Catalytic Conversion of Biomass-Derived Molecules Into Mono- and Dicarboxylic Acids and Esters

Home , Alcohol oxidation, Trifluoroperacetic acid

Catalytic Conversion of Biomass-Derived Molecules into Mono- and Dicarboxylic Acids and Esters

by Yuran Wang

B.S., Tsinghua University (2010) M.S., Massachusetts Institute of Technology (2014)

Submitted to the Department of Chemical Engineering in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

at the

MASSACHUSETTS INSTITUTE OF TECHNOLOGY

June 2016

Signature of Author ………………………………………………………………………………... Department of Chemical Engineering May 12, 2016

Certified by ………………………………………………………………………………………... Yuriy Román-Leshkov Associate Professor of Chemical Engineering Thesis Supervisor

Accepted by ………………………………………………………………………………………... Richard D. Braatz Edwin R. Gilliland Professor of Chemical Engineering Chairman, Committee for Graduate Students

Catalytic Conversion of Biomass-Derived Molecules into Mono- and Dicarboxylic Acids and Esters

by Yuran Wang

Submitted to the Department of Chemical Engineering on May 12, 2016 in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Chemical Engineering

ABSTRACT Biomass can serve as a renewable alternative to the inevitably depleting fossil fuel resources and provide feedstocks for the production of fuels and chemicals. Mono- and dicarboxylic acids and esters are key intermediates to chemical products, especially for biodegradable polymers. This thesis has addressed the challenges in chemocatalytic synthesis of mono- and dicarboxylic acids and esters, including gluconic acid, succinic acid, itaconic acid and their esters, from biomass- derived molecules. Gold (Au) catalysts have been rarely investigated for the oxidation of glucose in the absence of a base. These conditions are critical, however, to enable the sequential one-pot combination of cellulose hydrolysis and glucose oxidation. The study provides insights into the deactivation of the catalysts caused by leaching and hydrothermal sintering of Au nanoparticles, as well as by adsorption of reaction species. We found that lowering the surface density of Au on metal oxides decreases the sintering rate of the Au nanoparticles and hence enhances the stability and activity of the catalyst. Levulinate derivatives are an attractive platform for the production of renewable chemicals. We report on the oxidation of methyl levulinate into dimethyl succinate with peroxides under mild conditions using Brønsted and Lewis acid catalysts and focuses on the reaction selectivity control. While the molecular structure (i.e., carbon chain length and branching around the C=O group) and the oxidant type affect the product distribution, solvent choice has the strongest impact on changing the location of oxygen insertion into the carbon backbone. In contrast to Brønsted acids, for water-tolerant Lewis acidic triflate salts, the reaction selectivity is affected by the size of the metal cation. We have developed a novel approach to synthesize unsaturated dicarboxylic acid esters via aldol condensation of keto esters catalyzed by Lewis acidic zeolites. Hafnium-containing BEA (Hf- BEA) zeolites are highly active, selective and stable for the condensation of ethyl pyruvate into itaconic acid ester analogues. Analysis of the dynamic behavior of Hf-BEA under flow conditions and studies with Na-exchanged zeolites suggest that Hf(IV) open sites possess dual functionality for Lewis and Brønsted acid catalysis. Thesis Supervisor: Yuriy Román-Leshkov Title: Associate Professor of Chemical Engineering

Acknowledgements

I am deeply grateful for all the help and support I have received throughout my Ph.D. that have made the work presented in this thesis possible. First of all, I would like to express my sincere gratitude to my thesis advisor Prof. Yuriy Román-Leshkov for the opportunity to work with him. It has been a great experience. Yuriy is always encouraging and supportive. I remembered the time when he was in the lab showing me how to install the HPLC column, his long email guiding me how to properly plan my research, and our conversations about career development. He also always challenged me to start with what I have in hand instead of waiting till everything is ready. Being one of his first students, I have seen and also really appreciated his constant efforts to equip us with the best resources to do research. I would also like to thank my thesis committee members: Prof. William H. Green, Prof. T. Alan Hatton and Prof. Klavs F. Jensen. Their insightful comments and hard questions prompted me to broaden my research from various perspectives. They also have their lab doors open to me so that I could have access to their facility to accomplish my research. I also appreciate their support and help in my career development. Throughout my Ph.D., I have had the opportunity to work with many brilliant collaborators. First, I would like to thank Dr. Krishna K. Sharma, who has set up the lab when there was nothing, taught me how to run my very first batch reactions and catalyst synthesis and established the foundation for the base-free glucose oxidation work. My sincere thanks go to Dr. Stijn Van de Vyver for his contribution in my very first first-author publication. He has challenged me to present my work and my ideas in an effective way and helped me grow as a researcher. I am grateful to Ferdinand Vogelgsang for his assistance in the Baeyer-Villiger oxidation project. I thank Jennifer D. Lewis, who has been extremely supportive and helpful. I also really appreciated the inspiring discussions we had about aldol condensation. Being one of the first members in the group, I have had the privilege to meet and work with each individual we have had. My sincere gratitude goes to Dr. Mark Mba Wright for his friendship; Dr. Herui Dou for sharing his ‘insights’ into research and life; Dr. Murat Kaya for synthesizing new catalysts for me and always being peaceful; Dr. Teerawit Prasomsri (Air) for all his help, support and patience. I also thank Dr. Mehmet Zahmakiran for setting up the lab; Dr. Tarit Nimmanwudipong for nicely inventorying our Swagelok fittings; Dr. Anthony Crisci for his encouragement and introducing the only beer that I would be happy to drink; Dr. David Simakov for sharing his knowledge about building reactors; Dr. Kenta Iyoki for taking lots of work during lab cleaning; Dr. Maria Millina for sharing her thesis as a template and being helpful and delightful; Dr. Christ Hendon for introducing Slack to the group. Many thanks go to Dr. William Gunther for synthesizing zeolite catalysts for us; Helen Luo for taking care of lab supplies and being very responsible, helpful and cheerful; Sean Hunt for teaching me some interesting math, sharing his knowledge on XPS and electrocatalysis and providing the Chinese candy and other snack supplies; Karthik Narsimhan for helping me with using the portable MS and coming to school from home to open the lab door for me after midnight; Manish Shetty and Karthick Murugappan for always trying to accommodating me in their busy GCMS schedule and many fun outside the lab; Aaron Garg for taking care of the glovebox; Eric Anderson for his creative way of taking care of the Parr reactors; Daniel Consoli for lending me his ID to open the lab door; Hoyoung Park for sharing his new interesting results with me; Mike Orella for bringing us

another refrigerator for sample storage; Kim Dinh for taking on many responsibilities in the group; Zhenshu Wang for being another person that I can talk with in Chinese in the lab; Caroline Odermatt for her friendship; Rong Liu for many fun outside lab; Marco Wonink for bringing fun to the lab, and many others who were great colleagues. I am so grateful to work with all the group members in Román Group. The group atmosphere has been positive, friendly, and collaborative. Everyone in the group has been so generous to teach me new things and so wonderful to inspire me. I have been very fortunate to have all of you in my life, who have made the Ph.D. experience a very enjoyable one even during the most difficult times. I would also like to thank all my friends in the ChemE Department and at MIT for their support and friendship. Particularly, I would like to thank Dr. Nicole Yang for being my mentor during my first year in ChemE, Dr. Rong Yang for her constant support and mentoring throughout my doctoral studies, Dr. Caleb Class for his help on Gaussian calculations, Dr. Monica Sircar for her advice and support, and Shengchang Tang for helping me with auto column separations. I would also like to acknowledge all the administrative staff in the ChemE Department, who have made my transition into a new country smooth and have been super helpful and supportive throughout my Ph.D. life at MIT. Last but not least, I must express my deep gratitude to my parents for being always supportive and giving me the courage to face all the challenges. I would also like to thank my sisters and brother for their care and support. This completion of my thesis work would not have been possible without them.

Table of Contents

Abstract 3

Acknowledgements 5

Table of Contents 7

1. Introduction 9

2. Insights into the stability of gold nanoparticles supported on metal 21 oxides for the base-free oxidation of glucose to gluconic acid

3. Acid-catalyzed oxidation of levulinate derivatives to succinates under 41 mild conditions

4. Synthesis of itaconic acid ester analogs via self-aldol condensation of 59 ethyl pyruvate catalyzed by hafnium BEA zeolites

5. Conclusions and outlook 79

Appendix 89

Chapter 1

Introduction

1. Biomass as an alternative renewable carbon source

The heavy dependence on fossil fuel resources and the ever growing plastic waste pollution are two grand challenges we are faced with towards a sustainable future. In the past century, the global economy has been heavily dependent on fossil fuel resources, including oil, coal and natural gas,1,2 which have provided the material and energy foundation for modern society. In addition to the economy prosperity, heavy dependence on fossil fuel resources has also caused significant problems. Fossil fuel production and consumption not only are the main sources of green-house

gases (e.g., CO2 and CH4) and leading contributors to climate change, but also cause significant pollutions to air, water, and soil, which in turn impose a high cost on human health.2 The dwindling supply of fossil fuels, the disturbing economic impacts of fossil fuel price fluctuations and the threatening effects of climate change have stimulated increasing interests and efforts in developing clean and sustainable alternatives. Among the renewable alternative energy resources such as solar, wind and wave power, biomass is currently the only source that provides organic carbon to serve as a renewable feedstock to the chemical industry. Compared to petroleum- derived molecules, most biomass-derived molecules are highly functional. By selectively modifying the functionality via depolymerization, hydrolysis, dehydration, deoxygenation and oxidation, etc., these molecules can be integrated into the current petroleum industry or converted into molecules with new functions.

In addition to alleviating the societal dependence on fossil fuel sources, biomass holds the promise to reduce plastics pollution. Plastics are ubiquitous in modern life. Among municipal waste in Europe and the United States (U.S.), 12-25% are plastics, only 1-2% of which are recycled.3 The estimated amount of plastic waste entered the ocean in 2010 is 4.8 to 12.7 million metric tons, and the prediction for 2025 is one order of magnitude higher.4 Since most of the plastic waste is chemically stable and not biodegradable, the exact properties contributing to the pervasiveness of plastic products, the waste can persist in the environment for hundreds of years, taking up more landfill space and threatening the ecosystem.3,5 An increasing number of studies focus on the impact of the plastic waste on the environment, particularly the effects of

Chapter 1 microplastic debris (smaller than 5 mm plastic waste) on wildlife.3,5 On the other hand, biodegradable and biocompatible polymeric materials, which have been of great interest for biomedical applications, can offer a feasible alternative,6 and there have been growing efforts to design and synthesize biodegradable polymers.6-9 Natural polymers, such as polysaccharides and proteins, and synthetic polymers containing bonds that are hydrolytically and/or enzymatically sensitive in their backbones are (potentially) biodegradable.7,8 For instance, poly(anhydrides), poly(esters) and poly(amides) are promising candidates for biodegradable polymers, as they contain hydrolyzable bonds.6,10 Biomass can act as a renewable source of not only natural polymers (e.g., polysaccharides) but also monomers of synthetic biodegradable polymers (e.g., lactones, polyols, dicarboxylic acids, amino acids, etc.).8,9

In conclusion, biomass can play a crucial role in solving these two grand challenges and serve as the foundation for a sustainable future.

2. Chemocatalytic synthesis of mono- and dicarboxylic acids and esters from biomass-derived molecules

In biorefinery, biomass can be transformed into a wide range of compounds for fuels and chemicals.11 Among all the available biomass, including starch, fats and vegetable oils, lignocellulosic biomass is the most abundant. Particularly, lignocellulosic biomass does not compete with food for land use and is cheap as it is normally produced as waste from agriculture and paper industry.12 Therefore, the valorization of lignocellulose has attracted considerable attention.8,11,13

Lignocellulose mainly consists of 35–50% cellulose, 20–35% hemicellulose and 10–25% lignin,8 among which cellulose is a crystalline polymer with glucose units connected via β-1,4-glycosidic bonds.8,11,13 In principle, hydrolysis of cellulose into glucose is the first step of cellulose valorization. A variety of platform molecules can be produced via further transformation of glucose. Scheme 1 shows the top value-added chemicals from biomass identified by the U.S. Department of Energy.14 Depending on the reaction conditions and catalysts, glucose can be then chemocatalytically converted into different compounds, e.g., sorbitol via hydrogenation, gluconic acid via oxidation, isomerization into fructose and then dehydration into 5-hydroxymethylfurfural, which can be further converted into levulinic acid and formic acid via hydrolysis; alternatively,

Introduction glucose, as the feedstock for many biocatalysts, can also be enzymatically transformed into valuable products, such as itaconic acid, succinic acid, and 3-hydroxylproponic acid, etc.

O OH HO O Fumaric Acid O OH O OH O HO amination OH OH HO O HO Malic Acid O NH2 O O Succinic Acid Aspartic Acid O HO OH HO O NH2 fermentation 3-Hydroxypropanoic acid O OH HO fermentation Levulinic acid O Aspartic Acid fermentation hydration OH Glucose O O O dehydration O OH O OH O oxidation O OH HO OH HO HO OH fermentation 5-Hydroxymethylfurfural 2,5-Furan dicarboxylic acid O O OH Itaconic acid oxidation

hydrogenation OH OH fermentation OH HO fermentation & OH OH O O O oxidation oxidation OH OH Gluconic Acid HO OH OH HO hy O O dr O OH OH NH2 OH OH ogenol OH Glutamic Acid Sorbitol y HO sis HO OH OH O 3-Hydroxybutyrolactone Glucaric Acid OH HO HO OH OH Glycerol Ethylene glycol

Scheme 1. Top value-added chemicals from glucose identified by the U.S. Department of Energy.14 Due to the polymeric nature of lignocellulose, it is imperative to break it down to smaller fragments to allow further transformation into various chemicals and fuels. Tremendous efforts have been made and various strategies are being developed, such as gasification of biomass into syngas, biomass pyrolysis to bio-oil, and depolymerization of cellulose to glucose, etc.13,15,16 More progress in dissembling biomass is expected, yet various smaller molecules have been derived from

13,15-17 lignocellulose, such as CO, H2, furans, phenols, and sugars, etc. Since the structures of many of these smaller molecules are rather different from the hydrocarbon compounds in fossil fuel resources, new catalytic systems have to be developed to further upgrade these molecules.17

Monocarboxylic acids and dicarboxylic acids (diacids) play a central role in the bio-based

Chapter 1 chemicals portfolio, as evidenced by their prevalence in the top platform molecules from biomass identified by the U.S. Department of Energy in Scheme 1.14 Carboxylic acids, particularly diacids, are building blocks in condensation polymerization reactions,18 and their ester forms serve in many applications as lubricants, plasticizers, and polymer intermediates.19 Therefore, the focus of this thesis is chemocatalytic synthesis of mono- and dicarboxylic acids and esters from biomass- derived molecules. The target molecules include gluconic acid, succinic acid, itaconic acid and their ester derivatives.

2.1. Aerobic oxidation of glucose into gluconic acid

As the most abundant monosaccharide, developing reactions to transform glucose is of critical importance in the overall biorefinery scheme.14 Aldehyde oxidation of glucose results in gluconic acid, which has a large worldwide consumption, mainly as a complexing or acidifying agent in the pharmaceutical and food industry as well as a monomer for biodegradable polymer synthesis.20 Glucaric acid, the product of oxidizing the aldehyde and primary alcohol groups of glucose, can serve as a starting point for the production of a wide range of chemicals with applicability in high volume markets such as chelating agents, surfactants and carbohydrate polymer building blocks. Moreover, development of efficient processes to convert glucose into gluconic acid and glucaric acid will also open up the possibility to utilize other inexpensive sugars, such as xylose or arabinose.14

Water is commonly used as the solvent for glucose oxidation due to the high solubility of glucose and its oxidation products in water and the benign nature of water as a preferred solvent

21 21 in production. In contrast to strong oxidants, such as HNO3 and NaIO4, air or O2 as oxidant for glucose oxidation only produces water as by-product, which is non-toxic and does not negatively impact the separation of the desired products. Therefore, it is desired to achieve aerobic oxidation of glucose in aqueous environment. In the context of catalytic conversion of biomass, glucose can be obtained via acid-catalyzed hydrolysis of cellulose.13 Further conversion of glucose in a base-free environment will be beneficial, as it eliminates pH adjustment between steps and all the associated costs.

Supported noble metals are well studied as the catalysts for aerobic oxidation of alcohols and aldehydes.22-26 In particular, supported nanoparticles of platinum (Pt), palladium (Pd), rhodium

Introduction (Rh), gold (Au) and their alloys have been shown active and selective for glucose oxidation into gluconic acid and even glucaric acid.27-36 The first step of aldehyde oxidation in water is the formation of a geminal diol, and then one of the hydroxyl group is deprotonated, followed by C- H scission, similar to those in alcohol oxidation.37,38 Most of the previous studies are carried out under alkaline conditions, as alkaline environment can promote the deprotonation of hydroxyl group, rendering C-H scission rate-limiting.37,39 Additionally, acid products are neutralized into the salt form with increased solubility in water. Consequently, the inhibition effects of acid products on the active metal sites can be mitigated under alkaline conditions. On the other hand, Zope et al. have shown that ketone intermediates and condensation products of ketones formed under alkaline oxidation conditions can severely deactivate the supported noble metal catalysts.26 Since common biomass-derived compounds, such as sugars, sugar alcohols and other oxygenates, can be converted into ketones under oxidation conditions and further transformed into condensation products in the presence of bases, alkaline oxidation conditions can be potentially problematic with biomass-derived feed streams.26 Therefore, an active and stable catalyst under base-free conditions is essential for oxidation processes that have high tolerance of impurities from biomass-derived feed streams.

Supported Au nanoparticles has been shown to be active for oxidation under base-free conditions.27,28 Furthermore, Au is also less prone to over oxidation than Pt and Pd.27 Therefore,

Au seems more promising as the active metal under base-free conditions and under high O2 pressure. The main challenge lies in the stability of supported Au catalysts under base-free conditions and even acidic conditions.

2.2. Baeyer-Villiger oxidation of levulinates into succinates

Levulinic acid is another compound that can be readily produced from cellulose, and significant progress has been made in recent years.8,16,40-42 Biofine Renewables, LLC. has developed a continuous process to convert biomass into levulinic acid. The earlier estimated price of Biofine levulinic acid upon scale-up is $0.10-0.15/kg.43 Recent estimation of the commercial price by Biofine is around $1.0/kg, still much lower than the current commercial price $3.5/kg.43 Therefore, new reaction schemes utilizing levulinic acid should be developed to take advantage of this new raw material.

Chapter 1

Scheme 2. A complementary reaction scheme for Baeyer-Villiger oxidation of levulinic acid with H2O2. The presence of a ketone group in the molecular structure of levulinic acid provides different potential reaction pathways to transform levulinic acid. Baeyer-Villiger (BV) oxidation of levulinic acid can lead to valuable compounds as shown in Scheme 2. The BV oxidation is generally accepted as a two-step reaction, where a peroxide is added to a C=O group via nucleophilic attack forming a Criegee intermediate, followed by migration of alkyl group or H adjacent to the C=O group and effectively inserting an oxygen into the C-C bond or C-H bond. As shown in Scheme 2, oxygen insertion between the δ-carbon and the γ-carbon followed by hydrolysis can produce succinic acid and methanol. Alternatively, when the oxygen insertion occurs between the γ-carbon and the β-carbon, 3-acetoxypropanoic acid is generated, which can readily hydrolyze into acetic acid and 3-hydroxypropropanoic acid. Both succinic acid and 3- hydroxypropanoic acid are listed in the top 12 building block molecules identified by the U.S. Department of Energy as shown in Scheme 1.14 Succinic acid can be used to synthesize biodegradable polymers, detergents, food additives, and pharmaceutical products, etc.44 In addition, it can also serve as the starting molecule to synthesize 1,4-butanediol, tetrahydrofuran, and gamma-butyrolactones, etc.14 Similarly, 3-hydroxypropanoic acid is a promising monomer for biodegradable polymer synthesis.14,45 Its dehydration can produce acrylic acid and its derivatives that represent a large market.14 It can also be used to produce other useful chemicals such as 1,3- propanediol.14 Yet the current available technologies to produce them mainly rely on biological pathways and suffer from the common limitations of biological production, such as delicate pH control, difficult separation of products from the fermentation broth, deactivation of enzymes, etc.14,46-50 Therefore, chemocatalytic conversion of levulinic acid can provide a valuable alternative

Introduction to produce these important chemicals.

Challenges for BV oxidation of levulinic acid are two-fold. First, to design an active catalytic system for BV oxidation of levulinic acid is challenging, since levulinic acid is a linear ketone with a carboxylic acid group and it is generally more difficult to convert linear ketones via BV oxidation, partially due to the lack of ring strain.51 Secondly, the oxygen can be inserted in either side of the C=O group, then to control the location of oxygen insertion is the key to tune the reaction selectivity.

2.3. Aldol condensation of ethyl pyruvate into itaconic acid ester analogs

Most of the current research focuses on the synthesis of a second carboxylic acid or both carboxylic acid groups via oxidation of the starting compounds, whereas many products derived from biomass contain one carboxylic acid group, such as acetic acid from bio-oil (obtained from biomass fast pyrolysis),52 pyruvic acid from biological lignin upgrading,53 and crotonic acid from the catalytic pyrolysis of the biopolymer poly(3-hydroxybutyrate).54 Then a dicarboxylic acid molecule can be generated by inheriting the carboxylic acid group from the reactant molecules via C-C coupling. Indeed, Flanagan et al. have shown the promise to produce dicarboxylic acid esters (diesters) via addition dimerization of crotonates.54

Among all the carboxylic acid molecules available from biomass, keto acids and esters are the intermediates in a myriad of metabolic pathways, and thus render biocatalysis exceptionally well suited to generate high titers of keto acids. Meanwhile, chemocatalysis is ideal for aldol condensation at scale. Therefore, the C-C coupling of keto acids or esters is a promising renewable route to generate a wide range of diacids or diesters using coupled bio- and chemocatalytic processes. In contrast to purely biological or chemical pathways, approaches that combine these routes better integrate biomass into the current paradigm of fuels and chemicals production. A large variety of diacids and diesters with different backbones can be produced via aldol condensation of keto acids and esters, respectively, with several examples shown in Scheme 3. The key challenge in catalyst design is to convert biosynthetic molecules efficiently and selectively, and to even handle the feed stream from biosynthetic processes directly.

Chapter 1

Scheme 3. Proposed production of diacids or diesters via C-C coupling of keto acids or esters, R = H, alkyl group.

3. Aim of the thesis

The primary goal of this thesis is to develop effective catalytic systems that convert biomass- derived molecules into mono- and dicarboxylic acids and esters, and to understand the factors governing the activity, selectivity and stability of the catalysts in the systems. Based on the different molecular characteristics of the reactants derived from biomass, different catalytic systems were devised and studied. The systems developed and the insights obtained will further enable a sustainable future built upon renewable biomass resources and biodegradable polymers.

4. Outline of the thesis

Chapter 2 focuses on the development of an active and stable catalytic system for the aerobic base-free oxidation of glucose. Special attention was paid on support materials, an important element of the catalyst, in the presence of acidic products and used in redox reactions.22,25,55 Au catalysts with different supports were synthesized and compared. Then the performance of supported Au nanoparticles under uncontrolled pH and controlled acidic conditions were evaluated. A detailed characterization of the catalysts was conducted, which, coupled with the reactivity study, shed light on the key deactivation mechanisms of the catalyst. With insights into the stability of these catalysts, a new approach to stabilize supported Au catalysts was proposed and proven to be effective.

Introduction Chapter 3 discusses the conversion of levulinates. No BV oxidation of levulinic acid occurs in the presence of strong Brønsted acids at a catalytic amount in water,56 and there was a lack of systematic understanding on the causes of the zero reactivity. Instead, in order to understand the factors affecting the reactivity, BV oxidation reactions of methyl levulinate with methanol and heptane as solvents were investigated. A series of Brønsted acids and Lewis acidic metal triflates were tested for the BV oxidation of methyl levulinate. In addition to the strength of Brønsted acids and the hydrolysis properties of the metal triflates, the effects of other reaction conditions on the reaction selectivity were also investigated, including the solvent choice, the oxidants, and the carbon backbones of the molecules. Key parameters governing the reaction selectivity were elucidated.

Chapter 4 explores the synthesis of diacids and diesters via aldol condensation of keto acids and esters with a focus on the synthesis of itaconic acid ester analogs from ethyl pyruvate. Compared to the traditional solid alkaline catalysts for aldol condensation, Lewis acidic zeolites with BEA topology containing framework heteroatoms were demonstrated to be active, selective and stable catalysts, among which Hf-BEA showed the highest selectivity at high conversion. The nature of the active sites involved in the reactions was further unraveled through analyses of the dynamic behavior of Hf-BEA under flow conditions and studies with Na-exchanged zeolites.

Concluding remarks and outlooks are addressed in Chapter 5.

Each chapter of this thesis was written based on one or more publications and can be read independently. Accordingly, some overlap between the chapter introductions occurs.

References

(1) Wigley, T. M.; Richels, R.; Edmonds, J. A. Nature 1996, 379, 240-243. (2) Payne, S.; Dutzik, T.; Figdor, E. “The High Cost of Fossil Fuels,” Environment America Research & Policy Center, 2009. http://www.environmentamerica.org/sites/environment/files/reports/The-High-Cost- of-Fossil-Fuels.pdf. (3) Huerta Lwanga, E.; Gertsen, H.; Gooren, H.; Peters, P.; Salanki,́ T. s.; van der Ploeg, M.; Besseling, E.; Koelmans, A. A.; Geissen, V. Environ. Sci. Technol. 2016, 50, 2685-2691. (4) Jambeck, J. R.; Geyer, R.; Wilcox, C.; Siegler, T. R.; Perryman, M.; Andrady, A.; Narayan, R.; Law, K. L. Science 2015, 347, 768-771.

Chapter 1 (5) Sussarellu, R.; Suquet, M.; Thomas, Y.; Lambert, C.; Fabioux, C.; Pernet, M. E. J.; Le Goïc, N.; Quillien, V.; Mingant, C.; Epelboin, Y. Proc. Natl. Acad. Sci. U.S.A. 2016, 113, 2430-2435. (6) Nair, L. S.; Laurencin, C. T. Prog. Polym. Sci. 2007, 32, 762-798. (7) Tschan, M. J. L.; Brulé, E.; Haquette, P.; Thomas, C. M. Polym. Chem. 2012, 3, 836- 851. (8) Isikgor, F. H.; Becer, C. R. Polym. Chem. 2015, 6, 4497-4559. (9) Vilela, C.; Sousa, A. F.; Fonseca, A. C.; Serra, A. C.; Coelho, J. F.; Freire, C. S.; Silvestre, A. J. Polym. Chem. 2014, 5, 3119-3141. (10) Göpferich, A. Biomaterials 1996, 17, 103-114. (11) Yabushita, M.; Kobayashi, H.; Fukuoka, A. Appl. Catal., B 2014, 145, 1-9. (12) Caspeta, L.; Buijs, N. A.; Nielsen, J. Energy Environ. Sci. 2013, 6, 1077-1082. (13) Van de Vyver, S.; Geboers, J.; Jacobs, P. A.; Sels, B. F. ChemCatChem 2011, 3, 82-94. (14) Werpy, T.; Petersen, G.; Aden, A.; Bozell, J.; Holladay, J.; White, J.; Manheim, A.; Eliot, D.; Lasure, L.; Jones, S. “Top value added chemicals from biomass. Volume 1-Results of screening for potential candidates from sugars and synthesis gas,” Department of Energy: Washington DC, 2004. (15) Huber, G. W.; Iborra, S.; Corma, A. Chem. Rev. 2006, 106, 4044-4098. (16) Van de Vyver, S.; Thomas, J.; Geboers, J.; Keyzer, S.; Smet, M.; Dehaen, W.; Jacobs, P. A.; Sels, B. F. Energy Environ. Sci. 2011, 4, 3601-3610. (17) Rinaldi, R.; Schüth, F. Energy Environ. Sci. 2009, 2, 610-626. (18) Dicarboxylic Acids, Aliphatic. Ullmann's Encyclopedia of Industrial Chemistry [Online]; Wiley-VCH Verlag GmbH & Co. KGaA, Posted NOV 19, 2014. http://onlinelibrary.wiley.com/doi/10.1002/14356007.a08_523.pub3/pdf. (19) Dicarboxylic Acids. Kirk-Othmer Encyclopedia of Chemical Technology [Online]; John Wiley & Sons, Inc., Posted Sep 17, 2010. http://onlinelibrary.wiley.com/doi/10.1002/0471238961.0409030110150814.a01.pub2/p df. (20) Marcincinova-Benabdillah, K.; Boustta, M.; Coudane, J.; Vert, M. Biomacromolecules 2001, 2, 1279-1284. (21) Chatterjee, C.; Pong, F.; Sen, A. Green Chem. 2015, 17, 40-71. (22) Abad, A.; Corma, A.; García, H. Chem. Eur. J. 2008, 14, 212-222. (23) Corma, A.; Garcia, H. Chem. Soc. Rev. 2008, 37, 2096-2126. (24) Casanova, O.; Iborra, S.; Corma, A. J. Catal. 2009, 265, 109-116. (25) Casanova, O.; Iborra, S.; Corma, A. ChemSusChem 2009, 2, 1138-1144. (26) Zope, B. N.; Davis, R. J. Green Chem. 2011, 13, 3484-3491. (27) Biella, S.; Prati, L.; Rossi, M. J. Catal. 2002, 206, 242-247. (28) Önal, Y.; Schimpf, S.; Claus, P. J. Catal. 2004, 223, 122-133. (29) Baatz, C.; Thielecke, N.; Prüße, U. Appl. Catal., B 2007, 70, 653-660. (30) Thielecke, N.; Aytemir, M.; Prüsse, U. Catal. Today 2007, 121, 115-120. (31) Ishida, T.; Kinoshita, N.; Okatsu, H.; Akita, T.; Takei, T.; Haruta, M. Angew. Chem. 2008, 120, 9405-9408. (32) Yin, H.; Zhou, C.; Xu, C.; Liu, P.; Xu, X.; Ding, Y. J. Phys. Chem. C 2008, 112, 9673- 9678. (33) Prüße, U.; Herrmann, M.; Baatz, C.; Decker, N. Appl. Catal., A 2011, 406, 89-93.

Introduction (34) Delidovich, I. V.; Moroz, B. L.; Taran, O. P.; Gromov, N. V.; Pyrjaev, P. A.; Prosvirin, I. P.; Bukhtiyarov, V. I.; Parmon, V. N. Chem. Eng. J. 2013, 223, 921-931. (35) Ma, C.; Xue, W.; Li, J.; Xing, W.; Hao, Z. Green Chem. 2013, 15, 1035-1041. (36) Zhang, H.; Toshima, N. Catal. Sci. Technol. 2013, 3, 268-278. (37) Zope, B. N.; Hibbitts, D. D.; Neurock, M.; Davis, R. J. Science 2010, 330, 74-78. (38) Davis, S. E.; Zope, B. N.; Davis, R. J. Green Chem. 2012, 14, 143-147. (39) Shang, C.; Liu, Z.-P. J. Am. Chem. Soc. 2011, 133, 9938-9947. (40) Lin, H.; Strull, J.; Liu, Y.; Karmiol, Z.; Plank, K.; Miller, G.; Guo, Z.; Yang, L. Energy Environ. Sci. 2012, 5, 9773-9777. (41) Weingarten, R.; Conner, W. C.; Huber, G. W. Energy Environ. Sci. 2012, 5, 7559-7574. (42) Wettstein, S. G.; Alonso, D. M.; Chong, Y.; Dumesic, J. A. Energy Environ. Sci. 2012, 5, 8199-8203. (43) Luterbacher, J.; Alonso, D. M.; Dumesic, J. Green Chem. 2014, 16, 4816-4838. (44) Succinic Acid and Succinic Anhydride. Kirk-Othmer Encyclopedia of Chemical Technology [Online]; Posted April 14, 2006. http://onlinelibrary.wiley.com/doi/10.1002/0471238961.1921030306211301.a01.pub2/p df. (45) Zhang, D.; Hillmyer, M. A.; Tolman, W. B. Macromolecules 2004, 37, 8198-8200. (46) Song, H.; Lee, S. Y. Enzyme Microb. Technol. 2006, 39, 352-361. (47) Beauprez, J. J.; De Mey, M.; Soetaert, W. K. Process Biochem. 2010, 45, 1103-1114. (48) Cao, Y.; Cao, Y.; Lin, X. J. Ind. Microbiol. Biotechnol. 2011, 38, 649-656. (49) Chen, K. Q.; Li, J.; Ma, J. F.; Jiang, M.; Wei, P.; Liu, Z. M.; Ying, H. J. Bioresour. Technol. 2011, 102, 1704-1708. (50) Li, J.; Zheng, X. Y.; Fang, X. J.; Liu, S. W.; Chen, K. Q.; Jiang, M.; Wei, P.; Ouyang, P. K. Bioresour. Technol. 2011, 102, 6147-6152. (51) Brink, G.-J. t.; Arends, I. W. C. E.; Sheldon, R. A. Chem. Rev. 2004, 104, 4105 - 4123. (52) Crisci, A. J.; Dou, H.; Prasomsri, T.; Román-Leshkov, Y. ACS Catal. 2014, 4, 4196- 4200. (53) Johnson, C. W.; Beckham, G. T. Metab. Eng. 2015, 28, 240-247. (54) Flanagan, J. C. A.; Kang, E. J.; Strong, N. I.; Waymouth, R. M. ACS Catal. 2015, 5, 5328-5332. (55) Casanova, O.; Iborra, S.; Corma, A. ChemSusChem 2009, 2, 1138-1144. (56) Choudhary, H.; Nishimura, S.; Ebitani, K. Appl. Catal., A. 2013, 458, 55-62.

Chapter 2

Insights into the Stability of Gold Nanoparticles Supported on Metal Oxides for the Base-Free Oxidation of Glucose to Gluconic Acid

1. Introduction

Aldonic and aldaric acids have emerged as highly attractive chemical intermediates that can be

used for a variety of applications.1-3 One such compound, D-gluconic acid (100,000 tons per year), is widely used in the food, pharmaceutical, paper and concrete industries.4-6 Gluconic acid and its

salts are currently produced by the enzymatic oxidation of D-glucose by Aspergillus niger and Gluconobacter suboxydans.1 One of the impediments to the large-scale application of the fermentation processes is that they necessitate the neutralization of the acid in order to avoid the deactivation of the enzymes.7 This means that there is a need to develop heterogeneous catalysts

that can catalyze the oxidation of D-glucose under base-free conditions.

The benefits of using gold (Au) catalysts for the aerobic oxidation of carbohydrates have been well researched and documented in recent years.2,8 Key features include exceptionally high catalytic activities and selectivities for the oxidation of pentoses9,10 and hexoses,4,6,10-21 even at relatively low temperatures (40–65 °C). Also, it has been shown that Au can catalyze the oxidation of glucose without pH control, thus allowing the production of gluconic acid, not its salt, under slightly acidic conditions.12 It has been demonstrated that such conditions can enable the one-pot combination of acid-catalyzed hydrolysis reactions with Au-catalyzed glucose oxidation.22-24 This kind of catalytic cascade reaction is primarily important for the depolymerization and valorization of cellulose,7,25-27 although it might also be relevant to the catalytic processing of other di-, oligo-, and polysaccharides.

Most studies on the stability of Au catalysts so far have concerned the oxidation of glucose in the presence of a base. Initially, Biella et al. studied the recycling of Au on activated carbon (Au/AC) at pH values from 7.0 to 9.5.12 They found a decrease in activity of more than 50% after four consecutive runs, mainly due to leaching and sintering of the Au particles. In contrast, Au on

metal oxide catalysts are typically more stable under basic conditions. For example, Au/Al2O3 and

Au/TiO2 catalysts could be successfully reused for the oxidation of glucose and disaccharides at

6,10,28,29 pH 9. Thielecke et al. even demonstrated the long-term stability of Au/Al2O3 in a

Chapter 2 continuous-flow system.4,13 However, comparatively little is known about the stability of such catalysts under acidic conditions.

In this work, we report on the use of Au supported on nano- or microsized metal oxides. Au

on nanosized ceria (Au/nCeO2), for example, has recently been shown to be a relatively stable catalyst for the oxidative esterification of 5-hydroxymethyl-2-furfural into 2,5-dimethylfuroate under base-free conditions.30 The choice of support influences both the dispersion and the electronic state of Au nanoparticles (Au NPs),31 leading us to investigate a series of metal oxides including ceria, titania and zirconia. The immobilization of Au NPs on such supports can be achieved by the deposition–precipitation method.32 Avoiding agglomeration and sintering of the Au NPs during thermal activation is a great challenge since the catalytic activity markedly decreases for Au particles larger than 10 nm.33 It is argued that Au species are comparatively more stable when formed by the deposition–precipitation method than when they are prepared by e.g. anionic adsorption, leading to significantly smaller, and hence more active, Au NPs.32 Another factor that determines the size of Au particles is the metal/support interaction.11 It is understandable that a higher Fermi level of the support material can lead to a stronger electronic interaction with the Au NPs and in turn a reduced tendency for agglomeration.31

The aim of this study was to evaluate the catalytic performance and stability of Au NPs supported on various metal oxides for the base-free oxidation of glucose to gluconic acid. The key aspect that distinguishes our approach from most of the previous studies is that the pH of the

reaction solution was either unadjusted or lowered by the addition of H2SO4.

2. Experimental section

2.1. Preparation of Au catalysts

Nanosized ceria and zirconia, hereafter referred to as nCeO2 and nZrO2, were synthesized

30 according to a previously reported method. Briefly, an aqueous solution of Ce(NO3)3·6H2O or

ZrO(NO3)2·xH2O (0.8 M, 375 mL) was added to a solution of NH4OH in deionized water (0.8 M, 1.1 L). After stirring for 30 min, the solution was aged at 100 °C for 24 h in a polyethylene vessel. The mixture was cooled down, filtered and washed with an excess amount of deionized water. The resultant particles were dried in vacuum and calcined under an air flow of 100 mL min−1 at 400 °C for 4 h. The other support materials were purchased from Sigma Aldrich and

Base-Free Oxidation of Glucose to Gluconic Acid used as received.

Au was deposited on the metal oxides using the deposition–precipitation method. A solution of

HAuCl4·3H2O (175 mg) in deionized water (80 mL) was brought to pH 10 by adding a NaOH solution (0.2 M). This solution was mixed with a suspension containing the metal oxide (2 g) in deionized water (25 mL). After stirring for 18 h, the suspension was filtered and washed with an excess amount of deionized water until no more chlorine could be detected with the AgCl test. The supported Au catalyst was dried overnight at 80 °C and then reduced for 4 h at 225 °C in 5%

−1 H2 in N2 (100 mL min ). Calcination of metal oxide supported Au NPs were carried out in a tube furnace under air at 225, 325 and 425 oC after reduction at 225 oC.

The Au/AC catalyst was prepared according to the procedure reported by Biella et al.12 A

−1 solution of HAuCl4·3H2O (1 L, 100 μg mL ) was mixed with 2.5 g of a 2 wt% solution of

poly(vinyl alcohol) (PVA, MW ~ 10000). To this solution, 0.1 M NaBH4 solution (20 mL) was added dropwise, leading to the formation of metallic Au NPs. The PVA stabilized Au NPs were then immobilized on activated carbon (Darco®, 100 mesh) by adding 2 g of the support. The as- synthesized Au/AC catalyst was filtered after 2 h and washed with deionized water.

34 Cyanide leaching of Au/nCeO2 was adopted based on the work reported by Yang et al.

o Briefly, Au/nCeO2 was calcined at 400 C for 4 h, and then it was dispersed in 0.05 wt% NaCN solution (pH ~ 12) for 30 min under vigorously stirring. The leached sample was washed with deionized water and dried at room temperature under vacuum and tested within two days.

2.2. Characterization of the catalysts

Transmission electron microscopy (TEM) measurements were performed on either a JEOL TEM-200CX microscope operated at 120 kV or a JEOL TEM 2011 microscope operated at 200 kV. Samples were deposited on the TEM grids after ultrasonic dispersion in ethanol. The Au loading of the metal oxide catalysts was determined by inductively coupled plasma atomic emission spectrometry (ICP-AES) performed with an ICP analyzer (HORBIAJOBIN YVON). The Au loading of the Au/AC catalyst was analyzed by X-ray fluorescence with a Bruker Tracer III-SD Handheld X-Ray Fluorescence spectrometer. Nitrogen adsorption/desorption isotherms were recorded with a Quantachrome Autosorb iQ at liquid nitrogen temperature (−196 °C). Prior to the physisorption measurements, the samples were degassed for 2 h under vacuum at 400

Chapter 2 °C. X-ray photoelectron spectroscopy (XPS) was performed on a PHI Versaprobe II equipped with a monochromatic aluminium anode X-ray source and a dual-beam charge neutralization system with an electron neutralizer bias of 1.2 eV and an Ar ion beam energy of 10 eV. The C 1s peak was used as a reference and shifted to 284.8 eV for charge correction. Themogravimetric

−1 analysis (TGA) was performed by heating the catalyst under a flow of 10 mL min N2 and 90 mL min−1 air, using a Q500 thermal analysis system (TA Instruments). About 30–50 mg of the sample was heated at 2 °C min−1 up to 600 °C.

The dispersion of Au NPs was calculated as the ratio of the number of external Au atoms

(Nsurface Au) to the total number of Au atoms (Ntotal Au). The values of Nsurface Au and Ntotal Au were estimated based on the assumption that the Au NPs can be modelled as an fcc crystal lattice:35

= 10 − 20 + 12 (1)

() = (2)

= 1.105 × × (3)

With m equal to the numer of shells, Dm equal to the TEM-determined mean diameter of the Au

NPs, and Datom equal to the atomic diameter of Au (0.288 nm).

2.3. Oxidation reactions and product analysis

The catalytic reactions were carried out with solutions of glucose in deionized water (167 mmol L−1) at an initial pH of 4. In some experiments the initial pH was adjusted to 1.6 by the addition of small amounts of H2SO4. The reaction mixture was pressurized with O2 to 2.3 bar and stirred at 65 °C in a pressure reaction vessel (Andrews Glass Co.). The catalysts were recovered at the end of each run by centrifugation and washed with deionized water for at least 5 times until the supernatant reached the pH of deionized water. The alkaline wash was done with a 0.1 M NaOH solution. In both cases, the washed catalysts were dried at 80 °C prior to reuse. Regeneration of

the Au/μCeO2 catalyst was performed by calcination for 4 h at the temperatures specified in

−1 −1 Section 3.4 with a flow of 10 mL min N2 and 90 mL min air.

Liquid samples were syringe filtered (0.2 μm PTFE membrane) and analyzed by an Agilent 1260 HPLC. The carbohydrates were separated on a Bio-Rad Aminex HPX-87C column at 80

Base-Free Oxidation of Glucose to Gluconic Acid °C and detected by an evaporative light scattering detector, using deionized water as the mobile phase at a flow rate of 0.6 mL min−1. In parallel, organic acids were separated on a Bio-Rad Aminex HPX-87H column at 30 °C and detected by a UV-Vis detector at 210 nm, using 8 mM

−1 H2SO4 in deionized water as the mobile phase at a flow rate of 0.6 mL min . Dried aliquots of the samples were trimethylsilylated according to a previously reported protocol,36,37 and analyzed by GC–MS. The conversion of glucose and the selectivity to gluconic acid were calculated on a molar basis.

3. Results and discussion

3.1. Catalyst screening

A series of Au supported on reducible metal oxides was prepared, characterized and tested for the base-free oxidation of glucose to gluconic acid. Table 1 shows the Au loading, the mean particle size and dispersion of the Au NPs, as well as typical catalytic results. Prefixes n and μ refer to nano- and microsized metal oxide supports with particle sizes smaller than 30 nm and larger than 100 nm, respectively. All of the Au NPs supported on metal oxides were highly active for glucose oxidation under base-free conditions, with conversions of glucose ranging between 24% and 91% after 2 h reaction. For comparison, a benchmark Au/AC catalyst with 0.6 wt% Au loading showed only 7% conversion. As for the selectivity to gluconic acid, the values in Table 1 agree well with previously reported data for Au-based catalysts under acidic conditions.12 A GC–MS

analysis after trimethylsilylation of the product mixture showed the presence of glycolic acid, C5 aldonic acids and the δ-lactone of gluconic acid as the most important by-products in this reaction.

Ishida et al. previously reported that the size of Au particles, and hence their catalytic activity, depend strongly on the preparation method.11 The deposition–precipitation method used in this study gave Au NPs with mean particle sizes ranging from 1.7 nm to 3.0 nm (fourth column in

Table 1). The most promising results overall were obtained for TiO2 and CeO2 — the latter being known for its capability to store oxygen and to become reduced.38-41 There is, however, no correlation in Table 1 between the size of the Au NPs and the catalytic activity.

The formation of gluconic acid induced a significant pH decrease of the reaction solution. For the experiments under unadjusted pH, typical pH values measured at room temperature ranged from 4.0 to 2.5 before and after reaction, respectively. To evaluate the performance of the

Chapter 2 catalysts under acidic conditions, parallel experiments were performed at an initial pH of 1.6 by

the addition of small amounts of H2SO4 (last two columns in Table 1). Lowering the pH caused a decrease of the reaction rate, which is in agreement with the study by Biella et al.12 The inhibiting effect of the acidic conditions is reflected in lower glucose conversions; however, the impact on the catalytic activity is relatively small for Au on nanosized metal oxides. The selectivity to gluconic acid remained ≥88% regardless of the pH condition used. Entries 1–2 and 5–6 in Table 1

show the effect of Au loading on the catalytic activity of Au/nCeO2 and Au/μCeO2. When decreasing the Au loading of these catalysts, the conversion of glucose remained approximately constant for Au/nCeO2; however, for Au/μCeO2 under acidic conditions, the conversion increased from 37% to 59% upon decreasing the loading from 1.1% to 0.6%. Hence, lowering the

Au loading seems to increase the activity of Au/μCeO2 under acidic conditions. This result prompted us to study the stability of both catalysts under the given conditions.

Table 1. Base-free catalytic oxidation of glucose with Au NPs supported on different metal oxidesa

Unadjusted pH Initial pH of 1.6 Au loadingb Entry Catalyst support Dc (nm) Dispersiond Conv. (%) Select. (%) Conv. (%) Select. (%) (wt%)

1 nCeO2 0.5 1.8 0.60 74 95 67 96

2 nCeO2 1.4 2.5 0.47 75 99 75 96

3 nZrO2 1.6 1.7 0.62 89 99 86 97

4 nTiO2 1.3 3.0 0.40 91 99 81 98

5 μCeO2 0.6 2.0 0.55 76 96 59 90

6 μCeO2 1.1 2.9 0.41 77 98 37 90

7 μZrO2 0.8 2.1 0.53 65 98 24 88

8 μTiO2 1.2 2.0 0.55 89 98 54 >99

a -1 Reaction conditions: glucose 167 mmol L , 12 mL, pO2 = 2.3 bar, 65 °C, glucose/Au = 140, unadjusted pH, 2 h. b Au loading determined by ICP-AES. c Mean Au particle diameter determined by TEM analysis and based on a count of at least 100 Au NPs. d Dispersion of the Au NPs calculated assuming a predominantly cuboctahedral structure of the Au NPs (see the Experimental Section).

3.2. Comparison of the stability of Au/μCeO2 and Au/nCeO2

Recycling and reuse experiments provided further information on the stability of the Au catalysts.

Figure 1 compares the evolution of the activity of 0.6 wt% Au/μCeO2 and 0.5 wt% Au/nCeO2 over consecutive reaction cycles. The catalysts were filtered, washed with deionized water and dried after each run.

Base-Free Oxidation of Glucose to Gluconic Acid

a 100 90 0.6 wt% Au/μCeO2 80 0.5 wt% Au/nCeO2 70 After NaOH wash 60 50 40 30

Glucose conversion (%) Glucose conversion 20 10 0 Run 1 Run 2 Run 3 Run 4 Run 5

b 100 90 0.6 wt% Au/μCeO2 After NaOH 80 0.5 wt% Au/nCeO2 wash 70 60 50 40 30

Glucose conversion (%) Glucose conversion 20 10 0 Run 1 Run 2 Run 3 Run 4

Figure 1. Conversion of glucose over Au/μCeO2 and Au/nCeO2 measured after 4 h reaction (a) under unadjusted pH and (b) at an initial pH of 1.6. The catalysts were washed with NaOH before the last -1 reaction cycle. Reaction conditions: glucose 167 mmol L , 12 mL, pO2 = 2.3 bar, 65 °C, glucose/Au = 140. The experiments of Figure 1(a) were carried without pH adjustment of the reaction solution. It

can be seen that by the fourth cycle, the conversion of glucose over Au/μCeO2 decreased from

87% to 33%. Au/nCeO2 was less prone to deactivation, as in this case the conversion decreased from 87% to 55%. Our initial strategy to recover the catalytic activity was based on a study by Abad et al.,35 who showed that a NaOH wash can remove strongly adsorbed carboxylic acid species. This method was applied to the catalysts recovered after the fourth reaction cycle. In our case, however, the NaOH wash did not result in significant changes of the catalytic activity.

Instead, we suspected that one of the major causes of the deactivation of Au/μCeO2 was Au leaching. ICP-AES analysis showed that this catalyst had lost about 55% of its original Au loading during the oxidation reaction. Notably, the Au/nCeO2 catalyst had lost less than 2% of its Au loading after five cycles.

Consistent with the approach in Table 1, we also investigated the effect of lowering the pH of

Chapter 2

the reaction solution. Figure 1(b) shows that the catalytic activity of 0.6 wt% Au/μCeO2 decreased drastically under acidic conditions compared to that of the unadjusted pH conditions.

Specifically, the conversion of glucose over 0.6 wt% Au/μCeO2 decreased from 69% (run 1) to 12%

(run 3) when adding H2SO4 to the reaction solution. In contrast, only a slight decrease in catalytic

activity was observed for 0.5 wt% Au/nCeO2. ICP-AES analysis showed no significant Au

leaching for Au/nCeO2, while Au/μCeO2 lost 74% of its Au loading after four reaction cycles. As such, the distinct stability properties of the two catalysts required additional studies.

3.3. Effect of the Au particle surface density on the catalyst stability

Au leaching cannot fully account for the observed decrease in activity, and hence we investigated the sensitivity of the Au NPs to sintering to gain further insight into the deactivation mechanisms.

Figure 2 shows TEM images of the 0.6 wt% Au/μCeO2 and 0.5 wt% Au/nCeO2 catalysts prior and after reaction, while Figure 3 displays the evolution of the Au particle size distribution during several reaction cycles. The TEM-based investigation reveals that the textural properties of the metal oxide support, such as its size and surface area, are key factors in determining the size and stability of the Au NPs.42

The histograms in Figure 3 indicate relatively narrow Au particle size distributions of the as-

synthesized catalysts, irrespective of whether μCeO2 or nCeO2 is used as support. However, there is a vast difference in the evolution of the particle size distributions between the reused 0.6 wt%

Au/μCeO2 and 0.5 wt% Au/nCeO2 samples. The former is notably more sensitive to sintering, which is consistent with the fact that this catalyst showed higher deactivation rates (Figure 1).

Acidic conditions accelerate Au sintering for both Au/nCeO2 and Au/μCeO2; however, the shift

towards larger Au NP sizes is more pronounced for Au/μCeO2 than for Au/nCeO2. Indeed, the mean Au particle size for Au/μCeO2 increased from 2.0 nm to 7.1 nm after the last reaction cycle

(Figure 3c), whereas the mean particle size for Au/nCeO2 increased only from 1.3 nm to 1.9 nm

(Figure 3d). We hypothesized that the relatively stable nature of Au/nCeO2 is due to the low

surface density of the Au NPs. N2 physisorption showed that the Brunauer–Emmett–Teller (BET)

2 2 surface area of nCeO2 (100 m /g) is 25 times larger than the surface area of μCeO2 (4 m /g) (Figure 4). Since both catalysts have a similar Au loading, the surface density of the Au NPs is about 25 times higher for Au/μCeO2 than for Au/nCeO2. This seems to suggest that the interparticle distance determines the coalescence rate of the Au NPs and hence their tendency to

Base-Free Oxidation of Glucose to Gluconic Acid sinter by the particle migration mechanism.42-44 An alternative mechanism that sustains our experimental observations is based on a solution-mediated Ostwald ripening via Au dissolution and redeposition. Ostwald ripening could explain the observed correlation between the growth of the Au NPs and the extent of Au leaching of Au/μCeO2 and Au/nCeO2. The Au NPs are more susceptible to sintering during reactions at a lower pH, which, again, is in line with the general expectations given that the solubility of Au is higher under more acidic conditions.

Figure 2. Representative TEM images of the 0.6 wt% Au/μCeO2 (a,b) and 0.5 wt% Au/nCeO2 catalysts (c,d), before (left) and after use in glucose oxidation (right).

The size of the CeO2 support has important implications on the stability of the Au NPs. Given

its higher density of defect sites (i.e., mainly oxygen vacancies), nCeO2 is known to have more

30 anchoring sites than μCeO2. One prevalent view is that these anchoring sites can further stabilize the Au NPs against particle growth.32,41,45 In this study, a higher number of defect sites on the surface of nCeO2 probably helps to stabilize a higher number of Au NPs per unit surface area. In

the case of Au/μCeO2, we anticipated that reducing the surface density of Au NPs could improve their stability because of (i) a more preferred nucleation and growth of the Au NPs on the relatively scarce defect sites of the μCeO2 support during the deposition–precipitation process, and (ii) an increase in the interparticle spacing (vide supra). To validate our hypotheses, we synthesized a Au/μCeO2 catalyst with a Au loading as low as 0.02 wt%. Since the dispersion for the 0.02 wt% Au/μCeO2 (0.7) and that for the 0.5 wt% Au/μCeO2 (0.6) were virtually identical,

Chapter 2 the former catalyst evidently had a significantly lower surface density of Au NPs. The

performance of these catalysts and 0.4 wt% Au/nCeO2 was investigated for a longer reaction time (72 h instead of 4–16 h) and a higher glucose/Au ratio (9000 instead of 140). Figure 5 shows the turnover number (TON, defined as moles of glucose converted per mol surface Au) as a function of reaction time. The results are in good agreement with the observations made in Figure 1 that

0.4 wt% Au/nCeO2 is found to be more stable than 0.5 wt% Au/μCeO2. Although Au NPs supported on microsized metal oxides were initially as active as Au NPs supported on nanosized supports, their activity decreased due to sintering. Consistent with our previous findings, the

increased maximum TON of 9090 for 0.02 wt% Au/μCeO2 shows that the lower surface density of the Au NPs indeed leads to a significantly higher activity and catalyst stability. A similar phenomenon was reported by Abad et al. while studying the influence of Au loading on the

35 activity of Au/nCeO2 for the base-free oxidation of cinnamyl alcohol in toluene. They observed that the turnover frequency (TOF, defined as moles of substrate converted per mol surface Au per

hour) was more than three times higher for 0.44 wt% Au/nCeO2 than for 1.80 wt% Au/nCeO2, although both catalysts had almost the same particle size distribution.

a 0.6 wt% Au/µCeO2 (unadjusted pH) b 0.5 wt% Au/nCeO2 (unadjusted pH) 90 80 80 As-synthesized 70 As-synthesized After run 1 After run 1 70 60 60 After run 5 After run 5 50 50 40 40 30 30 Frequency (%) Frequency Frequency (%) Frequency 20 20 10 10 0 0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Particle size (nm) Particle size (nm)

c 0.6 wt% Au/µCeO2 (initial pH of 1.6) d 0.5 wt% Au/nCeO2 (initial pH of 1.6) 90 90 80 As-synthesized 80 As-synthesized 70 After run 1 70 After run 1 60 After run 4 60 After run 4 50 50 40 40 30 30 Frequency (%) Frequency Frequency (%) Frequency 20 20 10 10 0 0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Particle size (nm) Particle size (nm)

Figure 3. Au particle size distribution histograms of 0.6 wt% Au/μCeO2 (a,c) and 0.5 wt% Au/nCeO2 (b,d). Reaction conditions: see Figure 1.

Base-Free Oxidation of Glucose to Gluconic Acid

120

100

80 at STP) -1 g

3 60 (cm 40 adsorbed

V 20

0 0.0 0.2 0.4 0.6 0.8 1.0 Relative pressure Figure 4. N2 adsorption/desorption isotherms of μCeO2 (green circles) and nCeO2 (red triangles).

12000 0.02 wt% Au/μCeO2

) 10000 0.5 wt% Au/μCeO2 -1 0.4 wt% Au/nCeO2 8000 surface Au mol 6000 glucose 4000

TON (mol 2000

0 0 1020304050607080 Reaction time (h)

Figure 5. Turnover number for glucose oxidation with 0.02 wt% Au/μCeO2, 0.5 wt% Au/μCeO2 and -1 0.4 wt% Au/nCeO2, as a function of reaction time. Reaction conditions: glucose 167 mmol L , 12 mL,

pO2 = 2.3 bar, 65 °C, glucose/Au = 9000. 3.4. Effect of the Au particle size on the turnover frequency

An important question concerning the impact of sintering is how the TOF relates to the mean diameter of the Au NPs.46 Studies by Ishida et al. have previously demonstrated the size dependence of the TOF of Au supported on metal oxides for glucose oxidation at pH 9.11 They inferred that the reaction might occur at the interface between the Au NPs and the metal oxide support rather than at the surface of the Au particles. Comparable trends were observed for the catalytic oxidation reactions performed under our reaction conditions (Figure 6), which points to a more general nature of these findings. Au/nCeO2, however, showed a significant decrease in TOF even for samples with almost the same mean diameter of Au NPs. This suggests that the deactivation may be caused by factors other than leaching or sintering of the Au NPs, such as the adsorption of reactive species on the Au surface.

Chapter 2 400 0.6 wt% Au/μCeO2 350 1 (unadjusted pH) )

-1 0.5 wt% Au/nCeO2 h

-1 300 (unadjusted pH) 1 2 3 0.6 wt% Au/μCeO2 250 (initial pH of 1.6) surface Au 1 0.5 wt% Au/nCeO2 mol 200 (initial pH of 1.6) 1

glucose 150 2 3 100 3

TOF (mol 50 2 2 3 0 012345678910 Mean diameter of the Au NPs (nm) Figure 6. TOF per surface Au atom as a function of the mean diameter of the Au NPs. The TOFs were measured after 20 min reaction. Numbers 1–3 indicate data points corresponding to the as-synthesized catalyst, the catalyst recovered after the first cycle and the catalyst recovered after the last cycle in Figure 1, respectively. Reaction conditions: see Figure 1. 3.5. Inhibition of the catalytic activity by adsorption of reactive species and regeneration by calcination

Compelling evidence for the adsorption of reactive species came from TGA and XPS analysis of

the Au/μCeO2 catalyst before and after reaction, as well as after calcination at 225 °C and 325 °C. The XPS spectra in Figure 7 indicate that Au0 is the predominant oxidation state of the Au NPs on the as-synthesized catalysts. The Au 4f signal shows two characteristic peaks at 87.5 eV and

83.8 eV, corresponding to Au 4f5/2 and Au 4f7/2 transitions, respectively. After reaction with glucose, both peaks shifted to lower binding energies compared to the as-synthesized sample. The negative shift (−0.5 eV) is tentatively attributed to the electron transfer from intermediate species to Au0 on the surface of the Au NPs. This electronic perturbation can be anticipated from previous studies that showed a severe deactivation of Au NPs by the strong adsorption of ketone intermediates and condensation products of ketones.47 The adsorption of species on the Au catalyst was also confirmed by TGA. Spent catalyst samples were washed with deionized water, dried at 80 °C and heated from room temperature to 600 °C under air while monitoring the weight losses (Figure 8). The decomposition occurred at temperatures up to 325 °C — which is why we chose to calcine the spent catalysts at this temperature. An additional calcination treatment was performed at 225 °C to diminish the possible effect of sintering.

Base-Free Oxidation of Glucose to Gluconic Acid

Au 4f5/2 Au 4f7/2

(iv)

(iii)

(ii)

(i)

92 91 90 89 88 87 86 85 84 83 82 81 80 Binding energy (eV)

Figure 7. XPS Au 4f spectra of (i) as-synthesized 0.02 wt% Au/μCeO2, (ii) after 72 h reaction with glucose and washing with deionized water, (iii) after regeneration by calcination at 225 °C and (iv) after regeneration by calcination at 325 °C. The open circles show raw experimental data. The C 1s signal was shifted to 284.8 eV for charge correction.

As shown in the XPS spectra (Figure 7), the Au 4f5/2 and Au 4f7/2 peaks of the calcined samples gradually shifted back to the band energies of the as-synthesized catalyst. According to ICP-AES measurements, the recovered and regenerated catalysts ((ii), (iii) and (iv) in Figure 7) had almost the same Au loading as the as-synthesized catalyst (i). Because of the similar electronic state and loading of Au for the as-synthesized catalyst and the one regenerated by calcination at 325 °C, it was expected that both catalysts would show a comparable catalytic activity. Our experiments, however, still showed a significant decrease in catalytic performance. Specifically, the initial TOF

-1 -1 of the as-synthesized catalyst was 3100 molglucose molsurface Au h , whereas the calcined sample

-1 -1 showed an initial TOF of 760 molglucose molsurface Au h . A reaction promoted by the calcined catalyst gave a conversion of 8% after 12 h, while the as-synthesized catalyst led to a conversion of 43% under the same reaction conditions. Note that the catalyst calcined at 225 °C showed no significant catalytic activity.

Although part of the deactivation of Au/μCeO2 can be ascribed to the inhibition by reactive species, the severe loss of activity after calcination at 325 °C is remarkable. Currently, we hypothesize that the most likely deactivation pathway is caused by agglomeration of the Au NPs, as the catalyst calcined at 325 °C (5.7 ± 5.5 nm) showed significantly larger particle sizes than the as-synthesized catalyst (1.7 ± 0.6 nm) (Figure 9). Consequently, larger particles may feature drastically lower amounts of the most active sites on the surface of the Au NPs. A future challenge

Chapter 2 to be addressed is to design the supported Au catalysts that resist agglomeration under aqueous reaction conditions.

100

99.5

99 Weight (%)

98.5

98 0 100 200 300 400 500 600 Temperature (°C)

Figure 8. TGA curve of the 0.02 wt% Au/μCeO2 catalyst after 72 h reaction with glucose. Reaction conditions: see Figure 5 above.

Frequency (%) Frequency 30

0 0 5 10 15 20 25 30 Particle size (nm)

Figure 9. Au particle size distribution histograms of 0.02 wt% Au/μCeO2: as-synthesized (black circles), after 72 h reaction with glucose and washing with deionized water (green diamonds), after regeneration by calcination at 225 °C (red triangles), and after regeneration by calcination at 325 °C (blue squares). Particle sizes were 1.7 ± 0.6, 5.3 ± 3.2, 5.2 ± 3.8 and 5.7 ± 5.5, respectively. Reaction conditions: see Figure 5 above. As surface adsorbed species are important in catalyst deactivation, we studied glucose conversion in the presence of molecules containing targeted functional groups to understand the

Base-Free Oxidation of Glucose to Gluconic Acid impact of binding of intermediates and products on catalyst activity. Glucose oxidation reactions were studied using glucose feed solutions mixed with the following additives: gluconic acid, acetic acid (a single carboxylic acid group), levulinic acid (a ketone group and a carboxylic acid group), hydroxyacetone (a ketone group and a primary alcohol group), acetone (a ketone group), and furfural (a ketone group and a furan ring). The glucose/additive molar ratio was 10. As shown in Figure 10, no obvious inhibition was caused by the presence of hydroxyacetone, while a decrease of 28% and 54% in glucose conversion after 20 min was observed with addition of acetic acid and gluconic acid, respectively (see Figure 10). Since the presence of gluconic acid lowers the solution pH (2.9), another reaction using H2SO4 to adjust the solution pH to 2.9 was carried out as a control and a 9% decrease in glucose conversion after 20 min was observed. These data suggest that under base-free conditions, carboxylic acid groups contribute to lowering catalyst activity while primary hydroxyl and carbonyl groups do not, likely due to competitive binding to active sites.

Figure 10. Glucose conversion after 20 min for reactions with additives. Reaction conditions: glucose 167 -1 mmol L , 12 mL, pO2 = 2.3 bar, 65 °C, 0.4 wt% Au/nCeO2, glucose/Au = 410, glucose/additive = 10. Selectivity to gluconic acid above 95% for each run. Given that Au NPs supported on nanosized metal oxides exhibited high activity and stability under acidic conditions, attractive opportunities exist to pursue cascade reactions involving

Chapter 2 acid/redox sequences, for instance, by combining cellulose hydrolysis catalyzed by homogeneous acids with glucose oxidation. However, cellulose hydrolysis streams often contain many by- products other than monosaccharides (e.g., formic acid, levulinic acid and furan-derivatives). In view of the potential interaction of by-products with active sites (vide supra), we performed proof-of-principle experiments to verify the feasibility of a cascade reaction. Thus, catalyst performance was evaluated in the presence of levulinic acid and furfural with a glucose/additive ratio of 10. Both levulinic acid and furfural led to a drop in conversion of ~50% when compared to the control experiment with no additives (see Figure 11). Product analysis showed that while levulinic acid was not consumed during the reaction, furfural was oxidized into 2-furoic acid. Note that full glucose conversion could still be achieved albeit at a slower rate than the control case, thus pointing towards a reversible competitive adsorption on the active sites (see Figure 12). Of all additives, furfural exhibited the strongest catalyst inhibition.

Figure 11. Glucose conversion after 20 min with by-products from cellulose hydrolysis/oxidation as -1 additives. Reaction conditions: glucose 167 mmol L , 12 mL, pO2 = 2.3 bar, 65 °C, 0.5 wt% Au/nCeO2, glucose/Au = 328, glucose/additive = 10. Selectivity to gluconic acid is above 95% for each run.

Base-Free Oxidation of Glucose to Gluconic Acid 100 Control Levulinic acid Furfural 80 Gluconic acid

40 Glucose Conversion (%) Conversion Glucose 20

0 024681012141618 Time (h) Figure 12. Glucose conversion profiles with by-products from cellulose hydrolysis/oxidation as additives. -1 Reaction conditions: glucose 167 mmol L , 12 mL, pO2 = 2.3 bar, 65 °C, 0.5 wt% Au/nCeO2, glucose/Au = 328, glucose/additive = 10. Selectivity to gluconic acid is above 95% for each run. 3.6. Stabilization of supported Au catalysts

Among the causes of Au catalyst deactivation, the activity loss caused by inhibition of adsorption species can be regenerated by calcination; however, the deactivation caused by Au particle size increase seems irreversible. Therefore, we then focused on design and synthesis of Au catalysts that are more resistant to agglomeration. As shown in Appendix, we demonstrated 1) that the atomically dispersed Au atoms on different metal oxide supports are not stable for glucose oxidation in aqueous environment, even though they were shown to be active and stable catalysts for gas-phase reactions, e.g. water-gas shift reaction,48 2) the potential to stabilize the Au NPs supported on metal oxides via strong metal-support interaction.49-54

4. Conclusions

We have shown that Au nanoparticles supported on metal oxides are active and selective catalysts for the oxidation of glucose to gluconic acid under base-free conditions. The stability study offers some clues into how the irreversible deactivation of the catalysts occurs through leaching and hydrothermal sintering of the Au nanoparticles. An easily applicable approach for improving the catalyst's stability against sintering is lowering the Au loading on the metal oxides. Our results indicate that the surface density of the Au nanoparticles affects their tendency to agglomerate

Chapter 2 during the oxidation reaction. Under the applied conditions, the highest stability was found for a

0.02 wt% Au/μCeO2 catalyst prepared by the deposition–precipitation method. The reversible deactivation of this catalyst is ascribed to the adsorption of reactive species, which could be removed by calcination of the spent catalyst at 325 °C. Carboxylic acid groups in the reactive species contribute the most to the reversible deactivation, likely due to competitive binding to the active sites. Since the inhibition of Au catalysts caused by the by-products from acid-catalyzed cellulose hydrolysis is reversible, it is possible to couple glucose production from cellulose hydrolysis with glucose stabilization via oxidation into gluconic acid without an intense purification step or even in a one-pot reaction system. We also showed that atomically dispersed Au supported on metal oxides are not stable under liquid conditions and therefore are not suitable for glucose oxidation in aqueous environment. Nonetheless, it is promising to apply the strong metal-support interaction between Au and metal oxide supports under oxidation conditions to stabilize the Au nanoparticles.

Acknowledgements

Dr. Stijn Van de Vyver and Dr. Krishna K. Sharma are gratefully acknowledged for their major contributions to this chapter.

References

(1) Gluconic Acid. Ullmann's Encyclopedia of Industrial Chemistry [Online]; Wiley-VCH Verlag GmbH & Co. KGaA, Posted June 15, 2000. http://onlinelibrary.wiley.com/doi/10.1002/14356007.a12_449/pdf. (2) Kusema, B. T.; Murzin, D. Y. Catal. Sci. Technol. 2013, 3, 297-307. (3) Van de Vyver, S.; Roman-Leshkov, Y. Catal. Sci. Technol. 2013, 3, 1465-1479. (4) Thielecke, N.; Aytemir, M.; Prüsse, U. Catal. Today 2007, 121, 115-120. (5) Xavier, N. M.; Rauter, A. l. P.; Queneau, Y. Top. Curr. Chem. 2010, 295, 19-62. (6) Baatz, C.; Prüße, U. J. Catal. 2007, 249, 34-40. (7) Kobayashi, H.; Fukuoka, A. Green Chem. 2013, 15, 1740-1763. (8) Pina, C. D.; Falletta, E.; Rossi, M. Chem. Soc. Rev. 2012, 41, 350-369. (9) Kusema, B. T.; Campo, B. C.; Mäki-Arvela, P.; Salmi, T.; Murzin, D. Y. Appl. Catal., A 2010, 386, 101-108. (10) Mirescu, A.; Pruesse, U. Appl. Catal., B 2007, 70, 644-652. (11) Ishida, T.; Kinoshita, N.; Okatsu, H.; Akita, T.; Takei, T.; Haruta, M. Angew. Chem., Int. Ed. 2008, 47, 9265-9268. (12) Biella, S.; Prati, L.; Rossi, M. J. Catal. 2002, 206, 242-247. (13) Thielecke, N.; Vorlop, K.-D.; Prüße, U. Catal. Today 2007, 122, 266-269.

Base-Free Oxidation of Glucose to Gluconic Acid (14) Ishida, T.; Kuroda, K.; Kinoshita, N.; Minagawa, W.; Haruta, M. J. Colloid Interface Sci. 2008, 323, 105-111. (15) Matveeva, V.; Bykov, A.; Doluda, V.; Sulman, M.; Kumar, N.; Dzwigaj, S.; Marceau, E.; Kustov, L.; Tkachenko, O.; Sulman, E. Top. Catal. 2009, 52, 387-393. (16) Önal, Y.; Schimpf, S.; Claus, P. J. Catal. 2004, 223, 122-133. (17) Ma, C.; Xue, W.; Li, J.; Xing, W.; Hao, Z. Green Chem. 2013, 15, 1035-1041. (18) Yin, H.; Zhou, C.; Xu, C.; Liu, P.; Xu, X.; Ding, Y. J. Phys. Chem. C 2008, 112, 9673- 9678. (19) Okatsu, H.; Kinoshita, N.; Akita, T.; Ishida, T.; Haruta, M. Appl. Catal., B 2009, 369, 8- 14. (20) Delidovich, I. V.; Moroz, B. L.; Taran, O. P.; Gromov, N. V.; Pyrjaev, P. A.; Prosvirin, I. P.; Bukhtiyarov, V. I.; Parmon, V. N. Chem. Eng. J. 2013, 223, 921-931. (21) Zhang, H.; Toshima, N. Catal. Sci. Technol. 2013, 3, 268-278. (22) Tan, X.; Deng, W.; Liu, M.; Zhang, Q.; Wang, Y. Chem. Commun. 2009, 7179-7181. (23) An, D.; Ye, A.; Deng, W.; Zhang, Q.; Wang, Y. Chem. Eur. J. 2012, 18, 2938-2947. (24) Zhang, J.; Liu, X.; Hedhili, M. N.; Zhu, Y.; Han, Y. ChemCatChem 2011, 3, 1294- 1298. (25) Van de Vyver, S.; Geboers, J.; Jacobs, P. A.; Sels, B. F. ChemCatChem 2011, 3, 82-94. (26) Geboers, J. A.; Van de Vyver, S.; Ooms, R.; Op de Beeck, B.; Jacobs, P. A.; Sels, B. F. Catal. Sci. Technol. 2011, 1, 714-726. (27) Yabushita, M.; Kobayashi, H.; Fukuoka, A. Appl. Catal., B 2014, 145, 1-9. (28) Baatz, C.; Thielecke, N.; Prüße, U. Appl. Catal., B 2007, 70, 653-660. (29) Prüße, U.; Herrmann, M.; Baatz, C.; Decker, N. Appl. Catal., A 2011, 406, 89-93. (30) Casanova, O.; Iborra, S.; Corma, A. J. Catal. 2009, 265, 109-116. (31) Murdoch, M.; Waterhouse, G. I. N.; Nadeem, M. A.; Metson, J. B.; Keane, M. A.; Howe, R. F.; Llorca, J.; Idriss, H. Nat. Chem. 2011, 3, 489-492. (32) Corma, A.; Garcia, H. Chem. Soc. Rev. 2008, 37, 2096-2126. (33) Comotti, M.; Della Pina, C.; Matarrese, R.; Rossi, M. Angew. Chem., Int. Ed. 2004, 43, 5812-5815. (34) Yang, M.; Allard, L. F.; Flytzani-Stephanopoulos, M. J. Am. Chem. Soc. 2013, 135, 3768-3771. (35) Abad, A.; Corma, A.; Garcia, H. Chem-Eur J 2008, 14, 212-222. (36) Van de Vyver, S.; Helsen, S.; Geboers, J.; Yu, F.; Thomas, J.; Smet, M.; Dehaen, W.; Román-Leshkov, Y.; Hermans, I.; Sels, B. F. ACS Catal. 2012, 2, 2700-2704. (37) Van de Vyver, S.; Geboers, J.; Helsen, S.; Yu, F.; Thomas, J.; Smet, M.; Dehaen, W.; Sels, B. F. Chem. Commun. 2012, 48, 3497-3499. (38) Guan, Y.; Ligthart, D. M.; Pirgon-Galin, Ö.; Pieterse, J. A.; van Santen, R. A.; Hensen, E. J. Top. Catal. 2011, 54, 424-438. (39) Fu, Q.; Weber, A.; Flytzani-Stephanopoulos, M. Catal. Lett. 2001, 77, 87-95. (40) Lee, Y.; He, G.; Akey, A. J.; Si, R.; Flytzani-Stephanopoulos, M.; Herman, I. P. J. Am. Chem. Soc. 2011, 133, 12952-12955. (41) Si, R.; Flytzani-Stephanopoulos, M. Angew. Chem., Int. Ed. 2008, 47, 2884-2887. (42) Prieto, G.; Meeldijk, J. D.; de Jong, K. P.; de Jongh, P. E. J. Catal. 2013, 303, 31-40. (43) Prieto, G.; Zecevic, J.; Friedrich, H.; de Jong, K. P.; de Jongh, P. E. Nat. Mater. 2013, 12, 34-39. (44) Ono, L. K.; Roldan-Cuenya, B. Catal. Lett. 2007, 113, 86-94.

Chapter 2 (45) Ta, N.; Liu, J.; Chenna, S.; Crozier, P. A.; Li, Y.; Chen, A.; Shen, W. J. Am. Chem. Soc. 2012, 134, 20585-20588. (46) Corma, A.; Concepcion, P.; Boronat, M.; Sabater, M. J.; Navas, J.; Yacaman, M. J.; Larios, E.; Posadas, A.; Lopez-Quintela, M. A.; Buceta, D.; Mendoza, E.; Guilera, G.; Mayoral, A. Nat. Chem. 2013, 5, 775-781. (47) Zope, B. N.; Davis, R. J. Green Chem. 2011, 13, 3484-3491. (48) Flytzani-Stephanopoulos, M. Acc. Chem. Res. 2013, 47, 783-792. (49) Tauster, S.; Fung, S.; Baker, R.; Horsley, J. Science 1981, 211, 1121-1125. (50) Tauster, S. Acc. Chem. Res. 1987, 20, 389-394. (51) Mei, B.; Wiktor, C.; Turner, S.; Pougin, A.; van Tendeloo, G.; Fischer, R. A.; Muhler, M.; Strunk, J. ACS Catal. 2013, 3, 3041-3049. (52) Liu, X.; Liu, M.-H.; Luo, Y.-C.; Mou, C.-Y.; Lin, S. D.; Cheng, H.; Chen, J.-M.; Lee, J.-F.; Lin, T.-S. J. Am. Chem. Soc. 2012, 134, 10251-10258. (53) Kuwauchi, Y.; Yoshida, H.; Akita, T.; Haruta, M.; Takeda, S. Angew. Chem. Int. Ed. 2012, 51, 7729-7733. (54) Akita, T.; Tanaka, S.; Tanaka, K.; Haruta, M.; Kohyama, M. J. Mater. Sci. 2011, 46, 4384-4391.

Chapter 3

Acid-Catalyzed Oxidation of Levulinate Derivatives to Succinates under Mild Conditions

1. Introduction

The conversion of renewable sources into commodity and specialty chemicals through chemical or biological routes has attracted considerable attention.1-5 Thermocatalytic routes offer an attractive alternative to biological routes, with less stringent requirements for temperature and pH control, as well as potentially less energy-intensive product separation and purification processes.1,6 Succinic acid (SA) has been identified as one of the top 12 building blocks from biomass by the US Department of Energy.7 The market for SA is expected to exceed $1.1 billion in revenue by 2020.8 Succinates formed via esterification of SA with monoalcohols are important plasticizers, lubricants and intermediates.9,10 A thermocatalytic route to produce SA and succinates from renewable resources is therefore highly desirable.

Levulinic acid (LA) is a key platform molecule that can be readily produced from lignocellulosic carbohydrates.11-15 Upon scale up, the estimated LA price is expected to decrease to less than $1/kg from the current price of approximately $3.5/kg.16 The carbonyl group in LA and its esters can be oxidized to produce SA and its esters, respectively; however, linear aliphatic ketones are generally difficult to oxidize selectively. Consequently, only a few studies, mostly based on vanadium-, ruthenium- and manganese-based catalysts, have been reported on the oxidation of LA derivatives.17-20 These catalytic systems suffer from several challenges, including necessitating

high reaction temperatures, using toxic reagents, or releasing a stoichiometric amount of CO2. Recently, Podolean et al. demonstrated that Ru-based magnetic nanoparticles are efficient

2 catalysts for the oxidation of LA into SA under 10 bar O2 at 150 °C. The authors hypothesized that strong Brønsted acid sites are responsible for catalyzing the oxidation of the LA carbon backbone via a Baeyer-Villiger (BV) mechanism. Unfortunately, no systematic studies on the role of Brønsted acids in the BV oxidation of LA into SA were reported. This is particularly important given that Choudhary et al. showed that strong Brønsted acids cannot catalyze the oxidation of

5 LA into SA with H2O2 in water.

Here we investigate the oxidation of methyl levulinate into dimethyl succinate under mild

Chapter 3 conditions using Brønsted and Lewis acid catalysts. Given the low reactivity of LA in water,5 we preform our reactivity studies in solvents other than water to gain insight into the oxidation process. As shown in Scheme 1, the BV oxidation mechanism is generally accepted to proceed in two steps where first, a peroxide is added to a C=O group forming a Criegee intermediate, followed by migration of R1 group adjacent to the C=O group to insert an oxygen into a C-C bond in a concerted manner. Both Brønsted and Lewis acids catalyze this process via interacting with the carbony group, the peroxide and/or the Criegee intermediate to facilitate the addition and/or the migration step to influence the final product selectivity.21-24

Scheme 1. Acid-catalyzed oxidation of ketones using hydroperoxides via a Criegee intermediate.

Scheme 2. Products obtained from the acid-catalyzed oxidation of methyl levulinate with H2O2 in methanol.

As shown in Scheme 2, the oxidation of methyl levulinate (1) can follow two different

Baeyer-Villiger Oxidation of Levulinates to Succinates pathways. For a pathway involving oxygen insertion between the δ-carbon and the γ-carbon, dimethyl succinate (2) is produced. Alternatively, when the oxygen insertion occurs between the γ-carbon and the β-carbon, methyl 3-acetoxypropanoate is generated, which readily hydrolyzes into methyl acetate (3) and 3-hydroxypropanoate (4) in the presence of an acid catalyst. We note that propanoate 4 may undergo further oxidation to generate methyl 3-methoxypropanoate (5), methyl 3,3-dimethoxypropanoate (6) and dimethyl malonate (7). Although both strong Brønsted and Lewis acids are active for the oxidation of methyl levulinate, we demonstrate that the product distribution strongly depends on the solvent polarity and, for Lewis acidic triflates, the metal cations of the triflate salts.

2. Experimental section

2.1. Catalytic tests

Oxidation reactions were carried out in 5 mL thick-walled glass reactors. In a typical oxidation reaction of methyl levulinate (ML), hydrogen peroxide solution (50 wt % in water) was added into ML (61 mmol L-1) in methanol at a 1:2 ML/peroxide molar ratio. After catalysts were added at a 10:1 ML/cat. molar ratio, the reactor was sealed and heated at 80 °C for 6 h. Liquid samples were syringe filtered (0.2 μm PTFE membrane) and analyzed by either HPLC or GC. For HPLC analysis, compounds were separated using an Agilent 1200 HPLC with a Bio-Rad Aminex HPX-

87H column at 30 °C and detected by refractive index detector, using 10 mM H2SO4 in deionized water as the mobile phase at a flow rate of 0.6 mL min−1. For GC analysis, the oxidation products were separated and identified using an Agilent 7980 GC equipped with an Agilent DB-1701 column and an Agilent 5975C mass selective detector (MSD). The quantification was based on the flame ionization detector (FID) signal. The GC parameters used were as follows: inlet temperature 280 °C; split ratio 25:1; helium as the carrier gas with a flow rate of 1.2 mL min−1 (35 °C). The temperature program started at 35 °C and held for 3 minutes; then the temperature was increased to 165 °C at a ramp of 10 °C min−1, and further increased to

−1 255 °C at a ramp of 15 °C min . The reaction mixture was also diluted 5 times with CH3OD to characterize using 1H nuclear magnetic resonance (NMR) spectroscopy (Bruker 400) to further

confirm the presence of the oxidation products. The H2O2 decomposition was determined by

-1 titration with 0.25 mol L Ce(SO4)2 aqueous solution.

Chapter 3 Oxidation reactions of aliphatic ketones were conducted in a similar way. The reactions were quantified by an Agilent 7980 GC equipped with an Agilent DB-1701 column using FID. With the same GC parameters as above, the temperature program used started at 35 °C and held for 3 minutes; then the temperature was increased to 60 °C at 5 °C min−1, and further increased to 240 °C at 15 °C min−1. Reactions in heptane employed tert-butyl hydroperoxide solution (~5

-1 mol L in decane) as oxidant instead of H2O2 aqueous solution due to the low water solubility in heptane. All the products were separated and identified with the GC using mass spectroscopy. 1,3,5-Tri-tert-butylbenzene was used as an internal standard.

2.2. Quantification of CO2 formed

ML oxidation was conducted in a 25 mL Parr reactor with a PTFE liner. Hydrogen peroxide

-1 solution (50 wt % in water) was added into ML (61 mmol L ) in methanol at a 1:2 ML/H2O2 molar ratio. After catalysts, p-TsOH, were added at a 10:1 ML/acid molar ratio, the reactor was purged using 1 % Ar/He for 5 times and pressurized at 2 atm, and then heated at 80 °C for 6 h. Then the reaction was quenched using an ice bath. The gas sample was quantified using a Hiden Analytical HPR 20 mass spectrometer with 1 % Ar/He as the carrier gas at a flow rate of

−1 100 mL min . A control experiment without ML was also conducted and CO2 formed was quantified in the same way. Liquid samples were characterized in the same way as described above.

CO2 formed is less than 2 % of ML converted.

3. Results and discussion

3.1. Brønsted acid catalyst screening

Table 1 shows the results of the BV oxidation of methyl levulinate using both homogeneous and

heterogeneous Brønsted acid catalysts with H2O2 as oxidant and methanol as solvent. Our data indicate that acid strength has a drastic effect on activity. Acetic acid, a weak Brønsted acid, was not active even at a levulinate/acid molar ratios of 1 (Table 1, entry 5). However, strong Brønsted

acid catalysts, such as p-toluenesulfonic acid (p-TsOH), methanesulfonic acid (MeSO3H), sulfuric acid (H2SO4), and TfOH, all catalyzed the levulinate oxidation effectively, generating similar product distributions (entries 1-4). Similar conversions (defined as moles of subtracted converted/initial moles of substrate × 100 %) and selectivities (defined as moles of product

generated/moles of substrate converted × 100 %) obtained using p-TsOH, MeSO3H and H2SO4

Baeyer-Villiger Oxidation of Levulinates to Succinates 25 are expected based on their similar Hammett acidity functions H0 of ≈ 2.0 (measured in ethanol and mainly due to solvent leveling26). For instance, p-TsOH converted 56 % of levulinate 1 after 6 h with a selectivity of 61 % and 39 % to succinate 2 and methyl acetate 3, respectively (entry 1). Along with methyl acetate produced is propanoate 4, most of which is further converted into 6 (58 %) and 7 (35 %). Due to the nature of the reaction, a concomitant 22 % yield (defined as moles of product generated/initial moles of substrate × 100 %) of 4 is obtained. The propanoate 4 is further converted into 6 and 7 resulting in yields of 13 % and 8 %, respectively. Similar

amounts of CO2 were detected when analyzing the headspace of reactions using p-TsOH and

H2O2 in methanol in the presence and in the absence of methyl levulinate, indicating that although terminal hydroxyl groups are readily oxidized, carbon-carbon cleavage is unlikely under the reaction conditions investigated.20 Overall, the heterogeneous acids generated lower levulinate conversions compared to their homogeneous counterparts. Specifically, Amberlyst-15 and a polymer-supported p-TsOH material featured conversions of 13 and 24 %, respectively (entries 6 and 8). Grinding the resins into powders did not increase the activities, but increasing the amount of catalyst from a levulinate/acid molar ratio of 10:1 to 1:1 resulted in similar conversions and product yields as those obtained with the homogeneous acids (entry 7), demonstrating that although the polymer-supported acids have lower reactivity (likely due to changes of the interaction between the acid sites and their chemical environment after immobilization),27,28 the

a Table 1. Brønsted acids for methyl levulinate oxidation in methanol with aqueous H2O2 as oxidant

b c Entry Catalyst Conv. (%) SM (%) SA (%) 1 p-TsOH 56 61 39

2 MeSO3H 48 54 42

3 H2SO4 54 61 37 4 TfOH 49 61 29 d 5 CH3COOH <1 n.d. n.d. 6 Amberlyst-15 13 60 36 7d Amberlyst-15 67 61 39 8e p-TsOH 24 55 29 9 Al-H-BEA 47 22 32 a Reaction conditions: methyl levulinate 61 mmol/L, levulinate/peroxide molar ratio = 1:2, b levulinate/acid molar ratio = 10:1, 80°C, 6 h. SM, selectivity to succinate, mol of succinate c generated/mol of levulinate converted. SA, selectivity to acetate, moles of methyl acetate generated/moles of levulinate converted. d levulinate/acid molar ratio = 1. e Polymer-supported p-TsOH, levulinate/acid molar ratio = 9:1.

Chapter 3 product selectivity is not affected. Interestingly, the zeolite H-BEA generated a levulinate conversion of 47 % but with a succinate selectivity of 22 % (entry 9), which is much lower than that obtained with the other Brønsted acids tested.

3.2. Effects of reaction conditions and molecular structure on reaction selectivity with Brønsted acids as catalyst

For the BV oxidation, the reported migratory aptitude of the R groups generally follows the order: tertiary alkyl > secondary alkyl > primary alkyl > methyl.23,29 However, our data shows that the selectivity to dimethyl succinate, the product of methyl group migration, is higher than the selectivity to methyl acetate. To rationalize this effect, we investigated the influence of the substrate’s molecule structure (i.e., carbon chain length and branching) and reaction conditions (i.e., different oxidants and solvents) on the product distribution using both a heterogeneous (i.e., Amberlyst-15) and a homogeneous (i.e., p-TsOH) Brønsted acid as catalysts. Figure 1 shows the product distribution obtained from reacting 2-pentanone (8a), 2-hexanone (8b) and C6 methyl ketones with a branched secondary alkyl and a tertiary alkyl group (8c and 8d, respectively). In

methanol with H2O2 as oxidant, both p-TsOH (red, Table A3.3) and Amberlyst-15 (A-15, black, Table A3.4) generated methyl ester 9/acetate 10 molar ratios higher than 1. Furthermore, the data show that branching of the carbon backbone also results in higher methyl ester 9/acetate 10 molar ratios. For instance, the 9d/10d molar ratio with Amberlyst-15 as the catalyst is 10.5, a value over six times as high as the molar ratio observed for 9c/10c (1.6) and about nine times as high as the one for 9b/10b (1.2). When tert-butyl hydroperoxide (TBHP) was used as oxidant in methanol (blue bars in Figure 1, Table A3.5), methyl group migration continued to dominate the product distribution under the reaction conditions investigated. However, the branched ketones yielded

lower methyl ester 9/acetate 10 molar ratios when compared to the H2O2 system. For instance, the 9d/10d molar ratio decreased from 10.5 to 0.9. Moreover, when the solvent is switched to heptane (yellow bars in Figure 1, Table A3.6), all the methyl ester 9/acetate 10 molar ratios became lower than 1.

Evidently, the solvent choice seems to critically impact the product selectivity. With Amberlyst- 15 as catalyst and TBHP in decane as oxidant, all methyl ester 9/acetate 10 molar ratios were below 1 when heptane (yellow), instead of methanol (blue), was used as solvent. The migration ratios obtained in heptane are consistent with the general migratory aptitude for BV oxidation.23,29

Baeyer-Villiger Oxidation of Levulinates to Succinates Selectivity shifts during the BV oxidation of ketones in different solvents were investigated by Lehtinen et al.30 Specifically, solvents capable of forming hydrogen bonds, such as methanol and 1-propanol, were shown to favor the migration of hydrogen over branched alkyl groups, while the migration of branched alkyl groups was preferentially observed in non-hydrogen-bonding solvents, such as toluene and CH2Cl2. Indeed, when 1-propanol and 1-butanol were used as solvents for the oxidation of 8a and 8b, respectively (Table A3.3, entries 5-6), product distributions obtained were similar to those in methanol (Table A3.3, entries 1-2). Considering the highly oxygenated nature of our reactive species, alcohol solvents are likely to interact strongly with the Criegee intermediate, possibly altering the stability of the transition state. We further verified this effect by oxidizing methyl levulinate in heptane with TBHP under the same reaction conditions as those shown in Table 1. In agreement with the product distribution changes shown in Figure 1, a lower selectivity of 14 % to dimethyl succinate was obtained (Table A3.6, entry 5).

10 p-TsOH, H2O2, methanol A-15, H2O2, methanol 8 A-15, TBHP, methanol A-15, TBHP, heptane 3

Methylester 9/Acetate(mol/mol) 10 0 8a 8b 8c 8d Methyl ketone 8

Figure 1. Product distribution for the oxidation of methyl ketones expressed as methyl ester 9/acetate 10 molar ratios. Reaction conditions: substrate 61 mmol/L, substrate/peroxide molar ratio = 1:2,

substrate/acid molar ratio = 1, 80°C, 6 h. p-TsOH as catalyst, 50 wt % aqueous H2O2 as oxidant in

methanol (red). Amberlyst-15 (A-15) as catalyst, 50 wt % aqueous H2O2 as oxidant in methanol (black). A- 15 as catalyst, tert-butyl hydroperoxide (TBHP) in decane as oxidatnt in methanol (blue). A-15 as catalyst, TBHP in decane as oxidant in heptane (yellow). We observe that different oxidants also affect the migratory preferences during BV oxidation.

With Amberlyst-15 in methanol, aqueous H2O2 (black) generated higher methyl ester 9/acetate

Chapter 3 10 ratios than TBHP (blue), indicating the methyl migration is less favourable with TBHP as oxidant. Hawthorne et al. showed a higher differentiation of the migration groups with peroxyacetic acid as oxidant than trifluoroperacetic acid for the BV oxidation of cyclohexyl phenyl ketone.31 It was hypothesized that the impact of the migrating groups on the energy barrier is larger for the less reactive intermediate.31 In our case, the different electronic and steric

effects between TBHP and H2O2 may be responsible for the different product selectivities observed.

As a consequence, we observe that the molecular structure of the ketones, i.e., carbon chain

length and branching, also affect the migration preference. In methanol with H2O2 as oxidant, a higher degree of branching in the carbon backbone resulted in higher methyl ester 9/acetate 10 molar ratios (Figure 1, red and black bars). Although in methanol with TBHP as oxidant the trend was reversed (Figure 1, blue bars), the ratios of 9a/10a (C5) were higher than 9b/10b (C6) under all the reaction conditions investigated. To our knowledge, virtually no studies exist for the BV oxidation of asymmetric ketones in alcohol solvents. Our data indicate that the migratory aptitude trend obtained in methanol is opposite to that in heptane. Mora-Diez et al. showed that the polarity of solvents affects the reaction mechanism.32 Hence, our studies encourage more investigation, both experimental and theoretical, to understand the effects of oxidants, catalysts and the molecular structure on the reaction selectivity in alcohol solvents.

3.3. Metal triflates as Lewis acidic catalysts

Lewis acids are known as active BV oxidation catalysts.21-24 Some metal triflates have been shown to maintain their Lewis acid activity in the presence of water, which allows their use with

33,34 aqueuous solutions of H2O2. Table 2 shows the activity of metal triflates for the oxidation of methyl levulinate. Compared to Brønsted acids, metal triflates generate lower levulinate conversions and yield a wider spectrum of product selectivies. Among the metal triflates tested,

Zn(OTf)2 showed the lowest conversion (7 %) with no significant succinate formation after 6 h (Table 2, entry 11). Increasing the reaction time to 24 h resulted in a succinate yield of 14 %.

The highest levulinate conversions were obtained with Hf(OTf)4 (36 %), Hg(OTf)2 (40 %) and

Sc(OTf)3 (38 %), generating selectivities to succinate exceeding 47 % with the rest of the carbon ending up in acetate-derived products (entries 1, 5, and 7). In comparison, triflates based on Y, Yb, and Er showed lower levulinate conversions of ≈ 16 % with succinate selectivities of ≈ 40 %.

Baeyer-Villiger Oxidation of Levulinates to Succinates

Interestingly, In(OTf)3 showed the highest selectivity to acetate (70 %) at a levulinate conversion of 16 % (entry 6). Moreover, Group IIIB and IVB metal cations showed higher selectivities to dimethyl succinate than to methyl acetate (entries 1-3, 7-10), whereas Group IIB and IIIA metal cations showed higher selectivities to methyl acetate (entries 5, 6, and 11).

The catalytic activity of metal triflates can be correlated with their hydrolysis constant Kh and water exchange rate constant (WERC, defined as the exchange rate for substitution of inner-

sphere water ligands) (Table 2 and Figure 2). A lower pKh value translates to a stronger hydrolysis

34 tendency of the metal salt. As such, Zn(OTf)2 possessing the highest pKh value (8.96), showed

the lowest levulinate conversion. Similarly, triflate salts of Yb, Er and Y, featuring pKh values ranging from 7.7 to 7.9, showed moderate activities. For salts with pKh lower than 4.3, including

Hf(OTf)4, Sc(OTf)3, and Hg(OTf)2, higher levulinate conversions were obtained. Although

In(OTf)3 has a pKh of 4.0, it converted only 16% of the substrate (Table 2, entry 6). Unlike other triflates, In(OTf)3 has a WERC orders of magnitude smaller. Given that the dissociation rate of the triflate ligand is proportional to the WERC,35 a smaller WERC value translates into a slower dissociation rate of triflate ligand into triflic acid. The higher activity associated with metal triflates that possess higher dissociationand exchange constants provides strong evidence that the hydrolysis product, TfOH, is likely the true catalyst. This hypothesis was further confirmed by adding 2,6- di-tert-butylpyridine (DTBP), a steric hindered base,36-39 to the reaction vessel at different molar ratios (Table 2, entries 2-4). As expected, addition of DTBP progressively suppressed reactivity,

up to complete deactivation upon reaching a DTBP/Hf(OTf)4 molar ratio of 4.

The different migration preferences obtained with triflates of different groups of metals indicate a possible synergetic effect between the metal triflate and triflic acid formed in situ impacts product selectivity. The apparent cooperative effect between the Lewis acid centers and the generated Brønsted acid species of metal triflates has been often reported to improve the reaction activity and/or alter the selectivity in reactions, such as the etherification of glycerol with short chain alkyl alcohols and hydrothiolation of non-activated olefins.36,40,41 The Lewis acidic metal centers are expected to strong interact with the highly oxygenated compounds involved in the reaction, and likely affect the structure of the transition state and alter the reaction selectivity. Indeed, the reaction selectivity correlates with the size of the metal cations instead of the hydrolysis properties, as indicated in Figure 3. Group IIIB and IVB metal cations possessing a

Chapter 3 covalent radius above 1.70 Å showed a higher selectivity towards succinate, the methyl group migration products, whereas Group IIB and IIIA metal cations with a covalent radius less than 1.70 Å preferentially catalyzed the alkyl group migration.

a Table 2. Metal triflates for methyl levulinate oxidation in methanol with aqueous H2O2 as oxidant

b c Entry Catalyst Conv. (%) SM (%) SA (%) pKh WERC

1 Hf(OTf)4 36 57 39 0.25 - d 2 Hf(OTf)4 25 61 38 0.25 - e 3 Hf(OTf)4 15 61 39 0.25 - f 4 Hf(OTf)4 <1 n.d. n.d. 0.25 - 9 5 Hg(OTf)2 40 47 48 3.4 2×10 4 6 In(OTf)3 16 26 70 4.0 4.0×10 7 7 Sc(OTf)3 38 50 33 4.3 4.8×10 7 8 Y(OTf)3 17 39 22 7.7 1.3×10 7 9 Yb(OTf)3 16 44 21 7.7 8×10 8 10 Er(OTf)3 16 43 31 7.9 1.4×10 8 11 Zn(OTf)2 7 n.d. 41 8.96 5×10 a Reaction conditions: methyl levulinate 61 mmol/L, levulinate/peroxide molar ratio = 1:2, levulinate/cat. b molar ratio = 10:1, 80°C, 6 h. SM, SA, same as in Table 1. pKh = - log(Kh), Kh is the hydrolysis constant 34 c 34 d of the metal triflate. WERC, water exchange rate constant of the metal triflate. DTBP/Hf(OTf)4 e f molar ratio = 0.5. DTBP/Hf(OTf)4 molar ratio = 1.4. DTBP/Hf(OTf)4 molar ratio = 4.0.

Figure 2. The effect of hydrolysis constant of metal triflates on ML conversion. Reaction conditions: methyl levulinate 61 mmol/L, levulinate/peroxide molar ratio = 1:2, levulinate/cat. molar ratio = 10:1, 80°C, 6 h.

Baeyer-Villiger Oxidation of Levulinates to Succinates

Figure 3. The effect of covalent radius of metal triflates on BV oxidation selectivity. Reaction conditions: methyl levulinate 61 mmol/L, levulinate/peroxide molar ratio = 1:2, levulinate/cat. molar ratio = 10:1, 80°C, 6 h. 3.4. Computational study on the reaction mechanisms

Scheme 3. Baeyer-Villiger reaction of 2-hexanone in the presence of H2O2 catalyzed by acid catalysts. For acid-catalyzed BV oxidation in alcohol solvents, there are many remaining questions. For instance, for Brønsted acid catalyts, why acetic acid did not show any activity, why solvent polarity significantly affects the migratory preference, and why Brønsted acids are not active in water, etc. For Lewis acidic metal triflates, how the metal center interacts with the transition state and affect the reaction selectivity. Here we only focus on the first question about the inactivity of acetic acid. In an attempt to answer this question, we have conducted density functional theory (DFT) calculations with Gaussian 03 software package. Considering that the calculations may be extended to study other properties of the system, PBE1PBE, a dispersion corrected functional was

Chapter 3 used, which is also suitable for studying hydrogen bonding effects.42 To study the solvent effects, a continuum solvation model, IEF-PCM, was used for single-point energy calculation corrections.32 The basis set used in this study is 6-311++G(d,p). The standard state for the gas-phase calculation is 298 K and 1 atm. For liquid-phase calculation, further corrections, such as Benson’s correction,43 was not applied given the purpose of this preliminary study. a b c

Figure 4. Ground states (GS) for a) no catalyst, b) acetic acid as catalyst, and c) MeSO3H as catalyst. Red, oxygen atom; dark grey, carbon atom; light grey, hydrogen atom; yellow, sulfur atom. a b c

d e f

Figure 5. Criegee intermediates for alkyl group migration (I) when a) no catalyst, b) acetic acid as catalyst, and c) MeSO3H as catalyst, and for methyl group migration (MI) when d) no catalyst, e) acetic acid as catalyst, and e) MeSO3H as catalyst. Red, oxygen atom; dark grey, carbon atom; light grey, hydrogen atom; yellow, sulfur atom.

Baeyer-Villiger Oxidation of Levulinates to Succinates Mora-Diez et al. showed that a concerted non-ionic pathway is the most energetic favored in non-polar solvents, and the rate-determining step in water is a concerted non-ionic migration step in a pathway with an ionic addition step.32 Considering the low acidity of acetic acid and relatively lower polarity of methanol compared to water, the concerted non-ionic pathway was chosen to study first. The BV oxidation of 2-hexanone studied is shown in Scheme 3. For the alkyl migration, the transition state for addition is labelled as TS1, and TS2 is for the migration step. For the methyl group migration, the addition and migration steps are labelled as MTS1 and MTS2, respectively. The optimized structures of the reactants, intermediates and transition states are shown in Figure 4-7.

a b c

d e f

Figure 6. Transition states to form Criegee intermediates (TS1) for alkyl group migration when a) no

catalyst, b) acetic acid as catalyst, and c) MeSO3H as catalyst, and for methyl group migration (MTS1)

when d) no catalyst, e) acetic acid as catalyst, and e) MeSO3H as catalyst. Red, oxygen atom; dark grey, carbon atom; light grey, hydrogen atom; yellow, sulfur atom. As shown in Table 3, the addition step from gound state (GS) to Criegee intermediate is mostly the rate-limiting step without any catalysts. Its ΔG (174.1 kJ/mol) in the alkyl migration pathway

44 under vacuum is consistent with literature. Using either acetic acid or MeSO3H significantly

Chapter 3 decreases ΔG for each step compared to the case without any catalysts, and leaves the migration step rate-limiting, as shown in previous studies.32,44 However, ΔG values for migration step catalyzed by MeSO3H are at least 22 kJ/mol lower than those for migration step catalyzed by acetic acid, which translates into a reaction rate at least three orders of magnitude higher for

MeSO3H catalyzed migration than that for acetic acid catalyzed migration. Therefore, it is expected that no reactivity should be observed for acetic acid in the reaction. a b c

d e f

Figure 7. Transition states for alkyl group migration (TS2) when a) no catalyst, b) acetic acid as catalyst, and c) MeSO3H as catalyst, and for methyl group migration (MTS2) when d) no catalyst, e) acetic acid as catalyst, and e) MeSO3H as catalyst. Red, oxygen atom; dark grey, carbon atom; light grey, hydrogen atom; yellow, sulfur atom. We also note that the ΔG for methyl group migration (methyl group migration Criegee intermediate (MI)  MTS2) is always higher than ΔG for alkyl group migration (Alkyl migration Criegee intermediate (I)  TS2) and the methyl group migration gets more energetically unfavorable as the solvent gets more polar. This is opposite from what we have observed. One possible reason is that a continuum solvation model is used and it can mainly capture the overall electronic effect of the solvation on the reaction pathways but may overlook

Baeyer-Villiger Oxidation of Levulinates to Succinates the specific interaction between solvent molecules and the reactive species. Since the reactive species are highly oxygenated, the molecular interaction between solvent molecules and the reactive species are likely important to be accounted for in the overall energistics. This can be achieved by additing explicit solvent molecules into the computation models, e.g. one or two solvent molecules32 or the first solvation shell, in addition to the continuum solvent model. Another factor that may cause this discrepency is the cencerted non-ionic pathway employed in

the study above. When strong Brønsted acids, such as MeSO3H, are used as catalysts in a protic polar solvent, i.e. methanol, an inonic reaction pathway is more viable, along which the stabilization effects of solvation of reactive species, especially the transition state, may be more favored. Therefore, the ionic pathway should be investigated to elucidate the interaction between the solvent and the reactive species in alcohol solvents.

Table 3. DFT calculations for BV oxidation of 2-hexanonea

ΔG (kJ/mol) Catalyst GS TS1b I  TS2c GS  MTS1 MI  MTS2d None 174.1 171.5 176.6 177.6 vacuum (ε = 1, 20oC) Acetic acid 65.5 105.7 73.9 113.3

MeSO3H 53.1 83.2 66.3 87.0 None n.a.e n.a. n.a. n.a. heptane (ε = 2, 20oC) Acetic acid 64.1 104.4 73.1 112.8 f MeSO3H n.d. 79.1 63.9 n.d. None 208.4 178.4 211.3 191.1 methanol (ε = 33, 20oC) Acetic acid 65.9 102.1 73.9 109.7

MeSO3H n.d. 65.1 n.d. n.d. a PBE1PBE/6-311++G(d,p), solvent effect with model IEF-PCM. b GS, ground state, the configuration of c 2-hexanone, H2O2 and the acid catalysts (if applicable) with the lowest Gibbs free energy. I, Criegee intermediate for alkyl migration. d MI, Criegee intermediate for methyl migration. e n.a., not available as not calculated. f n.d., not determined, as not converged.

4. Conclusions

In summary, the oxidation of methyl levulinate into dimethyl succinate was performed using peroxide-based oxidants in methanol. Both Brønsted and Lewis acids catalyzed the reaction

without significant CO2 formation. With strong Brønsted acids, a selectivity of ≈ 60 % to dimethyl succinate was obtained. Upon switching from methanol to heptane as solvent, the selectivity decreased to 14 %, indicating the strong solvent effect on directing product selectivity.

Chapter 3 Lewis acidic metal triflates generated a broader spectrum of product selectivities in methanol, which is dictated by the nature of the metal cations (e.g., size) in the triflate salt.

Acknowledgements

Ferdinand Vogelgsang is thanked for his assistance with the catalytic reactions.

References

(1) Song, H.; Lee, S. Y. Enzyme Microb. Technol. 2006, 39, 352-361. (2) Podolean, I.; Kuncser, V.; Gheorghe, N.; Macovei, D.; Parvulescu, V. I.; Coman, S. M. Green Chem. 2013, 15, 3077. (3) Beauprez, J. J.; De Mey, M.; Soetaert, W. K. Process Biochem. 2010, 45, 1103-1114. (4) Chen, K. Q.; Li, J.; Ma, J. F.; Jiang, M.; Wei, P.; Liu, Z. M.; Ying, H. J. Bioresour. Technol. 2011, 102, 1704-1708. (5) Choudhary, H.; Nishimura, S.; Ebitani, K. Appl. Catal., A. 2013, 458, 55-62. (6) Li, J.; Zheng, X. Y.; Fang, X. J.; Liu, S. W.; Chen, K. Q.; Jiang, M.; Wei, P.; Ouyang, P. K. Bioresour. Technol. 2011, 102, 6147-6152. (7) Werpy, T.; Petersen, G.; Aden, A.; Bozell, J.; Holladay, J.; White, J.; Manheim, A.; Eliot, D.; Lasure, L.; Jones, S. “Top value added chemicals from biomass. Volume 1-Results of screening for potential candidates from sugars and synthesis gas,” Department of Energy: Washington DC, 2004. (8) Clark, S. “Bio-succinic Acid Market - Size, Share, Trends, Opportunities, Global Demand, Insights, Analysis, Research, Report, Company Profiles, Segmentation and Forecast, 2013 - 2020,” Allied Market Research, 2014. (9) Dicarboxylic Acids, Aliphatic. Ullmann's Encyclopedia of Industrial Chemistry [Online]; Posted November 19, 2014. http://onlinelibrary.wiley.com/doi/10.1002/14356007.a08_523.pub3/pdf. (10) Succinic Acid and Succinic Anhydride. Kirk-Othmer Encyclopedia of Chemical Technology [Online]; Posted April 14, 2006. http://onlinelibrary.wiley.com/doi/10.1002/0471238961.1921030306211301.a01.pub2/p df. (11) Bui, L.; Luo, H.; Gunther, W. R.; Roman-Leshkov, Y. Angew. Chem. Int. Ed. 2013, 52, 8022-8025. (12) Lin, H.; Strull, J.; Liu, Y.; Karmiol, Z.; Plank, K.; Miller, G.; Guo, Z.; Yang, L. Energy Environ. Sci. 2012, 5, 9773-9777. (13) Van de Vyver, S.; Thomas, J.; Geboers, J.; Keyzer, S.; Smet, M.; Dehaen, W.; Jacobs, P. A.; Sels, B. F. Energy Environ. Sci. 2011, 4, 3601-3610. (14) Weingarten, R.; Conner, W. C.; Huber, G. W. Energy Environ. Sci. 2012, 5, 7559-7574. (15) Wettstein, S. G.; Alonso, D. M.; Chong, Y.; Dumesic, J. A. Energy Environ. Sci. 2012, 5, 8199-8203. (16) Luterbacher, J.; Martin Alonso, D.; Dumesic, J. Green Chem. 2014, 16, 4816-4838. (17) Dunlop, A. P.; Smith, S. United States, 2676186, 1954.

Baeyer-Villiger Oxidation of Levulinates to Succinates (18) Van Der Klis, F.; Van ES, D. S.; Van Haveren, J. United States, WO2012044168 A1, 2012. (19) Pandey, S. K.; Yadav, S. P. S.; Prasad, M.; Prasad, J. Asian J. Chem. 1999, 11, 203-206. (20) Liu, J.; Du, Z.; Lu, T.; Xu, J. ChemSusChem 2013, 6, 2255-2258. (21) Roy, S.; Bakhmutsky, K.; Mahmoud, E.; Lobo, R. F.; Gorte, R. J. ACS Catal. 2013, 3, 573-580. (22) Uyanik, M.; Ishihara, K. ACS Catal. 2013, 3, 513-520. (23) Brink, G.-J. t.; Arends, I. W. C. E.; Sheldon, R. A. Chem. Rev. 2004, 104, 4105 - 4123. (24) Luo, H. Y.; Bui, L.; Gunther, W. R.; Min, E.; Román-leshkov, Y. ACS Catal. 2012, 2, 2695-2699. (25) García-Suárez, E. J.; Khokarale, S. G.; van Buu, O. N.; Fehrmann, R.; Riisager, A. Green Chem. 2014, 16, 161-166. (26) Mihichuk, L. M.; Driver, G. W.; Johnson, K. E. ChemPhysChem 2011, 12, 1622-1632. (27) Siegel, R.; Domingues, E.; De Sousa, R.; Jérôme, F.; Morais, C. M.; Bion, N.; Ferreira, P.; Mafra, L. J. Mater. Chem. 2012, 22, 7412-7419. (28) Mbaraka, I.; Shanks, B. J. Catal. 2006, 244, 78-85. (29) Renz, M.; Meunier, B. Eur. J. Org. Chem. 1999, 1999, 737-790. (30) Lehtinen, C. N., Vesa; Brunow, Gösta. Tetrahedron 2001, 55, 4741-4751. (31) HAWTHORNE, M. F.; EMMONS, W. D. J. Am. Chem. Soc. 1958, 80, 6393 - 6398. (32) Mora-Diez, N.; Keller, S.; Alvarez-Idaboy, J. R. Org. Biomol. Chem. 2009, 7, 3682 - 3690. (33) Román-Leshkov, Y.; Davis, M. E. ACS Catal. 2011, 1, 1566-1580. (34) Kobayashi, S.; Nagayama, S.; Busujima, T. J. Am. Chem. Soc. 1998, 120, 8287-8288. (35) Martell, A. E. Coordination Chemistry; American Chemical Society: Washington, DC, 1978. (36) Liu, F.; De Oliveira Vigier, K.; Pera-Titus, M.; Pouilloux, Y.; Clacens, J.-M.; Decampo, F.; Jérôme, F. Green Chem. 2013, 15, 901-909. (37) Dang, T. T.; Boeck, F.; Hintermann, L. J. Org. Chem. 2011, 76, 9353-9361. (38) Wabnitz, T. C.; Yu, J. Q.; Spencer, J. B. Chem. Eur. J. 2004, 10, 484-493. (39) Brown, H. C.; Kanner, B. J. Am. Chem. Soc. 1966, 88, 986-992. (40) Weiwer, M.; Coulombel, L.; Dunach, E. Chem. Commun. 2006, 332-334. (41) Yamamoto, H.; Futatsugi, K. Angew. Chem. Int. Ed. 2005, 44, 1924-1942. (42) Ireta, J.; Neugebauer, J.; Scheffler, M. J. Phys. Chem. A 2004, 108, 5692-5698. (43) Mora-Diez, N.; Keller, S.; Alvarez-Idaboy, J. R. Org. Biomol. Chem. 2009, 7, 3682- 3690. (44) Bach, R. D. J. Org. Chem. 2012, 77, 6801-6815.

Chapter 4

Synthesis of Itaconic Acid Ester Analogs via Self-Aldol Condensation of Ethyl Pyruvate Catalyzed by Hafnium BEA Zeolites

1. Introduction

Dicarboxylic acids (diacids) play a central role in the biobased chemicals portfolio, as evidenced by their prevalence in the top 12 chemicals from biomass identified by the U.S. Department of Energy.1 Diacids are building blocks in condensation polymerization reactions,2 and their ester forms (diesters) can serve as lubricants, plasticizers, and polymer intermediates.3 In particular, the unsaturated diacid itaconic acid is a potential biodegradable substitute for high-volume petroleum-derived chemicals such as acrylic acid, maleic anhydride, or acetone cyanohydrin.4 It can also be used in the production of superabsorbent polymers, synthetic latex, and laminating resins.4 Although some diacids (e.g. succinic acid) are already produced industrially from biomass, most, including itaconic acid, suffer from prohibitively high production costs.5-8 Current catalytic routes to synthesize diacids and diesters from biomass-derived molecules, such as Baeyer-Villiger oxidation,9,10 C-C bond cleavage,11 and noble metal catalyzed aerobic oxidation,12,13 suffer from poor selectivity,9,10 inefficient carbon utilization,11 and/or catalyst deactivation.12 Flanagan et al. reported a C-C coupling strategy to produce unsaturated dicarboxylic esters through the addition dimerization of crotonates;14 however, the system depends on homogeneous catalysts.

An alternative route to generate diacids or diesters is via the C-C coupling of keto acids or esters, which are common intermediates in a myriad of metabolic pathways and can be produced through biocatalysis on a large scale.7,15-20 While this coupling strategy is promising, only a few homogeneous catalysts have been shown to activate the carbonyl group adjacent to the ester group.21-23 However, these catalysts lactonize the aldol adducts instead of generating linear condensation products. Solid base catalysts typically used for aldol condensation are not ideal for the proposed system, as they easily deactivate in the presence of organic acids.24-26

In this work, we demonstrate a general strategy to synthesize unsaturated dicarboxylic acid esters via the C-C coupling of keto esters with Lewis acidic zeolites. We emphasize the synthesis of itaconic acid ester analogs from the condensation of ethyl pyruvate (EP, 1) (Scheme 1) due to the low thermal stability of pyruvic acid.27 Zeolites with BEA topology containing Lewis acidic

Chapter 4 framework heteroatoms, including Sn(IV), Hf(IV) and Zr(IV), can catalyze C-C coupling of biomass-derived oxygenates.28-31 These materials promote aldol condensation via a soft enolization pathway reminiscent to that observed in type II aldolases.24 Importantly, these zeolites feature remarkable tolerance to carboxylic acid groups24 and water,32,33 making them prime candidates for the biomass-based production of diacids and diesters.

2. Experimental section

2.1. Catalyst synthesis

Zeolites were synthesized based on a procedure reported by Corma et al.34 using the following precursors: hafnium(IV) chloride, zirconium(IV) oxychloride octahydrate, and tin(II) chloride

dihydrate. Tin(II), which oxidizes to tin(IV) in water, was used instead of SnCl4·5H2O since it has

35 been shown to result in Sn-Beta consistently free of extra-framework SnO2. Briefly, Hf-Beta was synthesized as follows: aqueous tetraethylammonium hydroxide (27.158 g) and tetraethyl orthosilicate (23.968 g) were added to a Teflon [polytetrafluoroethylene (PTFE)] dish, which was magnetically stirred at 250 rpm and room temperature for 90 min. Additional deionized water (15 mL) was added, and the dish was cooled in an ice bath. Then, hafnium(IV) chloride (0.3747 g) dissolved in ethanol (2 mL) was added drop wise while stirring. The solution was left uncovered on the stir plate for 12 h to reach a total mass of 33.147 g after evaporation of ethanol and some of the water. Next, aqueous hydrofluoric acid (2.620 g) was added drop wise, and the mixture was homogenized using a PTFE spatula, resulting in a thick gel. Si-Beta (0.364 g) was seeded into the mixture, and the sol-gel was allowed to evaporate to 33.956 g. The final molar composition

was 1 SiO2:0.01 HfCl4:0.56 TEAOH:0.56 HF:7.5 H2O. The thick paste was transferred to a 45 mL PTFE-lined stainless steel autoclave and heated to 140 oC for 7-20 days under static conditions. The solids were recovered by filtration, washed with deionized water, and dried at 100 oC. Zeolites were calcined by heating at 580 oC for 10 h with a 1 oC min−1 ramp rate and 1 h isothermal steps at 150 and 350 oC. All catalysts were synthesized to achieve a silicon/metal ratio of ca. 100. Actual ratios as reported by inductively coupled plasma mass spectrometry (ICP-MS) are given in Table 3.

All the metal oxides supported on Si-BEA were prepared via incipient wetness impregnation of Si-Beta with the aqueous solutions of the same metal precursors used above, followed by drying at

Self-Aldol Condensation of Ethyl Pyruvate to Itaconic Acid Ester Analogs 110 oC and calcination in air flow at 260 oC.36

The Na+ and H+ ion exchange Hf-BEA zeolites were prepared using methods adapted from

37 Bermejo-Deval et al. Briefly, in Treatment 1, 100 mg zeolites were mixed with 15 mL NaNO3 (1 M) solution at room temperature for 18 h. Then the solids were separated by centrifuge and dried under ambient conditions. The solids recovered were denoted as Hf-BEA-Na-1. Similarly, in Treatment 2, 100 mg zeolites were mixed with 15 mL NaOH (pH = 10) solution at room temperature for 18 h. Then the solids were separated by centrifuge and washed with deionized water once and dried under ambient conditions. The solids recovered were denoted as Hf-BEA-

Na-2. 100 mg of each Na-exchanged Hf-BEA was mixed with 15 mL H2SO4 (1 M) at room temperature for 1 h. Then the zeolites were recovered using centrifuge and washed with deionized water for 5 times before drying under ambient conditions. These materials were then calcined under air at a flow rate of 100 mL min−1 at 550 oC for 10 h and denoted as Hf-BEA- AW-1 and Hf-BEA-AW-2, respectively.

Nanosized zirconia, hereafter referred to as nZrO2, were synthesized according to a previously

13 reported method. Briefly, an aqueous solution of ZrO(NO3)2·xH2O (0.8 M, 375 mL) was added

to a solution of NH4OH in deionized water (0.8 M, 1.1 L). After stirring for 30 min, the solution was aged at 100 °C for 24 h in a polyethylene vessel. The mixture was cooled down, filtered and washed with an excess amount of deionized water. The resultant particles were dried in vacuum and calcined under an air flow of 100 mL min−1 at 400 oC for 4 h.

Hydrotalcite was synthesized by co-precipitation modified from a method previously reported.38

Briefly, after adding 100 mL solution of Al(NO3)3 (0.25 M) and Mg(NO3)2 (0.75 M) into 100 mL

o solution of Na2CO3 (0.5 M) and NaOH (1.25 M), the suspension was aged at 65 C for 18 h. Then the solids were filtered and washed with deionized water after cooling the suspension to room temperature, and then dried overnight at 110 oC.

Dealuminated Al-BEA (DeAl-BEA) was synthesized by stirring 5 g of Al-BEA (Si/Al ratio of 12.5 from Zeolyst International) in 100 mL of 70% nitric acid at 90 oC for 4 h. The material was then washed with deionized water until the filtrate reached neutral pH, dried at 100 oC, and calcined using the procedure for zeolites stated above. The Si/Al ratio after acid treatment was measured by ICP-MS.

Chapter 4 2.2. Catalyst characterization

ICP-MS measurements were conducted using an Agilent 7900 ICP-MS in helium mode. Powder

X-ray diffraction (PXRD) patterns were collected using a Bruker D8 diffractometer with CuKα radiation. N2 adsorption–desorption isotherms were measured on a Quantachrome Autosorb iQ

o apparatus at liquid N2 temperature (-196 C). All samples were degassed under vacuum prior to use (350 oC, 12 h). Micropore volumes were analyzed by the t-plot method. Thermogravimetric analysis (TGA) was conducted by heating the catalyst with a 2 oC min-1 ramp with 1 h isothermal

o o -1 -1 holds at 110 C and 3 h isothermal holds at 550 C under 10 mL min N2 and 90 mL min air flow using a Q500 thermal analysis system (TA Instruments). FTIR spectra were acquired from 4000 to 400 cm-1 using a Bruker Vertex 70 spectrophotometer by averaging 64 scans at 2 cm-1 resolution. Samples (~2 mg) were pressed into 7 mm diameter self-supporting pellets and placed in a Harrick high temperature transmission cell equipped with KBr windows. Samples were calcined in situ under flowing dry air (50 mL min−1), with a temperature ramp of 5 oC min-1 to 300 oC and 1 oC min-1 to 500 oC, and held at 500 oC for 5 h. After the cell was cooled to 30 oC, dynamic vacuum of roughly 0.1 Pa was established. A reference spectrum of the bare material was then acquired. Under a static vacuum, the cell was dosed with CD3CN vapor until a prominent peak at 2268 cm-1 appeared. Spectra were collected as the cell was exposed to vacuum to monitor the disappearance of peaks until the pressure reached 0.1 Pa. The spectra were referenced to the bare material and normalized by the combination and overtone modes of zeolite Si-O-Si stretches (1750–2100 cm-1).

2.3. Catalytic reactions

Batch reactions. Batch reactions were carried out under autogenous pressure using 5 mL glass reactors sealed with PTFE-lined silicone septa. The reaction mixture was kept at the reaction temperature with a temperature-controlled oil bath equipped with magnetic stirring. Typical reaction solutions consisted of 3 wt% ethyl pyruvate (EP) in toluene with 0.1 wt% 1,3,5-tri-tert- butylbenzene as an internal standard. Catalyst was added to the reaction solution such that the molar ratio of ethyl pyruvate to catalyst metal sites was 195:1. The reactor was purged with N2 before the reaction started. Samples were collected periodically. Solids were removed by a 0.2 μm Millipore PTFE syringe filter before analysis by gas chromatography.

Self-Aldol Condensation of Ethyl Pyruvate to Itaconic Acid Ester Analogs Flow reactions. Flow reactions were carried out in a tubular stainless steel reactor (6.35 mm o.d., 4.57 mm i.d., 10 mL volume), fixed inside an aluminum block within an insulated single-zone furnace (850 W/115 V, Applied Test Systems Series 3210). The catalyst bed consisted of the calcined catalyst loaded between quartz wool plugs and supported by borosilicate glass beads (Sigma-Aldrich, 1 mm diameter). A K-type thermocouple (Omega) was used to measure the reactor temperature at the catalyst bed, which was regulated by a temperature controller (Digi- Sense, model 68900-10). The system was pressurized with nitrogen and controlled with a back- pressure regulator (Swagelok SS BP Regulator, 0 to 500 psig). A HPLC pump (Acuflow Series II) was used to introduce the feed solution into the up-flow reactor. We used stainless steel tubing (1.59 mm o.d., 0.57 mm i.d., 0.6 mL volume) to connect the pump to the reactor inlet. The effluent from the reactor was condensed at room temperature in a gas/liquid separator (Gage & Valve Co.), allowing for periodic sampling of the liquid product stream. Reactions were carried out at 120 oC and 12 bar.

Product analysis. Reaction mixtures were analyzed by a 7890A GC System from Agilent Technologies equipped with an Agilent DB-1701 column, flame ionization detector (FID), and 5975C MSD. Compounds were quantified using 1,3,5-tri-tert-butylbenzene as an internal standard. Pure products were obtained through separation of reaction mixtures using automatic column separation with Biotage IsoleraTM flash purification system. Products were identified by NMR and by MS. 1H and 13C NMR spectra were recorded on a Bruker AVANCE III HD (13C, 400 MHz) spectrometer at ambient temperature.

3. Results and discussion

3.1. Reactions catalyzed by Lewis acidic zeolites

As depicted in Scheme 1, the self-aldol addition of EP 1 followed by dehydration of the aldol addition product, diethyl 2-hydroxy-2-methyl-4-oxopentanedioate 2, results in two diester isomers: diethyl 2-methyl-4-oxopent-2-enedioate 3 and diethyl 2-methylene-4-oxopentanedioate 4. The latter is a functional analog of itaconic acid ester. In addition to the main coupling reaction, several undesired side reactions are triggered by the presence of acid sites and water generated from the dehydration of aldol adducts. These side reactions (Scheme 2) include hydrolysis of EP to pyruvic acid and ethanol, Meerwein−Ponndorf−Verley (MPV) reduction of EP accompanied

Chapter 4 with Oppenauer oxidation of ethanol, and aldol reaction between EP and these side products. Toluene was used as the solvent to suppress side reactions that can occur between EP and alcohols (MPV) or water (hydrolysis).

O O O O O O O

-H2O 2 O O O O O + O O O OH O O O 1 234

Side products O O O O O O O O O O O O O OH O O 9 11 O O OH O O O 13 5 678 O O

O O 10 12

Scheme 1. Self-aldol condensation of ethyl pyruvate catalyzed by Lewis acidic zeolites with products from side reactions listed. 3.2. Catalyst screening in batch reactions

The catalytic performance of Lewis acidic zeolites for the self-aldol condensation of EP is shown in Table 1. Hf-BEA and Zr-BEA generate the highest EP conversions (>80%) with comparable selectivities (>64%) to diester products (3 - 5 in Scheme 1) after 1 h at 120 oC (entries 1 and 2). Sn-BEA shows a lower EP conversion of 19% under identical conditions (entry 3), consistent with previous studies on aldol reactions.24 Only a ca. 1% conversion is observed from Si-BEA or metal oxides supported on Si-BEA, and a 17% conversion (with 12% selectivity) is obtained for Al-BEA (Table 2, entries 1-7). Collectively, these results indicate that the catalytic activity most likely originates from framework Lewis acidic heteroatoms. In contrast to previously reported homogeneous catalytic systems,21-23 the lactonization of 2 is not observed with these Lewis acidic zeolites. Nonetheless, compound 13 is likely produced from the lactonization of a C7 molecule (reaction f in Scheme 2), indicating that although lactonization is feasible, the confining effects of the zeolite pores may limit this pathway for larger molecules.39 Metal oxides with Brønsted basicity typically employed for aldol reactions, including MgO and hydrotalcite,40 show EP conversions of less than 13% and selectivities of less than 18% (Table 2, entries 8-12). In these systems, pyruvic acid produced by the hydrolysis of EP is expected to deactivate the basic sites.24

Self-Aldol Condensation of Ethyl Pyruvate to Itaconic Acid Ester Analogs

Scheme 2. Main reactions promoted by Lewis acidic zeolite catalysts in toluene. Reaction a is aldol condensation to produce diesters, as shown in Scheme 1. Side reactions are initiated by hydrolysis (b) of EP to form lactic acid and ethanol. The cascade of side reactions include Meerwein−Ponndorf−Verley reaction (c), condensation (d,e), lactonization (f), and etherification (g,h). Compound 5 is quantified as a diester in addition to compounds 3 and 4 in Scheme 1. Compounds 6-13 are quantified as side products. Pyruvic acid 14 could not be detected by GC and was not quantified.

Table 1. Self-aldol condensation of ethyl pyruvate catalyzed by Lewis acidic zeolitesa

Selectivity (%)b Entry Catalyst EP conversion (%) Diesters Side products 1 Hf-BEA 80 68 15 2 Zr-BEA 81 64 11 3 Sn-BEA 19 67 6 4 Hf-BEAc 78 81 9 a Reaction conditions: 3 wt% EP in toluene, EP/metal = 195 (mol/mol), 120 oC, 1 h.

 NnCi, i Selectivity =×i 100% b Nn()− n CEP,, EPo EP , where NC,i is the number of carbon atoms in compound i, ni is the number of moles of compound i in the reaction mixture, and nEP,o is the initial moles of EP added. Diesters: 3 - 5 in Scheme 1. Side products: 6 - 13 in Scheme 1; distribution for a typical reaction is shown in Figure A4.7. c Calcined in dry air at 550 oC for 5 h after four consecutive runs as shown in Figure A4.8.

Chapter 4 Hf-BEA was investigated in further detail due to its superior diester selectivity (68%) at high conversion (80%). The recyclability of Hf-BEA was probed by washing with toluene and reusing the material for five consecutive batch reactions at 120 °C. After a gradual conversion decrease from 80% (run 1) to 60% (run 4), the catalyst shows no further conversion drop in run 5 (Figure A4.8). Remarkably, calcination of the catalyst under dry air after run 4 improves the diester selectivity from 68% to 81% in addition to recovering the initial activity (Table 1, entry 4). This unique behavior prompted us to further investigate the stability and regeneration of Hf-BEA for EP coupling under flow conditions.

Table 2. Self-aldol condensation of ethyl pyruvate catalyzed by solid catalystsa

Selectivity (%)b Entry Cat. EP conv. (%) Diesters Side products

1 HfO2/Si-BEA 1 n.d. n.d.

2 ZrO2/Si-BEA 1 n.d. n.d.

3 SnO2/Si-BEA 1 n.d. n.d. 4 Si-BEA 1 n.d. n.d. 5 Si-BEAc 1 n.d. n.d. 6 DeAl-BEAd 20 23 20 7 Al-BEA 17 12 13 8 MgO 5 4 7

9 TiO2 6 18 8

10 ZrO2 6 5 16 11 HAP 2 n.d. n.d. 12 HT 13 13 10 13 Nonee 0 n.d. n.d. a Reaction conditions: 3 wt% ethyl pyruvate (EP) and 0.1 wt% 1,3,5-tri-tert-butylbenzene in toluene, 120 oC, reaction time 1 h. For entries 1-3 and 7, EP/metal = 195 (mol/mol); for entries 4-6 and 8-12, 10 mg catalyst per g EP solution. HAP, hydroxyapatite; HT, hydrotalcite (Mg/Al = 3 mol/mol).

 NnCi, i Selectivity =×i 100% − b NnCEP,,() EPo n EP , where NC,i is the number of carbon atoms in compound i, ni is the number of moles of compound i in the reaction mixture, and nEP,o is the initial moles of EP added. Diesters include compounds 3 - 5 in Scheme 1, and side products include compounds 6 - 13 in Scheme 1. n.d., c product selectivities were not quantified for reactions with conversion values lower than 3%. EP/H2O = d 2 (mol/mol) Dealuminated Al-BEA, with a Si/Al molar ratio of 906 and EP/H2O = 2 (mol/mol). e Reaction time 17 h.

Self-Aldol Condensation of Ethyl Pyruvate to Itaconic Acid Ester Analogs Table 3. Elemental analysis and micropore volumes for BEA zeolites used

Zeolites Si/M molar ratio Micropore volume (cm3 g-1)a Hf-BEAb 129 0.19 Hf-BEA-spent 129 0.17 Zr-BEA 104 0.18 Sn-BEA 97 0.18 Al-BEA 12.5 n/a a n/a, not applicable. b The batch used in all the flow studies and screening studies. 3.3. Flow studies of Hf-BEA

Figure 1 shows EP conversion and selectivities to diesters and side products catalyzed by Hf-BEA as a function of time on steam (TOS) in a packed bed reactor (see Figure A4.10 for yields). Hf- BEA was calcined in situ under dry air at 550 oC for 5 h prior to feeding the reactants. The catalyst bed was regenerated by calcination under identical conditions over the course of the experiment; each cycle is indicated by the dashed lines in Figure 1. The reactor was operated at 120 oC with a weight hourly space velocity (WHSV) of 32.2 h-1. Steady-state conversions of ca.

-1 60%, selectivities towards diesters of ca. 80%, and a total turnover number of 5110 molEP molHf over 132 h on stream were observed. Elemental analysis of the spent catalyst showed no detectable change in Hf content (Table 3), and powder X-ray diffraction confirmed that the long-

range crystallinity was preserved (Figure A4.13). N2 adsorption data of the spent catalyst showed a slightly smaller micropore volume (0.17 cm3 g-1) than that of the pristine catalyst (0.19 cm3 g-1) (Table 3, Figure A4.16).

After each calcination, transient and steady-state regimes are clearly observed in the reactivity data (Figure 1 and Figure A4.10). During the transient periods, the EP conversion decreases by ca. 10% over 10 h TOS, while the diester selectivity increases by ca. 20% and the selectivity to side products decreases by ca. 5%. The decrease in EP conversion and the increase in diester selectivity appear to be correlated with a decrease in the rates of side reactions that are initiated by hydrolysis (Scheme 2), indicating the deactivation of active sites responsible for hydrolysis reactions. The steady-state period, which accounts for the remaining 14-26 h on stream, is characterized by less than 5% decrease in EP conversion and stable selectivity to diesters (Figure 1). Calcination of the catalyst bed results in a partial recovery of hydrolysis activity, demonstrating that deactivation of hydrolysis active sites is partially reversible. The regeneration by calcination coupled with the 5 wt%

Chapter 4 loss observed by thermogravimetric analysis of the catalyst recovered after 132 h TOS suggests that interaction of organic molecules with the sites contributes to the apparent decrease in hydrolysis rates.

The increased diester yield coupled with the concurrent decrease of hydrolysis activity in the transient regimes suggests that different moieties in the active site ensemble catalyze the aldol condensation and hydrolysis reactions. Control batch reactions with Si-BEA (Table 2, entries 4-5) showed no activity, thus pointing to the involvement of Hf in generating these sites. In addition to Lewis acidity, spectroscopic evidence41 and computational studies42-44 suggest the presence of weak Brønsted acidity associated with the silanol moiety of monohydrolyzed framework heteroatom sites (i.e. open sites) in Lewis acidic zeolites. Thus, these weak Brønsted acid sites are likely responsible for hydrolysis, and selective deactivation of the acidic proton due to alkoxide formation could explain the transient deactivation of hydrolysis. Indeed, similar deactivation behavior has been observed in the flow studies of Lewis acidic zeolites.36,45 Lewis et al. showed through Sn- or Hf-BEA catalyzed tandem reactions between 5-(hydroxymethyl)furfural and alcohols that Lewis acid-catalyzed transfer hydrogenation activity remained unchanged, while Brønsted acid-catalyzed etherification activity decreased over time and could be partially recovered by calcination.36 Importantly, 13C{1H} CP MAS NMR of the spent catalyst showed alkoxide species formation on silanol groups in the zeolite.36

An important feature demonstrated in the flow data is the increase in selectivity toward diesters after the first regeneration by calcination (Figure 1, after 24 h TOS) that is consistent with the increase in selectivity observed after calcination in the batch studies. Specifically, the steady-state diester selectivity increases from 65% to 79% (yield: 42% to 48%), and the steady-state selectivity to side products decreases from 17% to 10% (yield: 11% to 5%). However, no further improvement in diester selectivity is observed after successive calcinations. This irreversible shift in selectivity after the first regeneration is indicative of permanent changes to the catalyst active sites that contribute to the decrease in hydrolysis activity, the increase in aldol condensation activity, or a combination of both.

Self-Aldol Condensation of Ethyl Pyruvate to Itaconic Acid Ester Analogs 100 EP conversion Diester selectivity 90 Side product selectivity

Conversion/Selectivity (%) 20

0 0 20 40 60 80 100 120 140 TOS (h)

Figure 1. EP conversion and selectivity to diesters and side products as a function of time on stream (TOS) for the flow reaction with Hf-BEA. Reaction conditions: 120 oC, 12 bar, 320 mg Hf-BEA, 3 wt% EP in toluene, flow rate 0.20 mL min-1, WHSV 32.2 h-1. Dashed lines represent regeneration of Hf-BEA: flush o o o with toluene at 120 C, dry under N2 at 150 C, and calcine with dry air at 550 C for 5 h. Diesters: 3 - 5 in Scheme 1. Side products: 6 - 13 in Scheme 1; distributions are shown in Figure A4.9. The dynamic behavior exhibited by Hf-BEA in flow is consistent with prior studies of Lewis acidic zeolites that suggest amorphization and site restructuring are responsible for changes in activity and selectivity after extended TOS.30,36,45 Specifically, for aqueous dihydroxyacetone isomerization, Lari et al. observed an 18% loss in crystallinity for Sn-BEA and a 16% decrease in tetrahedral coordination for Sn-MFI after 24 h TOS.45 Lewis et al. demonstrated with 119Sn MAS and 119Sn{1H} CP/MAS NMR that open and distorted framework Sn sites were generated in 119Sn-BEA during the transfer hydrogenation and etherification of 5-(hydroxymethyl)furfural with ethanol, while only closed sites were observed in the pristine catalyst.36 Studies by Boronat et al. and Bermejo-Deval et al. also suggest that the distribution of open and closed sites in Sn-BEA changes with post-synthesis treatment, e.g. calcination or exposure to reaction conditions.46,47

3.4. Na exchange studies of Hf-BEA

The molecular connectivity of active sites, i.e. open versus closed, has been suggested to affect

Chapter 4 catalytic activity and selectivity due to differences in Lewis acidity, flexibility, and functionality of the neighboring silanol group.37,46-49 Bermejo-Deval et al. demonstrated with Na-treated Sn-BEA that the cations likely exchange onto silanol groups proximal to Sn(IV) open sites.37 The presence of Na+ alters the glucose isomerization reaction pathway from a fructose-producing 1,2-hydride shift to a mannose-producing 1,2-carbon shift, and the effect can be reversed by washing with a

+ 37 H2SO4 solution to remove the Na cations. Computational studies indicate that the change in product distribution is a consequence of the electrostatic stabilization of the carbon shift relative to hydride shift caused by the presence of Na+ ions at the open Sn center.50 To probe the nature of the active sites in Hf-BEA, we conducted a preliminary study with Na-exchanged Hf-BEA. Specifically, two Na-exchanged Hf-BEA catalysts were prepared using methods adapted from the work of Bermejo-Deval et al. (see Section S1).37 These samples were denominated as Hf-BEA- Na-1 and Hf-BEA-Na-2, with Na/Hf molar ratios of 0.69 and 0.02, respectively (Table 4). Batch reactions catalyzed by Hf-BEA-Na-1 resulted in 15% EP conversion and ca. 34% selectivity to diesters (Table 4 and 5, entry 2). This significant reduction in activity and selectivity is unlikely a result of pore blockage (Section S2). Despite its low Na content, Hf-BEA-Na-2 resulted in a diminished EP conversion of 59% when compared to the 77% conversion generated by pristine Hf-BEA (Tables 2 and 4, entry 3). The catalyst performance and pore volume of both Na-

exchanged zeolites were recovered by washing with H2SO4 (Tables 4 and 5, entries 4-5). This reversible activity drop is likely caused by a strong interaction between Na+ and the active sites, which we investigate further using Fourier transform infrared (FTIR) spectroscopy with

deuterated acetonitrile (CD3CN) adsorption.

FTIR spectra of CD3CN adsorbed on pristine Hf-BEA show the characteristic bands seen with

37,46,51 -1 M-BEA zeolites (Figure 2a). The C≡N stretching vibrations of CD3CN at 2310 cm , 2275

-1 -1 cm , and 2268 cm are representative of CD3CN that is strongly bound to Lewis acid sites (i.e. tetrahedral Hf(IV) centers), coordinated to silanol groups, and physisorbed, respectively.37,46 The

-1 band at 2310 cm appears to be a single peak, which differs from previous reports of two bands corresponding to open and closed sites for Sn-BEA.37,46 However, these data are consistent with studies of Zr-BEA that show only one band in this region.41 Na exchange results in a decrease in the band at 2310 cm-1 and the appearance of a new band at 2284 cm-1, which has previously been

37 assigned as CD3CN adsorbed on Na-exchanged open sites for Sn-BEA. This change is more

Self-Aldol Condensation of Ethyl Pyruvate to Itaconic Acid Ester Analogs prominent for the material with a higher degree of Na exchange (Hf-BEA-Na-1, Figure 2b) than for the material with low Na loading (Hf-BEA-Na-2, Figure 2d). Controls with Na-exchanged Si-BEA (Si-BEA-Na) confirm that framework Hf(IV) is required to observe this feature (Figure 2e).52 After acid wash, the FTIR spectra resemble again those of the pristine material, demonstrating the reversibility of the Na exchange (Figure 2c).

Table 4. Aldol condensation of ethyl pyruvate catalyzed by Na-exchanged and acid-washed Hf-BEA at 120 oCa

Selectivity (%)b Si/Hf Na/Hf Micropore EP Side Entry Catalyst Treatment Diesters ratioc ratiod volume (cm3 g-1) conv. (%) products 1 Hf-BEA None 115 n/a 0.20 77 69 7 e 2 Hf-BEA-Na-1 NaNO3 (1 M) 116 0.69 0.17 15 34 10 3 Hf-BEA-Na-2f NaOH (pH = 10) 113 0.02 0.20 59 70 7 g 4 Hf-BEA-AW-1 H2SO4 (1 M) 115 n.d. 0.20 74 69 12 g 5 Hf-BEA-AW-2 H2SO4 (1 M) 119 n.d. 0.19 77 79 7 a Reaction conditions same as in Table 1, but with a different batch of Hf-BEA zeolite. b Selectivity same as defined in Table 1. c No significant changes in Si/Hf ratios after different treatments, indicating the treatments do not cause Hf leaching. d n/a, not applicable. n.d., not detected. e Hf-BEA after Na exchange treatment 1 f g with NaNO3 (1 M). Hf-BEA after Na exchange treatment 2 with NaOH (pH =10). Hf-BEA-Na-1 or Hf- o BEA-Na-2 washed with H2SO4 (1 M) at room temperature for 1 h, recovered and then calcined at 550 C. Full details on material synthesis available in Experimental section.

Table 5. Aldol condensation of ethyl pyruvate catalyzed by Na-exchanged and acid-washed Hf-BEA at 80 oCa

Selectivity (%)b -1 -1 c d Entry Catalyst EP conv. (%) Diesters Side products TOF (molEP ∙ molHf ∙ h ) TOF ratio 1 Hf-BEA 13 43 5 101 1 2 Hf-BEA-Na-1e 6 5 0 24 0.23 3 Hf-BEA-Na-2 9 33 4 74 0.73 4 Hf-BEA-AW-1 11 43 5 88 0.88 5 Hf-BEA-AW-2 12 46 6 94 0.93 a Reaction conditions: 3 wt% EP and 0.1 wt% 1,3,5-tri-tert-butylbenzene in toluene, EP/metal = 200

 NnCi, i Selectivity =×i 100% − o b NnCEP,,() EPo n EP (mol/mol), 80 C, 15 min. , where NC,i is the number of carbon atoms in compound i, ni is the number of moles of compound i in the reaction mixture, and nEP,o is the initial moles of EP added. Diesters include compounds 3 - 5 in Scheme 1. Side products include compounds 6 - 13 in Scheme 1. c TOF, turnover frequency, the number of moles of EP converted per mole of Hf per hour. d TOF ratio = TOF of a catalyst/TOF of Hf-BEA. e 30 min.

Chapter 4 Assuming that Na+ interacts with the open framework Hf(IV) site in a similar manner to that proposed by Bermejo-Deval et al. for Sn-BEA,37 we hypothesize that the open sites are the main active sites for aldol condensation. Thus, open site ensembles likely have twofold functionality: the Lewis acidic character of the heteroatom catalyzes aldol reaction, while the Brønsted acidic silanol group catalyzes hydrolysis. The change in the chemical environment of the open site upon

Na exchange—evidenced by the change in the FTIR band corresponding to CD3CN adsorbed on Hf sites from 2310 to 2284 cm-1—may explain the lower reactivity of Na-exchanged Hf-BEA for aldol condensation. However, the presence and catalytic contribution of Hf(IV) closed sites cannot be excluded. Detailed characterization and kinetics studies are currently underway to confirm these hypotheses.

Figure 2. FTIR spectra with decreasing CD3CN coverage on a) Hf-BEA, b) Hf-BEA-Na-1, c) Hf-BEA- AW-1, d) Hf-BEA-Na-2, e) Si-BEA-Na, and f) Si-BEA. Signals are referenced to the bare material and normalized by the combination and overtone modes of zeolite Si-O-Si stretches (1750–2100 cm-1). -1 -1 Reference lines are for physisorbed CD3CN (2268 cm ) and CD3CN adsorbed on silanols (2275 cm ), Na-exchanged open Hf(IV) sites (tentative, 2284 cm-1), and Lewis acidic Hf sites (2310 cm-1).

Self-Aldol Condensation of Ethyl Pyruvate to Itaconic Acid Ester Analogs 3.5. Hf-BEA catalyzed aldol reactions with other substrates

A variety of keto substrates can be used to produce diacids and diesters. In addition to ethyl pyruvate, we have also studied the reaction with other substrates, including ethyl glyoxylate (EG), ethyl acetoacetate (EAA) and pyruvic acid. As shown in Scheme 3, diesters of different carbon backbones can be produced by aldol reactions (a – c). A preliminary test shown in Table 6 demonstrates the wide substrate scope of Hf-BEA. We note that diethyl tartrate can be generated even with only EG. In light the formation of diethyl oxalate in the reaction mixture, we proposed a reaction mechanism as shown in Scheme A4.1, where a Cannizzaro reaction occurs to convert EG into ethyl glycolate and oxalic acid, mono ethyl ester, and ethyl glycolate provides the α-C necessary for aldol addition. Cannizzaro reactions are well known to be catalyzed by Lewis acids.53,54 Both reaction a in Scheme A4.1 and reaction d in Scheme 3 indicate that Hf-BEA can activate the α-C adjacent to a carboxylic acid ester group in addition to the α-C adjacent to a carbonyl group, rendering Hf-BEA an excellent catalyst for aldol addition for a wide range of substrates.

OH O O O O R O R HO R R O R=Ethyl O O (a) O OH

O O O O O O -H O R R R 2 R R O R O O O O O O (b) O OH O O

OH OO OO O R O R R O R O O O O OO OO -H2O OO O R R R R O O (c) O O O O OH O O R R

O O O O O O O O -H2O OH OH O OH HO OH HO OH (d) O O OH O O O

Scheme 3. Proposed production of diacids or diesters via C-C coupling of keto acids or esters, R = H, alkyl group. In addition, after a closer examination of the products from aldol condensation of EG and EAA,

Chapter 4 we found that only branched addition and condensation products are formed (reaction c in Scheme 3). Since the H on the C between two carbonyl groups is more acidic than the H on the terminal C in EAA, it reacts more readily and favors the formation of branched aldol products instead of the linear ones.

Table 6. Hf-BEA catalyzed aldol condensation between ethyl glyxolyate and other ethyl keto esters.a

Conversion (%) Selectivity (%)b Entry Reactantc EG/Hf ratio Reactant/EG ratio EG Reactant Diesters Carbon yield (%) 1 n.a. 239 n.a. 74 n.a. 79 90 2 EP 119 1.6 95 57 46 66 3 EAA 118 1.7 100 59 80 94 a Reaction conditions: 1.5 wt% ethyl glyxolyate (EG) and 0.1 wt% 1,3,5-tri-tert-butylbenzene in toluene,

 NnCi, i o b 120 C, 1 h, under 1 atm N2. Selectivity =×i 100% , where NC,i is the −+ − NnCE,G()() EGo , n EG Nnn CR , Ro , R

number of carbon atoms in compound i, ni is the number of moles of compound i in the reaction mixture, nEG,o is the initial moles of EG added, and R represents the other reactant. Diesters are the aldol addition products and condensation products in Scheme 3. c EAA, ethyl acetoacetate.

When pyruvic acid was used as a reactant, instead of C6 diacids, C5 compounds, such as 2- methylfumaric acid, 2-methylenesuccinic acid, and 2-methylsuccinic acid, are formed as shown by GC-MS after derivatization with either silylation or esterification. Since pyruvic acid is not stable at high temperature, especially at the presence of silicate materials,27 the decarboxylation of

pyruvic acid is likely occur to generate CO2 and acetaldehyde, which can react with the remaining pyruvic acid to form C5 aldol addition and condensation products. While more optimization of the reaction conditions is required, we have demonstrated the promise to directly convert keto acids into useful products with hydrophobic Lewis acidic zeolites.

4. Conclusions

In summary, we have developed a new approach to synthesize unsaturated dicarboxylic acid esters from keto esters via aldol condensation catalyzed by Lewis acidic zeolites that is applicable to a wide range of substrates. In particular, we demonstrate that Hf-BEA can catalyze the self-aldol condensation of EP in toluene to produce itaconic acid ester analogs. The catalyst is stable for 132 h TOS in a packed-bed reactor with a diester selectivity of ca. 80% at 120 oC. Na exchange

substantially decreases Hf-BEA activity and significantly alters the FTIR band for CDCN3

Self-Aldol Condensation of Ethyl Pyruvate to Itaconic Acid Ester Analogs adsorbed on the framework Hf(IV) sites, indicating a strong interaction between Na+ and the active sites. The open framework Hf(IV) site is hypothesized to possess strong Lewis acidity and weak Brønsted acidity that are responsible for the dual functionality of the catalyst. However, further investigation is needed to fully elucidate the nature of the active site and guide catalyst and reaction optimization.

Acknowledgements

Jennifer D. Lewis is gratefully acknowledged for her major contributions to this chapter.

References

(1) Werpy, T.; Petersen, G.; Aden, A.; Bozell, J.; Holladay, J.; White, J.; Manheim, A.; Eliot, D.; Lasure, L.; Jones, S. “Top value added chemicals from biomass. Volume 1-Results of screening for potential candidates from sugars and synthesis gas,” Department of Energy: Washington DC, 2004. (2) Dicarboxylic Acids, Aliphatic. Ullmann's Encyclopedia of Industrial Chemistry [Online]; Wiley-VCH Verlag GmbH & Co. KGaA, Posted NOV 19, 2014. http://onlinelibrary.wiley.com/doi/10.1002/14356007.a08_523.pub3/pdf. (3) Dicarboxylic Acids. Kirk-Othmer Encyclopedia of Chemical Technology [Online]; John Wiley & Sons, Inc., Posted Sep 17, 2010. http://onlinelibrary.wiley.com/doi/10.1002/0471238961.0409030110150814.a01.pub2/p df. (4) “WP 8.1. Determination of market potential for selected platform chemicals. Itaconic acid, succinic acid, 2,5-furandicarboxylic acid,” WEASTRA s.r.o., 2012. http://www.bioconsept.eu/wp-content/uploads/BioConSepT_Market-potential-for- selected-platform-chemicals_report1.pdf. (5) Okabe, M.; Lies, D.; Kanamasa, S.; Park, E. Y. Appl. Microbiol. Biotechnol. 2009, 84, 597- 606. (6) Steiger, M. G.; Blumhoff, M. L.; Mattanovich, D.; Sauer, M. Front. Microbiol. 2013, 4. (7) Straathof, A. J. Chem. Rev. 2014, 114, 1871-1908. (8) Mondala, A. H. J. Ind. Microbiol. Biotechnol. 2015, 42, 487-506. (9) Wang, Y. R.; Vogelgsang, F.; Roman-Leshkov, Y. ChemCatChem 2015, 7, 916-920. (10) Dutta, S.; Wu, L. L.; Mascal, M. Green Chem. 2015, 17, 2335-2338. (11) Liu, J.; Du, Z.; Lu, T.; Xu, J. ChemSusChem 2013, 6, 2255-2258. (12) Wang, Y. R.; Van de Vyver, S.; Sharma, K. K.; Roman-Leshkov, Y. Green Chem. 2014, 16, 719-726. (13) Casanova, O.; Iborra, S.; Corma, A. ChemSusChem 2009, 2, 1138-1144. (14) Flanagan, J. C. A.; Kang, E. J.; Strong, N. I.; Waymouth, R. M. ACS Catal. 2015, 5, 5328- 5332. (15) Wieschalka, S.; Blombach, B.; Bott, M.; Eikmanns, B. J. Microb. Biotechnol. 2013, 6, 87- 102. (16) Li, Y.; Chen, J.; Lun, S. Y. Appl. Microbiol. Biotechnol. 2001, 57, 451-459.

Chapter 4 (17) Schwartz, T. J.; Shanks, B. H.; Dumesic, J. A. Curr. Opin. Biotechnol. 2016, 38, 54-62. (18) Dusselier, M.; Van Wouwe, P.; Dewaele, A.; Makshina, E.; Sels, B. F. Energy Environ. Sci. 2013, 6, 1415-1442. (19) Il’Chenko, A.; Shcherbakova, V. Microbiology 2008, 77, 430-435. (20) Johnson, C. W.; Beckham, G. T. Metab. Eng. 2015, 28, 240-247. (21) Gathergood, N.; Juhl, K.; Poulsen, T. B.; Thordrup, K.; Jorgensen, K. A. Org. Biomol. Chem. 2004, 2, 1077-1085. (22) Dambruoso, P.; Massi, A.; Dondoni, A. Org. Lett. 2005, 7, 4657-4660. (23) Wang, Q.; Sun, Y.; Chen, H.; Li, Z.; Tao, F. Synthesis 2008, 2008, 589-593. (24) Lewis, J. D.; Van de Vyver, S.; Roman-Leshkov, Y. Angew. Chem. Int. Ed. 2015, 54, 9835- 9838. (25) Podrebarac, G. G.; Ng, F. T. T.; Rempel, G. L. Chem. Eng. Sci. 1997, 52, 2991-3002. (26) New Solid Acids and Bases: Their Catalytic Properties; Tanabe, K.; Misono, M.; Ono, Y.; Hattori, H., Eds.; Elsevier Science Publishing Company, Inc.: New York, NY, 1989, p 339. (27) Frey, H. “Thermal decomposition of pyruvic acid and its esters leading to CO2,” Univeristy of California, Berkeley, Radiation Laboratory., 1956. (28) Mahmoud, E.; Yu, J. Y.; Gorte, R. J.; Lobo, R. F. ACS Catal. 2015, 5, 6946-6955. (29) Van de Vyver, S.; Román-Leshkov, Y. Angew. Chem. Int. Ed. 2015, 54, 12554-12561. (30) Van de Vyver, S.; Odermatt, C.; Romero, K.; Prasomsri, T.; Román-Leshkov, Y. ACS Catal. 2015, 5, 972-977. (31) Holm, M. S.; Pagan-Torres, Y. J.; Saravanamurugan, S.; Riisager, A.; Dumesic, J. A.; Taarning, E. Green Chem. 2012, 14, 702-706. (32) Moliner, M.; Roman-Leshkov, Y.; Davis, M. E. Proc. Natl. Acad. Sci. U S A 2010, 107, 6164-6168. (33) Roman-Leshkov, Y.; Moliner, M.; Labinger, J. A.; Davis, M. E. Angew. Chem. Int. Ed. 2010, 49, 8954-8957. (34) Corma, A.; Nemeth, L. T.; Renz, M.; Valencia, S. Nature 2001, 412, 423-425. (35) Gunther, W. R.; Duong, Q.; Román-Leshkov, Y. J. Mol. Catal. A: Chem. 2013, 379, 294- 302. (36) Lewis, J. D.; Van de Vyver, S.; Crisci, A. J.; Gunther, W. R.; Michaelis, V. K.; Griffin, R. G.; Roman-Leshkov, Y. ChemSusChem 2014, 7, 2255-2265. (37) Bermejo-Deval, R.; Orazov, M.; Gounder, R.; Hwang, S. J.; Davis, M. E. ACS Catal. 2014, 4, 2288-2297. (38) Fang, W.; Chen, J.; Zhang, Q.; Deng, W.; Wang, Y. Chem. Eur. J. 2011, 17, 1247-1256. (39) De Clercq, R.; Dusselier, M.; Chr istiaens, C.; Dijkmans, J.; Iacobescu, R. I.; Pontikes, Y.; Sels, B. F. ACS Catal. 2015, 5, 5803-5811. (40) Sacia, E. R.; Balakrishnan, M.; Deaner, M. H.; Goulas, K. A.; Toste, F. D.; Bell, A. T. ChemSusChem 2015, 8, 1726-1736. (41) Sushkevich, V. L.; Vimont, A.; Travert, A.; Ivanova, I. I. J. Phys. Chem. C 2015, 119, 17633- 17639. (42) Montejo-Valencia, B. D.; Salcedo-Pérez, J.; Curet-Arana, M. C. J. Phys. Chem. C 2016, 120, 2176-2186. (43) Yang, G.; Zhou, L.; Han, X. J. Mol. Catal. A: Chem. 2012, 363, 371-379. (44) Yang, G.; Lan, X. J.; Zhuang, J. Q.; Ma, D.; Zhou, L. J.; Liu, X. C.; Han, X. W.; Bao, X. H. Appl. Catal., A 2008, 337, 58-65.

Self-Aldol Condensation of Ethyl Pyruvate to Itaconic Acid Ester Analogs (45) Lari, G.; Dapsens, P.; Scholz, D.; Mitchell, S.; Mondelli, C.; Pérez-Ramírez, J. Green Chem. 2016, 18, 1249-1260. (46) Boronat, M.; Concepcion, P.; Corma, A.; Renz, M.; Valencia, S. J. Catal. 2005, 234, 111- 118. (47) Bermejo-Deval, R.; Assary, R. S.; Nikolla, E.; Moliner, M.; Roman-Leshkov, Y.; Hwang, S. J.; Palsdottir, A.; Silverman, D.; Lobo, R. F.; Curtiss, L. A.; Davis, M. E. Proc. Natl. Acad. Sci. U S A 2012, 109, 9727-9732. (48) Sushkevich, V. L.; Palagin, D.; Ivanova, I. I. ACS Catal. 2015, 5, 4833-4836. (49) Harris, J. W.; Cordon, M. J.; Di Iorio, J. R.; Vega-Vila, J. C.; Ribeiro, F. H.; Gounder, R. J. Catal. 2016, 335, 141-154. (50) Christianson, J. R.; Caratzoulas, S.; Vlachos, D. G. ACS Catal. 2015, 5, 5256-5263. (51) Roy, S.; Bakhmutsky, K.; Mahmoud, E.; Lobo, R. F.; Gorte, R. J. ACS Catal. 2013, 3, 573- 580. (52) Yu, J.; Luo, J.; Zhang, Y.; Cao, J.; Chang, C.-C.; Gorte, R.; Fan, W. Microporous Mesoporous Mater. 2016, 225, 472-481. (53) Russell, A. E.; Miller, S. P.; Morken, J. P. J. Org. Chem. 2000, 65, 8381-8383. (54) Dusselier, M.; Van Wouwe, P.; De Smet, S.; De Clercq, R.; Verbelen, L.; Van Puyvelde, P.; Du Prez, F. E.; Sels, B. F. ACS Catal. 2013, 3, 1786-1800.

Chapter 5

Conclusions and Outlook

1. Conclusions

Biomass plays a critical role as a renewable alternative to the inevitably depleting fossil fuel resources and provides the feedstock for chemicals production. The drastic differences on the molecular level between the fossil fuel and biomass feedstocks have fostered a growing need for technologies to produce renewable bio-based chemicals. Among the extensive ongoing efforts, this thesis provides a unique contribution on the chemocatalytic conversion of biomass-derived molecules into mono- and dicarboxylic acids and esters, which are the key intermediates to various important chemicals, especially biodegradable polymers. Targeting at gluconic acid, succinic acid, and itaconic acid analogs and their esters, this thesis has devised catalytic systems to effectively synthesize these chemicals and provided fundamental insights into the activity, selectivity and stability of the catalysts.

The molecular structures of the starting compounds from biomass have directed the selection of reaction pathways to achieve target molecules. Consequently, very different catalysts were investigated, including supported noble metal nanoparticles for aerobic oxidation of monosaccharides in aqueous environment, homogeneous Lewis and Brønsted acids for Baeyer- Villiger oxidation of linear ketones, and Lewis acidic zeolites for aldol condensation of keto esters. The key factors that affect the performance of these catalysts in liquid-phase biomass processing have been identified and studied.

We have shown that gold (Au) nanoparticles (NPs) supported on metal oxides are active and selective catalysts for the oxidation of glucose to gluconic acid under base-free conditions. The main challenge lies in the deactivation of Au catalysts under the reaction conditions. We evaluated the catalytic performance and stability of Au NPs supported on metal oxides for the oxidation of glucose to gluconic acid under unadjusted pH and acidic conditions. The study provides insights into the irreversible deactivation of the catalysts caused by leaching and hydrothermal sintering of Au NPs, as well as reversible deactivation caused by adsorption of reaction species. A key strategy proposed to improve the stability of support Au NPs is to lower the surface density of Au NPs, which not only increases the travel distance for Au NPs to migrate and coalesce, but also results in

Chapter 5 better anchoring of Au NPs on the oxygen vacancies of the metal oxide supports, enhancing the both stability and activity of the catalyst. Further studies showed that carboxylic acid group exhibits the highest inhibition effect among all the functional groups existing in the reaction mixture, and calcination under air can recover the activity loss caused by adsorption of reaction species. In the presence of competitive inhibition caused by common by-products in the glucose feed stream from acid-catalyzed cellulose hydrolysis, Au NPs supported on nanosized metal oxides is active, enabling the sequential one-pot combination of cellulose hydrolysis and glucose oxidation.

Levulinate derivatives are an attractive platform for the production of renewable chemicals. The Baeyer-Villiger (BV) oxidation of methyl levulinate into dimethyl succinate with peroxides was achieved using Brønsted and Lewis acid catalysts under mild conditions. Importantly, the impact of different reaction parameters on the oxidation selectivity was investigated, providing knobs to tune the reaction selectivity. Selectivities to succinate and acetate derivatives of ca. 60% and 40%, respectively, were obtained with strong Brønsted acids in methanol. While the molecular structure (i.e., carbon chain length and branching around the C=O group) and the oxidant type affect the product distribution, solvent choice has the strongest impact on changing the location of oxygen insertion into the carbon backbone. Specifically, switching the solvent from methanol to heptane resulted in a decrease in the succinate/acetate ratio from 1.6 to 0.3. In contrast to Brønsted acids, we demonstrate that for water-tolerant Lewis acidic triflate salts, the nature of the metal cation, particularly its size, significantly affects the reaction selectivity.

Furthermore, we have developed a novel approach to synthesize unsaturated dicarboxylic acid esters from keto esters via aldol condensation catalyzed by Lewis acidic zeolites. We demonstrated hafnium-containing BEA (Hf-BEA) zeolites catalyze the condensation of ethyl pyruvate into diethyl 2-methyl-4-oxopent-2-enedioate and diethyl 2-methylene-4-oxopentanedioate (an itaconic acid ester analog) with a selectivity of ca. 80% at ca. 60% conversion in a packed-bed reactor. The catalyst is stable for 132 h on stream, reaching a turnover number of 5110 molEP

-1 molHf . Analyses of the dynamic behavior of Hf-BEA under flow conditions and studies with Na- exchanged zeolites suggest that Hf(IV) open sites possess dual functionality for Lewis and Brønsted acid catalysis.

Conclusions and Outlook 2. Future directions

2.1. Stabilization of supported metal nanoparticles for liquid-phase biomass processing

To design and synthesize an active and stable supported metal nanoparticle catalyst for liquid- phase biomass processing is critical but challenging. Recent work by O’Neill et al.1 and Lee et al.2 has demonstrated that metal oxide decoration on the supported base metal nanoparticles via atomic layer deposition can effectively stabilize the catalyst in aqueous phase hydrogenation of biomass-derived molecules by eliminating the common deactivation via sintering and leaching. Due to versatility of this approach, it can be potentially applied to stabilizing supported Au NPs. On the other hand, the strong metal-support interaction (SMSI) has been known between reducible metal oxides and noble metals.3-5 A recent report by Tang et al. showed similar metal- support interaction even exist between Au NPs and hydroxyapatite (HAP), a nonoxide.6 Partial encapsulation of Au NPs after calcination with air at 500 oC significantly improved the stability of Au/HAP for solvent-free selective oxidation of benzyl alcohol.6

Our preliminary studies have also demonstrated the potential to incorporate the SMSI effect into catalyst design and synthesis to enhance the Au catalyst stability. One challenge is to identify the right treatment conditions under which the activity of the catalyst is preserved and even improved and the stability of the catalyst is enhanced. Another challenge is about the evaluation of the stability of the catalyst. In the case of glucose oxidation, different deactivation mechanisms are concomitant through the reaction. Right experimental design is necessary to decouple these effects. One possible approach is to evaluate the catalyst in a flow reactor, since the inhibition of the reaction species can be controlled with the flow reaction conditions.

2.2. Redispersion of gold catalysts

Many approaches have been developed to redisperse supported noble metal NPs in order to recover their activity.7 The basic principle is to disintegrate the large particles by subjecting them under reactive conditions, e.g. O2 at high temperature, to change the surface composition of the metal NPs and cause fragmentation followed by migration.7,8 Similar gas-phase approaches can be directly used to redisperse supported noble metal nanoparticles sintered under liquid phase reactions, or new approaches can be designed in the liquid phases. In this thesis, Au/nCeO2 has shown higher stability under pH 1.6 than under uncontrolled pH (starting pH about 4.0 and final

Chapter 5 pH about 2.5).9 Given the higher solubility of Au species in more acidic solutions and the large

number of oxygen vacancies on nCeO2, one reasonable hypothesis is that Au may get better redispersed during the recyclability test, a smaller particle size distribution is maintained and a higher fraction of the Au NPs is anchored on the oxygen vacancies during redispersion under pH 1.6 than under uncontrolled pH. But again, the challenges are to optimize the treatment conditions to achieve a significant redispersion.

2.3. Mechanistic understanding of selectivity control in BV oxidation

BV oxidation of levulinates provides a complementary reaction schemes to produce important chemicals. Experimentally, we have demonstrated the influence of different reaction parameters on the reaction activity and selectivity.10 However, a thorough mechanistic understanding will better inform fine reaction control. Recently, Mascal group has demonstrated the conversion of levulinic acid into succinic acid and 3-hdyroxypropanoic acid using excessive hydrogen peroxide in a concentrated trifluoroacetic acid or sulfuric acid, and a concentrated aqueous potassium hydroxide solution, respectively.11,12 Contrarily, we have showed that a catalytic amount of sulfuric acid (a levulinic acid/sulfuric acid molar ratio of 10) cannot catalyze the BV oxidation of levulinic acid in water, consistent with observation by Choudhary et al.13 Therefore, the conversion of levulinic acid in concentrated strong acids is surprisingly interesting. In light of the significant solvent effect observed in our study, a theoretical study on the interaction between different solvents and the reactive species will likely shed light on the reactivity and selectivity control. Our preliminary studies suggest that the ionic pathway should be considered in protic solvents in the presence of strong Brønsted acids, and explicit solvent molecules should be involved in the computation to account for the interaction between solvent molecules and the reactive species. Additionally, charge distribution analysis and hydrogen bonding analysis should be informative.

2.4. Active site characterization for aldol condensation in Hf-BEA zeolites

Our Na exchange studies suggest the important role of Hf(IV) open site in the reaction.14 However, due to the lack of effective characterization techniques for Hf(IV), it has been challenging to elucidate the structure of the active sites in Hf-BEA. Recently, our group has developed a new characterization approach based on solid-state nuclear magnetic resonance

Conclusions and Outlook spectroscopy that can probe different Lewis acidic sites in the zeolites and is not heteroatom- specific. This technique can be easily extended for characterization of the Lewis acidic sites in Hf- BEA to illuminate the active site structure.

2.5. Expanding the substrate scope for aldol condensation catalyzed by Lewis acidic zeolites

The ultimate goal is to couple the biocatalytic production of keto acids and esters with the chemocatalytic synthesis of dicarboxylic acids and esters. We have demonstrated the capability of these Lewis acidic zeolites for aldol condensation of various keto esters, and showed the promise to directly convert keto acids. However, the first challenge is the thermal instability of many keto acids. Tuning the reaction conditions, e.g. lower reaction temperature, may be necessary for the effective conversion of keto acids. Another challenge is created by the presence of different impurities in the aqueous biosynthetic feed stream, e.g. buffer salts. One possible solution is a biphasic system (e.g. water-toluene), where keto acids are in the aqueous layer and the zeolites stay in the organic layer, minimizing the impact of water and the impurities. Our preliminary work has shown silylation of these zeolites can lead to higher hydrophobicity while preserving the framework structure and maintaining the activity.

Furthermore, Lewis acidic zeolites with different framework topologies could be developed to expand the substrate scope for aldol condensation. First, different pore confinement comes with different topologies and could serve to accommodate larger molecules or enhance the shape selectivity with specific pore structures. Secondly, different topologies could also offer an additional approach to control the reactivity, as the reaction activity originates from the Lewis acidity of the framework heteroatom, which can also be influenced by the zeolite framework structure.15 Preliminary studies in our group has shown the incorporation of Hf(IV) in MFI topology, which is indeed active for aldol reactions.

2.6. Further processing of dicarboxylic acids and esters

One carbonyl group and one C=C bond or one hydroxyl group are present in the aldol reaction products in addition to the two dicarboxylic acid or ester groups. These functional groups will allow further modification of the polymer synthesized. Moreover, if a direct replacement of a monomer used in the current polymer synthesis is preferred, it can be achieved, for example, via

Chapter 5 hydrodeoxygenation of these products in liquid phase. For example, hydrodeoxygenation of the condensation products of glyoxylate and pyruvate can produce glutarate. Catalysts developed in our lab, such as ruthenium on carbon (Ru/C) or platinum coated tungsten carbide (Pt@WC) core-shell materials, are promising candidates. Furthermore, it is interesting and important to investigate the polymerization of these molecules from aldol addition or hydrodeoxygenation, since not only polymers with new properties may be produced, but also the insights into the monomer properties during the polymerization process will be valuable guidance for the monomer synthesis.

2.7. New targets in the top 12 building block list

O OH O O O OH OH levulinic acid HO O O HO O -H O -H O formaldehyde 2 2

O O O O O O oxepane-2,5-dione 4-acetyldihydrofuran-2(3H)-one

Baeyer-Villiger oxidation + hydrolysis

O O O OH HO 3-Hydroxybutyrolactone

Scheme 1. Synthesis of 3-hydroxybutyrolactone. Based on the knowledge obtained in this thesis, pathways to synthesize 3-hydroxybutyrolactone and its esters are proposed as shown in Scheme 1. Aldol reaction of levulinic acid and formaldehyde followed by lactonization can produce oxepane-2,5-dione and 4- acetyldihydrofuran-2(3H)-one (thermodynamically favored). The former can be used as a monomer for the synthesis of biodegradable (co)polymers.24 The latter is an isomer of alpha- acetyl-gamma-butyrolactone, an important pharmaceutical intermediate.25 BV oxidation with peroxide of 4-acetyldihydrofuran-2(3H)-one followed by hydrolysis can produce acetic acid and 3-hydroxybutyrolactone. In this case, our preliminary tests have shown the formation of the two lactonization products after aldol addition with Hf-BEA. BV oxidation of 4-acetyldihydrofuran-

Conclusions and Outlook 2(3H)-one can be catalyzed by strong Brønsted acids or Lewis acids. The challenges mainly are the selectivity control in the reaction series, namely, how to steer the addition location of formaldehyde on levulinic acid, how to encourage lactonization instead of dehydration, and how to make sure the ring structure migrates instead of methyl group, etc.

3. Outlook

Throughout this thesis, we have developed catalytic systems to effectively convert biomass- derived molecules into mono- and dicarboxylic acids and esters and provided fundamental understanding of the performance of the catalysts. Although we focused on mainly three target molecules, the insights gained in these representative systems can be generalized in liquid-phase processing of biomass-derived molecules and open up new avenues for chemocatalytic synthesis of more mono- and dicarboxylic acids and esters from biomass.

3.1. Top 12 building blocks and beyond

Biomass-derived molecules are highly functional and pose great challenges in selective conversion. However, access to a large product space is also granted by different transformation pathways of a variety of functional groups. In this thesis, we demonstrated mono- and dicarboxylic acids and esters can be generated via transformation of very diverse biomass-derived molecules: selective oxidation of aldehyde group of glucose forms carboxylic acid group while preserving the carbon backbone, BV oxidation employs oxygen insertion to break methyl levulinate into valuable fragments, and aldol condensation of ethyl pyruvate assembles carboxylic acid groups from two molecules into one molecule.

Diverse starting compounds can be obtained via current biological and chemical biomass processing, such as glucose and 5-hydroxymethylfurfural from acid-catalyzed cellulose hydrolysis, furans from biomass pyrolysis, aromatics from lignin depolymerization, and amino acids from enzymatic conversion, etc. Molecule structures that are not feasible based on hydrocarbon feedstocks can become possible by converting, cutting and/or stitching molecules derived from biomass. The key to the target molecules is activation of a specific functional group of reactants, which necessitates a fundamental understanding of the catalytic mechanism, and the capabilities and limits of the catalyst.

Chapter 5 3.2. A process perspective

The work in this thesis is a local optimization in the context of biorefinery. Constraints imposed by the whole process from raw biomass to target molecules have motivated the work on base-free oxidation. Then conversion of levulinates is driven by new opportunities created by progress made in the field. Finally aldol condensation of ethyl pyruvate has stemmed from the discoveries of new catalyst capacities to take advantage of the feedstocks provided by biological transformation and to better integrate biocatalytic and chemocatalytic pathways.

To keep a process perspective is important. A reaction can only make a real difference in the world if the whole process is economically feasible and environmental friendly. Target applications set the property requirements for molecules, which actually allow flexibility in terms of the molecular structures. Therefore, the target molecules can be a direct or functional replacement of a molecule widely used in industry. Moreover, not only can the process perspective of biorefinery set the targets for the design of a catalytic system, but also can the advance in catalysis create new possibilities and even change the paradigms of biorefinery.

References

(1) O'Neill, B. J.; Jackson, D. H.; Crisci, A. J.; Farberow, C. A.; Shi, F.; Alba‐Rubio, A. C.; Lu, J.; Dietrich, P. J.; Gu, X.; Marshall, C. L. Angew. Chem. Int. Ed. 2013, 52, 13808- 13812. (2) Lee, J.; Jackson, D. H.; Li, T.; Winans, R. E.; Dumesic, J. A.; Kuech, T. F.; Huber, G. W. Energy Environ. Sci. 2014, 7, 1657-1660. (3) Tauster, S.; Fung, S.; Baker, R.; Horsley, J. Science 1981, 211, 1121-1125. (4) Tauster, S. Acc. Chem. Res. 1987, 20, 389-394. (5) Liu, X.; Liu, M.-H.; Luo, Y.-C.; Mou, C.-Y.; Lin, S. D.; Cheng, H.; Chen, J.-M.; Lee, J.-F.; Lin, T.-S. J. Am. Chem. Soc. 2012, 134, 10251-10258. (6) Tang, H.; Wei, J.; Liu, F.; Qiao, B.; Pan, X.; Li, L.; Liu, J.; Wang, J.; Zhang, T. J. Am. Chem. Soc. 2015. (7) Morgan, K.; Goguet, A.; Hardacre, C. ACS Catal. 2015, 5, 3430-3445. (8) Ouyang, R.; Liu, J.-X.; Li, W.-X. J. Am. Chem. Soc. 2013, 135, 1760-1771. (9) Wang, Y. R.; Van de Vyver, S.; Sharma, K. K.; Roman-Leshkov, Y. Green Chem. 2014, 16, 719-726. (10) Wang, Y. R.; Vogelgsang, F.; Roman-Leshkov, Y. ChemCatChem 2015, 7, 916-920. (11) Dutta, S.; Wu, L. L.; Mascal, M. Green Chem. 2015, 17, 2335-2338. (12) Wu, L.; Dutta, S.; Mascal, M. ChemSusChem 2015, 8, 1167-1169. (13) Choudhary, H.; Nishimura, S.; Ebitani, K. Appl. Catal., A. 2013, 458, 55-62. (14) Wang, Y.; Lewis, J. D.; Román-Leshkov, Y. ACS Catal. 2016, 6, 2739-2744.

Conclusions and Outlook (15) Montejo-Valencia, B. D.; Salcedo-Pérez, J.; Curet-Arana, M. C. J. Phys. Chem. C 2016, 120, 2176-2186. (16) Wildeman, S. M. D.; Sonke, T.; Schoemaker, H. E.; May, O. Acc. Chem. Res. 2007, 40, 1260-1266. (17) Zhu, D.; Hua, L. Biotechnol. J 2009, 4, 1420-1431. (18) Wei, Y.; Wang, C.; Jiang, X.; Xue, D.; Li, J.; Xiao, J. Chem. Commun. 2013, 49, 5408- 5410. (19) Wei, Y.; Wang, C.; Jiang, X.; Xue, D.; Liu, Z.-T.; Xiao, J. Green Chem. 2014, 16, 1093-1096. (20) Huang, Y. B.; Dai, J. J.; Deng, X. J.; Qu, Y. C.; Guo, Q. X.; Fu, Y. ChemSusChem 2011, 4, 1578-1581. (21) Du, X. L.; He, L.; Zhao, S.; Liu, Y. M.; Cao, Y.; He, H. Y.; Fan, K. N. Angew. Chem. 2011, 123, 7961-7965. (22) Ogiwara, Y.; Uchiyama, T.; Sakai, N. Angew. Chem. Int. Ed. 2015. (23) Touchy, A. S.; Hakim Siddiki, S.; Kon, K.; Shimizu, K.-i. ACS Catal. 2014, 4, 3045- 3050. (24) Latere, J.-P.; Lecomte, P.; Dubois, P.; Jérôme, R. Macromolecules 2002, 35, 7857-7859. (25) Koehler, G.; Uhlenbrock, W., US 5789603, 1998.

Appendix

Chapter 2

Activity of atomically dispersed gold species

Series of work by Flytzani-stephanopoulos et al. has demonstrated that atomically dispersed Au atoms on different metal oxide supports are active and stable catalysts for low temperature water- gas shift reactions.1 These metal oxide supported single Au atoms are synthesized via cyanide leaching of Au catalysts prepared using the deposition-precipitation method. It is shown that Au cations are strongly bound to the supports and resistive to cyanide leaching, whereas metallic Au

- nanoparticles can react with cyanide to form [Au(CN)2] and be dissolved and removed from the supports. To explore the possibility of Au adatoms well dispersed on metal oxides as stable catalysts for glucose oxidation under base-free aqueous environment, we has first determined whether these gold adatoms are active for glucose oxidation. Leached Au/nCeO2 samples were prepared (see Experimental section in Chapter 2) and tested. However, only 3% glucose conversion was obtained after 20 min compared to a conversion of over 30% for non-leached

-1 Au/nCeO2 under identical reaction conditions: glucose 167 mmol L , 12 mL, pO2 = 2.3 bar, 65 °C, glucose/Au = 420. Under the reaction conditions, the leached samples quickly turned into purple from the color of the support (grey yellow), indicating the formation of gold nanoparticles.

Further reduction under H2 of the leached Au/nCeO2 showed comparable activities as the fresh catalyst with the same glucose/Au molar ratio, indicating metallic Au species are mainly responsible for the glucose oxidation activity of the catalyst. Indeed, Guan et al. also have shown the sole presence of Au cations cannot catalyst the benzylacohol oxidation in toluene.2 Therefore, despite the high activity and stability of these single gold atom centers for gas phase reactions, they are not stable under aqueous reaction conditions, rendering them poor catalysts for oxidation reaction in liquid phase.

Stabilization of Au NPs via strong metal-support interaction

Strong metal-support interaction (SMSI) is well known to affect the properties of supported metal catalysts. Tauster et al. have pioneered the studies on reductive SMSI effects between reducible

3,4 transition metal oxides (e.g. TiO2, Ta2O5, etc.) and noble metals (e.g. Pt, Pd, Rh, etc.). Upon reduction, metal oxide decoration can be formed on the surface of the noble metals, which can

Appendix block the active sites or form new interfaces that have special catalytic activity.3,4 SMSI has also been reported for Au catalyst, both under reductive5 and oxidative conditions.6 As the technology for transmission electron miscopy (TEM) has been significantly advanced, high resolution TEM (HRTEM),6 environmental TEM (ETEM)7 and high-angle annular dark-field scanning TEM (HAADF-STEM)8 are widely available and used for characterization of these SMSI phenomena. Since SMSI is characterized by the (partially) encapsulation of metal NPs by the support metal oxides, we hypothesize that the partial encapsulation by the metal oxide support can mitigate the sintering of Au NPs and greatly enhance the stability of the catalysts. The challenges lie in the optimization of the post-treatment conditions so that the catalyst activity is preserved or even enhanced.

A C

E F

o Figure A2.1. HRTEM images of Au/μCeO2 (calcined under air at 425 C for 1 h: A, initial, B, after 30 o sec under a strong beam), Au/μZrO2 (reduced under H2 at 225 C for 4 h: C, initial, D, under a strong o beam), Au/μTiO2 (reduced under H2 at 225 C for 4 h: E, initial, F, under a strong beam).

First, we confirmed the SMSI effect between Au NPs and metal oxides used, i.e., CeO2, ZrO2 and TiO2 with HRTEM. As shown in Figure A2.1, after exposure under the strong beam for a short period (~ 30 sec), the encapsulation of Au NPs occurs, which can be reversed once the

Appendix sample moves out of the beam center, consistent with the observation by Kuwauchi et al.7 and Akita et al..8

Second, we studied the effect of calcination temperature on the reactivity of the catalysts. For

Au/nCeO2 with different post-synthesis treatments, the Au NP sizes increase slightly as shown in

o Table A2.1. However, the TOFs of Au/nCeO2 calcined under air at 325 and 425 C were only 70% and 30% of the TOFs of original catalysts that were reduced at 225 oC (entries 3 and 4), respectively. There seems no negative effect on the catalytic activity for calcination at 225 oC, which even showed 14% increase in TOF. Similar trends were also observed for Au/nTiO2 and

Au/nZrO2 (See Figure A2.1 for R225C and O225C). One hypothesis is that if (partial) encapsulation indeed occurs and increases as temperature increases, the total Au surfaces exposed to reactant will decrease even if corrected by dispersion, and then calcination at high temperature may cause low usage efficiency of Au. This potential effect could also contribute to the low

o activity of Au/μCeO2 after regeneration at 325 C in air compared to the fresh catalyst. We note that Au/nTiO2 and Au/nZrO2 showed higher activities than Au/nCeO2 in Figure A2.2, consistent with the batch studies in Table 1 at a lower glucose/Au ratio under uncontrolled pH.

a Table A2.1. Catalytic activity of Au/nCeO2 with different treatments

b c d -1 e Entry Au/nCeO2 D (nm) Dispersion TOF (molglu · molsurface Au · h) TOF/ TOFR250C 1 R225C 1.6 0.66 172 1 2 O225C 1.5 0.67 202 1.1 3 O325C 1.7 0.63 115 0.7 4 O425C 1.8 0.60 47 0.3 a -1 Reaction conditions: glucose 167 mmol L , 12 mL, pO2 = 2.3 bar, 65 °C, glucose/Au = 140, unadjusted b o pH. 0.8 wt% Au loading determined by ICP-AES. R225C represents reduction under H2 at 225 C for 4 h; O225C, O325C and O425C represent calcination under air at 225, 325, 425 oC for 1 h after o reduction under H2 at 225 C for 4 h, respectively. This batch of Au/nCeO2 is less active than previous batches. c Mean Au particle diameter determined by TEM analysis and based on a count of at least 100 Au NPs. d Dispersion of the Au NPs calculated assuming a predominantly cuboctahedral structure of the Au NPs (see the Experimental section in Chapter 2). e TOF, turnover frequency, calculated after 20 min reaction with a maximum glucose conversion of 17%.

Appendix

100 Au/nTiO2 R250C Au/nTiO2 O250C Au/nZrO2 R250C 90 Au/nZrO2 O250C Au/uCeO2 Au/nCeO2 80

30 Glucose conversion Glucose conversion (%) 20

0 0 102030405060708090 Time (h)

Figure A2.2. Glucose conversion for 0.7 wt% Au/nTiO2, 0.3 wt% Au/nZrO2, 0.02 wt% Au/μCeO2, and -1 0.4 wt% Au/nCeO2 as a function of reaction time. Reaction conditions: glucose 167 mmol L , 12 mL, pO2 = 2.3 bar, 65 °C, glucose/Au = 9000. As the reaction profiles in Figure A2.2 are affected by deactivation caused by both the increase of particle size and the adsorption of reactive species, it is necessary to decouple these two effects to study each pathway. To test the catalysts in a fixed bed flow reactor will be helpful to alleviate the inhibition caused by adsorption and enable the investigation of the true effect of the change in Au particle size.

Appendix Chapter 3

1 Table A3.1. H NMR (400 MHz, CD3OD) chemical shifts of major compounds in the reaction

Compound Chemical Shifts δ (ppm)

A, 3.67 (s, 3H) B B, 2.56 (t, 2H)

D C A C, 2.81 (t, 2H) D, 2.18 (s, 3H)

B A, 3.70 (s, 6H) B, 2.64 (s, 4H) A

A, 3.62 (s, 3H) A B B, 2.05 (s, 3H)

A, 3.69 (s, 3H) D B B, 2.55 (t, 2H) C A C, 3.81 (t, 2H)

D, missing in CD3OD A, 3.70 (s, 3H) B B, 2.59 (t, 2H) D C A C, 3.67 (t, 2H) D, 3.35 (s, 3H) A, 3.70 (s, 3H) B B, 2.66 (d, 2H) C A C, 4.81 (t, 1H) D D, 3.37 (s, 6H)

B A, 3.71 (s, 6H) A B, 3.50 (s, 2H)

Appendix

1 Figure A3.1. H NMR spectrum of the reaction mixture in methanol diluted with CD3OD.

1000000

500000

Abundance (Arb.)

0 3.0 9 10 11 12 13 Time (min)

Figure A3.2. A typical GC chromatogram of methyl levulinate oxidation with H2O2 in methanol.

Appendix

Figure A3.3. Mass spectra of the substrate and products in methyl levulinate oxidation with H2O2 in methanol, obtained from Agilent 5975C MSD.

Appendix

Scheme A3.1. Oxidation of aliphatic ketones using H2O2 in methanol.

Scheme A3.2. Acid-catalyzed decomposition of acetates formed in methanol. Products of further oxidation of alcohols and/or methyl esters with H2O2 are not listed.

Appendix Table A3.2. Decomposition of acetates formed under reaction conditionsa

Entry Substrate Conv. (%) MA selectivity (%) 1 Propyl acetate 100 96 2 Butyl acetate 100 98 3 Sec-butyl acetate 84 92 4 Tert-butyl acetate 100 102 a -1 Reaction conditions: acetate 61 mmol L in methanol, acetate/H2O2 molar ratio = 1:2, acetat/:p-TsOH = 10:1, 80 °C, 6 h.

a Table A3.3. Oxidation of acyclic aliphatic ketones using H2O2 in methanol with p-TsOH

b c d Entry Substrate Conv. (%) SM (%) SA (%) EH2O2 (%) 1 2-pentanone 61 55 28 63 2 2-hexanone 59 45 29 59 3 3-methyl-2-pentanone 59 48 16 53 4 3,3-dimethyl-2-butanone 53 67 10 77 5e 2-pentanone 35 64 32 35 6f 2-hexanone 38 47 27 28 a -1 Reaction conditions: substrate 61 mmol L in methanol, substrate/H2O2 molar ratio = 1:2, substrate/acid b c molar ratio = 10:1, 80°C, 6 h. SM is the selectivity to methyl ester shown in Scheme A3.1. SA is the selectivity to acetate shown in Scheme A3.1. Acetates formed further decompose as shown in Scheme A3.2. The formation of acetate products is based on the amount of MA (Table A3.2), accounting for over d 92 % of the yield of acetates. EH2O2, H2O2 efficiency = moles of substrate converted/moles of H2O2 consumed × 100 %. e 1-propanol used as solvent. f 1-butanol used as solvent.

a Table A3.4. Oxidation of acyclic aliphatic ketones using H2O2 in methanol with Amberlyst-15

Entry Substrate Conv. (%) SM (%) SA (%) 1 2-pentanone 57 42 35 2 2-hexanone 56 39 33 3 3-methyl-2-pentanone 51 34 21 4 3,3-dimethyl-2-butanone 34 60 6 a Reaction conditions: catalyst Amberlyst-15, substrate 61 mmol L-1 in methanol, substrate/peroxide molar ratio = 1:2, substrate/acid molar ratio = 1:1, 80°C, 6 h. SM, SA seen in Table A3.3.

Appendix Table A3.5. Oxidation of acyclic aliphatic ketones with TBHP in methanol with Amberlyst-15a

Entry Substrate Conv. (%) SM (%) SA (%) 1 2-pentanone 51 38 32 2 2-hexanone 40 34 27 3 3-methyl-2-pentanone 43 17 14 4 3,3-dimethyl-2-butanone 43 14 17 a Reaction conditions seen in Table A3.4. SM and SA seen in Table A3.3.

Table A3.6. Oxidation of acyclic aliphatic ketones using TBHP in heptane with Amberlyst-15a

Entry Substrate Conv. (%) SM (%) SA (%) 1 2-pentanone 49 12 25 2 2-hexanone 39 14 63 3 3-methyl-2-pentanone 43 16 36 4 3,3-dimethyl-2-butanone 44 6 15 5b Methyl levulinate 36 14 42 a -1 Reaction conditions: substrate 61 mmol L in heptane, substrate/H2O2 molar ratio = 1:2, substrate/acid molar ratio = 1:1, 80°C, 6 h. SM and SA seen in Table A3.3. We note that the total carbon yield is less than 100 % in these experiments. To rule out the possibility that different product ratios are obtained due to preferential degradation of methyl esters over acetates in heptane, all the methyl esters were subjected to identical oxidation conditions. In all cases, product degradation of less than 5 % was observed confirming that the results are mainly due to primary reactions. bML/acid molar ratio = 10:1.

Table A3.7. Brønsted acids for methyl levulinate oxidation in methanol with aqueous H2O2 as oxidanta

Entry Cat. Conv. (%) SM (%) SA (%) EH2O2 (%) 1 p-TsOH 56 61 39 44

2 MeSO3H 48 54 42 39

3 H2SO4 54 61 37 38 4 TfOH 49 61 29 37 b 5 CH3COOH <1 n.d. n.d. n.d. 6 Amberlyst-15 13 60 36 47 7b Amberlyst-15 67 61 39 38 8c p-TsOH 24 55 29 46 9 Al-H-BEA 47 22 32 26 a b c Reaction conditions, SM, SA, and EH2O2, same as in Table A3.3. ML/acid molar ratio = 1. Polymer- supported p-TsOH purchased from Sigma Aldrich, ML/acid molar ratio = 9:1.

Appendix

Table A3.8 Metal triflates for methyl levulinate oxidation in methanol with aqueous H2O2 as oxidanta

Entry Cat. Conv. (%) SM (%) SA (%) EH2O2 (%)

1 Hf(OTf)4 36 57 39 25 b 2 Hf(OTf)4 25 61 38 22 c 3 Hf(OTf)4 15 61 39 11 d 4 Hf(OTf)4 <1 n.d. n.d. n.d.

5 Hg(OTf)2 40 47 48 19

6 In(OTf)3 16 26 70 11

7 Sc(OTf)3 38 50 33 30

8 Y(OTf)3 17 39 22 9

9 Yb(OTf)3 16 44 21 9

10 Er(OTf)3 16 43 31 8

11 Zn(OTf)2 7 n.d. 41 17 a -1 Reaction conditions: ML 61 mmol L in methanol, ML/H2O2 molar ratio = 1:2, ML/cat. molar ratio = b c 10:1, 80°C, 6 h. SM, SA, and EH2O2, same as in Table A3.3. DTBP/Hf(OTf)4 molar ratio = 0.5. d DTBP/Hf(OTf)4 molar ratio = 1.4. DTBP/Hf(OTf)4 molar ratio = 4.0.

100

Appendix Chapter 4

Catalytic reactions a. Batch reactions with Hf-BEA

Figure A4.1. FID signal for a typical batch reaction for EP self-aldol condensation in toluene with Hf- BEA. Compounds in black are diesters and compounds in blue are counted as side products. Compounds 1 and 7 were identified with commercial standards and the NIST MS database. Compounds 3 and 4 were identified with the NIST MS database and NMR characterization (Figure A4.2). Compound 5 was identified with the NIST MS database (Figure A4.5). Compounds 9-13 were assigned based on the possible reaction pathways and MS signals in reference to the NIST MS database as shown in Figures A4.3, A4.4, and A4.6 below.

101

. 3 C NMR in CDCl 13 , c. 3 H NMR in CDCl 1 Figure A4.1. a. mass mass spectrum, b. Figure A4.1. a. Characterization for diesters 3 and 4 in Characterization for diesters 3 and

Appendix Appendix Figure A4.2. 102 103 103 kely

Appendix Appendix b) with a retention time of 14.4 min (li on time of 13.9 min (likely compound 9) and Mass spectra of compounds a) with a retenti Figure A4.3. compound 10) in Figure A4.1.

ound 11 and 12 in Scheme A4.1. n in Figure A4.1, likely comp 1 min in Figure A4.1, likely compound 5 in Scheme A4.1. Mass spectrum of the compound with a retention time of 15.1 mi Mass spectrum of the compound with a retention time of 17.

Appendix Appendix Figure A4.4. Figure A4.5. 104 105 105 wt%

Appendix Appendix

O O 13 O O O O O O O 11 12 ene. Reaction conditions: EP 3 O O O O O O compound 13 in Figure A4.1. O 9 10 O Time (h) 0.5 1.0 O 8 C, 1 h. Diesters include compounds 3 - 5 in Figure o O 7 O OH 13 11+12 9+10 8 7 6 OH 6

5 0

15 10 Yield (%) Yield self-aldol condensation in tolu b 4 min in Figure A4.1, likely

Diester Side products Conversion Time (h) ylbenzene in toluene, Hf-BEA, EP/Hf = 195 (mol/mol), 120 ylbenzene in toluene, Hf-BEA, EP/Hf = 195 (mol/mol), 0.00.20.40.60.81.0 0

80 60 40 20

100 Conversion/Selectivity (%) Conversion/Selectivity a A typical time profile a) side product distribution b) for EP time profile a) and A typical Mass spectrum of the compound with a retention time of 17. Figure A4.6. Figure A4.7. and 0.1 wt% 1,3,5-tri-tert-but A4.1. Appendix

Conversion 100 Diesters

20 Conversion/Selectivity (%)

0 123455C Run

Figure A4.8. Recyclability test for Hf-BEA. Reaction conditions: EP 3 wt% and 0.1 wt% 1,3,5-tri-tert- butylbenzene in toluene, Hf-BEA, EP/Hf = 195 (mol/mol), 120oC, 1 h; after each run, the catalyst was washed with toluene once and dried under air, and then a fresh EP solution was applied; after the fourth run, the catalyst was calcined at 580 oC for 3 h under air and used in run 5C. Diesters include compounds 3 - 5 in Figure A4.1.

106

Appendix b. Flow reactions

13 15 11+12 9+10 8 7 6 10 Yield (%)

0 0 20 40 60 80 100 120 140 TOS (h)

Figure A4.9. Yield to side products as a function of TOS for the flow reaction with Hf-BEA. Reaction conditions: EP 3 wt% and 1,3,5-tri-tert-butylbenzene 0.1 wt% in toluene, 120 oC, 12 bar. The operating conditions of the flow reactor were: 320 mg Hf-BEA, EP solution flow rate 0.20 mL min-1, weight hourly space velocity 32.2 h-1. The grey lines represent regeneration of Hf-BEA: flushing the catalyst bed with o o o toluene at 120 C, drying under N2 at 150 C, and calcination with air at 550 C for 5 h.

107

Appendix 100

30 Conversion/Yield (%) Conversion/Yield 20

0 0 20 40 60 80 100 120 140 TOS (h) Figure A4.10. EP conversion ( ), yield to diesters ( ), side products ( ), and unaccounted carbon ( ) as a function of TOS for the flow reaction with Hf-BEA. Reaction conditions: EP 3 wt% and 1,3,5-tri-tert- butylbenzene 0.1 wt% in toluene, 120 oC, 12 bar. The operating conditions of the flow reactor were: 320 mg Hf-BEA, EP solution flow rate 0.20 mL min-1, weight hourly space velocity 32.2 h-1. The dashed line o represents regeneration of Hf-BEA: flushing the catalyst bed with toluene at 120 C, drying under N2 at 150 oC, and calcination with air at 550 oC for 5 h. Diesters include compounds 3 - 5 in Figure A4.1. Side products include compounds 6 - 13 in Figure A4.1 and their distribution is shown in Figure A4.9. Unaccounted carbon includes compounds that could not be detected with GC analysis, including pyruvic acid and degradation products.

108

Appendix

60 Total diesters Main diesters

30 Yield (%) 20

0 0 20406080100120140 TOS (h)

Figure A4.11. Yield to total diesters ( ), and the main diester ( ) as a function of TOS for the flow reaction with Hf-BEA. Reaction conditions: EP 3 wt% in toluene, 1,3,5-tri-tert-butylbenzene 0.1 wt% in toluene, 120 oC, 12 bar. The operating conditions of the flow reactor were: 320 mg Hf-BEA, EP solution flow rate 0.20 mL min-1, weight hourly space velocity 32.2 h-1. The dashed line represents regeneration of o o Hf-BEA: flushing the catalyst bed with toluene at 120 C, drying under N2 at 150 C, and calcination with air at 550 oC for 5 h. Total diesters include compounds 3 - 5 in Figure A4.1, and compounds 3 and 4 in Figure A4.1 are quantified as main diesters.

109

Appendix c. Batch reactions with Na-exchanged Hf-BEA

Discussion on the impact of pore volume decrease on the catalytic activity

Plausible causes for the reversible activity drop with Na exchange and H2SO4 wash include strong interaction between Na+ and the active sites or pore blockage due to Na salt solids. However, the

FTIR spectra of CD3CN adsorbed on the catalysts after Na exchange (Figure 2b) shows a -1 decrease in the band at 2310 cm (representative of CD3CN that is strongly bound to tetrahedral -1 Hf(IV) centers) and the appearance of a new band at 2284 cm (previously assigned as CD3CN adsorbed on Na-exchanged open sites for Sn-BEA), which demonstrates that the Lewis acidic

Hf(IV) sites are accessible to CD3CN for Hf-BEA-Na-1. So the 16% pore volume decrease is likely not the sole cause of the 80% activity drop observed. Indeed, for Hf-BEA-Na-2 with a minimal pore volume change and a Na/Hf ratio of 0.02, a 24% conversion drop was still observed (Table 2, entry 3 and Table A4.2, entry 3).

110

Appendix Catalyst characterization data

Hf-BEA Sn-BEA Zr-BEA Intensity (a.u.)

10 20 30 40 2θ (degree) Figure A4.12. PXRD for pristine Lewis acidic BEA zeolites. Catalysts were calcined for 10 h at 580 oC before analysis.

Pristine Spent Intensity (a.u.)

10 20 30 40 2θ (degree) Figure A4.13. PXRD for Hf-BEA. The spent Hf-BEA was on stream for 132 h, flushed with toluene, o dried under N2 and then calcined under air for 10 h at 580 C.

111

Appendix

Hf-BEA-AW-2 Hf-BEA-Na-2 Hf-BEA-AW-1 Hf-BEA-Na-1 Hf-BEA Intensity (a.u.)

10 20 30 40 θ 2 (degree)

Figure A4.14. PXRD for pristine Hf-BEA, Hf-BEA-Na-1 (after Na exchange with NaNO3), Hf-BEA- Na-2 (after Na exchange with NaOH pH = 10 solution), Hf-BEA-AW-1 and Hf-BEA-AW-2 (Hf-BEA- Na-1 and Hf-BEA-Na-2 after acid wash, respectively).

500 /g)

3 Zr-BEA 400

Sn-BEA 300

Hf-BEA 200

100 Adsorbed Amount at STP (cm STP at Amount Adsorbed

0 0.0 0.2 0.4 0.6 0.8 1.0 Relative Pressure (P/P ) 0

Figure A4.15. N2 adsorption (open symbols) and desorption (closed symbols) isotherms of pristine Lewis acidic BEA zeolites. The isotherms are offset by 100 cm3 g-1. Catalysts were calcined for 10 h at 580 oC before analysis.

112

Appendix

300 /g) 3

Spent

200

Pristine

100 Adsorbed amount Adsorbed amount at STP (cm

0 0.0 0.2 0.4 0.6 0.8 1.0 Relative pressure (P/P ) 0

Figure A4.16. N2 adsorption (open symbols) and desorption (closed symbols) isotherms of Hf-BEA. The isotherms for spent catalyst are offset by 100 cm3 g-1. The spent catalyst was on stream for 132 hours, o flushed with toluene, dried under N2 and then calcined under air for 10 h at 580 C.

600

Hf-BEA-AW-2 /g) 3

Hf-BEA-Na-2 400

Hf-BEA-AW-1

200 Hf-BEA-Na-1

Hf-BEA Adsorbed Amount at STP (cm at Amount Adsorbed

0 0.0 0.2 0.4 0.6 0.8 1.0 Relative Pressure (P/P ) 0

Figure A4.17. N2 adsorption (open symbols) and desorption (closed symbols) isotherms of Hf-BEA, Hf-

BEA-Na-1 (after Na exchange with NaNO3), Hf-BEA-Na-2 (after Na exchange with NaOH pH = 10 solution), Hf-BEA-AW-1 and Hf-BEA-AW-2 (Hf-BEA-Na-1 and Hf-BEA-Na-2 after acid wash, respectively). The isotherms are offset by 100 cm3 g-1.

113

Appendix Hf-BEA catalyzed aldol reactions with other keto acids and esters

Scheme A4.1. Hf-BEA catalysed reactions with ethyl glyoxylate in toluene, (a) Cannizzaro reaction and (b) aldol addition reaction. Esterification reactions of acids generated are not listed above.

Reference

(1) Flytzani-Stephanopoulos, M. Acc. Chem. Res. 2013, 47, 783-792. (2) Guan, Y.; Ligthart, D. M.; Pirgon-Galin, Ö.; Pieterse, J. A.; van Santen, R. A.; Hensen, E. J. Top. Catal. 2011, 54, 424-438. (3) Tauster, S.; Fung, S.; Baker, R.; Horsley, J. Science 1981, 211, 1121-1125. (4) Tauster, S. Acc. Chem. Res. 1987, 20, 389-394. (5) Mei, B.; Wiktor, C.; Turner, S.; Pougin, A.; van Tendeloo, G.; Fischer, R. A.; Muhler, M.; Strunk, J. ACS Catal. 2013, 3, 3041-3049. (6) Liu, X.; Liu, M.-H.; Luo, Y.-C.; Mou, C.-Y.; Lin, S. D.; Cheng, H.; Chen, J.-M.; Lee, J.-F.; Lin, T.-S. J. Am. Chem. Soc. 2012, 134, 10251-10258. (7) Kuwauchi, Y.; Yoshida, H.; Akita, T.; Haruta, M.; Takeda, S. Angew. Chem. Int. Ed. 2012, 51, 7729-7733. (8) Akita, T.; Tanaka, S.; Tanaka, K.; Haruta, M.; Kohyama, M. J. Mater. Sci. 2011, 46, 4384-4391.

114