence time and (b) HMF yield vs. glucose conversion at 200 °C in biphasic and single-phase microreactors. The aqueous feed in both setups contains 1 wt % glucose, 1.7 mM CrCl3, and 0.056 M HCl (pH=1.25). In the biphasic setup, 2-pentanol was used as the organic solvent in 1:1 flow rate ratio to the aqueous phase.

Next, we report the HMF productivity and compare it with literature studies.

The reaction conditions used in each study are summarized in Table 6.1. In Figure 6.3, the works of Binder et al.,92 Nikolla et al.90 and Choudhary et al.58 were done using batch reactors, and the works of Guo et al.,65 and Muranaka et al.,97 were done using flow reactors. At 200 °C in 92 s, we achieve the highest HMF yield of 51.4%, which corresponds to the productivity of 33.6% yield/min (Figure 6.3b) among all studies on glucose substrate with biphasic systems. Both the fast heat and mass transfer of our microreactor and the effectiveness of the tandem CrCl3/HCl catalysts contributed to the observed high HMF productivity. Increased productivity translates directly to lower reactor volumes and lower capital cost.

Table 6.1. Summary of literature on glucose conversion to HMF in biphasic solvents. Reference Reactor Catalyst Solvent T Other Time Max. Conversion Productivity

# (°C) conditions (min) HMF (%) (% HMF

Yield yield/min)

(%)

92 1 Batch CrCl3 Water/DMA 100 10 wt % 300 81 N/A 0.27

GLU

290 Batch Sn-BEA Water/THF/NaCl 180 10 wt% 70 57 79 0.81 GLU

152 58 3 Batch CrCl3/HCl Water/THF/NaCl 140 10 wt% 180 60 95 0.335

GLU

497 Flow PBS Water/SBP 180 1 wt% 47 76 100 1.62

buffer GLU

65 5 Flow AlCl3/HCl Water/MIBK/ 160 18 wt% 16 66 83 4.14

NaCl GLU

6 (This Flow CrCl3/HCl Water/2- 200 1 wt% 1 42 85.3 42.0 work) Pentanol GLU

(a) 100 Batch reactor Flow Reactor 80

60

40

20 Max. HMF Yield (%) Max. Yield HMF

0 Binder Nikolla Choudhary Muranaka Guo This Work

153 (b) 100 This Work

10 Guo et al. Muranaka et al.

1 (% (% Yield/min) Nikolla et al.

HMF Productivity Productivity HMF Binder et al.

0.1 80 100 120 140 160 180 200 220 Temperature (oC)

Figure 6.3. (a) Maximum HMF yield and (b) maximum HMF yield per time (productivity) vs. temperature from this work and relevant literature. All these prior works used glucose in aqueous/organic biphasic solvent mixtures. Experiments in flow reactors (red triangles) and batch reactors (blue squares).

6.3.3 Effect of Organic to Aqueous Flow Rate Ratio

Next, we study the effect of organic to aqueous flow rate ratio (O/A). The glucose conversion and HMF yield are given in Figure 6.4. Interestingly, the glucose conversion is similar at each residence time for O/A values of 1, 2, and 3. The HMF yields are also similar for all three O/A values. As can be seen in Figure 6.4b, the HMF yield for all three O/A exhibits a maximum at 92 s, and decreases at 120 s.

Unfortunately, we were not able to collect reliable data for residence times longer than

120 s at 200 °C due to the pressure build up and flow instability caused by the formation of humins. The partition coefficient of HMF (R) between the 2-pentanol and water phases ranges from 1.3 to 1.4 for all residence times and O/A values. These R values

154 are in good agreement with the value of 1.4 determined in fructose dehydration in water/2-pentanol system by Leshkov et al.89 The yields of other products are shown in

Figures E2 (c) and E4.

(a) 100 (b) 60 O/A=1 O/A=1 O/A=2 O/A=2 50 80 O/A=3 O/A=3 40 60 30 40

20 HMF Yield (%) Yield HMF

20 10 GLU Conversion (%) Conversion GLU 0 0 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 Residence Time (s) Residence Time (s)

(c) 60 O/A=1 O/A=2 50 O/A=3

40

30

20 HMF Yield (%) Yield HMF 10

0 0 20 40 60 80 100 GLU Conversion (%)

Figure 6.4. (a) Glucose conversion and (b) HMF yield vs. residence time and (c) HMF yield vs. glucose conversion at 200 °C at organic to aqueous flow rate ratios of 1:1, 2:1, and 3:1. The aqueous feed in all three reactions contains 1 wt % glucose, 1.7 mM CrCl3 and 0.056 M HCl (pH=1.25).

155

Although no literature on the HMF production from sugars using water/2- pentanol biphasic systems is available, studies have been carried out using other water/organic biphasic systems with various O/A. Gomes et al.185 conducted fructose dehydration at 180 °C with H3PO4 catalyst using water/acetone, water/butanol and water/diethyl ether biphasic systems where the aqueous phase was saturated with NaCl to create phase separation and improve the HMF extraction efficiency. The glucose conversion was close to 100% for all experiments. Upon increasing O/A from 1 to 2, the HMF yield increased from 59% to 80% for the water/acetone system and increased from 18% to 46% for the water/diethyl ether system. However, the HMF yield was very similar for O/A=1 and 2 in the water/2-butanol system. Muranaka et al.97 conducted glucose dehydration in water/SBP solvents using acidic PBS as the catalyst at 180 °C in a flow reactor. Both the glucose conversion and the HMF yield improved significantly when O/A was brought from 1 to 3. The same authors also used fructose as the substrate and investigated more O/A values between 1 and 4. Both the glucose conversion and the HMF yield improved when O/A was brought from 1 to 2, but stopped increasing when O/A was further raised to 3 and 4. No explanation for these trends was provided.

It appears that the change in sugars conversion and HMF yield with O/A is complex and specific to the substrate, the type of organic solvent and the O/A. It is unclear whether their processes were limited by mass transfer or reaction kinetics. The type of flow pattern, which plays a key role in the transport of HMF from the aqueous to the organic phase, was also unclear. According to a flow characterization study by Desir et al.,192

156 where laser-induced fluorescence (LIF) was used to elucidate the biphasic flow in capillary microchannels, the flow pattern changes from slug flow at low total flow rates to droplet flow to slug-and-droplet flow and to annular or dispersed flow as the flow rate increases. The specific type of flow pattern and the flow rate for the transition between patterns is affected by the O/A (Figure E5). For easy comparison, the flow patterns were plotted against the residence times used in our reactions based on the flow rates used (Table E1), as shown in Figure E6a. For residence times of 12, 24, 30, 40, and 60 s, different flow patterns are seen at O/A=1, 2 and 3 at the same residence time.

Given that the glucose conversion and the HMF yield are not strongly affected by the

O/A, it is most likely that our process is limited by reaction kinetics rather than mass transfer.

6.3.4 Comparison between Water/2-Pentanol and Water/MIBK Systems

(a) (b) 100 60 2-PeOH, GLU 2-PeOH MIBK, GLU MIBK 2-PeOH, HMF 50 80 MIBK, HMF 40 60 30

40 20

20 (%) Yield HMF 10

Conversion or Yield (%) Yield or Conversion 0 0 0 20 40 60 80 100 120 0 20 40 60 80 100 Residence Time (s) GLU Conversion (%)

157 (c) (d) 10 2-PeOH 60 2-PeOH MIBK MIBK 8 50

6 40 30 4 20

MAN Yield (%) Yield MAN 2 Humins Yield (%) Yield Humins 10

0 0 0 20 40 60 80 100 0 20 40 60 80 100 GLU Conversion (%) GLU Conversion (%)

Figure 6.5. (a) Glucose conversion and HMF yield vs. residence time in the water/2- pentanol and water/MIBK systems. (b) HMF, (c) mannose, and (d) humins yield vs. glucose conversion. Reaction conditions: O/A=1:1 at 200 °C. The aqueous feed in both reactions contain 1 wt % glucose, 1.7 mM CrCl3 and 0.056 M HCl (pH=1.25). To compare the reactivity of glucose in water/2-pentanol and water/MIBK, an additional experiment was performed in water:MIBK 1:1 flow rate ratio. As shown in

Figure 6.5a, both the glucose conversion and the HMF yield are slightly higher at residence times <60 s in the water/MIBK system compared to water/2-pentanol system.

The HMF yields for the water/MIBK system reaches a maximum of 43.0% at 60 s and starts dropping at 92 s. We were not able to collect reliable data for residence time of

120 s for the water/MIBK system due to the pressure and flow instability. As can be seen in Fig 6.4a, the HMF yield for the 2-pentanol/water system at O/A=1 reaches a maximum of 51.4% at 92 s and decreases at 120 s. Hence, compared to the 2- pentanol/water system, the maximum HMF yield for the water/MIBK system is lower and appears at shorter residence time. The partition coefficient of HMF (R) between 2- pentanol and water phases ranges from 1.3-1.4, and the R value for MIBK/water phases are 1.1-1.2. The HMF yield vs. glucose conversion data overlaps in Figure 6.5b,

158 indicating similar pathways to HMF formation in the two biphasic systems. However, the yield of other major products vs. glucose conversion do not overlap in the two biphasic systems. Specifically, the yield of mannose monotonically decreases with glucose conversion in the water/2-pentanol system and goes through a maximum in the water/MIBK system. Overall, the mannose yield is higher in the water/2-pentanol system than in the water/MIBK system. In Figure 6.5d, the yield of humins, calculated as the balance between the conversion and the sum of the carbon yields of all HPLC- detectable products, is also lower in the water/2-pentanol system. The different yield vs. conversion trend for mannose suggests that the glucose epimerization step is affected by the organic solvent. Since mannose is only present in the aqueous phase after the reaction, it is not expected that the difference in extraction efficiency between water/2- pentanol and water/MIBK systems would affect its yield or selectivity. The yields of other products as a function of residence time are given in Figure E7.

The difference in the flow patterns between water/2-pentanol and water/MIBK systems was also discussed in the work of Desir et al.192 As shown in Figure E5, at

O/A=1, when the total flow rate is below 0.8 mL/min, both systems exhibit slug flow pattern and transition into slug-droplet flow starting from 1 mL/min. However, as the flow rate further rises, the water/2-pentanol system transitions into annular flow, while the water/MIBK system first transitions into parallel flow and then into irregular flow at

3 mL/min. For the water/MIBK system at O/A=1, the flow patterns were also plotted against the residence times used in our reactions, as shown in Figure E6b. Taking

Figures E6a and E6b together, it can be observed that the flow patterns in the two

159 organic/water systems are the same for residence times of 24 s and higher, but different for residence times of 6 s and 12 s.

Besides the differences in the flow pattern and mass transfer caused by different organic solvents, the small amount of organic solvent dissolved in the aqueous phase may also affect the reaction kinetics to different extents. The respective solubility limits of 2-pentanol and MIBK in water at 25 °C are 45 g/L (4.5 wt %) and 19 g/L (1.9 wt %), and these values are expected to increase significantly at the reaction temperature of

200 °C. Several works193-196 have shown that homogeneous mixtures of organic solvents with water could affect the conversion of sugars and the selectivity of products by affecting the reactant’s surrounding environment, the catalyst, as well as the transition state. Vasudevan et al.196 modeled glucose dissolved in dimethyl sulfoxide

(DMSO)/water, THF/water and dimethylformamide (DMF)/water mixtures. It was found that with the presence of an organic solvent in water, the mobility of glucose molecules was reduced, forming longer-lived hydrogen bonds with decreasing water content. This was attributed to the increased hydrogen-bonding strength of glucose with water and with organic solvents in the presence of increasing amount of organic solvents.195 The authors suggested that the probability of a glucose molecule encountering another glucose molecule or an HMF molecule is therefore reduced, resulting in a reduction in intermolecular condensation reactions. Dumesic and coworkers194 conducted a fundamental study using combined experimental and Density

Functional Theory (DFT) simulations for Brønsted acid catalyzed fructose dehydration in water/organic solvent mixtures. It was found that the hydroxyl groups adjacent to the

160 sites of dehydration in fructose promote the formation of water clusters and hydrophilic domains at the dehydration site of the fructose reactant in the organic solvent mixtures with water, thereby allowing the acidic proton catalyst to interact directly with the hydroxyl groups within the same hydrophilic domains.194 The authors also proposed that co-solvents, such as DMSO and acetonitrile, alter the relative stabilities of initial states and transition states for dehydration of fructose. Therefore, it is likely that the small amounts of organic solvent dissolved in water affect the glucose chemistry, including the epimerization and humins formation steps (Figure 6.5).

6.4 Conclusions

In this work, we achieved fast glucose conversion to HMF using tandem

CrCl3/HCl catalysts in water/2-pentanol biphasic solvent mixtures using a flow microreactor. After optimizing the reaction temperature and HCl concentration, the highest ever HMF productivity of 33.6 % HMF yield/min (an eight-fold increase over existing literature) was achieved at 200 °C using 1.7 mM CrCl3 and HCl with a pH of

1.25 in the aqueous phase and an organic/aqueous volume ratio of 1:1.

The effect of varying organic/aqueous phase ratios (O/A) was also discussed.

For water/2-pentanol with O/A values ranging from 1 to 3, the glucose conversion and

HMF yields were similar until 2 min, the longest residence time studied, where the conversion exceeds 95%. For all three O/A values, the HMF yield reaches a maximum at 92 s and decreases at 2 min. These trends reveal the complexity of biphasic flow

161 systems where the HMF yield depends upon many factors such as the relative rates of reaction kinetics and mass transfer.

Lastly, the reactivity in water/2-pentanol and water/MIBK systems was compared. The two systems gave similar conversion and HMF yields, but different yields to mannose and humins. In addition to the differences in flow patterns and mass transfer, the reactivity may also be affected by the small amounts of the organic solvent dissolved in water as suggested by literature works. 193-196

162 Chapter 7

CONCLUSIONS, PERSPECTIVES, AND FUTURE RESEARCH DIRECTIONS

The first part of this dissertation provides detailed structural characterization and valorization paths of humins derived during the aqueous processing of carbohydrate biomass. The results could provide insights into designing more efficient and economically viable processes. The second part focuses on process intensification for glucose conversion to 5-hydroxymethylfurfural (HMF) in a microchannel flow reactor as another means of improving economic viability of bioprocesses. As the results and technical summaries have been provided in the previous chapters, this section is reserved to provide a holistic set of conclusions and their implications on future research directions.

7.1 Overall Summary and Key Conclusions

In this dissertation, we studied the structural characterization, formation and growth kinetics, and valorization of humins derived from the aqueous processing of fructose. In Chapter 2, we showed through room-temperature dissolution studies together with liquid chromatography-mass spectrometry (LC-MS), gel permeation chromatography (GPC) and Fourier transform infrared spectroscopy (FT-IR) characterizations that humins are complex, spatially and chemically heterogeneous materials composed of macromolecular networks weakly bonded with small, low-

163 molecular weight fragments. Among various solvent properties, the donor number was found to correlate with the amount of humins the solvent could dissolve. The implication of this finding is twofold: first, a suitable solvent can be selected for dissolving and upgrading humins for valorization; second, this suggests the presence of an electron donating-accepting (EDA) mechanism between the solvent and humins, therefore offering fundamental insights for the molecular structure of humins. The characterization work goes beyond functional group identification and deepens the current understanding of the macromolecular structure of humins. Armed with these insights, we were able to move on to the upgrading of humins using methanol, which is one of the highest donor numbers or “good” solvents and a hydrogen donor. In Chapter

4, four noble metal on carbons were first screened and Rh/C was selected due to its good performance on converting humins to light, gas chromatography (GC)-detectable products. Several classes of products were detected while the three major classes were aromatics, phenolics and esters. At 350-425 °C, H2 pressures between 10-45 bar, reaction times up to 6 h, and catalyst to humins mass ratios between 0 and 0.2, we show that the product distributions can be manipulated by changing the aforementioned reaction conditions. For the first time, we demonstrated the solvent to participate in the reaction; a 13C labeling study indicated that the humins-derived aromatics and phenolics precursors were alkylated by the -CH3 group of methanol to form the final products; and the ester precursors were esterified with the -OCH3 group of methanol.

In Chapter 3, we introduced a new method for operando growth experiments, ultra-small angle X-ray scattering (USAXS) measurements. We found that the

164 predominant route of humins growth is through HMF, rather than fructose. Porod analysis showed the humins had rough, dense surfaces that grow more compact as growth continues, findings which are consistent with previous scanning electronic microscopy (SEM) results in literature. 27, 110 The time evolution of the number of particles revealed a competition between the formation of new particles from soluble oligomers and aggregation of existing ones and percipitation. This aggregation continues until after all of the monomers, i.e. HMF, has been consumed. Building upon the schematic proposed in the FT-IR and dynamic light scattering (DLS) study by

Tsilomelekis et al. 110 (Scheme 3.1), we summarize our new findings by the updated network shown in schematic in Figure 3.8 where primary particles, larger particles, soluble oligomers and aggregates were distinguished from one another and the competing pathways were indicated.

In Chapter 5, building upon our group’s prior work on fructose dehydration to

HMF using an in-house built microchannel reactor,108 we carried out glucose isomerization-dehydration to HMF using the microreactor setup in aqueous phase. Our flow reactor operated stably and gave reliable and reproducible results. Faster glucose conversion and products formation at similar selectivity were achieved, in agreement with literature where batch and flow reactor performances were compared.94 Compared to existing literature on the single-phase glucose reaction in flow reactors, we achieved

-1 the highest ever HMF productivity of 15.2% min using a tandem CrCl3/HCl catalyst at

200 °C. The activity of the CrCl3 catalyst was found to be first increasing, and then decreasing with preheating time, revealing an interplay between Cr3+ hydrolysis to the

165 2+ catalytically active [Cr(H2O)5(OH)] ion and its subsequent decomposition to soluble

Cr oligomers and Cr(OH)3 solids. This was confirmed by both reactivity studies and

UV-Vis measurements on CrCl3 solutions preheated at different time intervals. Adding

HCl could suppress the formation of soluble oligomers and Cr(OH)3, but the

Brønsted/Lewis acid ratio should be tuned to maximize the concentration of the

2+ catalytically active [Cr(H2O)5(OH)] for the best reactivity.

In Chapter 6, we studied the glucose isomerization-dehydration reaction in biphasic solvent systems to further improve the HMF yield and productivity. The highest ever HMF productivity of 33.6 % HMF yield/min (an eight-fold increase over existing literature) was achieved at 51.4 % yield in 92 s, at 200 °C using 1.7 mM CrCl3 and a pH of 1.25 in the aqueous phase and an 2-pentanol/water volume ratio of 1:1. For

2-pentanol/water with flow rate ratios ranging from 1 to 3, the glucose conversion and

HMF yields were similar. Lastly, we compared the reactivity in water/2-pentanol and water/MIBK systems. The two systems gave similar conversion and HMF yields, but different yields to mannose and humins.

7.2 Future Research Directions

7.2.1 Addressing Challenges in the Valorization of Humins

Characterization of humins has received recent attention. Others27, 29, 30, 57, 116, 197

198and us110, 199 have used vibrational spectroscopy, microscopy, mass spectrometry

166 (MS), chromatography and other analytical tools to understand the chemical and macromolecular structure of humins. Methods to quantify the functional groups, especially the oxygen-containing ones, have been developed for lignin7, 200-202, a similar heterogeneous polymeric material; however, such methods for characterizing humins are still lacking. An initial effort to quantify the C=O functional groups of humins has been presented in the work of Constant et al.116 Developing these methods may be laborious and require deep fundamental chemistry understanding but can lay the groundwork for understanding the structural differences of humins obtained from a variety of starting materials and reaction conditions. The knowledge of the relative amounts of the functional groups in humins can help constructing structure-reactivity relationships, which can ultimately assist in their valorization.

Isolating the oligomers of different sizes of humins using liquid chromatography

(LC) and identifying their molecular weights and structures using mass spectrometry and nuclear magnetic resonance (NMR) would have been beneficial. Further efforts may focus on the visualization of humins particles to complement the X-ray scattering studies. In this regard, SEM would be desirable, but the evaporation of the solvent inside its vacuum chamber can cause aggregation of humins particles. More intricate experimental methods which can allow the visualization of primary humins particles and their actual size, such as cryogenic transmission electron microscopy (cryo-TEM), could be explored. In general, familiarity with polymer and colloidal science fundamentals is highly encouraged.

167 A current drawback in the chemical valorization routes of humins, including pyrolysis and hydrotreatment, is the low liquid product yields (10-20 wt%), especially the yield of low molecular weight, high value chemicals. Ways to improve the economic viability of the humins HDO process include further catalyst development, and catalyst separation and recycling. So far, only single noble metal catalysts have been used for catalytic hydrotreatment. 37-39, 199To lower the cost, inexpensive HDO metal catalysts such as Ni, Co and Mo could be employed. Also, bimetallic catalysts such as Ni-Mo and Co-Mo catalysts, could be used so that two metals could act synergistically to improve the humins conversion and the selectivity of oil products.

Catalyst supports other than activated carbon, for example, metal oxides, can be exploited so that the catalyst can be recycled by filtration followed by calcination after the reaction. Another approach to toward recycling is introducing a magnetically active catalyst support. For example, Trautmann et al. 203 synthesized a Co catalyst supported on magnetically active SiO2-Fe3O4 for the direct liquefaction of low rank coals. Adding a second HDO step, which is commonly used in the upgrading of biomass pyrolysis oil, may be able to further reduce the molecular weight of the current humins oil as well as improving its storage stability. 204 At the same time, solvents compatible with the catalysts and reaction conditions should be selected. Familiarity with upgrading of lignin, which is also recalcitrant, polymeric, and rich in oxygen containing functional groups (Scheme 1.2c), with some similarities to humins, is recommended. Finally, to gauge the economic feasibility of the process, a techno-economic analysis (TEA) should be conducted on any valorization route being developed.

168

Scheme 7.1 Proposed model for the molecular structure of alkali-treated, glucose- derived humins by van Zandvoort et al. 57 based on 2D-NMR data. Reprinted with permission. Copyright (2015) Royal Society of Chemistry.

As humins are already oxygen-rich, oxidation instead of hydrogenation can also be carried out to convert humins to oxygen-rich valuable chemicals. Currently, the only work on the oxidation of humins was done by Kang et al.205 A two-step process was proposed. First, humins were treated with NaOH hydrothermally at 180 °C. During this treatment, it has been reported that humins converted to soluble polymers with aromatic backbones and carboxylic side chains as shown in Scheme 6.1.57, 206 Then, these were

° precipitated using H2SO4, and underwent wet oxidation using H2O2 at 150 C. The target product acetic acid was obtained with the highest yield of 26.2% on carbon basis. The authors proposed that wet air oxidation (WAO) method may be used to replace H2O2 to lower the materials cost. In fact, WAO has been commonly used for treatment of

169 industrial wastes and domestic sludges. This process typically operates at temperatures of 200-350 °C and pressures up to 175 bar, 207and may be considered for the treatment of NaOH solubilized humins.

7.2.2 Process Intensification for HMF Production and Extraction

As an integral part of the U.S. Rapid Advancement of Process Intensification and Development (RAPID) initiative, work is underway to connect the microreactor with a micro-separator for HMF. A challenge in our current microreactor design is the clogging of the narrow tubing caused by the formation of humins, especially at high temperatures of >180 °C and long residence times beyond a minute. For this reason, dilute glucose solutions of 1 wt% have been used as the feed that limits the mass flow of HMF produced per unit time. Two approaches may be explored to improve the HMF productivity: (1) Minimize humins formation by using biphasic solvent systems so that a more concentrated glucose solution can be used; (2) Scale out, i.e. manufacture multiple microreactors so that many streams of dilute glucose solutions can be used for reaction at the same time. First, biphasic solvent systems for glucose conversion to

HMF needs to be investigated, following our group’s work with fructose conversion to

108 HMF in water/ethyl acetate solvents. The partitioning of the CrCl3 catalyst in the organic phase should be measured and then minimized to prevent HMF degradation and side products from the solvent. Also, heterogeneous Lewis acid catalysts such as Sn- beta zeolites could be used to further improve the HMF yield since these catalysts not

170 only deliver promising HMF yields in batch systems 90 but also could be recycled and reused. The back-pressure regulator for the system needs to be rebuilt accordingly to enable the handling of slurries. Lastly, biphasic liquid flow patterns and heat and mass transfer properties in the flow microchannel reactors also need to be well characterized, as to decrease the pressure drop, improve mass transfer and facilitate product separation from the reaction mixture.186

The kinetics data generated in this dissertation could be be used for extending our group’s prior glucose conversion model developed using batch reactor data generated at temperatures equal to or less than 140 °C. 16 Fructose, mannose and HMF could be used as starting materials with CrCl3 as the catalyst. To incorporate the

Brønsted-catalyzed glucose dehydration pathway, glucose reactions with HCl catalyst at different temperatures and pH values should also be conducted.

171 REFERENCES

1. Lee, R. A.; Lavoie, J.-M., From first-to third-generation biofuels: Challenges of producing a commodity from a biomass of increasing complexity. Animal Frontiers 2013, 3 (2), 6-11. 2. Naik, S. N.; Goud, V. V.; Rout, P. K.; Dalai, A. K., Production of first and second generation biofuels: a comprehensive review. Renewable and sustainable energy reviews 2010, 14 (2), 578-597. 3. Alexandridis, P.; Ghasemi, M.; Furlani, E. P.; Tsianou, M., Solvent processing of cellulose for effective bioresource utilization. Current Opinion in Green and Sustainable Chemistry 2018, 14, 40-52. 4. Zakzeski, J.; Bruijnincx, P. C. A.; Jongerius, A. L.; Weckhuysen, B. M., The Catalytic Valorization of Lignin for the Production of Renewable Chemicals. Chemical Reviews 2010, 110 (6), 3552-3599. 5. Chatterjee, C.; Pong, F.; Sen, A., Chemical conversion pathways for carbohydrates. Green Chemistry 2015, 17 (1), 40-71. 6. Saha, B. C., Hemicellulose bioconversion. Journal of Industrial Microbiology and Biotechnology 2003, 30 (5), 279-291. 7. Lin, S. Y.; Dence, C. W., Methods in lignin chemistry. Springer Science & Business Media: 2012. 8. Murzin, D. Y.; Salmi, T., Catalysis for Lignocellulosic Biomass Processing: Methodological Aspects. Catalysis Letters 2012, 142 (6), 676-689. 9. Mukasekuru, M. R.; Hu, J. G.; Zhao, X. Q.; Sun, F. F.; Pascal, K.; Ren, H. Y.; Zhang, J. H., Enhanced High-Solids Fed-Batch Enzymatic Hydrolysis of Sugar Cane Bagasse with Accessory Enzymes and Additives at Low Cellulase Loading. Acs Sustainable Chemistry & Engineering 2018, 6 (10), 12787-12796. 10. Deng, W. H.; Kennedy, J. R.; Tsilomelekis, G.; Zheng, W. Q.; Nikolakis, V., Cellulose Hydrolysis in Acidified LiBr Molten Salt Hydrate Media. Industrial & Engineering Chemistry Research 2015, 54 (19), 5226-5236. 11. Dornath, P.; Cho, H. J.; Paulsen, A.; Dauenhauer, P.; Fan, W., Efficient mechano-catalytic depolymerization of crystalline cellulose by formation of branched glucan chains. Green Chemistry 2015, 17 (2), 769-775. 12. Liu, S.; Dutta, S.; Zheng, W.; Gould, N. S.; Cheng, Z.; Xu, B.; Saha, B.; Vlachos, D. G., Catalytic hydrodeoxygenation of high carbon furylmethanes to renewable jet‐fuel ranged alkanes over a rhenium‐modified iridium catalyst. ChemSusChem 2017, 10 (16), 3225-3234. 13. Lilga, M. A.; Hallen, R. T.; Gray, M., Production of oxidized derivatives of 5- hydroxymethylfurfural (HMF). Topics in Catalysis 2010, 53 (15-18), 1264-1269.

172 14. Gilkey, M. J.; Mironenko, A. V.; Vlachos, D. G.; Xu, B., Adipic acid production via metal-free selective hydrogenolysis of biomass-derived tetrahydrofuran- 2, 5-dicarboxylic acid. ACS Catalysis 2017, 7 (10), 6619-6634. 15. Caratzoulas, S.; Davis, M. E.; Gorte, R. J.; Gounder, R.; Lobo, R. F.; Nikolakis, V.; Sandler, S. I.; Snyder, M. A.; Tsapatsis, M.; Vlachos, D. G., Challenges of and Insights into Acid-Catalyzed Transformations of Sugars. Journal of Physical Chemistry C 2014, 118 (40), 22815-22833. 16. Swift, T. D.; Bagia, C.; Choudhary, V.; Peklaris, G.; Nikolakis, V.; Vlachos, D. G., Kinetics of homogeneous Brønsted acid catalyzed fructose dehydration and 5- hydroxymethyl furfural rehydration: a combined experimental and computational study. ACS Catalysis 2013, 4 (1), 259-267. 17. Gruter, G. J. M.; Manzer, L. E., Hydroxymethylfurfural ethers from sugars or HMF and mixed alcohols. US20100058650A1: 2010. 18. Alonso, D. M.; Wettstein, S. G.; Dumesic, J. A., Gamma-valerolactone, a sustainable platform molecule derived from lignocellulosic biomass. Green Chemistry 2013, 15 (3), 584-595. 19. Pileidis, F. D.; Titirici, M. M., Levulinic Acid Biorefineries: New Challenges for Efficient Utilization of Biomass. Chemsuschem 2016, 9 (6), 562-582. 20. van Putten, R. J.; van der Waal, J. C.; de Jong, E.; Rasrendra, C. B.; Heeres, H. J.; de Vries, J. G., Hydroxymethylfurfural, A Versatile Platform Chemical Made from Renewable Resources. Chemical Reviews 2013, 113 (3), 1499-1597. 21. Rice, J. A., Humin. Soil Science 2001, 166 (11), 848-857. 22. Huang, X. M.; Koranyi, T. I.; Boot, M. D.; Hensen, E. J. M., Catalytic Depolymerization of Lignin in Supercritical Ethanol. Chemsuschem 2014, 7 (8), 2276- 2288. 23. Wang, X.; Rinaldi, R., Solvent effects on the hydrogenolysis of diphenyl ether with Raney nickel and their implications for the conversion of lignin. ChemSusChem 2012, 5 (8), 1455-1466. 24. Rasrendra, C. B.; Windt, M.; Wang, Y.; Adisasmito, S.; Makertihartha, I.; van Eck, E. R. H.; Meier, D.; Heeres, H. J., Experimental studies on the pyrolysis of humins from the acid-catalysed dehydration of C6-sugars. Journal of Analytical and Applied Pyrolysis 2013, 104, 299-307. 25. Horvat, J.; Klaić, B.; Metelko, B.; Šunjić, V., Mechanism of levulinic acid formation. Tetrahedron Letters 1985, 26 (17), 2111-2114. 26. Karwa, S.; Gajiwala, V. M.; Heltzel, J.; Patil, S. K.; Lund, C. R., Reactivity of levulinic acid during aqueous, acid-catalyzed HMF hydration. Catalysis Today 2016, 263, 16-21. 27. Patil, S. K.; Lund, C. R., Formation and growth of humins via aldol addition and condensation during acid-catalyzed conversion of 5-hydroxymethylfurfural. Energy & Fuels 2011, 25 (10), 4745-4755. 28. Sumerskii, I.; Krutov, S.; Zarubin, M. Y., Humin-like substances formed under the conditions of industrial hydrolysis of wood. Russian Journal of Applied Chemistry 2010, 83 (2), 320-327.

173 29. van Zandvoort, I.; Wang, Y.; Rasrendra, C. B.; van Eck, E. R.; Bruijnincx, P. C.; Heeres, H. J.; Weckhuysen, B. M., Formation, molecular structure, and morphology of humins in biomass conversion: influence of feedstock and processing conditions. ChemSusChem 2013, 6 (9), 1745-1758. 30. Patil, S. K.; Heltzel, J.; Lund, C. R., Comparison of structural features of humins formed catalytically from glucose, fructose, and 5- hydroxymethylfurfuraldehyde. Energy & Fuels 2012, 26 (8), 5281-5293. 31. Yang, G.; Pidko, E. A.; Hensen, E. J., Mechanism of Brønsted acid-catalyzed conversion of carbohydrates. Journal of catalysis 2012, 295, 122-132. 32. Pilath, H. M.; Nimlos, M. R.; Mittal, A.; Himmel, M. E.; Johnson, D. K., Glucose reversion reaction kinetics. Journal of agricultural and food chemistry 2010, 58 (10), 6131-6140. 33. Mija, A.; Van Der Waal, J.; Pin, J.-M.; Guigo, N.; De Jong, E. In Humins as promising material for producing sustainable polyssacharide-derived building materials, First International Conference on Bio-based Building Materials, 2015. 34. Pin, J.-M.; Guigo, N.; Mija, A.; Vincent, L.; Sbirrazzuoli, N.; van der Waal, J. C.; De Jong, E., Valorization of biorefinery side-stream products: combination of humins with polyfurfuryl alcohol for composite elaboration. ACS Sustainable Chemistry & Engineering 2014, 2 (9), 2182-2190. 35. Hoang, T. M. C.; Lefferts, L.; Seshan, K., Valorization of Humin‐Based Byproducts from Biomass Processing—A Route to Sustainable Hydrogen. ChemSusChem 2013, 6 (9), 1651-1658. 36. Hoang, T.; Van Eck, E.; Bula, W.; Gardeniers, J.; Lefferts, L.; Seshan, K., Humin based by-products from biomass processing as a potential carbonaceous source for synthesis gas production. Green chemistry 2015, 17 (2), 959-972. 37. Agarwal, S.; van Es, D.; Heeres, H. J., Catalytic pyrolysis of recalcitrant, insoluble humin byproducts from C6 sugar biorefineries. Journal of Analytical and Applied Pyrolysis 2017, 123, 134-143. 38. Wang, Y. H.; Agarwal, S.; Kloekhorst, A.; Heeres, H. J., Catalytic Hydrotreatment of Humins in Mixtures of Formic Acid/2-Propanol with Supported Ruthenium Catalysts. Chemsuschem 2016, 9 (9), 951-961. 39. Wang, Y.; Agarwal, S.; Heeres, H. J., Catalytic Liquefaction of Humin Substances from Sugar Biorefineries with Pt/C in 2-Propanol. Acs Sustainable Chemistry & Engineering 2017, 5 (1), 469-480. 40. Wang, Y.; Agarwal, S.; Tang, Z.; Heeres, H. J., Exploratory catalyst screening studies on the liquefaction of model humins from C6 sugars. Rsc Advances 2017, 7 (9), 5136-5147. 41. Pagan-Torres, Y. J.; Wang, T. F.; Gallo, J. M. R.; Shanks, B. H.; Dumesic, J. A., Production of 5-Hydroxymethylfurfural from Glucose Using a Combination of Lewis and Bronsted Acid Catalysts in Water in a Biphasic Reactor with an Alkylphenol Solvent. Acs Catalysis 2012, 2 (6), 930-934. 42. Swift, T. D.; Nguyen, H.; Anderko, A.; Nikolakis, V.; Vlachos, D. G., Tandem Lewis/Brønsted homogeneous acid catalysis: conversion of glucose to 5-

174 hydoxymethylfurfural in an aqueous chromium (III) chloride and hydrochloric acid solution. Green Chemistry 2015, 17 (10), 4725-4735. 43. Zhao, H.; Holladay, J. E.; Brown, H.; Zhang, Z. C., Metal Chlorides in Ionic Liquid Solvents Convert Sugars to 5-Hydroxymethylfurfural. Science 2007, 316 (5831), 1597-1600. 44. Choudhary, V.; Pinar, A. B.; Lobo, R. F.; Vlachos, D. G.; Sandler, S. I., Comparison of Homogeneous and Heterogeneous Catalysts for Glucose-to-Fructose Isomerization in Aqueous Media. Chemsuschem 2013, 6 (12), 2369-2376. 45. Qi, T.; He, M.-F.; Zhu, L.-F.; Lyu, Y.-J.; Yang, H.-Q.; Hu, C.-W., Cooperative Catalytic Performance of Lewis and Brønsted Acids from AlCl3 Salt in Aqueous Solution toward Glucose-to-Fructose Isomerization. The Journal of Physical Chemistry C 2019, 123 (8), 4879-4891. 46. Tang, J. Q.; Guo, X. W.; Zhu, L. F.; Hu, C. W., Mechanistic Study of Glucose- to-Fructose Isomerization in Water Catalyzed by Al(OH)(2)(aq) (+). Acs Catalysis 2015, 5 (9), 5097-5103. 47. Ennaert, T.; Van Aelst, J.; Dijkmans, J.; De Clercq, R.; Schutyser, W.; Dusselier, M.; Verboekend, D.; Sels, B. F., Potential and challenges of zeolite chemistry in the catalytic conversion of biomass. Chemical Society Reviews 2016, 45 (3), 584-611. 48. Román-Leshkov, Y.; Moliner, M.; Labinger, J. A.; Davis, M. E., Mechanism of Glucose Isomerization Using a Solid Lewis Acid Catalyst in Water. Angewandte Chemie International Edition 2010, 49 (47), 8954-8957. 49. Moliner, M.; Roman-Leshkov, Y.; Davis, M. E., Tin-containing zeolites are highly active catalysts for the isomerization of glucose in water. Proceedings of the National Academy of Sciences of the United States of America 2010, 107 (14), 6164- 6168. 50. Locas, C. P.; Yaylayan, V. A., Isotope labeling studies on the formation of 5- (hydroxymethyl)-2-furaldehyde (HMF) from sucrose by pyrolysis-GC/MS. Journal of Agricultural and Food Chemistry 2008, 56 (15), 6717-6723. 51. Antal Jr, M. J.; Mok, W. S.; Richards, G. N., Mechanism of formation of 5- (hydroxymethyl)-2-furaldehyde from D-fructose and sucrose. Carbohydrate research 1990, 199 (1), 91-109. 52. Chidambaram, M.; Bell, A. T., A two-step approach for the catalytic conversion of glucose to 2,5-dimethylfuran in ionic liquids. Green Chemistry 2010, 12 (7), 1253- 1262. 53. Harris, D. W.; Feather, M. S., STUDIES ON MECHANISM OF INTERCONVERSION OF D-GLUCOSE, D-MANNOSE, AND D-FRUCTOSE IN ACID SOLUTION. Journal of the American Chemical Society 1975, 97 (1), 178-182. 54. Kabyemela, B. M.; Adschiri, T.; Malaluan, R. M.; Arai, K., Kinetics of glucose epimerization and decomposition in subcritical and supercritical water. Industrial & Engineering Chemistry Research 1997, 36 (5), 1552-1558.

175 55. Orazov, M.; Davis, M. E., Tandem catalysis for the production of alkyl lactates from ketohexoses at moderate temperatures. Proceedings of the National Academy of Sciences of the United States of America 2015, 112 (38), 11777-11782. 56. Heltzel, J.; Patil, S. K.; Lund, C. R., Humin Formation Pathways. In Reaction Pathways and Mechanisms in Thermocatalytic Biomass Conversion II, Springer: 2016; pp 105-118. 57. van Zandvoort, I.; Koers, E. J.; Weingarth, M.; Bruijnincx, P. C.; Baldus, M.; Weckhuysen, B. M., Structural characterization of 13 C-enriched humins and alkali- treated 13 C humins by 2D solid-state NMR. Green Chemistry 2015, 17 (8), 4383-4392. 58. Choudhary, V.; Mushrif, S. H.; Ho, C.; Anderko, A.; Nikolakis, V.; Marinkovic, N. S.; Frenkel, A. I.; Sandler, S. I.; Vlachos, D. G., Insights into the Interplay of Lewis and Bronsted Acid Catalysts in Glucose and Fructose Conversion to 5-(Hydroxymethyl)furfural and Levulinic Acid in Aqueous Media. Journal of the American Chemical Society 2013, 135 (10), 3997-4006. 59. Li, H. X.; Xu, W. J.; Huang, T. Y.; Jia, S. Y.; Xu, Z. W.; Yan, P. F.; Liu, X. M.; Zhang, Z. C., Distinctive Aldose Isomerization Characteristics and the Coordination Chemistry of Metal Chlorides in 1-Butyl-3-methylimidazolium Chloride. Acs Catalysis 2014, 4 (12), 4446-4454. 60. Wang, T. F.; Glasper, J. A.; Shanks, B. H., Kinetics of glucose dehydration catalyzed by homogeneous Lewis acidic metal salts in water. Applied Catalysis a- General 2015, 498, 214-221. 61. Nguyen, H.; Nikolakis, V.; Vlachos, D. G., Mechanistic Insights into Lewis Acid Metal Salt-Catalyzed Glucose Chemistry in Aqueous Solution. Acs Catalysis 2016, 6 (3), 1497-1504. 62. Rasrendra, C. B.; Makertihartha, I.; Adisasmito, S.; Heeres, H. J., Green Chemicals from d-glucose: Systematic Studies on Catalytic Effects of Inorganic Salts on the Chemo-Selectivity and Yield in Aqueous Solutions. Topics in Catalysis 2010, 53 (15-18), 1241-1247. 63. Dutta, S.; De, S.; Alam, M. I.; Abu-Omar, M. M.; Saha, B., Direct conversion of cellulose and lignocellulosic biomass into chemicals and biofuel with metal chloride catalysts. Journal of Catalysis 2012, 288, 8-15. 64. Norton, A. M.; Nguyen, H.; Xiao, N. L.; Vlachos, D. G., Direct speciation methods to quantify catalytically active species of AlCl3 in glucose isomerization. Rsc Advances 2018, 8 (31), 17101-17109. 65. Guo, W.; Heeres, H. J.; Yue, J., Continuous Synthesis of 5- Hydroxymethylfurfural from Glucose Using a Combination of AlCl3 and HCl as Catalyst in a Biphasic Slug Flow Capillary Microreactor. Chemical Engineering Journal 2019, 122754. 66. Weingarten, R.; Cho, J.; Xing, R.; Conner Jr, W. C.; Huber, G. W., Kinetics and reaction engineering of levulinic acid production from aqueous glucose solutions. ChemSusChem 2012, 5 (7), 1280-1290.

176 67. Matsumura, Y.; Yanachi, S.; Yoshida, T., Glucose decomposition kinetics in water at 25 MPa in the temperature range of 448− 673 K. Industrial & engineering chemistry research 2006, 45 (6), 1875-1879. 68. Bienkowski, P.; Ladisch, M.; Narayan, R.; Tsao, G.; Eckert, R., Correlation of glucose (dextrose) degradation at 90 to 190 C in 0.4 to 20% acid. Chemical engineering communications 1987, 51 (1-6), 179-192. 69. Saha, B.; Abu-Omar, M. M., Advances in 5-hydroxymethylfurfural production from biomass in biphasic solvents. Green Chemistry 2014, 16 (1), 24-38. 70. Romo, J. E.; Bollar, N. V.; Zimmermann, C. J.; Wettstein, S. G., Conversion of Sugars and Biomass to Furans Using Heterogeneous Catalysts in Biphasic Solvent Systems. ChemCatChem 2018, 10 (21), 4805-4816. 71. Ramesh, P.; Kritikos, A.; Tsilomelekis, G., Effect of Metal Chlorides on Glucose Mutarotation and Possible Implications on Humin Formation. Reaction Chemistry & Engineering 2018. 72. Bermejo-Deval, R.; Orazov, M.; Gounder, R.; Hwang, S. J.; Davis, M. E., Active Sites in Sn-Beta for Glucose Isomerization to Fructose and Epimerization to Mannose. Acs Catalysis 2014, 4 (7), 2288-2297. 73. Kruger, J. S.; Nikolakis, V.; Vlachos, D. G., Aqueous-phase fructose dehydration using Brønsted acid zeolites: Catalytic activity of dissolved aluminosilicate species. Applied Catalysis A: General 2014, 469, 116-123. 74. Swift, T. D.; Nguyen, H.; Erdman, Z.; Kruger, J. S.; Nikolakis, V.; Vlachos, D. G., Tandem Lewis acid/Bronsted acid-catalyzed conversion of carbohydrates to 5- hydroxymethylfurfural using zeolite beta. Journal of Catalysis 2016, 333, 149-161. 75. Rajabbeigi, N.; Torres, A. I.; Lew, C. M.; Elyassi, B.; Ren, L. M.; Wang, Z. P.; Cho, H. J.; Fan, W.; Daoutidis, P.; Tsapatsis, M., On the kinetics of the isomerization of glucose to fructose using Sn-Beta. Chemical Engineering Science 2014, 116, 235-242. 76. Kruger, J. S.; Choudhary, V.; Nikolakis, V.; Vlachos, D. G., Elucidating the roles of zeolite H-BEA in aqueous-phase fructose dehydration and HMF rehydration. ACS Catalysis 2013, 3 (6), 1279-1291. 77. Otomo, R.; Yokoi, T.; Kondo, J. N.; Tatsumi, T., Dealuminated Beta zeolite as effective bifunctional catalyst for direct transformation of glucose to 5- hydroxymethylfurfural. Applied Catalysis a-General 2014, 470, 318-326. 78. Tong, X.; Ma, Y.; Li, Y., Biomass into chemicals: conversion of sugars to furan derivatives by catalytic processes. Applied Catalysis A: General 2010, 385 (1-2), 1-13. 79. Watanabe, M.; Aizawa, Y.; Iida, T.; Nishimura, R.; Inomata, H., Catalytic glucose and fructose conversions with TiO2 and ZrO2 in water at 473 K: Relationship between reactivity and acid–base property determined by TPD measurement. Applied Catalysis A: General 2005, 295 (2), 150-156. 80. Wang, L.; Xiao, F.-S., Nanoporous catalysts for biomass conversion. Green Chemistry 2015, 17 (1), 24-39.

177 81. Ordomsky, V. V.; van der Schaaf, J.; Schouten, J. C.; Nijhuis, T. A., Glucose Dehydration to 5‐Hydroxymethylfurfural in a Biphasic System over Solid Acid Foams. ChemSusChem 2013, 6 (9), 1697-1707. 82. Ordomsky, V. V.; van der Schaaf, J.; Schouten, J. C.; Nijhuis, T. A., Fructose Dehydration to 5‐Hydroxymethylfurfural over Solid Acid Catalysts in a Biphasic System. ChemSusChem 2012, 5 (9), 1812-1819. 83. Wang, Y.; Zhu, L. W.; Zhang, Y.; Cui, H. Y.; Yi, W. M.; Song, F.; Zhao, P. P.; Sun, X. Y.; Xie, Y. J.; Wang, L. H.; Li, Z. H., AlNb/SBA-15 Catalysts with Tunable Lewis and Bronsted Acidic Sites for Glucose Conversion to HMF. Chemistryselect 2018, 3 (12), 3555-3560. 84. Xia, H. A.; Hu, H.; Xu, S. Q.; Xiao, K. H.; Zuo, S. L., Catalytic conversion of glucose to 5-hydroxymethyfural over Fe/beta zeolites with extra-framework isolated Fe species in a biphasic reaction system. Biomass & Bioenergy 2018, 108, 426-432. 85. Hu, S. Q.; Zhang, Z. F.; Song, J. L.; Zhou, Y. X.; Han, B. X., Efficient conversion of glucose into 5-hydroxymethylfurfural catalyzed by a common Lewis acid SnCl4 in an ionic liquid. Green Chemistry 2009, 11 (11), 1746-1749. 86. Stahlberg, T.; Rodriguez-Rodriguez, S.; Fristrup, P.; Riisager, A., Metal-Free Dehydration of Glucose to 5-(Hydroxymethyl)furfural in Ionic Liquids with Boric Acid as a Promoter. Chemistry-a European Journal 2011, 17 (5), 1456-1464. 87. Tsilomelekis, G.; Josephson, T. R.; Nikolakis, V.; Caratzoulas, S., Origin of 5- Hydroxymethylfurfural Stability in Water/Dimethyl Sulfoxide Mixtures. Chemsuschem 2014, 7 (1), 117-126. 88. Wang, T. F.; Pagan-Torres, Y. J.; Combs, E. J.; Dumesic, J. A.; Shanks, B. H., Water-Compatible Lewis Acid-Catalyzed Conversion of Carbohydrates to 5- Hydroxymethylfurfural in a Biphasic Solvent System. Topics in Catalysis 2012, 55 (7- 10), 657-662. 89. Roman-Leshkov, Y.; Dumesic, J. A., Solvent Effects on Fructose Dehydration to 5-Hydroxymethylfurfural in Biphasic Systems Saturated with Inorganic Salts. Topics in Catalysis 2009, 52 (3), 297-303. 90. Nikolla, E.; Roman-Leshkov, Y.; Moliner, M.; Davis, M. E., "One-Pot" Synthesis of 5-(Hydroxymethyl)furfural from Carbohydrates using Tin-Beta Zeolite. Acs Catalysis 2011, 1 (4), 408-410. 91. Roman-Leshkov, Y.; Chheda, J. N.; Dumesic, J. A., Phase modifiers promote efficient production of hydroxymethylfurfural from fructose. Science 2006, 312 (5782), 1933-1937. 92. Binder, J. B.; Raines, R. T., Simple Chemical Transformation of Lignocellulosic Biomass into Furans for Fuels and Chemicals. Journal of the American Chemical Society 2009, 131 (5), 1979-1985. 93. Nguyen, H.; DeJaco, R. F.; Mittal, N.; Siepmann, J. I.; Tsapatsis, M.; Snyder, M. A.; Fan, W.; Saha, B.; Vlachos, D. G., A Review of Biorefinery Separations for Bioproduct Production via Thermocatalytic Processing. Annual Review of Chemical and Biomolecular Engineering, Vol 8 2017, 8, 115-137.

178 94. Tuercke, T.; Panic, S.; Loebbecke, S., Microreactor Process for the Optimized Synthesis of 5-Hydroxymethylfurfural: A Promising Building Block Obtained by Catalytic Dehydration of Fructose. Chemical Engineering & Technology 2009, 32 (11), 1815-1822. 95. Kashid, M. N.; Agar, D. W., Hydrodynamics of liquid-liquid slug flow capillary microreactor: Flow regimes, slug size and pressure drop. Chemical Engineering Journal 2007, 131 (1-3), 1-13. 96. Hommes, A.; Heeres, H. J.; Yue, J., Catalytic Transformation of Biomass Derivatives to Value‐Added Chemicals and Fuels in Continuous Flow Microreactors. ChemCatChem. 97. Muranaka, Y.; Nakagawa, H.; Masaki, R.; Maki, T.; Mae, K., Continuous 5- Hydroxymethylfurfural Production from Monosaccharides in a Microreactor. Industrial & Engineering Chemistry Research 2017, 56 (39), 10998-11005. 98. Salak Asghari, F.; Yoshida, H., Acid-catalyzed production of 5-hydroxymethyl furfural from D-fructose in subcritical water. Industrial & Engineering Chemistry Research 2006, 45 (7), 2163-2173. 99. Asghari, F. S.; Yoshida, H., Kinetics of the decomposition of fructose catalyzed by hydrochloric acid in subcritical water: formation of 5-hydroxymethylfurfural, levulinic, and formic acids. Industrial & Engineering Chemistry Research 2007, 46 (23), 7703-7710. 100. Kabyemela, B. M.; Adschiri, T.; Malaluan, R. M.; Arai, K., Glucose and fructose decomposition in subcritical and supercritical water: detailed reaction pathway, mechanisms, and kinetics. Industrial & Engineering Chemistry Research 1999, 38 (8), 2888-2895. 101. Aida, T. M.; Tajima, K.; Watanabe, M.; Saito, Y.; Kuroda, K.; Nonaka, T.; Hattori, H.; Smith Jr, R. L.; Arai, K., Reactions of d-fructose in water at temperatures up to 400 C and pressures up to 100 MPa. The Journal of Supercritical Fluids 2007, 42 (1), 110-119. 102. Aida, T. M.; Sato, Y.; Watanabe, M.; Tajima, K.; Nonaka, T.; Hattori, H.; Arai, K., Dehydration of d-glucose in high temperature water at pressures up to 80 MPa. The Journal of supercritical fluids 2007, 40 (3), 381-388. 103. Qi, J.; Xiuyang, L., Kinetics of non-catalyzed decomposition of glucose in high- temperature liquid water. Chinese Journal of Chemical Engineering 2008, 16 (6), 890- 894. 104. Zhang, X.; Hewetson, B. B.; Mosier, N. S., Kinetics of maleic acid and aluminum chloride catalyzed dehydration and degradation of glucose. Energy & Fuels 2015, 29 (4), 2387-2393. 105. Watanabe, M.; Aizawa, Y.; Iida, T.; Aida, T. M.; Levy, C.; Sue, K.; Inomata, H., Glucose reactions with acid and base catalysts in hot compressed water at 473 K. Carbohydrate research 2005, 340 (12), 1925-1930. 106. Siqueira, B.; Silva, M.; Moraes, C., Synthesis of HMF from glucose in aqueous medium using niobium and titanium oxides. Brazilian Journal of Petroleum and Gas 2013, 7 (2).

179 107. Hansen, T. S.; Woodley, J. M.; Riisager, A., Efficient microwave-assisted synthesis of 5-hydroxymethylfurfural from concentrated aqueous fructose. Carbohydrate research 2009, 344 (18), 2568-2572. 108. Desir, P.; Saha, B.; Vlachos, D. G., Ultrafast flow chemistry for the acid- catalyzed conversion of fructose. Energy & Environmental Science 2019. 109. Reiche, S.; Kowalew, N.; Schloegl, R., Influence of Synthesis pH and Oxidative Strength of the Catalyzing Acid on the Morphology and Chemical Structure of Hydrothermal Carbon. Chemphyschem 2015, 16 (3), 579-587. 110. Tsilomelekis, G.; Orella, M. J.; Lin, Z.; Cheng, Z.; Zheng, W.; Nikolakis, V.; Vlachos, D. G., Molecular structure, morphology and growth mechanisms and rates of 5-hydroxymethyl furfural (HMF) derived humins. Green Chemistry 2016, 18 (7), 1983- 1993. 111. van Putten, R. J.; Soetedjo, J. N.; Pidko, E. A.; van der Waal, J. C.; Hensen, E. J.; de Jong, E.; Heeres, H. J., Dehydration of Different Ketoses and Aldoses to 5‐ Hydroxymethylfurfural. ChemSusChem 2013, 6 (9), 1681-1687. 112. Hansen, C. M., The Three Dimensional Solubility Parameter—Key to Paint Component Affinities: II and III—II. dyes, Emulsifiers, Mutual Solubility and Compatibility, and Pigments. J. Paint Technology 1967, 39 (511), 505-510. 113. Avolio, R.; Bonadies, I.; Capitani, D.; Errico, M.; Gentile, G.; Avella, M., A multitechnique approach to assess the effect of ball milling on cellulose. Carbohydrate Polymers 2012, 87 (1), 265-273. 114. Given, P. H.; Marzec, A.; Barton, W. A.; Lynch, L. J.; Gerstein, B. C., The concept of a mobile or molecular phase within the macromolecular network of coals: a debate. Fuel 1986, 65 (2), 155-163. 115. Shinn, J. H., From coal to single-stage and two-stage products: a reactive model of coal structure. Fuel 1984, 63 (9), 1187-1196. 116. Constant, S.; Lancefield, C. S.; Weckhuysen, B. M.; Bruijnincx, P. C., Quantification and Classification of Carbonyls in Industrial Humins and Lignins by 19F NMR. ACS Sustainable Chemistry & Engineering 2016. 117. Shimadzu Separation Conditions: 1. Fundamentals of a First-Choice Mobile Phase. https://www.shimadzu.com/an/lcms/support/lib/lctalk/60/60lab.html (accessed Feb 17). 118. Herod, A. A., Limitations of mass spectrometric methods for the characterization of polydisperse materials. Rapid Communications in Mass Spectrometry 2010, 24 (17), 2507-2519. 119. Prokai, L.; Simonsick, W. J., Electrospray ionization mass spectrometry coupled with size‐exclusion chromatography. Rapid communications in mass spectrometry 1993, 7 (9), 853-856. 120. Sevilla, M.; Fuertes, A. B., The production of carbon materials by hydrothermal carbonization of cellulose. Carbon 2009, 47 (9), 2281-2289. 121. Sevilla, M.; Fuertes, A. B., Chemical and structural properties of carbonaceous products obtained by hydrothermal carbonization of saccharides. Chemistry–A European Journal 2009, 15 (16), 4195-4203.

180 122. Klein, R.; Meakin, P., Universality in colloid aggregation. Nature 1989, 339 (3). 123. Jurkiewicz, A.; Marzec, A.; Piślewski, N., Molecular structure of bituminous coal studied with pulse nuclear magnetic resonance. Fuel 1982, 61 (7), 647-650. 124. Aharoni, S.; Edwards, S., Gels of rigid polyamide networks. Macromolecules 1989, 22 (8), 3361-3374. 125. Vahrman, M., The smaller molecules derived from coal and their significance. Fuel 1970, 49 (1), 5-16. 126. Painter, P. C.; Park, Y.; Coleman, M. M., Thermodynamics of coal solutions. Energy & fuels 1988, 2 (5), 693-702. 127. Bozell, J. J.; Moens, L.; Elliott, D.; Wang, Y.; Neuenscwander, G.; Fitzpatrick, S.; Bilski, R.; Jarnefeld, J., Production of levulinic acid and use as a platform chemical for derived products. Resources, conservation and recycling 2000, 28 (3), 227-239. 128. Rasrendra, C.; Windt, M.; Wang, Y.; Adisasmito, S.; Makertihartha, I.; van Eck, E.; Meier, D.; Heeres, H., Experimental studies on the pyrolysis of humins from the acid-catalysed dehydration of C6-sugars. Journal of Analytical and Applied Pyrolysis 2013, 104, 299-307. 129. Ilavsky, J.; Jemian, P. R.; Allen, A. J.; Zhang, F.; Levine, L. E.; Long, G. G., Ultra-small-angle X-ray scattering at the Advanced Photon Source. Journal of Applied Crystallography 2009, 42 (3), 469-479. 130. Sivia, D. S., Elementary scattering theory: for X-ray and neutron users. Oxford University Press: 2011. 131. Ilavsky, J.; Jemian, P. R., Irena: tool suite for modeling and analysis of small- angle scattering. Journal of Applied Crystallography 2009, 42 (2), 347-353. 132. Hammouda, B. Probing Nanoscale Structures-the SANS Toolbox. https://www.ncnr.nist.gov/staff/hammouda/the_SANS_toolbox.pdf (accessed March 4th). 133. Kuster, B.; Der Van Steen, H., Preparation of 5‐Hydroxymethylfurfural Part I. Dehydration of Fructose in a Continuous Stirred Tank Reactor. Starch‐Stärke 1977, 29 (3), 99-103. 134. Girisuta, B.; Janssen, L.; Heeres, H., A kinetic study on the decomposition of 5- hydroxymethylfurfural into levulinic acid. Green Chemistry 2006, 8 (8), 701-709. 135. Carothers, W. H., Polymers and polyfunctionality. Transactions of the Faraday Society 1936, 32, 39-49. 136. Kuster, B. F.; Temmink, H. M., The influence of pH and weak-acid anions on the dehydration of D-fructose. Carbohydrate research 1977, 54 (2), 185-191. 137. Kuster, B. F.; van der Baan, H. S., The influence of the initial and catalyst concentrations on the dehydration of D-fructose. Carbohydrate research 1977, 54 (2), 165-176. 138. Kimura, H.; Nakahara, M.; Matubayasi, N., Solvent effect on pathways and mechanisms for D-fructose conversion to 5-hydroxymethyl-2-furaldehyde: in situ 13C NMR study. The Journal of Physical Chemistry A 2013, 117 (10), 2102-2113.

181 139. Enslow, K. R.; Bell, A. T., The role of metal halides in enhancing the dehydration of xylose to furfural. ChemCatChem 2015, 7 (3), 479-489. 140. Ramesh, P.; Kritikos, A.; Tsilomelekis, G., Effect of metal chlorides on glucose mutarotation and possible implications on humin formation. Reaction Chemistry & Engineering 2019, 4 (2), 273-277. 141. Galkin, K. I.; Krivodaeva, E. A.; Romashov, L. V.; Zalesskiy, S. S.; Kachala, V. V.; Burykina, J. V.; Ananikov, V. P., Critical Influence of 5‐ Hydroxymethylfurfural Aging and Decomposition on the Utility of Biomass Conversion in Organic Synthesis. Angewandte Chemie International Edition 2016, 55 (29), 8338-8342. 142. Boukari, H.; Lin, J. S.; Harris, M. T., Small-angle X-ray scattering study of the formation of colloidal silica particles from alkoxides: Primary particles or not? Journal of Colloid and Interface Science 1997, 194 (2), 311-318. 143. Green, D. L.; Lin, J. S.; Lam, Y. F.; Hu, M. Z. C.; Schaefer, D. W.; Harris, M. T., Size, volume fraction, and nucleation of Stober silica nanoparticles. Journal of Colloid and Interface Science 2003, 266 (2), 346-358. 144. Bogush, G. H.; Zukoski, C. F., UNIFORM SILICA PARTICLE- PRECIPITATION - AN AGGREGATIVE GROWTH-MODEL. Journal of Colloid and Interface Science 1991, 142 (1), 19-34. 145. Matsoukas, T.; Gulari, E., MONOMER-ADDITION GROWTH WITH A SLOW INITIATION STEP - A GROWTH-MODEL FOR SILICA PARTICLES FROM ALKOXIDES. Journal of Colloid and Interface Science 1989, 132 (1), 13-21. 146. Boukari, H.; Lin, J. S.; Harris, M. T., Probing the dynamics of the silica nanostructure formation and growth by SAXS. Chemistry of Materials 1997, 9 (11), 2376-2384. 147. Van Blaaderen, A.; Van Geest, J.; Vrij, A., Monodisperse colloidal silica spheres from tetraalkoxysilanes: particle formation and growth mechanism. Journal of colloid and interface science 1992, 154 (2), 481-501. 148. Werpy, T.; Petersen, G.; Aden, A.; Bozell, J.; Holladay, J.; White, J.; Manheim, A.; Eliot, D.; Lasure, L.; Jones, S. Top value added chemicals from biomass. Volume 1-Results of screening for potential candidates from sugars and synthesis gas; Department of Energy Washington DC: 2004. 149. Chheda, J. N.; Román-Leshkov, Y.; Dumesic, J. A., Production of 5- hydroxymethylfurfural and furfural by dehydration of biomass-derived mono-and poly- saccharides. Green Chemistry 2007, 9 (4), 342-350. 150. Panagiotopoulou, P.; Vlachos, D. G., Liquid phase catalytic transfer hydrogenation of furfural over a Ru/C catalyst. Applied Catalysis A: General 2014, 480, 17-24. 151. Jae, J.; Zheng, W.; Lobo, R. F.; Vlachos, D. G., Production of dimethylfuran from hydroxymethylfurfural through catalytic transfer hydrogenation with ruthenium supported on carbon. ChemSusChem 2013, 6 (7), 1158-1162.

182 152. Cheng, Z.; Everhart, J.; Tsilomelekis, G.; Nikolakis, V.; Saha, B.; Vlachos, D., Structural Analysis of Humins Formed in the Brønsted-Catalyzed Dehydration of Fructose. Green Chemistry 2018, 20 (5), 997-1006. 153. Hu, J.; Shen, D. K.; Wu, S. L.; Zhang, H. Y.; Xiao, R., Composition Analysis of Organosolv Lignin and Its Catalytic Solvolysis in Supercritical Alcohol. Energy & Fuels 2014, 28 (7), 4260-4266. 154. Sitthisa, S.; Resasco, D. E., Hydrodeoxygenation of Furfural Over Supported Metal Catalysts: A Comparative Study of Cu, Pd and Ni. Catalysis Letters 2011, 141 (6), 784-791. 155. Guvenatam, B.; Kursun, O.; Heeres, E. H. J.; Pidko, E. A.; Hensen, E. J. M., Hydrodeoxygenation of mono- and dimeric lignin model compounds on noble metal catalysts. Catalysis Today 2014, 233, 83-91. 156. Vorotnikov, V.; Mpourmpakis, G.; Vlachos, D. G., DFT study of furfural conversion to furan, furfuryl alcohol, and 2-methylfuran on Pd (111). ACS Catal. 2012, 2 (12), 2496-2504. 157. Vorotnikov, V.; Vlachos, D. G., Group additivity and modified linear scaling relations for estimating surface thermochemistry on transition metal surfaces: application to furanics. J. Phys. Chem. C 2015, 119 (19), 10417-10426. 158. Wang, S.; Vorotnikov, V.; Sutton, J. E.; Vlachos, D. G., Brønsted–Evans– Polanyi and transition state scaling relations of furan derivatives on Pd (111) and their relation to those of small molecules. ACS Catal. 2014, 4 (2), 604-612. 159. Panagiotopoulou, P.; Martin, N.; Vlachos, D. G., Effect of hydrogen donor on liquid phase catalytic transfer hydrogenation of furfural over a Ru/RuO2/C catalyst. Journal of Molecular Catalysis a-Chemical 2014, 392, 223-228. 160. Mironenko, A. V.; Vlachos, D. G., Conjugation-Driven "Reverse Mars-van Krevelen"-Type Radical Mechanism for Low-Temperature C-O Bond Activation. Journal of the American Chemical Society 2016, 138 (26), 8104-8113. 161. Kepp, K. P., A Quantitative Scale of Oxophilicity and Thiophilicity. Inorganic Chemistry 2016, 55 (18), 9461-9470. 162. Sanders, E. B.; Goldsmith, A. I.; Seeman, J. I., A model that distinguishes the pyrolysis of D-glucose, D-fructose, and sucrose from that of cellulose. Application to the understanding of cigarette smoke formation. Journal of analytical and applied pyrolysis 2003, 66 (1), 29-50. 163. Ma, X. L.; Ma, R.; Hao, W. Y.; Chen, M. M.; Iran, F.; Cui, K.; Tian, Y.; Li, Y. D., Common Pathways in Ethanolysis of Kraft Lignin to Platform Chemicals over Molybdenum-Based Catalysts. Acs Catalysis 2015, 5 (8), 4803-4813. 164. Vasireddy, S.; Morreale, B.; Cugini, A.; Song, C.; Spivey, J. J., Clean liquid fuels from direct coal liquefaction: chemistry, catalysis, technological status and challenges. Energy & Environmental Science 2011, 4 (2), 311-345. 165. Thring, R. W.; Katikaneni, S. P.; Bakhshi, N. N., The production of gasoline range hydrocarbons from Alcell® lignin using HZSM-5 catalyst. Fuel Processing Technology 2000, 62 (1), 17-30.

183 166. Tajbl, D.; Simons, J.; Carberry, J., Heterogeneous Catalysis in Continuous Stirred Tank Reactor. Industrial & Engineering Chemistry Fundamentals 1966, 5 (2), 171-175. 167. Parsell, T.; Yohe, S.; Degenstein, J.; Jarrell, T.; Klein, I.; Gencer, E.; Hewetson, B.; Hurt, M.; Kim, J. I.; Choudhari, H.; Saha, B.; Meilan, R.; Mosier, N.; Ribeiro, F.; Delgass, W. N.; Chapple, C.; Kenttamaa, H. I.; Agrawal, R.; Abu-Omar, M. M., A synergistic biorefinery based on catalytic conversion of lignin prior to cellulose starting from lignocellulosic biomass. Green Chemistry 2015, 17 (3), 1492- 1499. 168. Dumesic, J. A.; Rom, Y.; Chheda, J. N., Catalytic process for producing furan derivatives in a biphasic reactor. US7572925B2: 2009. 169. Ehrfeld, W.; Hessel, V.; Haverkamp, V., Microreactors. Ullmann's encyclopedia of industrial chemistry 2000. 170. Mason, B. P.; Price, K. E.; Steinbacher, J. L.; Bogdan, A. R.; McQuade, D. T., Greener approaches to organic synthesis using microreactor technology. Chemical reviews 2007, 107 (6), 2300-2318. 171. Jensen, K. F., Microreaction engineering—is small better? Chemical Engineering Science 2001, 56 (2), 293-303. 172. Stroock, A. D.; Dertinger, S. K. W.; Ajdari, A.; Mezić, I.; Stone, H. A.; Whitesides, G. M., Chaotic Mixer for Microchannels. Science 2002, 295 (5555), 647- 651. 173. Cook, K. J.; Fan, Y.; Hassan, I., Mixing Evaluation of a Passive Scaled-Up Serpentine Micromixer With Slanted Grooves. Journal of Fluids Engineering 2013, 135 (8). 174. Onjia, A. E.; Milonjić, S. K.; Čokeša, D.; Čomor, M.; Miljević, N., Characterization of colloidal chromia particles obtained by forced hydrolysis. Materials research bulletin 2003, 38 (8), 1329-1339. 175. Stuenzi, H.; Marty, W., Early stages of the hydrolysis of chromium (III) in aqueous solution. 1. Characterization of a tetrameric species. Inorganic Chemistry 1983, 22 (15), 2145-2150. 176. Bell, A.; Matijević, E., Complex chemistry of hydrous chromium (III) oxide sol formation. Journal of Inorganic and Nuclear Chemistry 1975, 37 (4), 907-912. 177. Avena, M. J.; Giacomelli, C. E.; De Pauli, C. P., Formation of Cr (III) hydroxides from chrome alum solutions: 1. Precipitation of active chromium hydroxide. Journal of Colloid and interface science 1996, 180 (2), 428-435. 178. Rotzinger, F. P.; Stuenzi, H.; Marty, W., Early stages of the hydrolysis of chromium (III) in aqueous solution. 3. Kinetics of dimerization of the deprotonated aqua ion. Inorganic Chemistry 1986, 25 (4), 489-495. 179. Tsuchida, R., Absorption spectra of co-ordination compounds. I. Bulletin of the Chemical Society of Japan 1938, 13 (5), 388-400. 180. Elving, P. J.; Zemel, B., Absorption in the ultraviolet and visible regions of chloroaquochromium (III) ions in acid media. Journal of the American Chemical Society 1957, 79 (6), 1281-1285.

184 181. Jia, S.; Liu, K.; Xu, Z.; Yan, P.; Xu, W.; Liu, X.; Zhang, Z. C., Reaction media dominated product selectivity in the isomerization of glucose by chromium trichloride: From aqueous to non-aqueous systems. Catalysis Today 2014, 234, 83-90. 182. Lennartson, A., The colours of chromium. Nature Chemistry 2014, 6, 942. 183. De, S.; Dutta, S.; Saha, B., Microwave assisted conversion of carbohydrates and biopolymers to 5-hydroxymethylfurfural with aluminium chloride catalyst in water. Green Chemistry 2011, 13 (10), 2859-2868. 184. Roman-Leshkov, Y.; Barrett, C. J.; Liu, Z. Y.; Dumesic, J. A., Production of dimethylfuran for liquid fuels from biomass-derived carbohydrates. Nature 2007, 447 (7147), 982-U5. 185. Gomes, F.; Pereira, L.; Ribeiro, N.; Souza, M., Production of 5- hydroxymethylfurfural (HMF) via fructose dehydration: Effect of solvent and salting- out. Brazilian Journal of Chemical Engineering 2015, 32 (1), 119-126. 186. Dessimoz, A.-L.; Cavin, L.; Renken, A.; Kiwi-Minsker, L., Liquid–liquid two- phase flow patterns and mass transfer characteristics in rectangular glass microreactors. Chemical Engineering Science 2008, 63 (16), 4035-4044. 187. Sotowa, K.-I., Fluid behavior and mass transport characteristics of gas–liquid and liquid–liquid flows in microchannels. Journal of Chemical Engineering of Japan 2014, 47 (3), 213-224. 188. Ghaini, A.; Mescher, A.; Agar, D. W., Hydrodynamic studies of liquid–liquid slug flows in circular microchannels. Chemical engineering science 2011, 66 (6), 1168- 1178. 189. Shimanouchi, T.; Kataoka, Y.; Tanifuji, T.; Kimura, Y.; Fujioka, S.; Terasaka, K., Chemical conversion and liquid–liquid extraction of 5‐ hydroxymethylfurfural from fructose by slug flow microreactor. AIChE Journal 2016, 62 (6), 2135-2143. 190. Lueckgen, J.; Vanoye, L.; Philippe, R.; Eternot, M.; Fongarland, P.; de Bellefon, C.; Favre-Réguillon, A., Simple and selective conversion of fructose into HMF using extractive-reaction process in microreactor. Journal of Flow Chemistry 2018, 8 (1), 3-9. 191. Brasholz, M.; Von Kaenel, K.; Hornung, C. H.; Saubern, S.; Tsanaktsidis, J., Highly efficient dehydration of carbohydrates to 5-(chloromethyl) furfural (CMF), 5- (hydroxymethyl) furfural (HMF) and levulinic acid by biphasic continuous flow processing. Green Chemistry 2011, 13 (5), 1114-1117. 192. Desir, P.; Chen, T.-Y.; Bracconi, M.; Saha, B.; Maestri, M.; Vlachos, D. G., Experiments and computations of microfluidic liquid–liquid flow patterns. Reaction Chemistry & Engineering 2019. 193. Yang, Y.; Hu, C.; Abu-Omar, M. M., Conversion of glucose into furans in the presence of AlCl3 in an ethanol–water solvent system. Bioresource technology 2012, 116, 190-194. 194. Mellmer, M. A.; Sanpitakseree, C.; Demir, B.; Bai, P.; Ma, K.; Neurock, M.; Dumesic, J. A., Solvent-enabled control of reactivity for liquid-phase reactions of biomass-derived compounds. Nature Catalysis 2018, 1 (3), 199.

185 195. Shuai, L.; Luterbacher, J., Organic solvent effects in biomass conversion reactions. ChemSusChem 2016, 9 (2), 133-155. 196. Vasudevan, V.; Mushrif, S. H., Insights into the solvation of glucose in water, dimethyl sulfoxide (DMSO), tetrahydrofuran (THF) and N, N-dimethylformamide (DMF) and its possible implications on the conversion of glucose to platform chemicals. Rsc Advances 2015, 5 (27), 20756-20763. 197. Filiciotto, L.; de Miguel, G.; Balu, A. M.; Romero, A. A.; van der Waal, J. C.; Luque, R., Towards the photophysical studies of humin by-products. Chemical Communications 2017, 53 (52), 7015-7017. 198. Filiciotto, L.; Balu, A. M.; Romero, A. A.; Angelici, C.; van der Waal, J. C.; Luque, R., Reconstruction of humins formation mechanism from decomposition products: A GC-MS study based on catalytic continuous flow depolymerizations. Molecular Catalysis 2019, 479, 110564. 199. Cheng, Z.; Saha, B.; Vlachos, D. G., Catalytic Hydrotreatment of Humins to Bio‐Oil in Methanol over Supported Metal Catalysts. ChemSusChem 2018, 11 (20), 3609-3617. 200. Calvo-Flores, F.; Dobado, J.; Isac-García, J.; Martín-MartíNez, F., Functional and Spectroscopic Characterization of Lignins. Lignin and Lignans as Renewable Raw Materials: Chemistry, Technology and Applications 2015, 145-187. 201. Li, J.; Hu, H.-C.; Chai, X.-S., Rapid method for determination of carbonyl groups in lignin compounds by headspace gas chromatography. Journal of Chromatography A 2015, 1404, 39-43. 202. Zakis, G. F. s. e., Functional analysis of lignins and their derivatives. Tappi Press: 1994. 203. Trautmann, M.; Lang, S.; Traa, Y., Direct liquefaction of lower-rank coals and biocoals with magnetically separable catalysts as a sustainable route to fuels. Fuel 2015, 151, 102-109. 204. Busetto, L.; Fabbri, D.; Mazzoni, R.; Salmi, M.; Torri, C.; Zanotti, V., Application of the Shvo catalyst in homogeneous hydrogenation of bio-oil obtained from pyrolysis of white poplar: new mild upgrading conditions. Fuel 2011, 90 (3), 1197-1207. 205. Kang, S.; Zhang, G.; Yang, Q.; Tu, J.; Guo, X.; Qin, F. G.; Xu, Y., A new technology for utilization of biomass hydrolysis residual humins for acetic acid production. BioResources 2016, 11 (4), 9496-9505. 206. van Zandvoort, I.; van Eck, E. R.; de Peinder, P.; Heeres, H. J.; Bruijnincx, P. C.; Weckhuysen, B. M., Full, reactive solubilization of humin byproducts by alkaline treatment and characterization of the alkali-treated humins formed. Acs sustainable chemistry & engineering 2015, 3 (3), 533-543. 207. Debellefontaine, H.; Foussard, J. N., Wet air oxidation for the treatment of industrial wastes. Chemical aspects, reactor design and industrial applications in Europe. Waste management 2000, 20 (1), 15-25.

186

Appendix A

SUPPLEMENTARY INFORMATION FOR CHAPTER 2

STRUCTURAL ANALYSIS OF HUMINS FORMED IN THE BRØNSTED-

CATALYZED DEHYDRATION OF FRUCTOSE

16 14 12 10 8 6 4

2 SolubilizedHumins

Concentration (mg/mL) Concentration 0 0 20 40 60 80 100 120 Time (h) Figure A. 1 Solubilized humins concentration in ACN as a function of time. 0.45 g humins were mixed with 3 mL of ACN and stirred at room temperature at 1000 rpm. Three replicates were done for each time point.

187 10

8

6

4

2 SolubilizedHumins

Concentration (mg/mL) Concentration 0 0.01 0.1 1 10 Filter Size (m)

Figure A. 2 Solubilized humins concentration in ACN as a function of the pore size of the filter. 0.45 g humins were mixed with 3mL of ACN and stirred at room temperature at 1000 rpm. Vacuum filtration couldn’t be used with 0.02, 0.2 and 0.45 μm pore sized filters, so a different procedure was used to investigate the effect of filter pore size. The slurry was filtered using syringe filters of pore sizes 0.02, 0.2, 0.45, and 11 μm. A known amount of filtrate was transferred into a pre-weighed glass vial and put into the vacuum oven at room temperature for 24 h. Subsequently, the vial containing the residue was weighed again. The concentration of solubilized humins was calculated as the mass of the residue divided by the volume of the solvent put into the vial. Three replicates were done for each condition. The concentration of solubilized humins using the 11 μm filter determined using this method is 1-2 mg/mL less than that determined by vacuum filtering the slurry and weighing the undissolved humins.

50 70 (a) (b) 40 60 50 30 40

20 30 Response Response 20 10 10 0 0 4 6 8 10 12 14 16 0 5 10 15 20 25 Time (min) Time (min) Figure A. 3 HPLC chromatogram for humins in (a) acetonitrile (ACN) solution and (b) methanol solution. The diode array detector (DAD) was set at 254 nm.

188 1.8x105 1.6x104 (a) (b) 5 1.6x10 1.2x104 1.4x105

8.0x103 Counts 1.2x105 Counts 4.0x103 1.0x105 0.0 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 Time (min) Time (min) Figure A. 4 Total ion chromatogram of humins in acetonitrile solution. (a) 10-300 m/z. (b) 301-500 m/z.

5x105 1.0x105 (a) (b) 4 4x105 8.0x10 6.0x104 3x105

4 Counts Counts 4.0x10 2x105 2.0x104

1x105 0.0 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40 45 50 Time (min) Time (min) Figure A. 5 Total ion chromatogram of humins in methanol solution. (a) 10-600 m/z. (b) 601-1200 m/z.

1.4x107 2.5x106 (a) (b) 1.2x107 A=1.971x106C+2.962x106 FCA R2=0.999 2.0x106 A=4.828x106C+1.095x105 1.0x107 R2=0.991 FCA 6

1.5x10 8.0x106

Area Area 1.0x106 6 A=1.949x106C+1.350x106 HMF 6.0x10 2 R =0.992 A=3.729x106C+1.473x104 HMF 5 2 6 5.0x10 R =0.999 4.0x10

0.0 2.0x106 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0 1 2 3 4 5 6 Concentration (mM) Concentration (mM) Figure A. 6 Calibration plots of 2-(4-formylphenoxymethyl) furan-3-carboxylic acid (FCA) and HMF (a) 0-0.5 mM and (b) 1-5 mM in acetonitrile.

189 40 40 (a) DMSO 35 (b) DMSO 30 THF 30 THF 25 Acetone 20 Acetone MeOH 20 15 ACN MeOH ACN 10 10 DCM

5 DCM Water Solubilized Humins Solubilized

Solubilized Humins Solubilized Cyclohexane Concentration (mg/mL) Concentration 0 Cyclohexane (mg/mL) Concentration 0 Water -5 0 10 20 30 40 50 60 7.5 8.0 8.5 9.0  Acceptor Number d 40 40 (d) (c) DMSO DMSO 30 30 THF THF Acetone 20 20 Acetone MeOH MeOH ACN ACN 10 10 DCM DCM Water Water

0 Cyclohexane Humins Solubilized 0

Solubilized Humins Solubilized Cyclohexane

Concentration (mg/mL) Concentration Concentration (mg/mL) Concentration 0 2 4 6 8 10 -5 0 5 10 15 20 25 p  h Figure A. 7 Solubilized humins concentration versus solvent properties. (a) Acceptor number (AN), (b)~(d) Hansen solubility parameters δd, δp, and δh. No clear correlation between the solubilized humins concentration and AN, δd, δp, or δh was found.

190

Bottom Absorbance Absorbance Top fraction

3500 3000 2500 2000 1500 1000 -1 Wavenumber (cm ) Figure A. 8 FTIR spectra of humins floating on top and settling to the bottom when placed in water. The nearly identical spectra indicate these fractions have the same chemical structure but are inhomogeneous in density.

20 Stage 1 5 6.0x10 Stage 2 15 Stage 6 4.0x105 10 2.0x105

5 Signal MS (Cnts) SolubilizedHumins Concentration (mg/mL) Concentration 0 1 2 3 4 5 6 0.0 5 10 15 20 25 Stage Time (min) Figure A. 9 (a) Concentration of the methanol solubilized humins at stages 1-6. 0.45 g of insoluble humins obtained from the previous stage was mixed with 3 mL of methanol and stirred at 25 oC for 24 h at 1000 rpm. The number of replicates was 10 at the first stage and decreased to 3 at the 6th stage. (b) Total ion chromatogram (TIC) of methanol solubilized humins for stages 1, 2, and 6 at 10-600 m/z.

191

Table A. 1 Concentrations of dominant species of solubilized humins detected by LC- MS. ACN solvent Methanol solvent Mass (m/z) Concentration (mM) Mass (m/z) Concentration (mM) 99 2.80 99 1.42 109 0.54 127 1.76 127 6.48 141 16.4 217 0.24 195 0.35 225 0.66 185 1.87 235 3.88 239 0.42 252 1.54 235 0.18 207 0.61 249 0.30 151 0.05 295 0.69 219 0.14 349 1.23 191 0.28 219 0.46 315 0.31 303 0.41 343 0.37 340 0.48 247 0.19 679 0.04 241 0.11 701 0.03

Table A. 2 Concentrations of prevalent, MeOH soluble humins masses of Soxhlet extracted and manually washed humins. Soxhlet Extracted Mass (m/z) Manually Washed Humins (mM) Humins (mM) 99 1.42 0.19 127 1.76 0.22 141 16.36 0.94 195 0.35 0.11 185 1.87 0.13 239 0.42 0.13 235 0.18 0.07 249 0.30 0.09 295 0.69 0.11 349 1.23 0.18 219 0.46 0.12 303 0.41 0.44 340 0.48 0.26 679 0.04 0.05

192 701 0.03 0.04

Table A. 3 Mole percentage (%) of each mass in the total amount of monitored species, determined by LC-MS analysis. m/z Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6 99 15.37 4.49 6.92 10.56 13.19 16.82 109 2.96 4.18 3.91 4.56 4.88 3.99 127 35.58 8.19 14.67 19.49 17.22 26.79 217 1.32 3.55 4.00 3.69 4.31 3.23 225 3.61 5.35 4.76 4.06 5.53 4.58 235 21.30 8.59 16.01 8.39 7.99 6.13 252 8.45 37.77 21.56 16.98 12.04 7.73 207 3.38 3.28 3.68 4.96 6.04 4.92 151 0.29 0.79 1.03 1.49 1.75 2.46 219 0.79 7.60 3.61 3.61 3.00 0.00 191 1.56 5.38 7.76 11.11 11.93 12.35 315 1.70 3.31 3.80 3.25 3.35 2.97 343 2.04 3.89 3.98 3.30 3.30 3.05 247 1.04 2.22 2.66 3.06 3.77 3.71 241 0.62 1.41 1.65 1.48 1.70 1.27

60 THF Acetone 50 MeOH ACN 40 30

20 RID Intensity RID 10 0 100 1000 10000 M (Da) p Figure A. 10 Refractive index detector (RID) intensity as a function of molecular weight for the acetone, ACN, methanol, and THF solubilized humins. Here the humins did not undergo Soxhlet extraction before dissolution in organic solvents.

193 60 (a) Manually Washed 60 (b) Manually Washed Soxhlet Extracted Soxhlet Extracted 50 50 40 40 30 30

20 20

RID Intensity RID RID Intensity RID 10 10 0 0 100 1000 10000 100 1000 10000 M (Da) Mp (Da) p

60 (c) Manually Washed 60 (d) Manually Washed Soxhlet Extracted Soxhlet Extracted 50 50 40 40 30 30

20 20

RID Intensity RID RID Intensity RID 10 10 0 0 100 1000 10000 100 1000 10000 M (Da) M (Da) p p Figure A. 11 Molecular weight distribution for the (a) acetone, (b) acetonitrile, (c) methanol, and (d) THF-soluble portions of humins that were Soxhlet extracted (red) and only manually washed with DI water before dissolution (black).

194 2925 cm-1 Stage 6

Stage 5

Stage 4

Stage 3

Stage 2

Absorbance Stage 1

Initial

3100 3000 2900 2800 Wavenumber (cm-1)

Figure A. 12 Zoomed in of the high wavenumber region of Figure 2.4a showing the 2925 cm-1 (C-H stretching) peak for the initial humins and the decreasing intensity of this peak in the insoluble humins from dissolution stages 1~6.

195 (a)

Stage 6 Stage 5 Stage 4 Stage 3

Absorbance Stage 2 Stage 1 Initial

3500 3000 1500 1000 Wavenumber (cm-1)

(b)

Stage 6 Stage 5 Stage 4 Stage 3

Stage 2 Absorbance Stage 1 Initial

3500 3000 1500 1000 Wavenumber (cm-1)

Figure A. 13 FTIR spectra of initial humins, and (a) the methanol solubilized humins at each stage of dissolution, and (b) insoluble humins remaining after each stage of dissolution.

196 Appendix B

SUPPLEMENTARY INFORMATION FOR CHAPTER 3 GROWTH

KINETICS OF HUMINS STUDIED VIA X-RAY SCATTERING

1000 Model Raw data

) 100 -1

10

Intensity (cm Intensity 1

0.1 1E-4 1E-3 0.01 0.1 q (Å-1)

Figure B. 1 Raw and fitted data in a typical experiment. Data shown here were taken from a 10 wt% pH=0 fructose solution heated at 80 °C at 80 min of reaction time. A Gaussian size distribution model is a good fit for the data.

(a) (b) 5x108 1x109

80 oC 70 oC )

o ) o 3 85 C 3 80 C 8 8x108 4x10 90 oC 95 oC 6x108 3x108

4x108 2x108

2x108 # of Particles (1/cm Particles of # 1x108 (1/cm Particles of # 0 20 40 60 80 100 120 0 100 200 300 400 500 600 Time (min) Time (min)

197 Figure B. 2 Number of particles per volume of solution as a function of time for (a) 10 wt % initial fructose solution at various temperatures indicated and (b) 50 wt % initial fructose at solution 70 and 80 °C. Reaction pH=0.

(a) 0.7 (b) 30 Fructose 0.6 HMF FA 0.5 LA 20 Balance (humins) 0.4 Fructose 0.3 FA LA 10 0.2 HMF

0.1

Concentration (mol/L) Concentration Conversion or Yield (%) Yield or Conversion 0.0 0 0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200 Time (min) Time (min) Figure B. 3 (a) Fructose and FA, LA, HMF concentrations and (b) calculated fructose conversion and product yields as functions of time. Reaction conditions: 10 wt% fructose, pH=0, 80 °C, no stirring.

(a) 1.0 (b) 45 HMF HMF 40 FA FA 0.8 LA 35 LA Balance (humins) 30 0.6 25 20 0.4 15

0.2 10

Concentration (mol/L) Concentration 5 Conversion or Yield (%) Yield or Conversion 0.0 0 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 Time (h) Time (h) Figure B. 4 (a) HMF, FA, and LA concentrations and (b) calculated HMF conversion and product yields as a function of time. Reaction conditions: 10 wt% HMF, pH=0, 70 °C, no stirring.

198 Appendix C

SUPPLEMENTARY INFORMATION FOR CHAPTER 4

CATALYTIC HYDROTREATMENT OF HUMINS TO BIO-OIL IN

METHANOL OVER SUPPORTED METAL CATALYSTS

Table C. 1 Detailed reaction conditions used for hydrotreatment of humins in methanol. # Catalysts Catalyst (g) T (°C) H2 (bar) Time (h) 1 Ru/C 0.055 350 30 3

2 Pt/C 0.055 350 30 3

3 Pd/C 0.055 350 30 3

4 Rh/C 0 350 30 3

5 Rh/C 0.014 350 30 3

6 Rh/C 0.055 350 30 3

7 Rh/C 0.11 350 30 3

8 Rh/C 0.055 375 30 3

9 Rh/C 0.055 400 30 3

10 Rh/C 0.055 400 30 3

11 Rh/C 0.055 425 30 3

12 Rh/C 0.055 400 30 0a

13 Rh/C 0.055 400 30 1.5

14 Rh/C 0.055 400 30 6

199 15 Rh/C 0.055 400 0 3

16 Rh/C 0.055 400 10 3

17 Rh/C 0.055 400 20 3

18 Rh/C 0.055 400 45 3

Table C. 2 Oxyophilicity of metals used as catalysts and the O/C, H/C ratio of the corresponding humins oil. Metal Oxophilicity Humins Oil-O/C Ratio Humins Oil-H/C Ratio Ru 0.4 0.219 0.965 Rh 0.3 0.154 1.35 Pt 0.1 0.276 1.29 Pd 0.0 0.108 1.35

Table C. 3 Reproducibility results in the yield of detected products from humins hydrotreatment in methanol over Rh/C. Run 1 Run 2

Solids before reaction (humins, Rh/C) (g) 0.60 0.60

Gas before reaction (pure H ) (g) 0.22 0.22 2

Liquid before reaction (pure MeOH) (g) 7.9 7.9

Solids after reaction (g) 0.18 0.19

Humins conversion (%) 77 75

Gas after reaction (g) 4.71 4.38

Humin oil, MeOH and water after reaction (g) 3.02 3.09

Total humins oil yield (% by weight) 21 19.7

GC-detectable oil yield (% by weight) 11.5 10.1

200 Total mass before reaction (g) 8.72 8.72

Total mass after reaction (g) 7.92 7.67

Mass balance closure (wt%) 90.9 87.9

Reaction conditions: 1:10 Rh/C catalyst to humins mass ratio, 400 °C, 30 bar H2, 3 h.

4x105

4 3x105

2x105 2 11 Counts 10 9 12 8 5 1 1x10 3 5 6 7

0 5 10 15 20 25 Time (min)

Figure C. 1 GC-MS total ion chromatogram of Humins-oil from the Ru/C catalyzed reaction. Reaction conditions: 1:10 Ru/C catalyst to humins mass ratio, 350 °C, 30 bar initial H2, 3 h. Some representative compounds are labeled.

201 (a) 3x105

2x105

Counts 1x105

0 5 10 15 20 25 30 Time (min)

(b) 3x105

2x105

1x105 Counts

0 5 10 15 20 25 30 Time (min)

Figure C. 2 GC-MS total ion chromatograms of compounds obtained in blank experiments using (a) Ru/C and (b) Rh/C catalysts without humins. The results show no products. Reaction conditions: 1:10 catalyst to humins mass ratio, 350 °C and 30 bar initial H2 for 3h.

202 10

8

6

4

2

GC-detectable Yield (wt%) Yield GC-detectable 0 Ru/C Pt/C Pd/C Rh/C Catalyst

■ Aromatics ■ Phenols ■ Esters ■ Ketones ■ Alcohols ■ Alkanes ■ Furans

Figure C. 3 Comparison of GC-detectable oil yield over four commercial catalysts. Reaction conditions: 1:10 catalyst to humins mass ratio, 350 °C, 3h with 30 bar of initial H2.

Table C. 4 GC-MS identified compounds in humins-oil. Compound CAS Name Compound CAS Name 1 554-12-1 Methyl propionate 44 90-00-6 Phenol, 2-ethyl- 1H-Indene, 2,3-dihydro-4- 2 71-43-2 Benzene 45 824-22-6 methyl- 3 64-19-7 Acetic acid 46 105-67-9 Phenol, 2,4-dimethyl- Benzene, (1-methyl-2- 4 71-36-3 1-Butanol 47 propen-1-yl)- 5 625-86-5 2,5-Dimethylfuran 48 576-26-1 Phenol, 2,6-dimethyl- 6 6032-29-7 2-Pentanol 49 526-75-0 Phenol, 2,3-dimethyl- Butanoic acid, 7 623-42-7 methyl ester 50 95-65-8 Phenol, 3,4-dimethyl 1,3,5- 56253- Benzene, [(1E)-2-methyl-1- 8 544-25-2 Cycloheptatriene 51 64-6 buten-1-yl]- 9 71-41-0 1-Pentanol 52 697-82-5 Phenol, 2,3,5-trimerhyl- 10 589-38-8 3-Hexanone 53 644-35-9 Phenol, 2-propyl- 1687-64- 11 591-78-6 2-Hexanone 54 5 Phenol, 2-ethyl-6-methyl-

203 1687-61- 12 626-93-7 2-Hexanol 55 2 Phenol, 2-ethyl-5-methyl- 2416-94- 13 624-24-8 Methyl valerate 56 6 Phenol, 2,3,6-trimethyl- Cyclopentanone, 2- 14 1120-72-5 methyl- 57 875-85-4 Phenol, 3,4-diethyl Cyclopentanol, 3- 3855-26- 15 18729-48-1 methyl- 58 3 Phenol, 2-ethyl-4-methyl- Cyclopentanone, 2- 16 1120-72-5 methyl- 59 N/A Phenol, 3-ethyl-5-methyl 17 100-41-4 Ethylbenzene 60 527-54-8 Phenol, 3,4,5-trimethyl- 18 108-38-3 m-Xylene 61 527-60-6 Phenol, 2,4,6-trimethyl- Benzene, (1,2-dimethyl-1- 19 22104-79-6 2-Nonen-1-ol 62 769-57-3 propenyl)- 6682-06- 1H-Indene, 2,3-dihydro-4,7- 20 543-49-7 2-Heptanol 63 0 trimethyl- 1-Ethyl-3- 21 3728-55-0 methylcyclohexane 64 90-12-0 1-methylnaphthalene Dihydrofuran- 3520-52- 22 96-48-0 2(3H)-one 65 3 2-methyl-6-propylphenol Hexanoic acid, 23 106-70-7 methyl ester 66 91-57-6 2-methylnaphthalene Cyclohexane, 2613-76- 1H-Indene, 2,3-dihydro- 24 1678-92-8 propyl- 67 5 1,1,3-trimethyl- Cyclopentane, 22531- naphthalene, 6-ethyl-1,2,3,4- 25 2040-95-1 butyl- 68 20-0 tetrahydro- Cyclohexanol, 3- 26 591-23-1 methyl- 69 937-30-4 Ethanone, 1-(4-ethylphenyl)- 27 502-42-1 Cycloheptanone 70 527-35-5 Phenol, 2,3,5,6-tetramethyl- 2(3H)-Furanone, 1,4-benzenediol, 2,6- 28 108-29-2 dihydro-5-methyl- 71 654-42-2 dimethyl- Benzene, 1-ethyl-2- Acenaphthylene,1,2,2,a,3,4,5- 29 611-14-3 methyl 72 480-72-8 hexahydro- 30 589-63-9 4-Octanone 73 571-61-9 1,5-dimethylnaphthalene 1731-86- Undecanoic acid, methyl 31 108-95-2 Phenol 74 8 ester 32 111-13-7 2-Octanone 75 581-42-0 2,6-dimethylnaphthalene 1,2,3- or 1,2,4- 2,3,5-trimethyl-1,4- 33 trimethylbenzene 76 700-13-0 benzenediol Heptanoic acid, 34 106-73-0 methyl ester 77 581-40-8 2,3-dimethylnaphthalene Butanedioic acid, 35 106-65-0 dimethyl ester 78 575-41-7 1,3-dimethylnaphthalene 1000195- 5-Isopropylidene-4,6- 36 95-48-7 Phenol, 2-methyl- 79 14-9 dimethylnona-3,6,8-trien-2-ol Ethanone, 1-(2- methyl-1- 2131-42- 37 3168-90-9 cyclopenten-1-yl)- 80 2 1,4,6-trimethylnapthalene

204 38 106-44-5 p-Cresol 81 102-25-0 1,3,5-triethylbenzene 2131-41- 39 1120-21-4 Undecane 82 1 1,4,5-trimethylnapthalene Cyclohexene, 1- 2245-38- 40 3282-53-9 butyl- 83 7 1,6,7-trimethylnaphthalene Phenol, 2,6- 41 576-26-1 dimethyl- 84 610-48-0 Anthracene, 1-methyl- Benzofuran, 2- 2444-68- 42 4265-25-2 methyl- 85 0 Anthracene, 9-ethenyl Octanoic acid, 43 111-11-5 methyl ester 86 129-00-0 Pyrene Reaction conditions: 1:10 Rh/C catalyst to humins mass ratio, 400 °C, 30 bar H2, 3 h

Table C. 5 Methanol degradation as a function of temperature.

350 °C (lowest T used) 425 °C (highest T used) MeOH input (mL) 10 10 MeOH after reaction (mL) 6.92 2.83 MeOH conversion (%) 30.8 71.7 Yield-CO (% by C) 14.78 41.57

Yield-CO2 (% by C) 1.70 1.91

Yield-CH4 (% by C) 5.02 11.15

Yield-C2H6 (% by C) 0.01 0.06 Yield-other gaseous HCs (% 0.36 0.64 by C) Carbon balance closure (%) 91.1 83.7

Reaction conditions: No humins, 0.0545g Rh/C catalyst, 30 bar H2, 3 h.

205 50 MeOH Solution Humins Oil 40

30

20 RID Intensity RID 10

0 100 1000 10000 M p Figure C. 4 Molecular weight distribution of species in methanol solubilized humins and humins oil. Humins were solubilized in methanol at 25 °C (red line). Humins o hydrotreatment was conducted at 400 C, 30 bar H2 using 1:10 Rh/C catalyst to humins mass ratio for 3 h (blue line).

Table C. 6 Polydispersity indices (PDI) of humins oil obtained from hydrotreatment using Rh/C at different reaction conditions. Catalyst (g) T (°C) H2 Pressure Time (h) PDI of Humins Oil (bar) 0 350 30 3 1.58

0.014 350 30 3 1.60

0.055 350 30 3 1.51

0.11 350 30 3 1.32

0.055 375 30 3 1.46

0.055 400 30 3 1.53

0.055 425 30 3 1.51

0.055 400 30 1.5 1.64

206 0.055 400 30 6 1.53

0.055 400 10 3 1.44

0.055 400 20 3 1.31

0.055 400 45 3 1.50

Table C. 7 GC-MS identified compounds derived from IPA in humins oil. Compound CAS # Name Compound CAS # Name 1 67-64-1 Acetone 17 526-73-8 1,2,3-trimethylbenzene 2 108-10-1 Methyl isobutyl ketone 18 538-93-2 i-Butylbenzene 3 591-78-6 2-hexanone 19 95-36-3 1,2,4-trimethylbenzene 4 141-79-7 4-methyl-3-peten-2-one 20 873-94-9 3,3,5- trimethylcyclohexanone 5 2213-23-2 2,4-dimethylheptane 21 1074-43- 1-methyl-3-propylbenzene 7 6 1072-05-5 2,6-dimethylheptane 22 N/A 1-ethyl-2,5- dimethylbenzene 7 3073-66-3 1,1,3-trimethylcyclohexane 23 108-39-4 3-methylphenol 8 110-12-3 5-methyl-2-hexanone 24 N/A 1-Isobutyl-4-methylbenzene 9 108-38-3 m-Xylene 25 4706-90- 1,3-dimethyl-5- 5 isopropylbenezene 10 N/A 1,1,3,5- 26 95-93-2 1,2,4,5-tetramethylbenzene tetramethylcyclohexane 11 106-42-3 p-Xylene 27 N/A 1-(i-propyl)-2,4- dimethylbenzene 12 626-33-5 2-methyl-4-heptanone 28 108-68-9 3,5-dimethylphenol 13 98-82-8 Isopropylbenzene 29 N/A 1-(i-butyl)-2,5- dimethylbenzene 14 622-96-8 1-ethyl-4-methylbenzene 30 4175-53- 5

1,3-Dimethylindan

207 15 108-67-8 1,3,5-trimethylbenzene 31 2416-94- 2,3,6-trimethylphenol 6 16 19549-80- 4,6-dimethyl-2-heptanone 5

Reaction conditions: No humins, 0.0545g Rh/C catalyst, 30 bar N2, 400 °C, 3 h

Table C. 8 GC-MS identified compounds in humins oil obtained only from humins during hydrotreatment in IPA. Compound CAS # Name Compound CAS # Name 1 6032-29- 2-pentanol 24 3228-02-2 3-methyl-4-(i-propyl)phenol 7 2 1757-42- 3-methylcyclopentanone 25 1470-94-6 5-indanol 2 3 108-38-3 m-Xylene 26 488-87-9 2,5-dimethyl-1,3- benzenediol 4 123-19-3 4-heptanone 27 1127-76-0 1-ethylnaphthalene 5 108-95-2 Phenol 28 582-16-1 2,7-dimethylnaphthalene 6 95-48-7 2-methylphenol 29 N/A 6-methyl-4-indanol 7 N/A 1-(2-Methyl-1-cyclopenten- 30 581-40-8 2,3-dimethylnaphthalene 1-yl) ethanone

8 576-26-1 2,6-dimethylphenol 31 9 4265-25- 2-methylbenzofuran 32 35946-91- 2,5-diisopropyphenol 2 9 10 90-00-6 2-ethylphenol 33 527-55-9 4,5-dimethyl-1,3- benzenediol 11 344-07-7 1-ethenyl-4-ethylbenzene 34 2245-38-7 1,6,7-trimethylnaphthalene 12 105-67-9 2,4-dimethylphenol 35 2131-42-2 1,4,6-trimethylnaphthalene 13 7344-34- 4-methylindene 36 N/A 7-Methyl-1-naphthol 5 14 526-75-0 2,3-dimethylphenol 37 613-33-2 Chapter 84,4'- Dimethylbiphenyl

15 1687-64- 2-ethyl-6-methylphenol 38 N/A 2,4,6-Trimethylbiphenyl 5 16 91-20-3 Naphthalene 39 N/A 2-Methyl-1-naphthol 17 N/A 1,2-dimethylindene 40 85-01-8 Phenanthrene 18 644-35-9 2-propylphenol 41 N/A 19 99-89-8 4-isopropylphenol 42 832-69-9 1-methylphenanthrene 20 2416-94- 2,3,6-trimethylphenol 43 779-02-2 9-methylanthracene 6 21 3855-26- 2-ethyl-4-methylphenol 44 N/A Chapter 95,6-Dihydro- 3 4H-benzo[de]anthracene

208 22 90-12-0 1-methylnaphthalene 45 2381-21-7 3-methylpyrene 23 608-25-3 2-methyl-1,3-benzenediol

Reaction conditions: No humins, 0.0545g Rh/C catalyst, 30 bar N2, 400 °C, 3 h.

(a) 100 (b) Conversion 8 Total Oil Yield 80 6

60 4

40 2

Conversion (%) Conversion 20

0 GC-detectable yield (wt%) yield GC-detectable 0 Humins-derived IPA and Humins-derived products, in IPA humins-derived products, in MeOH IPA solvent MeOH solvent products, in IPA

■ Aromatics ■ Phenols ■ Esters ■ Ketones ■ Alcohols ■ Alkanes ■ Furans

Figure C. 5 (a) Conversion of humins and total oil yield and (b) breakdown GC- detectable class of compounds and their yield in the oil obtained from hydrotreatment in IPA and MeOH without initial H2. Reaction Conditions: 400 °C, 3 h, 30 bar of initial N2 pressure, and Rh/C to humins mass ratio of 1: 10.

209 Appendix D

SUPPLEMENTARY INFORMATION FOR CHAPTER 5 GLUCOSE

CONVERSION TO 5-HYDRXYMETHYL FURFURAL AT SHORT

RESIDENCE TIMES IN A MICROREACTOR

140

120

C) o 100

80

60 Temperature ( Temperature 40 Microreactor Batch reactor in oil bath 20 0 200 400 600 800

Time (s) Figure D. 1 The heat-up profiles for microreactor and batch glass vial reactors heated in oil bath. The temperature set point for both reactors is 140 °C. For the microreactor, a thermocouple was inserted into a tee at the reactor outlet to measure the effluent temperature. For the glass vial reactor, a thermocouple was tightly inserted through the top hole of a capped 5mL vial filled with 2mL DI water to measure the temperature.

210 0.04

0.03 Glucose Mannose Fructose 0.02 HMF

0.01 Concentration (mol/L) Concentration

0.00 0 10 20 30 40 50 60 70 80 Time on Stream (min)

Figure D. 2 Concentrations of glucose, fructose, and mannose as a function of time-on- stream. Time zero was defined as when the reactor outlet temperature reached ±0.1 °C of the set point. Reaction conditions: 1 wt% glucose and 1.7 mM CrCl3 at inlet, 140 °C, 5 min residence time.

211 60 Glucose Mannose 50 Fructose HMF FA 40 LA

30

20

10 Conversion or Yield (%) Yield or Conversion 0 0 10 20 30 40 50 60 Contact Time (min)

Figure D. 3 Glucose conversion and product yields vs. residence time. The error bars were obtained from two repeat experiments. Reaction conditions: 1 wt% glucose and 1.7 mM CrCl3 at inlet, 140 °C.

(a) 20 Glucose (b) Glucose Mannose 8 Mannose Fructose 15 Fructose HMF HMF 6

10 4

5 2

Conversion or Yield (%) Yield or Conversion Conversion or Yield (%) Yield or Conversion 0 0 0 10 20 30 40 50 0 10 20 30 40 50 Contact Time (min) Contact Time (min) Figure D. 4. Control experiments: (a) 1 wt% glucose with HCl catalyst at pH=1.25 and 160 °C (no chromium chloride); (b) 1 wt% glucose with no catalyst at 160 °C.

212 Table D1. UV-Vis Peak Positions for Preheated CrCl3 Solutions. Preheating time Peak 2 Position (nm) Peak 3 Position (nm) 0 min 440 630 0.6 min 413 578 5 min 421 586 15 min 414 581 1 h 421 589 4 h 423 593 24 h 424 594

(a) 10 Before filtration Permeate 8 Retentate, diluted 5X

6

4 Absorbance 2

0 200 250 300 350 Wavelength (nm) (b) 0.20 Before filtration Permeate 0.16 Retentate, diluted 5X

0.12

0.08 Absorbance 0.04

0.00 300 400 500 600 700 800 Wavelength (nm) Figure D. 5 UV-Vis spectra of the 3.4 mM of CrCl3 solution heated for 24 h at 140 °C, and the permeate and retentate from ultracentrifugation at (a) 200-350 nm region and (b) 350-800 nm region. The ultracentrifugation is done using an Amicon centrifuge tube with a 10000 kDa molecular weight cutoff filter insert at 10000 rpm for 10 min. After ultracentrifugation, the retentate was diluted 5 times using deionized water to gather enough volume for UV-Vis analysis.

213 Sample Calculation for Obtaining Glucose Reaction Rate The reaction rate of glucose follows first-order kinetics and has the following rate expression: 푟퐺퐿푈 = −푘[퐺퐿푈] where rGLU denotes the rate, k is the rate constant, and [GLU] is the glucose concentration at any time. At conversions less than 15%, one can assume the rate is constant. Therefore, a linear fit for the glucose concentration vs. time data at <15% conversion, as shown in Figure D5 below, provides the rate constant:

0.060

C =-3.49x10-4t+0.0546 0.055 GLU R2=0.98

0.050

0.045

0.040

Glucose Concentration (mol/L) Glucose Concentration 0 5 10 15 20 25

Residence time (s) o Figure D. 6 Glucose concentration (CGLU) vs. time (t) at 160 C using a 1.7 mM CrCl3 catalyst which was not preheated. A linear fit (red line) was used to obtain the glucose reaction rate of 3.49×10-4 mol/L-s.

214 Appendix E

SUPPLEMENTARY INFORMATION FOR CHAPTER 6 GLUCOSE

CONVERSION TO 5-HYDROXYMETHYL FURFURAL IN BIPHASIC

SOLVENT MIXTURES IN A MICROREACTOR

Table E. 1 Total Flow Rates at Different Residence Times and Organic/Aqueous Flow Rate Ratios (O/A) Residence Time (s) Total Flow Rate (mL/min) O/A=1 O/A=2 O/A=3 6 4 6 8 12 2 3 4 24 1 1.5 2 30 0.8 1.2 1.6 40 0.6 0.8 1.2 60 0.4 0.6 0.8 92 0.26 0.39 0.52 120 0.2 0.3 0.4

60 Glucose Fructose 50 HMF

40

30

20

10 Conversion or Yield (%) Yield or Conversion 0 0 50 100 150 200 250 300 350 Residence Time (s)

Figure E. 1 Glucose conversion and selected product (fructose and HMF) yields vs. residence time. The error bars were obtained from two repeat experiments. Reaction

215 conditions: 1 wt% glucose and 1.7 mM CrCl3 in the aqueous phase, and 2-pentanol in equal flow rate as the aqueous phase, 160 °C.

(a) (b) 14 14 MAN MAN FRU FRU 12 LG 12 LG FA FA 10 LA 10 LA 8 8

6 6 Yield (%) Yield Yield (%) Yield 4 4 2 2 0 0 0 50 100 150 200 250 300 0 50 100 150 200 Residence Time (s) Residence Time (s) (c) 14 MAN FRU 12 LG FA 10 LA 8 6

Yield (%) Yield 4 2 0 0 20 40 60 80 100 120 140 Residence Time (s)

Figure E. 2 Carbon yields of minor products (mannose, fructose, levoglucosan, formic acid, levulinic acid) vs. residence time at temperatures of (a) 160, (b) 180 and (c) 200 °C. Aqueous feed composition: 1 wt % glucose, 1.7mM CrCl3 and 0.056M HCl (pH=1.25). 2-pentanol was used as the organic solvent in 1:1 flow rate ratio to the aqueous phase.

216 100 GLU MAN 80 FRU FA LA 60 HMF

40

20 Conversion or Yield (%) Yield or Conversion 0 0 100 200 300 400

Time (min) Figure E. 3 Glucose conversion and product yields vs. time in a batch reactor. Reaction conditions: 10 wt% glucose, 20 wt% NaCl, 0.1M of HCl and 18.6 mM CrCl3 in the aqueous phase, and tetrahydrofuran (THF) in 2:1 volume ratio to the aqueous phase, 140 °C.

(a) 8 (b) 8 MAN Yield MAN Yield FRU Yield FRU Yield LG Yield LG Yield FA Yield FA Yield 6 LA Yield 6 LA Yield

4 4

Yield (%) Yield (%) Yield 2 2

0 0 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 Residence Time (s) Residence Time (s) Figure E. 4 Yields of minor products (mannose, fructose, levoglucosan, formic acid, levulinic acid) vs. residence time at 200 °C using 2-pentanol/aqueous ratios (O/A) of (a) 2 and (b) 3. Aqueous feed composition: 1 wt % glucose, 1.7mM CrCl3 and 0.056M HCl (pH=1.25).

217

Figure E. 5 Maps of flow pattern vs. the org/aq (v/v) ratio and the total volumetric flow rate using various organic solvents: (a) 2-pentanol and (b)MIBK. Adapted with permission from Desir et al.1 Royal Society of Chemistry 2019. (a) (b) 3 3

Annular Irregular Slug-Droplet Parallel 2 Droplet 2 Droplet

Slug Slug MIBK/Water MIBK/Water (v/v)

2-Pentanol/Water (v/v) 2-Pentanol/Water 1 1

0 20 40 60 80 100 120 0 20 40 60 80 100 Residence Time (s) Residence Time (s)

Figure E. 6 Maps of flow pattern at the residence times used in our experiments for (a) 2-pentanol/water solvent system with flow rate ratios of 1, 2 and 3 (b) MIBK/water system at flow rate ratios of 1:1.

218 10 MAN FRU LG 8 FA LA

6

4 Yield (%) Yield

2

0 0 20 40 60 80 100 120 Residence Time (s)

Figure E. 7 Carbon yields of minor products (mannose, fructose, levoglucosan, formic acid, levulinic acid) vs. residence time at 200 °C using MIBK/aqueous ratios of 1:1. Aqueous feed composition: 1 wt % glucose, 1.7mM CrCl3 and 0.056M HCl (pH=1.25).

REFERENCES 1. Desir, P.; Chen, T.-Y.; Bracconi, M.; Saha, B.; Maestri, M.; Vlachos, D. G., Experiments and computations of microfluidic liquid–liquid flow patterns. Reaction Chemistry & Engineering 2019.

219 Appendix F

ADDITIONAL DATA FOR SINGLE, AQUEOUS PHASE GLUCOSE

CONVERSION IN A MICROREACTOR

All data below was generated in the single aqueous phase flow microreactor following the reaction and analysis procedures outlined in Chapter 5 under sections 5.2.2 and 5.2.4. Abbreviations GLU=glucose MAN=mannose FRU=fructose FA=formic acid LA=levulinic acid HMF=5-hydroxymethylfurfural LG=levoglucosan

220

Glucose Isomerization with Varying CrCl3 Catalyst Concentration

(a) 100 (b) 20 17mM CrCl3

8.5mM CrCl3 2.5mM CrCl 80 3 15 1.7mM CrCl3

0.85mM CrCl3 60 10 40

17mM CrCl3 8.5mM CrCl 5 20 3 (%) Yield HMF 2.5mM CrCl3

GLU Conversion (%) Conversion GLU 1.7mM CrCl3 0.85mM CrCl 0 3 0 0 20 40 60 80 100 120 0 20 40 60 80 100 120 Residence Time (min) Residence Time (min)

(c) 12 (d) 17mM CrCl3 25 17mM CrCl3 8.5mM CrCl3 10 8.5mM CrCl3 2.5mM CrCl3 20 2.5mM CrCl3 1.7mM CrCl3 1.7mM CrCl3 8 0.85mM CrCl3 0.85mM CrCl3 15 6

4 10

2 5

Mannose Yield (%) Yield Mannose Fructose Yield (%) Yield Fructose

0 0 0 20 40 60 80 100 120 0 20 40 60 80 100 120 Residence Time (min) Residence Time (min) Figure F. 1 (a) Glucose conversion and (b) mannose, (c) fructose, and (d) HMF yields vs. residence time at 140 oC at varying catalyst concentrations. Starting material: 1 wt% glucose and 0.85-17 mM CrCl3 catalyst in water. The yields of formic and levulinic acids are negligible and therefore are not shown. The error bars are obtained from the standard deviation of two repeat runs.

221 (a) 12 (b) 17mM CrCl3 24 17mM CrCl3 8.5mM CrCl3 8.5mM CrCl3 2.5mM CrCl3 10 20 2.5mM CrCl 1.7mM CrCl 3 3 1.7mM CrCl 0.85mM CrCl 3 3 0.85mM CrCl 8 16 3

6 12

8

4

MAN Yield (%) Yield MAN FRU Yield (%) Yield FRU 4 2 0 10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100 GLU Conversion (%) GLU Conversion (%) (c) 20 17mM CrCl3

8.5mM CrCl3 2.5mM CrCl 16 3 1.7mM CrCl3

0.85mM CrCl3 12

8

HMF Yield (%) Yield HMF 4

0 10 20 30 40 50 60 70 80 90 100 GLU Conversion (%) Figure F. 2 (a) Mannose, (b) fructose, and (c) HMF yields vs. glucose conversion at 140 oC at varying catalyst concentrations. Starting material: 1 wt% glucose and 0.85-17 mM CrCl3 catalyst in water. The yields of formic and levulinic acids are negligible and therefore are not shown. The error bars are obtained from the standard deviation of two repeat runs.

Kinetics Data for the Glucose Reaction Network over the CrCl3 Catalyst Kinetic experiments were carried out with 1 wt% of glucose, mannose, fructose, and HMF as separate reactants and 1.7 mM CrCl3 as the catalyst at low conversions (<15%).

222 (a) (b) 3.0 16 140 oC 140 oC o o 160 C 2.5 160 C 180 oC 180 oC 12 200 oC 200 oC 2.0

8 1.5 1.0 4

MAN Yield (%) Yield MAN 0.5 GLU Conversion (%) Conversion GLU 0 0.0 0 1 2 3 4 0 1 2 3 4 Residence Time (min) Residence Time (min)

(c) 5 (d) 2.0 140 oC 160 oC 4 o 1.5 180 C 200 oC 3 1.0 2

140 oC 0.5 FRU Yield (%) Yield FRU 1 160 oC (%) Yield HMF 180 oC 200 oC 0 0.0 0 1 2 3 4 0 1 2 3 4 Residence Time (min) Residence Time (min)

Figure F. 3 (a) Glucose conversion and (b) mannose, (c) fructose, and (d) HMF yields o vs. residence time at 140-200 C. Starting material: 1 wt% glucose and 1.7 mM CrCl3 catalyst in water. The yields of formic and levulinic acids are negligible and therefore are not shown.

223 (a) 20 (b) o 1.5 140 C o o 140 C 160 C o o 160 C 170 C o 15 o 170 C 180 C o 1.0 180 C 10

0.5

5

GLU Yield (%) Yield GLU MAN Conversion (%) Conversion MAN 0 0.0 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 Residence Time (s) Residence Time (s) (c) 8 (d) 140 oC 1.0 140 oC 160 oC 160 oC 170 oC 6 0.8 170 oC 180 oC 180 oC 0.6 4 0.4

2 FRU Yield (%) Yield FRU HMF Yield (%) Yield HMF 0.2

0 0.0 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 Residence Time (s) Residence Time (s) Figure F. 4 (a) Mannose conversion and (b) glucose, (c) fructose, and (d) HMF yields vs. residence time at 140-180 oC. Reaction conditions: 1 wt% mannose and 1.7 mM CrCl3 catalyst in water. The yields of formic and levulinic acids are negligible and therefore are not shown.

224 (a) (b) 12 0.8

10 0.6 8

6 0.4

4 140 oC 0.2 140 oC 2 160 oC (%) Yield GLU 160 oC

170 oC 170 oC FRU Conversion (%) Conversion FRU o o 0 180 C 0.0 180 C 0 5 10 15 20 25 0 5 10 15 20 25 Residence Time (s) Residence Time (s) (c) (d) 1.6 0.8

1.2 0.6

0.8 140 oC 0.4 160 oC 170 oC o

0.4 180 oC 0.2 140 C HMF Yield (%) Yield HMF MAN Yield (%) Yield MAN 160 oC 170 oC 180 oC 0.0 0.0 0 5 10 15 20 25 0 5 10 15 20 25 Residence Time (s) Residence Time (s) Figure F. 5(a) Fructose conversion and (b) glucose, (c) mannose, and (d) HMF yields o vs. residence time at 140-180 C. Starting material: 1 wt% fructose and 1.7 mM CrCl3 catalyst in water. The yields of formic and levulinic acids are negligible and are therefore not shown.

225 (b) 10 (a) 16 o o 140 C 140 C o o 160 C 160 C 8 o o 180 C 12 180 C o o 200 C 200 C 6 8 4

4 (%) Yield LA 2

HMF Conversion (%) Conversion HMF 0 0 0 10 20 30 40 50 0 10 20 30 40 50 Residence Time (min) Residence Time (min)

(c) 2.0 140 oC 160 oC o 1.5 180 C 200 oC

1.0

FA Yield (%) Yield FA 0.5

0.0 0 10 20 30 40 50 Residence Time (min)

Figure F. 6 (a) HMF conversion, (b) formic acid and (c) levulinic acid yields vs. o residence time at 140-180 C. Starting material: 1 wt% fructose and 1.7 mM CrCl3 catalyst in water.

226 Glucose Conversion Using Brønsted Acid (HCl) Catalyst

(a) (b) 60 30 GLU GLU MAN MAN 50 FRU FRU FA FA LA 20 LA 40 HMF HMF LG LG 30

10 20

10 Conversion or Yield (%) Yield or Conversion Conversion or Yield (%) Yield or Conversion 0 0 0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 Residence time (min) Residence time (min) Figure F. 7 Glucose conversion and mannose, fructose, HMF, FA, LA and LG carbon yields vs. residence time at (a) 140 °C and (b) 160 °C. Starting material: 1 wt% glucose and 0.2 mol/L HCl catalyst in water (pH=0.7).

227 (a) 10 (b) 40 GLU GLU MAN MAN 8 FRU FRU FA 30 FA LA LA 6 HMF HMF LG 20 LG 4

10

2 Conversion or Yield (%) Yield or Conversion Conversion or Yield (%) Yield or Conversion 0 0 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Residence time (min) Residence time (min)

(c) 60 GLU MAN 50 FRU FA 40 LA HMF 30 LG

20

10 Conversion or Yield (%) Yield or Conversion 0 0 5 10 15 20 Residence time (min) Figure F. 8 Glucose conversion and mannose, fructose, HMF, FA, LA and LG carbon yields vs. residence time at (a) 140 °C, (b) 160 °C, and (c) at 180 oC. Starting material: 1 wt% glucose and 0.1 mol/L HCl catalyst in water (pH=1).

(a) 30 (b) 60 GLU GLU MAN MAN FRU 50 FRU FA FA 20 LA 40 LA HMF HMF LG 30 LG

10 20

10

Conversion or Yield (%) Yield or Conversion (%) Yield or Conversion 0 0 0 10 20 30 40 50 60 0 10 20 30 Residence time (min) Residence time (min)

228 Figure F. 9 Glucose conversion and mannose, fructose, HMF, FA, LA and LG carbon yields vs. residence time at (a) 160 °C and (b) at 180 oC. Starting material: 1 wt% glucose and 0.056 mol/L HCl catalyst in water (pH=1.25).

(a) 30 (b) 60 GLU GLU MAN MAN FRU 50 FRU FA FA 20 LA 40 LA HMF HMF LG 30 LG

10 20

10

Conversion or Yield (%) Yield or Conversion (%) Yield or Conversion 0 0 0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 Residence time (min) Residence time (min) Figure F. 10 Glucose conversion and mannose, fructose, HMF, FA, LA and LG carbon yields vs. residence time at (a) 160 °C and (b)180 oC. Starting material: 1 wt% glucose and 0.03 mol/L HCl catalyst in water (pH=1.52).

(a) 30 (b) 50 GLU GLU MAN MAN FRU 40 FRU FA FA 20 LA LA HMF 30 HMF LG LG 20 10

10 Conversion or Yield (%) Yield or Conversion Conversion or Yield (%) Yield or Conversion 0 0 0 10 20 30 40 50 0 5 10 15 20 Residence time (min) Residence time (min) Figure F. 11 Glucose conversion and mannose, fructose, HMF, FA, LA and LG carbon yields vs. residence time at (a) 180 °C and (b) 200 °C. Starting material: 1 wt% glucose and 0.01 mol/L HCl catalyst in water (pH=2).

229 Glucose Conversion Using Tandem CrC3/HCl Catalysts

(b) (a) 30 40 GLU GLU MAN MAN FRU FRU FA 30 FA 20 LA LA HMF HMF 20 10

10 Conversion or Yield (%) Yield or Conversion Conversion or Yield (%) Yield or Conversion 0 0 0 10 20 30 40 50 60 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Residence Time (min) Residence Time (min) (c) (d) 70 GLU GLU 40 MAN MAN 60 FRU FRU FA FA 50 LA 30 LA HMF HMF 40 20 30 20 10

10 Conversion or Yield (%) Yield or Conversion Conversion or Yield (%) Yield or Conversion 0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 Residence Time (min) Residence Time (min) Figure F. 12 Glucose conversion and mannose, fructose, HMF, FA, and LA carbon yields vs. residence time at (a) 140 °C, (b) 160 °C, (c) 180 °C and (d) 200 °C. Starting material: 1 wt% glucose, 1.7mM CrCl3 and 0.1 mol/L HCl catalyst in water (pH=1).

230 (a) (b) 20 50 GLU GLU MAN MAN FRU 40 FRU FA FA LA LA HMF 30 HMF 10 20

10 Conversion or Yield (%) Yield or Conversion Conversion or Yield (%) Yield or Conversion 0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Residence Time (min) Residence Time (min) (c) 70 GLU 60 MAN FRU 50 FA LA 40 HMF 30 20 10

Conversion or Yield (%) Yield or Conversion 0 0.0 0.5 1.0 1.5 2.0 2.5 Residence Time (min) Figure F. 13 Glucose conversion and mannose, fructose, HMF, FA, and LA carbon yields vs. residence time at (a) 160 °C, (b) 180 °C and (c) 200 °C. Starting material: 1 wt% glucose, 1.7mM CrCl3 and 0.056 mol/L HCl catalyst in water (pH=1.25).

231 (a) (b) 50 80 GLU GLU MAN MAN 40 FRU FRU FA 60 FA LA LA 30 HMF HMF 40 20

20

10

Conversion or Yield (%) Yield or Conversion (%) Yield or Conversion 0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Residence Time (min) Residence Time (min) (c) 80 GLU MAN FRU 60 FA LA HMF 40

20 Conversion or Yield (%) Yield or Conversion 0 0.0 0.5 1.0 1.5 2.0 2.5 Residence Time (min) Figure F. 14 Glucose conversion and mannose, fructose, HMF, FA, and LA carbon yields vs. residence time at (a) 160 °C, (b) 180 °C and (c) 200 °C. Starting material: 1 wt% glucose, 1.7mM CrCl3 and 0.017 mol/L HCl catalyst in water (pH=1.77).

232 (a) 15 (b) GLU 50 MAN GLU FRU MAN FRU FA 40 FA 10 LA LA HMF 30 HMF

5 20

10 Conversion or Yield (%) Yield or Conversion 0 (%) Yield or Conversion 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Residence Time (min) Residence Time (min) (c) 50 GLU MAN 40 FRU FA LA 30 HMF

20

10

Conversion or Yield (%) Yield or Conversion 0 0.0 0.5 1.0 1.5 2.0 2.5 Residence Time (min) Figure F. 15 Glucose conversion and mannose, fructose, HMF, FA, and LA carbon yields vs. residence time at (a) 160 °C, (b) 180 °C and (c) 200 °C. Starting material: 1 wt% glucose, 1.7mM CrCl3 and 0.01 mol/L HCl catalyst in water (pH=2).

233

Appendix G

PARTITIONING OF SOLID ACID CATALYST PARTICLES BETWEEN

WATER AND 2-PENTANOL PHASES

Objective The objective of this experiment is to determine the partitioning of solid acid catalysts commonly used for the dehydration of sugars between water/organic biphasic solvent systems. The desired catalyst should have minimal partitioning into the organic phase, so that side reactions of the organic solvent and the extracted 5- hydroxymethylfurfural (HMF) is minimal.

Materials The H-beta (H-BEA) zeolite used in this experiment has a Si/Al ratio of 12.5 and was obtained by calcining commercially available NH4-beta. NH4-beta was purchased from Zeolyst International. Then, it was calcined in air in a muffler oven. The oven temperature first went from 25 to 90 oC in steps of 2 oC/min, and then was held at 90 oC for 1 h. Next, the temperature was brought up to 450 oC in steps of 2 oC/min and held at 450 °C for 8 h. Finally, heat was turned off and the oven was naturally cooled off. Sn-beta (Sn-BEA) was synthesized using the solid-state ion-exchange (SSIE) method and has a Si/Sn ratio of 73. Zr-beta (Zr-BEA) was also synthesized using the SSIE method and has a Si/Zr ratio of 200. Titanium oxide (anatase) and zirconium oxide were purchased from Alfa Aesar.

Procedure for Partitioning Experiments 0.1g of the solid acid catalyst particle, 5mL of water, 5mL of 2-pentanol and a stir bar were added to a 20mL glass vial and stirred at 700 rpm at 20 °C. After 45 min, stirring was turned off and the two phases were allowed to settle and separate. The water phase (bottom layer) was turbid and the 2-pentanol phase (top layer) was mostly clear. After 5 minutes of settling, 3mL of the 2-pentanol layer was transferred into a 1cm pathlength cuvette for turbidity measurements. Turbidity measurements were conducted on a Cary 600 UV-Visible Spectrophotometer. All scans were performed in the 200-800 nm range. 2-pentanol was used for 100% transmission baseline reference and zero absorbance calibrations. The absorbance value at 600 nm was used as the turbidity of the sample. Samples used for turbidity calibration were prepared by adding known quantities of solids in 2-pentanol followed by vortex shaking. The calibration plots that relate the turbidity with the mass concentration of particles (in mg/mL) are shown in Figure G1.

234 0.8

0.6

0.4

Turbidity H-BEA 0.2 Sn-BEA Zr-BEA

ZrO2 TiO 0.0 2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Particle Concentration (mg/mL)

Figure G. 1 Turbidity vs. mass concentration for H-BEA, Sn-BEA, Zr-BEA, TiO2, and ZrO2 particles suspended in 2-pentanol. Linear fits with zero y-intercepts were used to obtain the response factors which are summarized in Table F1.

Table G. 1 Response Factors for Solid Acid Catalyst Particles Suspensions in 2- Pentanol Material Response Factor (1/mg-mL) H-BEA 0.4352 Sn-BEA 0.3752 Zr-BEA 0.6965 TiO2 1.795 ZrO2 2.771

Results After measuring the turbidity of the solids suspended in 2-pentanol, they were converted to mass concentrations using the response factors above. Assuming the volume change of the 2-pentanol phase was negligible after mixing, the total mass of the particles partitioned into the 2-pentanol phase can be calculated as shown in the example below: 0.264푚푔 1푔 × 5푚퐿 × ( ) = 1.320 × 10−3푔 푚퐿 1000푚푔

235 Then, this was divided by the mass of initial particle input (0.1g) to give the percentage of particles partitioned into the 2-pentanol phase: 1.320 × 10−3푔 × 100% = 1.32% 0.100푔 The % dispersion of all five solids acid catalyst materials in 2-pentanol are given in Figure G2. All of them have very low dispersions in 2-pentanol (<2%), while Sn- BEA is the most hydrophobic. Therefore, all five catalyst can be used for glucose or fructose dehydration reactions in water/2-pentanol biphasic systems with minimal partitioning into the organic phase. 1.5

1.0

0.5

% Partitioning in 2-Pentanol 2-Pentanol in Partitioning % 0.0 Sn-BEA H-BEA ZrO2 Zr-BEA TiO2

Material Figure G. 2 Percent partitioned into the 2-pentanol phase for H-BEA, Sn-BEA, Zr-BEA, TiO2, and ZrO2 particles mixed with water/2-pentanol solvents.

236 Appendix H

PERMISSION Chapter 2 was reproduced with permission from the Royal Society of Chemistry152 Chapter 3 was reproduced with permission from John Wiley and Sons199

237 Figure H. 3 Permission for Chapter 2.

Figure H. 4 Permission for Chapter 4.

238

239