Selective C-O Hydrogenolysis and Decarboxylation of Biomass-Derived Heterocyclic Compounds Over Heterogeneous Catalysts

Selective C-O Hydrogenolysis and Decarboxylation of Biomass-Derived Heterocyclic Compounds over Heterogeneous Catalysts

Mei Chia

A dissertation submitted in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

(Chemical Engineering)

At the

UNIVERSITY OF WISCONSIN-MADISON

2013

Date of final oral examination: 11th July 2013

The dissertation is approved by the following members of the Final Oral Committee: James A. Dumesic, Professor, Chemical and Biological Engineering Thomas F. Kuech, Professor, Chemical and Biological Engineering Manos Mavrikakis, Professor, Chemical and Biological Engineering Brian F. Pfleger, Assistant Professor, Chemical and Biological Engineering Dane Morgan, Associate Professor, Materials Science and Engineering

i Selective C-O Hydrogenolysis and Decarboxylation of Biomass-Derived Heterocyclic Compounds over Heterogeneous Catalysts

Mei Chia

Under the supervision of Professor James A. Dumesic

At the University of Wisconsin-Madison

The catalytic deoxygenation of biomass-derived compounds through selective C-O hydrogenolysis, catalytic transfer hydrogenation and lactonization, and decarboxylation to value- added chemicals over heterogeneous catalysts was examined under liquid phase reaction conditions. The reactions studied involve the conversion or production of heterocyclic compounds, specifically, cyclic ethers, lactones, and 2-pyrones.

A bimetallic RhRe/C catalyst was found to be selective for the hydrogenolysis of secondary C-O bonds for a broad range cyclic ethers and polyols. Results from experimentally- observed reactivity trends, NH3 temperature-programmed desorption, fructose dehydration reaction studies, and first-principles density functional theory (DFT) calculations are consistent with the hypothesis of a bifunctional catalyst which facilitates acid-catalyzed ring-opening and dehydration coupled with metal-catalyzed hydrogenation. C-O hydrogenolysis and fructose dehydration activities were observed to decrease with an increase in reduction temperature and a decrease in the number of surface metallic Re atoms measured by in situ X-ray absorption spectroscopy. No C-O hydrogenolysis activity was detected over RhRe/C under water-free

ii conditions. The activation of water molecules by Re atoms on the surface of metallic Rh is suggested to result in the formation of Brønsted acidity over RhRe/C.

The catalytic transfer hydrogenation and lactonization of levulinic acid and its esters to γ- valerolactone was accomplished through the Meerwein-Ponndorf-Verley reaction over metal oxide catalysts using secondary alcohols as the hydrogen donor. ZrO2 was a highly active material for CTH under batch and continuous flow reaction conditions; the initial activity of the catalyst was repeatedly regenerable by calcination in air, with no observable loss in catalytic activity.

Lastly, the 2-pyrone, triacetic acid lactone, is shown to be a promising biorenewable platform chemical from which a wide range of chemical intermediates and end products can be obtained using heterogeneous catalysts or by thermal decomposition. Mechanistic insights from experimentally-observed reactivity trends and results from DFT calculations indicate that 2- pyrones undergo reactions unique to their structure such as keto-enol tautomerization, retro

Diels-Alder, and nucleophilic attack by water. Ring-opening and decarboxylation reactions were found to be governed by key structural features such as the degree of saturation in the ring (e.g.,

C4=C5 bond), nature of the solvent, and presence of an acid catalyst.

Approved by ______Professor James A. Dumesic

Date ______

iii Acknowledgements

This thesis would not have been possible without the generous contributions of many people that I have had the privilege to work with and learn from. I would first like to thank my thesis advisor, Professor James Dumesic, for giving me the opportunity to be part of a legendary research group; his guidance, (contagious) wry humor, and unfailing enthusiasm for science have played a central role in shaping this thesis and my PhD experience.

I am very grateful to my labmates, both past and present Dumesic group members, for their help and friendship. I especially would like to thank Mark and Yomaira for showing me the ropes when I first joined the group, and for mentoring me and giving valuable advice over the years. I would also like to thank the many individuals who have contributed to this thesis in so many ways, specifically, Drew, Christian, Ryan, David, Elif, Jesse, Stephanie, Jean, Max, Eric,

Gretchen, Ronald, Ana, and Carrie for sharing their knowledge and for technical assistance in the lab, and especially Ricky, Tom S., and Brandon, with whom I have had the opportunity to collaborate on projects with. Many long afternoons in the lab have been made enjoyable through the numerous scientific (and non-scientific/ philosophical) discussions I have had with all of you.

I would like to express my gratitude to the many administrators and staff at the Chemical

Engineering department, especially Donna Bell, Kathy Heinzen, John Cannon, John Ames, Joel

Lord, Eric Codner, and Todd Ninman for their administrative, mechanical, instrumental, and technical support, respectively. Special thanks to Judy Lewison, whose assistance with administrative matters, baked treats, and chats in the hallways have brightened many slow mornings.

iv A large part of the work here was only possible through the contributions of many external collaborators. Specifically, the computational work herein was performed by our collaborators at the University of Virginia; I would especially like to thank Professor Matthew

Neurock for sharing valuable knowledge and ideas during our project discussions, and through whom I, as an experimentalist, have been able to gain some insight into computational chemistry.

I would also like to acknowledge Dr. M. Ali Haider, David Hibbitts and Qiaohua Tan, all of whose contributions are instrumental to the fruition of much of the work here. I am grateful for generous help from Professor George Kraus and Gerald Pollock (Iowa State University) who synthesized organic compounds for reaction studies that enabled us to tell a complete “story” for the pyrone chemistry. The microscopy work presented in this thesis was performed by collaborators at the University of New Mexico, Professor Abhaya Datye and Dr. Hien Pham. I would also like to thank Dr. Jeffery Miller (Argonne National Laboratory) who was invaluable in obtaining and interpreting XAS data, and Professor Fabio Riberio and Paul Dietrich (Purdue

University) for access to and assistance with equipment for performing XAS experiments. I would also like to acknowledge funding from the NSF Engineering Research Centre for

Biorenewable Chemicals (CBiRC), and Professor Brent H. Shanks for sharing his perspective of the CBiRC vision which has influenced the work here. Also, I am especially grateful to the late

Dr PK Wong, who gave me invaluable professional advice over the years.

Finally, I would like to thank my family and friends for their unwavering support over the years. In particular, I would like to thank my parents, who have always encouraged me to pursue my aspirations and interests, and my aunt, April, who has been so supportive of my pursuits.

Lastly, I would like to thank my brother, who inspired me to take the road less travelled.

To my family

There is a pleasure in the pathless woods

vii Table of Contents

Abstract i Acknowledgements iii Dedication v List of Figures xi List of Tables xviii

1. Introduction 1 1.1 Current production and consumption of fuel and chemicals 1 1.2 Future outlook for fossil-based resources 4 1.3 Biomass as feedstock for the production of fuel and chemicals 6 1.4 Catalytic strategies for the production of chemicals from biomass-based feedstocks 13 1.4.1 5-hydroxymethylfurfural as a platform chemical 13 1.4.2 Bimetallic catalysts and applications for biomass conversion 16 1.4.3 Catalytic upgrading of biologically-produced compounds 21 1.5 Research overview and strategy 24 1.6 References 25 2. Experimental techniques 30 2.1 Catalyst preparation and synthesis methods 30 2.1.1 Supported metal catalysts 30 2.1.2 Metal oxide catalysts 31 2.1.3 Ion-exchange resins and zeolites 31 2.2 Reaction studies 31 2.2.1 Batch reactions 31 2.2.2 Continuous flow reactions 32 2.3 Analytical methods 33 2.3.1 High performance liquid chromatography 33 2.3.2 Gas chromatography 33 2.3.3 Nuclear magnetic resonance spectroscopy 34 2.4 Catalyst characterization 34

viii 2.4.1 Temperature-programmed methods 36 2.4.2 CO adsorption 36 2.4.3 Electron microscopy 36 2.4.4 In situ and operando X-ray absorption spectroscopy 36 2.5 References 38 3. Selective C-O hydrogenolysis over bimetallic catalysts 39 3.1 Introduction 39 3.1.1 Importance of C-O hydrogenolysis as a deoxygenation strategy 39 3.1.2 Current state of the art 41 3.2 Catalyst development 46 3.2.1 Highly reducible metals 46 3.2.2 Effect of varying catalyst composition 47 3.2.3 Precursor effects for Mo-promoted catalysts 48 3.2.4 Support effects 49 3.2.5 Initial reactivity studies 50 3.3 Characterization studies of RhRe/C 52 3.3.1 Temperature-programmed reduction 52 3.3.2 HAADF-STEM and EDS 53 3.3.3 CO adsorption 55

3.3.4 NH3 temperature-programmed desorption 56 3.4 Catalyst pretreatment and stability 57 3.5 Reactivity trends and Density Functional Theory calculations 59 3.5.1 Cyclic ethers 59 3.5.2 Diols and polyols 69 3.5.3 Governing principles of substrate reactivity and selectivity 70 3.5.4 Reaction kinetics 73 3.5.5 Role of rhenium and nature of the active site 75 3.6 Conclusions 78 3.7 Chemical notations 79 3.8 Computational methods 80

ix 3.9 References 82 4. Nature of the active site over rhodium-rhenium catalysts 86 4.1 Introduction 86 4.2 Effect of reduction temperature on C-O hydrogenolysis activity and acid site 88 density 4.3 Fructose dehydration to HMF 94 4.4 Mass transfer effects 98 4.5 X-ray absorption studies 99 4.5.1 Monometallic catalysts Rh/C and Re/C under in situ conditions 100 4.5.2 RhRe/C under in situ and operando C-O hydrogenolysis conditions 103 4.6 Nature of the active site 112 4.7 Effect of solvent on catalytic activity 117 4.8 Conclusions 118 4.9 References 120 5. Selective C=O hydrogenation 122 5.1 Introduction 122 5.1.1 Selective hydrogenation of unsaturated aldehydes and ketones over 123 heterogeneous metal catalysts 5.1.2 The Meerwein-Ponndorf-Verley reaction 124 5.1.3 Hydrogenation of levulinic to γ-valerolactone 126 5.2 Initial catalyst studies using metal oxide catalysts and levulinate esters 128

5.3 Catalytic transfer hydrogenation using ZrO2 129 5.3.1 Effect of levulinate esters and hydrogen donor on catalyst activity 129 5.3.2 CTH of levulinate esters in the presence of sec-butylphenol 130 5.3.3 Effect of levulinic acid on catalyst activity 132 5.3.4 Effect of water on catalyst activity 133

5.3.5 Regenerability of ZrO2 134 5.4 Conclusions 136 5.5 References 136

x 6. Triacetic acid lactone as a biorenewable platform chemical 140 6.1 Introduction 140 6.1.1 Combination of biological and chemical catalysis – platform chemical 140 approach 6.1.2 Triacetic acid lactone 142 6.2 Chemical diversification of triacetic acid lactone 143 6.3 Conclusions 150 6.4 Chemical notations 151 6.5 Analytical methods and NMR data 151 6.6 Computational methods 153 6.7 References 154 7. Mechanistic insights into ring-opening and decarboxylation of 2-pyrones 156 7.1 Introduction 156 7.2 Decarboxylation of triacetic acid lactone 158 7.3 Decarboxylation of 5,6-dihydro-4-hydroxy-6-methyl-2H-pyran-2-one 162 7.4 Reactivity of 4-hydroxy-6-methyltetrahydro-2-pyrone 168 7.5 General reactivity rules 173 7.6 Conclusions 175 7.7 Chemical notations 176 7.8 Synthesis of isoparasorbic acid 176 7.9 Computational methods 177 7.10 References 179

xi List of Figures

Figure 1-1 (a) Global energy consumption by fuel in 2011.1 Abbreviations “GTL” and 1 “CTL” refer to gas-to-liquid and coal-to-liquid, respectively. (b) 2011 US primary energy consumption estimates by source2. All numbers are in units of quadrillion Btu.

Figure 1-2 Primary energy consumption in the US by source and sector, 20112. 2 Numbers are in units of quadrillion Btu. “Renewables” refers to conventional hydroelectric power, geothermal, solar/ photovoltaic, wind power, and biomass.

Figure 1-3 Breakdown of components per barrel of crude oil. Numbers are in units of 4 gallons/ barrel.

Figure 1-4 General composition15 of lignocellulosic biomass and representative 8 structures of constituent components.

Figure 1-5 Illustration of the conversion of biomass through thermochemical and 10 aqueous phase processing (i.e., hydrolysis) strategies to fuel and platform chemicals.

Figure 1-6 Illustration of the fossil- and biomass-based chemicals industries and the 11 critical role of catalysis in both.

Figure 1-7 Possible chemicals from 5-(hydroxymethyl)furfural (HMF, 1). Compounds 15 are as follows: 2,5-(dihydroxymethyl) tetrahydrofuran (DHMTHF, 2); 2,5- bis(aminomethyl)-furan (3); 2,5-furandicarboxylic acid (FDCA, 4); 1,6- hexanediol (1,6-HDO, 5); adipic acid (6); hexamethylenediamine (HMDA, 7).

Figure 1-8 Proposed catalytic route for the production of 1,6-hexanediol (1,6-HDO, 5) 16 from 5-(hydroxymethyl)-furfural (HMF, 1). Compounds are as follows: 2,5-(dihydroxymethyl)tetrahydrofuran (DHMTHF, 2); 1,2,6-hexanetriol (3); 2-(hydroxymethyl) tetrahydropyran (HMTHP, 4).

Figure 1-9 Conversion of glycerol to various chemicals over bimetallic catalysts. 20 Adapted from Alonso et al.64

Figure 1-10 Conversion of lactic acid to chemicals. Compounds are as follows: lactic 23 acid (1), 1,2-propanediol (2), lactates (3), lactide (4), pyruvic acid (5), acrylic acid (6), acetaladehyde (7), and 2,3-pentadione (8). Adapted from Corma et al.33

xii Figure 2-1 Experimental set-up for batch reaction studies. 32

Figure 3-1 Reaction pathways for the C-O hydrogenolysis of (2- 40 hydroxymethyl)tetrahydropyran to hexanediol, hexanol and hexane.

Figure 3-2 Temperature-programmed reduction profiles for (a) 4 wt% Rh/C, (b) 3.6 53 wt% Re/C, (c) 1.8 wt% Mo/C, (d) 4 wt% RhRe/C (1:0.5), (e) 4 wt% RhMo/C (1:0.5).

Figure 3-3 Representative HAADF STEM images and EDS spot-beam analysis results 54 of (a) as-prepared and (b) spent 4 wt% RhRe/C (1:0.5) catalysts.

Figure 3-4 Representative HAADF-STEM images and particle size distribution 55 histograms for (a, c) as-prepared and (b, d) spent 4 wt% RhRe/C (1:0.5) catalysts. Number-averaged particle sizes are presented in the figures.

Figure 3-5 NH3 temperature-programmed desorption profiles for (a) 3.6 wt% Re/C, (b) 57 4 wt% Rh/C, and (c) 4 wt% RhRe/C (1:0.5). Catalyst samples were pretreated in flowing H2 (100 cm3 (STP) min-1) at 523 K for 4 h prior to dosing of NH3.

Figure 3-6 Results for the hydrogenolysis of 2-(hydroxymethyl)tetrahydropyran 1 over 59 4 wt% RhRe/C (1:0.5) in a continuous flow system. Conversion of 1 (■), selectivities to 1,6-hexanediol 2 (○) and 1-hexanol (▲), and reaction rates (×) at 393 K, 34 bar H2, WHSV = 0.52 h-1, feed: 5 wt% 1 in water. The 3 -1 catalyst was pretreated in flowing H2 (60 cm (STP) min ) at 523 K for 4 h and cooled to the reaction temperature prior to initiation of liquid feed flow.

Figure 3-7 DFT-calculated carbenium and oxocarbenium ion formation or reaction 65 energies for 2,3-butanediol 16, 2,4-pentanediol 15, and tetrahydrofurfuryl alcohol 3. The dotted lines refer to sequential paths which proceed via the formation of the OH-stabilized three (oxirane) and four (oxetene) ring intermediates that subsequently form the corresponding oxocarbenium ion intermediates whereas the solid lines refer to concerted protonation/hydride transfer steps that result in the direct formation of the oxycarbenium ion intermediates.

Figure 3-8 DFT-calculated reactant and transition state structures and corresponding 67 activation barriers for the acid catalyzed ring-opening of: (a) water- stabilized tetrahydrofurfuryl alcohol 3, (b) tetrahydrofurfuryl alcohol 3 and (c) 2-methyltetrahydrofuran 7 on a model Rh200Re1OH cluster.

Figure 3-9 DFT-calculated reactant and transition state structures for the concerted 67 protonation, hydride transfer and ring-opening of the water-stabilized

xiii tetrahydrofurfuryl alcohol 3 over a model Rh(111) surface with well dispersed Re-OH sites.

Figure 3-10 DFT-calculated reaction path and energies for the metal catalyzed ring- 68 opening of 2-methyltetrahydrofuran 7 at the substituted and unsubstituted carbon centers over Rh.

Figure 3-11 Comparison of the ratio between specific reaction rate (μmol g-1 min-1) and 72 initial reactant concentration (μmol mL-1) (on a logarithmic scale) and DFT-calculated carbenium and oxocarbenium ion energies for selected cyclic ethers and linear polyols. Each region shows distinct shifts in activity due to the stability of the carbenium and oxocarbenium ions as discussed in Table 3-6.

Figure 3-12 DFT-calculated deprotonation energies for various surfaces and atom 76 clusters.

Figure 4-1 Reactions discussed in this chapter: C-O hydrogenolysis of 2- 88 (hydroxymethyl)tetrahydropyran to 1,6-hexanediol and dehydration of glucose to 5-hydroxymethylfurfural.

Figure 4-2 Hydrogenolysis of HMTHP over RhRe/C under continuous flow reaction 90 conditions. Catalyst pretreated in flowing H2 at 393 K (×), 523 K (○) and 723 K (■) prior to initiation of liquid feed. Conversion levels of HMTHP at 8 h time-on-stream were 9%, 9% and 10% for 393 K (×), 523 K (○), and 723 K (■), respectively. Selectivities to 1,6-HDO were >90% for all data points. Reaction conditions: 5 wt% HMTHP in water as feed, -1 -1 -1 WHSV393K=4.2 h , WHSV523K=2.2 h , WHSV723K=0.5 h , 393 K, 500 psi H2.

Figure 4-3 Hydrogenolysis of HMTHP over RhRe/C under continuous flow reaction 91 conditions. Catalyst pretreated in flowing H2 at 393 K prior to initiation of liquid feed. Reaction conditions: 5 wt% HMTHP in water as feed, WHSV -1 = 4.2 h , 393 K, 34 bar H2.

Figure 4-4 NH3 temperature-programmed desorption profiles for: (a) Vulcan carbon, 92 (b) Rh/C pretreated in flowing H2 at 523 K, (c) Re/C pretreated in flowing H2 at 523 K, (d) RhRe/C (1:0.5) pretreated in flowing H2 at 723 K, (e) RhRe/C (1:0.5) pretreated in flowing H2 at 523 K, and (f) RhRe/C (1:0.5) pretreated in flowing H2 at 393 K.

Figure 4-5 Conversion of fructose (×), selectivity to 5-hydroxymethylfurfural (●), and 95 specific formation rate of HMF (■) as a function of time on stream over RhRe/C catalyst pretreated at 523 K. Reaction conditions: 300 psi He, 403 K, 2 wt% fructose in THF/water (mass ratio of THF: water = 4:1) as

xiv feed, WHSV =0.1h-1 .

Figure 4-6 Fructose dehydration in a continuous flow reaction system over RhRe/C 96 catalyst pretreated at different reduction temperatures of 523 K (■) and 723 K (×). Reaction conditions: 300 psi He, 403 K, 2 wt% fructose in THF/water (mass ratio of THF: water = 4:1) as feed, WHSV =0.1h-1.

Figure 4-7 Conversion of fructose (×), selectivity to HMF (●), and specific formation 97 rate of HMF (■) as a function of time on stream over ZSM-5. Reaction conditions: 300 psi He, 403 K, 2 wt% fructose in THF/water (mass ratio of THF: water = 4:1) as feed, WHSV =1.4 h-1.

Figure 4-8 Specific formation rates of HMF over RhRe/C pretreated at 523 K (×) and ZSM-5 (■). Reaction conditions: 300 psi He, 403 K, 2 wt% fructose in 98 THF/water (mass ratio of THF: water = 4:1) as feed.

Figure 4-9 (a) Rh K-edge XANES (23.19-23.27 keV) of Rh2O3 standard (black) and 101 4 wt% Rh/C catalyst reduced at 298 K (green), 313 K (blue), and 723 K (red). (b) Magnitude of k2-weighted FT of EXAFS data of Rh K-edge of Rh foil (black) and 4 wt% Rh/C catalyst reduced at 298 K (green) and 723 K (red) (Δk = 2.7 – 12.5 Å-1).

Figure 4-10 (a) Re LIII-edge XANES (10.51 - 10.56 keV) of 4 wt% Re/C reduced at 102 298 K (red) and 723 K (blue). (b) Magnitude of k2-weighted FT of EXAFS data of Re LIII-edge of 4 wt% Re/C catalyst reduced at 298 K (red) and 723 K (blue).

Figure 4-11 Rh K-edge XANES (23.19 – 23.27 keV) of (a) Rh foil (black), Rh/C (red), 105 and RhRe/C catalyst reduced at 723 K (blue), and (b) RhRe/C catalyst reduced at 393 K (black), 523 K (red), and 723 K (blue); profiles for catalyst reduced at 393 K and 523 K overlap and are not apparent in figure. Figure 4-12 Re LIII-edge XANES (10.51 – 10.56 keV) of Re foil (red), ReO2 (blue), 105 ReO3 (green), and Re2O7 (black).

Figure 4-13 Magnitude of k2-weighted FT of EXAFS data of (a) Rh K-edge of RhRe/C 110 catalyst reduced at 393 K (black), 523 K (red), and 723 K (blue) (Δk = 2.5 – -1 13.8 Å ). Re LIII edge of RhRe/C catalyst reduced at (b) 298 K (black), 313 K (red), and 333 K (blue) (Δk = 2.8 – 11.9 Å-1), and (c) 393 K (black), 523 K (red), and 723 K (blue) (Δk = 2.7 – 13.3 Å-1). All data was collected at room temperature.

Figure 4-14 (a) Rh K-edge XANES (23.19 – 23.27 keV) of RhRe/C reduced at 523 K 111 with spectra acquired under in situ conditions (black), and RhRe/C reduced at 523 K with spectra acquired under operando C-O hydrogenolysis

conditions (red). (b) Re LIII-edge XANES (10.51 – 10.56 keV) of RhRe/C reduced at 523 K with spectra acquired under in situ conditions (black), and RhRe/C reduced at 523 K with spectra acquired under operando C-O hydrogenolysis conditions (red).

Figure 4-15 Rate of hydrogenolysis of HMTHP over RhRe/C with varying THF/ water 114 mixture compositions under batch reaction conditions. Catalyst pretreated under 34 bar H2 at 393 K prior to introduction of liquid feed. Selectivities to 1,6-HDO were >90% for all data points. Reaction conditions: 5 wt% HMTHP in THF/ water mixtures as feed, mass ratio of catalyst:HMTHP = 1:7, 4 h, 393 K, 34 bar H2.

Figure 4-16 Rate of 1,6-hexanediol (1,6-HDO) formation over 4 wt% RhRe/C (1:0.5) as 118 a function of water concentration. Solvents were mixtures organic solvents with water. Reactions were performed under batch conditions with 5 wt% HMTHP in water as feed, 393 K, 500 psi H2, 4 h, dry catalyst pretreated under 500 psi H2 at 393 K prior to introduction of liquid feed. Selectivities to 1,6-HDO were more than 95% in all experiments.

Figure 5-1 Reaction pathways for the hydrogenation of unsaturated ketones (R2 = H) 122 or aldehydes (R2 = CxHy) (1) to unsaturated alcohols (2), saturated ketones or aldehydes (3), and saturated alcohols (4).

Figure 5-2 Mechanism for the Meerwein-Ponndorf-Verley reaction proceeding 125 through a six-membered transition state (2) over a metal alkoxide. Adapted from Chuah et al. and de Graauw et al.15, 17

Figure 5-3 Catalytic transfer hydrogenation of levulinic acid (1, R1 = H) and its esters 127 (1, R1 = CxH2x+1) to γ-valerolactone (5) using a secondary alcohol as the hydrogen donor (2, R2 = CyH2y+1).

Figure 5-4 Catalytic transfer hydrogenation of BL to GVL over ZrO2 in IPA at varying 134 water concentrations. Reaction conditions: batch reactions, 5 wt% BL in IPA/water as feed, 4 h, mass ratio catalyst:BL = 1:5.

Figure 5-5 Plot of γ-valerolactone (GVL) formation rate (■) and yield of GVL (○) as a 135 function of time-on-stream for the catalytic transfer hydrogenation of butyl levulinate (BL) to GVL over ZrO2 in a continuous flow reaction system. Reaction conditions were 5 wt% BL in 2-butanol as feed, WHSV = 0.18 h- 1, 423 K, 300 psi He. The catalyst was regenerated in-situ by calcination in 60 cc (STP)/min flowing air at 723 K for 4 h, at approximately 150 and 300 h, as indicated by the dotted lines.

Figure 6-1 Reactions discussed in this chapter. Compounds are as follows: 4-hydroxy- 143 6-methyl-2-pyrone/ triacetic acid lactone (1); 2,4-pentanedione/

xvi acetylacetone (2); 5,6-dihydro-4-hydroxy-6-methyl-2H-pyran-2-one (3); 3- penten-2-one (4); 4-hydroxy-2-pentanone (5); 4-hydroxy-6- methyltetrahydro-2-pyrone (6); 6-methyl-5,6-dihydro-2-pyrone/ parasorbic acid (7); 2,4-hexadienoic acid/ sorbic acid (8); 1,3-pentadiene (9); δ- hexalactone (10); hexenoic acid (11); γ-caprolactone (12).

Figure 6-2 Ring-opening of 10 to 11 and 12. ΔG° and ΔH° (in parentheses) for each 148 reaction in kJ/mol is shown at standard conditions.

Figure 7-1 (a) Reactions discussed in this chapter. Compounds are as follows: 4- 157 hydroxy-6-methyl-2-pyrone/ triacetic acid lactone (1); 2,4-pentanedione (2); 5,6-dihydro-4-hydroxy-6-methyl-2H-pyran-2-one (3); 3-penten-2-one (4); 4-hydroxy-pentanone (5); 4-hydroxy-6-methyltetrahydro-2-pyrone (6); parasorbic acid (7); sorbic acid (8); 1,3-pentadiene (9); (b) Nomenclature for the ring-carbon and ring-oxygen atoms of 1.

Figure 7-2 (a) Proposed mechanism for the ring-opening/ hydration and 160 decarboxylation of 1 to 2 in water, (b) DFT-calculated energy diagram for the reaction pathway of 1 to 2 in water, numbers indicate energy in kJ/mol, (c) measured rates of thermally-activated ring-opening and decarboxylation of 1 at various reaction temperatures (no catalyst). 21 bar He, space time = 70 min. Measured apparent activation energy barrier = 58 ± 12 kJ/mol (95% confidence interval).

Figure 7-3 Reactant, transition and product states in the (a) ring-opening of 1a to 1b, 161 and (b) decarboxylation of 1c to 1d. Bond lengths are given in Å. For clarity, only the local water molecules are shown.

Figure 7-4 (a) Proposed mechanism for the ring-opening and decarboxylation of 3 in 163 water, (b) DFT-calculated energy diagram for the reaction pathway of 3 to 4 in water, numbers indicate energy in kJ/mol, (c) rates of thermally- activated ring-opening and decarboxylation of 3 at various reaction temperatures in water (no catalyst). 21 bar He, space time = 13 min. Measured apparent activation energy barrier = 42 ± 18 kJ/mol (95% confidence interval), (d) rates of thermally-activated ring-opening and decarboxylation of 3 at various reaction temperatures in water over Amberlyst 70. 21 bar He, WHSV = 15 h-1. Measured apparent activation energy barrier = 18 ± 4 kJ/mol (95% confidence interval).

Figure 7-5 Reactant, transition and product state structures for the ring-opening and 165 decarboxylation of 3b in (a) gas phase, and (b) solution phase (27 water molecules/unit cell). Bond lengths are given in Å. For clarity, only the local water molecules are shown.

Figure 7-6 Structures of reactants, transition states and products of (a) un-catalyzed 167

xvii (without acid) tautomerization of 1 to 1a, and (b) acid-catalyzed tautomerization of 3d to 4. For clarity, only the local water molecules are shown.

Figure 7-7 (a) Proposed mechanism for the dehydration, ring-opening and 169 decarboxylation of 6 in water, (b) DFT-calculated energy diagram for the reaction pathway of 6 to 7 to 9 (solid line) and 6 to 6a to 9 (dashed line) in water, numbers indicate energy in kJ/mol, (c) reactant, product and transition state for the dehydration step of 6 to 7. Bond lengths are given in Å. (d) Ring-opening and decarboxylation of 3,6-dihydro-4,6,6-trimethyl- 2H-pyran-2-one (10) to 2,4-dimethyl-1,3-pentadiene (11), (e) DFT- calculated energy diagram for the reaction pathway of 10 to 11 in water.

Figure 7-8 Reactant, transition and product states for the dehydration step of 6 to 7. 169 Bond lengths are given in Å. For clarity, only the local water molecules are shown.

Figure 7-9 Structures of reactants, products and transition states of decarboxylation of 170 8 to 9. Bond lengths are given in Å. For clarity, only the local water molecules are shown.

Figure 7-10 Reactant, product and transition state structures for rDA reactions of (a) 6a 172 and (b) 10 in aqueous solution. Bond lengths are given in Å. For clarity, only the local water molecules are shown.

Figure 7-11 Overview of structure-reaction relationships for 2-pyrones. Abbreviations: 174 retro-Diels Alder (rDA), keto-enol tautomerization (KET), and apparent activation energy barrier (Eapp).

xviii List of Tables

Table 1-1 Top 30 industrial organic chemicals used in the US by weight produced.3 3

Table 1-2 Proposed top value-added chemicals from biorefinery carbohydrates. 22 Biologically-produced molecules are in bold.

Table 3-1 Effect of different catalyst compositions on catalytic activity and selectivity 47 of 2-(hydroxymethyl)tetrahydropyran 1 ring-opening to 1,6-hexanediol 2. The loadings of Rh, Ru, Pd and Pt were 4 wt%; all catalysts were supported on Vulcan XC-72.a

Table 3-2 Effect of metal-loading ratio on catalytic activity and hydrogenolysis 48 selectivity of 2-(hydroxymethyl)tetrahydropyran 1 to 1,6-hexanediol 2 over promoted Rh and Pt catalysts. The loadings of Rh and Pt were 4 wt%; all catalysts were supported on Vulcan XC-72a.

Table 3-3 Effect of different metal oxide precursors on catalyst activity and the 49 selectivity of 2-(hydroxymethyl)tetrahydropyran 1 ring-opening to 1,6- hexanediol 2. The loading of Rh was 4 wt%; atomic ratio of Rh:Mo = 1:0.5. all catalysts were supported on Vulcan XC-72.a

Table 3-4 Support effects on the selectivity of 2-(hydroxymethyl)tetrahydropyran 1 50 ring-opening to 1,6-hexanediol 2. The loadings of Rh and Pt were both 4 wt%; M1:M2 = 1:0.5 (mol:mol, M1 = Rh, Pt, M2 = Re, Mo).a

Table 3-5 Hydrogenolysis of tetrahydrofurfuryl alcohol 3 and 2- 51 (hydroxymethyl)tetrahydropyran 1 over Re- and Mo-promoted Rh catalysts and mono-metallic catalysts.a

Table 3-6 Quantification of CO adsorption on monometallic and promoted Rh 56 catalysts at 298 K.

Table 3-7 Effect of catalyst pretreatment temperature on hydrogenolysis activity of 2- 58 (hydroxymethyl)tetrahydropyran 1 over 4 wt% RhRe/C (1:0.5) and extent of rhenium leaching.a

Table 3-8 Hydrogenolysis of cyclic ethers and polyols over 4 wt% RhRe/C (1:0.5).a 60

Table 3-9 DFT-calculated gas phase reaction energies for the formation of carbenium 66 and oxocarbenium ion intermediates involved in the ring-opening of cyclic ethers and dehydration of polyols.

Table 3-10 Comparison of specific hydrogenolysis rates over 4 wt% RhRe/C (1:0.5) 70

xix and DFT-calculated carbenium ion energies for various cyclic ethers and polyols.

Table 3-11 Hydrogenolysis rates of 2-(hydroxymethyl)tetrahydropyran 1 to 2 over 4 74 wt% RhRe/C (1:0.5) in a continuous flow reaction system.a

Table 3-12 Effect of homogeneous acid and base on hydrogenolysis rate and product 77 selectivity of 2-(hydroxymethyl)tetrahydropyran 1 to hexanediols over Rh- containing catalysts.a

Table 4-1 Effect of pretreatment temperature on catalytic activity for C-O 89 hydrogenolysis of HMTHP to 1,6-HDO and amount of Re leached into solution under batch reaction conditions over RhRe/C.a

Table 4-2 Effect of pretreatment temperature on NH3 titrated site density and catalytic 93 activity for hydrogenolysis of HMTHP to 1,6-HDO over RhRe/C (1:0.5) under continuous flow reaction conditions.a

Table 4-3 Fit of Rh K and Re LIII-edge XANES and EXAFS for monometallic Rh/C 103 and Re/C catalysts reduced at varying reduction temperatures.a

Table 4-4 Fit of Rh K and Re LIII-edge XANES and EXAFS for Rh and Re standards, 108 RhRe/C catalyst reduced at varying reduction temperatures.a

Table 5-1 Catalytic transfer hydrogenation of levulinate esters using various alcohols 129 as the hydrogen donor.a

Table 5-2 Catalytic transfer hydrogenation of ethyl levulinate over ZrO2 in the 131 presence of sec-butyl phenol with varying molar ratios of isopropanol to ester.a

Table 5-3 Catalytic transfer hydrogenation of levulinic acid and levulinate esters 133 using 2-butanol as the solvent and hydrogen donor.a

Table 6-1 Ring-opening and decarboxylation of 1 to 2.a 144

Table 6-2 Ring-opening and decarboxylation of 3 to 4.a 146

Table 6-3 Hydrogenation of 1 to 6 over 10 wt% Pd/C.a 146

Table 6-4 Dehydration of 6 to 7 over Amberlyst 70.a 147

Table 6-5 Ring-opening of 10 to 11 and 12 over Amberlyst 70.a 148

xx Table 6-6 Dehydration of 6 and ring-opening of 7 over Amberlyst 70.a 149

Table 7-1 Results for batch reactions with 1, 3 and 6 as reactants in water and THF 158 solvents.a

Chapter 1 Introduction

1.1 Current production and consumption of fuel and chemicals

Fossil fuels have been and still are the primary feedstock for the production of fuel and chemicals. In 2011, global energy consumption was dominated by crude oil, natural gas, and coal

(Figure 1-1a, 87%), while renewable sources such as hydroelectric, wind power, solar electric, and bio-fuels contributed to only 9% of global consumption1. Similarly, energy consumption in the US in 2011 was largely based on fossil fuels (Figure 1-1b, 83%), and only 9% was attributable to renewable (non-nuclear) sources2.

Figure 1-1 (a) Global energy consumption by fuel in 2011.1 Abbreviations “GTL” and “CTL” refer to gas-to-liquid and coal-to-liquid, respectively. (b) 2011 US primary energy consumption estimates by source2. All numbers are in units of quadrillion Btu.

A breakdown of the supply sources and corresponding demand sectors for the US primary energy consumption in 2011 is shown in Figure 1-2. The transportation sector relies heavily on petroleum sources (93%), while renewable sources account for only 4% of the sector2.

Natural gas is equally distributed as a source of energy between the industrial, residential and

2 commercial, and electric power sectors at 31-33% per sector. Both coal and nuclear sources are largely used to generate electric power. As shown in Figure 1-2, renewable and nuclear electric sources account for 9% and 8% of total energy consumption in the US, respectively.

Figure 1-2 Primary energy consumption in the US by source and sector, 20112. Numbers are in units of quadrillion Btu. “Renewables” refers to conventional hydroelectric power, geothermal, solar/ photovoltaic, wind power, and biomass.

The production of fuel and chemicals are closely related as more than 90% of industrial organic chemicals produced are derived from crude oil and natural gas. The chemicals industry is historically operated on a platform chemical approach3, where more than 90% of organic chemicals are obtained from seven basic chemical building blocks derived from petroleum refining, namely, methane (or methanol), ethylene, propylene, C4 olefins (i.e., butenes and butadiene), benzene, toluene, and xylenes4. Methane (or methanol) is obtained from natural gas, the olefins (i.e., ethylene, propylene, and C4 olefins) from steam and catalytic cracking of crude

3 oil, and the aromatics from catalytic reforming of crude oil over Pt catalysts modified with Re or

Ir. The US produces over 500 billion lbs of organic chemicals annually, of which ethylene is the base for more than 200 billion lbs of chemicals and polymers, making it by far the most important chemical building block4. It is notable that the development of the chemicals industry was based on the need to utilize light hydrocarbon by-products from crude oil refining.3 These light ends undergo thermal cracking to produce ethylene, propylene, and benzene, from which approximately 75% of all organic chemicals are currently produced. Much of the conversion technologies in the chemicals industry therefore involve the introduction of functionality to these building blocks such as the incorporation of heteroatoms through oxidation, amination, and chlorination, alkylation, and oligomerization reactions.4 Table 1-1 lists the top 30 industrial organic chemicals used in the US, and the high utility of the seven platform chemicals as evidenced from their presence in this list underscores their importance as key chemical building blocks.

Table 1-1 Top 30 industrial organic chemicals used in the US by weight produced.3

Ethylene Phenol Propylene Acetic acid Ethylene dichloride Acrylonitrile Methanol α-olefins Vinyl chloride Propylene oxide Benzene Vinyl acetate Ethylbenzene Cyclohexane Styrene Acetone Terephthalic acid Acrylic acid Formaldehyde Adipic acid Ethylene oxide Nitrobenzene p-xylene Bisphenol-A Cumene n-butanol Ethylene glycol Caprolactam Butadiene Aniline

4 As shown from a typical breakdown of components per barrel of crude oil in Figure 1-3, it is evident that while crude oil is largely used for the production of liquid fuels, a small but significant portion is channeled into the production of chemicals. It is notable that although chemicals utilize a relatively small volumetric fraction of crude oil, the sale of chemicals generates revenues in the US of around USD 375 billion/ yr which is comparable to that from high volume, low value fuels (approximately USD 385 billion/ yr)5. Consequently, the profitability of oil refineries is closely tied to that of chemicals.

Figure 1-3 Breakdown of components per barrel of crude oil. Numbers are in units of gallons/ barrel.

1.2 Future outlook for fossil-based resources

The rise in global demand for fossil-based resources is driven mainly by population and income growth. Estimates indicate that by 2030, the world population would reach 8.3 billion and global income is projected to be double that in 2011 in real terms. 90% of this population

5 growth is expected to originate from low and medium income economies outside the

Organization for Economic Co-operation and Development (OECD), and rapid industrialization of these economies are expected to account for 70% global GDP growth and more than 90% of global energy demand increases1. Non-OECD energy consumption is estimated to grow at a rate of 2.5% p.a., and reach levels 61% above that in 2011 by 2030. Conversely, energy consumption in the OECD in 2030 is expected to be only 6% higher than that in 2011. The growth in primary energy consumption is largely attributable to increases in power generation, with renewable energy (including bio-fuels) projected to be the most rapidly growing fuel sector at 7.6% p.a. in the period of 2011 to 2030. It is also expected that while the highest growth in energy consumption will originate from non-OECD economies, the main increases in energy production would also arise from these countries and overall world primary energy production will match consumption at a growth rate of 1.6% p.a..

The global fuel mix is projected to change based on rising oil prices, implementation of new technology, and policy changes. In particular, tight oil and shale gas are expected to possess an increased share of primary energy sources, with shale gas trebling and tight oil increasing six- fold in production for the period of 2011 to 2030. This increased importance of tight oil and shale gas follows a continued and long decline in the market share of oil. In general, fossil fuel prices are currently at record levels: oil, coal, and gas prices over the period of 2007 to 2011 were 220%, 141%, and 95% above the average for the period of 1997 to 2001, respectively.1

Although energy production is expected to grow, studies indicate that global production of liquid fuels from conventional and non-conventional sources is likely to peak within the next decade.6 This depletion of crude oil resources and continual decline in its market share in the

6 energy sector is likely to impact chemicals production significantly due to the close ties between fuel and chemicals production from crude oil, as previously described. Additionally, the recent changes in East Asia and Middle East towards the use of lighter cracker feeds is likely to lead to critical shortfalls in chemicals such as butadiene.7 Therefore, alternative methods and feedstock for the production of chemicals will be needed in the near future.

1.3 Biomass as feedstock for the production of fuel and chemicals

The development of sustainable technologies for the production of carbon-based chemicals and alternative energy has gained tremendous importance with increasing concern about environmental change, and rising energy demand. These drivers behind the interest in biomass as an alternative feedstock for carbon based fuel and chemicals are directly related: it is now widely accepted that the increase in global temperatures since pre-industrial times (ca. 1750) is positively correlated with the rise in release of anthropogenic greenhouse gases (GHG) due to

8 9 fossil fuel use . Also, there is consensus that the continued release of GHG, particularly CO2 , has to be substantially mitigated to prevent disastrous changes in global climate. Apart from climate change issues, studies also indicate that global production of oil and from conventional and non-conventional sources is likely to peak within the next decade as described in Section

1.26. A critical need therefore exists to develop technologies capable of using renewable resources to displace fossil-based feedstock as the primary source of liquid fuels and chemicals.

There are three general classes of biomass: starches, triglycerides, and lignocellulosic biomass.10 Starches are biopolymers of glucose linked by α-glycosidic bonds, and may be readily hydrolyzed into their monomers; these have been extensively used as feedstock for the production of first generation bio-fuels (i.e., bio-ethanol11) from starchy components of food

7 crops such as wheat, corn, and sugar cane. Triglycerides can be derived from both plant and animals sources (e.g., vegetable oils, animal fat, algae), and is used as feedstock for the production of biodiesel through the transesterification (i.e., alcoholysis) of lipids with an alcohol

(e.g., methanol) to form a mixture of mono-alkyl esters of long-chain fatty acids (e.g., Fatty Acid

Methyl Ester (FAME)), forming glycerol as a byproduct. Transesterification is conventionally carried out using homogeneous bases such as NaOH and KOH.12, 13 The use of edible starches and triglycerides as feedstock for the production of fuel and chemicals is generally regarded as non-sustainable and not scalable due to encroachment of the demand for energy and chemicals on limited food resources. Additionally, processes based on edible starches and triglycerides as feedstock are consistently more costly compared to fossil-based ones. Consequently, recent effort in the utilization of starches and triglycerides has been to exploit sources of waste, such as non-edible oils and waste cooking oil, as alternative feedstock.14

Lignocellulosic biomass is the inedible portion of biomass and most abundant of the three sources. As shown in Figure 1-4, lignocellulosic biomass is composed of approximately 40-50% cellulose (glucose polymer with β-glycosidic bonds), 25-35% hemicelluloses (amorphous

15 heteropolymer of C6 and C5 sugars), and 15-20% lignin (amorphous phenolic polymer). While abundant and inexpensive, lignocellulosic biomass is difficult to fractionate and selective conversion strategies require pretreatment steps for the isolation of its components prior to hydrolysis to its monomeric sugars. Pretreatment methods employ a combination of physical

(e.g., milling) and chemical processes (e.g., acid hydrolysis), and are among the most costly steps in the utilization of lignocellulosic biomass as a feedstock for fuel and chemicals.15

Figure 1-4 General composition15 of lignocellulosic biomass and representative structures of constituent components.

Lignocellulosic biomass may be converted non-selectively using thermochemical methods, or selectively through catalytic means (i.e., aqueous phase processing) (Figure 1-5).

The thermochemical routes, i.e., gasification and pyrolysis, facilitate the deconstruction of whole lignocellulosic biomass (i.e., cellulose, lignocelluloses, and lignin fractions) using direct application of heat16, and are relatively inexpensive methods compared to catalytic routes. In gasification, a controlled amount of oxygen and/or steam is present during heating, while for pyrolysis, thermal decomposition is carried out in the absence of oxygen. Gasification occurs at temperatures higher than 773 K, resulting in the production of synthesis gas (i.e., CO and H2) which is a versatile feedstock for the production of methanol, aldehydes/ alcohols from olefins

(i.e., hydroformylation), and hydrocarbons through the Fischer-Tropsch process4. The reactions

9 that occur during gasification include partial oxidation of carbon, complete combustion of carbon, gasification of carbon with CO2, steam and water, water-gas shift, and methanation. Pyrolysis occurs at a temperature range of 573-600 K and involves reactions such as dehydration, cracking, isomerization, dehydrogenation, coking, condensation and aromatization. Depending on the heating rate employed, pyrolysis may be broadly classified as slow or fast 17. Slow pyrolysis is a method conventionally used to make charcoal, and uses a low temperature and heating rate to achieve high char yields. Fast pyrolysis requires high heating rates (103-104 K/s) and rapid quenching of products to obtain a complex mixture consisting of more than 300 oxygenated compounds called bio-oil18, 19. Bio-oil has to be upgraded either by hydrodeoxygenation or catalytic cracking to obtain fuel range hydrocarbons; due to high liquid yields and low cost, fast pyrolysis is regarded as one of the more promising methods to obtain liquid fuels from biomass19-22.

In the following chapters, we investigate means to convert molecules produced from lignocellulosic biomass (e.g., cyclic ethers, polyols) to chemicals. Such molecules may be obtained through the selective conversion of lignocellulosic biomass (Figure 1-5), which is more generally known as the aqueous phase-processing method or hydrolysis pathways10. Unlike thermochemical methods, current aqueous phase-processing technologies focus on the utilization of the cellulose and hemicellulose fractions of lignocellulosic biomass; lignin is not as amenable to selective deconstruction under aqueous phase-processing conditions and remains a difficult fraction to upgrade. C6 and C5 sugars may be obtained by the catalytic hydrolysis of lignocellulosic biomass using acid catalysts or through enzymatic means23-25 (Figure 1-5), and these sugars can then be selectively converted to a handful of intermediates through catalytic means. Unlike thermochemical methods, hydrolysis pathways therefore result in the formation of

10 a few specific products, i.e., platform molecules, such as 5-hydroxymethylfurfural, levulinic acid,

γ-valerolactone, furfural, organic acids, or 2-pyrones (Figure 1-5). Thus, these catalytic routes are inherently suited for the production of biorenewable platform chemicals in that high specificity of the target product and high carbon efficiency may be achieved.

Figure 1-5 Illustration of the conversion of biomass through thermochemical and aqueous phase processing (i.e., hydrolysis) strategies to fuel and platform chemicals.

It is notable that unlike that in strategies for alternative energy, the use of biomass as a source of renewable carbon remains the sole option available for the sustainable production of carbon-based chemicals that are used in the manufacture of products pervasive in today’s society.

The production of bio-based chemicals is strategically sound for many reasons:26 (1) chemicals

11 are relatively high value-added commodities and therefore more likely to overcome economic barriers of entry than bio-fuels when compared on a carbon basis; (2) industrial chemicals are currently manufactured using methods that frequently involve the co-production of

27 28 comparatively large amounts of waste (Environmental Factor ~ 1-5) and high CO2 emissions , and the development of alternative bio-based technologies would greatly reduce the environmental footprint of the chemicals industry; (3) chemicals are produced and consumed on a scale compatible with the potential capacity of biomass production and conversion. Based on estimates by the USDA and DOE29, it is evident that biomass can theoretically replace petroleum as a feedstock for the chemical industry30, 31. On a global context, the use of biomass as a source of renewable carbon for chemicals is also regarded as the most economically feasible and non- wasteful strategy for the use of limited amounts of renewable carbon32-34; (4) many chemicals and their intermediates possess some degree of oxygenation, implying that extensive chemical transformation of biomass will not be necessary and the inherent functionality of biomass can be advantageously exploited.

Figure 1-6 Illustration of the fossil- and biomass-based chemicals industries and the critical role of catalysis in both.

12 In contrast to petroleum, biomass is overfunctionalized (i.e., oxygenated), and its use requires pretreatment and catalytic strategies unique to its chemistry. This fundamental difference in feedstock chemistry implies that technologies developed and optimized for the petroleum-based chemicals industry cannot be directly translated and a completely new “tool chest” of catalysts and processes are needed. Although specific technologies are not immediately transferable, the general operational framework of the petroleum-based chemicals industry is likely to translate well to a bio-based one. Specifically, it has been suggested that the opportunities for acceptance of a bio-based chemicals industry are very much enhanced if the platform chemical and direct-replacement approach is taken3 , where end products identical to those produced in petroleum refineries today are to be made from biomass through a handful of key intermediates (Figure 1-6). As previously discussed, the aqueous-phase processing of lignocellulosic biomass results in the selective production of a handful of chemical intermediates and therefore would be very amenable to the implementation of this direct-replacement, platform chemical approach.

As historically shown in the petroleum-based chemicals industry, catalysis will undoubtedly play a central role in the success of a bio-based strategy (Figure 1-6); the use of catalysis is imperative for achieving high efficiencies and productivity levels in chemical transformations which ultimately translates into economic viability of production processes.

However, unlike that in conventional petrochemical processes, the catalytic routes for biomass conversion often require condensed/ liquid-phase reaction conditions, accompanied with high amounts of water due to the high hydrophilicity of biomass and its derivatives. Additionally, the unique structure of lignocellulosic biomass requires catalysts which are selective for C-O scission and formation reactions which are historically of relatively minor importance, and thus

13 less developed in classical petrochemical literature. Consequently, an urgent need exists for the development of new catalysts and strategies tailored for the processing of biomass and its derivatives.

1.4 Catalytic strategies for the production of chemicals from biomass-based

feedstocks

1.4.1 5-hydroxymethylfurfural as a platform chemical

5-hydroxymethylfurfural (HMF) is a versatile platform chemical for the production of fuel (e.g., liquid alkanes35, diesel36, 2,5-dimethylfuran37) and chemical intermediates38 for fine chemicals and plastics. Recent work in the literature has demonstrated that good yields of HMF may be obtained from glucose or fructose using monophasic39, 40 or biphasic41-43 reaction systems.

Solvents used as the extracting layer include methylisobutylketone, 2-butanol44, sec- butylphenol42, and γ-valerolactone43. Homogeneous or heterogeneous acid catalysts (e.g., HCl,

40 H2SO4, sulfonated ion-exchange resins or mesoporous silica ) and small amounts of phase modifiers (dimethylsulfoxide or poly(1-vinyl-2-pyrrolidinone)) have also been effective in increasing the overall yield of HMF from sugars. Reports in the literature have also demonstrated that polysaccharides such as starch and cellobiose can be converted to HMF with good yields.

Recent reports in literature indicate that HMF may be directly obtained from cellulose and lignocellulosic biomass (e.g., corn stover) using solvents such as ionic liquids with chloride salts

45 (e.g., CrCl2) as catalysts.

As described by Nicolau et al.3 and Christensen et al.28, either the direct or functional replacement approach can be undertaken in production of biorenewable chemicals. In the direct replacement approach, biomass-derived chemicals identical to current petrochemicals are

14 produced, while the functional replacement approach seeks to manufacture novel molecules that could be substitutes for current petrochemicals. The strategies for obtaining chemicals from

HMF and its derivatives span both the direct and functional replacement approaches (Figure

1-7). For instance, HMF-derived furans such as 2,5-furandicarboxylic acid (FDCA) have been recognized as one of the top value-added chemicals obtainable from glucose for the production of furanic polymers38. The alternative direct replacement approach which is described in

Chapters 3 and 4, is to obtain molecules, for example 1,6-hexanediol (1,6-HDO, Figure 1-7), which are identical to those currently utilized by the chemicals industry; α,ω-diols such as 1,6-

HDO are especially useful in the production of high volume polymers like polyesters and polyurethanes.

As seen in Figure 1-7, other chemicals vital in the production of polymers may be derived from 1,6-HDO, specifically adipic acid and hexamethylenediamine (HMDA). It is notable that the current production and use of α,ω-functionalized linear compounds such as adipic acid, HMDA, and 1,6-HDO are also closely linked, and typically starts with the oxidative cleavage of cyclohexane to a mixture of cyclohexanol and cyclohexanone, followed by further

4 oxidation of these intermediates to adipic acid using HNO3 . Adipic acid is then hydrogenated over Cu-, Co-, or Mn- catalysts to 1,6-HDO. HMDA is mostly produced through the reduction amination of adipic acid. Therefore, as shown in Figure 1-7, 1,6-HDO, adipic acid, and HMDA can be produced from biorenewable HMF instead of cyclohexene, in a closely linked manner that mimics current industrial routes. The main technological challenge faced in the implementation of this biorenewable route lies in identifying catalysts capable of performing these reactions with high selectivity and productivity under conditions suitable for scale-up,

15 thereby establishing sufficient economic competitiveness with present fossil-based routes (i.e., cyclohexene-based strategies).

OH OH 1 O OH 2 OH 5 O O

NH NH OH OH 2 3 2 4 O O O O

NH2 OH 7 6 O O

NH2 OH

Figure 1-7 Possible chemicals from 5-(hydroxymethyl)furfural (HMF, 1). Compounds are as follows: 2,5-(dihydroxymethyl) tetrahydrofuran (DHMTHF, 2); 2,5-bis(aminomethyl)-furan (3); 2,5-furandicarboxylic acid (FDCA, 4); 1,6-hexanediol (1,6-HDO, 5); adipic acid (6); hexamethylenediamine (HMDA, 7).

A catalytic route for the production of 1,6-HDO from HMF is shown in Figure 1-846. In this four-step strategy, HMF is hydrogenated over a metal catalyst to 2,5-

(dihydroxymethyl)tetrahydrofuran (DHMTHF). DHMTHF then undergoes selective ring- opening to 1,2,6-hexanetriol, which subsequently ring-closes to the stable cyclic intermediate, 2-

(hydroxymethyl)tetrahydropyran (HMTHP). Finally, selective ring-opening of HMTHP yields the desired end product, 1,6-HDO. While the steps from HMF through to HMTHP occur at relatively high selectivities over metal47 and acid catalysts46, the fourth step in which HMTHP is

16 ring-opened to 1,6-HDO does not occur selectively over conventional mono-metallic catalysts.

To address this bottleneck, Chapters 3 and 4 will detail our study of a rhenium-promoted rhodium catalyst supported on carbon which performs highly selective C-O hydrogenolysis of

HMTHP, and the elucidation of the nature of the active site over this catalyst.

Figure 1-8 Proposed catalytic route for the production of 1,6-hexanediol (1,6-HDO, 5) from 5- (hydroxymethyl)-furfural (HMF, 1). Compounds are as follows: 2,5- (dihydroxymethyl)tetrahydrofuran (DHMTHF, 2); 1,2,6-hexanetriol (3); 2-(hydroxymethyl) tetrahydropyran (HMTHP, 4).

1.4.2 Bimetallic catalysts and applications for biomass conversion Bimetallic catalysts often display electronic and chemical properties that are distinct from their parent metals, and the accompanying enhancements in catalytic selectivity, activity, and stability have been an area of much study. Pioneering contributions to the use of bimetallic catalysts in the area of petrochemistry, specifically hydrogenolysis and dehydrogenation of alkanes over Cu bimetallic catalysts, were by Sinfelt et al.48, 49. Currently, some of the most commercially relevant petrochemical refining reactions are performed over bimetallic catalysts, such as the catalytic reforming of naphtha fractions to aromatics over PtRe or PtIr catalysts

17 supported on alumina50. A vast number of reactions have been reported to occur over bimetallic catalysts at selectivities and activities different from their parent metals, such as oxidation51-54, hydrogenolysis37, 55-57, hydrogenation58-60, and reforming61-63 reactions, all of which are highly relevant for the conversion of biomass to fuel and chemicals.64 On a fundamental level, the unique changes in catalytic behavior upon addition of one metal to another to form an alloy have been attributed to modifications in the ensemble size, ensemble composition, and/ or a ligand/ electronic effect.58 These alloying effects are believed to result in an alteration in the configuration and binding strength of molecules (i.e., reactants, intermediates, transition states, products) adsorbed onto the catalyst surface, thereby affecting overall catalytic behavior. As such, given the latitude for modification of catalytic behavior through the manipulation of catalyst composition and particle morphology, bimetallic catalysts are a promising class of materials which can be tailored for the conversion of biomass-derived molecules.64

The mixing patterns between metals can vary greatly, and it is these structural differences that affect nanoparticle composition and consequently catalytic behavior. Ferrando et al. suggested that there are generally four possible mixing patterns between two metals in nanoparticles65: (a) core-shell segregated, (b) subclusters sharing a mixed interface, (c) ordered or random mixture, and (d) layered or onion-like alternating multishell. The type of mixing pattern is influenced by the relative strength of bonding between the two metals as opposed to the pure metals, the surface energy of the two metals, the relative atomic sizes, whether there is electron transfer between the two metals, and electronic or magnetic effects.65 Recently, it has suggested that the extent of modification of the parent metal’s properties is most pronounced when the admetal coverage is in the submonolayer to monolayer regime. Current work in density functional theory calculations66 (DFT) and ultrahigh vacuum experiments suggest that,

18 depending on experimental conditions, the admetal monolayer forms three different structures with the host metal: as a surface monolayer on the host (i.e., core-shell segregated), as a subsurface monolayer (i.e., core-shell segregated), and as an intermixed bimetallic surface.

Experimental methods to determine the surface composition of supported bimetallic

67 catalysts are varied, and include chemisorption techniques (e.g., CO and H2), temperature- programmed methods (e.g., desorption, oxidation, and reduction), transmission electron microscopy (e.g., high annular dark field for high Z-contrast, tomography), X-ray photoelectron spectroscopy, energy-dispersive X-ray spectroscopy, infrared spectroscopy, and X-ray absorption spectroscopy. Due to limitations in individual methods, a combination of these techniques is typically required in the determination of the average surface composition of supported bimetallic nanoparticles.

It is notable that the prediction of surface composition or compositional segregation of bimetallic particles from first principle DFT calculations is a challenge. The difficulties in obtaining accurate predictions lie in the lack of reliable experimental data for the improvement of existing theoretical models, and sensitivity of surface composition to exposed environment (e.g., temperature, pressure, gases) which can result in effects such as adsorbate-induced segregation.

Ruban et al.68 have used DFT calculations to estimate the surface composition of an extensive array of transition-metal alloys in vacuum through the computation of the surface segregation energy (Eseg), which is the energy required to move an impurity atom from the bulk to the surface of the host metal (i.e., difference in total energies in the systems with the impurity in the bulk

68 and on the surface). Ruban et al. reported that the surface composition of an alloy can be estimated based on the magnitude of Eseg. If Eseg is negative the impurity tends to segregate to the

19 surface, while a positive value of Eseg indicates that the impurity is likely to reside in the bulk.

Similarly, Christensen et al.69 used DFT to predict the stability of surface alloy phases through the construction of surface phase diagrams of surface energy as a function of surface composition. To better estimate surface composition in the presence of an adsorbate, Greeley et al.66 have reported methods to predict the tendency of hydrogen to induce surface segregation in various transition metal alloys. In these estimates, Eseg values were compared against the difference in hydrogen binding energies between the pure solute and pure host. If Eseg is positive and H binds more strongly to the host than the solute, then no hydrogen-induced segregation is expected occur.

Biomass-derived compounds are composed of a variety of C-C and C-O bonds, and achieving chemoselectivity in reactions involving C-C and C-O bond scission, and C=O and

C=C bond hydrogenation are some of the most relevant issues in the production of biorenewable chemicals which have been addressed by bimetallic catalysts. Some general examples of the use of bimetallic catalysts for the conversion of biomass-derived compounds include: aqueous phase reforming of oxygenates and glycerol over NiSn70 and PtRe71, hydrogenolysis of glycerol, sugars, and furans over PtRe56, IrRe72, and CuRu37, hydrogenation of levulinic acid, furfural, and glucose over RuRe73, PtSn74, and NiMo75, and oxidation of sugars over AuPd53 and PdBi76.

Figure 1-9 shows the various reactions studied for the conversion of glycerol to chemicals and the various bimetallic catalysts reported in the literature that have been used to achieve chemoselectivity in these transformations. Glycerol has been extensively studied as a model compound due to its structural similarity to sugars, and reactions of specific interest such as C-O hydrogenolysis to form propanediols, oxidation to acids, and reforming have been found to proceed selectively over a number of bimetallic catalysts, as depicted in Figure 1-9.

Figure 1-9 Conversion of glycerol to various chemicals over bimetallic catalysts. Adapted from Alonso et al.64

Due to the complexities involved in establishing structure-function relationships in bimetallic catalysts, a multi-pronged approach combining experimental reaction kinetics, catalyst characterization, and theoretical calculations using DFT are presented in Chapters 3 and 4 for our studies on a bimetallic RhRe catalyst supported on carbon. We investigate the ability of this catalyst to perform selective C-O hydrogenolysis, and based on reactivity trends obtained for a wide range of polyols and cyclic ethers, a reaction mechanism is proposed. Results from DFT calculations are used to provide a basis for our proposed reaction mechanism, and several experimental catalyst characterization techniques are employed to determine the nature of the active site.

21 1.4.3 Catalytic upgrading of biologically-produced compounds There are several ways in which synergies between chemical and biological catalysis can be exploited for the production of biorenewable chemicals, such as the combination of chemical catalysts with enzymes77 or whole cells77 in a single pot to effect cascade reactions mimicking biological systems78, 79. Although one-pot methods are promising, difficulties in achieving effective partitioning of catalysts from one another and the limited range of suitable reaction solvents have relegated these techniques to highly niche applications and limited reaction classes, typically for the production of fine chemicals and pharmaceuticals.80, 81 Instead, an alternative strategy that has immediate potential for the production of large volume industrial chemicals is through the catalytic upgrading of biologically-produced molecules. This strategy is prominently featured in the US DOE 2004 report 38, in which six out of fifteen proposed target molecules from biorefinery carbohydrates are biologically-produced organic acids (Table 1-2).

More recently, an update to the US DOE report by Bozell et al.82 identified ten carbohydrate-derived molecules as the most likely targets to be produced commercially; as shown in Table 1-2, five out of ten of the molecules on the list are biologically-produced molecules, once more indicating that biologically-produced compounds are consistently regarded as highly promising targets from carbohydrates. These molecules were selected based on nine different criteria82: (1) the amount of research activity/ attention attracted to date, (2) broadness of technology application to other products, (3) potential for direct replacement of existing petrochemicals, (4) potential market impact of product (i.e., high or low volume product), (5) potential as a platform chemical, (6) whether existing scale-up from benchtop to pilot, demonstration or full-scale plant is underway, (7) whether the compound is an existing commercial product, (8) the compound’s ability to serve as a primary chemical building block,

22 and (9) whether the commercial production of the compound from renewable carbon is already well established.

Table 1-2 Proposed top value-added chemicals from biorefinery carbohydrates. Biologically- produced molecules are in bold.

2004 US Department of Energy report38 2010 report update by Bozell et al.82 Succinic, fumaric and malic acid Succinic acid 2,5-furan-dicarboxylic acid Furans 3-hydroxypropionic acid 3-hydroxypropionic acid or aldehyde Aspartic acid -- Glucaric acid -- Glutamic acid -- Itaconic acid -- Levulinic acid Levulinic acid 3-hydroxybutyrolactone -- Glycerol Glycerol and derivatives Sorbitol Sorbitol Xylitol/arabinitol Xylitol -- Lactic acid -- Ethanol -- Biohydrocarbons (e.g., isoprene)

Of the molecules proposed as biorenewable platform chemicals, lactic acid and ethanol are biologically-produced molecules that currently enjoy significant commercial success and substantial impact on the chemicals industry. Lactic acid may be obtained through the fermentation of sugars such as glucose (derived from starch), maltose (derived from starch), sucrose (derived from syrups and molasses), and lactose (derived from whey). As discussed by

Corma et al.33 (Figure 1-10), lactic acid is an excellent platform from which a diverse number of end products may be obtained through catalytic dehydration, hydrogenation, esterification, oxidation, and most importantly, polymerization to form poly(lactic acid). The commercial production of ethanol through fermentation of sugars is well-known, and bio-ethanol is currently blended with gasoline for use as a liquid transportation fuel. Ethanol may also be used as a precursor for the production of industrial chemicals such as ethylene, butanol83-86, and

23 butadiene87, 88, although these technologies have not yet been commercialized due to lack of cost competitiveness.

OH OH 2 Hydrogenation

6 O Hydrogenation OH Dehydration Esterification O OH OR 7 1 O O 3 O

Oxidation Esterification O O 8 O polylactic O acid O O OH 4

5 O

Figure 1-10 Conversion of lactic acid to chemicals. Compounds are as follows: lactic acid (1), 1,2-propanediol (2), lactates (3), lactide (4), pyruvic acid (5), acrylic acid (6), acetaladehyde (7), and 2,3-pentadione (8). Adapted from Corma et al.33

In line with this approach, Chapters 6 and 7 will detail our work in the upgrading of 2- pyrones, specifically, triacetic acid lactone (TAL), to various industrially relevant end products and chemical intermediates. TAL may be produced from glucose using genetically-modified

Escherichia coli or Saccharomyces cerevisiae through heterologous expression of 2-pyrone synthase89 or other genetically modified polyketide synthases90, 91. The work described in

Chapter 6 demonstrates that TAL is a promising platform chemical and may serve as a primary building block for the direct replacement of various industrial chemicals. The chemistries unique

24 to 2-pyrones are systematically investigated in Chapter 7 in order to develop structure-reactivity relationships that may find broader use in the manipulation of compounds with structures analogous to TAL.

1.5 Research overview and strategy Despite recent research in the catalytic transformation of biomass, the technological challenges faced in the transition to a bio-based chemicals industry remain formidable and unsolved. Some challenges include: relatively low catalytic efficiency of chemical transformations, lack of scalability and significant dependence on use of external hydrogen in proposed processes, need for fundamental understanding and exploitation of catalysis in non- conventional solvents, and insufficient and ineffective integration and exploitation of synergies between disciplines (e.g., biological and chemical catalysis). The work presented in the following chapters will attempt to address some of these issues, with an overarching emphasis on the study of reactions relevant to heterocyclics. This work will deal with the development of highly chemoselective bimetallic catalysts (Chapters 3 and 4), the use of non-conventional solvents as a hydrogen source (Chapter 5), and catalytic upgrading of a novel biologically- produced platform chemical (Chapters 6 and 7). Much of the material in this work deals with the deoxygenation of compounds through the selective scission of C-O bonds, either with the aid of molecular hydrogen (i.e., C-O hydrogenolysis, Chapters 3 and 4) or without (i.e., decarboxylation, Chapters 6 and 7); many of the molecules examined are heterocyclics which can be derived from biomass. An emphasis is placed on understanding on structure-reactivity relationships, either with regard to the catalyst structure in the case of bimetallic catalysts, or reactant structure in the study of 2-pyrones. The experimental methods generally applicable to

25 most of the work are presented in Chapter 2, and more chapter-specific experimental and computational methods may be found at the end of each relevant chapter.

1.6 References 1. BP Energy Outlook 2030 (www.bp.com) 2. United States Energy Information Adminstration 3. Nikolau, B. J., Perera, M. A. D. N., Brachova, L. & Shanks, B. Platform biochemicals for a biorenewable chemical industry. The Plant Journal 54, 536-545 (2008). 4. Weissermel, K. & H.-J., A. Industrial Organic Chemistry (Wiley-VCH, 2003). 5. Rinaldi, R. & Schuth, F. Design of solid catalysts for the conversion of biomass. Energy & Environmental Science 2, 610-626 (2009). 6. Boyle, G. Renewable Energy (Oxford University Press, 2004). 7. www.icis.com. 8. Solomon; et al. IPCC, 2007: Summary for Policymakers. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (Cambridge University Press, Cambridge, 2007). 9. Lacis, A. A., Schmidt, G. A., Rind, D. & Ruedy, R. A. Atmospheric CO2: Principal Control Knob Governing Earth's Temperature. Science 330, 356-359 (2010). 10. Alonso, D. M., Bond, J. Q. & Dumesic, J. A. Catalytic conversion of biomass to biofuels. Green Chemistry 12, 1493-1513 (2010). 11. Lynd, L. R., Cushman, J. H., Nichols, R. J. & Wyman, C. E. Fuel Ethanol from Cellulosic Biomass. Science 251, 1318-1323 (1991). 12. Melero, J. A., Iglesias, J. & Morales, G. Heterogeneous acid catalysts for biodiesel production: current status and future challenges. Green Chemistry 11, 1285-1308 (2009). 13. Chouhan, A. P. S. & Sarma, A. K. Modern heterogeneous catalysts for biodiesel production: A comprehensive review. Renewable and Sustainable Energy Reviews 15, 4378-4399 (2011). 14. Atabani, A. E. et al. A comprehensive review on biodiesel as an alternative energy resource and its characteristics. Renewable and Sustainable Energy Reviews 16, 2070- 2093 (2012). 15. Wyman, C. E. et al. Coordinated development of leading biomass pretreatment technologies. Bioresource Technology 96, 1959-1966 (2005). 16. de Wild, P. J., Uil, H. d., Reith, J. H., Kiel, J. H. A. & Heeres, H. J. Biomass valorisation by staged degasification: A new pyrolysis-based thermochemical conversion option to produce value-added chemicals from lignocellulosic biomass. Journal of Analytical and Applied Pyrolysis 85, 124-133 (2009). 17. Bridgwater, A. V. Catalysis in thermal biomass conversion. Applied Catalysis A: General 116, 5-47 (1994). 18. Furimsky, E. Catalytic hydrodeoxygenation. Applied Catalysis A: General 199, 147-190 (2000). 19. Wang, H. M., Male, J. & Wang, Y. Recent Advances in Hydrotreating of Pyrolysis Bio- Oil and Its Oxygen-Containing Model Compounds. Acs Catalysis 3, 1047-1070 (2013).

26 20. Zhang, S., Yan, Y., Li, T. & Ren, Z. Upgrading of liquid fuel from the pyrolysis of biomass. Bioresource Technology 96, 545-550 (2005). 21. Czernik, S. & Bridgwater, A. V. Overview of Applications of Biomass Fast Pyrolysis Oil. Energy & Fuels 18, 590-598 (2004). 22. Elliott, D. C. Historical Developments in Hydroprocessing Bio-oils. Energy & Fuels 21, 1792-1815 (2007). 23. Sun, Y. & Cheng, J. Hydrolysis of lignocellulosic materials for ethanol production: a review. Bioresource Technology 83, 1-11 (2002). 24. Dhepe, P. L. & Fukuoka, A. Cellulose Conversion under Heterogeneous Catalysis. ChemSusChem 1, 969-975 (2008). 25. Sahu, R. & Dhepe, P. L. A One-Pot Method for the Selective Conversion of Hemicellulose from Crop Waste into C5 Sugars and Furfural by Using Solid Acid Catalysts. ChemSusChem 5, 751-761 (2012). 26. Gallezot, P. Conversion of biomass to selected chemical products. Chemical Society Reviews 41, 1538-1558 (2011). 27. Sheldon, R. A. The E Factor: fifteen years on. Green Chemistry 9, 1273-1283 (2007). 28. Claus Hviid, C., Jeppe, R.-H., Charlotte , C. M., Esben, T. & Kresten, E. The Renewable Chemicals Industry. ChemSusChem 1, 283-289 (2008). 29. DOE & USDA, Biomass as Feedstock for a Bioenergy and Bioproducts Industry: The Technical Feasibility of a Billion-Ton Annual Supply (2005). 30. DOE. 31. Dodds, D. R. & Gross, R. A. Chemistry: Chemicals from Biomass. Science 318, 1250- 1251 (2007). 32. Lange, J.-P. in Catalysis for Renewables 21-51 (Wiley-VCH Verlag GmbH & Co. KGaA, 2007). 33. Corma, A., Iborra, S. & Velty, A. Chemical Routes for the Transformation of Biomass into Chemicals. Chemical Reviews 107, 2411-2502 (2007). 34. Michel, H. Editorial: The Nonsense of Biofuels. Angewandte Chemie International Edition 51, 2516-2518 (2012). 35. Chheda, J. N. & Dumesic, J. A. An overview of dehydration, aldol-condensation and hydrogenation processes for production of liquid alkanes from biomass-derived carbohydrates. Catalysis Today 123, 59-70 (2007). 36. Balakrishnan, M., Sacia, E. R. & Bell, A. T. Etherification and reductive etherification of 5-(hydroxymethyl)furfural: 5-(alkoxymethyl)furfurals and 2,5-bis(alkoxymethyl)furans as potential bio-diesel candidates. Green Chemistry 14, 1626-1634 (2012). 37. Roman-Leshkov, Y., Barrett, C. J., Liu, Z. Y. & Dumesic, J. A. Production of dimethylfuran for liquid fuels from biomass-derived carbohydrates. Nature 447, 982-985 (2007). 38. Werpy, T. & Petersen, G. Top Value Added Chemicals from Biomass. Volume I - Results of Screening for Potential Candidates from Sugars and Synthesis Gas. U. S. D. o. Energy (2004). 39. Tucker, M. H. et al. Sustainable Solvent Systems for Use in Tandem Carbohydrate Dehydration Hydrogenation. ACS Sustainable Chem. Eng. 1, 554-560 (2013).

27 40. Tucker, M. H. et al. Acid-Functionalized SBA-15-Type Periodic Mesoporous Organosilicas and Their Use in the Continuous Production of 5-Hydroxymethylfurfural. ACS Catalysis 2, 1865-1876 (2012). 41. Roman-Leshkov, Y., Chheda, J. N. & Dumesic, J. A. Phase modifiers promote efficient production of hydroxymethylfurfural from fructose. Science 312, 1933-1937 (2006). 42. Pagan-Torres, Y. J., Wang, T., 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 Catal. 2, 930-934 (2012). 43. Gallo, J. M. R., Alonso, D. M., Mellmer, M. A. & Dumesic, J. A. Production and upgrading of 5-hydroxymethylfurfural using heterogeneous catalysts and biomass- derived solvents. Green Chem. 15, 85-90 (2013). 44. Chheda, J. N., Roman-Leshkov, Y. & Dumesic, J. A. Production of 5- hydroxymethylfurfural and furfural by dehydration of biomass-derived mono- and polysaccharides. Green Chemistry 9, 342-350 (2007). 45. Dutta, S., De, S. & Saha, B. Advances in biomass transformation to 5- hydroxymethylfurfural and mechanistic aspects. Biomass and Bioenergy (2013). 46. Buntara, T. et al. From 5-Hydroxymethylfurfural (HMF) to Polymer Precursors: Catalyst Screening Studies on the Conversion of 1,2,6-hexanetriol to 1,6-hexanediol. Topics in Catalysis 55, 612-619 (2012). 47. Alamillo, R., Tucker, M., Chia, M., Pagan-Torres, Y. & Dumesic, J. The selective hydrogenation of biomass-derived 5-hydroxymethylfurfural using heterogeneous catalysts. Green Chemistry 14, 1413-1419 (2012). 48. Sinfelt, J. H., Carter, J. L. & Yates, D. J. C. Catalytic hydrogenolysis and dehydrogenation over copper-nickel alloys. Journal of Catalysis 24, 283-296 (1972). 49. Sinfelt, J. H. Supported "bimetallic cluster" catalysts. Journal of Catalysis 29, 308-315 (1973). 50. Carter, J. L., McVinker, G. B., Weissman, W., Kmak, M. S. & Sinfelt, J. H. Bimetallic catalysts; application in catalytic reforming. Applied Catalysis 3, 327-346 (1982). 51. Della Pina, C., Falletta, E. & Rossi, M. Highly selective oxidation of benzyl alcohol to benzaldehyde catalyzed by bimetallic gold-copper catalyst. Journal of Catalysis 260, 384- 386 (2008). 52. Sandoval, A., Aguilar, A., Louis, C., Traverse, A. s. & Zanella, R. Bimetallic Au- Ag/TiO2 catalyst prepared by deposition-precipitation: High activity and stability in CO oxidation. Journal of Catalysis 281, 40-49 (2011). 53. Ketchie, W. C., Murayama, M. & Davis, R. J. Selective oxidation of glycerol over carbon-supported AuPd catalysts. Journal of Catalysis 250, 264-273 (2007). 54. Marx, S. & Baiker, A. Beneficial Interaction of Gold and Palladium in Bimetallic Catalysts for the Selective Oxidation of Benzyl Alcohol. The Journal of Physical Chemistry C 113, 6191-6201 (2009). 55. Nakagawa, Y., Shinmi, Y., Koso, S. & Tomishige, K. Direct hydrogenolysis of glycerol into 1,3-propanediol over rhenium-modified iridium catalyst. Journal of Catalysis 272, 191-194 (2010). 56. Daniel, O. M. et al. X-ray Absorption Spectroscopy of Bimetallic Pt–Re Catalysts for Hydrogenolysis of Glycerol to Propanediols. ChemCatChem 2, 1107-1114 (2010).

28 57. Ruppert, A. M., Weinberg, K. & Palkovits, R. Hydrogenolysis Goes Bio: From Carbohydrates and Sugar Alcohols to Platform Chemicals. Angewandte Chemie International Edition 51, 2564-2601 (2012). 58. Ponec, V. On the role of promoters in hydrogenations on metals; α,β-unsaturated aldehydes and ketones. Applied Catalysis A: General 149, 27-48 (1997). 59. Mäki-Arvela, P., Hájek, J., Salmi, T. & Murzin, D. Y. Chemoselective hydrogenation of carbonyl compounds over heterogeneous catalysts. Applied Catalysis A: General 292, 1- 49 (2005). 60. Hermans, S. et al. Solvent-Free, Low-Temperature, Selective Hydrogenation of Polyenes using a Bimetallic Nanoparticle Ru–Sn Catalyst. Angewandte Chemie International Edition 40, 1211-1215 (2001). 61. Kunkes, E. L. et al. The role of rhenium in the conversion of glycerol to synthesis gas over carbon supported platinum-rhenium catalysts. Journal of Catalysis 260, 164-177 (2008). 62. Simonetti, D. A., Kunkes, E. L. & Dumesic, J. A. Gas-phase conversion of glycerol to synthesis gas over carbon-supported platinum and platinum-rhenium catalysts. Journal of Catalysis 247, 298-306 (2007). 63. Parera, J. M. & Beltramini, J. N. Stability of bimetallic reforming catalysts. Journal of Catalysis 112, 357-365 (1988). 64. Alonso, D. M., Wettstein, S. G. & Dumesic, J. A. Bimetallic catalysts for upgrading of biomass to fuels and chemicals. Chemical Society Reviews (2012). 65. Ferrando, R., Jellinek, J. & Johnston, R. L. Nanoalloys: From Theory to Applications of Alloy Clusters and Nanoparticles. Chemical Reviews 108, 845-910 (2008). 66. Greeley, J. & Mavrikakis, M. Alloy catalysts designed from first principles. Nat Mater 3, 810-815 (2004). 67. Yu, W., Porosoff, M. D. & Chen, J. G. Review of Pt-Based Bimetallic Catalysis: From Model Surfaces to Supported Catalysts. Chemical Reviews 112, 5780-5817 (2012). 68. Ruban, A. V., Skriver, H. L. & Norskov, J. K. Surface segregation energies in transition- metal alloys. Physical Review B 59, 15990-16000 (1999). 69. Christensen, A. et al. Phase diagrams for surface alloys. Physical Review B 56, 5822- 5834 (1997). 70. Shabaker, J. W., Huber, G. W. & Dumesic, J. A. Aqueous-phase reforming of oxygenated hydrocarbons over Sn-modified Ni catalysts. Journal of Catalysis 222, 180- 191 (2004). 71. King, D. L. et al. Aqueous phase reforming of glycerol for hydrogen production over Pt- Re supported on carbon. Applied Catalysis B: Environmental 99, 206-213 (2010). 72. Chen, K., Tamura, M., Yuan, Z., Nakagawa, Y. & Tomishige, K. One-Pot Conversion of Sugar and Sugar Polyols to n-Alkanes without C-C Dissociation over the Ir-ReOx/SiO2 Catalyst Combined with H-ZSM-5. ChemSusChem 6, 613-621 (2013). 73. Braden, D. J., Henao, C. A., Heltzel, J., Maravelias, C. C. & Dumesic, J. A. Production of liquid hydrocarbon fuels by catalytic conversion of biomass-derived levulinic acid. Green Chemistry (2011). 74. Merlo, A. B., Vetere, V., Ruggera, J. F. & Casella, M. n. L. Bimetallic PtSn catalyst for the selective hydrogenation of furfural to furfuryl alcohol in liquid-phase. Catalysis Communications 10, 1665-1669 (2009).

29 75. Gallezot, P., Cerino, P. J., Blanc, B., Fleche, G. & Fuertes, P. Glucose hydrogenation on promoted raney-nickel catalysts. Journal of Catalysis 146, 93-102 (1994). 76. Karski, S., Paryjczak, T. & Witonnska, I. Selective Oxidation of Glucose to Gluconic Acid over Bimetallic Pd-Me Catalysts (Me = Bi, Tl, Sn, Co). Kinetics and Catalysis 44, 618-622 (2003). 77. Andreana, P. R., McLellan, J. S., Chen, Y. & Wang, P. G. Synthesis of 2,6- Dideoxysugars via Ring-Closing Olefinic Metathesis. Organic Letters 4, 3875-3878 (2002). 78. Marr, A. C. & Liu, S. Combining bio- and chemo-catalysis: from enzymes to cells, from petroleum to biomass. Trends in Biotechnology 29, 199-204 (2011). 79. Bruggink, A., Schoevaart, R. & Kieboom, T. Concepts of Nature in Organic Synthesis: Cascade Catalysis and Multistep Conversions in Concert. Organic Process Research & Development 7, 622-640 (2003). 80. Mäki-Arvela, P. et al. Utilization of cascade chemo-bio catalysis for the synthesis of R - 1-phenylethyl acetate. Reaction Kinetics and Catalysis Letters 94, 281-288 (2008). 81. Långvik, O. et al. Dynamic Kinetic Resolution of rac-2-Hydroxy-1-indanone by using a Heterogeneous Ru(OH)3/Al2O3 Racemization Catalyst and Lipase. ChemCatChem 2, 1615-1621 (2010). 82. Bozell, J. J. & Petersen, G. R. Technology development for the production of biobased products from biorefinery carbohydrates-the US Department of Energy's "Top 10" revisited. Green Chemistry 12, 539-554 (2009). 83. Carlini, C. et al. Selective synthesis of isobutanol by means of the Guerbet reaction: Part 2. Reaction of methanol/ethanol and methanol/ethanol/n-propanol mixtures over copper based/MeONa catalytic systems. Journal of Molecular Catalysis A: Chemical 200, 137- 146 (2003). 84. Tsuchida, T. et al. Reaction of ethanol over hydroxyapatite affected by Ca/P ratio of catalyst. Journal of Catalysis 259, 183-189 (2008). 85. Kozlowski, J. T. & Davis, R. J. Sodium modification of zirconia catalysts for ethanol coupling to 1-butanol. Journal of Energy Chemistry 22, 58-64 (2013). 86. Leon, M., Diaz, E. & Ordonez, S. Ethanol catalytic condensation over Mg-Al mixed oxides derived from hydrotalcites. Catalysis Today 164, 436-442 (2011). 87. Kitayama, Y. & Michishita, A. Catalytic activity of fibrous clay mineral sepiolite for butadiene formation from ethanol. Journal of the Chemical Society, Chemical Communications, 401-402 (1981). 88. Makshina, E. V., Janssens, W., Sels, B. F. & Jacobs, P. A. Catalytic study of the conversion of ethanol into 1,3-butadiene. Catalysis Today (2012). 89. Eckermann, S. et al. New pathway to polyketides in plants. Nature 396, 387-390 (1998). 90. Xie, D. M. et al. Microbial synthesis of triacetic acid lactone. Biotechnology and Bioengineering 93, 727-736 (2006). 91. Zha, W., Shao, Z., Frost, J. W. & Zhao, H. Rational Pathway Engineering of Type I Fatty Acid Synthase Allows the Biosynthesis of Triacetic Acid Lactone from d-Glucose in Vivo. Journal of the American Chemical Society 126, 4534-4535 (2004).

Chapter 2 Experimental techniques

2.1 Catalyst preparation and synthesis methods

2.1.1 Supported metal catalysts

Supported metal catalysts were prepared by incipient wetness impregnation of their respective supports (e.g., Vulcan XC-72, Norit SX-1G, Cab-o-sil fumed silica) with aqueous solutions of RhCl3 (Mitsubishi), H2PtCl6(H2O)6 (STREM), Ru(NO)(NO3)3 (STREM),

Pd(NO3)2·xH2O (Aldrich), NH4ReO4 (Alfa Aesar), and (NH4)6Mo7O24·4H2O (Sigma). Where otherwise explicitly stated, Vulcan XC-72 was used as the support for RhRe/C catalysts. Rh/C catalysts with oxophilic additives were obtained by successive impregnation of dried, unreduced

Rh/C with the corresponding precursor of the oxophilic additive. The loading amount of the second metal is stated as a atomic ratio (e.g., 4 wt% RhRe/C (1:0.5) catalyst designates a metal atomic ratio of Rh:Re = 1:0.5). Prior to use in experiments, catalysts were dried in air (393 K), reduced in flowing H2 (723 K), and passivated with flowing 2% O2 in He (298 K). Reduced and passivated catalysts are described as “as-prepared” catalysts herein.

5 wt% Pd/Nb2O5 was prepared by incipient wetness impregnation of an aqueous solution of Pd(NO3)2·xH2O (Aldrich) with commercial niobia (HY-340, CBMM, Brazil). The catalyst was dried in air at 373 K, calcined in 60 cm3(STP)/min flowing air at 573 K for 5 h, reduced in

3 3 60 cm (STP)/min flowing H2 at 623 K for 5 h, and then passivated with 60 cm (STP)/min flowing 2% O2 in He for 2 h at room temperature. 10 wt% Pd/C (Alfa Aesar) was used as- received.

31 2.1.2 Metal oxide catalysts

The γ-Al2O3 catalyst was used as-received (STREM, >97%, low soda). The MgO/Al2O3 catalyst was obtained by calcination of hydrotalcite (Mg6Al2(CO3)(OH)16·4H2O, Aldrich) in

60 cc (STP)/min flowing air at 723 K for 4 h prior to use in batch experiments. The CeZrOx and

ZrO2 catalysts were prepared by precipitation of Ce(NO3)3·6H2O and ZrO(NO3)2 with NH4OH

(Aldrich), according to Serrano-Ruiz et al.1 Solids were calcined in 60 cc (STP)/min flowing air at 723 K for 4 h prior to use in batch experiments. The MgO/ZrO2 catalyst was prepared by precipitation of Mg(NO3)2·6H2O and ZrOCl2·6H2O with NaOH (Aldrich), according to

Aramendia et al.2 Solids were calcined in 60 cc (STP)/min flowing air at 873 K for 4 h prior to use in batch experiments.

2.1.3 Ion-exchange resins and zeolites

Amberlyst 70 and Amberlyst 15 (Rohm and Haas) were washed with deionized water, dried in air at 373 K, and mechanically crushed to a fine powder. The powders were then sifted through a sieve (standard US size 45) to remove large particulates. ZSM-5 (Engelhard, Si/Al =30) was calcined in situ at 723 K for 4 h in flowing air prior to introduction of liquid feed in continuous flow studies.

2.2 Reaction studies

2.2.1 Batch reactions

Reactions were carried out using a 50 mL pressure vessel (Hastelloy C-276, Model 4792,

Parr Instrument, Figure 2-1). After loading the reactant solution, catalyst, and magnetic stirrer bar into the reactor, the vessel was sealed, purged with 34 bar He, and pressurized with He or H2.

Mechanical stirring was maintained using a magnetic stirrer plate (500 rpm).

Figure 2-1 Experimental set-up for batch reaction studies.

2.2.2 Continuous flow reactions

The flow reactor was a 6.4 mm (0.25 inch) outer diameter stainless steel tube with wall thickness of 0.7 mm (0.028 inch). The catalyst bed consisted of catalyst loaded between a quartz wool plug and fused SiO2 granules (-4+16 mesh, Sigma-Aldrich). The reactor was heated with a furnace composed of close-fitting aluminium blocks externally heated by an insulated furnace

(1450 W/ 115 V, Applied Test Systems Series 3210), and temperature was monitored using a K type thermocouple (Omega) placed outside the reactor immediately next to the position of catalyst bed. Temperature was controlled with a PID temperature controller (Love Controls,

Series 16A). Liquid feed solution was introduced into the reactor with co-feeding of gas (H2 or

He) in an up-flow configuration. A HPLC pump (LabAlliance, Series 1) was used to control the liquid feed rate. Gas flow was controlled with a mass-flow controller (Brooks 5850 model), and the system pressure was maintained by passing the effluent gas stream through a back-pressure

33 regulator (GO Regulator, Model BP-60). The effluent liquid was collected in a gas-liquid separator and drained periodically for analysis by HPLC. Carbon balances closed within 5% for all data points for C-O hydrogenolysis reactions.

2.3 Analytical methods

2.3.1 High performance liquid chromatography

Quantitative analyses were performed using a Waters 2695 separations module high performance liquid chromatography instrument equipped with a differential refractometer

(Waters 410) and a photodiode array detector (Waters 996). Depending on the nature of the sample mixture, several different columns were employed. For samples containing organic acids, an ion-exclusion column (Aminex HPX-87H, 300 mm×7.8 mm) with 0.005 M H2SO4 at a flow- rate of 0.6 mL min-1 as the mobile phase was used. For pH neutral mixtures of polyols and alcohols, compounds were separated using an Aminex HPX-87P column (Biorad) at 358 K, with

Milli-Q water at a flow rate of 0.6 mL min-1 as the mobile phase. For studies involving the quantification of triacetic acid lactone and 5,6-dihydro-4-hydroxy-6-methyl-2H-pyran-2-one, the

HPLC column was a reversed-phase Agilent Zorbax SB-C18 (4.6 x 300mm, 5µm) using 5 mM

H2SO4 as the aqueous phase with acetonitrile as the organic modifier. For the fractionation/ isolation of products from reaction product mixtures for NMR analysis, the HPLC column was a reversed-phase Agilent Zorbax SB-C18 (4.6 x 300mm, 5µm) using Milli-Q water as the aqueous phase with methanol as the organic modifier.

2.3.2 Gas chromatography

Gas phase products from batch reactions were collected in a gas bag and analysed using a

Varian GC (Saturn 3) equipped with a FID detector and a GS-Q column (J&W Scientific).

34 Quantitative analyses of liquid phase products were performed by gas chromatography

(Shimadzu GC2010 with an FID and DB-5 column). Identification of products was performed using a gas chromatograph-mass spectrometer (Shimadzu Corp., GCMS-QP2010S) equipped with a SHRXI-5MS capillary column (30 m × 0.25 mm × 0.25 μm), and matched to at least 95% similarity with purchased compounds (Sigma Aldrich), references from the NIST MS library, or the literature.

2.3.3 Nuclear magnetic resonance spectroscopy

Unless otherwise stated, isolated samples were dissolved in CDCl3 (Aldrich, 100%,

≥99.96 atom% D), and NMR spectra were obtained using a Varian Mercury 300 (1H NMR, 300

MHz; 13C NMR, 75 MHz) and referenced to tetramethyl silane (Sigma-Aldrich, 99.9+%, NMR

Grade).

2.4 Catalyst characterization

2.4.1 Temperature-programmed methods

Temperature–programmed experiments were carried out using an apparatus comprising of a tube furnace connected to a variable power-supply and PID temperature controller (Love

Controls) with a K-type thermocouple (Omega). For temperature-programmed reductions, the

3 -1 reducing gas consisted of 5% H2 in He, at a flow-rate of approximately 15 cm (STP) min . Dried, unreduced catalyst samples (0.2 g) were loaded into a 12.6 mm outer diameter, fritted quartz tube, purged with the reducing gas for 30 min, followed by initiation of a temperature ramp at 10 K

-1 min to 973 K. For NH3 temperature-programmed desorption studies, reduced and passivated metal catalyst samples (0.4 g) were loaded into the fritted quartz tube, pretreated in flowing H2

(100 cm3(STP) min-1) at 523 K for 4 h, and purged with He (200 cm3(STP) min-1) at 523 K for

90 min. NH3 adsorption was performed at 298 K using 1% NH3 in He at a flow-rate of 100

3 -1 3 cm (STP) min . After NH3 adsorption, the sample was purged with flowing He (200 cm (STP) min-1) at 298 K for 90 min. The He flow-rate adjusted to approximately 30 cm3(STP) min-1 followed by initiation of a temperature ramp at 10 K min-1 from 298 to 1073 K. Vacuum for the mass spectrometer was provided by a diffusion pump connected in series to a rotary pump. ZSM-

5 (0.1 g) was calcined in flowing air (100 cm3(STP) min-1) at 723 K for 4 h, cooled to 423 K

3 -1 under flowing He (100 cm (STP) min ), followed by adsorption of NH3 at 423 K (1% NH3 in He,

3 -1 100 cm (STP) min ). After NH3 adsorption, the sample was purged with flowing He (200 cm3(STP) min-1) at 423 K for 90 min, and then cooled to 298 K under He flow. The He flow-rate was adjusted to 60 cm3(STP) min-1 followed by initiation of a temperature ramp at 10 K min-1 from 298 to 1073 K. In Chapter 3, the effluent from the reactor tube was introduced into a vacuum chamber (5x10-5 Torr) via a constricted quartz capillary, and the composition of the gas was monitored by a mass spectrometer system (quadruple residual gas analyzer, Stanford

Instruments RGA 200). In Chapter 4, the effluent from the reactor was analyzed using an

OmniStar Gas Analyzer (Pfeiffer Vacuum, Model GSD 320) using the SEM detector.

The NH3 desorption profiles reported are based on the mass 15 (NH) signal which was verified to be directly indicative of NH3 desorption alone and did not experience any interference from the fragmentation of contaminants such as H2O and CO2. The acid site density for ZSM-5

-1 (Engelhard) was determined to be 524 µmolg using NH3 TPD analysis.

36 2.4.2 CO Adsorption

Prior to CO adsorption measurements, RhRe/C catalysts were pretreated in 150 cm3(STP)

-1 min flowing H2 at 393, 523 or 723 K for 4 h. The CO adsorption uptakes at 298 K were measured on a standard gas adsorption apparatus described elsewhere3.

2.4.3 Electron microscopy

Catalysts were dispersed in ethanol and mounted on holey carbon grids for examination in a JEOL 2010F 200 kV transmission electron microscope equipped with Oxford Energy

Dispersive Spectroscopy (EDS) system for elemental analysis. Images were recorded in high angle annular dark field (HAADF) mode with a probe size of 1.0 nm.

2.4.4 In situ and operando X-ray absorption spectroscopy

Rh K-edge (23.220 keV) and Re LIII-edge (10.534 keV) X-ray absorption measurements were conducted on the beamlines of the Materials Research Collaborative Access Team

(MRCAT, 10-ID and BM) at the Advanced Photon Source (APS) at Argonne National

Laboratory. Ionization chambers were optimized for the maximum current with linear response

(ca. 1010 photons detected s-1). A third detector in series simultaneously collected a foil reference spectrum with each measurement for energy calibration. The X-ray beam was 0.5 × 0.5 mm2 at the ID beamline and 1.0 x 1.0 mm2 at the BM beamline, and data was collected in transmission mode.

Catalysts were reduced in a continuous-flow reactor, which consisted of a quartz tube (1 inch OD, 10 inch length) sealed with Kapton windows by two Ultra-Torr fittings. Ball valves were welded to each Ultra-Torr fitting and served as the gas inlet and outlet. An internal K type thermocouple (Omega) was placed against the catalyst sample holder to monitor temperature.

37 Catalyst samples were gently pressed into a cylindrical sample holder consisting of six wells, forming a self-supporting wafer. The catalyst amount used was calculated to give an absorbance

3 (µx) of approximately 1.0. The catalysts were reduced in flowing 3.5% H2 in He (50 cm (STP) min-1) at temperatures indicated in the text , purged with flowing He for 10 min and then cooled to RT. XAS spectra were collected on the reduced samples. Traces of oxygen and moisture in the gases were removed by means of a purifier (Matheson PUR-Gas Triple Purifier Cartridge).

Operando XAS measurements were performed under continuous flow C-O hydrogenolysis reaction conditions, identical to those used in kinetic studies at WHSV = 0.9 h-1.

The reactor was a glassy carbon Sigradur G-Tube (10 mm/4 mm OD/ID, Hochtemperatur-

Werkstoffe GmbH) connected to standard Swagelok fittings with graphite ferrules. Catalyst (0.2 g, 4 wt% RhReOx/C (1:0.5)) was loaded and held in place by quartz wool plugs. A K-type thermocouple was inserted into the top of the bed to monitor the reaction temperature. The reactor was placed in an aluminum heater block equipped with a PID temperature controller

(Omega). A small slit (25.4 mm x 2.0 mm) in the heater block allowed X-rays to pass through the reactor. Liquid feed solution (5 wt% HMTHP in water) was introduced into the reactor with co-feeding of H2 in an up-flow configuration using a syringe pump (Teledyne ISCO). The effluent liquid was collected in a gas-liquid separator (Jurgusen) and drained periodically for analysis by HPLC; conversion and product selectivities were ascertained to be similar to that obtained in the laboratory. System pressure was controlled by a back-pressure regulator (GO regulator). The Debye–Waller factor (Δσ2) at reaction temperature was determined by obtaining the XAS spectra at reaction temperature and room temperature and assuming the same coordination number at both temperatures.

The normalized, energy calibrated, Rh K and Re LIII-edge XANES spectra were obtained by standard methods. The edge energy was determined from the maximum in the first peak of the first derivative of the XANES. Experimental phase shift and backscattering amplitudes were obtained from the Rh and Pt foils for Rh–Rh (12 at 2.68 Å) and Pt–Pt (12 at 2.77 Å), respectively, and NH4ReO4 for Re–O (4 at 1.74 Å). Theoretical phase and amplitude functions for Rh–Re and

2 2 Re–Rh were calculated with FEFF6 using a two atom calculation. The values for So and Δσ were determined by fitting the foils with FEFF. Standard procedures using WINXAS 3.1 software were employed to fit the XAS data. The EXAFS parameters were obtained by a least square fit in R-space of the k2-weighted Fourier transform (FT) data.

2.5 References

1. Serrano-Ruiz, J. C., Luettich, J., Sepúlveda-Escribano, A. & Rodríguez-Reinoso, F. Effect of the support composition on the vapor-phase hydrogenation of crotonaldehyde over Pt/CexZr1-xO2 catalysts. Journal of Catalysis 241, 45-55 (2006). 2. Aramendía, M. A. et al. Synthesis and textural-structural characterization of magnesia, magnesia-titania and magnesia-zirconia catalysts. Colloids and Surfaces A: Physicochemical and Engineering Aspects 234, 17-25 (2004). 3. Spiewak, B. E., Shen, J. & Dumesic, J. A. Microcalorimetric Studies of CO and H2 Adsorption on Nickel Powders Promoted with Potassium and Cesium. The Journal of Physical Chemistry 99, 17640-17644 (1995).

Chapter 3 Selective C-O hydrogenolysis over bimetallic catalysts

3.1 Introduction

A central challenge in biomass conversion lies in developing catalysts for the selective deoxygenation of highly functionalized molecules, such as sugars, polyols, and cyclic ethers, often under conditions where water is used as the solvent1, 2. Catalytic deoxygenation may be accomplished through reactions such as decarboxylation, decarbonylation, dehydration, and C-O hydrogenolysis. Of these reactions, selective C-O hydrogenolysis of polyols and cyclic ethers over heterogeneous catalysts represents an important class of reactions for the production of high value chemicals, such as the conversion of 2-(hydroxymethyl)tetrahydropyran 1 to 1,6- hexanediol 2 (Figure 3-1). In this chapter, the use of bimetallic catalysts, specifically highly reducible metals promoted with oxophilic components are evaluated for their potential to facilitate selective C-O hydrogenolysis of cyclic ethers and polyols in liquid water.

3.1.1 Importance of C-O hydrogenolysis as a deoxygenation strategy

The importance of selective C-O hydrogenolysis has been highlighted by Schlaf2, who demonstrated that regardless of the complexity of specific chemical structures, all chemical transformations for the deoxygenation of biomass-derived molecules may be generalized to the following three basic reactions: (a) dehydration of vicinal diols and hydrogenation of the resulting carbonyl functional group, (b) dehydration of alcohols and hydrogenation of the resulting carbon-carbon double bond, and (c) condensation of diols to oxacycles and hydrogenolysis of the resulting cyclic ether. It is notable that while reactions (a) and (b) are relatively well-understood, reaction (c) remains a non-trivial chemical transformation for which

40 good chemoselectivity and activity is difficult to attain over conventional monometallic catalysts.

More generally, the ability to perform selective C-O hydrogenolysis provides new routes for the production of biorenewable chemicals through the selective ring-opening of heterocyclics and deoxygenation of polyols such as sugars.

+H OH 2 +H2 HO HO +H b a OH a 1,6-hexanediol 1-hexanol 2 O +H2

b hexane 2-hydroxymethyl +H2 +H +H2 tetrahydropyran 2 HO OH OH 2-hexanol 1,2-hexanediol

Figure 3-1 Reaction pathways for the C-O hydrogenolysis of (2-hydroxymethyl)tetrahydropyran to hexanediol, hexanol and hexane.

Recent experimental work suggests that platinum, ruthenium, rhodium, and iridium catalysts promoted with rhenium display not only high activity, but also high selectivity in the hydrogenolysis of tetrahydrofurfuryl alcohol 33, 14, and glycerol 45-8 to their corresponding α,ω- diols. However, despite this growing body of experimental work, a systematic understanding remains elusive of the fundamental reaction mechanisms governing the rate and selectivity of C-

O bond scission reactions over these interesting catalysts. In this chapter, we explore selective C-

O hydrogenolysis of a wide range of biomass-derivable oxygenated feedstocks and identify a highly effective RhRe/C catalyst for these reactions. We then obtain fundamental understanding of the nature of the active site over this bimetallic catalyst using a combination of results from experimental work, preliminary catalyst characterization, and quantum chemical calculations.

41 Finally, the general reactivity rules elucidated are then shown to be translatable to a general class of catalysts consisting of the unique pairing between a highly reducible metal and an oxophilic component.

3.1.2 Current state of the art

Selective hydrogenolysis of cyclic ethers

The most relevant literature to this work concerns the selective hydrogenolysis of 3 to form 1,5-pentanediol 11. Compound 3 is a structural analogue of 1, and can be obtained from biomass through well-established processes such as the acid-catalyzed thermolysis of pentosans to furfural9, and subsequent hydrogenation of furfural to 3. Compound 3 ring-opens more easily than 1 mainly due to relatively higher strain in its five-membered ring. Notably, there has been enduring interest in obtaining 11 from 3. One of the earliest reports by Conner et al. involve the use of a copper chromite catalyst (prepared by coprecipitation) for the liquid phase

10 hydrogenation of 3 (175 °C, 100-150 atm H2) . Although the combined yield of diols (11 and

1,2-pentanediol 13) was 70%, selectivity to 11 was only 38%. To circumvent these losses associated with non-selective ring-opening of 3, Schniepp et al. examined a process in which 3 is

11 catalytically converted (Al2O3, 375 °C) to dihydropyran . Dihydropyran is then hydrolyzed with acid (0.02 M HCl) to form δ-hydroxyvaleraldehyde, which is hydrogenated over CuCr2O4 to 11; the final yield of 11 from 3 was 70%. It is therefore evident that obtaining the α,ω-diol from 3, or any of its structural analogues, is by no means a trivial undertaking using conventional catalysts.

Another example underscoring this point is work by Tike et al. who demonstrated that during the hydrogenation of furfural to 3 over 5 wt% Ru/TiO2, the only by-product formed by initial hydrogenolysis of the heterocyclic is 1312.

More recently, Koso et al. reported the use of ReOx-modified Rh/SiO2 for the chemoselective hydrogenolysis of 3 to 113 where a maximum selectivity of 94% to 11 (57% conversion) was achieved. Based on characterization studies using transmission electron microscopy (TEM) and X-ray diffraction spectroscopy (XRD), it was proposed that due to particle size similarities between Rh/SiO2 and RhReOx/SiO2, Rh metal particles were covered by

ReOx clusters. Extended X-ray absorption fine structure spectroscopy (EXAFS) analysis suggested that Re was not fully reduced to its metallic state, and that RhRe coordination was present. Other studies for the selective hydrogenolysis of 3 to 11 suggested that RhMo was also an effective catalyst, although lower specific activities were observed compared to Re13. 90% selectivity to 11 at 94% conversion of 3 was reported over a 4 wt% RhMoOx/SiO2 catalyst, which was recycled five times with no significant loss in activity and selectivity to 11.

Selective hydrogenolysis of glycerol

As mentioned in Chapter 1 (Section 1.4.2), the hydrogenolysis of 4 to propanediols over heterogeneous catalysts has been extensively studied in the literature. The preferred metals for the hydrogenolysis of 4 to propanediols are Ru, Pt, Rh, Ni14 and Cu15-17; selective C-O scission without C-C cleavage remains a challenge over these metals. Factors such as the nature of the support, nominal loading of metal, and catalyst synthesis method have been shown to affect selectivity to 1,3-propanediol 17 significantly18. Although Ru-based catalysts appear to be effective in the hydrogenolysis of 4 to propanediols, Ru also facilitates a significant extent of C-

C scission resulting in the production of considerable amounts of alkanes and ethylene glycol.

Recent reports using a variety of bimetallic catalysts and oxide promoters to increase C-O hydrogenolysis activity have been promising. More importantly, the literature is consistent in

43 showing that 1,2-propanediol is produced in favor of 17 over most catalysts, and that selectivities to 1,2-propanediol are typically around 4-5 times higher than that to 17.

Several reaction mechanisms for the hydrogenolysis of 4 have been proposed, most notably, that involving acetol as an intermediate. Acetol was successfully isolated by Dasari et al.15 and shown to be produced in the presence of an acid catalyst (e.g. Amberlyst 1519, 20). This mechanism with acetol as the intermediate was further validated by Dasari et al. who studied the hydrogenolysis over Cu-Cr catalysts under relatively mild reaction conditions (200 °C, 200 psi

15 H2) . It was found that the yield of propanediols increased with decreasing water content, but a minimum of 10-20 wt% solvent (water or methanol) is needed to prevent excessive polymerization and product degradation. Another reaction mechanism under high pH reaction conditions has been recently proposed by Maris et al., who examined the effect of base addition

(NaOH, CaO) on catalyst activity and selectivity21. The proposed mechanism is a modified mechanism by Montassier et al.22, and involves initial dehydrogenation of 4 to glyceraldehydes as the first step in the hydrogenolysis of 4. In the presence of water or under basic conditions, glyceraldehyde is in equilibrium with its enolic tautomers which undergoes hydrogenation to finally form saturated glycols, methane and lactic acid. Using Ru/C and Pt/C catalysts, it was found that although Pt/C is less active, Pt/C catalyzed propylene glycol formation with higher selectivity then Ru/C. Also, base addition led to a larger increase in activity of Pt/C than for

Ru/C. It was therefore suggested that the scission of C–C bonds leading to the formation of ethylene glycol occurs through a base-catalyzed reaction in the presence of Pt, while this reaction occurs through a metal-catalyzed route for Ru/C.

Several approaches have been made to increase catalyst activity and selectivity to propanediols, such as modification with sulfur, use of bimetallic catalysts23, and addition of solid

44 acid catalysts to facilitate initial dehydration and improve propanediol yields19, 20. Initial attempts by Montassier et al. to modify Ru/C using sulfur led to decreased activity of both hydrogenation and hydrogenolysis reactions, and only low yields of propanediols were achieved22. However, it was found that the selectivity to 17 increased with increasing sulfur loading. Similarly, Lahr et al., reported a decrease in apparent activation energy of hydrogenolysis of 4 with increasing values of the ratio of S:Ru, as well as a linear relationship between selectivity to propanediols and the ratio of S:Ru24. The former was attributed to site-blocking by sulfur and the latter to hindered diffusion of absorbed hydrogen by sulfur and long-range electronic effects.

Of greater relevance to this study is the addition of an oxophilic promoter (Re, W, Mo) to a highly reducible metal as catalysts for the hydrogenolysis of 4 to propanediols. Several

25, 26 investigators have reported that Re had a significant effect on the activity of Rh/SiO2 and

6, 7 Ru/SiO2 catalysts. In all studies, catalyst characterization using X-ray photoelectron spectroscopy (XPS) and X-ray absorption spectroscopy (XAS) suggested that that Re was present as an oxide; Shimi et al. estimated from the Re LIII edge XANES that Re in a RhRe/SiO2 catalyst had an average oxidation state of +2 to +2.526. Ma et al. showed that although the activity of RuRe/SiO2 catalysts were substantially higher than monometallic Ru/SiO2, the rate of product (i.e., propanediols) degradation was lower over these Re-promoted catalysts6.

Additonally, the addition of Re2(CO)10 to monometallic Ru catalysts was observed to significantly increase catalytic activity for the hydrogenolysis of 46, 7. Shimao et al., reported that while the reactivity order of various polyols on monometallic Rh/SiO2 was 4 ≈ 17 > 1,2-

25 propanediol , the order was modified to 4 > 13 > 17 over Re-promoted Rh/SiO2 catalysts, cumulating in augmented final yields of 17. More recently, Nakagawa et al reported the selective hydrogenolysis of 4 to 17 over a Re-promoted Ir catalyst supported on SiO2, showing that 38%

45 yield of 17 was attained (81% conversion of 4)27. These results collectively show that Re- promoted catalysts are highly selective in cleaving specific C-O bonds on a molecule.

Hydrogenolysis of Tetrahydrofuran and Methyltetrahydrofuran

The fundamental factors affecting the ring-opening of oxacycloalkanes can be rationalized from studies using structural analogues such as tetrahydrofuran 8 and 2- methyltetrahydrofuran 7. The role of the support in C-O bond scission was demonstrated by

Kreuzer et al. who used supported Pt in the hydrogenolysis of 8 (423 -629 K, 900-250 mbar H2, batch reactor)28. It was found that catalytic ring-opening activity to butanol and subsequent selectivity to butane was of the following order: Pt black << Pt/SiO2 < Pt/Al2O3 < Pt/TiO2. The high selectivity to butane over Pt/TiO2 was attributed to the superior ability of adlineation sites at the Pt/TiO2 interface to activate C-O bonds. Also, it was observed that CO preferentially inhibits these adlineation active sites. Gennnari et al. specifically chose 7 as a probe molecule to demonstrate adlineation effects on supported Pt29; ring-opening of the heterocyclic was purported to be a good reaction for investigating structure sensitivity as the O atom is more susceptible to adsorption by acidic sites. Infrared measurements showed that 7 adsorbs on the catalyst as a slightly dehydrogenated reactive intermediate (loss of 1-2 hydrogen atoms from the ring), with the methyl group unaltered. It was postulated that this is evidence of a diadsorbed species attached to the surface via the O atom in the ring and a neighboring C atom, which subsequently leads to C-O scission. Using a flow reactor and similar Pt catalysts (Pt black and

Pt/SiO2), Gennari et al. obtained product distributions indicating that the selectivity of C-O

29 cleavage was governed by adlineation effects (425-530 K, 10 and 43 mbar H2) . Specifically, Pt black displayed considerably lower selectivity to scission of the more-substituted C-O bond

(~7%), than Pt/SiO2 (~30%). This suggested that C-O scission between the secondary C and O

46 atom occurs at the metal-support phase boundary, while Pt sites are responsible for that between

the primary C and O atom.

3.2 Catalyst development

3.2.1 Highly reducible metals

Initial work in this study focused on screening several highly reducible metal-metal oxide

pairings that gave the most active catalysts for the ring-opening of 1. In exploratory experiments,

Rh was chosen as the highly reducible metal and paired with several oxophilic metals (Fe, W,

Ni, Mo, and Re) to obtain promoted catalysts. It was found that only Mo and Re conferred

appreciable ring-opening activity under the reaction conditions employed (i.e., 393 K, 500 psi

H2), and thus further work focused on the use of these metal oxides as promoters. The superior

promotive effect of Mo and Re over other oxophilic metals agrees well with the literature, where

it has been noted that Fe30, 31 and W13 are less effective promoters for C-O scission reactions on

supported Rh. Also, the marked promotive effect of Mo for 1 ring-opening here parallels that in

published literature13, 32-35 wherein the addition of Mo was found to significantly enhance C-O

dissociation.

Ru, Pd, and Pt were used as alternative highly reducible metals to Rh. It was observed

that only Ru-, Rh- and Pt-containing bimetallic catalysts (Table 3-1) were able to effectively

ring-open 1; while Ru displayed activity towards ring-opening, selectivities to 2 were low

compared to Rh and Pt catalyst, with significant amounts of 2-methyltetrahydropyran 5 being

formed. Pd catalysts were found to be inactive for C-O hydrogenolysis under the reaction

conditions employed here. Generally, Rh-based catalysts displayed the highest specific activity,

particularly for 4 wt% RhRe/C (1:0.5) where a specific activity of 81 μmolg-1min-1 was

47 observed. Interestingly, the only by-products detected over Rh catalysts were degradation

products of 2 (1-hexanol, 1-pentanol), i.e., no 1,2-hexanediol was detected, indicating that ring-

opening was highly selective at the sterically-hindered C-O bond in the pyran ring.

Table 3-1 Effect of different catalyst compositions on catalytic activity and selectivity of 2- (hydroxymethyl)tetrahydropyran 1 ring-opening to 1,6-hexanediol 2. The loadings of Rh, Ru, Pd and Pt were 4 wt%; all catalysts were supported on Vulcan XC-72.a

Total Catalyst M1:M2 M2 metal Time Catalyst:1 Conv Sel to 2 Rateb atomic Precursor loading (h) mass ratio (%) (%) (μmolg-1min-1) M1 M2 ratio (wt%) Rh Re NH4ReO4 1:0.5 7.4 4 1:9 24.7 91.0 81 (NH ) Mo Rh Mo 4 6 7 1:0.5 5.8 12 2:7 4.2 93.6 2 O24∙4H2O

0.7 Ru Re NH4ReO4 1:0.5 7.4 12 2:7 87.9 (5 – 37 23.9) 6.3 4 2:7 62.3 (5 – 77 26.6)

c Pd Re NH4ReO4 1:0.5 7.2 12 2:7 NR -- --

(NH ) Mo Pt Mo 4 6 7 1:0.5 4.9 12 2:7 9.0 85.4 4 O24∙4H2O

Pt Re NH4ReO4 1:0.5 5.8 12 2:7 1.3 68.4 1 a Reaction conditions: 393 K, 500 psi H2, feed mixtures were 5 wt% 1 in water as feed, catalysts not pretreated prior to introduction of liquid feed. bSpecific rate was calculated as the number of moles of 1 reacted per gram of catalyst per minute. cNR: no reaction.

3.2.2 Effect of varying catalyst composition

A carbon-supported Rh catalyst was modified with two different oxophilic promoters,

namely Re and Mo, and initial reaction kinetics studies were carried out using 1 as the reactant

(Table 3-2) to identify the amount of promoter required to maximize catalytic activity. For both

promoters used, the specific activity of the promoted Rh catalysts passes through a maximum as

the amount of promoter is increased. Catalysts consisting of 4 wt% RhRe/C (1:0.5 atomic ratio)

48 and 4 wt% RhMo/C (1:0.1 atomic ratio) were found to be the most active materials, and these

catalysts were selected for use in subsequent reaction kinetics studies. In contrast, it was found

that for Pt catalysts, an increase in Pt:Re atomic ratio led to a steady increase in catalytic activity

for C-O hydrogenolysis. However, all PtRe catalysts were consistently less active than RhRe

catalysts.

Table 3-2 Effect of metal-loading ratio on catalytic activity and hydrogenolysis selectivity of 2- (hydroxymethyl)tetrahydropyran 1 to 1,6-hexanediol 2 over promoted Rh and Pt catalysts. The loadings of Rh and Pt were 4 wt%; all catalysts were supported on Vulcan XC-72a.

b Catalyst M1:M2 Time Catalyst: 1 PH2 Conv Sel to 2 Rate M1 M2 (mol:mol) (h) mass ratio (psi) (%) (%) (μmolg-1min-1) Rh Re 1:0.25 5 1:9 900 48 82 132 1:0.5 5 1:9 900 55 86 153 1:1 5 1:9 900 38 92 109 6 1:9 900 46 89 99 Rh Mo 1:0.05 12 2:7 900 40 81 16 1:0.1 12 2:7 900 55 84 22 1:0.25 12 2:7 900 48 85 19 1:0.5 12 2:7 900 48 77 20 Pt Re 1:0.5 12 2:7 500 1.3 68.4 1 1:1 12 2:7 500 17.0 92.4 7 1:2 12 2:7 500 29.4 83.3 12 1:2.5 12 2:7 500 36.2 87.4 15 aReaction conditions were 393 K, catalysts not pretreated prior to use, 5 wt% 1 in water as feed. bSpecific rate defined as the moles of 1 reacted per gram of catalyst per minute.

3.2.3 Precursor effects for Mo-promoted catalysts

To improve the activity of Mo-promoted catalysts, two different oxophilic metal

precursors were used. As shown in Table 3-3, there is a significant disparity in catalytic

activities depending on the precursor employed. Specifically, a five-fold rise in specific activity

-1 -1 was observed when (NH4)6Mo7O24∙4H2O is used (20 μmolg min ), compared to that for

-1 -1 Mo(CO)6 (4 μmolg min ). A slight increase in selectivity was also noted when

(NH4)6Mo7O24∙4H2O is the precursor. Such a marked effect of the precursor on catalyst

49 performance could be due to the differences in precursor dispersion on the support and/ or nature

of alloying with the highly reducible metal, i.e., Rh.

Table 3-3 Effect of different metal oxide precursors on catalyst activity and the selectivity of 2- (hydroxymethyl)tetrahydropyran 1 ring-opening to 1,6-hexanediol 2. The loading of Rh was 4 wt%; atomic ratio of Rh:Mo = 1:0.5. all catalysts were supported on Vulcan XC-72.a

Rateb Catalyst Mo precursor Time (h) Conv (%) Sel to 2 (%) (μmol∙ g-1 ∙min-1)

RhMo/C Mo(CO)6 12 9 68 4 (NH4)6Mo7O24∙4H2O 12 48 77 20 a Reaction conditions were 393 K, 900 psi H2, 5 wt% 1 in water as feed , mass ratio of catalyst:1 = 2:7. bSpecific rate was calculated as the number of moles of 1 reacted per gram of catalyst per minute.

3.2.4 Support effects

Carbon and silica supported Rh-based catalysts were found to differ in activity and

selectivity in the hydrogenolysis of 1 to 2 (Table 3-4). For Re-promoted catalysts, selectivities to

2 were higher on Vulcan carbon than silica, while selectivities to 2 over Mo- promoted catalysts

prepared using Mo(CO)6 as the precursor were higher on silica. Several different carbon supports

were used to prepare RhRe/C catalysts, namely, Vulcan XC-72, Norit SX-1G, and CMK-3. Of

these, Vulcan XC-72 and CMK-3 supported catalysts displayed similar C-O hydrogenolysis

activities, while Norit SX-1G supported catalysts were relatively inactive (i.e., 4-5 fold decrease

in C-O hydrogenolysis rate compared to Vulcan XC-72 and CMK-3). The reasons underlying the

lower activity of RhRe catalysts supported on Norit SX-1G is unclear, and could be due to the

higher ash content of activated carbon (i.e., Norit SX-1G) compared to carbon black (i.e., Vulcan

XC-72) and CMK-3 (ash-free). C-O hydrogenolysis activity for a RhRe catalyst prepared on a

carbide support, i.e., β-SiC, was found half that with Vulcan XC-72 as the support; the lower

hydrogenolysis activity observed over carbide supported RhRe catalyst is likely due to the low

50 surface area of the support resulting in poor metal dispersion (observed using STEM, results not

shown). As Vulcan XC-72 was observed to yield the catalyst with the highest activity and is

commercially available, it was therefore used for all subsequent studies.

Table 3-4 Support effects on the selectivity of 2-(hydroxymethyl)tetrahydropyran 1 ring-opening to 1,6-hexanediol 2. The loadings of Rh and Pt were both 4 wt%; M1:M2 = 1:0.5 (mol:mol, M1 = Rh, Pt, M2 = Re, Mo).a

Promoter Time P Sel to Rateb Catalyst H2 Conv (%) precursor (h) (psi) 2 (%) (μmolg-1min-1)

RhMo/Vulcan C Mo(CO)6 4.5 900 3 47 8 RhRe/Vulcan C NH4ReO4 4.5 900 55 86 153 RhMo/SiO2 Mo(CO)6 4.5 900 17 78 48 RhRe/SiO2 NH4ReO4 4.5 900 36 66 111 c RhRe/β-SiC NH4ReO4 4 500 12 87 40 RhRe/Vulcan C NH4ReO4 4 500 27 97 90 RhRe/Norit C NH4ReO4 4 500 6 94 20 RhRe/CMK-3 NH4ReO4 4 500 29 94 96 a Reaction conditions were 393 K, mass ratio of catalyst:1 = 1:9, 5 wt% 1 in water as feed. bSpecific rate was calculated as the number of moles of 1 reacted per gram of catalyst per minute. cD10273 β-SiC UHP3 (SICAT R&D, Germany), crushed to a fine powder prior to impregnation of metals.

3.2.5 Initial reactivity studies

The effect of oxophilic promoters, such as Re or Mo on the C-O hydrogenolysis activities

of Rh catalysts is evident in experiments with reactants 1 and 3 (Table 3-5). Specifically, the

hydrogenolysis rate of 3 was increased fifteen-fold over 4 wt% RhRe/C (1:0.5) compared to the

monometallic 4 wt% Rh/C catalyst. The promoting effect of Mo is not as pronounced as for Re:

only a six-fold increase in the hydrogenolysis rate of 3 was observed for 4 wt% RhMo/C (1:0.1).

Importantly, the Re or Mo promoters lead to a remarkable enhancement of selectivities to the

α,ω-diols for both cyclic ethers: scission of the C-O was observed to occur primarily at the more

sterically hindered secondary carbon-oxygen bond for the promoted Rh catalysts, resulting in

high selectivities to the respective α,ω-diols. This behavior is in contrast to non-selective C-O

51 hydrogenolysis observed over monometallic 4 wt% Rh/C, as well as metal-catalyzed ring opening of substituted cyclopentane and cyclohexane derivatives over Pt and other transition metals which preferentially occur at the least-substituted carbon-carbon bonds36. The increases in hydrogenolysis activity and selectivity for Re-promoted Rh/C are in agreement with results reported by Chen, et al..4 Experiments using monometallic 3.6 wt% Re/C, and 1.8 wt% Mo/C catalysts indicate that the Re and Mo promoters in the absence of a highly reducible metal, such as Rh, do not possess hydrogenolysis activity, and this result is consistent with the literature3, 4,

13.

Table 3-5 Hydrogenolysis of tetrahydrofurfuryl alcohol 3 and 2- (hydroxymethyl)tetrahydropyran 1 over Re- and Mo-promoted Rh catalysts and monometallic catalysts.a

Catalyst: Time Product selectivity Rateb Catalyst reactant Reactant Conv (%) (h) (μmolg-1min-1) (g:g) Compound (%) 4 wt% RhRe/C 1:9 4 3 47.2 1,5-pentanediol 97.2 180 (1:0.5) 1-pentanol 2.8 1 27.3 1,6-hexanediol 97.0 90 1-hexanol 3.0 4 wt% 2:7 4 3 51.6 1,5-pentanediol 91.3 72 RhMo/C 1-pentanol 3.2 (1:0.1) 1-butanol 0.1 Othersc 5.4 1 25.8 1,6-hexanediol 88.6 32 1-hexanol 3.5 1-pentanol 7.2 Othersc 0.7 4 wt% Rh/C 2:7 4 3 8.5 1,5-pentanediol 59.1 12 1,2-pentanediol 20.7 1-pentanol 7.1 2-pentanol 10.0 Othersc 3.1 1 3.3 1,6-hexanediol 43.5 4 1,2-hexanediol 11.1

1-hexanol 2.8 Othersc 42.6 3.6 wt% Re/C 2:7 12 3 NRd - - - 1 NRd - - -

52 1.8 wt% Mo/C 2:7 12 3 NRd - - - 1 NRd - - - a Reaction conditions were 393 K, 500 psi H2, 5 wt% reactant in water as feed. bSpecific rate defined as the moles of 1 or 3 reacted per gram of catalyst per minute. cAlkanes in gas phase and monoalcohols at trace levels. dNR: no reaction.

3.3 Characterization studies of RhRe/C

3.3.1 Temperature-programmed reduction

Temperature-programmed reduction (TPR) profiles of dried, unreduced catalysts suggest that contact between Rh and the Re and Mo promoters is established during catalyst preparation

(Figure 3-2), in view of the observation that the H2 consumption peaks for 4 wt% RhRe/C

(1:0.5) and 4 wt% RhMo/C (1:0.5) coincide with the H2 consumption peak of monometallic 4 wt% Rh/C, in agreement with results from the literature32, 33, 37. These TPR profiles show that the presence of Rh lowers the reduction temperature of the Re and Mo precursors, suggesting the

38, 39 formation of metal alloys . Quantitative analysis of H2 consumption by 4 wt% RhRe/C

(1:0.5) reveals that Re is present in a highly reduced state after reduction at temperatures above

600 K: approximately 0.24 mmol H2 was consumed compared to 0.25 mmol H2 required to completely reduce the Rh and Re to the metallic state. The low valence of Re in 4 wt% RhRe/C

4, 40 (1:0.5) after reduction with H2 is consistent with the literature .

(e)

(d) (c)

(b)

consumption (a.u.)

2 H

(a)

300 400 500 600 700 800 900 1000 T (K)

Figure 3-2Temperature-programmed reduction profiles for (a) 4 wt% Rh/C, (b) 3.6 wt% Re/C, (c) 1.8 wt% Mo/C, (d) 4 wt% RhRe/C (1:0.5), (e) 4 wt% RhMo/C (1:0.5).

3.3.2 HAADF-STEM and EDS

High angle annular dark field scanning transmission electron microscopy (HAADF-

STEM) studies of the 4 wt% RhRe/C (1:0.5) catalyst before and after reaction revealed that the metallic nano-particles typically contained both Rh and Re, as seen in electron dispersive X-ray spectroscopy (EDS) microanalyses of Figure 3-3. The average metallic particle size of the 4 wt% RhRe/C (1:0.5) catalyst before and after reaction was estimated to be 2.1 nm from HAADF-

STEM analysis of over 700-900 randomly selected particles, as displayed in Figure 3-4.

54 a b

Figure 2a (as-prepared) Figure 2b (spent) Particle A B C D E F A B C D E F Rh 0.76 0.73 0.02 5.45 0.27 0.03 1.14 0.06 6.39 0.14 5.32 0.77 (at%) Re 0.13 0.2 0.02 1.33 0.31 0.01 0.23 0.10 4.58 0.00 0.77 0.14 (at%) Re:Rh 0.2 0.3 1.0 0.2 1.2 3.0 0.2 1.7 0.7 - 0.1 0.2 Figure 3-3 Representative HAADF STEM images and EDS spot-beam analysis results of (a) as- prepared and (b) spent 4 wt% RhRe/C (1:0.5) catalysts.

a b

350 (d) (c) 300 200

250 Average particle size: 150 Average particle size: 200 2.13 ± 0.65 2.12 ± 0.59 150 100

100 Particle counts Particle counts 50 50 0 0 0 1 2 3 4 5 6 0 1 2 3 4 5 6 Particle size (nm) Particle size (nm)

Figure 3-4 Representative HAADF-STEM images and particle size distribution histograms for (a, c) as-prepared and (b, d) spent 4 wt% RhRe/C (1:0.5) catalysts. Number-averaged particle sizes are presented in the figures.

3.3.3 CO adsorption

The extent of irreversible CO adsorption at 298 K on the various supported carbon

catalysts was used to estimate the dispersions of the metallic particles (where dispersion is

defined as the fraction of the metallic atoms that are on the surface), and these results are shown

in Table 3-6. The metal dispersion was found to be approximately 50% for both monometallic 4

wt% Rh/C and 4 wt% RhRe/C (1:0.5). The particle size of the as-prepared 4 wt% RhRe/C

(1:0.5) catalyst determined from metal dispersion values was found to be in the range of 2.2-

2.3 nm (Table 3-6).

56 Table 3-6 Quantification of CO adsorption on monometallic and promoted Rh catalysts at 298 K.

Part- Rh Re Molar Irreversible CO:Rh Pretreat- Dispersion icle Catalyst loading loading ratio CO uptake (mol:mo ment T(K) (%)a size (wt%) (wt%) Rh:Re (µmolg-1) l) (nm)b Rh/C 4 - - 393 151 0.39 52 2.1 Re/C - 3.6 - 723 29 0.15c - - RhRe/C: as- 4 3.6 1:0.5 393 143 0.38 51 2.2 prepared 523 145 0.39 51 2.2 723 135 0.36 48 2.3 RhRe/C: 4 3.6 1:0.5 393 7 0.02 - - spent 723 74 0.20 - - aDispersion calculated with respect to total Rh loading and assuming 0.75 ML coverage of CO at full saturation. bParticle size = 1.1/Dispersion; factor of 1.1 calculated from the literature41-43. cCO:Re.

In general, estimates of the metal particle sizes obtained from HAADF-STEM and metal dispersion values compare well with one another, and indicate that no significant changes in particle size take place upon exposure to reaction conditions.

3.3.4 NH3 temperature-programmed desorption

Temperature-programmed desorption (TPD) profiles for NH3 for 4 wt% RhRe/C (1:0.5) and the monometallic catalysts 4 wt% Rh/C and 3.6 wt% Re/C were obtained. All catalyst samples were pretreated in flowing H2 at 523 K for 4 h prior to the adsorption of NH3. As shown in Figure 3-5, the NH3 desorption profile for 4 wt% RhRe/C (1:0.5) displays a peak with

-1 maximum at 498 K. The number of NH3-titrated sites on 4 wt% RhRe/C (1:0.5) was 40 μmol g , and molar ratio of NH3-titrated sites to metal sites was equal to 0.28. Assuming that the standard

-1 -1 44, 45 entropy change of adsorption (ΔS°ads) is approximately -150 J K mol , which corresponds to loss of the gas phase translational entropy, and that first-order desorption and rapid

57 readsorption occurs, the heat of adsorption of NH3 (ΔH°ads) on 4 wt% RhRe/C (1:0.5) can be

46 -1 estimated to be -100 kJ mol . In contrast, no discernible NH3 desorption peaks were observed for monometallic 4 wt% Rh/C and 3.6 wt% Re/C under the conditions employed. The lack of acidity of Re/C here could be due to a combination of low Re loading and poor metal dispersion.

(c)

(b)

desorption (a.u.)

3 NH (a)

400 600 800 1000 T (K)

Figure 3-5 NH3 temperature-programmed desorption profiles for (a) 3.6 wt% Re/C, (b) 4 wt% 3 Rh/C, and (c) 4 wt% Rh-Re/C (1:0.5). Catalyst samples were pretreated in flowing H2 (100 cm -1 (STP) min ) at 523 K for 4 h prior to dosing of NH3.

3.4 Catalyst pretreatment and stability

Rhenium is oxophilic and some of its oxides (e.g., Re2O7 and ReO3) are soluble in water, leading potentially to the leaching of rhenium from the catalyst under aqueous reaction environments and to possible homogeneous catalysis. To optimize catalyst pretreatment conditions for maximum activity and stability of the 4 wt% RhRe/C (1:0.5) catalyst, the effect of catalyst reduction temperature on the rate of hydrogenolysis for reactant 1 was first determined,

58 and the extent of rhenium leaching monitored (Table 3-7). These experiments were conducted in a batch reactor, and pretreatment of the dry catalyst was performed under H2 atmosphere at temperatures of 393 K, 523 K, and 723 K prior to addition of the aqueous feed. After reaction for 4 h at 393 K, the reactor was cooled to 298 K, the vessel was depressurized, and the reaction solution was filtered to remove the catalyst. Elemental analysis of the filtrates was performed by inductively-coupled plasma atomic emission spectroscopy (ICP-AES). The results in Table 3-7 show that the catalytic activity and the extent of rhenium leached into the reaction solution were highest without pretreatment of the catalyst (2%) or with a low temperature pretreatment (393 K,

1%). The extent of Re leaching was substantially lower (< 0.5%, which is at the detection limit of our system) after pretreatment at 523 K, and therefore, this condition was chosen for use in continuous flow experiments to further study catalyst stability versus time-on-stream.

Table 3-7 Effect of catalyst pretreatment temperature on hydrogenolysis activity of 2- (hydroxymethyl)tetrahydropyran 1 over 4 wt% RhRe/C (1:0.5) and extent of rhenium leaching.a

H pretreatment 2 Conv Sel to 2 Rateb Re leached temperature (%) (%) (μmolg-1min-1) (%) (K) No pretreatment 27 97 90 2.0 393 25 93 86 1.2 523 16 99 51 < 0.5 723 5 99 15 < 0.5 a Reaction conditions: 393 K, 34 bar H2, 4 h, mass ratio of catalyst:1 = 1:9. Reactant mixtures were 5 wt% 1 in water. bSpecific rate defined as the moles of 1 reacted per gram of catalyst per minute.

Figure 3-6 shows results for the hydrogenolysis of 1 over 4 wt% RhRe/C (1:0.5) in a continuous packed-bed reactor system. The catalyst was pretreated under flowing H2 for 4 h at

523 K prior to initiating the flow of liquid feed. Under these continuous flow conditions, the performance of the catalyst stabilized at conversion level of 32% and a selectivity to 2 of 90%.

After 120 h time-on-stream, the activity of the catalyst declined by only 16%. The stability of the

59 catalyst under continuous liquid flow conditions provides evidence that we have established

catalyst pretreatment and reaction conditions under which 4 wt% RhRe/C (1:0.5) functions fully

as a heterogeneous catalyst.

100 40

80 Rate 30

( 60 

molg 20

-1

40 min

-1 10 )

20 Conversion or selectivity (%) 0 0 0 20 40 60 80 100 120 Time on stream (h)

Figure 3-6 Results for the hydrogenolysis of 2-(hydroxymethyl)tetrahydropyran 1 over 4 wt% RhRe/C (1:0.5) in a continuous flow system. Conversion of 1 (■), selectivities to 1,6-hexanediol -1 2 (○) and 1-hexanol (▲), and reaction rates (×) at 393 K, 34 bar H2, WHSV = 0.52 h , feed: 5 3 -1 wt% 1 in water. The catalyst was pretreated in flowing H2 (60 cm (STP) min ) at 523 K for 4 h and cooled to the reaction temperature prior to initiation of liquid feed flow.

3.5 Reactivity trends and Density Functional Theory calculations

3.5.1 Cyclic ethers

To better understand the high hydrogenolysis selectivities of 1 to 2 over 4 wt% RhRe/C

(1:0.5), the hydrogenolysis rates of cyclic ethers with different functional groups were studied in

a batch reactor, and the results are shown in Table 3-8. The reactivity trend for the six-

membered cyclic ethers was observed to be 1 >> 5 > 6. The reactivity trend of the five-

60 membered ring cyclic ethers is analogous to that of the six-membered ring molecules, where 3

>> 7 > 8. For both classes of cyclic ethers, the absence of the hydroxyl group leads to a significant decrease in C-O hydrogenolysis activity. These results suggest that the hydroxyl groups in 1 and 3 are responsible for the high C-O hydrogenolysis rates and selectivities to their respective α,ω-diols over 4 wt% RhRe/C (1:0.5). This selectivity towards ring-opening at the more substituted carbon center is not consistent with the preferential ring-opening of unsubstituted sites of substituted cyclohexane and cyclopentane intermediates over metal surfaces.36 As is described below with regard to the ring-opening of compounds 5 and 7, metal- catalyzed ring opening first occurs through dehydrogenation, causing the unsubstituted sites to be preferentially activated.

Table 3-8 Hydrogenolysis of cyclic ethers and polyols over 4 wt% RhRe/C (1:0.5).a

React- Rate/ Reactant Ti Product selectivity Rateb ant Conv initial reactant me (μmolg- Struc- conc (%) conc Name (h) Compound (%) 1min-1) ture (wt%) (mL g-1min-1) Cyclic ethers 2-(hydroxyl 1,6-hexanediol 97.0 0.21 methyl) 5 4 27.3 90 tetrahydro- 1-hexanol 3.0 pyran 1 c 2-methyl 1 4 NR - - - tetrahydropyran 2-hexanol 9 24.3 0.08 1d 20 33.6 7 5 Otherse 75.7 tetrahydropyran 1 4 NRc - - - - 6 Tetrahydro- 1,5- 97.2 0.36 furfuryl 5 4 47.2 pentanediol 180 alcohol 3 1-pentanol 2.8 5 4 1.4 2-pentanol 10 >99.9 7 0.01 2-pentanol 10 58.4 2-methyltetra- 1-pentanol 1.6 5 d 20 20.5 7 hydrofuran 7 2-butanol 2.9 Otherse 37.1 tetrahydrofuran 5 4 NRc - - - - 8

61 Triols 5 4 8.1 1,6-hexanediol >99.9 23 0.06 1,6-hexanediol 61.9 1,5-hexanediol 8.7 1-hexanol 3.6 1,2,6- 2-hexanol 9 0.2 hexanetriol 5 d 14 59.3 1-pentanol 5.0 17 (1)f 2-pentanol 10 4.4 2-methyltetra- 3.5 hydropyran Otherse 12.7 1,2-butanediol 8.9 47 0.10 1,3-butanediol 14.2 1,4-butanediol 62.8 1,2,4-butanetriol 5 4 13.1 1-butanol 5.6 2-butanol 3.4 Otherse 5.1 1,3- 24.0 0.08 propanediol

OH 1,2- HO OH 37.2 glycerol 4 5 4 9.8 propanediol 45 1-propanol 27.3 2-propanol 11.4 1,3- 5.3 propanediol 1,2- d ,g 17.7 5 12 10.5 propanediol 6 1-propanol 8.2 2-propanol 68.8

Diols c 1,6-hexanediol 2 3 4 NR - - - - 1-hexanol 79.5 0.15 1,2-hexanediol 3 4 12.4 2-hexanol 9 19.0 40 e Others 1.5 1,5-pentanediol 1-pentanol 97.8 0.01 3 4 1.0 2 11 Otherse 2.2 1,2-pentanediol 1-pentanol 87.8 0.11 3 4 8.8 32 13 2-pentanol 10 12.2 2,4-pentanediol 2-pentanol 10 88.9 0.21 3 4 30.9 e 58 15 Others 11.1 1-butanol 67.6 0.01 1,4-butanediol 5 4 2.1 tetrahydro- 4 (2)f 12 32.4 furan 1,2-butanediol 1-butanol 78.4 0.11 5 4 14.9 62 14 2-butanol 21.6 2,3-butanediol 2-butanol 80.1 0.15 5 4 38.3 78 16 Otherse 19.9

HO ethanol 95.0 0.15 OH ethylene glycol 5 4 40.0 116 Otherse 5.0 a Reaction conditions were 393 K, 500 psi H2, mass ratio of catalyst:reactant =1:9. bSpecific rate defined as the moles of reactant consumed per gram of catalyst per minute. Total specific rates are divided by two to account for symmetry in molecules 11, 12, 15, 16, and ethylene glycol. cNR: no reaction. dMass ratio of catalyst:reactant = 2:7. eAlkanes in gas phase and monoalcohols at trace levels. fRing-closing rate denoted in parentheses. gCatalyst used was 4 wt% Rh/C.

We suggest instead that the experimentally observed hydrogenolysis selectivity trends are consistent with a bifunctional mechanism in which the catalyst facilitates initial acid-catalyzed ring-opening at the more substituted carbon center followed by metal-catalyzed hydrogenation.

To elucidate the observed differences in reactivity and product selectivity, a quantitative measure of the reactivity of these different molecules over solid acids is needed. Accordingly, a Born cycle analysis can be used to establish a relationship between the changes in the activation barrier and changes in the catalyst or the molecules reacted. More specifically the activation barrier (Ea) for a reaction can be written in terms of the deprotonation energy (DPE) of the acid, which provides a measure of solid acidity, the gas phase carbenium ion reaction or formation energies (ΔHrxn(R+)), and the interaction energy (ΔEint) between the gas phase carbenium ion and the negative charge of the conjugate base that forms in the transition state47, 48:

Ea =DPE+ΔH rxn(R+) +ΔE int (1)

As the catalyst is identical for all of the reactions considered in Table 3-8, the reactions can therefore be described by the changes in the gas phase carbenium ion reaction energies,

ΔHrxn(R+), and changes in the interaction energies, ΔEint. If the catalyst is the same, the changes in the interaction energies that result from different reactants are often linearly related to the changes in the proton affinity of the reactant or the gas phase carbenium ion reaction energies49.

The acid-catalyzed C-O activation is thought to proceed via the simultaneous protonation of the

63 oxygen atom in the ring and ring-opening, thus resulting in the formation of a secondary carbenium which can be stabilized through a Coulombic interaction between the positive charge on the carbenium ion center and the neighboring oxygen on the hydroxyl group through a back- side interaction as shown in Figure 3-8. This secondary carbenium ion intermediate can rapidly undergo a hydride transfer with the neighboring –CH-OH intermediate to form the more stable

RCH2-C=O(+)H oxocarbenium intermediate depicted in Figure 3-8 which more effectively stabilizes the charge on the oxygen atom. Alternatively, the ring-opening can proceed directly with a concerted hydride transfer from the neighboring CH-OH group to form the more stable

RCH2-C=O(+)H intermediate. While we have calculated the reaction energies to form the oxirane (3-membered ring) and oxetene (4- membered ring) as well as the RCH2-C=O(+)H oxocarbenium ions that result from the stabilization of the OH intermediates, we herein discuss only the results for the energetically more stable oxocarbenium ion intermediates. To further demonstrate and ensure the validity of gas phase carbenium ion calculation results, more detailed analyses of the reaction paths and corresponding activation barriers were performed for the five- membered ring structures 3 and 7 over a model Rh200Re1OH cluster, as outlined below.

The results from DFT calculations for simple gas phase carbenium ion reaction energies presented in Figure 3-7 and Table 3-5 show that the high selectivities for ring-opening 1 and 3 to the α,ω-diol can be attributed to the greater stability (30 kJ mol-1) of the secondary carbenium ion structures leading to the α,ω-diol, versus the primary carbenium ion structures which result in the formation of the 1,2-diol. Furthermore, the presence of hydroxyl group at the α or β position allows for the concerted protonation and hydride transfer to form the more stable oxocarbenium

(RCH2-CH=O(+)H) ion intermediate. For cyclic ethers, the presence of the hydroxyl substituent

α to the dehydration center is shown to increase the stability of the carbenium ion, such that the

64 carbenium ion formation energies for 1 and 3 are -845 and -852 kJ mol-1, respectively, while the corresponding structures without the α-OH substituent, 5 and 7, are significantly less stable at -

742 and -743 kJ mol-1, respectively (Figure 3-7 and Table 3-9 DFT-calculated gas phase reaction energies for the formation of carbenium and oxocarbenium ion intermediates involved in the ring-opening of cyclic ethers and dehydration of polyols.). This trend was confirmed by more rigorous transition state simulations over model Rh200Re1OH clusters, displayed in Figure

3-8, which show that the activation barrier to ring-open 3 is lower by 14 kJ mol-1 compared to the barrier to open 7 as a result of the back-side stabilization of the positive charge in the carbenium ion transition state by the OH group and by stabilization with water. The concerted protonation and hydride transfer reaction to form the oxocarbenium ion transition state lowers the activation barrier by an additional 16-84 kJ mol-1 as shown in the results displayed in Figure

3-9. While these calculations explain the selectivities reported in Table 3-8, the absolute values of the activation barriers are likely over-predicted due to the simplicity of the model system; both the presence of water and acid sites stronger than those in the modeled system would lead to lower activation barriers.

Figure 3-7 DFT-calculated carbenium and oxocarbenium ion formation or reaction energies for 2,3-butanediol 16, 2,4-pentanediol 15, and tetrahydrofurfuryl alcohol 3. The dotted lines refer to sequential paths which proceed via the formation of the OH-stabilized three (oxirane) and four (oxetene) ring intermediates that subsequently form the corresponding oxocarbenium ion intermediates whereas the solid lines refer to concerted protonation/hydride transfer steps that result in the direct formation of the oxycarbenium ion intermediates.

Table 3-9 DFT-calculated gas phase reaction energies for the formation of carbenium and oxocarbenium ion intermediates involved in the ring-opening of cyclic ethers and dehydration of polyols.

Figure 3-8 DFT-calculated reactant and transition state structures and corresponding activation barriers for the acid catalyzed ring-opening of: (a) water-stabilized tetrahydrofurfuryl alcohol 3, (b) tetrahydrofurfuryl alcohol 3 and (c) 2-methyltetrahydrofuran 7 on a model Rh200Re1OH cluster.

Figure 3-9 DFT-calculated reactant and transition state structures for the concerted protonation, hydride transfer and ring-opening of the water-stabilized tetrahydrofurfuryl alcohol 3 over a model Rh(111) surface with well dispersed Re-OH sites.

Figure 3-10 DFT-calculated reaction path and energies for the metal catalyzed ring-opening of 2-methyltetrahydrofuran 7 at the substituted and unsubstituted carbon centers over Rh.

The ring-opening of cyclic ethers 5 and 7 is characteristically different from the functionalized cyclic structures discussed above, in that the C-O activation occurs at the sterically less-hindered primary carbon-oxygen bond, leading to the formation of the corresponding secondary alcohols (i.e., 9 and 2-pentanol 10, respectively). This preferential hydrogenolysis of 7 at the primary carbon-oxygen bond is similar to that observed by Gennari, et al. who performed gas-phase reactions over unpromoted Pt catalysts29. The activation of the ring for these cyclic ethers, however, cannot be explained through acid chemistry, as it would result in the formation of a highly unstable primary carbenium ion. Instead, these experimental trends are consistent with DFT results for ring-opening of cyclic ethers via a solely metal-catalyzed route over Rh (Figure 3-10), showing that metal-catalyzed insertion is favored at the less

69 substituted C-O bond by 30 kJ mol-1 due to steric hindrance at the substituted C-O bond, preventing the initial dehydrogenation which leads to C-O cleavage.

3.5.2 Diols and polyols

To study further the role of hydroxyl groups in the hydrogenolysis of neighboring C-O bonds, the reactivity profiles for several straight-chain polyols over the 4 wt% RhRe/C catalyst were examined, and these results are shown in Table 3-8. For all diols studied, it is apparent that

1,2-diols are more reactive than α,ω-diols. Furthermore, the selectivity is significantly higher towards hydrogenolysis of the hydroxyl group at the secondary carbon atom for all 1,2-diols, leading to the formation of the corresponding primary alcohol as the predominant product. The high reactivity for hydrogenolysis of a secondary hydroxyl group adjacent to a primary hydroxyl group is also exhibited in the reactivity trends of the triols. In addition, a secondary hydroxyl group in close proximity to another secondary hydroxyl group shows high reactivity for undergoing hydrogenolysis over 4 wt% RhRe/C (1:0.5). These reactivity trends show that the position of hydroxyl groups in a reactant molecule is instrumental in dictating its hydrogenolysis reactivity over 4 wt% RhRe/C (1:0.5).

As previously shown for cyclic ethers, the above experimental trends for polyols are also consistent with a bifunctional catalyst which facilitates initial acid-catalyzed carbenium ion chemistry followed by metal-catalyzed hydrogenation. The low hydrogenolysis rates of α,ω-diols

2, 1,5-pentanediol 11, and 1,4-butanediol 12 are in agreement with the lower stability of primary carbenium ions which would form upon dehydration of these reactants. The higher reactivity and selective hydrogenolysis of 1,2-diols, such as 1,2-pentanediol 13 and 1,2-butanediol 14, at the secondary C-O bond can be explained by the formation of the more stable RCH2-CH=O(+)H oxocarbenium ion intermediates resulting from the concerted elimination of water and hydride

70 transfer from the primary –CH2OH group as reported in Table 3-9. Additionally, the higher reactivities of β-diols such as 2,4-pentanediol 15 and 2,3-butanediol 16 over that for 1,2-diols can be similarly explained by the formation of the secondary oxocarbenium ions in 15b and 16b which are over 100 kJ mol-1 more stable than the secondary carbenium ions in 15c and 16c.

3.5.3 Governing Principles of Substrate Reactivity and Selectivity

The experimental reactivity and selectivity trends shown in Table 3-8 are in agreement with the carbenium ion formation energies presented in Table 3-9. The reactants studied are ranked from the most to the least reactive in Table 3-10, along with their corresponding carbenium ion reaction energies. The results indicate that the most reactive molecules are the cyclic ethers which contain hydroxyl substituents that are located α to the carbon atom in the C-

O bond where the ring is opened. These reactants can undergo concerted ring-opening and hydride transfer from the neighboring –CH2OH group to form stable primary oxocarbenium ion transition state structures which are further stabilized by the OH groups that result upon protonation of the ring. As such, structures 1 and 3 are two of the most reactive structures examined with specific reaction rates of 90 and 180 μmol g-1 min-1, respectively.

Table 3-10 Comparison of specific hydrogenolysis rates over 4 wt% RhRe/C (1:0.5) and DFT- calculated carbenium ion energies for various cyclic ethers and polyols.

Reactant Specific rate Carbenium Ion Oxocarbenium Ion -1 -1 -1 -1 Structure Name (μmolg min ) Energies (kJ mol ) Energies (kJ mol )

Cyclic Ethers with α-OH Groups tetrahydrofurfuryl alcohol 3 180 -762 -852

2-(hydroxymethyl) 90 -756 -845 tetrahydro-pyran 1 β-Diols

2,4-pentanediol 15 117 -754/-799 -857

71 2,3-butanediol 16 156 -762 -857

1, 2-Diols and Polyols

1,2,4-butanetriol 47 -743 -835 OH HO OH glycerol 4 45 -732 -810 1,2-butanediol 14 62 -736 -830 1,2,6-hexanetriol 23 -733 -825 1,2-hexanediol 40 -734 -832 1,2-pentanediol 13 32 -741 -833 Substituted Cyclic Ethers 2-methyltetrahydropyran 5 7 -742 -

2-methyltetrahydrofuran 7 7 -743 - Cyclic Ethers and α,ω-Diols tetrahydropyran 6 - -710 -

tetrahydrofuran 8 - -715 - 1,6-hexanediol 2 - -712 - 1,5-pentanediol 11 4 -710 - 1,4-butanediol 12 8 -707 -

β-diols (15 and 16) also lead to stable oxocarbenium ion structures that can effectively stabilize the carbenium ion charge that results upon dehydration. These structures appear to be just as reactive as those for 1 and 3 with specific rates of 58 (15) and 78 (16) μmol g-1 min-1. In agreement with this experimental finding, the theoretical results show similar carbenium ion formation energies. The structures of carbenium ions from linear polyols with α,β-OH groups are less stable, as the oxocarbenium ions that form are now delocalized over primary carbon atoms in the oxocarbenium ion structure without the additional OH stabilization in the transition state; the reactivities of these reactants lie between 23-62 μmol g-1 min-1. Linear polyols that do not contain α,β or β,β-OH substituents are unable to form an oxocarbenium ion structures and must effectively stabilize the charge on the secondary carbon centers. As such the reactivities of these reactants are significantly lower and found to be in the range of 10 μmol g-1 min-1. Finally,

72 all of the primary carbenium ion centers from linear and cyclic polyols are unstable, and as such these reactants do not show significant reactivity.

Figure 3-11 Comparison of the ratio between specific reaction rate (μmol g-1 min-1) and initial reactant concentration (μmol mL-1) (on a logarithmic scale) and DFT-calculated carbenium and oxocarbenium ion energies for selected cyclic ethers and linear polyols. Each region shows distinct shifts in activity due to the stability of the carbenium and oxocarbenium ions as discussed in Table 3-6.

The results from Table 3-8 (with the exception of the data for the ring opening of 5 and 7 which are metal catalyzed reactions, and data from the unreactive α,ω-polyols) are plotted in

Figure 3-11, showing a correlation between the carbenium ion formation energy and the logarithm of the normalized rate expressed as the ratio between specific reaction rate (μmol g-1 min-1) and initial reactant concentration (μmol mL-1). This ratio is employed to reflect an appropriately normalized reaction rate which is independent of the minor differences in the

73 initial molar concentrations of reactants. It is also notable that specific rate values were divided by two to account for molecules with a plane of symmetry, namely, 11, 12, 15, 16, and ethylene glycol. The differences in reactivity between unsubstituted (primary carbenium ions), substituted

(secondary), and substituted and OH-stabilized cyclic ethers and polyols are highlighted by different colored regions in Figure 3-11.

We note here that we have included glycerol in the correlation of Figure 3-11, showing that the overall reactivity for conversion of this molecule is controlled by acid catalysis, consistent with the high selectivity for the formation of 1,3-propanediol, compared with the metal-catalyzed route over Rh/C. It is interesting that the rate of formation of 1,2-propandiol over the RhRe/C catalyst is also much higher than the corresponding reaction over Rh/C, suggesting that an acid-catalyzed pathway exists involving a secondary carbenium ion along the way to form the 1,2 oxirane that can subsequently react to form both 1,2 and 1,3-propanediol products close to the relative percentages that are seen experimentally.

3.5.4 Reaction kinetics

Although it was impractical to evaluate the complete reaction rate expression for all fifteen of the different substrates in Table 3-8, we evaluated the reaction kinetics for our prototypical cyclic ether, 2-(hydroxymethyl)tetrahydropyran (compound 1), and made the reasonable assumption that other substrates in this study will behave similarly. Importantly, we used a continuous flow reactor for this study to eliminate any possible complications that could be caused by initial transient phenomena (e.g., catalyst deactivation, initial leaching of rhenium) that could take place in a batch reactor. The result of these flow reactor studies, presented in

Table 3-11, show that the reaction is approximately first order with respect to species 1 and slightly less than first order (i.e., approximately equal to 0.9) with respect to hydrogen pressure.

74 Because the reactivity trends of this paper indicate that acid sites must be involved in a kinetically-significant part of the overall reaction scheme, one might have expected the reaction to be zero order in hydrogen. However, the observed positive order is consistent with the bifunctional nature of the catalyst and complexity of the mechanism, as illustrated by Boudart and Djega-Mariadassou in studies of the influence of hydrogen on the kinetics for isomerization of n-pentane to isopentane over a bifunctional Pt/alumina catalyst (containing metal sites and acid sites), even though hydrogen does not appear in the overall stoichiometric equation of alkane isomerization50. In that example, the rate determining step is the acid-catalyzed isomerization of n-pentene to i-pentene, and the observed effect of the hydrogen pressure is caused by the equilibrium of n-pentane with n-pentene on the metal sites prior to the rate determining acid-catalyzed isomerization step. In our case, the observed positive order in hydrogen results from the required hydrogenation of an unsaturated intermediate after the acid- catalyzed rate determining step.

Table 3-11 Hydrogenolysis rates of 2-(hydroxymethyl)tetrahydropyran 1 to 2 over 4 wt% RhRe/C (1:0.5) in a continuous flow reaction system.a

Concentration of 1 Conversion of 1 Selectivity to 2 Rateb P (psi) (μmol mL-1) H2 (%) (%) (μmolg-1min-1) 439 520 20 92 32 226 520 31 90 24 94 520 20 85 7

435 760 14 87 21 520 10 91 15 340 5 85 10 a 3 -1 Reaction conditions were 393 K, water as solvent. The catalyst was pretreated in flowing H2 (60 cm (STP) min ) at 523 K for 4 h and cooled to the reaction temperature prior to initiation of liquid feed flow. bSpecific rate defined as the moles of 1 reacted per gram of catalyst per minute.

75 In summary, a simple first order reaction rate constant cannot be calculated for the reactions in this paper since they occur over a bifunctional catalyst containing acid sites and metal sites, both of which contribute to the overall reaction. However, because the reaction appears to be approximately first order with respect to the concentration of the organic reactant, we have instead chosen to correlate observed rates divided by the reactant concentration, to approximate the rate at nearly identical experimental conditions. The trend in Figure 3-11 of this normalized rate versus the calculated carbenium ion formation energies is surprisingly good, given the fact that rates vary by more than an order of magnitude across the series of compounds.

3.5.5 Role of rhenium and nature of the active site

We suggest that the importance of acid chemistry over the rhodium-rhenium catalyst originates from the Brønsted acidity of hydroxylated Re atoms on Rh-Re nanoparticles, arising from strong Re-O bonds, resulting in a weak O-H bond as well as high electron affinity for the conjugate base. Other solid acids such as zeolites47 and heteropolyacids (HPAs)51 achieve this high electron affinity through the distribution of the negative charge across multiple oxygen atoms. As shown in Figure 3-12, hydroxylated rhenium species on Rh-Re nano-particles demonstrate deprotonation energies at the corner, edge and terrace sites of 1140 kJ mol-1, as compared to values of 1050-1200 kJ mol-1 for well-known solid acids such as zeolites47 and

HPAs51. A more detailed analysis of the electronic structure shows that there is electron transfer from Re to the oxygen. Also, it is evident from Figure 3-12 that the DFT-predicted acidities of hydroxylated molybdenum species on RhMo nanoparticles are markedly lower than that for

RhRe, consistent with the lower hydrogenolysis rates observed of 1 over the 4 wt% RhMo

(1:0.1) catalyst (Table 3-5). Additionally, we found that the hydrogenolysis of 1 over monometallic Rh/C in the presence of 0.1 M H2SO4 or HCl proceeds at rates and selectivities to 2

76 which are significantly lower than that over RhRe/C (Table 3-12). The above results thus suggest that the active sites for RhRe/C are located at the catalyst surface in the form of hydroxylated rhenium near rhodium; to find evidence for these sites, spectroscopy studies were performed and results are presented in the next chapter. Results from NH3 TPD studies (Figure

-1 3-5) show that the NH3-titrated site density over this catalyst is approximately 40 μmol g .

While it is not apparent what the exact nature of the interaction between NH3 and surface sites on

RhRe/C is, the significant increase in the extent of NH3 absorption by RhRe/C compared to the monometallic catalysts Rh/C and Re/C suggests that NH3 serves as a probe for surface sites unique to the bimetallic catalyst. A more detailed discussion of the significance of NH3 TPD results for RhRe/C is presented in the next chapter.

Figure 3-12 DFT-calculated deprotonation energies for various surfaces and atom clusters.

77 Table 3-12 Effect of homogeneous acid and base on hydrogenolysis rate and product selectivity of 2-(hydroxymethyl)tetrahydropyran 1 to hexanediols over Rh-containing catalysts.a

Sel to Sel to 1,2- Time Conv Specific rateb Catalyst 1,6-hexanediol 2 hexanediol (h) (%) (μmolg-1min-1) (%) (%) 4 wt% RhRe/C (1:0.5) 4 27 97 0 90 4 wt% RhRe/C (1:0.5) & 0.1M 4 NRc - - - NaOH 4 wt% Rh/C 20 NRc - - -

4 wt% Rh/C & 0.1M H2SO4 20 13 41 1 9 4 wt% Rh/C & 0.1M HCl 20 6 32 7 4 4 wt% Rh/C & 0.1M NaOH 20 NRc - - - a Reaction conditions were 393 K, 500 psi H2, mass ratio of catalyst: 1 = 1:9, 5 wt% 1 in water as feed. bSpecific rate defined as the moles of 1 reacted per gram of catalyst per minute. cNR: no reaction

We propose that the bifunctional nature of the RhRe/C catalyst described above can be generalized to other transition metal catalysts containing rhenium. For example, we have recently reported the promotion of the hydrogenolysis of 4 on Re-promoted Pt catalysts8. In that work, addition of Re to the Pt catalyst increased the turnover frequency of the hydrogenolysis of

4 by a factor of 20, with 1,3-propanediol 17 now appearing in the product mixture. Indeed, the selectivity of the hydrogenolysis reaction to 17 over Pt-Re was 34% at 443 K, a temperature at which Pt alone exhibited only trace conversion of glycerol.

Similarly, we observe here that the hydrogenolysis of 4 over 4 wt% RhRe/C (1:0.5) resulted in a marked increase in hydrogenolysis rates and selectivities to 17 compared to the performance of monometallic 4 wt% Rh catalyst (Table 3-5). The higher selectivities to 17 in the presence of rhenium are consistent with the premise that monometallic Pt or Rh are only able to facilitate metal-catalyzed C-O scission, while the addition of rhenium results in a bifunctional catalyst which is also capable of acid chemistry. King, et al.52 suggested the possible formation of acid sites associated with the presence formation of ReOx clusters in PtRe/C catalysts, and

78 they suggested the possible influence of these species in the aqueous phase reforming of glycerol. The basis of this proposed acidity over Pt-Re catalysts is clearly demonstrated by the

DFT data presented in Figure 3-12, wherein Pt-Re nano-particles are shown to display acidities comparable to that for RhRe nano-particles, suggesting that similar acid-catalyzed chemistry would be displayed in both systems.

Our proposed reaction scheme involving acid-catalyzed dehydration at the secondary carbon of a polyol, followed by hydrogenation to form the α,ω-diol, is consistent with the experimental result that the presence of NaOH completely suppressed the hydrogenolysis activity of 1 over 4 wt% RhRe/C (1:0.5) (Table 3-8). Also, previous work involving the hydrogenolysis of 4 over PtRe showed that addition of NaOH led to suppression of 20 formation8. Other investigators have also shown that promotion of Rh53 or Ru54 with Re enhances the formation of 17 from 4 in neutral water.

Finally, we note that it is possible to destroy the active sites for RhRe/C catalysts by reduction in H2 at elevated temperatures. In particular, the results in Figure 3-2 from temperature programmed reduction studies indicate that the Rh and Re in the RhRe/C catalyst are both highly reduced at temperatures above about 600 K. In addition, we have found that the catalytic activity of RhRe/C decreases when the catalyst is pre-treated in flowing H2 at increasing temperatures as shown for batch reactions (Table 3-7). The effect of reduction temperature on catalytic activity of RhRe/C will be examined in detail in Chapter 4.

3.6 Conclusions

In this chapter, we have demonstrated55 that a Re-promoted Rh/C catalyst is selective for the hydrogenolysis of secondary C-O bonds for a broad range cyclic ethers and polyols. It is also shown that a catalyst pretreatment strategy of reducing the catalyst in H2 at 523 K prior to

79 contact with the aqueous reactant results in the attainment of high catalytic activity and stability in a continuous flow reaction system. Importantly, results from experimentally-observed reactivity trends, NH3 TPD, and DFT calculations are consistent in supporting the hypothesis of a bifunctional catalyst which facilitates acid-catalyzed ring-opening and dehydration coupled with metal-catalyzed hydrogenation. Results from NH3 TPD indicates that the NH3-titrated site

-1 density and standard enthalpy of NH3 adsorption are approximately 40 μmol g and -100 kJ mol-1, respectively. Results from DFT calculations show that hydroxyl groups on rhenium associated with rhodium are acidic, thus enabling proton donation to reactant molecules and formation of carbenium ion transition states. The observed reactivity trends and DFT calculations are also consistent with the enhancement in stability of carbenium and oxocarbenium ion structures that form as a result of protonation and concerted protonation and hydride transfer steps, respectively. The experimentally-obtained rates of hydrogenolysis can be correlated with DFT-calculated carbenium ion formation energies. Results from this study suggest the bifunctional nature of active sites in metal catalysts promoted with oxophilic additives for hydrogenolysis reactions, and provides guidance for the use of this class of heterogeneous catalysts for the selective deoxygenation of biomass to fuels and chemicals. In the next chapter, we investigate the bifunctional nature of this RhRe/C catalyst in greater detail, through the use of probe reactions, NH3 TPD experiments, examination of solvent effects, and characterization of the catalyst under in situ and operando conditions using X-ray absorption spectroscopy in order to find evidence for the proposed hydroxylated rhenium species in

RhRe/C.

3.7 Chemical notations

IUPAC name Numerical notation

80 2-(hydroxymethyl)tetrahydropyran 1 1,6-hexanediol 2 tetrahydrofurfuryl alcohol 3 glycerol 4 2-methyltetrahydropyran 5 tetrahydropyran 6 2-methyltetrahydrofuran 7 Tetrahydrofuran 8 2-hexanol 9 2-pentanol 10 1,5-pentanediol 11 1,4-butanediol 12 1,2-pentanediol 13 1,2-butanediol 14 2,4-pentanediol 15 2,3-butanediol 16 1,3-propanediol 17 5-hydroxymethylfurfural 18

3.8 Computational methods

All gas phase energies were calculated using gradient-corrected DFT calculations as implemented in DMol3 38, 56, 57. Wavefunctions were represented by numerical basis sets of double numerical quality (DNP) with d-type polarization functions on each atom and expanded out to a 3.5 Å cutoff. The Perdew-Wang 9158 form of Generalized Gradient Approximation

(GGA) was used to model gradient corrections to the correlation and exchange energies. The electronic density for each self-consistent iteration converged to within 1×10-5 a.u. The energy in each geometry optimization cycle was converged to within 2×10-5 Hartree with a maximum displacement and force of 4x10-3 Å and 3x10-3 Hartree/Å, respectively.

The gas phase carbenium ion reaction or formation energies were calculated for ring structures as:

ROR + H+ → RORH+ROR H →RORH

81 where ROR and RORH+ refer to the initial ring structure and the ring-opened carbenium ion.

The energies for the linear polyols however result in dehydration and as such were calculated as:

+ + ROH + H → RH + H2OROH H →RH H2O where ROH and RH+ refer to the initial polyol and the resulting dehydrated carbenium ion.

The gas phase carbenium ion energies are used herein only as an initial probe of the activity and selectivity of the different molecules examined experimentally. This approach over simplifies the environment of the metal in solution. While the presence of the aqueous solution will clearly lower the overall reaction energies and activation barriers, we do not believe that selectivity patterns reported here will be altered.

All calculations on the metal surfaces and metal clusters were performed using gradient- corrected periodic plane wave DFT calculations as implemented in the Vienna Ab Initio

Simulation Program (VASP)59-61. The PW91 form of the GGA was used to determine gradient corrections to the exchange and correlation energies. Wavefunctions were constructed by expanding a series of plane waves within a cut-off energy of 400 eV where interactions between the core and valence electrons were modelled using Vanderbilt ultrasoft pseudopotentials62. For periodic calculations, a 4x4x1 Monkhorst-Pack k-point mesh63 was used to sample the first

Brillouin zone, for the calculations on the metal cluster, only the Γ-point was used. The electronic energies converged to within 1x10-4 eV and forces on each atom were optimized to within 0.05 eV Å-1. The 111 surfaces were modeled using a 3x3 unit cell with four metal layers and 16 Å vacuum region between each slab where the top two layers were allowed to relax and the bottom two held fixed to their lattice constant for Rh (3.8034 Å). The calculations on

RhRe201 clusters were carried out in a 30.4352 Å cubic unit cell to ensure a sufficient vacuum region to separate clusters in neighboring cells; all cluster atoms were allowed to relax in energy

82 calculations and transition state searches. Transition state searches were performed using the dimer64 method with the initial guesses for the transition state structure and the reaction trajectory obtained through the nudged elastic band method65.

3.9 References

1. Chheda, J. N., Huber, G. W. & Dumesic, J. A. Liquid-phase catalytic processing of biomass-derived oxygenated hydrocarbons to fuels and chemicals. Angewandte Chemie- International Edition 46, 7164-7183 (2007). 2. Schlaf, M. Selective deoxygenation of sugar polyols to ,-diols and other oxygen content reduced materials-a new challenge to homogeneous ionic hydrogenation and hydrogenolysis catalysis. Dalton Transactions, 4645-4653 (2006). 3. Koso, S. et al. Chemoselective hydrogenolysis of tetrahydrofurfuryl alcohol to 1,5- pentanediol. Chemical Communications, 2035-2037 (2009). 4. Chen, K., Koso, S., Kubota, T., Nakagawa, Y. & Tomishige, K. Chemoselective Hydrogenolysis of Tetrahydropyran-2-methanol to 1,6-Hexanediol over Rhenium- Modified Carbon-Supported Rhodium Catalysts. ChemCatChem 2, 547-555 (2010). 5. Nakagawa, Y., Shinmi, Y., Koso, S. & Tomishige, K. Direct hydrogenolysis of glycerol into 1,3-propanediol over rhenium-modified iridium catalyst. Journal of Catalysis 272, 191-194. 6. Ma, L. & He, D. Influence of catalyst pretreatment on catalytic properties and performances of Ru-Re/SiO2 in glycerol hydrogenolysis to propanediols. Catalysis Today 149, 148-156. 7. Ma, L., He, D. & Li, Z. Promoting effect of rhenium on catalytic performance of Ru catalysts in hydrogenolysis of glycerol to propanediol. Catalysis Communications 9, 2489-2495 (2008). 8. Daniel, O. M. et al. X-ray Absorption Spectroscopy of Bimetallic Pt–Re Catalysts for Hydrogenolysis of Glycerol to Propanediols. ChemCatChem 2, 1107-1114 (2010). 9. Kottke, R. H. in Kirk-Othmer Encyclopedia of Chemical Technology 259-286 (John Wiley & Sons, Inc, 2000). 10. Adkins, H. & Connor, R. The Catalytic Hydrogenation of Organic Compounds Over Copper Chromite. Journal of the American Chemical Society 53, 1091-1095 (2002). 11. Schniepp, L. E. & Geller, H. H. Preparation of Dihydropyran, Hydroxyvaleraldehyde and 1,5-Pentanediol from Tetrahydrofurfuryl Alcohol. Journal of the American Chemical Society 68, 1646-1648 (2002). 12. Tike, M. A. & Mahajani, V. V. Kinetics of Liquid-Phase Hydrogenation of Furfuryl Alcohol to Tetrahydrofurfuryl Alcohol over a Ru/TiO2 Catalyst. Industrial & Engineering Chemistry Research 46, 3275-3282 (2007). 13. Koso, S. et al. Promoting effect of Mo on the hydrogenolysis of tetrahydrofurfuryl alcohol to 1,5-pentanediol over Rh/SiO2. Journal of Catalysis 267, 89-92 (2009). 14. Perosa, A. & Tundo, P. Selective Hydrogenolysis of Glycerol with Raney Nickelâ€ Industrial & Engineering Chemistry Research 44, 8535-8537 (2005).

83 15. Dasari, M. A., Kiatsimkul, P.-P., Sutterlin, W. R. & Suppes, G. J. Low-pressure hydrogenolysis of glycerol to propylene glycol. Applied Catalysis A: General 281, 225- 231 (2005). 16. Chaminand, J. et al. Glycerol hydrogenolysis on heterogeneous catalysts. Green Chemistry 6, 359-361 (2004). 17. Wang, S. & Liu, H. C. Selective hydrogenolysis of glycerol to propylene glycol on Cu- ZnO catalysts. Catalysis Letters 117, 62-67 (2007). 18. Kurosaka, T., Maruyama, H., Naribayashi, I. & Sasaki, Y. Production of 1,3-propanediol by hydrogenolysis of glycerol catalyzed by Pt/WO3/ZrO2. Catalysis Communications 9, 1360-1363 (2008). 19. Miyazawa, T., Kusunoki, Y., Kunimori, K. & Tomishige, K. Glycerol conversion in the aqueous solution under hydrogen over Ru/C + an ion-exchange resin and its reaction mechanism. Journal of Catalysis 240, 213-221 (2006). 20. Miyazawa, T., Koso, S., Kunimori, K. & Tomishige, K. Development of a Ru/C catalyst for glycerol hydrogenolysis in combination with an ion-exchange resin. Applied Catalysis A: General 318, 244-251 (2007). 21. Maris, E. P. & Davis, R. J. Hydrogenolysis of glycerol over carbon-supported Ru and Pt catalysts. Journal of Catalysis 249, 328-337 (2007). 22. Montassier, C., Ménézo, J. C., Hoang, L. C., Renaud, C. & Barbier, J. Aqueous polyol conversions on ruthenium and on sulfur-modified ruthenium. Journal of Molecular Catalysis 70, 99-110 (1991). 23. Maris, E. P., Ketchie, W. C., Murayama, M. & Davis, R. J. Glycerol hydrogenolysis on carbon-supported PtRu and AuRu bimetallic catalysts. Journal of Catalysis 251, 281-294 (2007). 24. Lahr, D. G. & Shanks, B. H. Effect of sulfur and temperature on ruthenium-catalyzed glycerol hydrogenolysis to glycols. Journal of Catalysis 232, 386-394 (2005). 25. Shimao, A. et al. Promoting Effect of Re Addition to Rh/SiO2 on Glycerol Hydrogenolysis. Chemistry Letters 38, 540-541 (2009). 26. Shinmi, Y., Koso, S., Kubota, T., Nakagawa, Y. & Tomishige, K. Modification of Rh/SiO2 catalyst for the hydrogenolysis of glycerol in water. Applied Catalysis B: Environmental 94, 318-326. 27. Nakagawa, Y., Shinmi, Y., Koso, S. & Tomishige, K. Direct hydrogenolysis of glycerol into 1,3-propanediol over rhenium-modified iridium catalyst. Journal of Catalysis 272, 191-194 (2010). 28. Kreuzer, K. & Kramer, R. Support Effects in the Hydrogenolysis of Tetrahydrofuran on Platinum Catalysts. Journal of Catalysis 167, 391-399 (1997). 29. Gennari, U., Kramer, R. & Gruber, H. L. Hydrogenolysis of methyltetrahydrofuran on platinum: II. Effects of self-poisoning and evaluation of structure sensitivity. Applied Catalysis 44, 239-250 (1988). 30. Boffa, A. B., Lin, C., Bell, A. T. & Somorjai, G. A. Lewis Acidity as an Explanation for Oxide Promotion of Metals - Implications of Its Importance and Limits for Catalytic Reactions. Catalysis Letters 27, 243-249 (1994). 31. Boffa, A., Lin, C., Bell, A. T. & Somorjai, G. A. Promotion of CO and CO2 Hydrogenation over Rh by Metal-Oxides - the Influence of Oxide Lewis Acidity and Reducibility. Journal of Catalysis 149, 149-158 (1994).

84 32. Kip, B. J. et al. Hydrogenation of carbon monoxide over rhodium/silica catalysts promoted with molybdenum oxide and thorium oxide. Applied Catalysis 35, 109-139 (1987). 33. Reyes, P., Concha, I., Pecchi, G. & Fierro, J. L. G. Changes induced by metal oxide promoters in the performance of Rh-Mo/ZrO2 catalysts during CO and CO2 hydrogenation. Journal of Molecular Catalysis A: Chemical 129, 269-278 (1998). 34. Choi, S. H. & Lee, J. S. XAFS Characterization of Pt-Mo Bimetallic Catalysts for CO Hydrogenation. Journal of Catalysis 167, 364-371 (1997). 35. Johnston, P., Joyner, R., Pudney, P., Shpiro, E. & Williams, P. In Situ Studies of Supported Rhodium Catalysts. Faraday Discussions of the Chemical Society 89, 91-105 (1990). 36. Gault, F. G. & D.D. Eley, H. P. a. P. B. W. in Advances in Catalysis 1-95 (Academic Press, 1981). 37. Van't Blik, H. F. J. & Niemantsverdriet, J. W. Characterization of bimetallic FeRh/SiO2 catalysts by temperature programmed reduction, oxidation and Mössbauer spectroscopy. Applied Catalysis 10, 155-162 (1984). 38. Simonetti, D. A., Kunkes, E. L. & Dumesic, J. A. Gas-phase conversion of glycerol to synthesis gas over carbon-supported platinum and platinum-rhenium catalysts. Journal of Catalysis 247, 298-306 (2007). 39. Augustine, S. M. & Sachtler, W. M. H. On the mechanism for the platinum-catalyzed reduction of rhenium in PtRe/-Al2O3. Journal of Catalysis 116, 184-194 (1989). 40. Kunkes, E. L. et al. The role of rhenium in the conversion of glycerol to synthesis gas over carbon supported platinum-rhenium catalysts. Journal of Catalysis 260, 164-177 (2008). 41. Kunimori, K. et al. Percentage exposed of supported Pt, Pd and Rh catalysts studied by gas adsorption, tpr and tem methods. Applied Catalysis 4, 67-81 (1982). 42. Gatica, J. M. et al. Rhodium Dispersion in a Rh/Ce0.68Zr0.32O2 Catalyst Investigated by HRTEM and H2 Chemisorption. The Journal of Physical Chemistry B 104, 4667-4672 (2000). 43. Halttunen, M. E., Niemelä, M. K., Krause, A. O. I., Vaara, T. & Vuori, A. I. Rh/C catalysts for methanol hydrocarbonylation: I. Catalyst characterisation. Applied Catalysis A: General 205, 37-49 (2001). 44. Sharma, S. B., Meyers, B. L., Chen, D. T., Miller, J. & Dumesic, J. A. Characterization of catalyst acidity by microcalorimetry and temperature-programmed desorption. Applied Catalysis A: General 102, 253-265 (1993). 45. Niwa, M., Katada, N., Sawa, M. & Murakami, Y. Temperature-Programmed Desorption of Ammonia with Readsorption Based on the Derived Theoretical Equation. Journal of Physical Chemistry 99, 8812-8816 (1995). 46. Cardona-Martinez, N. & Dumesic, J. A. Thermochemical Characterization (eds. Ertl, G., Knozinger, H., Schuth, F. & Weitkamp, J.) (Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, 2008). 47. Janik, M. J., Macht, J., Iglesia, E. & Neurock, M. Correlating Acid Properties and Catalytic Function: A First-Principles Analysis of Alcohol Dehydration Pathways on Polyoxometalates. The Journal of Physical Chemistry C 113, 1872-1885 (2009).

85 48. Macht, J., Janik, M. J., Neurock, M. & Iglesia, E. Mechanistic Consequences of Composition in Acid Catalysis by Polyoxometalate Keggin Clusters. Journal of the American Chemical Society 130, 10369-10379 (2008). 49. Janik, M. J., Davis, R. J. & Neurock, M. The relationship between adsorption and solid acidity of heteropolyacids. Catalysis Today 105, 134-143 (2005). 50. Boudart, M. & Djega-Mariadassou, G. Kinetics of Heterogeneous Catalytic Reactions (Princeton University Press, 1984). 51. Brandle, M. & Sauer, J. Acidity Differences between Inorganic Solids Induced by Their Framework Structure. A Combined Quantum Mechanics/Molecular Mechanics ab Initio Study on Zeolites. Journal of the American Chemical Society 120, 1556-1570 (1998). 52. King, D. L. et al. Aqueous phase reforming of glycerol for hydrogen production over Pt- Re supported on carbon. Applied Catalysis B: Environmental 99, 206-213. 53. Shinmi, Y., Koso, S., Kubota, T., Nakagawa, Y. & Tomishige, K. Modification of Rh/SiO2 catalyst for the hydrogenolysis of glycerol in water. Applied Catalysis B: Environmental 94, 318-326 (2010). 54. Ma, L. & He, D. H. Hydrogenolysis of Glycerol to Propanediols Over Highly Active Ru- Re Bimetallic Catalysts. Topics in Catalysis 52, 834-844 (2009). 55. Chia, M. et al. Selective Hydrogenolysis of Polyols and Cyclic Ethers over Bifunctional Surface Sites on Rhodium-Rhenium Catalysts. Journal of the American Chemical Society 133, 12675-12689 (2011). 56. Delley, B. An all-electron numerical method for solving the local density functional for polyatomic molecules. The Journal of Chemical Physics 92, 508-517 (1990). 57. (Accelyrys, Inc., 2010). 58. Perdew, J. P. et al. Atoms, molecules, solids, and surfaces: Applications of the generalized gradient approximation for exchange and correlation. Physical Review B 46, 6671 (1992). 59. Kresse, G. & Hafner, J. Ab initio molecular-dynamics simulation of the liquid-metal- amorphous-semiconductor transition in germanium. Physical Review B 49, 14251 (1994). 60. Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Physical Review B 54, 11169 (1996). 61. Kresse, G. & Furthmüller, J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Computational Materials Science 6, 15- 50 (1996). 62. Vanderbilt, D. Soft self-consistent pseudopotentials in a generalized eigenvalue formalism. Physical Review B 41, 7892 (1990). 63. Monkhorst, H. J. & Pack, J. D. Special points for Brillouin-zone integrations. Physical Review B 13, 5188 (1976). 64. Henkelman, G. & Jonsson, H. A dimer method for finding saddle points on high dimensional potential surfaces using only first derivatives. The Journal of Chemical Physics 111, 7010-7022 (1999). 65. Jonsson, H., Mills, G. & Jacobsen, K. W. Nudged Elastic Band Method for Finding Minimum Energy Paths of Transitions (World Scientific, 1998).

Chapter 4 Nature of the active site over rhodium-rhenium catalysts

4.1 Introduction

In the previous chapter, we demonstrated that various bimetallic catalysts composed of a highly reducible metal and oxophilic component were effective in the deoxygenation of biomass- derived feedstock through C-O hydrogenolysis. It is notable that the role of oxophilic promoters

(e.g., Re) on the catalytic activity and selectivity of reducible metals (e.g., Pt, Rh) has been an area of recent interest in the literature, particularly for selective C-O scission reactions relevant to the conversion of biomass-derived feedstock. For example, the addition of Re to Pt catalysts has been shown to cause an increase in the activity for aqueous phase reforming (APR) of glycerol, and this effect was attributed to a decrease in the binding energy of CO to Pt and an increase in the water-gas-shift activity of the catalyst 1-3. Similarly, PtMo catalysts have been reported to be effective in the APR of glycerol 4. As mentioned in the previous chapter, various authors have also demonstrated that the addition of Re to Ru, Pt, and Ir catalysts causes an enhancement in C-O hydrogenolysis activities of glycerol and an increase in the selectivity to

1,3-propanediol compared to the corresponding monometallic catalysts 5-8. For instance, work by

Daniel et al. 8 reported that the hydrogenolysis of glycerol over Pt/C results in the formation of ethylene glycol, 1,2-propanediol, and 1-propanol as the main products with no 1,3-propanediol detected, while selectivities of 20-35% to 1,3-propanediol were observed over PtRe/C.

Furthermore, the total glycerol hydrogenolysis activity of PtRe/C was an order of magnitude higher than Pt/C.

Interestingly, besides C-O scission reactions, the addition of Re to Pt and Ru catalysts has been shown to enhance activities for the hydrogenation of carboxylic acids. For example,

Manyar et al. reported that the addition of Re to Pt/TiO2 increased the hydrogenation activity for the conversion of stearic acid to stearyl alcohol compared to Pt/TiO2, although selectivities to

9 stearyl alcohol over PtRe/TiO2 were decreased to 70%, compared to 93% over Pt/TiO2 . Also, the addition of Re to a Ru/C catalyst results in higher activity and stability for the hydrogenation of levulinic acid to γ-valerolactone compared to Ru/C in the presence of sulfuric acid, and the

Ru-Re catalyst displayed high activity for formic acid decomposition 10.

In the previous chapter, we demonstrated that the combination of an oxophilic promoter and a highly reducible metal results in the formation of a bifunctional catalyst possessing both acid and metal sites 11. We established this concept using results from experimental studies and first-principle density functional theory (DFT) calculations, where the reactivity trends over the

RhRe/C catalyst were consistent with a bifunctional mechanism in which selective C-O hydrogenolysis proceeds through initial acid-catalyzed ring-opening or dehydration followed by

11 metal-catalyzed hydrogenation . Temperature-programmed desorption (TPD) of NH3 was used as a means to quantify these active sites over RhRe/C, and the apparent Brønsted acidity of this catalyst was proposed to arise from the deprotonation of hydroxyl groups on rhenium atoms associated with rhodium. Similarly, recent work by Zhang et al. has reported the presence of acid sites on a PtRe catalyst under glycerol reforming conditions using NH3 TPD analysis and examination of product distributions 12; their work provides further evidence of the general bifunctional nature of this unique class of catalysts.

In this chapter, we show that the reduction temperature used to pretreat a RhRe/C (molar ratio Rh:Re = 1:0.5) catalyst prior to exposure to continuous-flow reaction conditions has a significant effect on the catalytic activity and the number of active sites quantified by NH3 TPD.

Additionally, we have studied this catalyst for an acid- catalyzed reaction that is important in

88 biomass conversion processes: the dehydration of biomass-derived fructose to 5- hydroxymethylfurfural (HMF) (Figure 4-1). 13-16 Importantly, we have studied the nature of the active sites on this RhRe/C catalyst using X-ray absorption spectroscopy (XAS) under in situ and operando conditions to probe the relationship between catalyst structure and catalytic activity.

Finally, the effect of the solvent and changes in water concentration on catalytic activity for C-O hydrogenolysis over RhRe/C is investigated.

O OH +H2 OH HO

2-hydroxymethyl- 1,6-hexanediol tetrahydropyran

HO OH HO O -3H2O O OH O H+ HO OH Fructose 5-hydroxymethylfurfural

Figure 4-1 Reactions discussed in this chapter: C-O hydrogenolysis of 2- (hydroxymethyl)tetrahydropyran to 1,6-hexanediol and dehydration of glucose to 5- hydroxymethylfurfural.

4.2 Effect of reduction temperature on C-O hydrogenolysis activity and acid

site density

Initial experiments were carried out in the batch reaction mode to examine the effect of pretreatment temperature on the catalytic activity of RhRe/C. Catalysts were pretreated in

500 psi H2 at temperatures of 393, 523, and 723 K prior to introduction of the reactant (5 wt%

HMTHP in water). Results displayed in Table 4-1 show that the catalytic activity for C-O hydrogenolysis of HMTHP and the amount of Re leached into solution decrease with increasing

89 pretreatment temperature. The selectivity to 1,6-hexanediol (1,6-HDO) is maintained at more than 90% for all pretreatment temperatures examined. It is observed that an intermediate pretreatment temperature of 523 K leads to undetectable leaching of Re, while maintaining significant catalytic activity for C-O hydrogenolysis. Interestingly, further increases in pretreatment temperature from 523 to 723 K lead to a sharp decline in C-O hydrogenolysis activity.

Table 4-1 Effect of pretreatment temperature on catalytic activity for C-O hydrogenolysis of HMTHP to 1,6-HDO and amount of Re leached into solution under batch reaction conditions over RhRe/C.a

Pretreatment T (K) Re leachedb (%) Conversion (%) Selectivity to 1,6-HDO (%) None 2.0 27.3 97.0 393 1.2 25.2 93.3 523 NDc 14.9 >99 723 NDc 4.6 >99 a Reaction conditions: 5 wt% HMTHP in water, 4 h, 500 psi H2, 393 K, mass ratio of catalyst:HMTHP = 1:9. b Percent of total Re detected in liquid as determined by ICP. cNot detectable.

The effect of reduction temperature on the catalytic activity for C-O hydrogenolysis was further investigated under continuous flow reaction conditions using RhRe/C pretreated at the same temperatures selected for batch reaction studies (Figure 4-2). Continuous flow reaction conditions were employed to better observe catalytic stability versus time and for the measurement of C-O hydrogenolysis rates in the absence of transient behavior (e.g., heating times to reaction temperature) present under batch reaction conditions. As shown in Figure 4-2, the rate of C-O hydrogenolysis of HMTHP decreased with increasing pretreatment temperatures, as observed in the batch reaction studies. For example, the initial rate of C-O hydrogenolysis of

HMTHP decreased from 30 to 7 µmolg-1min-1 with an increase in pretreatment temperature from

523 to 723 K. The selectivity to 1,6-HDO was maintained at >90% for all data points. It was

90 found that deactivation of the catalyst pretreated at 393 K was significant, while catalysts pretreated at 523 and 723 K were stable. The deactivation of the catalyst pretreated at 393 K is likely due to leaching of Re, which is more severe over long times under continuous flow conditions than that observed under batch reaction conditions (Table 4-1). Additionally, based on the monotonically decreasing activity of the catalyst with time-on-stream, TOS (as observed from a semi-log plot of rate as a function of time, Figure 4-3), the activity of the catalyst pretreated at 393 K does not appear to stabilize with time.

60 )

-1 50 min

-1 40

molg 

( 30

10 Specific rate 0 0 20 40 60 80 100 Time on stream (h)

Figure 4-2 Hydrogenolysis of HMTHP over RhRe/C under continuous flow reaction conditions. Catalyst pretreated in flowing H2 at 393 K (×), 523 K (○) and 723 K (■) prior to initiation of liquid feed. Conversion levels of HMTHP at 8 h time-on-stream were 9%, 9% and 10% for 393 K (×), 523 K (○), and 723 K (■), respectively. Selectivities to 1,6-HDO were >90% for all data -1 points. Reaction conditions: 5 wt% HMTHP in water as feed, WHSV393K=4.2 h , -1 -1 WHSV523K=2.2 h , WHSV723K=0.5 h , 393 K, 500 psi H2.

4.0

) -1

min 3.8

-1

molg 

( 3.6 ln ln rate

3.4 0 20 40 60 80 100 Time on stream (h)

Figure 4-3 Hydrogenolysis of HMTHP over RhRe/C under continuous flow reaction conditions. Catalyst pretreated in flowing H2 at 393 K prior to initiation of liquid feed. Reaction conditions: -1 5 wt% HMTHP in water as feed, WHSV = 4.2 h , 393 K, 500 psi H2.

In Chapter 311, we proposed that C-O hydrogenolysis occurs over RhRe/C through a bifunctional mechanism involving initial acid-catalyzed ring-opening followed by metal- catalyzed hydrogenation, and we used NH3 as a probe molecule to titrate sites associated with acidity over the catalyst. Accordingly, we measured the NH3 desorption profiles of RhRe/C pretreated at various temperatures, and the monometallic catalysts Rh/C and Re/C. These TPD spectra are presented in Figure 4-4. Desorption profiles for RhRe/C pretreated at 393 and 523 K were similar, with peak maxima centered at 430-445 K, indicating the presence of relatively weak NH3-adsorption sites. In contrast, Re/C displayed relatively low amounts of desorbed NH3, with peak maxima at 430 and 750 K. The NH3 desorption profiles of Rh/C and RhRe/C pretreated at 723 K were similar, with broad desorption peaks centered at 525 and 460 K, respectively, indicating a distribution of weak to moderately strong NH3-adsorption sites. The catalyst support, i.e., Vulcan carbon, showed a negligible amount of NH3 in the TPD spectrum.

0.4

0.3 f

0.2

desorption (a.u.) 3 d

NH 0.1 c b 0.0 a 400 600 800 1000 T (K)

Figure 4-4 NH3 temperature-programmed desorption profiles for: (a) Vulcan carbon, (b) Rh/C pretreated in flowing H2 at 523 K, (c) Re/C pretreated in flowing H2 at 523 K, (d) RhRe/C (1:0.5) pretreated in flowing H2 at 723 K, (e) RhRe/C (1:0.5) pretreated in flowing H2 at 523 K, and (f) RhRe/C (1:0.5) pretreated in flowing H2 at 393 K.

In general, the catalysts containing Rh (i.e., Rh/C and RhRe/C) display NH3-adsorption sites of similar strength, and significant differences are not observed between these catalysts apart from site densities. Acid site densities for all catalysts were determined by quantitative analyses of the NH3 desorption profiles (i.e., integration of area under desorption peaks), and the results are shown in Table 4-2. The site density over RhRe/C was found to decrease with increasing pretreatment temperatures. In our previous work, the number of metal sites over

RhRe/C was determined through measurement of the extent of irreversible CO uptake at 298 K for catalysts pretreated at the same temperatures employed here 11. Based on data from both CO and NH3 titrations, the molar ratio of CO titrated sites to sites titrated using NH3 for RhRe/C was determined to be 2.2, 4.3, and 16.9 after pretreatment at 393, 523, and 723 K, respectively. We note that the number of acid sites on the RhRe/C catalyst after various pretreatments was

93 calculated from the NH3 TPD spectrum by subtracting the amount of NH3 desorbed from the monometallic Rh/C catalyst; these values are denoted as “corrected site density” in Table 4-2.

The rationale for this correction of the NH3-titrated sites for RhRe/C is due to the known inactivity of Rh/C towards C-O hydrogenolysis of HMTHP under the reaction conditions of this study. Therefore, such a correction allows for the quantification of surface sites due to the presence of Re in RhRe/C.

The turnover frequency (TOF) values for C-O hydrogenolysis were calculated from the rates of HMTHP hydrogenolysis measured at 8 h time-on-stream (TOS) under continuous flow reaction conditions, by normalizing the rates for RhRe/C catalysts pretreated at various temperatures with the corrected site densities. As shown in Table 4-2, TOF values for C-O hydrogenolysis of HMTHP were found to be similar for all pretreatment temperatures examined, indicating that catalytic activity appears to be correlated with the number of NH3 titrated sites present over RhRe/C. The comparable TOF values obtained here are consistent with our previously proposed bifunctional C-O hydrogenolysis mechanism.

Table 4-2 Effect of pretreatment temperature on NH3 titrated site density and catalytic activity for hydrogenolysis of HMTHP to 1,6-HDO over RhRe/C (1:0.5) under continuous flow reaction conditions.a

Total Corrected Sel to Pretreat- Irreversible Specific rate NH site NH site Conv 1,6- TOFe Catalyst ment T CO uptakeb 3 3 at 8h TOS densityc densityd (%) HDO (min-1) (K) (µmolg-1) (µmolg-1 min-1) (µmolg-1) (µmolg-1) (%) 393 143 95 65 9 90 56 0.86 RhRe/C 523 145 64 34 9 96 29 0.85 723 135 38 8 10 95 7 0.93 Rh/C 523 -- 30 ------Re/C 523 -- 17 ------aReaction conditions described in Figure 4-2. bValues taken from Chia et al. 11. c Determined by NH3 TPD (Figure 4-4). dCorrected site density = (Total – Rh/C). e TOF calculated using corrected NH3 titrated site density.

94 4.3 Fructose dehydration to HMF

To further probe the apparent of acidity of RhRe/C, we employed an acid-catalyzed probe reaction, namely fructose dehydration to HMF13-16, a reaction of relevance to biomass conversion. Results from DFT calculations presented in our previous work 11 showed that the deprotonation energies of Re-OH species in contact with metallic Rh were in the range comparable to that for typical solid acid catalysts such as zeolites and heteropolyacids. Figure

4-5 shows results for the dehydration of fructose over RhRe/C that was pretreated at 523 K in flowing hydrogen prior to initiation of liquid feed flow. A pretreatment temperature of 523 K was informed by our results for C-O hydrogenolysis of HMTHP (Figure 4-2) showing that this pretreatment temperature leads to stable catalytic activity under continuous flow conditions. A single-phase reaction solvent consisting of a mixture of tetrahydrofuran (THF) and water (mass ratio of THF: water = 4:1) was employed, because this solvent system was previously shown to be effective for fructose dehydration to HMF under continuous flow reaction conditions 17.

Interestingly, the RhRe/C catalyst displayed high activity for conversion of fructose to HMF, with selectivity to HMF maintained at approximately 50% at a conversion level of 30% (Figure

4-5). While the selectivity to HMF was low for the first 24 h TOS, the selectivity increased and stabilized thereafter. The initial low selectivity to HMF is consistent with results previously reported by Tucker et al. 17 and Ordomsky et al. 18, 19 for a wide range of conventional solid acid catalysts such as zeolites, ZrPO4, sulfonated polystyrene and silica-based catalysts. Significantly, the RhRe/C catalyst here was stable with TOS and specific formation rates of HMF were maintained at approximately 1.8 µmolg-1min-1 for up to 250 h TOS. The formation of HMF from fructose here provides further evidence of Brønsted acid sites over RhRe/C.

60 10

HMF formation rate rate formation HMF

50 8

40 6 30 4

( 20 

mol

2 g

10 -1

min Conversion or Selectivity (%)

-1

0 0 ) 0 50 100 150 200 250 Time (h)

Figure 4-5 Conversion of fructose (×), selectivity to 5-hydroxymethylfurfural (●), and specific formation rate of HMF (■) as a function of time on stream over RhRe/C catalyst pretreated at 523 K. Reaction conditions: 300 psi He, 403 K, 2 wt% fructose in THF/water (mass ratio of THF: water = 4:1) as feed, WHSV =0.1h-1 .

The effect of the reduction temperatures on the catalytic activity for fructose dehydration was investigated, and these results are shown in Figure 4-6. The RhRe/C catalyst was pretreated in flowing hydrogen at 523 and 723 K, and fructose dehydration was conducted in the presence of an inert sweep gas (i.e., He). The lower pretreatment temperature of 393 K was not used here due to the instability of the catalyst for C-O hydrogenolysis (Figure 4-2). During the initial stages of the experiment (i.e., 48-72 h TOS), it is evident that the catalyst pretreated at the higher temperature is significantly less active, consistent with observations for C-O hydrogenolysis

(Figure 4-2). Interestingly, it was observed that the activity of the catalyst pretreated at 723 K increased after 100 h TOS, and eventually reached a similar rate of HMF formation rates as the catalyst that was pretreated at a lower temperature (i.e., 523 K). This increase in catalytic activity with TOS suggests that surface Re species, although significantly reduced in number after a high

96 temperature reduction pretreatment step (i.e., over-reduced), may repopulate the catalyst after prolonged exposure to liquid water and in the absence of a reducing environment. It is notable that this increase in catalytic activity with TOS for the catalyst pretreated at 723 K was not observed for experiments in the C-O hydrogenolysis of HMTHP (Figure 4-2), presumably due to the reducing environment these experiments were conducted under. The fructose conversion and HMF selectivity levels for the catalyst pretreated at 723 K from 120-215 h TOS were 30% and 45%, respectively.

) -1

min 1.5

-1

molg

 ( 1.0

0.5 HMF formation rate 0.0 50 100 150 200 Time on stream (h)

Figure 4-6 Fructose dehydration in a continuous flow reaction system over RhRe/C catalyst pretreated at different reduction temperatures of 523 K (■) and 723 K (×). Reaction conditions: 300 psi He, 403 K, 2 wt% fructose in THF/water (mass ratio of THF: water = 4:1) as feed, WHSV =0.1h-1.

ZSM-5 was used as a benchmark catalyst for fructose dehydration, for comparison with the dehydration activity of RhRe/C. Figure 4-7 shows the results for the dehydration of fructose to HMF over ZSM-5, using the same number of sites, as titrated by NH3 TPD, that were used as

97 for RhRe/C in the previous continuous flow experiments (Figure 4-5). The selectivity for production of HMF over ZSM-5 was similar to that over RhRe/C (i.e., approximately 40% selectivity at 20% conversion). Also, similar to what was observed for RhRe/C, the selectivity to

HMF was low initially, and then increased and stabilized thereafter. The HMF selectivity has been suggested to correlate with the strength and availability of Bronsted acid sites over solid acid catalysts 18, 19, and our results for RhRe/C and ZSM-5 thus suggest that the strength of

Brønsted acid sites over these two catalysts are similar. The turnover frequencies for production of HMF over RhRe/C and ZSM-5 are plotted versus TOS in Figure 4-8, using the number of sites titrated by NH3 TPD. On a rate-per-site basis, Figure 4-8 shows that RhRe/C displays HMF formation rates that are approximately twice as high as ZSM-5. This behavior is in agreement with the results from DFT calculations which indicate similar deprotonation energies for these catalysts.

60 8 rate formation HMF

50 6 40

30 4

(

 20 g mol 2

-1

10 min Conversion or Selectivity (%)

-1 0 0 ) 0 50 100 150 200 250 Time (h)

Figure 4-7 Conversion of fructose (×), selectivity to HMF (●), and specific formation rate of HMF (■) as a function of time on stream over ZSM-5. Reaction conditions: 300 psi He, 403 K, 2 wt% fructose in THF/water (mass ratio of THF: water = 4:1) as feed, WHSV =1.4 h-1.

3.5

) 3.0

-1 h -1 2.5

2.0 mol sites  1.5

1.0

HMF formation rate mol HMF

 0.5 (

0.0 0 50 100 150 200 Time on stream (h)

Figure 4-8 Specific formation rates of HMF over RhRe/C pretreated at 523 K (×) and ZSM-5 (■). Reaction conditions: 300 psi He, 403 K, 2 wt% fructose in THF/water (mass ratio of THF: water = 4:1) as feed.

4.4 Mass transfer effects

To verify that the reaction rates measured under flow conditions were not mass transfer

20 limited, the Weisz-Prater criterion was used. Accordingly, the Weisz-Prater (Nw-p) was calculated as follows:

-1 -3 Where is the reaction rate (mols cm ), Rp is the catalyst particle radius (cm), Cs is the

-3 reaction concentration at the particle surface (molcm ), and Deff is the effective diffusivity of reactant in the catalyst pores (cm2s-1).

99 In previous sections of this chapter, reaction rates were measured on a per gram catalyst basis, and therefore, these values were divided by the pore volume of the catalyst support to

-1 -3 obtain reaction rates in the appropriate units (i.e., mols cm )) for the calculation of Nw-p. The values for textural properties of Vulcan XC-72 are reported in the literature to be 228 m2g-1 (BET surface area) and 0.40 cm3g-1 (pore volume by the BJH method).21 The maximum reactions rates measured under flow conditions were 60 μmolg-1min-1 and 3 μmolg-1min-1, for C-O hydrogenolysis of HMTHP and fructose dehydration, respectively. The surface concentration of fructose and HMTHP were assumed to be equal to their bulk concentrations, at 2 wt% (1.125 ×

10-4 molcm-3) and 5 wt% (4.304 × 10-4 molcm-3), respectively. The particle radius was estimated by STEM images presented in Chapter 3 at around 50-100 nm; a particle radius of 100 nm was assumed for the most conservative estimate of Nw-p. The effective diffusivity of fructose in water-filled pores was estimated to be 2 × 10-9 cm2s-1 from values of glucose diffusivity in HY- zeolite. The value of the effective diffusivity of HMTHP in water-filled pores was not available in the literature, and was estimated from the bulk diffusivity of tetrahydropyran in water22 (i.e., 8

-6 2 -1 -9 2 -1 × 10 cm s ) to be 8 × 10 cm s . Accordingly, the Nw-p values for HMTHP and fructose in

-4 -5 Vulcan XC-72 were calculated to be 3 × 10 and 1 × 10 , respectively. As both values of Nw-p are lower than 0.3, the reaction rates measured herein are therefore free of mass transfer limitations.

4.5 X-ray absorption studies

X-ray absorption measurements were obtained to gain insight into the structure of the

RhRe/C catalyst after pretreatment and under C-O hydrogenolysis conditions. X-ray absorption spectra for the Rh K-edge and the Re LIII-edge were obtained for the monometallic catalysts

100

Rh/C and Re/C, and RhRe/C in He after pretreatment in flowing H2 at 298, 313, 333, 363, 393,

523, and 723 K, which is here denoted as in situ conditions. Absorption spectra at the Rh K and

Re LIII-edges were also measured under operando conditions for RhRe/C catalyst pretreated in flowing H2 at 523 K and 723 K and maintained under continuous flow C-O hydrogenolysis reaction conditions (i.e., 500 psi H2, 393 K, 5 wt% HMTHP in water as feed).

4.5.1 Monometallic catalysts Rh/C and Re/C under in situ conditions

Rh K and Re LIII-edge X-ray absorption spectra were obtained for the monometallic catalysts Rh/C and Re/C, and the values of fit parameters from the R-space of the EXAFS spectra are presented in Table 4-3. For the Rh/C catalyst, the Rh K-edge XANES (Figure 4-9a) and Fourier transforms (FTs) of the EXAFS spectra (Figure 4-9b) collected in He after pretreatment in flowing H2 at various temperatures in the range of 298 to 723 K indicate that Rh is highly reduced (~50%) after pretreatment at 298 K, and Rh is fully reduced after pretreatment at 363 K. A small amount of Rh-O coordination was observed for samples pretreated at temperatures lower than 363 K. Table 4-3 shows that the Rh-Rh contribution at a bond distance of 2.67-2.68 Å is observed, and this value is similar to that for Rh foil (Table 4-4).

For the Re/C catalyst, the Re LIII-edge XANES (Figure 4-10a) and FTs of the EXAFS spectra (Figure 4-10b) collected in He after pretreatment in flowing H2 indicate that Re is fully oxidized after pretreatment at 298 to 523 K. No remaining Re-O contribution after reduction at

723 K was observed, while a Re-Re contribution was detected with coordination number of 9.8 and bond distance of 2.74 Å, similar to that for Re foil (Table 4-4). The Re-Re coordination number corresponds to Re nanoparticles of about 5 nm. In general, these results indicate that monometallic Rh is highly reducible even at 298 K, while monometallic Re, as expected, is highly oxophilic and does not reduce easily to the metallic state.

101

1.0 Normalized Absorption

23.2 23.22 23.24 23.26

Photon Energy (keV)

)

2 0.06 -

0.04

× χ(k)] (Å χ(k)] ×

FT[k 0.02

0 1 2 3 4

R (Å)

Figure 4-9 (a) Rh K-edge XANES (23.19-23.27 keV) of Rh2O3 standard (black) and 4 wt% Rh/C catalyst reduced at 298 K (green), 313 K (blue), and 723 K (red). (b) Magnitude of k2- weighted FT of EXAFS data of Rh K-edge of Rh foil (black) and 4 wt% Rh/C catalyst reduced at 298 K (green) and 723 K (red) (k = 2.7 – 12.5 Å-1).

102 a

3.0

2.0

1.0 Normalized Absorption

10.52 10.53 10.54 10.55 10.56

Photon Energy (keV)

) 0.03

2 -

0.02

× χ(k)] (Å χ(k)] ×

FT[k 0.01

0 1 2 3 4

R (Å)

Figure 4-10 (a) Re LIII-edge XANES (10.51 - 10.56 keV) of 4 wt% Re/C reduced at 298 K (red) 2 and 723 K (blue). (b) Magnitude of k -weighted FT of EXAFS data of Re LIII-edge of 4 wt% Re/C catalyst reduced at 298 K (red) and 723 K (blue).

103

Table 4-3 Fit of Rh K and Re LIII-edge XANES and EXAFS for monometallic Rh/C and Re/C catalysts reduced at varying reduction temperatures.a

Treatment/ XANES fit Δσ2 Edge energy Absorber- ΔE Sample scan N R (Å) (×103 0 (keV) backscatterer (eV) conditions Å2) Rh Rh (III) Rh(0)

edge H 723 K/ - 1.0 Rh/C 2 23.2200 Rh-Rh 8.3 2.68 1.0 0.2 He RT H 523 K/ - 1.0 2 23.2202 Rh-Rh 8.2 2.68 1.0 0.9 He RT H 393 K/ - 1.0 2 23.2203 Rh-Rh 7.5 2.67 1.0 -0.3 He RT H 363 K/ - 1.0 2 Rh-Rh 8.4 2.68 1.0 0.3 He RT H 333 K/ 0.12 0.88 Rh-O 0.8 2.04 1.0 0.5 2 He RT Rh-Rh 6.8 2.67 1.0 0.1 H 313 K/ 0.14 0.86 Rh-O 0.8 2.03 1.0 0.3 2 He RT Rh-Rh 7.2 2.68 1.0 0.5 H 298 K/ 0.45 0.55 Rh-O 3.1 2.05 1.0 0.3 2 He RT Rh-Rh 3.6 2.68 1.0 0.9

Re Re(VII) Re(0)

edge H 723 K/ - 1.0 Re/C 2 10.5350 Re-Re 9.8 2.74 1.0 2.2 He RT H 523 K/ 0.95 0.05 Multiple overlapping Re-O 2 10.5393 Re-O He RT distances H 393 K/ 1.0 - 2 10.5399 Re-O 3.8 1.72 1.0 -1.4 He RT H 363 K/ 1.0 - 2 10.5399 Re-O 3.8 1.72 1.0 -1.9 He RT H 333 K/ 1.0 - 2 10.5340 Re-O 3.9 1.73 1.0 -1.8 He RT H 313 K/ 1.0 - 2 10.5340 Re-O 3.9 1.73 0.0 -2.4 He RT H 298 K/ 1.0 - 2 10.5399 Re-O 3.9 1.73 0.0 -1.6 He RT aThe estimated accuracies are: N, ±10%; R, ±0.02 Å.

4.5.2 RhRe/C under in situ and operando C-O hydrogenolysis conditions

Figure 4-11a shows that the Rh K-edge XANES spectra of Rh/C and RhRe/C exhibit differences in shape compared to a Rh foil, consistent with the formation of small nanoparticles even after reduction at 723 K. The position of the leading edge of the first peak in the XANES for Rh/C and RhRe/C is not significantly different from Rh foil, and the edge energies (i.e.,

104 energy at the first inflection point) are similar. Also, the whiteline intensities for Rh/C, RhRe/C, and Rh foil are comparable and suggest that Rh is fully reduced in each. Figure 4-11b shows that the Rh K-edge XANES spectra of RhRe/C reduced at 393, 523, and 723 K are similar and all exhibit edge energies and peak intensities indicative of fully metallic Rh.

The Re LIII-edge XANES of Re foil, ReO2, ReO3, and Re2O7 are shown in Figure 4-12.

The edge energy of the Re oxide references increases with increasing formal oxidation state

(Table 4-4). After pretreatment at 363 K, the edge energy of the Re LIII-edge for the RhRe/C catalyst is within experimental error of Re foil (Table 4-4), indicating that Re in RhRe/C is fully reduced. At the same reduction temperature, the edge energy, XANES and EXAFS spectra indicate that Re/C is Re(VII) oxide. This ease of reduction of Re under mild conditions for the

RhRe/C catalyst suggests that metallic Rh promotes the reduction of Re.

1.0

0.5 Normalized Absorption

0.0 23.20 23.22 23.24 23.26 Photon Energy (keV)

105

1.0

0.5 Normalized Absorption

0.0 23.20 23.22 23.24 23.26 Photon Energy (keV)

Figure 4-11 Rh K-edge XANES (23.19 – 23.27 keV) of (a) Rh foil (black), Rh/C (red), and RhRe/C catalyst reduced at 723 K (blue), and (b) RhRe/C catalyst reduced at 393 K (black), 523 K (red), and 723 K (blue); profiles for catalyst reduced at 393 K and 523 K overlap and are not

apparent in figure.

4.0

3.0

2.0

1.0 Normalized Absorption Normalized

10.52 10.53 10.54 10.55

Photon Energy (keV)

Figure 4-12 Re LIII-edge XANES (10.51 – 10.56 keV) of Re foil (red), ReO2 (blue), ReO3 (green), and Re2O7 (black).

106

The values of parameters from R-space fits of Rh K and Re-LIII-edge EXAFS spectra for

Rh and Re standards and RhRe/C are presented in Table 4-4. For the RhRe/C catalyst reduced at temperatures from 298 to 333 K, there is a Rh-O contribution with coordination number decreasing with increasing temperature from 3.2 to 0.4 indicative of partially oxidized Rh. It appears that Rh is fully reduced to the metallic state after pretreatment at temperatures higher than 333 K, which is close to the temperature required to fully reduce monometallic Rh/C

(Section 4.5.1).

The magnitude of the FTs of the Rh K-edge EXAFS spectra for the RhRe/C catalyst pretreated at 393 to 723 K are shown in Figure 4-13a. The Rh portion of the catalyst is fully reduced with no evidence of Rh-O bonds. The fits of the spectra indicate the presence of a Rh-Re scatter in addition to Rh-Rh. The Rh-Rh coordination (ca. 5-6) is larger than that of Rh-Re (ca. 2-

3). The Rh-Rh bond distance of 2.68 Å is similar to that for Rh foil; while the Rh-Re bond distance (i.e., 2.63 Å) is shorter than the average bond distance for Rh and Re foil references.

The coordination number of the Rh-Rh contribution here is significantly lower than that reported by Chen et al. for similar catalysts (6.4 compared to 9.2) 23 possibly due to lower average particle size in this study. The low coordination number of the Rh-Re contribution (1.6-2.8) at pretreatment temperatures of 313 to 723 K suggests that Re resides largely on surface of bulk Rh nanoparticles under these conditions. These results for the Rh-Rh and Rh-Re contributions and coordination numbers are consistent with RhRe nanoparticles having a core that is Rh-rich.

The magnitude of the FTs of the Re LIII-edge EXAFS spectra are shown in Figure 4-13b- c, and the fit of the EXAFS (Table 4-4) shows a Re-O contribution for the RhRe/C catalyst after pretreatment temperatures of 363 K and below, with an interatomic distance of 1.71Å and coordination number of 1.3 to 0.4 (298 K-333 K). The Re-O bond distance is consistent with a

107 Re(VII) oxide double bond, i.e., Re=O. The Re-O contribution and its corresponding low coordination number suggest that a mixture of Re states are present. The Re-O contribution declines and is not apparent after pretreatment temperatures above 363 K, suggesting that Re is reduced at relatively low temperatures (i.e. < 393 K), in agreement with results from the XANES edge energies and linear component analysis. The conclusion of fully reduced Re was also in agreement with delta-XANES and delta-EXAFS analyses. The ease of reduction of Re in

RhRe/C compared to monometallic Re/C suggests that all Re is in contact with metallic Rh in

RhRe/C. A Re-Re contribution was also observed, with a coordination number of 1.3-4.3, increasing with pretreatment temperature (298 - 723 K), and bond distance of 2.57 Å which is significantly contracted compared to that for Re foil (2.75 Å). In addition, a Re-Rh contribution was observed with coordination number of 2.4-3.9, increasing with temperature (298 K-723 K), with a bond distance of 2.63 Å. The Re-Rh (and Rh-Re) bond distance of 2.63 Å is equal to the sum of the atomic radii of Rh-Rh and Re-Re from the catalyst. The direct bonding between Rh and Re after reduction has been previously suggested for similar reduced catalysts (e.g.,

24 RhRe/SiO2) . At low pretreatment temperatures, the coordination of Re-Re is lower than the coordination of Re-Rh, consistent with the Rh-rich core and Re-rich shell structure suggested by the Rh K-edge EXAFS. At high pretreatment temperatures, (i.e., 523 and 723 K), the Re-Re coordination increases relative to the Re-Rh coordination, and the overall coordination of Re increases. These changes are consistent with the migration of Re from the surface into the Rh-

Re particle.

As noted previously, the reactivity trends of the RhRe/C catalyst for hydrogenolysis of

HMTHP and for dehydration of fructose indicate the presence of acid sites, suggesting the presence of Re-OH species on the surface of RhRe particles under reaction conditions.

108 Accordingly, we collected XAS measurements of this catalyst under operando conditions for hydrogenolysis of HMTHP in liquid water at 393 K using RhRe/C catalyst pretreated at 523 K and 723 K. Due to lack of high quality data over a large k range, the EXAFS fits were unable to distinguish differences in the catalysts reduced at the two temperatures, or irrefutably distinguish the results from the in situ experiments. However, the important result from the operando measurements is that the XANES edge energies were consistent with fully reduced catalysts.

Figure 4-14 shows the XANES of the RhRe/C catalyst reduced at 523 K under in situ and operando conditions at both the Rh and Re edges. The spectra are very similar and indicate that

Rh and Re are fully reduced. The matching XANES results and the inability to irrefutably distinguish the EXAFS results of the in situ and operando results suggests that the Re does not form Re-OH to a significant extent.

Table 4-4 Fit of Rh K and Re LIII-edge XANES and EXAFS for Rh and Re standards, RhRe/C catalyst reduced at varying reduction temperatures.a

Treatment/ Edge Δσ2 Absorber- ΔE Sample scan energy XANES fit N R (Å) (×103 0 backscatterer (eV) conditions (keV) Å2) Rh edge Rh (III) Rh(0) Rh foil - 23.2200 - - Rh-Rh 12 2.68 0.0 -0.3 H 723 K/ Rh-Rh 5.0 2.68 2.0 -0.8 RhRe/C 2 23.2205 - 1.0 He RT Rh-Re 2.8 2.63 2.0 -8.5 H 523 K/ Rh-Rh 5.6 2.68 2.0 0.2 RhRe/C 2 23.2205 - 1.0 He RT Rh-Re 2.3 2.63 2.0 -9.3 H 393 K/ Rh-Rh 6.1 2.68 2.0 1.0 RhRe/C 2 23.2203 - 1.0 He RT Rh-Re 2.0 2.63 2.0 -10.5 H 363 K/ Rh-Rh 6.0 2.68 2.0 0.6 RhRe/C 2 23.2200 - 1.0 He RT Rh-Re 1.7 2.63 2.0 -10.8 Rh-Rh 5.6 2.68 2.0 1.1 H 333 K/ RhRe/C 2 0.06 0.94 Rh-Re 1.7 2.63 2.0 -11.0 He RT Rh-O 0.4 2.05 1.0 2.2 Rh-Rh 5.3 2.68 2.0 -0.2 H 313 K/ RhRe/C 2 0.10 0.90 Rh-Re 1.6 2.63 2.0 -11.4 He RT Rh-O 0.4 2.05 1.0 2.2 H 298 K/ Rh-Rh 3.7 2.68 2.0 1.0 RhRe/C 2 0.40 0.60 He RT Rh-O 3.2 2.05 1.0 2.6

Re edge Re(VII) Re(0) Re foil - 10.5350 - Re-Re 12 2.75 0.0 0.6

109

ReO2 - 10.5386 - - ReO3 - 10.5395 - - Re2O7 - 10.5403 - - Re-O 4 1.74 0.0 -0.3 H 723 K/ Re-Re 4.3 2.57 2.0 -11.6 RhRe/C 2 10.5350 0 1.0 He RT Re-Rh 3.9 2.63 2.0 5.9 H 523 K/ Re-Re 3.6 2.57 2.0 -11.1 RhRe/C 2 10.5350 0 1.0 He RT Re-Rh 3.4 2.63 2.0 7.1 H 393 K/ Re-Re 2.5 2.57 2.0 -11.3 RhRe/C 2 10.5350 0 1.0 He RT Re-Rh 3.5 2.63 2.0 5.9 H 363 K/ Re-Re 2.2 2.66 2.0 -11.0 RhRe/C 2 10.5351 0 1.0 He RT Re-Rh 3.5 2.63 2.0 5.7 Re-O 0.4 1.71 1.0 -6.2 H 333 K/ RhRe/C 2 10.5352 0.10 0.90 Re-Re 1.4 2.57 2.0 -11.4 He RT Re-Rh 3.1 2.63 2.0 5.2 Re-O 0.8 1.71 1.0 -5.3 H 313 K/ RhRe/C 2 10.5355 0.20 0.80 Re-Re 1.8 2.57 2.0 -10.7 He RT Re-Rh 2.9 2.63 2.0 5.3 Re-O 1.3 1.71 1.0 -4.4 H 298 K/ RhRe/C 2 10.5358 0.30 0.70 Re-Re 1.3 2.57 2.0 -10.5 He RT Re-Rh 2.4 2.63 2.0 5.5 aThe estimated accuracies are: N, ±20%; R, ±0.02 Å

) 0.02

-2

(

]

(k)

 ×

2 0.01

[ FT

0.00 0 1 2 3 4 R (Å)

110

0.010

)

-2

(

]

(k)

 ×

2 0.005

[ FT

0.000 0 1 2 3 4 R (Å)

0.010

)

-2

(

]

(k)

 ×

2 0.005

[ FT

0.000 0 1 2 3 4 R (Å)

Figure 4-13 Magnitude of k2-weighted FT of EXAFS data of (a) Rh K-edge of RhRe/C catalyst -1 reduced at 393 K (black), 523 K (red), and 723 K (blue) (k = 2.5 – 13.8 Å ). Re LIII edge of RhRe/C catalyst reduced at (b) 298 K (black), 313 K (red), and 333 K (blue) (k = 2.8 – 11.9 Å- 1), and (c) 393 K (black), 523 K (red), and 723 K (blue) (k = 2.7 – 13.3 Å-1). All data was collected at room temperature.

111

1.0

0.5 Normalized Absorption

0.0 23.20 23.22 23.24 23.26 Photon Energy (keV)

b 3

1 Normalized Absorption

0 10.52 10.53 10.54 10.55 10.56 Photon Energy (keV)

Figure 4-14 (a) Rh K-edge XANES (23.19 – 23.27 keV) of RhRe/C reduced at 523 K with spectra acquired under in situ conditions (black), and RhRe/C reduced at 523 K with spectra acquired under operando C-O hydrogenolysis conditions (red). (b) Re LIII-edge XANES (10.51 – 10.56 keV) of RhRe/C reduced at 523 K with spectra acquired under in situ conditions (black), and RhRe/C reduced at 523 K with spectra acquired under operando C-O hydrogenolysis conditions (red).

112 4.6 Nature of the active site

The results from XAS measurements show that increasing pretreatment temperatures are accompanied by an increase in the coordination numbers of the Rh-Re (313 - 723 K, 1.6-2.8) and

Re-Rh (298 - 723 K, 2.4-3.9) contributions, indicative of an increased extent of alloying between the two metals. The increase in degree of alloying between Rh and Re atoms with increasing pretreatment temperature is also evident from the R-space of the Re LIII-edge EXAFS spectra displayed in Figure 4-13c. As presented above (Section 4.2-4.3), our reaction kinetics results demonstrate that a decrease in activity for C-O hydrogenolysis and dehydration over RhRe/C occurs with an increase in pretreatment temperature, and these changes are accompanied by a decrease in site density as measured by NH3 TPD. This decrease in catalytic activity and NH3 titrated site density correlates with the increase in extent of alloying between Rh and Re as determined through the fit of the Rh K and Re LIII-edge EXAFS. Therefore, the results from reaction kinetics and XAS indicate that the presence of Re on the surface of Rh particles is responsible for the high catalytic activity of RhRe/C for C-O hydrogenolysis and dehydration reactions. As reduction temperature increases, Re coordination increases indicating its migration into the nanoparticle which results in a decrease in catalytic activity.

The observed decrease in the Re-O contribution from the Re LIII-edge XANES and

EXAFS with increasing pretreatment temperatures is consistent with work reported by Koso et al. for similar catalysts 24. After pretreatment at 363 K, the EXAFS results for RhRe/C show no evidence of Re-O coordination and indicate the formation of bimetallic nanoparticles with a Rh- rich core and Re-rich surface. The presence of fully reduced Re is also evident under operando conditions for the RhRe/C catalyst pretreated at 523 and 723 K and exposed to liquid water and flowing H2 at 393 K, i.e., C-O hydrogenolysis conditions. Therefore, rhenium oxide does not

113 appear to be necessary to impart acidic functionality to the catalyst. In view of the oxophilic nature of Re compared to Rh, we suggest that the apparent acidity of RhRe/C may be generated by the activation of water molecules through interaction of the O atom in water with small clusters or even single atoms of Re on the surface of metallic Rh (i.e., Re-Rh species). The

25 activation of water over Re in similar catalysts (e.g., PtRe/TiO2) has been previously reported .

An increase in the pretreatment temperature causes a decrease in low coordination Re atoms on the surface of Rh particles (i.e., increased extent of alloying) which decreases the abundance of surface Re-Rh species, thereby leading to a decrease in activity for dehydration and C-O hydrogenolysis. Furthermore, as the pretreatment temperature is increased from 393 to 723 K, the total coordination to Re increased from 6.0 to 8.2 which would decrease the ability of Re to activate water.

To investigate the influence of water concentration on catalytic activity, we performed batch experiments for the C-O hydrogenolysis of HMTHP over RhRe/C using solvents consisting of THF/ water mixtures of varying compositions. As shown in Figure 4-15, C-O hydrogenolysis activity over RhRe/C steadily decreases with decreasing water concentration.

Importantly, no hydrogenolysis activity was observed when dry THF was the reaction solvent.

Therefore, the decline in C-O hydrogenolysis activity with decreasing water concentration, and the inactivity of RhRe/C under water-free conditions provides further evidence of the role of water in conferring an acidic functionality over this catalyst.

114

THF concentration (wt%) 100 80 60 40 20 0 120

100 )

-1 80

min 60 -1

molg 40

 (

20 Rate 0

0 20 40 60 80 100 Water concentration (wt%)

Figure 4-15 Rate of hydrogenolysis of HMTHP over RhRe/C with varying THF/ water mixture compositions under batch reaction conditions. Catalyst pretreated under 34 bar H2 at 393 K prior to introduction of liquid feed. Selectivities to 1,6-HDO were >90% for all data points. Reaction conditions: 5 wt% HMTHP in THF/ water mixtures as feed, mass ratio of catalyst:HMTHP = 1:7, 4 h, 393 K, 34 bar H2.

Interestingly, we found that the activity for fructose dehydration of the RhRe/C catalyst pretreated at 723 K increased after 100 h TOS, and eventually achieved similar HMF formation rates as the catalyst that was pretreated at a lower temperature (i.e., 523 K). This increase in

HMF formation rates with TOS suggests that the number of active sites increases with prolonged exposure to water at 393 K and in an inert atmosphere. We thus suggest that whereas Re migrates into the RhRe nanoparticle under reducing conditions, Re migrates back to the surface in the absence of a reducing atmosphere and in the presence of liquid water, increasing the concentration of surface Re active sites for fructose dehydration. This preferential occupation of oxophilic metals on the surface of nanoparticles enriched with a highly reducible metal under

115 liquid water conditions is consistent with work by Dietrich et al. for a PtMo/C catalyst4.

Additionally, the subsurface migration of Re in RhRe alloys under reducing conditions is consistent with work by Greeley and Mavrikakis, according to which surface Rh overlayers and subsurface Re have been predicted to form stable near surface alloy structures for RhRe bimetallics26.

In addition to being oxophilic, the low occupancy of the d-band in metallic Re also leads to Lewis acid like behavior and strong binding with nitrogen. Accordingly, NH3 serves as a probe molecule for the measurement of dispersed Re species on the surface of Rh particles. NH3

TPD results shown in Figure 4-4 and Table 4-2 indicate that the extent of NH3 adsorption is significantly higher for the bimetallic RhRe/C compared to the monometallic Re/C, which is consistent with the high dispersion and low coordination of Re on the surface of Rh particles in the RhRe/C catalyst as determined from the Re and Rh EXAFS. The Re XANES results for monometallic Re/C show that Re remains as rhenium oxide after pretreatment at 523 K, and is likely to be poorly dispersed, resulting in a low extent of NH3 adsorption. For the bimetallic

RhRe/C catalyst, an increase in pretreatment temperature has been shown from EXAFS to result in an increase in extent of alloying between Re and Rh, such that fewer Re atoms are present on the surface of Rh particles, consistent with a decrease in amount of NH3 adsorbed. Therefore, we suggest that NH3 serves as a probe molecule for measuring the dispersion of Re atoms on the surface of Rh particles, and thus the sites active for hydrogenolysis and dehydration.

The results from XAS indicate the onset of rhenium oxide reduction to metallic Re at room temperature in RhRe/C, which is consistent with the vast majority of Re being in contact with Rh. Therefore, the XAS results are representative of the behavior of Re in RhRe particles.

To probe the possible difference in catalytic activity for disparate RhRe alloy compositions, we

116 prepared and studied two catalysts with compositions determined from reported bulk RhRe alloy phase diagram (i.e., using compositions at the phase boundaries)26. The two catalysts that were prepared correspond to a Re-rich (i.e., 81 at% Re, 19 at% Rh) and a Rh-rich (i.e., 12 at% Re, 88 at% Rh) catalyst, both having the same total metal loading (molar basis) as the RhRe/C catalyst.

The two catalysts were pretreated at 393 K under 34 bar H2 prior to introduction of liquid feed (5 wt% HMTHP in water), and reactions were performed under batch conditions for 4 h at 393 K under 34 bar H2. The observed rates for C-O hydrogenolysis of HMTHP normalized based on the total metal loading (i.e., Rh and Re), the extent of irreversible uptake of CO adsorption, and the

-1 amount of NH3 adsorption were 0.69, 0.61, and 2.04 min , respectively, for the Re-rich catalyst, and 0.57, 0.24, and 1.00 min-1, respectively, for the Rh-rich catalyst. Therefore, while the higher activity of the Re-rich catalyst underscores the key role of Re in RhRe catalysts for C-O hydrogenolysis, the disparity in catalyst activity is small and demonstrates that large variations in catalyst composition do not have a remarkable effect on catalytic behavior.

Finally, we note that we have attributed the acidic properties of our RhRe/C catalyst to the presence of Brønsted acid sites formed in the presence of water, in view of our previous results from DFT calculations, the observations of HMF formation from fructose, and the inactivity of RhRe/C for C-O hydrogenolysis under water-free conditions (i.e., using dry THF as the solvent). Another area that could be explored in the future is that the presence of Re on the surface of Rh particles forms Lewis acid sites under anhydrous reaction conditions. Additionally, in light of the preliminary operando results and the work by Greeley and Mavrikakis, we cannot definitively rule out that the working state of the RhRe/C catalyst under reducing conditions may be a nanoparticle with a shell enriched in Rh on a RhRe subsurface alloy. In this case, the role of

Re would be to alter the catalytic properties of Rh at the surface26.

117 4.7 Effect of solvent on catalytic activity

To further examine the effect of varying water concentration on catalytic activity, mixtures of various organic solvents and water were used as the reaction medium. Figure 4-16 shows results for the hydrogenolysis of HMTHP over RhRe/C, where dry catalyst was pretreated at 393 K prior to introduction of liquid feed under batch reaction conditions. As previously shown in Figure 4-15, the rate of 1,6-HDO formation declined with a decrease in water concentration for THF/ water mixtures. Similarly, the use of other alcohols and 2- methyltetrahydrofuran led to a decrease in C-O hydrogenolysis activity over RhRe/C (Figure

4-16). Of the various alcohols examined, it was found that primary alcohols were the most effective in suppressing C-O hydrogenolysis activity, followed by furans. Secondary alcohols were observed to result in the smallest decrease in catalytic activity. Specifically, at a water concentration of 91 wt%, the organic solvents in order of the most to least effective in suppressing of catalytic activity was as follows: 1-butanol > methanol >1-propanol > THF > 2- methyltetrahydrofuran > 2-butanol > 2-propanol. This observed trend could be due to a combination of varying water concentration through simple dilution and the blocking of sites on the catalyst surface by the alcohols. While it is currently unclear why this specific trend is observed, it is evident that availability of water at the catalyst surface affects C-O hydrogenolysis activity significantly. Therefore, our postulate that water activation results in Brønsted acidity over RhRe/C is consistent with these results.

118

THF 100 water methanol 1-propanol 80 2-propanol

1-butanol )

-1 2-butanol

60 2-methylTHF

min -1

molg  ( 20

Rate of 1,6-HDO formation 0

0 20 40 60 80 100 Water concentration (wt%)

Figure 4-16 Rate of 1,6-hexanediol (1,6-HDO) formation over 4 wt% RhRe/C (1:0.5) as a function of water concentration. Solvents were mixtures organic solvents with water. Reactions were performed under batch conditions with 5 wt% HMTHP in water as feed, 393 K, 500 psi H2, 4 h, dry catalyst pretreated under 500 psi H2 at 393 K prior to introduction of liquid feed. Selectivities to 1,6-HDO were more than 95% in all experiments.

4.8 Conclusions

The pretreatment temperature of the RhRe/C catalyst has a significant effect on activity of the catalyst for the hydrogenolysis of HMTHP and fructose dehydration to HMF in liquid water under continuous flow reaction conditions. We show that a bimetallic catalyst displays substantial activity and selectivity for fructose dehydration to HMF (i.e., 50% selectivity to HMF at 30% conversion of fructose), providing evidence for Brønsted acidity over this RhRe/C catalyst in liquid water. The turnover frequency for production of HMF (with acid sites titrated by temperature programmed desorption of NH3) was found to be two-times higher over RhRe/C compared to ZSM-5. Catalytic activities for both C-O hydrogenolysis and fructose dehydration

119 were observed to decrease with an increase in pretreatment temperature, which coincides with a decrease in the number of low coordination metallic Re atoms on the surface of the catalyst measured by XAS and with the number of sites quantified using NH3 TPD. The rates of C-O hydrogenolysis of HMTHP normalized by the number of sites titrated by NH3 TPD were nearly identical (0.85-0.93 min-1) for the RhRe/C catalysts pretreated at 393, 523, and 723 K, which suggests a direct correlation between catalytic activity and sites titrated using NH3. Results for the characterization of RhRe/C using XAS after reduction in H2 are consistent with the formation of RhRe nanoparticles with Rh-rich cores and a shell of metallic Re islands. Furthermore, there was no evidence of rhenium oxide (i.e., Re-O contribution in the Re LIII-edge EXAFS) after the catalyst was pretreated at temperatures of 393 K and above in flowing H2 (i.e., in situ conditions) or under C-O hydrogenolysis operando conditions. Additionally, the EXAFS spectra are consistent with an increase in extent of alloying between Rh and Re in the RhRe/C catalyst with increasing pretreatment temperatures. Based on these results from reaction kinetics studies and

XAS measurements, we suggest that the acidity of RhRe/C in liquid water is generated from the activation of water molecules through the interaction of the O atom in water with Re atoms on the surface of metallic Rh (i.e., Re-Rh species), causing the formation of Brønsted acidity.

Importantly, we have also observed the inactivity of RhRe/C for C-O hydrogenolysis of HMTHP under water-free conditions, which provides further evidence of the role of water in conferring an acidic functionality over this catalyst. Additionally, NH3 serves as a probe molecule for the measurement of these highly dispersed Re species on the surface of Rh particles. An increase in the reduction pretreatment temperature decreases the abundance of these surface Re-Rh species through increased alloying between Re and Rh, causing a decrease in amount of NH3 adsorbed by the catalyst as determined through NH3 TPD, and leading to a decrease in catalytic activity

120 due to the decrease in number of active sites. Finally, we have observed that a decrease in water concentration results in a decline in C-O hydrogenolysis activity of RhRe/C, suggesting that the apparent Brønsted acidity over RhRe/C is related to the availability of water at the catalyst surface and consequently, extent of water activation.

4.9 References

1. Kunkes, E. L. et al. The role of rhenium in the conversion of glycerol to synthesis gas over carbon supported platinum-rhenium catalysts. Journal of Catalysis 260, 164-177 (2008). 2. Simonetti, D. A., Kunkes, E. L. & Dumesic, J. A. Gas-phase conversion of glycerol to synthesis gas over carbon-supported platinum and platinum-rhenium catalysts. Journal of Catalysis 247, 298-306 (2007). 3. King, D. L. et al. Aqueous phase reforming of glycerol for hydrogen production over Pt- Re supported on carbon. Applied Catalysis B: Environmental 99, 206-213 (2010). 4. Dietrich, P. J. et al. Aqueous Phase Glycerol Reforming by PtMo Bimetallic Nano- Particle Catalyst: Product Selectivity and Structural Characterization. Top. Catal. 55, 53- 69 (2012). 5. Ma, L., He, D. & Li, Z. Promoting effect of rhenium on catalytic performance of Ru catalysts in hydrogenolysis of glycerol to propanediol. Catalysis Communications 9, 2489-2495 (2008). 6. Nakagawa, Y., Shinmi, Y., Koso, S. & Tomishige, K. Direct hydrogenolysis of glycerol into 1,3-propanediol over rhenium-modified iridium catalyst. Journal of Catalysis 272, 191-194 (2010). 7. Shimao, A. et al. Promoting Effect of Re Addition to Rh/SiO2 on Glycerol Hydrogenolysis. Chemistry Letters 38, 540-541 (2009). 8. Daniel, O. M. et al. X-ray Absorption Spectroscopy of Bimetallic Pt–Re Catalysts for Hydrogenolysis of Glycerol to Propanediols. ChemCatChem 2, 1107-1114 (2010). 9. Manyar, H. G. et al. Highly selective and efficient hydrogenation of carboxylic acids to alcohols using titania supported Pt catalysts. Chemical Communications 46, 6279-6281 (2010). 10. Braden, D. J., Henao, C. A., Heltzel, J., Maravelias, C. C. & Dumesic, J. A. Production of liquid hydrocarbon fuels by catalytic conversion of biomass-derived levulinic acid. Green Chemistry (2011). 11. Chia, M. et al. Selective Hydrogenolysis of Polyols and Cyclic Ethers over Bifunctional Surface Sites on Rhodium-Rhenium Catalysts. Journal of the American Chemical Society 133, 12675-12689 (2011). 12. Zhang, L. et al. Correlation of Pt-Re surface properties with reaction pathways for the aqueous-phase reforming of glycerol. J. Catal. 287, 37-43 (2012). 13. Kuster, B. F. M. & Temmink, H. M. G. The influence of pH and weak-acid anions on the dehydration of d-fructose. Carbohydrate Research 54, 185-191 (1977).

121 14. Kuster, B. F. M. & S. van der Baan, H. The influence of the initial and catalyst concentrations on the dehydration of d-fructose. Carbohydrate Research 54, 165-176 (1977). 15. Román-Leshkov, Y., Chheda, J. N. & Dumesic, J. A. Phase Modifiers Promote Efficient Production of Hydroxymethylfurfural from Fructose. Science 312, 1933-1937 (2006). 16. Chheda, J. N., Roman-Leshkov, Y. & Dumesic, J. A. Production of 5- hydroxymethylfurfural and furfural by dehydration of biomass-derived mono- and polysaccharides. Green Chemistry 9, 342-350 (2007). 17. Tucker, M. H. et al. Acid-Functionalized SBA-15-Type Periodic Mesoporous Organosilicas and Their Use in the Continuous Production of 5-Hydroxymethylfurfural. ACS Catalysis 2, 1865-1876 (2012). 18. Ordomsky, V. V., van der Schaaf, J., Schouten, J. C. & Nijhuis, T. A. The effect of solvent addition on fructose dehydration to 5-hydroxymethylfurfural in biphasic system over zeolites. Journal of Catalysis 287, 68-75 (2012). 19. 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 5, 1812-1819 (2012). 20. Vannice, M. A. Kinetics of Catalytic Reactions (Springer, New York, 2005). 21. Soboleva, T. et al. On the Micro-, Meso-, and Macroporous Structures of Polymer Electrolyte Membrane Fuel Cell Catalyst Layers. ACS Applied Materials & Interfaces 2, 375-384 (2010). 22. Leaist, D. G., MacEwan, K., Stefan, A. & Zamari, M. Binary Mutual Diffusion Coefficients of Aqueous Cyclic Ethers at 25 °C. Tetrahydrofuran, 1,3-Dioxolane, 1,4- Dioxane, 1,3-Dioxane, Tetrahydropyran, and Trioxane. Journal of Chemical & Engineering Data 45, 815-818 (2000). 23. Chen, K., Koso, S., Kubota, T., Nakagawa, Y. & Tomishige, K. Chemoselective Hydrogenolysis of Tetrahydropyran-2-methanol to 1,6-Hexanediol over Rhenium- Modified Carbon-Supported Rhodium Catalysts. ChemCatChem 2, 547-555 (2010). 24. Koso, S., Watanabe, H., Okumura, K., Nakagawa, Y. & Tomishige, K. Stable Low- Valence ReOx Cluster Attached on Rh Metal Particles Formed by Hydrogen Reduction and Its Formation Mechanism. The Journal of Physical Chemistry C 116, 3079-3090 (2011). 25. Azzam, K. G., Babich, I. V., Seshan, K., Mojet, B. L. & Lefferts, L. Stable and Efficient Pt–Re/TiO2 catalysts for Water-Gas-Shift: On the Effect of Rhenium. ChemCatChem 5, 557-564 (2013). 26. Greeley, J. & Mavrikakis, M. Alloy catalysts designed from first principles. Nat Mater 3, 810-815 (2004).

122

Chapter 5 Catalytic transfer hydrogenation over metal oxides

5.1 Introduction

In this chapter, we examine another class of reactions of importance in biomass conversion: the selective hydrogenation of C=O bonds in the presence of C=C functional groups to obtain unsaturated alcohols as the target product (Figure 5-1). Although good chemoselectivity for the hydrogenation of unsaturated aldehydes to their corresponding unsaturated alcohols can be achieved over bimetallic catalysts with molecular hydrogen as the H source1, selectivity levels to the unsaturated alcohol are markedly lower when unsaturated ketones are used as the reactant2. Therefore, our approach towards accomplishing high chemoselectivity in such reactions is through the exploitation of the Meerwein-Ponndorf-Verley reaction which proceeds over metal oxide catalysts. Accordingly, alternative H sources (e.g., secondary alcohols) are used to achieve this chemistry and molecular hydrogen is not employed as a co-reactant.

R1 R2 2 OH O R R R R 1 2 1 2 4 1 O

R1 R2 3

Figure 5-1 Reaction pathways for the hydrogenation of unsaturated ketones (R2 = H) or aldehydes (R2 = CxHy) (1) to unsaturated alcohols (2), saturated ketones or aldehydes (3), and saturated alcohols (4).

123 5.1.1 Selective hydrogenation of unsaturated aldehydes and ketones over

heterogeneous metal catalysts

The selective hydrogenation of carbonyl functional groups in the presence of C=C bonds is a catalytically demanding reaction of significant importance for the production of fragrances and bio-active compounds. The selective production of allyl alcohols is also a highly relevant for the conversion of biomass-derived feedstock, and the ability to achieve high chemoselectivity is critical to overall process economics. Unfortunately, the thermodynamics for the hydrogenation of the C=O over a C=C functional group are generally quite unfavorable and it is well-known that the C=C bond is much more reactive than the C=O bond over transition metal surfaces. For example, the hydrogenation of methyl vinyl ketone (MVK) to 3-buten-2-ol instead of butanone is thermodynamically unfavorable by approximately 78 kJ/mol over Ru(0001), and with relatively higher energy barriers.2 The control of chemoselectivity in such reactions over transition metals is therefore generally difficult, as this requires the preferential binding of the C=O group over the

C=C bond3.

There are several approaches to promoting the hydrogenation of the C=O bond over the

C=C functional group over transition metals catalysts, such as the addition of a promoter or a second metal. The effect of promoter or metal addition has been attributed to the formation of an alloy, thereby resulting in ensemble size, ensemble composition, and/ or a ligand/ electronic effects. Promoters such as Sn and Fe have been shown to be effective in altering hydrogenation selectivities over monometallic catalysts when paired with transition metals such as Pt. In addition, the nature of the support and catalyst preparation methods have been shown to have an effect on catalytic selectivity.1 As previously mentioned, it is notable that the selective production of allyl alcohols from unsaturated ketones is more difficult than that for unsaturated

124 aldehydes; Ide et al. showed that while Au catalysts displayed modest selectivity towards the formation of the allyl alcohol when crotonaldehyde was the reactant (37-51% selectivity at 21-34% conversion), the same catalysts were completely unselective when MVK was the reactant.2

Theoretical calculations performed for a Ru(0001) surface indicated that the activation energy barrier towards C=C and C=O bond hydrogenation in MVK were 77-109 kJ/mol and 99-115 kJ/mol, respectively, consistent with experimental trends observed.2

An important and well-studied bimetallic alloy for the selective hydrogenation of unsaturated aldehydes and ketones is Pt catalysts modified with Sn. PtSn catalysts have been effective in the hydrogenation of unsaturated aldehydes such as citral4-6 and crotonaldehyde7, 8 to their corresponding unsaturated alcohols. Importantly, in terms of reactions relevant to biomass conversion to chemicals, PtSn catalysts have been reported to be highly selective for the hydrogenation of fufural to furfuryl alcohol.9-11 The high chemoselectivities for C=O bond hydrogenation over PtSn catalysts have been attributed to the formation of ionic Sn species (i.e.,

Snδ+) which act as Lewis acid sites that promote the hydrogenation of the C=O functional group12, 13; the addition of Sn has been suggested to result in the modification of the geometry of

Pt particles by surface enrichment of Sn, and decreased hydrogen adsorption capacity of the bimetallic14.

5.1.2 The Meerwein-Ponndorf-Verley reaction

The Meerwein-Ponndorf-Verley (MPV) reaction, also known as the Meerwein-Schmidt-

Ponndorf-Verley reduction, is a highly selective method for the hydrogenation of carbonyl functional groups to alcohols in the presence of other functional groups such as C=C, ester, and nitro groups.15 It is generally accepted that the reaction mechanism in the MPV reaction proceeds

125 through a six-membered cyclic transition state where the hydrogen donor (i.e., alcohol) and acceptor (i.e., aldehyde or ketone) are both coordinated to the same metal centre (Figure 5-2).

The alkoxide product formed undergoes alcoholysis during which a proton is abstracted from a H donor molecule (i.e., alcohol) from the bulk. This reaction mechanism is also known as the concerted or direct hydride-transfer pathway. Cohen et al. found that of the three possible reaction mechanisms for the MPV reaction (i.e., direct hydrogen transfer, hydridic, and radical), density functional theory calculations indicate that the direct hydrogen transfer pathway was the most favorable with the lowest energy barriers.16 The concerted mechanism was also found to be consistent with experimental kinetic isotope effect results, which indicated that C-H bond breaking and formation was the rate limiting step16. An unfavorable equilibrium for this reaction exists if the metal centre binds more strongly to the product than the reactant.

R R R2 R3 R2 3 R2 3 R R R1 H R4 R1 H 4 R1 H 4 C C C C C C O O O O O O M M M

1 2 3

Figure 5-2 Mechanism for the Meerwein-Ponndorf-Verley reaction proceeding through a six- membered transition state (2) over a metal alkoxide. Adapted from Chuah et al. and de Graauw et al.15, 17

The MPV reaction occurs over a wide variety of heterogeneous catalysts such as hydrous and calcined zirconia oxide18, 19, MgO17, 20, mixed metal oxides17, and zeolitic materials (e.g., Sn- beta21). The reduction of unsaturated carbonyl compounds by the MPV reaction has been hypothesized to occur over Lewis acid sites, basic sites, and acid-base pairs. Some of the main

126 side reactions observed in MPV reactions are aldol condensation, the Tischenko reaction

(producing carboxylic esters), and dehydration.

The main advantages of using the MPV reaction are in that the need for handling of hazardous gaseous hydrogen under high pressure or metal hydrides is eliminated, and close to quantitative chemoselectivity is often possible. In the latter, this is a significant advantage especially when a reaction is particularly challenging over metal catalysts, and a high value product is targeted. For example, the selective hydrogenation of MVK to buten-2-ol, while difficult over transition metals and using molecular hydrogen as the H source, can be performed with good selectivities through the exploitation of the MPV reaction. Specifically we have found that over MgO/Al2O3 and with isopropanol as the hydrogen donor, selectivity to buten-2-ol was

79% at 34% conversion (5 wt% MVK in isopropanol as feed, 355 K, 4h).

5.1.3 Hydrogenation of levulinic acid to γ-valerolactone

The hydrogenation of biomass-derivable levulinic acid (LA) and its esters22-26 to γ- valerolactone (GVL) (Figure 5-3) is a key reaction in the development of economically viable and carbon-efficient biorenewable routes to chemicals27, 28 and liquid transportation fuels29. LA and its esters possess a ketone group which may be hydrogenated to its corresponding hydroxyl form over suitable catalysts. Subsequent ring-closure of the hydroxyacid to GVL is typically rapid and occurs through intramolecular condensation. Group VIII metals, notably ruthenium28, have been shown to facilitate this hydrogenation step using molecular H2, and quantitative yields of GVL from LA easily attainable. The use of high reaction temperatures (> 473 K) in combination with Ni or Cu catalysts for the hydrogenation of LA has been shown to facilitate the formation of significant quantities of 2-methyltetrahydrofuran30-32 which can be used as a fuel

127 additive. It has recently been suggested33 that the use of precious metal catalysts is detrimental to the overall process economics for the production of high volume and relatively low value liquid transportation fuels. Thus, in this chapter we investigate a new route for the conversion of LA to

GVL using inexpensive heterogeneous catalysts, such as zirconium oxide.

Figure 5-3 Catalytic transfer hydrogenation of levulinic acid (1, R1 = H) and its esters (1, R1 = CxH2x+1) to γ-valerolactone (5) using a secondary alcohol as the hydrogen donor (2, R2 = CyH2y+1).

As an attractive alternative to the reduction of LA to GVL using H2 over metal catalysts, we have explored the reduction of LA by catalytic transfer hydrogenation (CTH), whereby a hydrogen source other than molecular H2 is used. Recent literature on the CTH of LA has focused on using formic acid (FA) as the hydrogen donor, since FA is a by-product of LA production from C6 sugars. However, the use of FA entails several unresolved disadvantages, such as the need for precious metals (e.g., Pd, Rh), homogeneous catalysts34, 35, and/ or harsh reaction conditions (i.e., hydrothermal conditions in the presence of salts36, 37) for both the generation of H2 from FA and the subsequent hydrogenation of LA to GVL.

128 In contrast to previous work, we note that CTH through the MPV reaction remains an unexplored alternative chemistry for this reduction step, offering important advantages for its application. For example, as mentioned in Section 5.1.2, the MPV reaction possesses exceptional chemoselectivity for the reduction of carbonyl groups under mild reaction conditions and in the presence of other functional groups (such as C=C double bonds). Significantly, the MPV reaction can take place over non-precious metal heterogeneous catalysts17. Moreover, the hydrogen donor in the MPV reaction, usually a secondary alcohol, can be recycled after hydrogenation over base metal catalysts such as nickel38-40 or copper41, 42, or even sold as a commodity chemical in its oxidized form (i.e., ketones, Figure 5-3, 4).

In this chapter, it is demonstrated that CTH via the MPV reaction is a viable means for the hydrogenation of LA and its esters over inexpensive, heterogeneous catalysts that are easily regenerable, with attainment of close to quantitative yields of GVL under appropriate reaction conditions. In this respect, the MPV reaction is uniquely able to meet the above-mentioned techno-economic demands for the production of GVL from LA and its esters. Importantly, it is shown that the MPV reaction can be exploited for the formation of GVL from LA and its esters over heterogeneous catalysts, which has hitherto been reported in only one instance to occur in the presence of homogenous catalysts43, and more recently, over zeolitic materials44.

5.2 Initial studies using metal oxide catalysts and levulinate esters

In initial reaction kinetic studies using batch reactors, it was found that a variety of metal oxides are active materials for the transfer hydrogenation of butyl levulinate (BL) to GVL (Table

5-1, Entries 1-5). Of the catalysts examined, ZrO2 was the most active material for CTH. We have further observed that the CTH reaction occurs at reaction temperatures > 373 K, and the

129 rate of formation of GVL increases with increasing reaction temperature (Table 5-1, Entries 9-

11), with no observable loss in GVL selectivity at temperatures of up to 493 K when secondary alcohols (e.g., isopropanol (IPA), 2BuOH) are employed as the hydrogen donor. The only by- products detected in the product mixtures were those formed through the transesterification of the levulinate ester with the secondary alcohol solvent (e.g, isobutyl levulinate as the product).

Table 5-1 Catalytic transfer hydrogenation of levulinate esters using various alcohols as the hydrogen donor.a

Time Mass ratio Conversion GVL yield GVL formation rate Entry Catalyst Solvent (h) catalyst: ester (%) (%) (μmol g-1 min-1) 1 MgO/Al2O3 16 1:2.4 2BuOH 17.0 14.6 1.9 2 MgO/ZrO2 16 1:2.4 2BuOH 12.6 8.0 1.1 3 CeZrOx 16 1:2.4 2BuOH 19.7 15.8 2.0 4 γ-Al2O3 16 1:2.4 2BuOH 37.0 29.6 3.9 5 ZrO2 16 1:2.4 2BuOH >99.9 84.7 -- 6 ZrO2 4 1:4.8 2BuOH 70.1 55.6 65.6 7 ZrO2 16 1:1.2 1BuOH 60.5 42.8 4.0 8 ZrO2 16 1:1.2 EtOH 98.2 49.2 -- b 9 ZrO2 4 1:4.8 IPA 36.6 12.2 14.2 10 ZrO2 4 1:4.8 IPA 77.2 55.0 70.8 c 11 ZrO2 4 1:4.8 IPA 93.2 80.5 108.4 d 12 ZrO2 4 1:4.8 IPA 70.8 62.4 86.0 e 13 ZrO2 16 1:1.2 THP 5.8 1.5 5.2 aBatch reactions, 5 wt% BL in respective solvents as feed, 423 K, 300 psi He. b393 K c453 K d5 wt% EL as feed. e5 wt% IL as feed. Entries 1-8 and 11-12: the only by-product detected was isobutyl levulinate, formed through the transesterification of BL or EL with the solvent.

5.3 Catalytic transfer hydrogenation using ZrO2

5.3.1 Effect of levulinate esters and hydrogen donor on catalyst activity

In contrast to using secondary alcohols as hydrogen donors, we found that although primary alcohols such as 1-butanol (Table 5-1, Entry 7, 1BuOH) and ethanol (Table 5-1, Entry 8,

EtOH) are able to facilitate CTH of levulinate esters leading to the formation of GVL, various

130 by-products are formed causing a decrease in GVL selectivity. These by-products were identified using GC-MS as forming through the self-condensation of the aldehyde formed after dehydrogenation of the hydrogen donor, or by forming through reaction of the aldehyde with levulinate esters. Significant differences are not observed in the rate of GVL formation when using different levulinate esters, e.g., BL and ethyl levulinate (EL) (Table 5-1, Entries 10 and

12). Additionally, 4-hydroxypentanoic acid (Table 5-1, 3) was not detected in any product mixtures throughout this study, indicating that rapid ring-closing (i.e., lactonization) to form

GVL occurs under the reaction conditions employed. To investigate whether CTH of levulinate esters could be occurring through direct intramolecular hydrogen transfer within the ester, a reaction mixture of isopropyl levulinate (IL) in tetrahydropyran (THP) was used (Table 5-1,

Entry 13). The formation of small amounts of GVL suggests that intramolecular CTH does appear to take place, but at a much lower rate (e.g., by a factor of approximately 20) than through the MPV reaction with a hydrogen donor.

5.3.2 CTH of levulinate esters in the presence of sec-butylphenol

A key advantage of the MPV reaction is that the high chemo-selectivity for hydrogenation of carbonyl groups allows the CTH of levulinate esters to GVL to occur in the presence of other highly functionalized molecules. For example, we have observed that the reduction of levulinate esters occurs effectively in the presence of an aromatic diluent such as sec-butyl phenol (Table 5-2, SBP), whilst leaving the aromatic functional group in SBP completely unconverted. This high selectivity for conversion of LA versus reduction of SBP has practical implications, because we have shown elsewhere that alkylphenol solvents can be used to extract levulinic acid from aqueous solutions of sulfuric acid used to achieve the deconstruction of cellulose26. Although the use of equimolar concentrations of hydrogen donor

131 (i.e., IPA) and ester result in relatively low reaction rates (Table 5-2, Entry 1), it was observed that modest increases in the amount of IPA were effective in raising the yield of GVL. For instance, an increase in molar ratio of IPA:EL from 1:1 to 4:1 (Table 5-2, Entries 1-2) increased the GVL yields from 18% to 73%. At a molar ratio of IPA:EL of 7:1 (Table 5-2, Entry 3), it was found that the rate of GVL formation was comparable to that when pure IPA is used as the solvent (Table 5-1, Entry 10), and close to quantitative yields of GVL were attained (Table 5-2,

Entry 4).

Table 5-2 Catalytic transfer hydrogenation of ethyl levulinate over ZrO2 in the presence of sec- butyl phenol with varying molar ratios of isopropanol to ester.a

GVL Molar ratio Conversion GVL formation rate Entry EL (wt%) SBP (wt%) Time (h) yield IPA:EL (%) (μmol g-1 min-1) (%) 1 9.6 86.4 1:1 16 28.0 18.1 2.2 2 8.6 77.1 4:1 16 77.3 72.6 -- 3 7.7 69.7 7:1 1 34.5 33.4 70.1 4 7.7 69.7 7:1 16 97.3 93.3 -- a Batch reactions, mass ratio ZrO2:EL = 1:2.4, 423 K, 300 psig He. Only by-product detected was isopropyl levulinate.

As previously discussed, it is notable that the high chemo-selectivity inherent in the MPV reaction, required in production processes for conversion of LA and its esters in the presence of highly functionalized extraction solvents, such as those recently reported by Alonso et al.26, is not easily attainable over conventional heterogeneous metal catalysts using molecular H2. In particular, the hydrogenation of C=C over C=O bonds is thermodynamically favoured, and typically requires the use of precious metals (e.g., Ru, Pt) modified with a base metal (e.g., Sn) at the expense of lower catalytic activity1, 26.

132 5.3.3 Effect of levulinic acid on catalyst activity

We have observed that LA undergoes CTH less readily than its esters, with GVL formation rates of 4 μmol g-1 min-1 (Table 5-3, Entry 3). In comparison, the rate of GVL formation from BL was approximately 70 μmol g-1 min-1. For all runs with LA (Table 5-3), the sole by-product detected was isobutyl levulinate, formed through the esterification of LA with the solvent over the catalyst. Entries 1-2 in Table 5-3 demonstrate that the addition of LA to BL

-1 and 2BuOH results in a decrease in the rate of GVL formation over ZrO2 from 42 to 15 μmol g min-1 as the concentration of LA is increased from 0.5 to 1 wt%.

We suggest that the observed inhibiting effect of LA on CTH is due to the strong binding

45, 46 of the acid functional group in LA to basic sites on ZrO2, which is known to be amphoteric ; the number of acidic and basic sites for the ZrO2 material used here has been previously reported as 296 and 212 μmol g-1, respectively47. Because basic sites have been proposed to be active sites for the MPV reaction, either by themselves or in a cooperative manner with acid sites leading to concerted direct hydrogen transfer through a six-membered cyclic transition state16, the blocking of basic sites by adsorbed LA would thus result in a decline in catalytic activity. We further found that addition of a base (i.e., MgO) to the reaction mixture was effective in mitigating catalyst inhibition by LA, and resulted in a four-fold increase in the GVL formation rate even with the use of lower amounts of ZrO2 (Table 5-3, Entry 5). Of the metal oxide catalysts examined, similar to what we observed for the CTH of levulinate esters to GVL, ZrO2 was found to be the most active material for the CTH of LA to GVL (Table 5-3, Entries 6-9). The observed high activity of ZrO2 for CTH of LA and its esters to GVL is consistent with the literature for the

MPV reaction over heterogeneous catalysts14, 26, which suggest that its strong amphoteric nature is one of the key factors underlying its exceptional activity.

133 Table 5-3 Catalytic transfer hydrogenation of levulinic acid and levulinate esters using 2-butanol as the solvent and hydrogen donor.a

Mass ratio GVL formation LA BL Time GVL yield Ester & LA Entry catalyst: Catalyst rate (wt%) (wt%) (h) (%) conv (%) LA (μmol g-1 min-1) ZrO 1 0.5 5 -- 2 4 30 30.1 42.1 ZrO 2 1 5 -- 2 4 13 26.0 14.9 ZrO 3 5 0 1:2 2 16 22 52.0 3.9 b ZrO 4 5 0 1:2 2 16 71 >99.9 -- c ZrO 5 5 0 1:5 2 16 39 >99.9 16.7 ZrO 6 1 0 2:1 2 16 92 >99.9 -- MgO/ZrO 7 1 0 2:1 2 16 54 >99.9 -- γ-Al O 8 1 0 2:1 2 3 16 56 79.9 2.0 9 1 0 2:1 CeZrOx 16 11 43.4 0.4 aBatch reactions, 423 K, 300 psig He. b493 K c MgO added to reaction mixture; mass ratio of MgO:ZrO2 = 1:1.

5.3.4 Effect of water on catalyst activity

The catalytic activity of ZrO2 for CTH was observed to decrease with an increase in water concentration. As shown in Figure 5-4, GVL formation and BL conversion rates declined steadily with increasing amounts of water. It is possible that competitive adsorption of water with the reactant for the active sites on the ZrO2 catalyst results in this decline in CTH activity, for instance, with water acting as a Lewis base. This effect of water on catalytic activity for hydrogenation by the MPV reaction was also observed by Corma et al., who found that zeolitic materials with a higher degree of hydrophobicity retained more activity in the presence of water, and that postsynthesis silylation of Sn-beta zeolite was found to result in increased water resistance.21

134

100 100 GVL formation rate BL conv rate 80 IL conv rate 80

BL conv (%) conversion BL

) -1

60 60

min -1

molg 40

 (

Rate 20

0 0 0 2 4 6 8 10 Water concentration (wt%)

Figure 5-4 Catalytic transfer hydrogenation of BL to GVL over ZrO2 in IPA at varying water concentrations. Reaction conditions: batch reactions, 5 wt% BL in IPA/water as feed, 4 h, mass ratio catalyst:BL = 1:5.

5.3.5 Regenerability of ZrO2

Having identified ZrO2 as a highly active catalyst for CTH of BL to GVL, we studied this

reaction using 2BuOH as the hydrogen donor in a continuous flow reactor to examine the

stability and regenerability of the catalyst. As shown in Figure 5-5, the ZrO2 catalyst deactivates

for the first 100 h of time-on-stream, and then appears to exhibit stable catalytic performance

thereafter. This stabilization in activity could be indicative of the presence of sites with different

activities, where initial rapid deactivation, possibly by coking, is proposed to first occur over

relatively active sites, followed by sustained catalysis on less active sites. Importantly, the initial

activity of the catalyst is fully regained after calcination in air at 723 K, and the catalyst was then

135 found to display the same deactivation profile as the fresh catalyst after each regeneration cycle.

After the second regeneration cycle and 500 h time-on-stream, it was observed that the catalyst stabilized to a similar activity level as the fresh catalyst (i.e., at 100 h). This ease of repeated catalyst regeneration with no loss in catalyst performance, and the stabilization in catalytic activity with time-on-stream, are significant advantages for the use of this catalyst system in industrial applications.

100 50

)

-1

min -1 40

GVL yield (%) yield GVL

molg

 (

10 30

20 GVL formation rate 1 10 0 100 200 300 400 500 Time (h)

Figure 5-5 Plot of γ-valerolactone (GVL) formation rate (■) and yield of GVL (○) as a function of time-on-stream for the catalytic transfer hydrogenation of butyl levulinate (BL) to GVL over ZrO2 in a continuous flow reaction system. Reaction conditions were 5 wt% BL in 2-butanol as feed, WHSV = 0.18 h-1, 423 K, 300 psi He. The catalyst was regenerated in-situ by calcination in 60 cc (STP)/min flowing air at 723 K for 4 h, at approximately 150 and 300 h, as indicated by the dotted lines.

136 5.4 Conclusions

In this chapter, it has been demonstrated that the reduction of LA and its esters to GVL can be accomplished by CTH through the MPV reaction over various metal oxide catalysts using secondary alcohols as the hydrogen donor. Moreover, close to quantitative yields of GVL can be achieved under appropriate reaction conditions (e.g., at 423 K). ZrO2 was demonstrated to a highly active material for CTH, in both batch and continuous flow reactor studies. The presence of LA or water was found to decrease the catalytic activity of ZrO2 for GVL formation. While the activity of this catalyst decreased and then stabilized during operation for 100 h of time-on- stream, the initial activity of the catalyst was repeatedly regenerable by calcination in air, with no observable loss in catalytic activity.

5.5 References

1. Mäki-Arvela, P., Hájek, J., Salmi, T. & Murzin, D. Y. Chemoselective hydrogenation of carbonyl compounds over heterogeneous catalysts. Applied Catalysis A: General 292, 1- 49 (2005). 2. Ide, M. S., Hao, B., Neurock, M. & Davis, R. J. Mechanistic Insights on the Hydrogenation of α,β-Unsaturated Ketones and Aldehydes to Unsaturated Alcohols over Metal Catalysts. ACS Catalysis 2, 671-683 (2012). 3. Ponec, V. On the role of promoters in hydrogenations on metals; α,β-unsaturated aldehydes and ketones. Applied Catalysis A: General 149, 27-48 (1997). 4. Vilella, I. M. J., Miguel, S. R. d., Lecea, C. n. S.-M. n. d., Linares-Solano, Ã. n. & Scelza, O. A. Catalytic performance in citral hydrogenation and characterization of PtSn catalysts supported on activated carbon felt and powder. Applied Catalysis A: General 281, 247- 258 (2005). 5. Zgolicz, P. D., Rodriguez, V. I., Vilella, I. M. J., de Miguel, S. R. & Scelza, O. A. Catalytic performance in selective hydrogenation of citral of bimetallic Pt-Sn catalysts supported on MgAl2O4 and -Al2O3. Applied Catalysis A: General 392, 208-217 (2011). 6. de Miguel, S. R., Roman-Martinez, M. C., Cazorla-Amoas, D., Jablonski, E. L. & Scelza, O. A. Effect of the support in Pt and PtSn catalysts used for selective hydrogenation of carvone. Catalysis Today 66, 289-295 (2001). 7. Ruiz-Martinez, J., Sepulveda-Escribano, A., Anderson, J. A. & Rodriguez-Reinoso, F. Influence of the preparation method on the catalytic behaviour of PtSn/TiO2 catalysts. Catalysis Today 123, 235-244 (2007).

137

8. Liberkova, K. & Touroude, R. Performance of Pt/SnO2 catalyst in the gas phase hydrogenation of crotonaldehyde. Journal of Molecular Catalysis A: Chemical 180, 221- 230 (2002). 9. Merlo, A. B., Vetere, V., Ruggera, J. F. & Casella, M. n. L. Bimetallic PtSn catalyst for the selective hydrogenation of furfural to furfuryl alcohol in liquid-phase. Catalysis Communications 10, 1665-1669 (2009). 10. Vetere, V., Merlo, A. B., Ruggera, J. F. & Casella, M. L. Transition Metal-based Bimetallic Catalysts for the Chemoselective Hydrogenation of Furfuraldehyde. Journal of the Brazilian Chemical Society 21, 914-920 (2010). 11. Merlo, A. B., Vetere, V., Ramallo-Lopez, J. M., Requejo, F. G. & Casella, M. L. Liquid- phase furfural hydrogenation employing silica-supported PtSn and PtGe catalysts prepared using surface organometallic chemistry on metals techniques. Reaction Kinetics Mechanisms and Catalysis 104, 467-482 (2011). 12. Margitfalvi, J. L., Vanko, G., Borbath, I., Tompos, A. & Vertes, A. Characterization of Sn-Pt/SiO2 Catalysts Used in Selective Hydrogenation of Crotonaldehyde by Mossbauer Spectroscopy. Journal of Catalysis 190, 474-477 (2000). 13. Santori, G. F., Casella, M. n. L. & Ferretti, O. A. Hydrogenation of carbonyl compounds using tin-modified platinum-based catalysts prepared via surface organometallic chemistry on metals (SOMC/M). Journal of Molecular Catalysis A: Chemical 186, 223- 239 (2002). 14. Torres, G. C., Ledesma, S. D., Jablonski, E. L., de Miguel, S. R. & Scelza, O. A. Hydrogenation of carvone on Pt-Sn/Al2O3 catalysts. Catalysis Today 48, 65-72 (1999). 15. Degraauw, C. F., Peters, J. A., Vanbekkum, H. & Huskens, J. Meerwein-Ponndorf- Verley Reductions and Oppenauer Oxidations - an Integrated Approach. Synthesis- Stuttgart, 1007-1017 (1994). 16. Cohen, R., Graves, C. R., Nguyen, S. T., Martin, J. M. L. & Ratner, M. A. The Mechanism of Aluminum-Catalyzed Meerwein-Schmidt-Ponndorf-Verley Reduction of Carbonyls to Alcohols. Journal of the American Chemical Society 126, 14796-14803 (2004). 17. Chuah, G. K., Jaenicke, S., Zhu, Y. Z. & Liu, S. H. Meerwein-Ponndorf-Verley Reduction over Heterogeneous Catalysts. Current Organic Chemistry 10, 1639-1654 (2006). 18. Liu, S. H., Jaenicke, S. & Chuah, G. K. Hydrous Zirconia as a Selective Catalyst for the Meerwein-Ponndorf-Verley Reduction of Cinnamaldehyde. Journal of Catalysis 206, 321-330 (2002). 19. Urbano, F. J., Aramendía, M. A., Marinas, A. & Marinas, J. M. An insight into the Meerwein-Ponndorf-Verley reduction of [alpha],[beta]-unsaturated carbonyl compounds: Tuning the acid-base properties of modified zirconia catalysts. Journal of Catalysis 268, 79-88 (2009). 20. Braun, F. & Di Cosimo, J. I. Catalytic and spectroscopic study of the allylic alcohol synthesis by gas-phase hydrogen transfer reduction of unsaturated ketones on acid-base catalysts. Catalysis Today 116, 206-215 (2006). 21. Corma, A., Domine, M. E., Nemeth, L. & Valencia, S. Al-Free Sn-Beta Zeolite as a Catalyst for the Selective Reduction of Carbonyl Compounds (Meerwein-Ponndorf- Verley Reaction). Journal of the American Chemical Society 124, 3194-3195 (2002).

138 22. Tominaga, K., Mori, A., Fukushima, Y., Shimada, S. & Sato, K. Mixed-acid systems for the catalytic synthesis of methyl levulinate from cellulose. Green Chemistry 13, 810-812 (2011). 23. Lange, J.-P., van de Graaf, W. D. & Haan, R. J. Conversion of Furfuryl Alcohol into Ethyl Levulinate using Solid Acid Catalysts. ChemSusChem 2, 437-441 (2009). 24. Gürbüz, E. I., Alonso, D. M., Bond, J. Q. & Dumesic, J. A. Reactive Extraction of Levulinate Esters and Conversion to γ-Valerolactone for Production of Liquid Fuels. ChemSusChem 4, 357-361 (2011). 25. Fitzpatrick, S. W. The Biofine Technology: A "Bio-refinery" Concept Based on Thermochemical Conversion of Cellulosic Biomass (eds. Bozell, J. J. & Patel, M. K.) (ACS, Washington, DC, 2005). 26. Alonso, D. M., Wettstein, S. G., Bond, J. Q., Root, T. W. & Dumesic, J. A. Production of Biofuels from Cellulose and Corn Stover Using Alkylphenol Solvents. ChemSusChem, DOI: 10.1002/cssc.201100256 (2011). 27. Horvath, I. T., Mehdi, H., Fabos, V., Boda, L. & Mika, L. T. -Valerolactone-a sustainable liquid for energy and carbon-based chemicals. Green Chemistry 10, 238-242 (2008). 28. Manzer, L. E. Catalytic synthesis of -methylene--valerolactone: a biomass-derived acrylic monomer. Applied Catalysis A: General 272, 249-256 (2004). 29. Bond, J. Q., Alonso, D. M., Wang, D., West, R. M. & Dumesic, J. A. Integrated Catalytic Conversion of -Valerolactone to Liquid Alkenes for Transportation Fuels. Science 327, 1110-1114 (2010). 30. Upare, P. P. et al. Direct Hydrocyclization of Biomass-Derived Levulinic Acid to 2- Methyltetrahydrofuran over Nanocomposite Copper/Silica Catalysts. ChemSusChem 4, 1749-1752 (2011). 31. Hayashi, I. Catalytic hydrogenation of levulinic acid esters. Kogyo Kagaku Zasshi 60, 280-2 (1957). 32. Hayashi, I., Negoro, E. & Hachihama, Y. Levulinic acid and its derivatives. II. Preparation and reduction of γ-valerolactone by catalytic hydrogenation. Nippon Kagaku Kaishi (1921-47) Ind. Chem. Sect. 57, 67-9 (1954). 33. Braden, D. J., Henao, C. A., Heltzel, J., Maravelias, C. C. & Dumesic, J. A. Production of liquid hydrocarbon fuels by catalytic conversion of biomass-derived levulinic acid. Green Chemistry (2011). 34. Deng, L., Li, J., Lai, D.-M., Fu, Y. & Guo, Q.-X. Catalytic Conversion of Biomass- Derived Carbohydrates into γ-Valerolactone without Using an External H2 Supply. Angewandte Chemie International Edition 48, 6529-6532 (2009). 35. Deng, L. et al. Conversion of Levulinic Acid and Formic Acid into γ-Valerolactone over Heterogeneous Catalysts. ChemSusChem 3, 1172-1175 (2010). 36. Kopetzki, D. & Antonietti, M. Transfer hydrogenation of levulinic acid under hydrothermal conditions catalyzed by sulfate as a temperature-switchable base. Green Chemistry 12, 656-660 (2010). 37. Shen, Z. et al. Hydrogen-Transfer Reduction of Ketones into Corresponding Alcohols Using Formic Acid as a Hydrogen Donor without a Metal Catalyst in High-Temperature Water. Industrial & Engineering Chemistry Research 49, 6255-6259 (2010).

139 38. Lemcoff, N. O. Liquid phase catalytic hydrogenation of acetone. Journal of Catalysis 46, 356-364 (1977). 39. Fouilloux, P. The nature of raney nickel, its adsorbed hydrogen and its catalytic activity for hydrogenation reactions (review). Applied Catalysis 8, 1-42 (1983). 40. Chang, N.-S., Aldrett, S., Holtzapple, M. T. & Davison, R. R. Kinetic studies of ketone hydrogenation over Raney nickel catalyst. Chemical Engineering Science 55, 5721-5732 (2000). 41. Rao, R. S., Walters, A. B. & Vannice, M. A. Influence of Crystallite Size on Acetone Hydrogenation over Copper Catalysts The Journal of Physical Chemistry B 109, 2086- 2092 (2004). 42. Yurieva, T. M. Mechanisms for activation of hydrogen and hydrogenation of acetone to isopropanol and of carbon oxides to methanol over copper-containing oxide catalysts. Catalysis Today 51, 457-467 (1999). 43. Wise, N. J. & Williams, J. M. J. Oxidation of alcohols by transfer hydrogenation: driving the equilibrium with an intramolecular trap. Tetrahedron Letters 48, 3639-3641 (2007). 44. Bui, L., Luo, H., Gunther, W. R. & Román-Leshkov, Y. Domino Reaction Catalyzed by Zeolites with Brønsted and Lewis Acid Sites for the Production of γ-Valerolactone from Furfural. Angewandte Chemie International Edition (2013). 45. Tanabe, K. & Yamaguchi, T. Acid-base bifunctional catalysis by ZrO2 and its mixed oxides. Catalysis Today 20, 185-197 (1994). 46. Yamaguchi, T. Application of ZrO2 as a catalyst and a catalyst support. Catalysis Today 20, 199-217 (1994). 47. Gürbüz, E. I., Kunkes, E. L. & Dumesic, J. A. Integration of C-C coupling reactions of biomass-derived oxygenates to fuel-grade compounds. Applied Catalysis B: Environmental 94, 134-141 (2010).

140

Chapter 6 Triacetic acid lactone as a biorenewable platform chemical

6.1 Introduction

6.1.1 Combination of biological and chemical catalysis – platform chemical

approach

As discussed in Section 1.4.3, a promising approach for the production of biorenewable chemicals is to convert biomass-derived carbohydrates to platform molecules, which in turn serve as building blocks for commercially-valuable end products. To provide guidance towards the selection of target biorenewable chemicals, the U.S. Department of Energy (DOE) released the DOE “Top 10” report1 in 2004 by Werpy and Petersen, which outlined a list of approximately twelve potential species deemed as the most promising platform molecules for a bio-refinery. More recently in 2009, an update to this report was provided by Bozell et al.2, which showed that research in the production of many of these standard platform molecules has been relatively extensive.

Nikolau et al. have previously suggested that an expanded array of biologically-derived platform chemicals offers unique opportunities for the conversion of renewable carbohydrates to families of commodity chemicals for use by society.3 These alternative bio-based platform chemicals are envisioned to be products of a single metabolic pathway (e.g., polyketide biosynthesis), with flexibility for the generation of a series of homologous molecules. These platform molecules can then undergo chemical catalytic upgrading, providing access to a diverse stream of end products that serve as functional and/or direct replacements of currently used petrochemicals. In this manner, combination of the inherently high chemo- and stereo-

141 selectivities of biosynthetic pathways with the superior efficiencies of chemical catalytic strategies can be fully exploited in an integrated bio-refinery.

Strategies for the effective exploitation of synergies between biological and chemical catalysis are currently uncommon in biomass processing. A major technological hurdle lies in the high cost of downstream processing (i.e., product separation and purification) of biologically- synthesized products due to the relatively low product titers of cultures and high heat of vaporization of water. Of the various reports of successful combinations of biological and chemical catalysis for the production of biorenewable chemicals, the catalytic dehydration of bio-ethanol and lactic acid to ethylene and polylactic acid, respectively, are most well-known.

Other biologically-produced chemicals that are have been studied as potential platform chemicals are largely organic acids, such as succinic acid, 3-hydroxypropionic acid, itaconic acid, and glutamic acid4.

It is notable that many studies typically examine the biological or chemical synthesis processes in isolation, and that actual process integration (e.g., catalytic upgrading of fermentation products) has been relatively rare. Process integration has most commonly been reported in cases where product isolation is easily accomplished in the laboratory, for instance, through extraction5, 6. Lennen et al. reported the production of saturated alkanes through the

7 catalytic decarboxylation of extracted C12 and C14 fatty acids over Pd/C . The fatty acids were synthesized by metabolically engineered E. coli, and a concentration of 0.44 gL-1 (culture volume) undecane was obtained after catalytic grading. More recently, Toste and coworkers described the conversion of acetone-n-butanol-ethanol (ABE) fermentation products into C7-C15 ketones through alkylation over Pd catalysts8. Catalytic upgrading was performed on distilled

142 products from the extractant phase, and ~38% of the carbon from the glucose feed was recovered

as ketones which can undergo hydrotreating to form alkanes (i.e., components for petrol, diesel,

and jet-fuel).

6.1.2 Triacetic acid lactone

In this chapter, we demonstrate that 2-pyrones, such as 4- hydroxy-6-methyl-2-pyrone

(also denoted as triacetic acid lactone, 1), are biomass-derived compounds that may serve as

potential chemical building blocks for biorenewable chemicals9. In particular, 1 can be obtained

from natural sources (plant and microbial), or synthetically produced from acetic acid10-12.

Significantly, 1 may be biologically synthesized from glucose using genetically modified

Escherichia coli and Saccharomyces cerevisiae through heterologous expression of 2-pyrone

synthase13 or other genetically modified polyketide synthases14, 15; enzyme engineering may

therefore be employed for the generation of an array of 2-pyrones with varying functionalities

derived from different starter molecules through a single polyketide biosynthetic pathway16.

The catalytic upgrading of 1 is presented in this chapter, with the purpose of

demonstrating that 1 can be converted to a variety of commercially valuable chemical

intermediates and end products (Figure 6-1). We describe methods for the conversion of 1 to

these target molecules, employing heterogeneous catalysts or through thermal processes, using

relatively mild reaction conditions and environmentally-friendly and biorenewable solvents. The

literature on the catalytic transformations of 1 has largely dealt with the enantioselective

synthesis of chemical intermediates, with possible applications as precursors for bioactive

compounds17-20. Some work in the derivation of aromatics from 1 has also been reported in the

literature21, 22. In contrast, our work focuses on the development of chemical derivatization

143 methods for the conversion of 1 to products that may serve as direct replacements of existing petrochemicals, allowing for integration into current industrial chemical markets. Notably, all these products possess two or more functional groups, and this retention of functionality has unique strategic advantages, such as in ensuring the usefulness of these molecules in downstream chemistry and requiring only minimal usage of external hydrogen, both key factors in the assessment of the economic viability of production processes for biorenewable chemicals.

Figure 6-1 Reactions discussed in this chapter. Compounds are as follows: 4-hydroxy-6-methyl- 2-pyrone/ triacetic acid lactone (1); 2,4-pentanedione/ acetylacetone (2); 5,6-dihydro-4-hydroxy- 6-methyl-2H-pyran-2-one (3); 3-penten-2-one (4); 4-hydroxy-2-pentanone (5); 4-hydroxy-6- methyltetrahydro-2-pyrone (6); 6-methyl-5,6-dihydro-2-pyrone/ parasorbic acid (7); 2,4- hexadienoic acid/ sorbic acid (8); 1,3-pentadiene (9); δ-hexalactone (10); hexenoic acid (11); γ- caprolactone (12).

6.2 Chemical diversification of triacetic acid lactone

We have found that 1 undergoes ring-opening and decarboxylation in the absence of catalyst when water is the solvent, giving 2,4-pentanedione (2) also known as acetylacetone, and

CO2 as the primary products. Compound 2 is a commercially important commodity chemical used as a fuel additive, in extraction processes for metals, in metal plating, in resin modification,

144 and as an intermediate in the synthesis of dyes. At present, 2 is produced in a multi-step process from acetone and ketene23. Table 6-1 shows that close to quantitative yields of 2 can be attained when water is used as the reaction solvent, while almost no conversion of 1 occurs when an aprotic solvent (i.e., tetrahydrofuran (THF)) is employed. Furthermore, the presence of an acid catalyst (i.e., Amberlyst 70) did not promote the reaction in THF. The low reactivity of 1 in THF can be related to the low concentration of water in THF, because stoichiometric amounts of water are needed for the reaction to proceed. Furthermore, we have found that the presence of water adsorbed on Amberlyst 70 (up to 15% w/w) can result in significant conversion of 1 to 2; therefore, the catalyst must be thoroughly dried at 373 K prior to use in these experiments.

Table 6-1 Ring-opening and decarboxylation of 1 to 2.a

Solvent Conversion of 1 (%) Selectivity to 2 (%) Water 94.3 95.6 THF 5.5 0.0 THFb 3.3 >99 aBatch reactions. Reaction conditions: 4 h, 373 K, 22 bar He, 1.8 wt% 1 in THF or 0.8 wt% 1 in water as feed. bDry Amberlyst 70 as catalyst, mass ratio 1:catalyst = 2:1.

Compound 1 can be hydrogenated over metal catalysts to yield either 5,6-dihydro-4- hydroxy-6-methyl-2H-pyran-2-one (3) or 4-hydroxy-6-methyltetrahydro-2-pyrone (6), through the hydrogenation of one or both C=C bonds in 1, respectively. Due to the susceptibility of 1 to undergo ring-opening in the presence of water, all hydrogenation reactions were performed in organic solvents. It was found that the maximum yield of the half-hydrogenated pyrone, 3, could be attained over 5 wt% Pd/Nb2O5. Specifically, at a reaction temperature of 343 K, batch time of

5 h, 8 bar H2, and using 1-butanol (BuOH) as the solvent, 92% yield of 3 was attained at complete conversion of 1, with δ-hexalactone (10) being the only other identifiable product

145 observed (~1% yield). It is likely that 10 forms through subsequent dehydration of 6 to 6-methyl-

5,6-dihydro-2-pyrone or parasorbic acid (7), and hydrogenation of 7 over the bifunctional

Pd/Nb2O5 catalyst. We found that the hydrogenation of 1 over 10 wt% Pd/C under the same reaction conditions resulted in 63% and 37% yield of 3 and 6, respectively. Therefore, an acidic support appears to be necessary for the attainment of high yields of 3, although its role remains unclear at present. We further note that the need for an acidic support is consistent with the

19 literature in which similar catalysts (e.g., Pd/Al2O3 and Pd/TiO2 ) have been shown to be suitable catalysts for this reaction.

Ring-opening and decarboxylation of 3 to 3-penten-2-one (4) and 4-hydroxy-2-pentanone

(5) were observed to occur in water and THF (Table 6-2). The formation of 4 and 5 from 3 has not been previously reported in the literature. 4 and 5, being bifunctional ketones, are valuable chemical intermediates. Similar to what was observed for 1, the ring-opening and decarboxylation of 3 are achievable in water, while the use of THF as the reaction solvent results in lower conversions of 1. However, it was found that the conversion of 3 in water was substantially higher in the presence of an acid catalyst, such as Amberlyst 70, with a significant amount of 5 being formed through hydration of the C=C bond in 4. Also, the conversion of 3 in

THF over an acid catalyst for was similar to that for reaction in water without a catalyst. The maximum yields of 4 were observed to be approximately 58% when an acid catalyst is used in combination with THF as the solvent. Using a continuous flow reaction system at similar reaction conditions, we determined through sampling of gas phase products using an in-line gas chromatograph that the molar ratio of CO2 evolved to converted 3 was 1:1, indicating that the loss of carbon (i.e., less than quantitative selectivity to 4) is likely due to the further reaction of 4 to by-products unidentifiable by GC-MS.

146 Table 6-2 Ring-opening and decarboxylation of 3 to 4.a

Solvent T (K) Time (h) Conversion of 3 (%) Selectivity to 4 (%) Water 343 1 67.8 49.9 Waterb 343 1 >99 15.7c THF 373 4 3.2 >99 THFb 373 4 >99 57.7 aBatch reactions. Reaction conditions: 1.5 wt% 3 as feed, 22 bar He, no catalyst. bAmberlyst 70 as catalyst, mass ratio 3:catalyst = 2:1. c5 was the major product at 77.2% selectivity.

In the absence of an acidic support, 1 can be fully hydrogenated over 10 wt% Pd/C to give close to quantitative yields of 6 (Table 6-3). The hydrogenation reactions appear to proceed more slowly in THF compared to that in BuOH, with lower conversions of 1 observed in THF than for BuOH under the same reaction conditions. As observed previously over Pd/Nb2O5, 10 was the only by-product detected. In particular, 10 was formed at low levels even at complete conversion of 1 and 3 in both BuOH and THF.

Table 6-3 Hydrogenation of 1 to 6 over 10 wt% Pd/C.a

Mass ratio Time Selectivity to 3 Selectivity to 6 Selectivity to 10 Solvent Conversion of 1 (%) catalyst: 1 (h) (%) (%) (%) BuOH 1:14 0.5 46.0 96.1 3.9 0.0 BuOH 1:14 4 >99 47.5 48.6 3.9 BuOH 1:14 12 >99 0.0 94.0 6.0 THF 1:14 0.5 37.1 98.4 1.6 0.0 THF 1:14 4 >99 71.6 37.4 1.0 THF 1:14 12 >99 24.7 72.4 2.9 THF 1:2 2 >99 0.0 96.4 3.9 a Batch reactions. Reaction conditions: 2 wt% 1 as feed, 323 K, 35 bar H2, 10 wt% Pd/C as catalyst.

The dehydration of 6 to 7 occurs over solid acid catalysts, such as Amberlyst 70.

Quantitative yields of 7 could be achieved in a batch reactor after of 12 h over Amberlyst 70 and using 2 wt% 6 in THF as feed, at 373 K, 22 bar He. Lower holding times of 4 and 8 h resulted in

147 6 conversions of 60% and 80%, respectively, with quantitative selectivity to 7 observed in both cases. To examine the recyclability of Amberlyst 70 for this reaction, catalyst recycling studies were performed. The rate of dehydration of 6 decreased from 66 μmolg-1min-1 to 37 μmolg-1min-

1 over three recycles (Table 6-4), indicating that moderate catalyst deactivation occurred, possibly by coking.

Table 6-4 Dehydration of 6 to 7 over Amberlyst 70.a

Specific rate of consumption of 6 Recycle No. Conversion of 6 (%)b (μmol g-1 min-1) b 0 27±2 66±4 1 24±1 55±3 2 20±1 41±1 3 18±3 37±3 aBatch reactions. Reaction conditions: 1 h, 373 K, 22 bar He, 2 wt% 6 in THF, mass ratio 6:catalyst = 2:1. Spent catalyst was filtered and dried at 100 °C before being reused for the next reaction cycle. Quantitative selectivity to 7 in all cases. bConversion and rate values averaged over two runs.

Although 7 is not an end-product in itself, one approach we propose is the hydrogenation of 7 over a metal catalyst to 10, and subsequent ring-opening of 10 to form the unsaturated acid, hexenoic acid (11). 11 can in turn undergo ring-closing to form γ-caprolactone (12). The applications of 11 and 12 are diverse, thus making them valuable end products. For instance, some isomers of 11 are currently used as flavouring agents in food18, and the presence of both the acid and C=C bond functional groups makes 11 a useful chemical intermediate; 12 is commercially important as a flavouring agent in food, tobacco, fragrances, and cosmetics, and as a potential intermediate for insecticides19-21. We have carried out ring-opening of 10 over a solid acid catalyst (i.e., Amberlyst 70), and obtained 11 and 12 at maximum selectivities of 52% and

68%, at reaction temperatures of 423 and 443 K, respectively (Table 6-5). A mixture of isomers

148 of 11 was obtained, consisting mostly of 4-hexenoic acid with small amounts of 3-hexenoic acid.

We further observed that the selectivity to HA decreases with lower reaction temperatures, while the selectivity for 12 increases. Based on gas-phase thermodynamic calculation estimates using

Gaussian software (Figure 6-2), we predict that the ring-opening of 10 to HA is endothermic and entropically favored, while that for the formation of 12 from 11 is exothermic. Based on these estimates, it is expected that the ratio of 11 to 10 would increase with reaction temperature, which is consistent with the experimental data.

Table 6-5 Ring-opening of 10 to 11 and 12 over Amberlyst 70.a

T (K) Time (h) Conversion of 10 (%) Selectivity to 11 (%) Selectivity to 12 (%) 413 4 39.8 56.1 18.3 423 4 62.3 51.8 22.4 423 8 68.7 58.2 23.3 433 8 91.8 40.0 50.1 443 8 95.9 24.5 58.5 443 12 96.6 21.6 68.2 aBatch reactions. Reaction conditions: 2 wt% 10 in THF as feed, 22 bar He, Amberlyst 70 as catalyst, mass ratio of 10:catalyst = 2:1.

Figure 6-2 Ring-opening of 10 to 11 and 12. ΔG° and ΔH° (in parentheses) for each reaction in kJ/mol is shown at standard conditions.

149 Another approach that we explored was the ring-opening of 7 over a solid acid catalyst, with the purpose of retaining the C=C bond in the carbon backbone and thus obtaining a dienoic acid as the end product. Since both the dehydration of 6 and ring-opening of 7 are acid-catalyzed, these reactions were performed sequentially in a batch reactor over Amberlyst 70. Because the dehydration of 6 to 7 takes place at a lower temperatures than the ring-opening of 7 to 8, the initial dehydration step at T1 (373 K) was followed by an increase in temperature to T2 to initiate ring-opening. Specifically, the reaction temperature was held at T1 for a duration of t1, and then held at T2 for t2. As shown in Table 6-6, it was found that a lower ring-opening reaction temperature (T2) resulted in higher selectivities to the sorbic acid (8), with respect to the amount of 7 converted. GC-MS analysis of the product mixture indicated that in addition to E,E-8 being formed as the major product, small amounts of other 8 isomers (1-5% yield with respect to 6) were also present. The maximum molar yield of 8 formed with respect to 1 was 64%.

Table 6-6 Dehydration of 6 and ring-opening of 7 over Amberlyst 70.a

Yield of 7wrt 6 Yield of 8 wrt 6 Yield of 9 b wrt 6 C balance Yield of 8 wrt 1 t (h) t (h) T (K) 1 2 2 (%) (%) (%) (%) (%) 12 12 443 19.5 65.8 0.9 86.2 63.2 0 12 443 7.8 66.7 1.1 75.6 64.1 12 18 443 8.9 66.1 1.5 76.5 63.4 12 12 453 3.3 65.4 1.7 70.4 62.8 0 4 473 3.5 59.2 1.9 64.6 56.8 aBatch reactions. Reaction conditions: 22 bar He, Amberlyst 70 as catalyst, mass ratio of 6:catalyst = 2:1, THF as solvent; the reaction temperature profile consisted of two steps: t1 at T1, and t2 at T2. T1 = 373 K in all experiments. Complete conversion of 6 in all runs. b Estimated based on amount of CO2 detected in gas phase.

Of the several products we have obtained from 1, the dienoic acid, 8, appears to be a particularly attractive end product in itself. 8 and its potassium salt are currently widely used as

150 preservatives for food and feed due to their ability to act as effective inhibitors of a wide spectrum of yeasts, mould, and bacteria while exhibiting low toxicity to mammals22. Currently, 8 is industrially produced from non-renewable fossil fuel derivatives, namely, crotonaldehyde and ketene, in a multi-step process22, 23. Besides its use as a preservative, 8 has potential for use as a monomer, and 8 has been demonstrated to form copolymers with acrylonitrile, butadiene, isoprene, acrylates, pentadiene, styrene, and polyethylene22, 24-26.

Analyses of the gas phase products after the reaction of 7 over Amberlyst 70 showed that

CO2 was formed, thereby indicating the occurrence of decarboxylation. GC-MS analyses of the liquid phase confirmed the presence of 1,3-pentadiene (9) as a minor product. Based on estimates of the amount of CO2 evolved, we calculate that the yield of 9 from 6 was 1-2% for the runs shown in Table 5. 9 is currently commercially obtained from the C5 fraction of naphtha cracking, and is a valuable chemical intermediate for making resins, adhesives and plastics24.

The yields of 9 observed here are relatively low due to the use of conditions optimized to obtain

8, and we expect that further work in the decarboxylation of 8 over appropriate catalysts should be effective in increasing 9 yields from 1.

6.3 Conclusions

In this chapter, we have demonstrated that a wide range of commercially valuable chemical intermediates and end products may be obtained from 1 over heterogeneous catalysts or through thermal decomposition in appropriate solvents. Accordingly, this work provides guidance in the use of 1, and more generally, 2-pyrones, as bio-based platform chemicals. In the next chapter, a detailed study of the mechanisms underlying the ring-opening and decarboxylation reactions of 1,

3, and 6 will be described.

151 6.4 Chemical notations

IUPAC/ common name Numerical notation 4-hydroxy-6-methyl-2-pyrone/ triacetic acid lactone 1 2,4-pentanedione/ acetylacetone 2 5,6-dihydro-4-hydroxy-6-methyl-2H-pyran-2-one 3 3-penten-2-one 4 4-hydroxy-2-pentanone 5 4-hydroxy-6-methyltetrahydro-2-pyrone 6 6-methyl-5,6-dihydro-2-pyrone/ parasorbic acid 7 2,4-hexadienoic acid/ sorbic acid 8 1,3-pentadiene 9 3,6-dihydro-4,6,6-trimethyl-2H-pyran-2-one 10 hexenoic acid 11 γ-caprolactone 12

6.5 Analytical methods and NMR data

Identification of all intermediates and products in the liquid phase, except 6, was performed using a gas chromatograph-mass spectrometer (Shimadzu Corp., GCMS-QP2010S) equipped with a SHRXI-5MS capillary column (30 m × 0.25 mm × 0.25 μm). MS were verified against purchased compounds (Sigma Aldrich) and the NIST MS library and confirmed to at least 95% similarity. For 7, the MS was compared against the literature24, and the fragmentation pattern was confirmed against all significant mass ions, such as the high mass ion peaks, m/z 112, m/z 97, m/z 68, and lower mass peaks centred around m/z 40. Identification of 6 was performed using NMR and the spectra compared to literature values25.

Identification of all products were verified by NMR. NMR spectra of all reaction products were obtained using a Varian Mercury 300 (1H NMR, 300 MHz), and referenced internally tetramethyl silane (TMS, Sigma-Aldrich, 99.9+%, NMR Grade), CDCl3 (Aldrich,

100%, ≥99.96 atom% D), or residual solvent impurities (shifts based on literature values26).

152

10% D2O (Aldrich) was added directly to the final reaction mixture for compounds 2, 4, and 5 to provide solvent lock, and Varian’s PRESAT solvent presaturation program was used to suppress the H2O resonance.

Compound 6 was synthesized from the hydrogenation of 1 over 10wt% Pd/C using ethanol as the solvent. 5 g of 1 was dissolved in 185 g ethanol and hydrogenated over 0.5 g 10wt%

Pd/C at 34 bar H2 for 20 h. 6 was isolated using a rotary evaporator, with the sample heated in a water bath controlled at 333 K. 4.9 g of 6 was isolated, corresponding to 95% molar yield with respect to 1, consistent with experimental results presented in Table 3. The isolated sample was dissolved in CDCl3 prior to analysis by NMR.

Compounds 3, 7, 8, and 12 were isolated using a Water HPLC equipped with a reversed- phase Agilent Zorbax SB-C18 (4.6 x 300mm, 5µm) using Milli-Q water as the aqueous phase with methanol as the organic modifier. Isolated compounds were collected using a Waters

Fraction Collector connected to the column outlet. Samples were extracted with an equal volume of dichloromethane. The organic phase was separated and dried over sodium sulfate prior to rotary evaporation. The residue was then dissolved CDCl3 prior to analysis by NMR.

Compound 11 was isolated by HPLC and fractionation as described above (i.e., for compounds 3, 7, 8, and 12) but the residue obtained after rotary evaporation was dissolved in

1 D2O, and spectra were taken using a Varian INOVA 500 ( H NMR, 500 MHz). Multiple solvent suppression (WET1D) was used to suppress resonance of both the H2O and methanol residue left from the sample isolation.

The results from NMR analysis are as follows:

153

2 δH(300 MHz; D2O; i-PrOH) 2.3 (6 H, s, 1-H, 5-H), 3.9 (3 H, s, 3-H). 3 δH(300 MHz; CDCl3;

Me4Si) 1.5 (2 H, d, J = 6.3 Hz, Me), 2.8-2.4 (2 H, m, 5-H), 3.7-3.3 (2 H, m, 3-H), 4.8 (1 H, dqd, J

= 11.3, 6.3, 2.8 Hz, 6-H). 4 δH(300 MHz; D2O; i-PrOH) 1.9 (3 H, dd, J = 6.8, 1.7 Hz, 5-H), 2.3 (3

H, s, 1-H), 6.2 (1 H, dq, J = 15.9, 1.7 Hz, 4-H), 7.1 (1 H, dq, J = 15.9, 6.8 Hz, 3-H). 5 δH(300

MHz; D2O; i-PrOH) 1.2 (3 H, d, J = 6.3 Hz, 5-H), 2.2 (3 H, s, 1-H), 2.7 (2 H, d, J = 6.3 Hz, 3-H),

4.3 (1 H, h, J = 6.3 Hz, 4-H). 6 δH(300 MHz; CDCl3; Me4Si) 1.4 (3 H, d, J = 6.3 Hz, Me), 1.6-1.5

(1 H, m, 5-HH), 2.3 (dddd, J = 13.7, 5.5, 3.0, 1.4 Hz, 5-HH), 2.5 (1 H, dd, J = 17.1, 7.8 Hz, 3-

HH), 2.9 (1 H, ddd, J = 17.1, 5.9, 1.4 Hz, 3-HH), 4.3-4.2 (1 H, m, 4-H), 4.4-4.3 (1 H, m, 6-H). 7

δH(300 MHz; CDCl3; CH2Cl2) 1.5 (3 H, d, J = 6.3 Hz, Me), 2.3-2.4 (2 H, m, 5-H), 4.7-4.5 (1 H, m,

6-H), 6.0-5.9 (1 H, m, 4-H), 6.9-6.8 (1 H, m, 3-H). 8 δH(300 MHz; D2O; MeOH) 1.8 (3 H, d, J =

4.9 Hz, Me), 6.1-5.8 (1 H, m, 5-H), 6.4-6.2 (2 H, m, 3-H & 4-H), 7.3-7. (1 H, m, 2-H). 11

δH(500 MHz; D2O; MeOH) 1.6 (3 H, d, J = 6.0 Hz, Me), 2.2 (2 H, q, J = 7.1 Hz, 3-H), 2.3 (2 H, t,

J = 7.1 Hz, 2-H), 5.6-5.4 (2 H, m, 4-H & 5-H). 12 δH(300 MHz; CDCl3) 1.0 (3 H, t, J = 7.4 Hz,

Me), 2.6-1.5 (6 H, m, 3 x CH2), 4.5-4.3 (1 H, m, 5-H).

6.6 Computational methods

Gaussian 0327 software was used for the calculation of gas-phase thermodynamic properties of δ-hexalactone, hexenoic acid, and γ-caprolactone. Geometry optimizations and frequency calculations were performed using B3LYP/6-31+G. Frequency calculations were used to provide estimates of standard changes in enthalpy, entropy, and Gibbs free energy.

154 6.7 References

1. Werpy, T. & Petersen, G. Top Value Added Chemicals from Biomass. Volume I - Results of Screening for Potential Candidates from Sugars and Synthesis Gas. U. S. D. o. Energy (2004). 2. Bozell, J. J. & Petersen, G. R. Technology development for the production of biobased products from biorefinery carbohydrates-the US Department of Energy's "Top 10" revisited. Green Chemistry 12, 539-554 (2009). 3. Nikolau, B. J., Perera, M. A. D. N., Brachova, L. & Shanks, B. Platform biochemicals for a biorenewable chemical industry. The Plant Journal 54, 536-545 (2008). 4. Corma, A., Iborra, S. & Velty, A. Chemical Routes for the Transformation of Biomass into Chemicals. Chemical Reviews 107, 2411-2502 (2007). 5. Brennan, T. C. R., Turner, C. D., Krömer, J. O. & Nielsen, L. K. Alleviating monoterpene toxicity using a two-phase extractive fermentation for the bioproduction of jet fuel mixtures in Saccharomyces cerevisiae. Biotechnology and Bioengineering 109, 2513-2522 (2012). 6. Stark, D. & Stockar, U. in Process Integration in Biochemical Engineering (eds. Stockar, U. et al.) 149-175 (Springer Berlin Heidelberg, 2003). 7. Lennen, R. M., Braden, D. J., West, R. M., Dumesic, J. A. & Pfleger, B. F. A process for microbial hydrocarbon synthesis: Overproduction of fatty acids in Escherichia coli and catalytic conversion to alkanes. Biotechnology and Bioengineering 106, 193-202 (2010). 8. Anbarasan, P. et al. Integration of chemical catalysis with extractive fermentation to produce fuels. Nature 491, 235-239 (2012). 9. Chia, M., Schwartz, T. J., Shanks, B. H. & Dumesic, J. A. Triacetic acid lactone as a potential biorenewable platform chemical. Green Chemistry (2012). 10. Collie, N. J. The lactone of triacetic acid. Journal of the Chemical Society 59, 607-617 (1891). 11. Goel, A. & Ram, V. J. Natural and synthetic 2H-pyran-2-ones and their versatility in organic synthesis. Tetrahedron 65, 7865-7913 (2009). 12. Taeschler, C. Ketenes, Ketene Dimers, and Related Substances (2010). 13. Eckermann, S. et al. New pathway to polyketides in plants. Nature 396, 387-390 (1998). 14. Xie, D. M. et al. Microbial synthesis of triacetic acid lactone. Biotechnology and Bioengineering 93, 727-736 (2006). 15. Zha, W., Shao, Z., Frost, J. W. & Zhao, H. Rational Pathway Engineering of Type I Fatty Acid Synthase Allows the Biosynthesis of Triacetic Acid Lactone from d-Glucose in Vivo. Journal of the American Chemical Society 126, 4534-4535 (2004). 16. Jez, J. M. et al. Structural control of polyketide formation in plant-specific polyketide synthases. Chemistry & Biology 7, 919-930 (2000). 17. Huck, W. R., Bürgi, T., Mallat, T. & Baiker, A. Asymmetric Hydrogenation of 4- Hydroxy-6-methyl-2-pyrone: Role of Acid-Base Interactions in the Mechanism of Enantiodifferentiation. Journal of Catalysis 200, 171-180 (2001). 18. Huck, W. R., Bürgi, T., Mallat, T. & Baiker, A. Palladium-catalyzed enantioselective hydrogenation of 2-pyrones: evidence for competing reaction mechanisms. Journal of Catalysis 219, 41-51 (2003).

155 19. Huck, W. R., Mallat, T. & Baiker, A. Potential and Limitations of Palladium-Cinchona Catalyst for the Enantioselective Hydrogenation of a Hydroxymethylpyrone. Journal of Catalysis 193, 1-4 (2000). 20. Huck, W. R., Mallat, T. & Baiker, A. Heterogeneous enantioselective hydrogenation of 2-pyrones over cinchona-modified palladium. New Journal of Chemistry 26, 6-8 (2002). 21. Huber, U. (Givaudan Corporation, U.S., 1978). 22. Hansen, C. A. & Frost, J. W. Deoxygenation of Polyhydroxybenzenes: An Alternative Strategy for the Benzene-Free Synthesis of Aromatic Chemicals. Journal of the American Chemical Society 124, 5926-5927 (2002). 23. Hwang, Y.-L. & Bedard, T. C. Ketones (John Wiley & Sons, Inc., 2000). 24. Stafford, A. E., Black, D. R., Haddon, W. F. & Waiss, A. C. Analysis and improved synthesis of parasorbic acid. J. Sci. Fd. Agric. 23, 771-776 (1972). 25. Le Sann, C. et al. Assembly intermediates in polyketide biosynthesis: enantioselective syntheses of -hydroxycarbonyl compounds. Organic & Biomolecular Chemistry 3, 1719-1728 (2005). 26. Gottlieb, H. E., Kotlyar, V. & Nudelman, A. NMR Chemical Shifts of Common Laboratory Solvents as Trace Impurities. The Journal of Organic Chemistry 62, 7512- 7515 (1997). 27. M. J. Frisch, G. W. T., H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, J. A. Montgomery, T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, O. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith, A. Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez, J. A. Pople. (2004).

156

Chapter 7 Mechanistic insights into ring-opening and decarboxylation

of 2-pyrones

7.1 Introduction

As mentioned in the previous chapter, 2-pyrones comprise an important class of compounds that can be derived from biomass by genetically-modified polyketide biosynthesis routes, thereby serving as intermediates in the sustainable production of biorenewable chemicals1-3. This integrated strategy is part of an effort to demonstrate that an expanded array of biologically-derived platform chemicals can be produced from a single metabolic pathway, thereby providing flexibility in generating a series of homologous molecules which are catalytically upgraded to functional and/or direct replacements of currently used petrochemicals3.

Representative 2-pyrones which were introduced in the previous chapter are shown in

Figure 7-1, and include 1, 5,6-dihydro-4-hydroxy-6-methyl-2H-pyran-2-one (3), and 4-hydroxy-

6-methyltetrahydro-2-pyrone (6). We demonstrated that hydrogenation of 1 over a Pd catalyst provides access to 3 and 6, and these three 2-pyrones can be converted to bifunctional chemicals including 2,4-pentanedione (2), 3-penten-2-one (4), 4-hydroxy-pentanone (5), and sorbic acid

(8)1. Surprisingly, we observed for the first time the thermally-activated ring-opening and decarboxylation of both 1 and 3 in liquid water at low reaction temperatures(< 373 K), which is in contrast to the high temperature and acidic reaction environments required for decarboxylation of alkyl or unsaturated carboxylic acids4, 5, and lactones (e.g., pentanoic acid, pentenoic acid, and

γ-valerolactone)6, 7.

157

Figure 7-1 (a) Reactions discussed in this chapter. Compounds are as follows: 4-hydroxy-6- methyl-2-pyrone/ triacetic acid lactone (1); 2,4-pentanedione (2); 5,6-dihydro-4-hydroxy-6- methyl-2H-pyran-2-one (3); 3-penten-2-one (4); 4-hydroxy-pentanone (5); 4-hydroxy-6- methyltetrahydro-2-pyrone (6); parasorbic acid (7); sorbic acid (8); 1,3-pentadiene (9); (b) Nomenclature for the ring-carbon and ring-oxygen atoms of 1.

In this chapter, we probe the mechanistic aspects that control the ring-opening and decarboxylation of 1, 3, and 6 using experimental reaction kinetics measurements together with first-principles density functional theory (DFT) calculations8. For example, we show that the

C5=C6 (Figure 7-1b) bond in the pyrone ring of 1 allows for ring–opening through the nucleophilic addition of water at the C2 lactone carbonyl, leading to decarboxylation by formation of a zwitterion intermediate in solution; and, we show that the presence of a C4=C5 bond in the isomer of 3 leads to a low barrier retro-Diels Alder reaction that eliminates the CO2 dienophile from the resulting diene. In contrast, 6 shows low reactivity. Accordingly, we establish reactivity trends for 2-pyrones based upon key structural features of the substrate molecule including the position of the C=C bond in the ring, as well as the degree of substitution and type of functional group at the C4 and C6 positions of the ring. Additionally, it is notable that

158 besides their potential as platform chemicals, 2-pyrones of natural and synthetic origin, including

1, have been demonstrated to possess synthetic potential and versatility as building blocks for aromatics9 and bioactive compounds10. Therefore, the findings here provide guidance in the selection of conditions (e.g., solvent, temperature) for conversion of these compounds.

7.2 Decarboxylation of triacetic acid lactone

Experimental reactivity trends were obtained in batch reactor studies (Table 7-1) of 1, 3, and 6 in liquid water or tetrahydrofuran (THF) solvents, at different temperatures and reaction times, and in the presence or absence of an acid catalyst (i.e., Amberlyst 70). Entries 3-4 in

Table 7-1 show that 1 undergoes ring-opening and decarboxylation in the absence of catalyst when water is used as the solvent, leading to the production of 2 and CO2, with yields of 2 higher than 90%.

Table 7-1 Results for batch reactions with 1, 3 and 6 as reactants in water and THF solvents.a

Reactant feed Reactant Product Catalyst Time Entry Solvent Reactant concentration T(K) conversion Selectivity (Y/N) (h) (wt%) (%) Product % 1 Water 1 0.8 Y 1 373 68.5 2 78.1 2 Water 1 0.8 Y 4 373 95.3 2 >99 3 Water 1 0.8 N 1 373 77.1 2 81.7 4 Waterb 1 0.8 N 4 373 94.3 2 95.6 5 THF 1 1.8 Y 4 373 23.6 2 71.5 6 THF 1 1.8 Y 16 373 62.4 2 65.3 7 THFb,c 1 1.8 Y 4 373 3.3 2 >99 8 THFb 1 1.8 N 4 373 5.5 2 0.0 9 Waterb 3 1.5 N 1 343 67.8 4 49.9 10 Water 3 1.5 N 1 373 >99 4 37.1 11 Water 3 1.5 N 4 373 >99 4 48.6 12 Waterb 3 1.5 Y 1 323 64.9 4 20.5 5 79.5 13 Water 3 1.5 Y 1 343 >99 4 15.7 5 77.2 14 THFb 3 1.5 N 4 373 3.2 4 >99 15 THF 3 1.5 Y 1 343 39.2 4 57.4 16 THFb 3 1.5 Y 4 373 >99 4 57.7 17 Water 6 2.0 Y 4 373 6.8 7 >99

159 18 Water 6 2.0 N 4 373 0.0 -- -- 19 Water 6 2.0 N 12 443 67.7 7 >99 20 THF 6 2.0 Y 4 373 50.6 7 >99 21 THF 6 2.0 N 4 373 0.0 -- -- 22 THF 6 2.0 N 12 443 0.0 -- -- aReaction conditions: 21 bar He, Amberlyst 70 as catalyst where indicated, mass ratio of reactant:catalyst = 2:1. bData taken from Chapter 6 and Chia et al.1. cCatalyst dried at 373 K prior to reaction. dFeed concentrations varied according to the solubility limits of reactants in respective solvents.

Our experimental results were used to guide the reaction pathways explored using first principle periodic gradient corrected density functional theory calculations, as implemented in the Vienna ab initio software program, VASP.11 The theoretical and experimental results were subsequently used to understand the ring-opening and decarboxylation mechanisms for the 2- pyrone molecules 1, 3, and 6. Several reaction mechanisms were explored, and the mechanism having the lowest energy path and supporting the experimentally observed reactivity trends is presented here. The energies reported herein do not take into account temperature corrections or entropy. The results from theory presented in Figure 7-2 suggest that ring-opening of 1 in water proceeds through the keto-enol tautomerization (KET) of 1 to 1a, followed by nucleophilic addition of water to the lactone carbonyl (1a to 1b) with activation energies of 42 and 52 kJ/mol, respectively. The transition state for the water addition step is shown in Figure 7-3a. The reaction proceeds by the coordination of water to the C2 carbon (C-O bond length = 1.54 Å).

The bond between C2 and ring-oxygen in the transition state is increased to 1.61 Å as compared to 1.38 Å in the reactant state. Although other mechanisms for the ring-opening of 1 and 1a are possible through nucleophilic addition of water at the C5=C6 bond (1 to 1f to 1g) or at the C2 position of the lactone carbonyl (1 to 1g)12, our calculations indicate that these alternative routes have higher activation barriers (Figure 7-2b).

160

Figure 7-2 (a) Proposed mechanism for the ring-opening/ hydration and decarboxylation of 1 to 2 in water, (b) DFT-calculated energy diagram for the reaction pathway of 1 to 2 in water, numbers indicate energy in kJ/mol, (c) measured rates of thermally-activated ring-opening and decarboxylation of 1 at various reaction temperatures (no catalyst). 21 bar He, space time = 70 min. Measured apparent activation energy barrier = 58 ± 12 kJ/mol (95% confidence interval).

Following ring-opening of 1a, the resultant β-keto acid (1b) undergoes proton transfer

(∆Ea = 20 kJ/mol), to form a stable (i.e., an optimized local energy minimum on the potential energy surface) zwitterion intermediate (1c) in solution, that can subsequently decarboxylate to produce 1,3-pentadiene-2,4-diol, (1d), with a calculated activation energy barrier of 30 kJ/mol.

The transition state for the decarboxylation of 1c depicted in Figure 7-3b is late along the reaction coordinate, with a C2-C3 separation of 2.26 Å. The proposed zwitterion intermediate is

161 consistent with the literature for the thermal decarboxylation of β-lactones13-14. 1d then undergoes subsequent KET reactions to form 1e (∆Ea = 18 kJ/mol), followed by the formation of the final product 2 (∆Ea = 29 kJ/mol). The apparent activation energy barrier (Eapp) for the overall reaction of 1 to 2 was calculated to be 59 kJ/mol (Figure 7-2b), which is in good agreement with the experimentally measured value of 58 ± 12 kJ/mol (for 95% confidence interval) (Figure

7-2c).

Figure 7-3 Reactant, transition and product states in the (a) ring-opening of 1a to 1b, and (b) decarboxylation of 1c to 1d. Bond lengths are given in Å. For clarity, only the local water molecules are shown.

162 Further evidence for water-assisted ring-opening of 1 is the lower reactivity of 1 when the reaction is carried out in THF, without the formation of 2 (Table 7-1, Entry 8). Additionally,

DFT calculations in water indicate that the rate controlling step is the nucleophilic addition of water to 1 to open the ring (1a to 1b), and thus reaction rates should not be influenced by the presence of an acid. This prediction is consistent with our experimental results which show that the addition of an acid catalyst does not result in an appreciable difference in the conversion of 1 or the production of 2 when water or THF is used as the solvent (Table 7-1, Entries 7-8, 1 and 3).

The catalytic conversion of 1 was observed to be sensitive to the amount of water adsorbed on the catalyst, which is consistent with the aforementioned results and the proposed mechanism. In particular, the conversion of 1 to 2 in the THF solvent becomes significant only when the catalyst is not dried prior to use (Table 7-1, Entries 5-6). According to thermogravimetric analyses, Amberlyst 70 adsorbs up to 15 wt% water when exposed to ambient conditions, which is consistent with the amount of 2 that is formed in the presence of catalyst that had not been dried prior to use (Entry 6).

7.3 Decarboxylation of 5,6-dihydro-4-hydroxy-6-methyl-2H-pyran-2-one

Similar to the behavior of 1, we observed that 3 undergoes ring-opening and decarboxylation to 4 in liquid water without a catalyst (Table 7-1, Entries 9-11). The presence of an acid catalyst used in combination with water as the solvent was observed to result in the formation of a significant amount of 5 (Table 7-1, Entries 12-13). The improved carbon balance accompanying the formation of 5 suggests that although decarboxylation occurs to produce 4, compound 4 is lost to unidentifiable side-products and to evaporation from the liquid phase (i.e.,

4 boils at 394 K if not further hydrated to 5 under acidic conditions in water).

163

Figure 7-4 (a) Proposed mechanism for the ring-opening and decarboxylation of 3 in water, (b) DFT-calculated energy diagram for the reaction pathway of 3 to 4 in water, numbers indicate energy in kJ/mol, (c) rates of thermally-activated ring-opening and decarboxylation of 3 at various reaction temperatures in water (no catalyst). 21 bar He, space time = 13 min. Measured apparent activation energy barrier = 42 ± 18 kJ/mol (95% confidence interval), (d) rates of thermally-activated ring-opening and decarboxylation of 3 at various reaction temperatures in water over Amberlyst 70. 21 bar He, WHSV = 15 h-1. Measured apparent activation energy barrier = 18 ± 4 kJ/mol (95% confidence interval).

164

Unlike 1, compound 3 does not possess a C=C bond at the C5-C6 position to facilitate nucleophilic addition of water at C2, and the stoichiometry of the reaction (3 to 4) indicates that water is not a co-reagent. Thus, we suggest that ring-opening of 3 proceeds through KET to 3a followed by formation of the enolic isomer, 3, 6-dihydro-4-hydroxy-6-methyl-pyran-2-one (3b),

Figure 7-4a. These KET reactions occur rapidly due to their low DFT-predicted activation barriers of 34 (3 to 3a) and 22 kJ/mol (3a to 3b), respectively (Figure 7-4b), compared to the higher barrier for decarboxylation (49 kJ/mol).

The intermediate 3b subsequently reacts through a retro-Diels-Alder (rDA) mechanism to produce 3-hydroxy-penta-1,3-diene (3d) and CO2. In the gas phase, the rDA reaction occurs in a single concerted step (3b to 3d) with an activation barrier of 123 kJ/mol. The transition state structure depicted in Figure 7-5a for the rDA reaction shows a ring-opened structure, with C2-C3 bond length of 1.7 Å and C6-O1 separation distance of 2.32 Å. We found that the solvation of 3b by water significantly decreases the activation barrier for rDA. Under conditions of full solvation

(27 water molecules/unit cell), the rDA reaction was found to proceed in two steps through a zwitterion intermediate (3b to 3c), followed by decarboxylation (3c to 3d), with activation energy barriers of 31 and 49 kJ/mol, respectively. This proposed zwitterion intermediate is consistent with the literature, for which both one-step and two-step Diels Alder mechanisms15 have been proposed to proceed through the formation of a zwitterions or biradical intermediate16-

20. The structures of the transitions states of the two-step rDA reaction in water are shown in

Figure 7-5b. The C6-O1 separation distance in the transition state of the ring-opening step (3b to

3c) (2.02 Å) is 0.3 Å shorter than that in the gas phase. While water is not directly involved in the transition states for ring-opening or decarboxylation in the rDA reaction, water accelerates the rate by stabilizing the highly polar or charged intermediate and transition states through

165 hydrogen bonding interactions21-23. Similarly, water is known to influence the Diels-Alder mechanism24-26. Consistent with the literature, the lower activation barriers for rDA in water calculated here are attributable to the reduction in the C6-O1 separation distance of the transition states for ring-opening of 3b in the gas phase (Figure 7-5a) compared to that in water (Figure

7-5b) and hydrogen bonding interactions of water and the transition state for 3b.

Figure 7-5 Reactant, transition and product state structures for the ring-opening and decarboxylation of 3b in (a) gas phase, and (b) solution phase (27 water molecules/unit cell). Bond lengths are given in Å. For clarity, only the local water molecules are shown.

The 3d diene that results from decarboxylation can undergo tautomerization (∆Ea =

27 kJ/mol) to 4 (Figure 7-4a). The overall activation barrier for the ring-opening and decarboxylation of 3 is estimated to be 50 kJ/mol (Figure 7-4b), which agrees with the experimentally measured value of 42±18 kJ/mol (95% confidence interval) (Figure 7-4c). The experimentally measured activation barrier for ring- opening and decarboxylation of 3 is

166 approximately 16 kJ/mol lower than that for 1, in agreement with the results from DFT calculations.

In contrast to the behavior of 1, the presence of an acid catalyst for the conversion of 3 in water results in a significant increase in the rate of reaction (Table 7-1, Entries 9 and 13).

Moreover, the increase in rate of conversion of 3 was insensitive to the amount of adsorbed water on the catalyst. Accordingly, the Brønsted acid sites of Amberlyst 7027 play a critical role in the overall ring-opening and decarboxylation of 3. The rate enhancement of the Diels-Alder reaction in water by an acid catalyst is well known28, and has been suggested for the rDA reaction29, 30. Similarly, KET reactions in water are thought to be initiated by trace amounts of acid or base31. Results from our DFT calculations for the KET of 2-pyrones give a high activation energy barrier (Ea = 82 kJ/mol for the KET of 1 to 1a) for the one-step un-catalyzed reaction which requires three fully-solvated water molecules within the solution phase to shuttle the proton from the hydroxyl group on the enol to the adjacent carbon (Figure 7-6a), consistent

32, 33 with the literature . In the case of 2-pyrones, the pKa values of the 4-hydroxy group of 3 and 1 are similar to that of acetic acid at 5.434 and 4.7335, respectively, indicating that the KET of 3d to

4 is likely to proceed through an acid-catalyzed route. The KET reaction is facilitated by protons present in neutral water and further enhanced by the addition of Amberlyst 70. Figure 7-6b shows the transition state in which a proton is added to C3 with C3-H separation of 1.39 Å. The experimentally measured activation barrier for ring-opening and decarboxylation of 3 in water over Amberlyst 70 was 18±4 kJ/mol (95% confidence interval) (Figure 7-4d), which is 24 kJ/mol lower than the barrier reported above in water alone (42 kJ/mol).

167 It was observed that the reactivity of 3 is lower when THF is used as the reaction solvent

(Table 7-1, Entry 14), further indicating that the rDA reaction for 3 is accelerated by water. DFT calculations indicate that the rDA reaction in THF proceeds through a one-step mechanism (3b to 3d) without the formation of a stable zwitterion intermediate. This one-step rDA reaction in

THF is similar to that in the gas phase and occurs with a higher activation energy barrier (90 kJ/mol) compared to the reaction in water as a result of the decreased stabilization of the polar transition state in moving from water to THF. The calculated activation energy barriers for KET in THF (Figure 7-4a) were also found to be high (>200 kJ/mol), consistent with the experimental results where low conversion of 3 was observed with THF as the solvent in the absence of catalyst (3%, Entry 14, Table 7-1).

Figure 7-6 Structures of reactants, transition states and products of (a) un-catalyzed (without acid) tautomerization of 1 to 1a, and (b) acid-catalyzed tautomerization of 3d to 4. For clarity, only the local water molecules are shown.

168 7.4 Reactivity of 4-hydroxy-6-methyltetrahydro-2-pyrone

It was observed that ring-opening and decarboxylation of 6 do not occur under the reaction conditions used for 1 and 3. Instead, 6 undergoes dehydration to form parasorbic acid (7) with quantitative selectivity in the presence of water or an acid catalyst. When THF is used as the solvent (Entries 20-22, Table 7-1), 6 undergoes dehydration only in the presence of an acid catalyst. Entries 17-19 show that with water as the solvent, the dehydration of 6 to 7 occurs at elevated reaction temperatures even in the absence of a catalyst. The dehydration of 6 to 7 in water observed here is consistent with the literature where the dehydration of alcohols is known to be catalyzed by protons that are present in water at elevated temperatures in the absence of catalyst36-38. Figure 7-7a shows two possible products that can form from the dehydration of 6, namely 7 and isoparasorbic acid (6a); 7 was calculated to be more stable than 6a (18 kJ/mol) and is formed with a low activation energy barrier (6 to 7, 88 kJ/mol, Figure 7-7b). The transition state for the dehydration step (Figure 7-8) shows a proton abstracted from C3 and water coordinated to C4 with a bond length of 1.5 Å. The activation energy barrier for the acid- catalyzed isomerization of 7 to 6a was calculated to be 142 kJ/mol. Therefore, the higher barriers for dehydration of 6 to 6a and isomerization of 7 to 6a are consistent with experimentally observed high selectivities from conversion of 6 to 7 (> 99%).

169

Figure 7-7 (a) Proposed mechanism for the dehydration, ring-opening and decarboxylation of 6 in water, (b) DFT-calculated energy diagram for the reaction pathway of 6 to 7 to 9 (solid line) and 6 to 6a to 9 (dashed line) in water, numbers indicate energy in kJ/mol, (c) reactant, product and transition state for the dehydration step of 6 to 7. Bond lengths are given in Å. (d) Ring- opening and decarboxylation of 3,6-dihydro-4,6,6-trimethyl-2H-pyran-2-one (10) to 2,4- dimethyl-1,3-pentadiene (11), (e) DFT-calculated energy diagram for the reaction pathway of 10 to 11 in water.

Figure 7-8 Reactant, transition and product states for the dehydration step of 6 to 7. Bond lengths are given in Å. For clarity, only the local water molecules are shown.

170 Compound 7 can undergo ring-opening to 8 through KET of 7 to the stable enol intermediate 7a, followed by electrocyclic ring-opening to 8 (Figure 7-7a). The activation barriers for the KET and ring-opening steps were calculated to be 81 (7 to 7a) and 54 (7a to 8) kJ/mol, respectively (Figure 7-7b). The results from the potential energy diagram shown in

Figure 7-7b indicate that 7 is the most stable intermediate in the conversion of 6 to 8, which is consistent with our experimental observations where 7 is the predominant product under mild reaction conditions (T = 373 K). In our previous work, we observed the formation of 8 using

THF as the solvent at elevated temperatures of 443 K1. However 8 does not undergo decarboxylation at 443 K, which is consistent with the DFT-estimated activation energy barrier of 149 kJ/mol for the decarboxylation of 8 to 1, 3-pentadiene (9) in water. In the presence of a β- keto group (as in the case of 1a and 3a), the zwitterion intermediate (1c and 3c) readily decarboxylates via the formation of a late product-like transition state (Figure 7-3b and Figure

7-5b). In contrast, due to the lack of a β-keto group, 8 is difficult to decarboxylate, and the transition state appears somewhat early along the reaction coordinate, with a bond length of 1.75

Å, Figure 7-9.

Figure 7-9 Structures of reactants, products and transition states of decarboxylation of 8 to 9. Bond lengths are given in Å. For clarity, only the local water molecules are shown.

171 Our proposed mechanism for conversion of 3 (presented in Figure 7-4) suggests that 9 could be formed from 6a in a similar way as a result of the C4=C5 bond which is predicted to allow for direct ring-opening and decarboxylation through a rDA reaction. To probe this prediction, a synthesized sample of 6a (0.5 wt% 6a as feed) was reacted at 373 K for 4 h in a

THF/water mixture (mass ratio of 1:1), and 22% conversion of 6a was observed, with significant amounts of highly reactive and volatile 9. In contrast, when THF was the solvent, 6a was not converted under the same reaction conditions. Therefore, the decarboxylation of 6a in water at low reaction temperatures is consistent with our proposed mechanism through the rDA reaction.

Based on the ring-opening and decarboxylation mechanisms we have proposed for 3b and 6a, respectively, it is suggested that analogous structures with the C4=C5 bond in the ring should undergo decarboxylation in the absence of a catalyst via the rDA reaction. To probe this hypothesis, we carried out studies of 3,6-dihydro-4,6,6-trimethyl-2H-pyran-2-one (10, Figure

7-7c). Results from DFT calculations indicate an activation barrier of 95 kJ/mol for the one-step concerted ring-opening and decarboxylation of 10 in water by rDA (Figure 7-7d). Similar to 6a, the absence of the hydroxyl group at C4 in 10 prevents formation of a stable zwitterion intermediate in solution and leads to a single step rDA mechanism. Independent of the presence of methyl groups at C4 and C6, the transition states (Figure 7-10) for the rDA reaction for both

6a and 10 are similar and show a decarboxylated molecule with C3-C2 separation of 2.2 Å

(Figure 7-10a) and 2.1 Å (Figure 7-10b), respectively. In agreement with theory, batch reactions with 10 in THF resulted in no observed conversion. With a THF/water mixture (mass ratio THF: water = 1:1) as the solvent, 35% conversion of 10 and 70% selectivity to 2,4- dimethyl-1,3-pentadiene (11) was observed (373 K, 4 h, 2 wt% 10 as feed).

172

Figure 7-10 Reactant, product and transition state structures for rDA reactions of (a) 6a and (b) 10 in aqueous solution. Bond lengths are given in Å. For clarity, only the local water molecules are shown.

The ring-opening and decarboxylation of 10 in liquid water without the aid of a catalyst is similar to reactivity trends observed for 3b and 6a. All three molecules undergo rDA as a result of the precursor C4=C5 bond and stabilization from the resulting diene by substituents at the C6 position. The trend in the calculated activation barriers for the single step rDA reaction of the three molecules in the gas phase order as:

6a (193 kJ/mol) > 10 (126 kJ/mol) > 3b (123 kJ/mol)

The trend of decreasing activation barrier is directly related to the lowering of the highest-occupied molecular orbital (HOMO) of the resulting diene upon the introduction of an electron-donating substituent at the C4 position. This donation stabilizes the resulting positive charge at the C6 position upon activation of the C-O bond and increases the overlap with the

173 lowest unoccupied molecular orbital (LUMO) on the CO2 dienophile thus promoting the rDA reaction. In contrast, the trend in the calculated activation barriers for the rDA reaction of the three molecules in water is of the following order, which suggests a significant influence of the solvent in lowering the activation barrier:

10 (95 kJ/mol) > 6a (89 kJ/mol) > 3b (31 and 49 kJ/mol)

This trend is consistent with the higher conversion and reactivity measured for 3 over 6a and 10.

7.5 General reactivity rules

The reactivity trends we have observed for 2-pyrone molecules can be related to key structural features of this family of molecules. An overview of the structure-reaction relationships is displayed in Figure 7-11 and provides a universal set of rules by which 2- pyrones react. The observed ring-opening and decarboxylation of 3b, 6a and 10 in water without the aid of a catalyst indicates that the presence of a C4=C5 bond in the ring is required to carry out the low energy rDA reaction and eliminate the CO2 dienophile from the resulting diene. The nature of the substituent attached to the C4 atom during the ring opening/decarboxylation reaction dictates whether the rDA reaction proceeds through a single concerted or a two-step mechanism through a stable zwitterion intermediate. Additionally, the polarized transition state and zwitterion intermediate in the rDA are stabilized by hydrogen-bonding in protic solvents like water. The overall rDA reactivity of the pyrone therefore depends on the electron-withdrawing ability of the substituent attached to the C4 site. Results from DFT calculations show that 2- pyrone molecules with two or more double bonds, as in the case of 1f and 7a, with one of the double bonds at the C4=C5 position, will undergo direct ring-opening in water, similar to the first step of the rDA reaction 3b to 3c.

174

Figure 7-11 Overview of structure-reaction relationships for 2-pyrones. Abbreviations: retro- Diels Alder (rDA), keto-enol tautomerization (KET), and apparent activation energy barrier (Eapp).

The presence of an unsaturated C=C bond at the C5-C6 position in the pyrone ring, such as in 1, allows for ring–opening through the nucleophilic addition of water at the C2 lactone carbonyl to stabilize the positively charged carbenium ion center that results upon the C-O rupture. The presence of a β-keto group, as in the case of 1b, facilitates decarboxylation through the formation of a zwitterion intermediate (1c) in solution. The decarboxylation of the zwitterion intermediate (1c to 1d) is similar to the second step of the rDA reaction (3c to 3d).

175 Although 1, 3 and 6 differ by only one degree of unsaturation from one another, they undergo ring-opening and/or decarboxylation through different mechanisms. The presence of the

C=C bonds adjacent to the C4 hydroxyl group in the ring allows for the KET reactions observed for 1 and 3. For 1, the keto form is more reactive for the nucleophilic addition of water, which results in the formation of the β-keto acid, while the enol at the C4=C5 position of 3b is responsible for the rDA reaction. Because 6 does not possess unsaturated C=C bonds within the ring, the C4 hydroxyl group is unable to convert into a β-keto group and 6 does not undergo decarboxylation. Species 6, however, can dehydrate in the presence of water and under acidic conditions to form 7.

7.6 Conclusions

Experimentally-observed reactivity trends and results from DFT calculations provide mechanistic insight into reactivity trends for 2-pyrones, an important class of molecules for the production of biorenewable chemicals. Our results indicate that the presence of a double bond at the C4=C5 position in the ring (3b, 6a, 10) leads to a rDA reaction that eliminates CO2 as the dienophile, and this rDA process is accelerated by electron-donating substituents on the resulting dienophile, protic solvents and acidic conditions. Therefore, 2-pyrones with a C4=C5 bond readily undergo ring-opening in water alone and under acidic conditions, but these 2-pyrones are stable in aprotic solvents like THF. A C5=C6 bond in the ring (1) allows for ring–opening through the nucleophilic addition of water at either the C6 or C2 lactone carbonyl, and acidic conditions are ineffective for the acceleration of these reactions. Therefore, 2-pyrones with a

C5=C6 bond will ring-open in water but are stable in THF and under acidic conditions. Finally, fully saturated molecules (6) do not ring-open easily regardless of solvent and acidity.

176 7.7 Chemical Notations

7.8 Synthesis of isoparasorbic acid

3-buten-2-ol (0.601 mL, 6.93 mmol) was added to a solution of 3-butenoic acid (0.589 mL, 6.93 mmol) dissolved in distilled chloroform (15 mL) under argon. p-Toluenesulfonic acid

(0.013 g, 1 mol %) was added and the reaction flask was fit with a Dean-Stark trap. The solution was heated to reflux overnight, cooled and concentrated in vacuo under reduced pressure. The yellow liquid obtained was sufficiently pure for use in the next reaction.

Titanium isopropoxide (0.089 mL, 0.3 mmol) was added to a solution of crude 3-butenoic acid and 1-methyl-2-propen-1-yl ester (0.140 g, 1 mmol) dissolved in dry methylene chloride under argon. The reaction mixture was heated to reflux for 1.5 h, cooled to room temperature, and Grubbs second generation catalyst (10 mg, 1 mol %) was added in one portion and heated to reflux for 12 h. After cooling, the reaction was filtered through a thin pad of Celite and concentrated in vacuo. The crude material was purified by flash column chromatography (10:1 –

3:1 hexanes:ethyl acetate) to afford the product as a light brown oil in 64% yield. We modified the conditions reported by Andreana et al.39. NMR analysis gave the following result:

177 1 H NMR (300 MHz, CDCl3) δ 1.36 (d, J = 6 Hz, 3H), 2.95-2.98 (m, 2H), 4.97-5.03 (m, 1H), 5.76

(s, 2H).

7.9 Computational methods

Periodic gradient corrected density functional theory (DFT) calculations as implemented in the Vienna ab initio Simulation Package (VASP)11 were carried out to examine the elementary steps involved in the ring-opening and decarboxylation for the 2-pyrone molecules 1, 3 and 6.

The Kohn-sham40 equations were solved by using plane wave basis functions expanded out to a cutoff energy of 396 eV. Non-local gradient corrections to the exchange and correlation energies were calculated using the Perdew and Wang form of the generalized gradient approximation

(GGA)41. Vanderbilt ultrasoft pseudopotentials42 were used to describe the core electrons and the nuclei of the atoms. The structural relaxations were performed until the forces on each atom were below 0.05eV/Å43. Converging the aqueous phase structures to higher tolerances was found to be considerably more difficult as a result of low energy modes in the water network. We were successful in converging some of the structures to within 0.025eV/Å and found that the resulting reaction and activation energies were less than 2 kJ/mol than those from the 0.05eV/Å convergence simulations. The wave functions were all converged to within 1 × 10-6 eV.

All the reactions discussed in this study were simulated in an aqueous solution which was modeled by filling the unit cell with explicit water molecules to simulate a density of 1 g/cm3.

The oxygenate reagent along with 27 water molecules were placed in 9.86 Å x 9.86 Å x 9.86 Å cubic unit cell. The initial configurations of the water molecules within the cell were obtained by simulating the system out to 5 ps using classical molecular dynamic simulations at 400 K as implemented in the Discover Molecular Dynamics program of the Materials Studio 5.0 (Accelrys

178 17Inc., USA). The complexity of the solid Amberlyst™ 70 catalyst in the presence of water made it too difficult to model directly. Acid-catalyzed reactions were modeled instead by placing a proton in aqueous solution to mimic the solid acid site. The energy to deprotonate a solid acid is a direct measure of its Brønsted acidity. As such, the solvated proton provides an upper limit on the reactivity of the Amberlyst 70 catalyst as well as other solid acids. In order to better understand and compare the effects of solvent, the calculations were also performed in the gas phase and, in some cases, an explicit THF solvent. In the gas phase systems, the isolated reactant was simulated in an 18 Å cubic unit cell. To simulate the reaction in the THF solvent system, the reactant molecule was placed in a cubic unit cell of side 15 Å surrounded by 15 THF molecules in order to match the 0.8892 g/cm3 density of THF. All the subsequent gas phase and solvent phase calculations were performed using non-spin polarized DFT simulations.

The activation barriers were calculated using a two-step approach involving the Nudged

Elastic Band (NEB) method44 to establish the reaction trajectories and the Dimer method45 to subsequently isolate the transition state. In the NEB simulations, a set of 16 images between the initial and the final states were optimized along the potential energy surface until the force on each atom converges to 0.25 eV/Å. The dimer method which uses the two closely spaced images derived from the highest energy NEB image, was subsequently used to “walk” the dimer up along the potential energy surface to the transition state. The dimer structures were converged to

0.05 eV/ Å. The activation barrier was calculated as the difference in the energy between the reactant and the transition state. The results reported herein do not explicitly consider the effects of temperature and entropy. We have, however, carried out more rigorous Car Parrinello molecular dynamic (CPMD) simulations for keto-enol tautomerization reaction that explicitly simulate the changes in entropy and establish free energy barriers. A comparison of the free

179 energy barrier for the KET reaction of 3d to 4 from the CPMD simulations (25 kJ/mol) and that found from the static calculations (27 kJ/mol) were in good agreement which suggests that the enthalpic barriers reported herein provide a reasonable first approximation to the actual barriers.

The calculated activation barriers of the intrinsic reaction steps may differ from the real system as the full complexity of Amberlyst catalyst and the actual solution phase cannot be explicitly simulate. Comparing the simulation results with experimental observations allows us to test the validity of the model structures and reaction environments used to simulate these reactions, and in addition, the ability to develop structure-reactivity relationships.

7.10 References

1. Chia, M., Schwartz, T. J., Shanks, B. H. & Dumesic, J. A. Triacetic acid lactone as a potential biorenewable platform chemical. Green Chemistry 14, 1850-1853 (2012). 2. Jacoby, M. Teaming up for bio-based chemicals. Chemical and Engineering News 90, 37-38 (2012). 3. Nikolau, B. J., Perera, M. A. D. N., Brachova, L. & Shanks, B. Platform biochemicals for a biorenewable chemical industry. The Plant Journal 54, 536-545 (2008). 4. Simakova, I., Simakova, O., Maki-Arvela, P. i. & Murzin, D. Y. Decarboxylation of fatty acids over Pd supported on mesoporous carbon. Catalysis Today 150, 28-31 (2010). 5. Bigley, D. B. & May, R. W. Studies in decarboxylation. Part IV. The effect of alkyl substituents on the rate of gas-phase decarboxylation of some [small beta][gamma]- unsaturated acids. Journal of the Chemical Society B: Physical Organic 0, 557-559 (1967). 6. Bond, J. Q., Wang, D., Alonso, D. M. & Dumesic, J. A. Interconversion between Î³- valerolactone and pentenoic acid combined with decarboxylation to form butene over silica/alumina. Journal of Catalysis 281, 290-299 (2012). 7. Bond, J. Q., Martin Alonso, D., West, R. M. & Dumesic, J. A. g-Valerolactone Ring- Opening and Decarboxylation over SiO2/Al2O3 in the Presence of Water. Langmuir 26, 16291-16298 (2010). 8. Chia, M. et al. Mechanistic Insights into Ring-Opening and Decarboxylation of 2- Pyrones in Liquid Water and Tetrahydrofuran. Journal of the American Chemical Society 135, 5699-5708 (2013). 9. Hansen, C. A. & Frost, J. W. Deoxygenation of Polyhydroxybenzenes: An Alternative Strategy for the Benzene-Free Synthesis of Aromatic Chemicals. Journal of the American Chemical Society 124, 5926-5927 (2002). 10. Goel, A. & Ram, V. J. Natural and synthetic 2H-pyran-2-ones and their versatility in organic synthesis. Tetrahedron 65, 7865-7913 (2009).

180 11. Kresse, G. & Hafner, J. Ab initio molecular dynamics for liquid metals. Physical Review B 47, 558-561 (1993). 12. Fabian, W. M. F. Site of Nucleophilic Attack and Ring Opening of Five-Membered Heterocyclic 2,3-Diones: A Density Functional Theory Study. The Journal of Organic Chemistry 67, 7475-7482 (2002). 13. Moyano, A., Pericas, M. A. & Valenti, E. A Theoretical-Study on the Mechanism of the Thermal and the Acid-Catalyzed Decarboxylation of 2-Oxetanones (-lactones). Journal of Organic Chemistry 54, 573-582 (1989). 14. Mulzer, J. & Zippel, M. Hammett-correlations, solvent effects and stereochemistry of the thermal decarboxylation of 2-oxetanones - evidence for a zwitterionic intermediate. Tetrahedron Letters 21, 751-754 (1980). 15. Wilsey, S., Houk, K. N. & Zewail, A. H. Ground- and Excited-State Reactions of Norbornene and Isomers: A CASSCF Study and Comparison with Femtosecond Experiments. Journal of the American Chemical Society 121, 5772-5786 (1999). 16. Horn, B. A., Herek, J. L. & Zewail, A. H. Retro-Diels-Alder femtosecond reaction dynamics. Journal of the American Chemical Society 118, 8755-8756 (1996). 17. Diau, E. W. G., De Feyter, S. & Zewail, A. H. Femtosecond dynamics of retro Diels- Alder reactions: the concept of concertedness. Chemical Physics Letters 304, 134-144 (1999). 18. Sustmann, R. & Sicking, W. Influence of Reactant Polarity on the Course of (4 + 2) Cycloadditions. Journal of the American Chemical Society 118, 12562-12571 (1996). 19. Sustmann, R., Tappanchai, S. & Bandmann, H. a(E)-1-methoxy-1,3-butadiene and 1,1- dimethoxy-1,3-butadiene in (4+2) cycloadditions. A mechanistic comparison. Journal of the American Chemical Society 118, 12555-12561 (1996). 20. Gassman, P. G. & Gorman, D. B. STEPWISE MECHANISM FOR THE FORMATION OF 2-PI + 4-PI CYCLOADDUCTS IN THE IONIC DIELS-ALDER REACTION. Journal of the American Chemical Society 112, 8624-8626 (1990). 21. Wijnen, J. W. & Engberts, J. Retro-Diels-Alder reaction in aqueous solution: Toward a better understanding of organic reactivity in water. Journal of Organic Chemistry 62, 2039-2044 (1997). 22. van der Wel, G. K., Wijnen, J. W. & Engberts, J. B. F. N. Solvent Effects on a Diels−Alder Reaction Involving a Cationic Diene: Consequences of the Absence of Hydrogen-Bond Interactions for Accelerations in Aqueous Media. The Journal of Organic Chemistry 61, 9001-9005 (1996). 23. Wijnen, J. W. & Engberts, J. B. F. N. A Hetero Retro Diels–Alder Reaction in Aqueous Solution: A Dramatic Water-Induced Increase of the Equilibrium Constant and Inhibition of Cycloreversion. Liebigs Annalen 1997, 1085-1088 (1997). 24. Otto, S. & Engberts, J. Diels-Alder reactions in water. Pure and Applied Chemistry 72, 1365-1372 (2000). 25. Chandrasekhar, J., Shariffskul, S. & Jorgensen, W. L. QM/MM simulations for Diels- Alder reactions in water: Contribution of enhanced hydrogen bonding at the transition state to the solvent effect. Journal of Physical Chemistry B 106, 8078-8085 (2002). 26. Acevedo, O. & Jorgensen, W. L. Understanding rate accelerations for Diels-Alder reactions in solution using enhanced QM/MM methodology. Journal of Chemical Theory and Computation 3, 1412-1419 (2007).

181 27. Siril, P. F., Cross, H. E. & Brown, D. R. New polystyrene sulfonic acid resin catalysts with enhanced acidic and catalytic properties. Journal of Molecular Catalysis A: Chemical 279, 63-68 (2008). 28. Otto, S., Bertoncin, F. & Engberts, J. B. F. N. Lewis Acid Catalysis of a Diels−Alder Reaction in Water. Journal of the American Chemical Society 118, 7702-7707 (1996). 29. Bunnelle, W. H. & Randall Shangraw, W. Acid catalysis of the retro-diels alder reaction. Formation and electrophilic reactivity of 2-methylene-1,3-cyclopentanedione. Tetrahedron 43, 2005-2011 (1987). 30. Demuynck, A. L. W. et al. Retro-Diels–Alder Reactions of Masked Cyclopentadienones Catalyzed by Heterogeneous Brønsted Acids. Advanced Synthesis & Catalysis 352, 3419-3430 (2010). 31. Smith, M. B. & March, J. (John Wiley & Sons). 32. Bernasconi, C. F., Fairchild, D. E. & Murray, C. J. Kinetic Solvent Isotope Effect and Proton Inventory Study of the Carbon Protonation of Amine Adducts of Benzylidene Meldrum Acid and Other Meldrum Acid-Derivatives - Evidence For Concerted Intramolecular Proton-Transfer. Journal of the American Chemical Society 109, 3409- 3415 (1987). 33. Chiang, Y., Kresge, A. J., Santaballa, J. A. & Wirz, J. Ketonization of acetophenone enol in aqueous buffer solutions. Rate-equilibrium relations and mechanism of the uncatalyzed reaction. Journal of the American Chemical Society 110, 5506-5510 (1988). 34. Padbury, G. et al. (eds. Borchardt, R., Freidinger, R., Sawyer, T. & Smith, P.) 211-232 (Springer US, 2002). 35. Fehr, M. J., Consiglio, G., Scalone, M. & Schmid, R. Asymmetric Hydrogenation of Substituted 2-Pyrones. The Journal of Organic Chemistry 64, 5768-5776 (1999). 36. Hunter, S. E., Ehrenberger, C. E. & Savage, P. E. Kinetics and Mechanism of Tetrahydrofuran Synthesis via 1,4-Butanediol Dehydration in High-Temperature Water. The Journal of Organic Chemistry 71, 6229-6239 (2006). 37. Kuhlmann, B., Arnett, E. M. & Siskin, M. Classical Organic Reactions in Pure Superheated Water. The Journal of Organic Chemistry 59, 3098-3101 (1994). 38. Xu, X., Antal, M. J. & Anderson, D. G. M. Mechanism and Temperature-Dependent Kinetics of the Dehydration of tert-Butyl Alcohol in Hot Compressed Liquid Water. Industrial & Engineering Chemistry Research 36, 23-41 (1997). 39. Andreana, P. R., McLellan, J. S., Chen, Y. & Wang, P. G. Synthesis of 2,6- Dideoxysugars via Ring-Closing Olefinic Metathesis. Organic Letters 4, 3875-3878 (2002). 40. Kohn, W. & Sham, L. J. Self-Consistent Equations Including Exchange and Correlation Effects. Physical Review 140, A1133-A1138 (1965). 41. Perdew, J. P. et al. Atoms, molecules, solids, and surfaces: Applications of the generalized gradient approximation for exchange and correlation. Physical Review B 46, 6671 (1992). 42. Vanderbilt, D. Soft self-consistent pseudopotentials in a generalized eigenvalue formalism. Physical Review B 41, 7892 (1990). 43. Chia, M. et al. Selective Hydrogenolysis of Polyols and Cyclic Ethers over Bifunctional Surface Sites on Rhodium-Rhenium Catalysts. Journal of the American Chemical Society 133, 12675-12689 (2011).

182 44. Sheppard, D., Terrell, R. & Henkelman, G. Optimization methods for finding minimum energy paths. The Journal of Chemical Physics 128, 134106 (2008). 45. Henkelman, G. & Jonsson, H. A dimer method for finding saddle points on high dimensional potential surfaces using only first derivatives. The Journal of Chemical Physics 111, 7010-7022 (1999).