Quick viewing(Text Mode)

Co-Crystallization Induced Spontaneous Deracemization

Co-Crystallization Induced Spontaneous Deracemization

I

UNIVERSITÉ CATHOLIQUE DE LOUVAIN

INSTITUTE OF CONDENSED MATTER AND NANOSCIENCES

(IMCN)

Laboratory of Crystal Engineering

Prof. Tom Leyssens

Laboratory of Organic

Prof. Olivier Riant

Development and Optimization of a new general thermodynamic deracemization method: Co- Induced Spontaneous Deracemization

Thèse présentée en vue de l'obtention du grade

de docteur en Sciences par

Michael Guillot

Louvain-la-Neuve

Septembre 2020 II

Jury Members Prof. Y. Cartigny

Université de Rouen Normandie

Prof. S. Kuhn

Katholieke Universiteit Leuven

Prof. M. Singleton

Université catholique de Louvain

Dr. K. Robeyns

Université catholique de Louvain

Prof. Y. Garcia (President)

Université catholique de Louvain

Prof. T. Leyssens (Promotor)

Université catholique de Louvain

Prof. O. Riant (Co-promotor)

Université catholique de Louvain

III

Acknowledgements Je tiens tout d’abord à remercier mon promoteur, le professeur Tom Leyssens de m’avoir offert la chance de faire partie de son laboratoire pendant ces 4 années qui furent épanouissantes sur tous les plans. Je le remercie pour son encadrement tout au long de ma thèse, pour son soutien tant scientifique que personnel (surtout pendant « the dark year » où plus rien ne semblait marcher, je pense que tu vois de quel moment je parle Tom ;)) et pour la proximité qu’il sait créer au sein de son équipe. Je le remercie pour les congrès et séjours scientifiques que j’ai pu réaliser grâce à lui et qui ont tant participé à mon apprentissage autant scientifique que personnel. Je tiens tout aussi à remercier mon co-promoteur, le professeur Olivier Riant, pour ses conseils et corrections tout au long de ces quatres années de thèse. Je remercie aussi tous les membres du jury pour le temps qu’ils ont passé à évaluer cette thèse. I would also like to thank Professor Joop ter Hoost for welcoming me into his team and allowing me to carry out part of this research in his laboratory.

Now, I will go on by thanking Vanessa aka Vanish aka my BFF for those almost 4 years by my side (almost because you abandon me right before the end you bitch <3). You were always there when I needed you. You helped me more than you can imagine and the end of my PhD without you was not as great as it could have been with you. Getting to know you was the best part of my PhD but not entirely… Oui Xa aka Chou aka Dr. Xavier Buol, je te mets à égalité avec Vanessa sur ce point, te rencontrer est la meilleure partie de ma thèse. Tu as été là pour moi pendant ces 4 années de doctorat qu’on a en gros réalisées ensemble. Tu as su m’épauler pendant les moments difficiles (j’espère que ton t-shirt a réussi à sécher à présent ;)) et supporter mon chant. Tu as su me calmer dans mes divagations et m’écouter quand j’en avais besoin. T’es mon bro avec ses gros bibis, ti amo <3.

A présent, parlons de toi Carole aka Carolus. Tu as été ma première mémorante mais surtout une très belle rencontre au sein du laboratoire. Et je suis content que tu aies décider de revenir faire une thèse parmi nous et de pouvoir dire que tu es mon amie et ce même si tu as filmé cette fameuse vidéo à la gaypride que tu as accidentellement envoyé à Tom (oups !). Pour continuer chez les mémorantes, même si Sarah tu n’étais pas ma mémorante officielle, je te remercie quand même pour tout ce que tu as su apporter, ainsi que comme Carole pour ton aide lors de l’organisation de la STEL. Le marteau de Thor est encore à la cafétéria et j’espère qu’il y restera en souvenir de notre passage. Enfin, merci d’avoir permis à Joséphine de devenir ma deuxième mémorante officielle. Jo, je te remercie pour toute l’aide que tu m’as apporté sur la fin de ma thèse. J’ai pu compter sur toi comme personne et je suis vraiment désolé que tu aies eu à payer mes problèmes relationnels avec certaines personnes durant ton mémoire. Je sais très bien que tu avais fermé cette bouteille d’azote, je n’en ai jamais IV douté . Je suis aussi content que tu commences ta thèse au sein du groupe de Tom et te souhaite tout le succès que tu mérites.

Je vais maintenant remercier Aurore aka Aurorus, la maman de l’étage, je ne pense pas t’avoir déjà appelé comme cela mais c’est ce que tu représentes pour tout l’étage je pense et ce que tu as représenté pour moi durant ces 4 années de doctorat. Au moindre problème administratif tu as su m’aider (et que Tom en soit témoin l’administratif et moi ça fait 2), dans mon rôle de président STEL comme de membre ACIM tu as aussi su m’être indispensable. Je te remercie aussi pour tout ce que tu es et la bonne humeur que tu sais transmettre au sein de l’étage. Si comme moi, certains ont voulu te mettre des bâtons dans les roues et même si tu y as perdu des plumes, tu es restée et reste l’oiseau de valeur de notre étage. Tu es une secrétaire et personne en or. On se revoit quand tu pars à la retraite <3.

Pour continuer avec l’étage B1, je remercie Véro (et son côté râleur qui m’aura bien fait rire, ne change pas, râler c’est bon pour la santé !), Claire, Camila, Paolo, Martin, Kévin, Max, Nicolas, Tim, Lixing, Jean-Baptiste, Ricky, Gabriel, Fabrice, Bram, Benjamin, mon stagiaire IPL Elysée et toutes autres personnes qui ont permis à cet étage d’être aussi agréable et bonne ambiance contrairement à d’autres étages. Je vais spécialement remercier Ricky et Jean-Baptiste, aka JB, mes collègues de bureau pendant 3 ans, pour ces bons moments partagés. Et surtout merci Ricky de m’avoir défendu comme tu l’as fait alors que tu n’y étais pas du tout obligé. Ça a beaucoup compté pour moi. JB, je sais qu’on a eu nos hauts et nos bas, on n’a pas toujours été d’accord sur tout, mais tu reste une très belle rencontre. Je souhaite aussi remercier toutes les personnes qui m’ont aidées et avec qui j’ai passé de très bons moments lors de mon séjour dans le laboratoire du Professeur Ter Hoost et tout spécialement Maxime, Charline, Erin, Georgia, Russel, Jenna & Andrew.

Je souhaite remercier tous ceux qui m’ont aidé à organiser ces deux STEL en tant que président (Sébastien, Michael (A)) et tous les membres ACIM que j’ai eu plaisir à rencontrer (remerciement tout spécial à Quentin). Je souhaite remercier tout spécialement Mathilde, Sébastien, Mathieu, Céline, Audric (M), Marine et Corentin (L) du groupe Riant pour leur aide et leur gentillesse. Audric (M), je suis très content d’avoir pu te rencontrer pendant ton mémoire, d’avoir appris à te connaître et d’être devenu ton ami, monsieur le MJ. Je remercie aussi Nathalie, Saroj, Kiran, Junjie et Yonghua du groupe Riant.

Je tiens aussi à remercier toutes les personnes qui ont contribué non directement et pourtant tout aussi essentiellement : Mes besties Claire et Anaïs, malgré la distance rien ne change entre nous et j’espère que cela continuera ainsi jusqu’à ce que l’on soit vieux. Cora qui même si elle avait quitté le club des célib fut mon modèle tinder ;) Stéphane, mon plus vieil ami, on ne se voit pas souvent mais le peu de fois, c’est comme si le temps n’avait pas passé. Michaele, Martin, Martin Hélène, Florence et Davide, les colocs des combattants pour tous ces bons moments et ses fous rires, big V

up à Mike pour ces moments de proses inoubliables et ses discours qui ne sont pas tombé dans l’oreille d’un sourd. Enfin, je remercie Ousam, Giulia, Florence et Max, les colocs de la ramée pour cette très bonne dernière année passée à leur côté. Je remercie spécialement Ousam et Giulia qui sont des personnes formidables que je suis content d’avoir pu rencontrer et devenir ami avec. Oh Djadja c’est pour toi Giu. Ousam, mon bon couillon, je te remercie de m’avoir proposé cette colocation et n’oublie pas, derrière ton côté beau parleur, tu es un mec en or.

Enfin, je remercie mes parents et ma sœur qui malgré la distance ont su être là pour moi. Vous devriez revenir en Belgique, on visitera Pairi daiza ! (Une fois cette histoire de Corona passée).

VI

List of Publications

1. T. Leyssens, R. A. Couch, M. Guillot, B. Harmsen T. R. Bailey, M. Appelmans. Solid Forms of Fasoracetam. (2019). WO2019143824A1.

2. M. Guillot, J. de Meester, S. Huynen, L. Collard, K. Robeyns, O. Riant, T. Leyssens, Angew. Chem. Int. Ed. 2020, 132, 28, 11399- 11402.

3. M. Guillot, J. de Meester, J. H. ter Host, O. Riant, T. Leyssens, Fungicide precursor racemization kinetics for deracemization in complex systems. Eur. J. Org. Chem. 2020.

VII

Abstract In pharmaceutical industry, 50% of the marketed drug compounds contain a chiral center, essential to their functioning. Where one has the desired pharmacological effect, the other might be inactive, equally active or have adverse effects. For this reason, development of enantiopure drugs is strongly recommended by regulation authorities to industries. The industrially most prominent way to enantiopure drugs still involves formation of a racemic compound, followed by a chemical or physical resolution. Such a resolution implies a maximum yield of 50%, as the unwanted enantiomer is discarded upon separation.

However, if the compound is racemizable, in principle the unwanted enantiomer can be transformed into the desired one, leading to a so-called deracemization process, which finally can lead to a 100% maximum yield. Such processes exist for compounds that either are intermediates (Dynamic ), can form salts (Crystallization Induced Transformation) or conglomerates (Viedma Ripening). However, there are still a considerable amount of racemizable that don’t meet either criteria and therefore cannot be deracemized.

In this work, the aim is to develop a novel tool within the library of deracemization techniques in order to touch a larger range of compounds, with the ultimate goal to develop a physical thermodynamic deracemization technique applicable to all racemizable compounds. To do so, crystal engineering and crystal growth tools are combined to develop a Co-crystallization Induced Spontaneous Deracemization method (CoISD). This process is based on co-crystallization in order to induce an imbalance in solution by precipitation of only one enantiomer while racemizing the excess of the other one in solution.

Doing so, we first identified a suitable system on which to develop the CoISD process by synthetizing a series of analog compounds and submitting them to a co-crystal screening. A suitable co-crystal system composed of the synthetized analog (R,S)-4,4- dimethyl-1-(4-fluorophenyl)-2-(1H-1,2,4-triazol-1-yl)-Pentan-3-one ((R,S)-BnFTP) and the identified co-former enantiopure 3-Phenylbutyric acid. This system forms diastereomeric co-crystals that can be separated upon crystallization.

Then, this system was submitted to and racemization studies to eventually develop the CoISD process in toluene with 1,8-Diazabicyclo[5.4.0]undéc- 7-ène (DBU) as the racemizing agent. In parallel, the kinetics of racemization of BnFTP with DBU with and without the presence of the co-former were studied. The developed process ran as a two-pot one-step deracemization with a crystallization cell at low temperature and a racemization cell at high temperature. Following the success of the development, the process was optimized. Several operational parameters were varied in order to assess their impact on the process yield and overall deracemization. At the end, an efficient process with a yield of 73% and an oval deracemization of 80% was achieved. VIII

Finally, the deracemized BnFTP was valorized by the reduction of its function by keeping the of the original chiral center while inducing a favored configuration for the newly formed chiral center. Reduced BnFTP is a closely related analog of Paclobutrazol, a fungicide and growth retardant.

IX

List of Abbreviations

API – Active Pharmaceutical Ingredient AS3PC – Auto-Seeded Polythermic Programmed Preferential Crystallization BnBAA – α-bromo-Benzenepropanoic acid BnBnOP – 2-(benzyloxy)-4,4-dimethyl-1-phenylpentan-3-one

BnClTAmBn2 – α-[4-chlorophenylmethyl]-N,N-bis(phenylmethyl)-1H-1,2,4- Triazole-1-acetamide BnClTP – 4,4-dimethyl-1-(4-chlorophenyl)-2-(1H-1,2,4-triazol-1-yl)-Pentan-3-one

BnFTAmBn2 – α-[4-fluorophenylmethyl]-N,N-bis(phenylmethyl)-1H-1,2,4- Triazole-1-acetamide BnFTP – 4,4-dimethyl-1-(4-fluorophenyl)-2-(1H-1,2,4-triazol-1-yl)-Pentan-3-one BnOP – 1-(benzyloxy)-3,3-dimethylbutan-2-one BnTAA – α-(1H-1,2,4-triazol-1-yl)-Benzenepropanoic acid

BnTAmBn2 – α-[phenylmethyl]-N,N-bis(phenylmethyl)-1H-1,2,4-Triazole-1- acetamide BnTAP – 1,3-diphenyl-2-(1H-1,2,4-triazol-1-yl)propan-1-one BnTP – 4,4-dimethyl-1-phenyl-2-(1H-1,2,4-triazol-1-yl)-Pentan-3-one CIAT – Crystallization-Induced Asymmetric Transformation CIDT – Crystallization-Induced Diastereomer Transformation CoISD – Co-crystallization Induced Spontaneous Deracemization CP – Chloropinacolone CPC1 – Continuous Preferential Crystallization CSD – Cambridge Structural Database CSP – Chiral Stationary Phase DBU – 1,8-Diazabicyclo(5.4.0)undec-7-ene DBUH+ – Protonated 1,8-diazabicyclo(5.4.0)undec-7-ene DFT – Density Functionnal Theory DKR – Dynamic Kinetic Resolution DMF – Dimethylformamide DMAP – 4-Diméthylaminopyridine DPC – Dynamic Preferential Crystallization DSC – Differential Scanning Calorimetry E – FDA – Food and Drug Administration GC – Gaz X

GRAS – Generally Regarded As Safe HPLC – High Performance Liquid Chromatography IR – Infra-Red KHMDS – Potassium HexaMethylDiSilazide KR – Kinetic Resolution LDA – Lithium Diisopropylamine MBnTA – Methyl-(1H-1,2,4-Triazole-phenylmethyl)-acetate MTA – Methyl-(1H-1,2,4-Triazole)-acetate n-BuLi – n-Butyl lithium NMPA – N-(2-methyl-benzylidene)-phenylglycine amide NMR – Nuclear Magnetic Resonance PBA – 3-Phenylbutyric acid PC – Preferential Crystallization PEA – Phenylmethylamine PgTP – 2,2-dimethyl-4-(1H-1,2,4-triazol-1-yl)hept-6-yn-3-one PTSA – p-Toluenesulfonic acid PXRD – Powder X-Ray Diffraction SC-XRD – Single Crystal X-Ray Diffraction SIPC – Seeded Isothermal Preferential Crystallization

SN2 – Substitution of order 2 S3PC – Seeded Polythermic Preferential Crystallization TAm – 1H-1,2,4-Triazole-1-acetamide

TAmBn2 – N,N-bis(phenylmethyl)-1H-1,2,4-Triazole-1-acetamide TAP – 1-phenyl-2-(1H-1,2,4-triazol-1-yl)ethan-1-one THF – Tetrahydrofuran TP – 3,3-dimethyl-1-(1H-1,2,4-triazol-1-yl)- Butan-2-one VR – Viedma Ripening XRD – X-Ray Diffraction

XI

Contents

General Introduction ...... 1 1. : a key-feature of life ...... 1 1.1. What is chirality? ...... 1 1.1.1. The configuration of a chiral center ...... 1 1.1.2. The different types of chirality ...... 2 1.1.3. Properties of ...... 4 1.1.4. ...... 5 1.1.5. Chirality in the macroscopic world: ...... 5 1.2. The impact of chirality on the human body ...... 6 1.2.1. Mechanism of biological activity ...... 6 1.2.2. Pharmacology ...... 8 1.3. The aftermath of the scandal ...... 10 1.3.1. Chirality and industry: the ...... 10 1.3.2. Chirality in a current context ...... 10 1.3.3. Development and expansion of chiral tools ...... 11 2. Different pathways to enantiopurity ...... 12 2.1. Upstream: ...... 12 2.1.1. For synthesis of enantiopure compounds ...... 14 2.1.2. of existing biological compounds ...... 15 2.1.3. As a tool in asymmetric synthesis and resolution ...... 17 2.2. In-stream: Asymmetric synthesis ...... 17 2.2.1. ...... 19 2.2.2. Chiral auxiliaries ...... 21 2.2.3. Chiral reagents ...... 22 2.2.4. Chiral catalysts ...... 23 2.3. Downstream: Resolution ...... 26 2.3.1. Chemical Separation: Kinetic Resolution ...... 26 2.3.2. Physical separation: Chiral chromatography ...... 28 2.3.3. Physical separation - Kinetically-based: Preferential Crystallization ...... 36 2.3.4. Physical separation - Thermodynamically-based: Diastereomeric resolution 42 2.4. Recycling the distomer: Deracemization ...... 46 2.4.1. Racemization ...... 47 2.4.2. Chemical deracemization: Dynamic Kinetic Resolution ...... 49 2.4.3. Physical deracemization – Kinetically based: ...... 50 2.4.4. Physical deracemization – Thermodynamically-based: crystallization-induced diastereomer transformation ...... 54 3. Co-crystallization: A solid-state discovery full of promises ...... 56 3.1. Solid-state state of the art ...... 56 3.2. What is co-crystallization? ...... 57 3.3. The importance of co-crystallization for the pharmaceutical industry ...... 60 XII

3.4. Chirality and co-crystallization: expansion of the chiral toolbox ...... 62 3.4.1. Co-crystallization, an alternative to diastereomeric salt resolution ...... 62 3.4.2. Co-crystallization, an alternative for deracemization? ...... 65 4. Bibliography ...... 65 Goal of this work ...... 76 Chapter 1- Identifying a suitable system ...... 78 1. Overview ...... 79 2. Target compounds: from design to synthesis...... 79 2.1. Introduction ...... 79 2.2. Material & methods ...... 81 2.2.1. Model compound: BnClTP ...... 82 2.2.2. BnFTP analog ...... 83 2.2.3. BnTP analog ...... 85 2.2.4. PgTP analog ...... 87 2.2.5. BnBnOP analog ...... 87 2.2.6. BnTAP analog ...... 89 2.2.7. BnClTAmBn2 analog ...... 90 2.2.8. BnFTAmBn2 analog ...... 92 2.2.9. BnTAmBn2 analog ...... 93 2.2.10. MBnTA analog ...... 93 2.3. Results & discussion ...... 95 2.3.1. Dialkylation: A competition between O- and C-alkylation ...... 95 2.3.2. Crystallization developments ...... 100 2.4. Conclusion ...... 101 3. Co-crystal screening ...... 101 3.1. Introduction ...... 101 3.2. Material & methods ...... 104 3.2.1. Co-crystal screening ...... 104 3.2.2. Solid-state analysis ...... 104 3.2.3. Single crystal Growth ...... 104 3.2.4. Co-crystal study ...... 104 3.3. Results and discussion...... 105 3.3.1. Solid state analysis of the target library ...... 105 3.3.2. Co-crystal screening ...... 106 3.3.3. Selecting a suitable Co-crystal for the development of the CoISD process .. 112 3.4. Conclusion ...... 114 4. Bibliography ...... 114 Chapter 2- The road to the development of the CoISD process ...... 117 1. Overview ...... 118 2. Chiral resolution: Finding a suitable solvent and method to break down the co- crystal ...... 118 2.1. Introduction ...... 118 2.2. Materials & methods ...... 119 2.2.1. Preparation of (S)-3-phenylbutyric acid ...... 119 2.2.2. Preparation of cocrystal (S)-BnFTP-(S)-PBA ...... 120 XIII

2.2.3. Preparation of (S)-BnFTP ...... 121 2.2.4. Choosing the solvent ...... 121 2.2.5. Ternary phase diagram ...... 121 2.2.6. Breaking the co-crystal ...... 122 2.3. Results & discussion ...... 122 2.3.1. Choosing the solvent ...... 122 2.3.2. Ternary phase diagram ...... 125 2.3.3. Diastereomeric system ...... 126 2.3.4. Resolution in toluene ...... 127 2.3.5. Breaking the co-crystal ...... 127 2.4 Conclusion ...... 129 3. Racemization: Finding a suitable racemizing agent and kinetic study of the racemization reaction ...... 130 3.1. Introduction ...... 130 3.2. Material and methods ...... 131 3.2.1. Material ...... 131 3.2.2. Racemizing agent screening ...... 131 3.2.3. Polarimetric analysis ...... 131 3.2.4. Protocol for study of the impact of temperature ...... 134 3.3. Results and Discussion ...... 134 3.3.1. Racemizing agent screening ...... 134 3.3.2. Kinetic study of DBU ...... 135 3.3.3. Kinetic study with DBU in presence of (R)-PBA ...... 143 3.4. Conclusion ...... 147 4. Bibliography ...... 147 Chapter 3- Development & optimization of the CoISD process ...... 149 1. Overview ...... 150 2. CoISD: A general innovative thermodynamic approach to deracemization: Proof of concept...... 150 2.1. Introduction ...... 150 2.2. Materials and methods ...... 152 2.2.1. The studied system ...... 152 2.2.2. The co-crystals ...... 152 2.2.3. Deracemization process ...... 152 2.3. Results and discussion...... 155 2.4. Conclusion ...... 160 3. Toward an optimal CoISD process: the impact of operational parameters ..... 160 3.1. Introduction ...... 160 3.2. Materials and method ...... 161 3.2.1. General procedure for all runs ...... 161 3.2.2. Kinetic measurements ...... 162 3.3. Results & discussion ...... 163 3.3.1. The system ...... 163 3.3.2. Water impact on the process...... 165 3.3.3. Alternative solvents ...... 167 XIV

3.3.4. DBU concentration ...... 167 3.3.5. Temperature of the crystallization cell ...... 169 3.3.6. Productivity of the process ...... 170 3.3.7. Summary ...... 170 3.4. Conclusion ...... 171 4. Bibliography ...... 171 Chapter 4 - Valorization of BnFTP ...... 174 1. Overview ...... 175 2. Introduction ...... 175 3. Materials and methods ...... 177 3.1. Reduction of BnFTP ...... 178 3.1.1. By copper with BDP ...... 178 3.1.2. By copper catalysis with BINAP ...... 178 3.1.3. With NaBH4 ...... 179 3.1.4. With LiAlH4 ...... 179 3.1.5. With Me4NBH4 ...... 179 3.2. Chiral HPLC ...... 179 4. Results and discussion ...... 179 4.1. Finding a suitable reduction ...... 179 4.2. Diastereoselectivity of the reduction ...... 182 5. Conclusion ...... 187 6. Bibliography ...... 187 Conclusion & Perspectives ...... 189 Conclusion ...... 190 Perspectives ...... 193 Bibliography ...... 195 Appendices ...... 197 Appendix A: SI for Chapter 1 ...... 198 Material & methods for part 2...... 198 TP: DSC ...... 198 BnClTP: DSC...... 198 BnFTP: DSC ...... 199 BnTP: DSC ...... 200 PgTP: DSC ...... 201 BnBnOP: DSC ...... 201 TAP: DSC ...... 202 BnTAP: DSC ...... 202 TAmBn2: DSC ...... 203 BnClTAmBn2: DSC ...... 203 BnFTAmBn2: DSC ...... 204 BnTAmBn2: DSC ...... 204 BnBAA: DSC + Structural information ...... 205 BnTAA: DSC + Structural information ...... 206 MBnTA: DSC ...... 207 XV

Results & discussion for part 2...... 207 By-products BnFTP ...... 207 Results and discussion for part 3...... 209 Structural Information of analogs ...... 209 Structural Information of co-crystals ...... 213 Angle and distance for hydrogen bonds ...... 218 Suitable Co-crystals for the development of the CoISD process ...... 218 Phase diagram ...... 219 Calibration curves ...... 219 Results ...... 219 Appendix B: Article on BnTAA synthesis ...... 221 Main Article...... 221 Supporting information...... 231 Appendix C: SI for chapter 2 ...... 233 Part 2...... 233 XRPD for congruence test ...... 233 Solubility curves ...... 233 Calculation for H°, S° and the solubility of the (S,S) co-crystal at 20°C ...... 235 HPLC for breaking the co-crystal ...... 236 Part 3...... 237 Results of the acid/base screening ...... 237 Further study of racemization of (S)-BnFTP with DBU ...... 238 Racemization study by polarimetry ...... 238 Measurement set-up ...... 238 Study of the influence of (S)-BnFTP concentration ...... 239 Linearization for the study of the influence of (S)-BnFTP concentration ...... 239 Value of k’ for the influence of DBU concentrations ...... 241 Linearization for study of the influence of DBU concentration ...... 241 Linearization for study of the influence of the solvent ...... 242 Linear model error ...... 243 Linearization for study of influence of (R)-3-phenylbutyric acid on racemization ...... 244 Racemization study by HPLC ...... 245 Linearization for study of influence of temperature on racemization in presence of (R)-3- phenylbutyric acid ...... 245 Appendix D: SI for chapter 3...... 246 Part 2...... 246 Structural Information of (R)-BnFTP + (S)-PBA co-crystal ...... 246 Deracemization n°1 ...... 248 Deracemization n°2 ...... 251 Deracemization n°3 ...... 258 Racemization of (R)-BnFTP with the 1eq of (S)-PBA when temperature is applied ...... 259 Part 3...... 259 HPLC results for all runs...... 259 Run 1 ...... 259 XVI

Run 2 ...... 260 Run 3 ...... 260 Run 4 ...... 261 Run 5 ...... 261 Run 6 ...... 262 Run 7 ...... 262 Run 8 ...... 263 Run 9 ...... 263 Run 10 ...... 264 Kinetics measurements ...... 264 Study of racemization kinetics of BnFTP and PBA co-crystal in Toluene-water mixtures ...... 264 Follow-up of run 8 ...... 267 HPLC for K calculation ...... 269

1 General Introduction General Introduction

1. Chirality: a key-feature of life 1.1. What is chirality? Chirality is the geometrical property of an object or a , not to be superimposable on its mirror image1. Historically, Louis Pasteur first discovered chirality in 1848 when separating, by hand, the enantiomeric crystals of sodium ammonium tartate.2,3 The word “chirality” comes from the Greek word χειρ (kheir) that stands for "hand" because indeed, per definition, our hands are chiral since they are images of each other but cannot be overlaid on each other (figure 1).

Figure 1 Illustration of one of the most common chiral object, our hands (A) and the fact they cannot be superimposed (B).

The reasoning behind this empirical observation is that hands and, other chiral objects and molecules lack a mirror plane and/or a center of symmetry. This is why they can exists as two mirror image forms, in chemical terms labeled enantiomers.1,4 Consequently, any structure containing at least one mirror plane or center of symmetry is superimposable on its mirror image and is thus called achiral, e.g. a glass bottle.

1.1.1. The configuration of a chiral center In chemistry, chirality often implies occurrence of a stereogenic center, which most of the time is a sp3 carbon. Nevertheless, boron, sulfur (as ), nitrogen (as quaternary amines), phosphorus (as phosphate esters) and atoms can also lead to stereogenic centers. For a quaternary carbon to be chiral, it needs to have four different substituents. Then, two configurations can exist, with one being the mirror image of the other. They are called enantiomers, one termed the (S)-enantiomer and the other the (R)-enantiomer. The symbol “S” stands for sinister (left in Latin) and the symbol “R” stands for rectus (right in Latin). The configuration of a stereogenic center is determined using Cahn-Ingol-Prelog rules in order to assign a priority number to each substituent. This priority number depends on the atomic number of the atoms composing each substituent. One first looks at the first atom of each substituent; the higher atomic number has priority. If there is a tie, then one must look one bond further 2 General Introduction

and so on until a different priority can be attributed to each substituent. This performed, one considers the order of priority groups when looking the molecule with the lowest priority group pointing backward. If this leads to a clockwise rotation, the stereogenic center is R, if counter-clockwise, it is coined S (figure 2).4,5

Figure 2 Illustration of the determination of the configuration of a chiral carbon for two enantiomers. The order of priority between each group is given as 1, 2, 3 and 4; group 4 being the lowest priority group.

1.1.2. The different types of chirality Though central chirality is the most widespread type of chirality in chemistry, it is not the only one. Indeed, some chiral molecules do not possess a stereogenic center and are still not superimposable on their mirror image due to the absence of a symmetry plane, like e.g. BINOL. There are two other types of structural arrangement leading to chirality: 4,6

Axial chirality (which includes the sub-category helicoidal chirality)

Axial chirality revolves around the presence of an axis blocking four different groups in a given configuration like e.g. a double bond blocking rotation. Looking along this axis, chirality can be determined in a similar way as for stereogenic centers. Molecules like, biaryls, spiropyrans and allenes are axially chiral. Figures 3 and 4 explain how to determine the configuration for allenes and biaryls. 3 General Introduction

Figure 3 Illustration of the axial chirality of allenes: to determine their configuration, one has to look along the axis after having ordered the groups. Those linked together are pointed 1 and 2 for the front ones and 3 and 4 for the back ones. Configuration is obtained by the direction from group 2 to group 3: clockwise aR, counter-clockwise aS.

Figure 4 Illustration of the axial chirality of BINAP: to determine its configuration, one has to look along the axis of chirality after having ordered the groups linked together as 1 and 2 for the front one and 3 and 4 for the back ones. Chirality is obtained by the direction from group 2 to group 3: clockwise aR, counter-clockwise aS.

Helicoidal chirality is a special case of axial chirality originating from a molecule helicoidally revolving around an axis. In this case, the configuration is determined like for stairs, depending on the direction of the helix. If it is going clockwise, the configuration is called M (minus) and for the counter clockwise revolution, it is called P (plus). An example is given in figure 5 with the case of the Hexahelicene molecule. 4 General Introduction

Figure 5 Illustration of the helicoidal chirality of Hexahelicene.

As for planar chirality, a molecule needs to possess parts of its atoms in plane and the others out of this plane. This is the case for or E-Cyclooctene derivatives. For ferrocene derivatives, the two non-coplanar rings must differ in substitution on at least one atom and be blocked in a certain configuration by the connecting them. Figure 6 explains how to determine the configuration of a ferrocene derivative.

Figure 6 Illustration of the planar chirality of Ferrocene derivatives: to determine their configuration, one must consider the ring with the most substituents. Then, from a Newman representation perspective, looking from above the ring, the substituents are ordered and their direction gives the configuration: clockwise pR, counter-clockwise pS.

1.1.3. Properties of enantiomers As they are mirror images, enantiomers possess the same physical and chemical properties, like solubility, melting point, crystalline structure… as long as they are in an achiral environment. However, when placed in a chiral environment e.g. in a chiral column, two enantiomers behave differently, in this case showing different retention 5 General Introduction times. When put under polarized , one enantiomer deviates the plane of rotation of light in one direction, the other in the opposite direction. One therefore says that chiral molecules are optically active, with the (+)-enantiomer deviating the light clockwise and the (−)-enantiomer counter-clockwise. Polarimetry is the oldest technique still in use to characterize enantiomers despite the fact that the direction of deviation does not correlate to the absolute configuration of the molecule, hence an (R)-configuration can be either (+) or (−).

1.1.4. Diastereomers When more than one chiral feature (generally chiral center) is present in a molecule, 2n stereoisomers can be formed providing no mirror plane or inversion center is present on the molecule. For instance, tartaric acid (figure 7), a molecule possessing 2 chiral centers, only possesses 3 stereoisomers because the (R,S) form is equivalent to the (S,R) form due to the presence of a mirror plane in the molecule for those two forms, yielding a .

Figure 7 Three forms of tartaric acid from left to right and top to bottom: (R,S)- and (S,R)- tartaric acid the meso form, (S,S)-tartaric acid and (R,R)-tartaric acid, which are enantiomers.

When multiple chiral centers are present, not all stereoisomers are enantiomerically related to each other. Only the ones for which each chiral feature has its configuration inversed are actual mirror images and thus enantiomers. Other pairs are referred to as diastereomers or epimers in the case of molecules with at least 3 chiral centers and only differing in configuration of one center. Epimers are a specific case of diastereomers. As diastereomers are not mirror images of each other, their physico- chemical properties differ, even in a non-chiral environment.

1.1.5. Chirality in the macroscopic world: Homochirality In our world, chirality is omnipresent at the macroscopic scale. Consider curved stairs for instance, which show axial chirality (figure 8). This is also true for labeled bottles 6 General Introduction

for example. In this case, it is not the bottle itself that is chiral but the label with the writing, which renders the ensemble chiral (absence of symmetry plane).

Figure 8 Illustration of the axial chirality of stairs.

Now, should one go into the microscopic world, chirality would also be omnipresent, with DNA and proteins showing chirality just as their constituting elements, the amino acids. Furthermore, their configuration display the phenomenon of homochirality, meaning only one enantiomer can be found biologically as a constituting unit, though the other can chemically exist. For instance, all constituting amino acid are L while all DNA constituting sugars are D. Though still investigated regarding how it came to be, homochirality is connected to the origin of life and several theories have been studied e.g. Circularly polarized UV light on extra-terrestrial meteorites, Circularly polarized UV light on earth, Weak force (parity violation) or Autocatalysis (model by crystal growth scientist Frank in 1953).7,8 As a result of this phenomenon, the human body is constituted of homochiral molecules including proteins, and receptor, which are at the top of its biological functions. Because of this, the human body is a chiral environment and thus chirality has a strong impact on human life.

1.2. The impact of chirality on the human body And what an impact! Administering a given enantiomer of a chiral drug can result in the desired pharmaceutical effect while taking the other can jeopardize your health. This is because of chiral recognition occurring within the homochiral body, leading to two enantiomers having different properties in this environement.9,10

1.2.1. Mechanism of biological activity Chiral recognition can be understood by looking at the receptors of enzymes, which are shape-dependent. These sites can be seen as making a three-point interaction with drugs following the key-lock model by Easson and Stedman.11–13 Two enantiomers do not have the same 3D-shape which leads to one enantiomer being able to interact with a specific receptor while the other cannot (figure 9). 7 General Introduction

Figure 9 Easson-Stedman hypothetical interaction of two enantiomers and a receptor at the binding site. For the left enantiomer, all substituents are able to interact with the binding site while for the right enantiomer two substituents cannot interact with the binding site. Consequently, the left enantiomer is active and the right one inactive regarding this receptor.

To make a long story short, chiral recognition is similar as putting on shoes, only the left foot can enter the left shoe and the right foot the right shoe. However, if a given enantiomer does not fit a target receptor, it does not mean it won’t be able to fit another receptor.11–14 Moreover, distribution and removal of the compound from the body will be impacted by chirality. Indeed, several molecules, including antiarrhythmic drugs show a different distribution, metabolism and renal excretion for both enantiomers.15 There are numerous examples of chiral molecules interacting differently with the human body. For instance, the natural molecule Carvone (figure 10) interacts differently with odor and taste receptors. (R)-carvone is present in spearmint oil and possesses a sweetish minty smell while (S)-carvone is present in caraway seeds (Persian cumin), and has a spicy taste.16

Figure 10 Schematic of (R)- and (S)-Carvone with their different properties.

Glucose is another example where natural D-glucose and synthetic L-glucose (figure 11) both taste the same. However, only D-glucose can be used as a source of energy by living organisms since the L-enantiomer cannot enter the first step of glycolysis due to enantiospecific recognition by the .17,18 8 General Introduction

Figure 11 Schematic of L- and D-glucose.

1.2.2. Pharmacology With this in mind, there are three possible cases, pharmacology speaking, for :19–21

1. Only one enantiomer (eutomer) is the main bioactive enantiomer with the second (distomer) either less active, inactive or toxic 2. Both enantiomers are equally active 3. One or both enantiomer(s) are subject to chiral inversion inside the body

In the first group, we can find all β-blockers for which the levorotary-isomer is more potent in blocking β-adrenoceptors. For instance, S-(−)- (figure 12) is 100 times more efficient than its mirror image.22,23

Figure 12 Schematic of (S)- and (R)-propranolol, which are used to treat high blood pressure.

Another example would be (figure 13) for which the (S)-enantiomer acts as a strong sedative or analgesic while the (R)-enantiomer is less active but responsible for the hallucination effect.24,25

Figure 13 Schematic of (R)- and (S)-Ketamine. 9 General Introduction

In the second group, the anti-depressant fluoxetine (figure 14) is an example of a chiral drug where both enantiomers have the same biological properties.26

Figure 14 Schematic of (R,S)-fluoxetine, which is used to treat depression.

Finally, the third category can be split in two examples, one where only one enantiomer is inversed in the body, leading to the transformation of the distomer into the eutomer. This is the case of (figure 15), where the (S)-enantiomer is 100 times more active than the (R)-one in inhibiting cyclooxygenase I. Moreover, in the body, only (R)-ibuprofen can be inversed by hepatic enzymes into (S)-Ibuprofen.14,27

Figure 15 Schematic of (S)- and (R)-Ibuprofen with the latter transforming into the first one in the body.

The second possibility is illustrated by the infamous Thalidomide (figure 16), which was commercialized in the 1950s and 1960s, as a sedative and an anti-nausea under its racemic form. Racemization of both enantiomers furthermore occurs in the body. Sadly, only the (R)-enantiomer possesses the desired effect, the (S)-enantiomer being teratogenic, which caused malformation of the fetus with pregnant women using the to counter morning sickness. Due to the racemization, even administering (R)-Thalidomide would have led to a similar outcome.28,29

Figure 16 Schematic of (S)- and (R)-Thalidomide being racemized in the human body. 10 General Introduction

1.3. The aftermath of the Thalidomide scandal 1.3.1. Chirality and industry: the Chiral switch The scandal related to Thalidomide, led to a series of changes. One of the first milestones occurred in 1992, when the Food and Drug Administration (FDA) published a new policy on stereo-isomeric drugs, inciting pharmaceutical companies to consider quality, safety and efficacy, eventually pushing them toward enantiopure drugs.30 From that moment, the chiral drug area blossomed with the organization of several conferences. In parallel, a chiral switch was mentally developed. Indeed, in 1994, Ibuprofen became the first-ever racemic non-steroidal anti-inflammatory drug to be switched to its enantiopure version, (S)-Ibuprofen. It was shortly followed in 1998, by whose (S)-enantiomer was also marketed. In total, during the period 1994-2011, 15 racemic drugs were switched from a racemic marketed form to an enantiopure one. The chiral switch was very popular for several ethical reasons. First, the enantiopure drug improves the therapeutic index by upping potency and selectivity while decreasing side effects. Second, an enantiopure drug possesses a faster onset of action. Thirdly, there is a reduced risk of drug-drug interactions. Finally, due to the increase in potency, the necessary amount to produce the same biological response as the racemic is lower and thus the dosage for a patient is lessened. A good chiral switch example is Ketamine. Its chiral switch to (S)-Ketamine allowed for lesser dosage and decrease of adverse effects like hallucination and agitation. Nevertheless, companies also took advantage of the chiral switch in order to re-patent the same medication with roughly no difference except for its enantiopurity and the new patent extending the protection against rival companies who could use the racemic molecule once the patent expired.30–32 This was the case of Esomeprazol (figure 17), the (S)-enantiomer of , a gastric anti-secretory proton pump inhibitor (PPI),33 even though the efficiency over the racemate was not proven.

Figure 17 Schematic of (R,S)-omeprazol’s chiral switch to Esomeprazol ((S)-omeprazol).

1.3.2. Chirality in a current context In the beginning of the 20th century, 56% of all pharmaceutical API were chiral but 88% of them were still sold as racemates. This was at the early stages of the chiral switch and development of enantiopure drugs. In 2015, of the 45 chiral drugs the FDA 11 General Introduction approved, only one was marketed as a racemate while all others had a defined stereochemistry.

As more and more racemic drugs switched to their enantiopure form and new drugs were immediately marketed as a single stereoisomer, there was a parallel growing need for expanding the methodologies leading up to stereoisomerically pure molecules.

1.3.3. Development and expansion of chiral tools To shorten the time between drug discovery and development of chiral drugs, it was also necessary to develop new chiral analytical tools while improving those already existing. With these, companies can efficiently determine the absolute configuration and enantiomeric purity of their future marketed drugs even at early stages. With this in mind, a “Chiral Technology” toolbox was built.34 This Chiral toolbox is defined as a collection of techniques for determining absolute stereochemistry, separating enantiomers and facilitating asymmetric transformations. This toolbox grew in size as the years passed. For the determination of the absolute configuration the following methods were developed:32

Single Crystal X-ray crystallography, is the most reliable method to determine the absolute configuration of a chiral molecule. The preferred method is simple, reacting the molecule of unknown configuration with another chiral enantiopure molecule of known configuration in order to form a diastereomer able to crystallize as a single crystal. Then, by analyzing the crystal, one can obtain the crystal structure of the diastereomer and deduce the stereochemistry of the compound of unknown configuration based on that of the added compound.35 However, due to the improvement in the quality of the measure, it is also possible to determine the configuration of an enantiomerically pure molecule.36,37 NMR techniques, though unable to distinguish enantiomers, can be used when either a chiral solvating agent is added to a non-chiral standard NMR solvent or when the pure enantiomer of unknown configuration is reacted with an enantiopure compound to produce two diastereomeric derivatives. In both cases, the enantiomers can be distinguished but only the second method allows for determination of the absolute chemistry. For instance, “Mosher’s method” was developed in 1973 for determining the absolute configuration of secondary using methoxytrifluoromethylphenylacetic acid (MTPA) as a reagent.38–40 Vibrational circular dichroism spectroscopy can be used for determining the absolute configuration of chiral molecules or their enantiomeric excess. The absolute stereochemistry determination is done through comparison of the experimental and circular dichroism spectra with the Density 12 General Introduction

Functionnal Theory (DFT) calculated ones of the molecule for a chosen configuration. It must be noted that several absolute configurations of chiral molecules, in the past twenty years, were determined by this method and that the FDA officially approved this method for determining absolute configurations.41,42 Polarimetry, as stated previously, measures the of a chiral molecule at a specific temperature and in a specific solvent, giving access to its specific rotation in this condition. This value is either positive or negative depending on the configuration. When the specific rotation of a chiral molecule is tabulated, polarimetry can efficiently be used to determine the enantiomeric ratio alongside the configuration of the enantiomer in excess.43 Furthermore, recently, DFT calculation of the optical rotation of a molecule allowed for determination of its absolute configuration.44,45 Enantioselective chromatography is likely the perfect technique for separating enantiomers and accessing their enantiomeric excess. Regarding determination of absolute configuration, it is possible but limited to compounds for which an analog already has its absolute configuration correlated with the elution order.46 Indeed, the method is based on comparing the elution order of the molecule of unknown configuration with a well- known analog.

Regarding separation of enantiomers and synthesis of enantiopure compounds, several methods were developed over the years and are still being developed nowadays. Both synthesis and separation are very useful tools to access enantiopurity, though the pathways may differ.

2. Different pathways to enantiopurity One generally distinguishes three different methodologies to yield enantiopure compounds. First, there is the chiral pool where one starts from an enantiopure natural material and derivatizes this latter. Then, there is asymmetric synthesis where one starts from pro-chiral molecules that react to yield preferably one single enantiomer. Finally, there are resolution/deracemization techniques where one starts from racemic molecules and isolates the targeted enantiomer.

2.1. Upstream: Chiral Pool The oldest method to access enantiopure material is without a doubt the chiral pool. With it, take advantage of the homochirality of our world to synthetize new chiral molecules for different applications. The chiral pool is a vast, though limited library of naturally occurring compounds at the ’s disposal. Molecules can range from small chiral molecules to large macromolecules that are more complex. 13 General Introduction

Chiral pool molecules can be classified in one of the six following categories (figure 18):

Alkaloids: Class of naturally occurring organic molecules only containing basic nitrogen atoms alongside their carbon-hydrogen backbone. Some can contain and sulfur atoms in addition. Amino acids: Class of naturally occurring organic molecules containing amine and carboxylic acid groups on a carbon-hydrogen backbone. Hydroxy acids: Class of naturally occurring organic molecules containing and carboxylic acid functions on a carbon-hydrogen backbone. Carbohydrates: Class of naturally occurring organic molecules only containing oxygen, hydrogen and carbon atoms, with a specific ratio of two hydrogen atoms for one oxygen atom in general. Terpenes: Large class of naturally occurring hydrocarbons produced by plants and insects. When derivatized with other atoms; mostly oxygen; they are called terpenoids. Miscellaneous: Category containing all other naturally occurring molecules that do not belong to any of the previous categories. Those molecules can be organic or inorganic.

Figure 18 Picture exemplifying each category of the chiral pool library: Hyoscyamine for the , L- for amino acids, R-mandelic acid for hydroxyl acids, D-fructose (fructopyranose) for carbohydrates, α-pinene for terpenes and quartz and B12 vitamin for miscellaneous.

Due to their natural origin, all these molecules have their stereogenic centers well defined with an enantiomeric excess higher than 99%. Because of this high enantiopurity, they are used as starting materials in synthesis, whether it is to 14 General Introduction

synthetize new chiral compounds or as a tool (e.g. catalyst) in other pathways to enantiopurity such as asymmetric synthesis and resolution.

2.1.1. For synthesis of enantiopure compounds There are many examples in which chiral natural molecules are being used as starting material in a longer synthetic pathway yielding new enantiopure molecules. As the world keeps growing, new compounds of interest are needed in all possible fields, especially in pharmacy, where chirality is almost unavoidable. Starting from an already chiral and enantiopure molecule becomes an advantage no chemist would turn his back on. In a non-exhaustive way, the four following examples will detail this application of the chiral pool.

First, David Y.-K. Chen et al.47 developed a synthesis of the right fragment of the antibiotic Platensimycin starting from bioavailable R-carvone (figure 19). Platensimycin is a potent and selective inhibitor of β-ketoacyl-(acyl-carrier protein) synthase II (FabF), which participates in fatty acid biosynthesis. Second, Xin-Zhi Chen et al.48 developed a three-step chiral pool synthesis of (S)-6-Nitroindoline-2- carboxylic Acid from L-Phenylalanine (figure 19), yielding the product with an overall yield of 53% and high enantiomeric excess (>99.5%). (S)-6-Nitroindoline-2- carboxylic Acid is a building block for several alkaloids but also pharmaceutical drugs.

Figure 19 Two examples of chiral pool syntheses yielding compounds of interest. On the left, a source of the molecule is illustrated while on the right it is the application of the compound of interest that is illustrated.

Carsten Bolm et al.49 developed a chiral pool four-step synthesis of an ionic liquid starting from L-valine (figure 20). Ionic liquids are green alternatives to organic 15 General Introduction solvents for several chemical reactions. As a final example, Wünsch Bernhard et al. developed a chiral pool five-step synthesis of a piperazine-alcanol starting from the conjugated base of L-aspartic acid (figure 20). The piperazine-alcanol synthetized showed activity as a σ1 receptor antagonist and can therefore be developed as an antidepressant or anti-amnesic drug.

Figure 20 Two examples of chiral pool syntheses yielding compounds of interest. On the left, a source of the molecule is illustrated while on the right it is the application of the compound of interest that is illustrated.

2.1.2. Total synthesis of existing biological compounds In a similar way, there are several examples of chiral pool syntheses leading to natural compounds.50–52 Chemists are always on the lookout to develop total synthesis pathways to biological compounds. Though these syntheses require the use of all the chiral tools at disposal, starting from an already enantiopure chiral molecule is a plus that decreases the numbers of step toward the final product. In a non-exhaustive way, the four following examples will detail this application of the chiral pool.

First of all, Henrik H. Jensen et al.53 developed a 17-step total synthesis of the nortropane iminosugar calystegine A3 starting from the sugar D-glucose (figure 21). Because of its structural resemblance to monosaccharides, calystegine A3 is a potential inhibitor of glycoside hydrolases, in particular β-glucosidase. It can be developed as a treatment for diabetes. A second example was developed by R. W. Longmore et al.54 showing a 6-step synthesis of the very important ascorbic acid molecule, aka vitamin C, starting from the bacteria-synthetized chlorodiol presented in figure 21. Vitamin C, though found in many aliments is often lacking in our diets and needs to be taken as dietary supplement. 16 General Introduction

Figure 21 Two examples of chiral pool syntheses yielding a naturally occurring molecule of interest. On the left, a source of the molecule is illustrated while on the right it is the application of the compound of interest that is illustrated.

Third, W. R. F. Goundry et al.55 developed a synthesis of the antibiotic (+)- Apiosporamide starting from (−)-citronellol (figure 22). As a last example, S. J. J. Danishefsky et al.56 developed a synthesis of Migrastin from commercially available (S,S)-tartaric acid (figure 22).

Figure 22 Two examples of chiral pool syntheses yielding a naturally occurring molecule of interest. On the left, a source of the molecule is illustrated while on the right it is the application of the compound of interest that is illustrated. 17 General Introduction

Migrastatin is a compound of interest, which shows a potential inhibiting of the metastasis of cancer cells. Thus, it could be used in chemotherapy.57

2.1.3. As a tool in asymmetric synthesis and resolution The chiral pool is very attractive for some resolution techniques, especially in the context of diastereomeric salt resolution. One of the resolution agents is often a compound from the chiral pool like tartaric acid58 or mandelic acid59, which, associated with the molecule of interest, reacts to form a diastereomeric salt.

In enantioselective heterogeneous catalysis, most of the chiral catalysts used either are from the chiral pool or derived from chiral pool molecules. This assessment is similar in the field of homogeneous catalysis, where several efficient chiral derived from chiral pool molecules. It must be added that usually, the elements of chirality are left unchanged while the molecules are being substituted with additional functional groups in order to reach the desired properties.60 Furthermore, all the categories of molecules stated in figure 17 are concerned. For instance, L-Proline can be derived as a chiral reagent for the isomerization of to allylic alcohols when associated to zinc.61 R-mandelic acid is derived to form a catalyst, which when combined with is used for the hydrogenation of imines62 and enamides.63 Fructose can also be used e.g. with ZnO for the bromination of carbon-carbon double bonds.64 A derivative of Pinene can be used alongside terbutyl-lithium to reduce with an excellent selectivity (ee=99%)65. Finally, Quartz in association with copper, nickel, and platinum can be employed as a support for the dehydrogenation of racemic alcohols, though selectivity is quite low.64

2.2. In-stream: Asymmetric synthesis As part of the chiral toolbox, asymmetric synthesis is an ever-growing set of reactions allowing chemists to access enantiopure compounds. Asymmetric synthesis can be defined as a set of reactions, which convert an achiral molecule into a chiral enantio- enriched -even enantiopure- one upon the action of a chiral reagent.66 Though generally true, this definition fails to account for asymmetric induction starting with a chiral molecule. To account for these latter also, asymmetric synthesis is better defined as a reaction introducing a chiral enantio-enriched feature where the molecule did not possess such a feature at the onset of the action. Several cases of asymmetric reactions can be distinguished. 18 General Introduction

Figure 23 The different possible asymmetric reactions explained using the addition of a nucleophile to the heterotopic faces of an .66These faces are prochiral and are called Si or Re depending on the rotation of the atoms in the face, when looked at from the considered side. If the atoms are going clockwise, the face is Re, if not it is called Si.

As a mean to illustrate, the addition of a nucleophile to an aldehyde is detailed (figure 23). An aldehyde is prochiral meaning it can become chiral in a one-step addition when an element of dissymmetry is added, e.g. reacting with a nucleophile (provided that the nucleophile differs from the already present functional groups). Upon the addition of the nucleophile to an achiral aldehyde, there are two possibilities: if the nucleophile is achiral, the possible transition states are enantiomerically related and there is no difference in reactivity, yielding a , but if the nucleophile is chiral, the transition states are diastereomerically related and consequently, show a different energy, ultimately leading to enantio-enrichment. The level of enantio- enrichment depends on the difference in stability between both transition states. This 19 General Introduction main idea of forming diastereomeric transitions states is the foundation of asymmetric synthesis. Nevertheless, real-live cases can be more complex, e.g. when the starting aldehyde is already chiral. In this case the transition states will always be diastereomerically related and lead to one configuration being favored over the other. When the added nucleophile is achiral, one uses the term asymmetric induction, with the resulting stereochemistry controlled inter-molecularly. If the nucleophile is chiral, the system is more complex, with the chirality of the nucleophile also affecting the resulting stereochemistry. Finally, a chiral catalyst (e.g. a metal catalyst) can be used, leading once more to diastereomerically related transition states whether the nucleophile is chiral or not. This is called catalytic asymmetric synthesis.

2.2.1. Asymmetric induction Asymmetric induction will be discussed first since it can be correlated to the chiral pool to some extent. Indeed, for an asymmetric induction to occur, reaction on a chiral substrate is required with this latter preferentially enantiopure, with a single chiral center. The chiral pool is a very practical library for asymmetric induction. However, one is not limited to this library, as asymmetric induction can be performed on any chiral substrate, whether naturally occurring or not. A good example of efficient asymmetric induction is the 1,2-asymmetric induction by addition of a nucleophile to carbonyl compounds with a chiral carbon in the alpha position (figure 24).

The major product can be determined according to the attack of the nucleophile. Indeed, the favored product will be the one with the more stable transition state i.e. the one where the nucleophile encounters the less steric hindrance. Depending whether there is a chelating agent present or not, two different models need be considered to efficiently ascertain the major product. In both cases, the major product remains the one obtained from the lowest energy transition state. When no chelating agent is present, the Felkin-Ahn model is to be considered with the largest or most electro-attractive group perpendicular to the carbonyl bond. If a chelating agent is present, then the Cram-chelate model needs to be taken into consideration with the chelated group next to the carbonyl function. The more RS and RM are sterically different, the higher the induced selectivity will be. Finally, with chelation, the selectivity of the favored product can be reversed, allowing to choose the desired outcome.

An application of this is illustrated in figure 25, as a step in the total synthesis of Zaragozic A,50 a potent inhibitor of mammalian squalene synthase.67 This enzyme is the first enzyme involved in sterol synthesis and consequently, the use of Zaragozic acid was shown to produce low in primates’ plasma.67,68 In this step, the chelate model is used to furnish a hydroxyl directed syn-1,2-reduction with an excellent yield of 93%.50 20 General Introduction

Figure 24 On top, the Felkin-Ann model showing the favored product for the nucleophilic attack of a carbonyl. On the bottom, the Cram-chelate model showing the favored product of the same reaction, but in presence of a chelating agent.

21 General Introduction

Figure 25 Hydroxyl directed syn-1,2-reduction of a ketone: a key step in the synthesis of Zaragozic acid.

2.2.2. Chiral auxiliaries In a similar way, the use of chiral auxiliaries was one of the first techniques developed in asymmetric synthesis, in 1978, by .69 The principle is similar to the one of asymmetric induction. Indeed, when using a , this latter will temporarily bond to the molecule on which one wishes to induce a defined stereochemistry. Once the asymmetric reaction done, the auxiliary is removed. The general schematic of a chiral auxiliary synthesis is shown in figure 26, illustrated by the synthesis of (R)-Limonene, where the chiral auxiliary is removed during the asymmetric reaction.

Figure 26 General scheme of the chiral auxiliary approach, with an application to the synthesis of a citrus fruit odorous molecule. S is the achiral substrate, A* the chiral auxiliary and P* the chiral product obtained. The parentheses mean that the reaction of the substrate-auxiliary entity can directly furnish the product by releasing the chiral auxiliary at the same time, like in the example below where the chiral auxiliary is R-BINOL. 22 General Introduction

This reaction provides a low yield of 29% with an enantiomeric excess of 64% and marks the first use of BINOL as a chiral auxiliary.70 Furthermore, a chiral auxiliary was elegantly used in the total synthesis of (+)-Discodermolide (figure 27),71 a very potent molecule that can only be found in small amount in a deep-sea sponge (0.002% of dry weight).72 (+)-Discodermolide has several application e.g. as highly potent antiproliferative agent,73 blocking the cell-division cycle in the mitosis part;74 or as a neuroprotective agent and could be used for treating Alzheimer’s disease.75

Figure 27 Chiral auxiliary synthesis of a precursor in (+)-Discodermolide total synthesis. Naturally occurring, (+)-Discodermolide was found to be a potent inhibitor of tumor cell growth.71

2.2.3. Chiral reagents When using a chiral reagent, stereoselectivity of the reaction can be controlled inter- molecularly as opposed to the asymmetric induction and the use of a chiral auxiliary. This part of asymmetric synthesis is the one with the least examples in the literature; mostly due to its poor efficiency because the two-stereogenic units making the diastereomeric transition state are not covalently bond. However, there are some efficient examples of chiral reagents used in the literature, like the one detailed in figure 28. 23 General Introduction

Figure 28 Part of the synthesis of Passifloricin A using a chiral reagent. Interestingly, the enantiomer of the chiral reagent is used later on in the reaction scheme to provide the alcohol with the opposite stereochemistry. S is the achiral substrate, R* the chiral reagent and P* the chiral product.

This is an allyl transfer to an aldehyde with the allyl directly connected to the chiral titanium reagent, with a yield of 89% and an enantiomeric excess higher than 98%76. This methodology was used for the synthesis of Passifloricin A, a natural compound, extracted from Passiflora foetida, which showed very good activity against leishmania amastigotes; a parasite responsible for the disease leishmaniasis; despite cytotoxic effects.77

2.2.4. Chiral catalysts Finally, the use of chiral catalysts is one of the main and most recent categories of asymmetric synthesis. When using chiral catalysts, the stereo-selectivity of the reaction is controlled intra-molecularly and enantio-enriched chiral products can be obtained from achiral substrates and reagents (figure 29). This method is very attractive because catalysts are generally present in small amounts, decreasing the cost. Moreover, recycling of the catalyst is possible. Finally, the major appeal remains the excellent results one can attain in selectivity.

Figure 29 General scheme of a catalytic asymmetric synthesis. S is the achiral substrate, R the achiral reagent, Cat* the chiral catalyst and P* the chiral product.

This domain was game-changing for asymmetric synthesis, to a point that the 2001 in Chemistry was awarded to three chemists working in this very domain, which we first illustrate using an example from one of these three leading scientists, Barry Sharpless. 24 General Introduction

Figure 30 Schematic displaying the enantioselectivity of Sharpless epoxidation depending on which chiral catalyst is used. An application of the reaction is then illustrated with the case of (S)-Propranolol synthesis.

The Sharpless epoxidation reaction,78 specific to allylic alcohols, allows the oxidation of to the corresponding with a total control in the stereochemistry of the product. The reaction generally uses tert-butyl hydroperoxyde (tBuOOH) and a i catalytic (5-10%) amount of titanium isopropoxide (Ti(O Pr)4) and diethyltartrate to form the chiral catalyst. Depending on the diethyltartrate configuration, the peroxide will be able to preferentially attack on either the Si or the Re face of the compound. Furthermore, by reversing the stereochemistry of the catalyst the other enantiomer can be accessed (figure 30). This reaction was applied to the synthesis of (S)-Propranolol, yielding a product with an enantiomeric excess higher than 95%.79 Propranolol is a drug used to treat high blood pressure and to prevent migraines. Out of the three chemists awarded the Nobel Prize in 2001, there was also Professor Ryori Noyori who worked on the enantioselective hydrogenation of β-keto-esters, using a chiral complex made of ruthenium and BINAP. The catalyst is in the same phase as the other reagents, making it a homogeneous chiral catalysis. Dihydrogen complexes the catalyst to yield the chiral ruthenium hydride specie, which reduces the ketone in the β position of the ester into an enantio-enriched chiral alcohol (figure 31). The chiral ruthenium hydride species displays a rigid structure, as drawn in figure 31. The β-keto-ester can complex the ruthenium from one of its two sides, each side leading to a given enantiomer. As shown, one side is preferred due to less steric hindrance, leading to the enantioselectivity of the reaction. This reaction was applied to a plethora of compounds,80 including the GABA receptor agonist GABOB, an anticonvulsant used to treat epilepsy.81

25 General Introduction

Figure 31 Schematic displaying the mechanism explaining the enantioselectivity of Noyori’s hydrogenation of β-keto-esters. An application of the reaction is then illustrated with the case of GABOB’s synthesis.

Last but not least, Professor William Standish Knowles was the third chemist to be award the Nobel prize in 2001 for his work in enantioselective hydrogenation. One of the most prominent examples of application of his work is the industrial synthesis of L-DOPA, whose configuration is fixed by the asymmetric hydrogenation (figure 32).82 L-DOPA is an amino acid, precursor to the neurotransmitter dopamine; used as a treatment in Parkinson’s disease.83 Finally, based on the previous reaction, enantioselective hydrogenation leading to (S)-Metolachlor was developed using a different catalyst.84 This reaction is industrially applied for the synthesis of tons of (S)-Metolachlor, a widely used herbicide.84

26 General Introduction

Figure 32 Examples of industrial applications of enantioselective hydrogenation: L-DOPA and (S)-Metolachlor synthesis.

2.3. Downstream: Resolution Instead of synthesizing an enantiopure compound, one can access enantiopurity by separating the enantiomers of a racemic mixture. This is called resolution. A racemic mixture can be resolved through using chemical and physical separation techniques. Chemical separation lies between true resolution and asymmetric synthesis since it uses the principle of asymmetric synthesis to “separate” the two enantiomers. Regarding physical separations, this domain is vaster with two sub-categories: kinetic- based and thermodynamic-based resolutions.

2.3.1. Chemical Separation: Kinetic Resolution Kinetic resolution was first achieved in 1858 by Louis Pasteur using a mold from Penicillium glaucum to isolate (S,S)-tartrate, the other being metabolized selectively.85 Synthetically speaking, the first kinetic resolution was reported by Marckwald and McKenzie in 1899 when they isolated (+)-mandelic acid from its racemic mixture by esterification with (−)-menthol. (+)-Mandelic acid was shown to react faster than its (−)-enantiomer toward the esterification. Upon, hydrolysis of the ester, (+)-mandelic acid was resolved (figure 33).86

Figure 33 Kinetic resolution of (R,S)-mandelic acid to (S)-(+)-mandelic acid using (−)- menthol. The wanted diastereomer is obtained with a conversion of 50% and a diastereomeric excess of 90%.

The principle behind Kinetic Resolution (KR) is the use of a chiral reagent or chiral catalyst (the latter being preferred for the same reasons as before) in order to have one enantiomer selectively react, forming an enantio-enriched –up to enantiopure– product, which is different in nature to the non-reacting enantiomer (figure 34).87 Consequently, the resulting mixture of the product and the unreacted enantiomer can be physically separated by crystallization, chromatography… 27 General Introduction

Figure 34 General scheme of kinetic resolution with below the energy diagram behind the principle of kinetic resolution. SR and SS are respectively the substrate (R)- and (S)-enantiomer. R* is a chiral reagent and R,Cat*is a chiral catalyst and a reagent. PS and PR are respectively the products of the reaction from the (S)- and (R)-enantiomer of the substrate. TSR and TSS are ‡ ‡ respectively the transition state of the reaction with the (R)- and (S)-enantiomer. ΔG R and ΔG S are respectively the difference of free energy between the enantiomer and that of its transition state. kS and kR are respectively the rate constant of the (S)- and (R)-enantiomer in the reaction.

To get one enantiomer to react selectively, the KR process is based on a strong difference of the reaction rate between each enantiomer. This is possible due to the use of a chiral reagent/catalyst. Just as asymmetric synthesis makes one enantiomer more favorable as the product of the reaction by using a chiral unit; KR makes one enantiomer more favorably react by using a chiral unit. The difference in reactivity of the two enantiomers resides in the difference in energy of the two possible transition states. It is without saying that the higher the difference of reaction rate between the two enantiomers, the more efficient the KR process becomes. Ideally, one enantiomer should be extremely slow reacting compared to the other one to ensure a good yield and enantiopurity. An example of the application of kinetic resolution (figure 35) can be found for the synthesis of the drug Mexiletine.88 Mexiletine is an analog of lidocaine,89 used for treating chronic pain; however it is also used to treat cardiac arrhythmia.90 28 General Introduction

Figure 35 An example of kinetic resolution: the synthesis of (R)-Mexiletine.

The kinetic resolution involves the attack of water to a chiral racemic epoxide. A cobalt catalyst, (R,R)-Salen Co (III) , is used to react with the unwanted enantiomer, leaving the desired epoxide unreacted. The mixture is purified by column chromatography to yield 43% of the epoxide with an enantiomeric excess higher than 99%. Thanks to the use of kinetic resolution, the racemic mixture was transformed into a mixture of two different compounds, which could easily be separated using chromatography. Chromatography can however, also be used to resolve a racemic mixture, as specified in the following section.

2.3.2. Physical separation: Chiral chromatography Chiral High Pressure Liquid Chromatography (HPLC) Chiral chromatography is one of the most important analytical tools developed as part of the chiral toolbox, allowing isolation of pure enantiomers. Normal chromatography cannot discriminate enantiomersa but by making the stationary phase chiral, resolutions became attainable. Chiral HPLC devices are now standard, even though industrially, the technique remains quite expensive. A given column can be efficient in separating two enantiomers, but completely fail to do so for two others. A vast library of different chiral columns therefore exist. To obtain a chiral environment for the stationary phase, enantiopure groups are immobilized onto the silica gel. The variety of chiral columns will depend on the type of compound one immobilizes on the silica gel. Throughout the years, there were eight different types of compounds that were immobilized on silica to yield chiral stationary phases (figure 36), either polar or non-polar:

1. Amylose / Cellulose 2. Cyclodextrins a Though true for racemates, in the case of an enantio-enriched mixture, it was observed that non-linear interactions can occur yielding fractions with different ee.84 29 General Introduction

3. Crown 4. Proteins 5. Amino-acid-copper complexes 6. π-acceptors / π-donors 7. Macrocyclic antibiotics 8. Synthetic polymers

Figure 36 The different types of molecules that can be immobilized on silica with a color code, green meaning the molecules can be found in the nature while red stands for molecules only obtained from synthesis. For 1, the molecular structure of cellulose is displayed. For 2, the structure of the α-cyclodextrin is shown. For 3, the structure of the 18-crown-6 crown is drawn. Though achiral, it can be derivated to yield CSP. For 4, the general structure of a protein was displayed. For 5, The general structure of an amino-acid-copper complex is shown. For 6, The structure of the π-acceptor N-(3,5-dinitrobenzoyl)–phenylglycine is drawn. For 7, the structure of Vancomycin is displayed as an example of macrocyclic antibiotic used in chiral columns. Finally for 8, the structure of poly(triphenylmethylmetacrylate) is shown.

The resolution works by diastereomeric interactions when the enantiomers enter in contact with the chiral stationary phase. The enantiomer forming the most stable diastereomer will be retained longer by the column, while the opposite enantiomer will be the first to elute.91,92 As a general rule of thumb, for a column to be able to discriminate between the two enantiomers, a minimum of three points of interaction are needed, as otherwise the column does not allow proper separation. This is based on the three-point interaction model applied to the formation of diastereomers.86 30 General Introduction

In the case of chiral chromatography, chiral separation relies on weak intermolecular interactions. The intermolecular forces at stake when it comes to chiral recognition are polar/ionic interactions, π-π interactions, hydrophobic effects and hydrogen bonding. In some cases, the chiral recognition happens through the formation of inclusion complexes. The quality of the separation strongly depends on the mobile phase and the temperature. Regarding the temperature, the lower the temperature, the better the chiral recognition. However, decreased temperature also leads to broadened chromatographic peaks, implying an optimal temperature needs to be sought for. It is stated that a difference in the diastereomer free energy of 0.03 kJ/mol will lead to resolution.93 As said earlier, not all types of column are suitable for each pair of enantiomers. As the efficiency of the column depends on the interactions of the molecules with the Chiral Stationary Phase (CSP), the different types of column are classified by the type of interaction they perform with the solute. This classification was proposed by Irving Wainer94 and owes its success to providing the type of molecule (or its required features) to be separated by a certain CSP:

Type I: These columns differentiate enantiomers by the formation of complexes through attractive interactions like hydrogen bonds, π-π interactions or dipole stacking. Type II: These columns differentiate enantiomers by a combination of attractive interactions and inclusion complexes. Type III: These columns differentiate enantiomers by formation of inclusion complexes. Type IV: These columns differentiate enantiomers by the formation of diastereomeric metal complexes. Type V: These columns differentiate enantiomers by a combination of hydrophobic and polar interactions.

1. Cellulose/amylose columns form the majority of type II columns in the Wainer classification. The different types of columns originate from derivation of cellulose/amylose polymers by various functional groups. Depending on these groups, the columns can be either normal or reverse phase. The most efficient column are the OD and AD phases (respectively cellulose and amylose derivated by tris-(3,5- dimethylphenylcarbamate), which are recognized for their ability to separate 80% of the most common chiral molecules. An example of column and efficient separation is given in figure 37.95 31 General Introduction

Figure 37 Chiral Pak IB column used for separating Terfenadine, an antihistaminic drug that used to be taken when dealing with allergies.96 The Chiral Pak IB column only differs from the Chiralcel OD column in the size of silica-gel used (5µm for Chiral Pak IB and 10µm for Chiralcel OD).

2. Cyclodextrin columns are type III columns and were first developed by Prof. D.W. Armstrong.97 These stationary phases result from binding cyclodextrin molecules to silica gel. Cyclodextrins are cyclic oligosaccharides that are synthetized by the action of Bacillus macerans amylase or cyclodextrin transglycosylase on starch. Three sizes are commercially available ,  and  respectively corresponding to 6, 7 and 8 glucopyranose units respectively (figure 38).

Figure 38 The three commercially available cyclodextrins : α, β and γ.98

The cyclodextrin molecule forms a truncated conical cavity whose diameter depends on the number of glucopyranose units.98 The cyclodextrin molecule has twice more secondary hydroxyl groups at the opening of the cavity than primary hydroxyl groups at the end of the cavity. Consequently, the cavity is relatively hydrophobic compared 32 General Introduction

to its borders. When a solute enters the cavity, its hydrophobic part will interact with it while its hydrophilic part will interact with the bordering hydroxyl, hence making a 3-point interaction. To be able to separate enantiomers on cyclodextrin the molecule ideally has an aromatic group for inclusion in the cavity. Cyclodextrin column selectivity depends on the size of the molecules to separate. α-Cyclodextrin columns separate single phenyl or napthyl groups. β-Cyclodextrin columns separate napthyl groups and heavily substituted phenyl groups. γ-Cyclodextrin columns are efficient for bulky -type molecules. The β-form has the widest application and an example is given in figure 39.99

Figure 39 Application of β-cyclodextrin columns for separating DL-Phenylalanine analogs.

Moreover, cyclodextrin columns can be used in an alternative Polar Organic Mode, similar to Non-Aqueous Reverse Phase (NARP) chromatography. This mode allows for separation of some molecules, which will not separate using aqueous mobile phases such as Propranolol. Finally, recently, cyclodextrins molecules were modified in order to expand the range of compounds they can resolve. The derivatives are formed by bonding various groups onto the surface hydroxyls of the Cyclodextrin cavity. For instance, they can be acetylated (analogous to acetylated cellulose columns, separated compounds include Atropine, Cocaine…)100 or reacted with 3,5- dimethylphenylcarbamate (analogous to chiralcell OD type column).

3. Crown ether columns are classified as type III. They are efficient for separating chiral primary amines by complexation. The amine must be protonated for the complex to form. Consequently, acidic mobile phases are used such as perchloric acid. The most commonly used crown ether is "18 Crown 6" (illustrated in figure 36) which has been made available commercially as the "Crownpak" column. Based on this structure, chiral Crownpak columns were developed using a chiral derivative of crown ether “18 Crown 6”, e.g. figure 40.101 33 General Introduction

Figure 40 Example of chiral crown ether based on “18 Crown 6”

4. Protein columns are Type V in Wainer’s classification where a protein is immobilized on silica gel. Several types of proteins were immobilized to yield a chiral phase, like for instance, Human α-Acid Glycoprotein (AGP), Human Serum Albumin (HSA) and Bovine Serum Albumin (BSA). Protein columns have possibly the broadest application of any chiral column but do not always offer the most efficient separation. In the case of AGP (a polypeptide with 181 amino acid residues and 40 sialic acid residues), an efficient separation of compounds with a ring close to the chiral center and at least one hydrogen bonding site not greater than 3 atoms away from the ring is achieved. AGP can also be covalently bonded to silica to produce a reverse-phase column. A very wide range of molecules have been separated using AGP columns.102

5. Chiral Exchange Chromatography (CLEC), developed by Davankov103 is classified as type IV. This method is mainly used for separation of Amino Acids by formation of diastereomeric metal complexes. The chiral phase is formed of a chiral Amino-acid-copper complex that is bound to silica or a polymeric stationary phase. Water stabilizes the complex by coordinating in an axial position. The most stable complex is determined due to steric factors: usually for one complex, water molecules are sterically hindered from coordinating to the copper. An example of CLEC is given in figure 41.104 34 General Introduction

Figure 41 O-Benzyl-L-serine cupper complex column used to efficiently separate (+/−)-3- hydroxy-4,5,6,6a-tetrahydro-3aH-pyrrolo[3,4-d]-isoxazole-4-carboxylic acid, a glutamate transporter inhibitor with neuroprotective effects.105

6. Brush type columns (π-donor/π-acceptor immobilized on silica) originate from the work of Bill Pirkle et al.106 and were developed to ensure a strong three-point interaction with one of the enantiomers. They are classified as Type I. There are two main types of stationary phases, π-acceptor or π-donor phases. An example of π- acceptor stationary phase is N-(3,5-dinitrobenzoyl)-phenyl glycine bonded to n- propylamino silica (figure 36). Brush type columns are able to separate a large range of compounds, including the ones containing π-donor/π-acceptor aromatic groups. Furthermore, they are easily synthesized but also commercially available at reasonable cost. Over the years, mixed-type columns (stationary phase mixing π- donor and π-acceptor properties) were developed. For instance, the WHELK-O1 phase possesses both a π-donor and a π-acceptor region within the same molecule, yielding very high selectivity when separating molecules with an aromatic group and a hydrogen-bonding group on the chiral center. This phase is known for separating compounds like Ibuprofen or Warfarin (figure 42).107

Figure 42 WHELK-O1 column used for separating racemic Warfarin, an anticoagulant. 35 General Introduction

7. Macrocyclic antibiotics have been immobilized on silica to form a new class of chiral column, pioneered by Dan Armstrong.108 Three antibiotics have already been successfully immobilized: Rifamycin, Vancomycin and Ticoplanin. These columns differentiate enantiomers based on π-π interactions, hydrogen bonding, inclusion complexation, ionic interactions and peptide binding. Vancomycin (figure 36) is a macromolecule with 18 chiral centers and three fused rings. It possesses a basket like structure with a single flexible sugar flap that can enclose a molecule sitting in the basket. A carboxylic acid and a secondary amine group, sitting on the rim of the basket, can take part in ionic interactions. The Vancomycin column is versatile, being available in Reverse, Polar and Normal phase mode. Vancomycin is efficient in separating amines, amides and esters. This chiral column has exceptional loading capacity and is suited for preparative applications.

8. Synthetic chiral polymers are used as type III chiral stationary phases. There are numerous types of polymers that are being used in chiral HPLC but they all fall into three categories depending on the type of polymerization: addition polymers, condensation polymers and cross-linked gels. A very efficient chiral polymer as a CSP is helical (M)-poly(triphenylmethylmetacrylate) (2D structure shown in figure 36). This polymer belongs to the first category, addition polymers. In the synthesis (anionic polymerization), the configuration of the polymer results from the use of either a chiral organolithium (initiator control) or a complex of organolithium with a chiral ligand (ligand control, example: n-BuLi with (−)-Sparteine).109,110

Chiral GC The first chiral gas chromatography (GC) was developed in 1966 by Gil-Av et al.111 Throughout the years, several chiral stationary phase were developed, until one in particular, Chirasil-Val made its way onto the market. This CSP results from the coupling of L-valine-tert-butylamide to dimethylsiloxane and (2-carboxypropyl) methylsiloxane (figure 43).

Figure 43 Structure of Chirasil-Val Chiral GC column. 36 General Introduction

However, chiral GC was limited in its application because of the racemization of the chiral selector alongside reduced selectivity and stability at high temperature. This was until the introduction of derivatized cyclodextrin-based chiral GC stationary phases in the late 80s. Their broad applicability and good thermal stability was a game-changer for chiral GC. Indeed, these columns were applied for separating various type of compound like for instance essential oils, terpenoids, chiral auxiliaries, chiral catalysts, amino acids… In total, over the years, three categories of chiral GC columns were developed: cyclodextrin derivatives, chiral amino acid derivatives and metal coordination complexes. Those CSP are highly efficient, sensitive and reproducible. Moreover, the analyses are quite fast and there is no liquid mobile phase used as opposed to chiral HPLC. Other less common CSP were developed, including chiral ionic liquids, polysaccharides, cyclopeptides, metal organic frameworks, porous organic cages and cyclofructans. However, only the cyclodextrin-based GC stationary phases and amino acid derivatives are commercially available and consequently the most widely used by far.112

2.3.3. Physical separation - Kinetically-based: Preferential Crystallization The following resolution technique thrives on kinetics like kinetic resolution, except that here the kinetic aspect is related to a crystallization process rather than a reaction. Preferential crystallization is based on the interesting though rare occurrence of conglomerates for a racemic mixture of enantiomers.

Solid-state of racemates In the solid state, chiral molecules can form either racemic compounds (both enantiomers crystallize together in a regularly-arranged crystalline structure), conglomerates (both enantiomers crystallize separately but with the same crystalline structure) or solid solutions (both enantiomers crystallize together in a crystalline structure with no definite order).113 Those three cases are illustrated in figure 44.

Figure 44 Schematic of the different crystalline states a racemic mixture can take.

A general definition of conglomerates was given by G. Coquerel114 as follows:

“A conglomerate is a mixture of mirror-image crystallized phases exhibiting symmetrical enantiomeric excesses.” 37 General Introduction

Generally, 5 to 10% of the racemic mixtures crystallize as a conglomerate while 90% form racemic compounds. Solid solutions are rare with an occurrence inferior to 1%.115 Whether a compound forms a conglomerate or a racemic compound depends on the affinity of one enantiomer for the other. Generally, the enantiomer prefers the one of opposite configuration for association, making an improved crystalline packing, leading to the formation of a racemic mixture. However, in some cases, the enantiomer has a better affinity with itself, leading to the formation of a conglomerate. For solid solutions, there is no significant difference depending on the nature of the enantiomers.

How to detect conglomerate forming systems

Figure 45 On the left, binary phase diagram of a conglomerate-forming compound. L stands for the molten phase, R for the (R)-enantiomer, S for the (S)-enantiomer. TR = TS is the melting temperature of the enantiopure compound and TC stands for the temperature when the racemic conglomerate melts. On the right, an isothermal cut of a ternary phase diagram of a conglomerate-forming compound in a solvent. L stands for the liquid phase containing both enantiomers dissolved in the solvent, R and S stands respectively for the (R)-enantiomer and the (S)-enantiomer in suspension. SR = SS is the solubility of the enantiopure compound in the solvent.

As conglomerates are rare, chemists had to develop methods to detect them. The first method is based on Louis Pasteur’s discovery working on tetrahydrated sodium ammonium tartrate,2 where shape recognition was used to separate macroscopic crystals. In this peculiar case, the {111} hemihedral faces were strongly expressed making the single crystals of opposite chirality macroscopically differentiable. A second method involves the establishment of a binary phase diagram or a ternary phase diagram (when adding a solvent to the mixture) (figure 45). Racemic conglomerates always show a lower melting temperature or increased solubility compared to the pure enantiomers (figure 45). However, constructing full phase diagrams is time-consuming when applied to more than one compound, and is therefore not an effective screening tool. A third method relies on isolating a single crystal of the racemic mixture and analyzing its enantiopurity by either polarimetry or 38 General Introduction

chiral GC/HPLC. A fourth method compares the solid spectroscopic properties (IR, PXRD, sNMR, raman) of the racemic mixture with that of a single enantiomer. When a conglomerate is formed, these properties are expected to be identical. For instance, in PXRD a conglomerate mixture has the same diffraction pattern as the pure enantiomer. Following the XRD method, a fifth method uses the analysis of single crystals by SC-XRD in order to determine the exact structure. A conglomerate always crystallizes in a chiral space group (no center of symmetry, mirror, glide mirror or inverted axis present as symmetry operation). It must be noted that racemic compounds can also crystallize in a chiral space group but this occurs very rarely. Finally, the last method to screen for conglomerates uses their chiral space group characteristic and is based on Second Harmonic Generation (SHG). SHG is the phenomenon where under irradiation with a light beam of wavelength λ, a material will generate a beam of wavelength λ/2 whose intensity depends on the χ(2) tensor. The interest lies in the fact that for centrosymmetric space groups the χ(2) tensor is zero. Then, this method allows for detection of non-centrosymmetric space group (no inversion center) from which chiral space group are a sub-category. This method is currently the most wide-spread for large conglomerate screening.114

The application of conglomerates to resolution Preferential crystallization (PC) is the principal resolution method used for resolving conglomerates, even though other less frequently used techniques exist (resolution by shape recognition,116 preferential nucleation,117 and replacing crystallization118).

Preferential crystallization114,119 is a stereoselective crystallization of only one enantiomer under conditions where the two enantiomers are expected to crystallize simultaneously. This kinetic crystallization of one enantiomer is possible by seeding the super-saturated racemic or nearly racemic (ee<20%) mixture with the pure enantiomer. Once the enantiomer crystallizes, the suspension is metastable and if let for too long, the other enantiomer will eventually start to crystallize. However, if filtered prior to this event, one can isolate enantiopure material (figure 46).

This method originates from 1866 when a student from Pasteur, Gernez, detected the entrainment effect but it was not until 1963 that the concept of preferential crystallization really bloomed when Secor resolved several molecules including DL- threonine using this approach.120 Later, at the end of the 20th century, researchers further developed PC in an industrial setting for the resolution of racemic drugs or intermediates in drug synthesis. An example is given in figure 46. A second example is the drug Omeprazole, which was resolved by preferential crystallization of its potassium salt ethanol solvate, in ethanol.121 39 General Introduction

Figure 46 Schematic of PC’s principle with an example of industrial application for preparation of D-4-hydroxyphenylglycine, an important intermediate in the synthesis of antibiotics like or Vancomycin.122

There are three major types of preferential crystallization, though this list is not exhaustive:114,119,123

Seeded Isothermal Preferential Crystallization (SIPC): Starting from a clear

solution at T1, the mixture is then quickly cooled down to T2 (temperature at which the solution is supersaturated and both enantiomers can crystallize) and seeded with the desired pure enantiomer. Seeded Polythermic Preferential Crystallization (S3PC): Starting from a

clear solution at T1, the mixture is cooled down at T3 (temperature at which the solution is saturated and only one enantiomer is stable in suspension- this implies that the starting mixture has to be slightly in excess of one enantiomer) and seeded with the desired pure enantiomer. Then, the

suspension is slowly cooled down to TE (temperature at which the solution is saturated in racemic conglomerate (eutectic point)). When in equilibrium

at TE, the solution is racemic with the pure desired enantiomer in suspension.

Then, this suspension is cooled down to T2 with an adapted cooling profile. 40 General Introduction

Auto-Seeded Polythermic Programmed Preferential Crystallization (AS3PC): Starting from a thermodynamic suspension of the desired pure

enantiomer at T4 (temperature at which only one enantiomer is stable in suspension- this implies that the starting mixture has to be slightly in excess

of one enantiomer), the mixture is then cooled down to TE and follows the same path as for S3PC. This method was also carried out by evaporation of the solvent instead of temperature range.124

Out of the three, AS3PC is the one giving the best results because of a better control and the use of auto-seeding, providing the system with a consequent initial population of crystals. For any of them, when carried out on larger scale, filtration becomes critical as the other enantiomer can always crystallize out at a given moment.114,119

Process-wise, preferential crystallization can be single batch but in order to improve process yield, two or three-parallel batch processes were developed.125 The two-batch continuous process (figure 47) works either with a saturated feed vessel with a racemic mixture and a vessel where crystallization of one enantiomer occurs or with two vessels in which preferential crystallization occurs. The feed vessel is a vessel continuously furnished with racemic conglomerate, keeping its concentration constant. In the first case, the liquid phase of the feed vessel is transferred to the crystallization vessel while the liquid phase of the crystallization vessel is transferred to the feed vessel. The feed vessel is a saturated suspension of the racemic conglomerate, generally kept at a higher temperature than the crystallization vessel. Only its liquid phase is fed to the crystallization vessel. The role of the feed tank is to provide the crystallization vessel with a supersaturated solution (since this vessel is kept at a lower temperature than the feed one). Doing so, it will increase the concentration of the crystallizing enantiomer in solution, improving yield and process time. In the latter case, one vessel crystallizes the R-enantiomer and the second vessel the (S)-enantiomer. The liquid phase of the vessel crystallizing the (R)-enantiomer is depleted in the (R)-enantiomer while the opposite occurs for the (S)-crystallizing vessel. As for the first case, this process results in both yield and process time improvement. The major difference between both set-ups is that with a feed tank the first one will be able to furnish a high quantity of a given enantiomer, dependent on how long the process will run, while the other gives access to both enantiomers. 41 General Introduction

Figure 47 On the left, the first case is drawn, with a feed tank and a crystallization vessel. On the right, the second case is drawn with two crystallization vessels. Each crystallization vessel is seeded with a given enantiomer at the start of the process.

Regarding the three-pot process (figure 48), it involves two crystallization vessels and one feed vessel. The feed vessel is transferred to both crystallization vessels and each crystallization vessel is transferring its liquid phase to the feed vessel. In a first vessel, the (R)-enantiomer is being crystallized while in the other vessel the (S)-enantiomer is sought. This process is a combination of both two-pot processes, combining both advantages: high amounts for both enantiomers with a decreased process time.

Figure 48 Three-pot preferential crystallization with a feed tank and two crystallization vessels. Each crystallization vessel is seeded with a different enantiomer at the beginning.

Finally, one-pot processes can also be fine-tuned to increase yield and reduce process time by avoiding nucleation of the other enantiomer. An example can be found with continuous preferential crystallization (CPC1). In this process, the liquid phase, where crystallization occurs, is continuously renewed with clear supersaturated racemic mixture while the excess of volume is regulated by overflow.125 CPC1 was successfully developed for the resolution of several amino acid, including glutamic acid,126 a non-essential amino acid serving under its anionic form as a precursor in the synthesis of gamma-aminobutyric acid (GABA) (figure 49).127 42 General Introduction

Figure 49 One-pot continuous preferential crystallization (CPC1) schematic with an example of industrial application for the resolution of glutamic acid.

2.3.4. Physical separation - Thermodynamically-based: Diastereomeric resolution As for asymmetric synthesis, the difference in properties between two diastereomers can also be used to develop physical separation. In this case, the use of diastereomeric salts by reacting a brønsted acid with a brønsted base allows separating enantiomers by e.g. crystallization, distillation… Indeed, mixtures of diastereomers can crystallize in three forms, analogous to the ones observed for simple chiral compounds. For the majority of cases, diastereomers crystallize separately in a conglomerate-like fashion. From time to time, however, they can crystallize as a quasi-racemate in a racemic compound-like behavior (1:1 association of two structurally similar and configurationally quasi-enantiomeric compounds) or as a solid solution.128 In the case of diastereomeric salts, one partner needs to be racemic and the other enantiopure in order to form a pair of enantiopure diastereomers.129 Interestingly, when an enantiomer is resolved using this method, the opposite enantiomer can be resolved under identical conditions using the resolving agent of opposite handedness (Marckwald’s rule). It must be noted that diastereomeric pairs can also be formed for systems other than base-acid systems, like e.g. the case of chiral alcohols where a chiral acid can be used to esterify the alcohol and form a pair of diastereomeric esters.130 However, this is far less developed and used. Depending on the physical means used to separate, there are different diastereomeric resolution schemes possible: 43 General Introduction

Mechanical resolution: The first ever diastereomeric resolution accomplished by Pasteur under the microscope.2,3 This method can be interesting for producing seed crystals. Resolution by distillation: By using half an equivalent of the compound to separate, only the most stable salt will form, leaving one enantiomer unreacted. If this latter is volatile, it can be distillated -preferentially under reduced pressure to avoid the use of high temperature, which will induce degradation-, and a pure enantiomer can be obtained from the distillate in the best scenario. For example, this method was used to separate 1- phenylethylamine enantiomers by using tartaric acid to yield an optical purity of 22%.131 1-phenylethylamine is a very useful resolving agent for diastereomeric salt formation. Resolution by extraction: This method is usually carried out with half an equivalent of resolving agent. Sometimes, an achiral additive can be added. The idea is to use two non-miscible solvents, one where the diastereomeric salt is soluble and one where the remaining unreacted enantiomer is soluble. was for instance resolved by this method, using sodium- hydrogen-(R,R)-tartrate in a mixture of water and . The salt went to the aqueous phase while the unreacted (S)-amphetamine went to the organic phase.131 Resolution by supercritical extraction: Using a supercritical fluid (for

example CO2) and half an equivalent of resolving agent, the salt remains undissolved while the free enantiomer is extracted. For example, (S)- Ibuprofen can be extracted with an optical purity of 41.7% by using 1- phenylethylamine as the resolving agent.131 Resolution by sublimation: This method can be carried out with either half an equivalent or one equivalent of resolving agent.131 Resolution by (fractional) crystallization: This is the major and most developed diastereomeric method.131

2.3.4.1. Resolution by (fractional) crystallization This method consists of reacting the racemic molecule with one equivalent of the enantiopure resolving agent in the appropriate crystallization solvent to form a pair of diastereomeric salts. The thermodynamically more stable diastereomer will be less soluble and crystallize out from the solution while the other remains dissolved (best- case scenario). By filtrating the solid in suspension, only one diastereomer (hence one enantiomer of the racemate) is retrieved. Then, the salt can be broken by adding either a strong base or acid and the pure enantiomer either crystallized or extracted. The chemical model behind diastereomeric resolution by crystallization is based on the solubility equilibrium constant of each diastereomeric salt, KsR and KsS. Those constants can be directly related to the solubility of each salt in the solvent (sR and sS), with the selectivity depending on the solubility difference between the two 44 General Introduction

diastereomers and the yield depending on the solubility of the most stable diastereomer. Furthermore, the selectivity is pH-dependent.

In order to qualify the efficiency of a resolution, the resolvability parameter is introduced, defined as the product of the yield y of the resolution and the enantiomeric excess E of the resolved compound.

푆 = 푦 × 퐸

푠R − 푠S 푆 = 0.5퐶0

132 With C0 the initial concentration of the salt.

An application of this method can be found in figure 50 with the resolution of DL- diacetyl-lysine by diastereomeric salt formation with (R)-1-Phenylethylamine as the resolving agent in an ethanol/acetone mixture (0.29:0.71) in order to produce L-lysine, an essential amino acid. The resolution produced a salt with an enantiomeric excess of 83% and a 29% yield.133

Figure 50 Diastereomeric salt resolution of DL-diacetyl-lysine with (R)-1-Phenylethylamine.

Though resolution by crystallization is most of the time carried out as described above, there are variants of it that also found application. The first one is to use half an equivalent of resolving agent in combination with an achiral additive. Generally, the additive is a strong base or acid like hydrochloric acid (HCl) or sodium hydroxide (NaOH). This is called the Pope-Peachy method and it was applied to the resolution 45 General Introduction of in a 1:1 water/toluene biphasic system at 80°C.134 The additive was NaOH and the resolving agent was Leelamine, a diterpene amine (figure 51). Naproxen is a nonsteroidal anti-inflammatory drug (NSAID) used to treat fever, pain…

Figure 51 Pope-peachy resolution in two immiscible solvents of (R,S)-Naproxen with Leelamine (dihydrobiethylamine).

Another variant was also used in the previous example: resolution using two immiscible solvents.135 This method is employed when not all the components can be solubilized in the crystallization solvent. Generally, racemic bases are soluble in organic solvents and meagerly in water while acidic resolving agents are more soluble in water than in organic solvents. A third variant is the Salt-Salt resolution. Here, the racemate and resolving agents are first reacted with an inorganic base or acid (depending if they are respectively acidic or basic) to form their inorganic salts, which are soluble in water. Then, when mixed together, the diastereomeric salts will form alongside an inorganic salt. The first one will precipitate while the latter will remain in water. For instance, (R)-desylamine, a building block for the synthesis of morphine- like analgesics,136 was resolved with (S)-mandelic acid using their respective salt with

HCl and NH4OH in water (figure 52). After recrystallization in ethanol, the hydrochloride salt of (R)-desylamine was obtained enantiopure.137

Figure 52 Salt-salt resolution of (R)-desylamine with (S)-mandelic acid.

A fourth variant uses not only the cake but also the filtrate for recovery. One removes the resolving agent from the filtrate and adds its mirror image to precipitate the other enantiomer. With this, both enantiomers of the racemate can be obtained.132 A fifth variant uses di-functional resolving agents like tartaric acid. In this case, resolution can be performed using either an acidic or neutral salt with ratios varying from 1:1 to 46 General Introduction

1:4 (acid:base). The acidic salt occurs when only one acid function of tartaric acid react with the base while the neutral salt occurs when both functions have reacted with the base. Depending on which one forms under which condition, resolution will be possible or not.132,138 A sixth variant is called reversed resolution. In this version, the enantiopure compound, resolved by using resolving agent A, will serve as the resolving agent to resolve racemic A. This was applied for example for the resolution of N-acetyl-DL-tryptophan with 1-phenylethylamine (figure 53). N-acetyl-L- tryptophan was first resolved with half an equivalent (S)-1-Phenylethylamine and half an equivalent of NaOH in ethanol, with an enantiomeric excess of 99% for a yield of 36.5%. Then, N-acetyl-L-tryptophan was used to resolve (R,S)-1-phenylethylamine in ethanol by the Pope-Peachy resolution previously described, using half an equivalent of HCl. N-acetyl-L-tryptophan is an intermediate in the synthesis of L- tryptophan, another essential amino-acid.139

Figure 53 Reversed resolution of L-tryptophan with (S)-1-Phenylethylamine.

Regarding this last variant, it is generally coupled with a racemization reaction (when applicable) for both molecules to recycle the filtrate of each resolution. This was the case in the example above. Furthermore, the use of racemization is not uncommon in resolution processes since this allows going beyond the 50% yield limitation to tackle a theoretical 100% yield in the eutomer, by recycling the distomer. The resolution process then becomes a deracemization process.

2.4. Recycling the distomer: Deracemization By definition, a deracemization process is the combination of a resolution process with a racemization reaction. Those two steps can be carried out one after the other, 47 General Introduction though deracemization becomes advantageous when both steps are run simultaneously, in a one-pot process. Obviously, to carry out a deracemization process, the molecule has to be racemizable, meaning the chiral center one wishes to resolve has to be able to interchange configuration under specific conditions. Consequently, not all molecules are able to undergo deracemization processes.

2.4.1. Racemization Racemization is the potential of a compound to go from one configuration to the opposite. For a compound with more than one chiral center with only one being affected, racemization is called epimerization. When this process is applied to an enantio-enriched chiral compound, it results in a statistical 50/50 mixture of both enantiomers and thus in production of a racemic mixture. The interest of racemization for a deracemization process lies in the potential to transform the distomer into the eutomer. Racemization processes can be classified in three different categories:

Chemical racemization: racemization through a . This category is the most developed and can be applied to different types of chiral centers. Enzymatic racemization: racemization under the influence of an enzyme, generally a racemase. This type of racemization is very selective to certain type of compounds but also sensitive to the reaction conditions. For instance, Lactate Racemase140 and Racemiase141 racemizes lactic acid, ω- Transaminase racemizes chiral primary amines through a ketone intermediate142 and Glutamate Racemase racemizes glutamic acid…143 Thermal/photochemical racemization: This type of racemization majorly consist in the use of UV or heat to induce racemization in axial chiral compounds like BINOLs,144 other biphenyl compounds,145 and axially chiral alkene146 or some terpene compounds like α-pinene.147 Some studies also showed the application of microwave148 or ultrasound149 as a racemization inducer combined with a chemical racemization reaction.

Chemical racemization Chemical racemization can be carried out over a large range of compounds through different types of reactions, though some more common than others. The following non-exhaustive list exemplifies possible reactions (figure 54):

Chiral carbonyls: when the chiral center is in the α-position, racemization generally occurs via the keto-enol equilibrium, this is one of the most common racemization reactions since it occurs simply via base or acid catalysis. When the chiral center is in β-position with an amino group on it, racemization can occur via an organocatalytic aza-Michael/retro-aza- Michael reaction.150 48 General Introduction

Chiral can be racemized under acidic catalysis (e.g. HCl) in several solvents.151 Chiral alcohols: Secondary alcohols can be racemized under the action of Ru(II) catalysts in the presence of a base like NaOH.152,153,154 Furthermore, benzylic alcohols can be racemized using Vanadium catalysts like vanadyl 155 sulfate (VOSO4) or by simple acid-catalysis like in the case of adrenaline.156 Chiral amines can be racemized under Palladium or Ruthenium complex catalysis, through a planar intermediate.154 Racemization through an imine intermediate can also be done using Schiff bases.157 Chiral Halides can be racemized by copper(I)-catalysed atom transfer 146 reactions or by reversible SN reaction. Chiral alkanes: Some alkanes as 3-methylhexane can be racemized in presence of a Ni/kieselguhr catalyst through hydrogen exchange.158

Figure 54 On top are represented the general structure of, in red a chiral carbonyl in α- position, in dark blue a chiral carbonyl in β-position, in pink a chiral sulfoxide, in purple a chiral alcohol, in green a chiral amine, in light blue for a chiral halide (chloride) and in orange a chiral alkane. On the bottom, the mechanism of the keto-enol equilibrium for chiral carbonyls in α-position is drawn with on top the acid catalyzed mechanism and at the bottom the base- catalyzed mechanism. 49 General Introduction

2.4.2. Chemical deracemization: Dynamic Kinetic Resolution Dynamic Kinetic Resolution (DKR) is literally the upgrade of kinetic resolution using a racemization reaction while the kinetic resolution occurs. The key parameters for DKR to be efficient is the rate of the racemization reaction. It has to be fast, in particular faster than the rate of the kinetic resolution reaction in order to ensure rapid equilibrium between the two enantiomers while one is being depleted by the chemical transformation reaction (figure 55). The rest of the process remains the same and follows the same path as KR. An efficient DKR process is illustrated in figure 55 with the use of Noyori’s hydrogenation of β-keto-esters combined to an in-situ racemization (keto-enol equilibrium) to produce protected L-threonine with an enantiomeric excess of 98% and a syn/anti selectivity of 99/1.159

Figure 55 On top, the general schematic of a DKR process. On the bottom, an example of DKR process leading to the synthesis of protected L-threonine. Upon de-protection, the essential amino acid L-threonine can be obtained.

It must be noted that Dynamic kinetic resolution of chiral amines and chiral alcohols has been extensively studied.154 Finally, there exists a variant of DKR where two diastereomers are equilibrating in solution while the equilibrating chiral site is being 50 General Introduction

attacked and inversed by an achiral reagent160 (figure 56). An example is given on the same figure, where tert-butyl 1-methyl-2-oxoimidazolidine-4-carboxylate is epimerized in solution (solvent: Hexamethylphosphoramide, HMPA) by triethylamine while the same chiral center is inversed by a SN2 reaction of benzylamine.

Figure 56 On top, DKR with an epimerization of two diastereomers while one reacts faster than the other with an achiral reagent or catalyst. On the bottom, an example of this DKR variant.

This deracemization produced the product with a yield of 96% and with a selectivity between the two diastereomers of 94/6.161

2.4.3. Physical deracemization – Kinetically based: In a similar manner, the use of conglomerate forming compounds is compatible with an in-situ racemization reaction to provide a deracemization processes. 51 General Introduction

Preferential crystallization with an in-situ racemization Preferential crystallization can be combined with an in-situ racemization to yield a Crystallization-Induced Asymmetric Transformation (CIAT) and overcome the 50% max yield limitation of the preferential resolution method.162–164 The presence of an in-situ fast racemization can be used as illustrated in figure 57.165

Figure 57 On the left, an isothermal section of the ternary phase diagram of the deracemization process when seeding with the (R)-enantiomer. The dashed lines are inaccessible regions of the ternary diagram due to the presence of the in-situ racemization. On the right, an example of application with a racemic precursor of Paclobutrazol is given. Paclobutrazol is a fungicide working as a mixture of enantiomers of diastereomers. Deracemization of the precursor would allow enantiopurity of each diastereomer.

Upon seeding with a pure enantiomer, these will grow, depleting the liquid composition in this very enantiomer (black arrow). Using the racemization, this solution imbalance is pulled back to the racemic composition, compensating the preferential crystallization of the enantiomer and preventing the crystallization of the complementary enantiomer (red arrow). The liquid composition slowly evolves up to the blue line during the process. Furthermore, the total composition of the system follows the yellow arrow; up to the point where the liquid composition has reached, the eutectic composition and no more solid can crystallize out. This type of process was for example used to deracemize a racemic precursor of Paclobutrazol, a plant growth retardant and fungicide,166 with a yield of 70% and enantiomeric excess of 96.1%.167 A second example of preferential crystallization combined with in-situ racemization can be found with the deracemization of racemic Narwedine to (−)- Narwedine (figure 58). 52 General Introduction

This process was carried out in a 9:1 mixture of ethanol/triethylamine at 40°C and in presence of 1 mol% of Galanthamine to catalyze the racemization reaction, which occurs through an opening of the chiral ether bond and re-closure in a Michael-like reaction. In this specific case, no seeding was used and the seed was supposed to be generated in solution by a chiral amplification of Galanthanine (one of the two enantiomers racemizes faster than the other one). After two cycles, a yield of 90% with an enantiopure product is reached.168 Recently, continuous processes of preferential crystallization have been developed for batch deracemization.169

Figure 58 Deracemization of racemic Narwedine, an intermediate in the synthesis of Galanthamine, a drug used to treat Alzheimer’s desease.170

Viedma Ripening A second physical deracemization process was recently discovered by C. Viedma in

2005 when strongly grinding a saturated solution of sodium chlorate (NaClO3). Though achiral, sodium chlorate is known for crystallizing in a chiral space group, making it chiral at the solid state with two enantiomeric crystals that can grow. Upon abrasive grinding (glass balls), Viedma showed that the racemic crystal composition was brought to a homochiral composition over time.171 This groundbreaking discovery was then developed for organic molecules able to racemize in solution.172– 176 For sodium chlorate, since it is achiral, when redissolved, chirality was lost and this could be affiliated to a “racemization in solution”. The phenomenon behind Viedma ripening is a combination of four effects:

Racemization in solution (r): renders possible the deracemization Ostwald ripening (1): Bigger crystal are more energetically favorable than smaller ones. Because of the Gibbs-Thompson effect (solubility depends on crystal size for a molecule), small crystals will have a tendency to dissolve while bigger crystals will grow. Attrition (2) : breaks the crystals in smaller clusters 53 General Introduction

Cluster incorporation (3): Small clusters instead of dissolving can be incorporated to a bigger crystal. The more large crystals there are, the higher the chance of incorporation.

When combined, these four effects give rise to the Viedma ripening (figure 59).177–180

Indeed, a small difference in the ratio between two chiral molecules or crystals in suspension starts the Viedma ripening (practically this can be done by seeding in order to control the configuration of the deracemized molecule/crystal). Then, because of the attrition, big crystal will break into smaller clusters. However, since one enantiomer had more crystals in suspension, it will statistically have more big crystals and clusters in suspension. These clusters will be more favorably incorporated while the other enantiomer will see its small clusters more favorably dissolve. This will create a solution imbalance between the two enantiomers. Racemization will restore the solution to a racemic mixture. However, the other processes mentioned above continue depleting the solution more strongly in the enantiomer that shows an excess of solid state in suspension.181 Though this process is quite recent, some variants have already been developed.

Figure 59 Mechanism behind the Viedma ripening of a chiral molecule toward the S enantiomer according to J. M. McBride.177

One variant replaces the abrasive grinding by temperature cycling (heating/cooling sequences) to yield an efficient process, with faster deracemization rates in general.182,183 The rate of both approaches depend on the size of the crystals, the initial enantiomeric excess, the growth and dissolution rates of each enantiomers and the 54 General Introduction

grinding speed/temperature cycling respectively. Figure 60 illustrates both processes with on the left, an example of application for abrasive grinding and on the right an example for temperature cycling.

Deracemization of (R,S)-2-(benzylideneamino)butyramide was carried out in toluene using DBU as base to racemize the compound in the liquid phase. Starting from an enantiomeric excess of 14% upon seeding, (S)-2-(benzylideneamino)butyramide was isolated with a yield of 77% and an enantiomeric excess of 99%. Upon hydrolysis, the precursor of levetiracetam was obtained with an enantiomeric excess >99%.184 Regarding the deracemization of (R,S)-(2-methylbenzylidene)-phenylglycinamide, it was carried out by temperature cycling after 22h in methanol using again DBU as a base to induce racemization in the liquid phase. (R)- or (S)-Phenylglycine was added as an additive (2%) to induce deracemization to either one of the two enantiomers starting from a zero enantiomeric excess. Using (R)-phenylglycine led to (S)-(2- methylbenzylidene)-phenylglycinamide while (S)-phenylglycine led to (R)-(2- methylbenzylidene)-phenylglycinamide.183

Figure 60 On the left, application of abrasive grinding Viedma ripening in the deracemization of (R,S)-2-(benzylideneamino)butyramide to yield (S)-2-aminobutyramide, a precursor of two racetam molecules: Levetiracetam & Brivaracetam. Those two compounds have applications as anti-convulsant drugs for treating epilepsy. On the right, application of temperature cycling Viedma ripening in the deracemization of (R,S)-(2-methylbenzylidene)-phenylglycinamide to either one of its enantiomers depending on the chirality of the additive Phenylglycine.

2.4.4. Physical deracemization – Thermodynamically- based: crystallization-induced diastereomer transformation Finally, diastereomeric resolution can also be upgraded to a deracemization process by adding a racemization reaction in solution, yielding the so-called Crystallization- Induced Diastereomer Transformation (CIDT). The thermodynamic basis of CIDT processes involves two diastereomers, D1 & D2 that can equilibrate in solution while 55 General Introduction one precipitates from the solution. Racemization is an entropy-driven transformation that reaches its equilibrium when the enantiomeric excess is null. Consider LD1 the 푎D2 solubility product of D1 in the solvent, LD2 the solubility product of D2, and K= the 푎D1 racemization equilibrium constant between D1 and D2 in solution. Following situations can occur:128,146,185

If KLD1 = LD2 the system is at the equilibrium and no racemization occurs anymore

If KLD1 > LD2 the mixture of D1 & D2 would eventually be transformed into

pure D2 providing that the deracemization is total

If KLD1 < LD2 the mixture of D1 & D2 would eventually be transformed into

pure D1 providing that the deracemization is total

In this process, the maximum yield obtainable depends on both the racemization equilibrium and the solubility constants. D1 and D2 can be chiral or non-chiral (E/Z double-bond isomers). Generally, D1 and D2 are a pair of diastereomeric salts but they can also be two chiral moieties covalently bonded together. Either way, one center needs to be able to equilibrate in solution while the other remains invariant. A review on CIDT displays a large range of examples of application with most of them carried out in a one-pot approach repeated several times to give access to higher yield by decreasing the volume of solvent used at every step.128 In general, the salts are able to racemize simply by heating186–189 with some cases where the racemization even occurs at room temperature.190,191 Some other case display the use of a racemizing agent often combined with heat (e.g. 1,8-diazabicyclo(5.4.0)undec-7-ene (DBU),192 Benzaldehyde193, Butanal193). Occasionally, the racemizing agent is an excess of the amine partner of the salt.193 Figure 61 shows two examples of CIDT.

Figure 61 Two examples of CIDT applications: On top, the deracemization of a precursor of a growth hormone secretagogue. Below, the deracemization of BINOL, an important chiral ligand for asymmetric catalysis.

The first example shows the deracemization of a precursor of a growth hormone secretagogue that was synthesized by Pfizer. When the precursor, a pyrazolopyridine 56 General Introduction

derivative was stirred at 38°C, overnight, with (S,S)-tartaric acid in a tri-mixture of dichloromethane, acetone and water, the (R)-enantiomer precipitated out as a diastereomeric salt with tartaric acid. The deracemization gave the enantiopure compound with a yield of 88%. Regarding the racemization, it is suspected to occur via a retro-Mannich reaction.194,195 The second example shows a deracemization process with an axial-chirality compound BINOL. The CIDT works by formation of a diastereomeric pair with CuCl2 and (S)-amphetamine. Moreover, the racemization of BINOL in solution is working because of the presence of a copper-amine complex when working at temperatures higher than 20°C. The process yielded 87% of (−)- BINOL with an enantiomeric excess of 91%.196

3. Co-crystallization: A solid-state discovery full of promises As displayed above, they are different pathways to enantiopurity, each with its pros and cons but also its range of application. In addition, more recently, diastereomeric resolution received an update with the use of co-crystallization. The oldest known co- crystal dates from the first half of the 19th century with the molecule quinhydrone (figure 62), which is known for its use in potentiometric titrations.197 But, it was during the end of the 19th and beginning of the 20th century that co-crystals were consistently studied198 up to 1922 and Pfeiffer’s Organische Molekulverbindungen199 followed in 2005 by Herbstein’s Crystalline Molecular Complexes and Compounds.200 Those contributions gave rise to the field of crystal engineering and more specifically to the co-crystal field as it is known nowadays.

Figure 62 Molecular structure of Quinhydrone.

3.1. Solid-state state of the art Before going into details on co-crystallization, let us first lay the foundation of the solid-state. A solid can either be amorphous or crystalline. A solid is crystalline when it is ordered at the solid-state. Indeed, a crystal is a three-dimensional arrangement in which a pattern (the asymmetric unit) is repeated through different symmetry operations giving rise to a unit cell. The regular repetition of this unit cell by 57 General Introduction translation yields the crystal. On the contrary, a solid is called amorphous when its order at the solid state is only local.

When a molecule can exist at the solid state in different crystalline forms, polymorphism occurs. For example, carbon can crystallize as diamond or graphite, and the properties of the substance changes alongside with the structure as e.g. diamond and graphite.198 The first occurrence of polymorphism dates back to 1823, when Mitscherlich identified different crystal structures for a compound in a series of arsenate and phosphate salts.201 There are three categories of polymorphism:201–203

Packing polymorphism: when molecules arrange themselves in different ways leading to different structures. Conformational polymorphism: when molecules have multiple conformational possibilities, different crystal structures can be obtained with the same molecule in different conformations. polymorphism: when molecules crystallizes with different supramolecular interactions. In crystal engineering, a supramolecular synthon is a group of atoms from at least two molecules involved in an electrostatic interaction.

During crystallization, polymorphism occurs because of competition between kinetic and thermodynamic factors. Indeed, one polymorph is always thermodynamically stable while the others are metastable. Consequently, for a metastable polymorph to crystallize, kinetic factors like long-range intermolecular interactions during nucleation must prevail over thermodynamic ones.

Then, when adding a second molecule to the mix, multi-component molecular crystals are obtained (solid solutions, solvates (and hydrates), salts and co-crystals).204 Solid solutions are observed when one molecule is incorporated into the crystalline system of the other, in ratio depending on the ratio between the two molecules before mixing. Solvates form when a solvent molecule (liquid molecule at room temperature) crystallizes with a solid one. If the solvent is water, the assembly is called a hydrate. Salts are formed when at least two organic molecules react upon a Bronsted acid-base reaction and crystallize together. Regarding co-crystals, its status requires more discussion.

3.2. What is co-crystallization? Indeed, co-crystallization has often been defined in different ways with not one definition being taken as the universal one.205 For instance, one version was given during a meeting sponsored by the Indo-U.S. Science and Technology Forum (IUSSTF) stating that ‘‘Co-crystals are solids that are crystalline single-phase materials composed of two or more different molecular and/or ionic compounds generally in a stoichiometric ratio which are neither solvates nor simple salts’’.206 This 58 General Introduction

definition is a relative definition of co-crystals with respects to their peers (solvates and salts), because salts and solvates can be called co-crystals depending on how strict, the definition is and vice versa. This led to a recent suggestion that co-crystal should just be employed as a synonym of “multi-component molecular crystal”.207

In this work, we use the term co-crystal for a supra-molecular structure where both components are solid at room temperature.207,208 This is the general accepted definition, but we will add a twist to it: when considering chiral components, they only need to be solid at room temperature for one of their possible forms racemic or enantiopure forms, no matter if they are conglomerate, racemic compound or solid solution forming. In order to differentiate co-crystals from co-salts, it is of utmost importance that the constituting components are in a neutral form (zwitterions being considered as neutral) and interact only by non-covalent non-ionic interactions such as hydrogen bonds, halogen bonds, π-π or van der Waals interactions. Furthermore, in a co-crystal each constituent has a definite stoichiometric ratio. Because of the strictness of this definition, one must consider hybrid classes of co-crystals in order to account for special cases of supra-molecular structures that are more complex than just a co-crystal, just a salt or just a solvate (figure 63):

When a co-crystal structure contains components that are not neutral, one speaks of an ionic co-crystal. For example, when combining Piracetam and lithium salts such as LiCl, ionic co-crystals can be obtained.209 When a co-crystal structure also contains a liquid component, one speaks of co-crystal solvates or if the component is water, one speaks of co-crystal hydrates.

59 General Introduction

Figure 63 Diagram detailing the difference between co-crystal, co-salts, ionic co-crystals, hydrates, co-crystal hydrates, solvates and co-crystal solvates. For the first question, when applying the twist, it includes chiral compound with their racemic or enantiopure form as liquid at room temperature (RT).

Finally, in some cases, polymorphs of co-crystals can exist, meaning that the same stoichiometry of compounds leads to a different crystalline structure. Polymorphs of co-crystals are not so common but an example can be found with the 1:1 co-crystal of caffeine and Anthranilic acid that displays two polymorphs (figure 64).210 There are different possible ratios when two molecules form a co-crystal, the more common being 1:1 & 2:1/1:2.

Figure 64 Hydrogen bonding system of the 1:1 co-crystal of caffeine and Anthranilic acid, which gives rise to two different polymorphs depending on how the system is organized (packing polymorphism). In blue are the atom and bonds involved in the different hydrogen bonds of the system.

As shown in figure 64, the hydrogen bonds involved are N-H···O and O-H···N types. For co-crystallization to occur intermolecular interactions are necessary. Generally, when speaking about co-crystallization the major interactions are hydrogen and halogen bonding and in a less important way π-π interactions (figure 65).

Figure 65 Illustration of the intermolecular forces at stakes in co-crystallization. 60 General Introduction

From those three intermolecular forces, originate the typical synthons for co- crystallization. The main ones are:211

Carboxylic acids, Amides, Alcohols & Amines: donor and acceptor of hydrogen bonding, acceptor of halogen bonding. Halogen (except for fluorine), Esters, Ketones, & (acceptor of hydrogen bonding & acceptor of halogen bonding) Iodine close to fluorine atoms (halogen bonding donor) Aromatic rings (acceptor and donor of π-π interactions)

Additionally, carboxylic acids and amides functions are able to carry out double hydrogen bond interactions either in a homo-synthon or hetero-synthon manner (figure 66).

Figure 66 Homo- and hetero-synthons of carboxylic acids and amides.

3.3. The importance of co-crystallization for the pharmaceutical industry Co-crystals are of importance for the pharmaceutical industry for several reasons:212,197

Forming a co-crystal with an Active Pharmaceutical Ingredient (API) changes its solid-state properties; hence, solubility and bioavailability, stability, dissolution rate and even compressability and tableting properties can be improved. Using the right co-former, one can tune the solid-state of the API to display the optimal physico-chemical properties, and this without changing covalent bonds. Co-crystal formation is not limited to acidic or basic compounds as is salt formation and can be applied in principle to a larger range of API. Regulatory-wise, a co-crystal is seen as a polymorph of the API and thus, there is no need for supplementary biological studies when patenting a co- crystal of an API, as opposed to patenting a salt of an API. Indeed, the FDA only demands an in-vitro evaluation based on dissolution and/or solubility to prove that the API dissociates completely from the co-former. There are numerous suitable co-formers that are “Generally Regarded As Safe” (GRAS) including food additives and other generic API like aspirin and acetaminophen. 61 General Introduction

Co-crystallization can also be used to combine two API in the same formulation. Multidrug co-crystals are interesting when developing combination therapies, trying to prevent multidrug resistance or increasing the action of one drug through a synergic effect with another one. Co-crystallization can be used for developing new chiral tools and then expand the chiral toolbox.

Figure 67 On top, hydrogen bonding system of the asymmetric unit of the 1:2 co-crystal of Nitrofurantoin and 4-hydroxybenzamide. At the bottom, hydrogen bonding system of the asymmetric unit of the 1:1 co-crystal monohydrate of Lamivudine (lower molecule) and Zidovudin (upper molecule). The pink, red and purple partial hydrogen bonds refer to hydrogen bonds that are formed with the repetition of this asymmetric unit. Each partial bond of a color is completed with the one of the same color by a translation.

An example of a co-crystal of an API improving its physico-chemical properties is given in figure 67. In this figure, the 1:2 co-crystal of the Antibiotic Nitrofurantoin with the GRAS co-former 4-hydroxybenzamide is displayed. Nitrofurantoin is an 62 General Introduction

antibacterial drug used for the oral treatment of genito-urinary tract infections. It possesses both low solubility and permeability. Furthermore, its dissolution rate and bioavailability in commercial tablets decreases over time depending on the temperature and relative humidity conditions. Upon co-crystallization with 4- hydroxybenzamide, aqueous solubility and dissolution rates were strongly improved while photo-stability was efficiently improved.213 A second example is given in figure 67 to illustrate drug-drug co-crystals, showing the 1:1 co-crystal hydrate between Lamivudine and Zidovudine, both are anti-viral drugs active against HIV. Both molecules are nucleoside reverse-transcriptase inhibitors (NRTI). They inhibit the enzyme reverse transcriptase that HIV uses to replicate and therefore decrease its spread.214,215

3.4. Chirality and co-crystallization: expansion of the chiral toolbox Though fine-tuning the physical properties of a molecule is very important, especially in pharmaceutical industries, what we will focus on further is the expansion of the chiral toolbox with the use of co-crystallization. Indeed, co-crystallization can be employed for chiral resolution:

When forming a conglomerate by co-crystallization, co-crystallization can be used in preferential crystallization. For instance, ionic co-crystals of DL-proline with lithium salts (LiCl, LiBr…) gave conglomerate co-crystals susceptible to be resolved using preferential crystallization.216 When forming a pair of diastereomers by co-crystallization of a racemate with an enantiopure molecule, co-crystallization is an alternative method to diastereomeric salt formation and can then be used in the context of resolution. This can also be achieved when only one of the enantiomers of the racemate is able to co-crystallize with the enantiopure compound. Whether the co-former is chiral or achiral, co-crystallization can be used to crystallize liquid chiral compounds or difficult-to-crystallize solid ones in order to determine their absolute configuration through single-crystal X-ray diffraction. Co-crystallization has the advantage over salt formation that co- crystals dissociate once in solution, rendering the HPLC analysis of a co- crystal possible with the same conditions as those of the separate compounds.217

3.4.1. Co-crystallization, an alternative to diastereomeric salt resolution

When forming a salt, the sine qua non condition is a difference in pKA value between the acid and the base (ΔpKA = pKA(Acid)-pKA(Base)) higher than 2-3. In an opposite manner, if ΔpKA<0, then a co-crystal will generally form. For values in between, the 63 General Introduction interaction between the basic and acid compound can be either ionic (salt) or electrostatic (co-crystal) or even a mix of both. Since a co-crystal is based on electrostatic interactions like hydrogen bonding and not strong interactions like ionic bonding, enantiospecific co-crystallization can occur whereas it cannot in salt formation. Indeed, when mixing an acid racemate with a basic chiral resolving agent two diastereomeric salts are formed, one being more stable than the other. In the case of co-crystallization, this can also occur but since the interactions are weaker, the energy difference between the diastereomers is less pronounced. Furthermore, in most cases one only forms a co-crystal with one of the enantiomers of the racemate. In this case, one speaks of enantiospecific co-crystallization, since only one of the diastereomers can form.216 For instance, this was the case for Levetiracetam [(S)-2- (2oxopyrrodin-1-yl)-butanamide], which made a 1:1 enantiospecific co-crystal with (S)-mandelic acid or with S-ibuprofen (figure 68). In both cases, R-mandelic acid and (R)-ibuprofen did not co-crystallize with Levetiracetam. Additionally, by mirror image effect, when mixing (R)-Etiracetam with the (R)-enantiomers of mandelic acid and ibuprofen, the (R-R)-co-crystal would form while the (R-S)-form would not.218,219

Figure 68 The asymmetric unit of the enantiospecific co-crystal of Levetiracetam with on the left (S)-mandelic acid and on the right (S)-Ibuprofen. Both were crystallized in acetonitrile. With (S)-Ibuprofen, the co-crystal is a drug-drug one. The pink and red partial hydrogen bonds refer to hydrogen bonds that are formed with the repetition of this asymmetric unit. Each partial bond of a color is completed with the other one of the same color by a repetition of the relevant molecule.

Then, resolution of (R,S)-2-(2oxopyrrodin-1-yl)-butanamide [(R,S)-Etiracetam] with (S)-mandelic acid was performed upon fulfillment of following two conditions, true for any co-crystallization resolution:

Co-crystallization should be able to occur under the chosen solvent and temperature conditions. 64 General Introduction

The stability zone of the enantiospecific co-crystal (or alternatively the co- crystal with one of the two starting compounds) has to cross with the racemic composition line; since the resolution starts with a racemic mixture.219

An example of ternary phase diagrams of an enantiospecific co-crystal system suitable for co-crystallization resolution is given in figure 69. The stability zone of the pure co-crystal that crosses the racemic composition line is highlighted with an orange frame.

Figure 69 Isoplethal section of a quaternary diagram of a suitable co-crystal for co- crystallization resolution. This diagram is the one of the 1:1 co-crystal of Levetiracetam and S- mandelic acid (LSMA co-crystal) in acetonitrile, based on experimental results at −10 °C. R-1 is (R)-Etiracetam, S-1 is (S)-Etiracetam also called Levetiracetam and S-2 is (S)-mandelic acid.219

The efficiency of the resolution will depend on the solubility of the co-crystal in the chosen solvent at the chosen temperature and on the difference of solubility between the two diastereomers – or alternatively the difference of solubility between the co- crystal and the non-crystallizing enantiomer. Both are strongly influenced by the crystallization solvent.220 Consequently, the conditions of the resolution are as important as the identity of the co-crystal for an efficient resolution to occur. Two examples of co-crystal resolutions are given in figure 70. 65 General Introduction

Figure 70 On the left the asymmetric unit of the 1:1 co-crystal between (R)-Praziquantel and L-Malic acid is displayed. Praziquantel is chiral drug used to treat several types of internal/gastrointestinal parasitic worm infections. (R)-Praziquantel was resolved for its racemate using this co-crystal in a resolution process in ethyl acetate.221 On the right, the asymmetric unit of the 1:1 co-crystal between (R)-BINOL and (R,R)-1,2-diaminocyclohexane is shown. Thanks to this co-crystal, R-BINOL could be resolved from its racemate.222 The pink and red partial hydrogen bonds refers to hydrogen bonds that are formed with the repetition of this asymmetric unit. Each partial bond of a color is completed with the other one of the same color by a repetition of the relevant molecule.

3.4.2. Co-crystallization, an alternative for deracemization? Up to this point, co-crystals were shown suitable for chiral resolution in a similar manner as diastereomeric salt formation. Nevertheless, diastereomeric resolution can be upgraded to a deracemization process if the racemic mixture to resolve can be racemized in solution. What we aim for in this thesis, is to use a similar thinking on the co-crystallization scheme, combining racemization with co-crystallization resolution, to develop a co-crystallization induced deracemization process, which is applicable to compounds that do not form salts, and has not been described in literature yet.

4. Bibliography 1. Wainer, I. W. & Marcotte, A. A. Stereochemical terms and concepts. An overview. in Drug stereochemistry. Analytical methods and pharmacology (ed. Wainer, I. W.) 25– 34 (Marcel Dekker Publisher, 1993). 2. Challener, C. A. Overview of chirality. in Chiral drugs 3–14 (Ashgate Publisher, 2001). 3. Drayer, D. E. The early history of stereochemistry. in Drug stereochemistry. Analytical methods and pharmacology (ed. Wainer, I. W.) 1–24 (Marcel Dekker Publisher, 1993). 4. Clayden, J., Greeves, N., Warren, S. & Wothers, P. Stereochemistry. in (ed. Oxford University Press) 381–406 (Oxford University Press, 2008). 5. Cahn, R. S., Ingold, C. K. & Prelog, V. The specification of asymmetric configuration in organic chemistry. Experienta 81–124 (1956). 6. Kalsi, P. S. Stereochemistry Conformation and Mechanism. (2005). 7. A. Pavlov, V. & I. Klabunovskii, E. Homochirality Origin in Nature: Possible 66 General Introduction

Versions. Curr. Org. Chem. 18, 93–114 (2014). 8. Blackmond, D. G. The Origin of Biological Homochirality. Cold Spring Harb. Perspect. Biol. 11, 1–10 (2019). 9. Nguyen, L. A., He, H. & Pham-huy, C. Chiral Drugs : An Overview. Int. J. Biomed. Sci. 2, 85–100 (2006). 10. Jayakumar, R., Vadivel, R. & Ananthi, N. Role of Chirality in Drugs. Org. Med. Chem. Int. J. 5, 555–661 (2018). 11. Burke, D. & Henderson, D. J. Chirality: a blueprint for the print. Br. J. Anesth. 88, 563–576 (2002). 12. McConalthy, J. & Owens, M. J. Stereochemistry in drug action. J. Clin. Psychiatry - Prim. Care Companion 5, 70–73 (2003). 13. Bentley, R. Chemical methods for the investigation of stereochemical problems in biology. in Stereochemistry (ed. Gamm CT) 49–112 (Elsevier Biomedical Press, 1982). 14. Landoni, M. F. & Soraci, A. Pharmacology of chiral compounds: 2-Arylpropionic acid derivatives. Curr. Drug Metab. 2, 37–51 (2001). 15. Mehvar, R., Brocks, D. R. & Vakily, M. Impact of stereoselectivity on the parmacokinetics and pharmacodynamics of antiarrhythmic drugs. Clin. Pharmacokinet. 41, 533–558 (2002). 16. Leitereg, T. J., Guadagni, D. G., Harris, J., Mon, T. R. & Teranishi, R. Chemical and sensory data supporting the difference between the odors of the enantiomeric carvones. J. Agric. Food Chem. 19, 785–787 (1971). 17. NASA. A natural way to stay sweet. https://spinoff.nasa.gov/Spinoff2004/ch_4.html. 18. Sasajima, K. & Sinskey, A. Oxidation of l-glucose by a Pseudomonad. Biochim. Biophys. Acta - Enzymol. 120–126 (1979). 19. Ariens, E. J. Stereoselectivity of bioactive agents: general aspects. in Stereochemistry and Biological activity of Drugs (eds. Ariens, E. J., Soudijn, W. & Timmermans, P. B. M. W. M.) 1–33 (Blackwell Scientific, 1983). 20. Drayer, D. E. Parmacodynamic and pharmacokinetic differences between drug enantiomers in human: an overview. Clin. Pharmacol. Ther. 40, 125–133 (1986). 21. Waldeck, B. Three-dimensional pharmacology, a subject ranging from ignorance to overstatements. Pharmacol. Toxicol. 93, 203–210 (2003). 22. Stoschitzky, K., Lindner, W. & Zernig, G. Racemic beta-blockers-fixed combinations of different drugs. J. Clin. Basic Cardiol. 1, 15–19 (1998). 23. Rahn, K., Hawhina, A., Kersting, F. & Planz, G. Studies on the antihypertensive action of propranolol in man. Naunyn-Schmirdeberg’s Arch. Pharmacol. 319–323 (1974). 24. Verster, J. C., Brady, K., Galanter, M. & Conrod, P. Drug Abuse and Addiction in Medical Illness: Causes, Consequences and Treatment. (Pringer Science & Business Media, 2012). 25. Powell, J. R., Ambre, J. J. & I., R. T. The efficacy and toxicity of drug stereoisomers. in Drug stereochemistry. Analytical methods and pharmacology (ed. Wainer, I. W.) 245–270 (Marcel Dekker Publisher, 1988). 26. Davies, N. M. & Teng, X. V. Importance of chirality in drug therapy and pharmacy practice. Implication for psychiatry. Adv. Pharm. 1, 242–252 (2003). 27. Marzo, A. & Heftman, E. Enantioselective analytical methods in pharmacokinetics with specific reference to genetic polymorphic metabolism. J. Biochem. Biophys. Methods 54, 57–70 (2002). 28. Raikumar, S. V. Thalidomide: Tragedy past and promising future. Comment. Mayo Clin. Proc. 79, 899–903 (2004). 29. Eriksson, T., Bjorkman, S., Roth, B., Fyge, A. & Hoglund, P. Stereospecific determination, chiral inversion, chiral inversion in vitro and pharmacokinetics in humans of the enantiomers of thalidomide. Chirality 44–52 (1995). 67 General Introduction

30. Hutt, A. J. & Valentová, J. The chiral switch : the development of single enantiomer drugs. Acta Fac. Pharm. Univ. Comen. L, 7–23 (2003). 31. Gellad, W. F., Choi, P., Mizah, M., Good, C. B. & Kesselheim, A. S. Assessing the Chiral Switch: Approval and Use of Single-Enantiomer Drugs, 2001 to 2011. Am. J. Manag. Care 20, 90–97 (2014). 32. Calcaterra, A. & Acquarica, I. D. Journal of Pharmaceutical and Biomedical Analysis The market of chiral drugs : Chiral switches versus de novo enantiomerically pure compounds. J. Pharm. Biomed. Anal. 147, 323–340 (2018). 33. Therapeutics Initiative. Comparative effectiveness of proton pump inhibitors. Ther. Lett. 99, 1–2 (2016). 34. McConnell, O. et al. Enantiomeric separation and determination of absolute stereochemistry of asymmetric molecules in drug discovery—Building chiral technology toolboxes. Chirality 19, 658–682 (2007). 35. Flack, H. D. & Bernardinelli, G. The use of X-ray crystallography to determine absolute configuration. Chirality 681–690 (2008). 36. Albright, A. L. & White, J. M. Determination of Absolute Configuration Using Single Crystal X-Ray Diffraction. in Metabolomics Tools for Discovery 149–162 (Humana Press, 2013). 37. Flack, H. D. & Bernardinelli, G. The Use of X-ray Crystallography to Determine Absolute Configuration. Chirality 20, 681–690 (2008). 38. Weisman, G. R. Nuclear Magnetic Resonance Analysis Using Chiral SolvatingAgents in Asymmetric Synthesis. (1983). 39. Rinaldi, P. L. The determination of absolute configuration using nuclearmagnetic resonance techniques. Prog. Nucl. Magn. Reson. Spectrosc. 291–352 (1982). 40. Sullivan, G. R., Dale, J. A. & Mosher, H. S. Correlation of configuration and 19F chemical shifts of alpha-methoxy-alpha-trifluoromethylphenylacetate derivatives. J. Org. Chem. 2143–2147 (1973). 41. Wesolowski, S. S. & Pivonka, D. E. A rapid alternative to X-ray crystallographyfor chiral determination: case studies of vibrational circular dichroism (VCD) to advance drug discovery projects. Bioorg. Med. Chem. Lett. 4019–4025 (2013). 42. Stephens, P. J., Devlin, F. J. & Cheeseman, J. VCD Spectroscopy for OrganicChemists. (CRC Press, 2012). 43. Polavarapu, P. L. Optical Rotation : Recent Advances in Determining the Absolute Configuration. Chirality 14, 768–781 (2002). 44. Stephens, P. J. et al. Determination of Absolute Configuration Using Ab Initio Calculation of Optical Rotation. Chirality 15, S57–S64 (2003). 45. Stephens, P. J., Devlin, F. J., Cheeseman, J. R., Frisch, M. J. & Rosini, C. Determination of Absolute Configuration Using Optical Rotation Calculated Using Density Functional Theory. Org. Lett. 4, 4595–4598 (2002). 46. Husain, P. A., Debnath, J. & May, S. W. HPLC-based method for determination of absolute configuration of alpha-chiral amines. Anal. Chem. 1456–1461 (1993). 47. Nicolaou, K. C., Pappo, D., Tsang, K. Y., Gibe, R. & Chen, D. Y. A Chiral Pool Based Synthesis of Platensimycin. Angew. Chem. Int. Ed. 944–946 (2008) doi:10.1002/anie.200705080. 48. Liu, J.-Q., Qian, C. & Chen, X.-Z. A Facile Chiral Pool Synthesis of (S)-6- Nitroindoline-2-carboxylic Acid from L -Phenylalanine. Synthesis (Stuttg). 403–406 (2010) doi:10.1055/s-0029-1217122. 49. Wasserscheid, P. & Bolm, C. Synthesis and properties of ionic liquids derived from the ‘ chiral pool ’† directly derived from the ‘ chiral pool ’ and therefore readily. Chem. Commu. 200–201 (2002). 50. Magauer, T. 2 . 10 Chiral Pool Synthesis : Chiral Pool Synthesis from Hydroxy Acids : Lactic Acid , Tartaric Acid , Malic Acid , and 2-Methyl-3-hydroxypropionic Acid. Comprehensive Chirality vol. 2 (Elsevier Ltd., 2012). 68 General Introduction

51. Gaich, T. 2 . 7 Chiral Pool Synthesis : Starting from Terpenes. Comprehensive Chirality vol. 2 (Elsevier Ltd., 2012). 52. Rinner, U. 2 . 9 Chiral Pool Synthesis : Chiral Pool Syntheses from cis - Cyclohexadiene Diols. Comprehensive Chirality vol. 2 (Elsevier Ltd., 2012). 53. Rasmussen, T. S. & Jensen, H. H. Chiral pool synthesis of calystegine A 3 from 2- deoxyglucose via a Brown allylation. Carbohydr. Res. 346, 2855–2861 (2011). 54. Banwell, M. G., Blakey, S., Harfoot, G. & Longmore, R. W. No Title. Aust. J. Chem. 137 (1999). 55. Williams, D. R., Kammler, D. C., Donnell, A. F. & Goundry, W. R. F. No Title. Angew. Chem. Int. Ed. 6715–6718 (2005). 56. Gaul, C., Njardarson, J. T. & Danishefsky, S. J. No Title. J. Am. Chem. Soc. 6042– 6043 (2003). 57. Shan, D. et al. Synthetic analogues of migrastatin that inhibit mammary tumor metastasis in mice. PNAS 102, 3772–3776 (2005). 58. Rosini, G. et al. Acid promoted CIDT for the deracemization of dihydrocinnamic aldehydes with Betti’s base. Green Chem. 12, 1747–1757 (2010). 59. Blacker, A. J. et al. No Title. Org. Process Res. Dev. 13, 1370–1378 (2009). 60. Blaser, H. The Chiral Pool as a Source of Enantloselective Catalysts and Auxiliaries. Chem. Rev. 935–952 (1992). 61. Tomioka, K. Asymmetric Synthesis Utilizing External Chiral Ligands. Synthesis (Stuttg). 541–549 (1990). 62. Kang, G., Cullen, W. R., Fryzuk, M. D., James, B. R. & Kutney, J. P. No Title. J. Chem. Soc., Chem. Commun. 1466–1467 (1988). 63. Koenig, K. E. Asymmetric Synthesis Vol. 5. (1985). 64. Blaser, H. Enantioselective synthesis using chiral heterogeneous catalysts. Tetrahedron: Asymmetry 2, 843–866 (1991). 65. Brown, H. C. & Ramachandran, P. V. The boron approach to asymmetric synthesis. Pure Appl. Chem. 63, 307–316 (1991). 66. Gawley, R. E. & Aube, J. Introduction, General Principales, and Glossary of Stereochemical Terms. in Principles of asymmetric synthesis 1–62 (Elsevier, 2012). 67. Bergstrom, J., Dufresne, C., Bills, G., Nallin-Omstead, M. & Byrne, K. Discovery, biosynthesis, and mechanism of action of the zaragozic acids: potent inhibitors of squalene synthase. Annu. Rev. Microbiol. 49, 607–639 (1995). 68. Do, R., Kiss, R., Gaudet, D. & Engert, J. Squalene synthase: a critical enzyme in the cholesterol biosynthesis pathway. Clin. Genet. 75, 19–29 (2009). 69. Corey, E. J. & Ensley, H. E. Preparation of an Optically Active Intermediate via Asymmetric Induction. J. Am. Chem. Soc. 97, 6908–6909 (1975). 70. Sakane, S., Fujiwara, J., Maruoka, K. & Yamamoto, H. Chiral leaving group. Biogenetic-type asymmetric synthesis of limonene and bisabolenes. J. Am. Chem. Soc. 105, 6154–6155 (1983). 71. Mickel, S. J. et al. Large-Scale Synthesis of the Anti-Cancer Marine Natural Product (+)-Discodermolide . Part 1 : Synthetic Strategy and Preparation of a Common Precursor. Org. Process Res. Dev. 8, 92–100 (2004). 72. Gunasekera, S. P., Gunasekera, M., Longley, R. E., Pierce, F. & Schulte, G. K. Discodermolide : A New Bioactive Polyhydroxylated from the Marine Sponge Discodermia dissoluta. J. Org. Chem. 55, 4912–4915 (1990). 73. Longley, R. E., Gunasekera, S. P., Faherty, D., McLane, J. & Dumont, F. Immunosuppression by Discodermolide. Ann. N.Y. Acad. Sci. 696, 94–107 (1993). 74. Taxol, P. T. et al. Discodermolide , A Cytotoxic Marine Agent That Stabilizes Microtubules More Potently Than Taxol. Biochemistry 35, 243–250 (1996). 75. Forman, M. S., Trojanowski, J. Q. & Lee, V. M.-Y. Neurodegenerative diseases: a decade of discoveries paves the way for therapeutic breakthroughs. Nat. Med. 10, 69 General Introduction

1055–1063 (2004). 76. Bouzbouz, S. A. & Cossy, J. Enantioselective allyltitanations: synthesis of the proposed structures for passifloricin A. Tetrahedron Lett. 44, 4471–4473 (2003). 77. Cardona, W. G., Quiñones, W. F. & Echeverri, F. L. Leishmanicidal Activity of Passifloricin A and Derivatives. Molecules 9, 666–672 (2004). 78. Borman, S. Asymmetric catalysis wins. Chemistry Nobel honors Knowles, Noyori, Sharpless for chiral syntheses. Chem. Eng. News 79, 5 (2001). 79. Katsuki, T. The synthesis of E-(2S,3S)-3-trimethylsilylglycidol and its conversion to (-)-propranolol. Tetrahedron Lett. 25, 2821–2822 (1984). 80. Noyori, R. Asymmetric Catalysis : Science and Opportunities (Nobel Lecture). Angew. Chem. Int. Ed. 41, 2008–2022 (2002). 81. Smith, M. B. Methods of Non-α-Amino Acid Synthesis. (CRC Press, 2013). 82. Knowles, W. S. Application of Organometallic Catalysis to the Commercial Production of L-DOPA. J. Chem. Ed. 63, 222–225 (1986). 83. Cools, R. Dopaminergic modulation of cognitive function-implications for L -DOPA treatment in Parkinson’s disease. Neurosci. Biobehav. Rev. 30, 1–23 (2006). 84. Allen, J. R. F. Metolachlor, S-metolachlor and their role within sustainable weed- management. Crop Prot. 17, 207–212 (1998). 85. Pasteur, L. C. . C. R. Acad. Sci. Paris 46, 615–618 (1858). 86. Marckwald, W.; McKenzie, A. Ueber eine principiell neue Methode zur Spaltung racemischer Verbindungen in die activen Bestandtheile. Ber. Dtsch. Chem. Ges. 32, 2130–2136 (1899). 87. Fiaud, J. C. & Kagan, H. B. Kinetic Resolution. in Topics in Stereochemistry (eds. Eliel, E. L. & Wilen, S. H.) 249–340 (John Wiley and Sons, Inc., 1988). 88. Sasikumar, M., Nikalje, M. D. & Muthukrishnan, M. Tetrahedron : Asymmetry A convenient synthesis of enantiomerically pure (R) -mexiletine using hydrolytic kinetic resolution method. Tetrahedron: Asymmetry 20, 2814–2817 (2009). 89. Canavero, S. & Bonicalzi, V. Drug dissection. in Central Pain Syndrome: Pathophysiology, Diagnosis and Management (ed. Press, C. U.) 286 (2011). 90. Manolis, A. S., Deering, T. F., Cameron, J. & Estes, N. A. Mexiletine: pharmacology and therapeutic use. Clin. Cardiol. 13, 349–359 (1990). 91. Earll, M. Online Guide to Chiral HPLC. http://serge.engi.tripod.com/Methods_files/Chiral.htm#8. (1999). 92. Lough, W. J. Classification of LC chiral stationary phases : Wainer Types I – V revisited ଝ. J. Chromatogr. B 968, 1–7 (2014). 93. Flanagan, R. J., Taylor, A. A., Watson, I. D. & Whelpton, R. High-Performance Liquid Chromatography. in Fundamentals of Analytical Toxicology 207–213 (John Wiley and Sons, Inc., 2007). 94. Wainer, I. W. Proposal for the classification of high performance liquid chromatographic chiral phases: how to choose the right column. Trends Anal. Chem. 6, 125–134 (1987). 95. Chiral Technologie Europe. Instruction manual for Chiralpak® IB columns. (2006). 96. Horak, F. Antialergic and Vasoactive Drugs for Allergic Rhinitis. in Allergy Frontiers:Therapy and Prevention (ed. Pawankar, R.) (Springer Science & Business Media, 2010). 97. Armstrong, D. W. Bonded phase material for chromatographic separations. (1985). 98. Crini, G. et al. Fundamentals and Applications of Cyclodextrins. in Cyclodextrin Fundamentals, Reactivity and Analysis 1–55 (Springer, 2018). 99. Taylor, P., Remsburg, J. W., Armstrong, D. W., Péter, A. & Tóth, G. LC Enantiomeric Separation of Unusual Amino Acids Using Cyclodextrin ‐ based Stationary Phases LC Enantiomeric Separation of Unusual. J. Liq. Chrom. Relat. Tech. 31, 219–230 (2007). 100. Armstrong, D. W., Han, S. M. & Han, Y. I. Separation of optical isomers of scopolamine, cocaine, homatropine, and atropine. Anal. Biochem. 167, 261–4 (1988). 70 General Introduction

101. Hichrom. Chiral Technologies: CROWNPAK. https://www.hichrom.com/Products/Manufacturers/ChiralTechnologies/CT- Brands/CROWNPAK.htm (2018). 102. Hermansson, J. Direct liquid chromatographic resolution of racemic drugs using α1- acid glycoprotein as the chiral stationary phase. J. Chromatogr. A 269, 71–80 (1983). 103. Davankov, V. A. Resolution of racemates by ligand exchange. in Chromatography Advances in chromatography Vol 18 (ed. Dekker, M.) 139 (1980). 104. Ianni, F. et al. Last ten years (2008–2018) of chiral ligand‐exchange chromatography in HPLC: An updated review. J. Sep. Sci. 42, 21–37 (2019). 105. Colleoni, S. et al. Neuroprotective effects of the novel glutamate transporter inhibitor (-)-3-hydroxy-4,5,6,6a-tetrahydro-3aH-pyrrolo[3,4-d]-isoxazole-4-carboxylic acid, which preferentially inhibits reverse transport (glutamate release) compared with glutamate reuptake. J Pharmacol. Exp. Ther. 326, 646–656 (2008). 106. Pirkle, W. H. & House, D. W. Chiral HPLC stationary phases 1. J.Org.Chem. 44, 1957–1979 (1979). 107. Zheng, J. & Shamsi, S. A. Brush-type chiral stationary phase for enantioseparation of acidic compounds: Optimization of chiral capillary electrochromatographic parameters. J. Chromatogr. A 1005, 177–187 (2003). 108. Armstrong, D. W. et al. Macrocyclic Antibiotics as a New Class of Chiral Selectors for Liquid Chromatography. Anal. Chem. 66, 1473–1484 (1994). 109. Nakano, T. Optically active synthetic polymers as chiral stationary phases in HPLC. J. Chromatogr. A 906, 205–225 (2001). 110. Okamoto, Y. Chiral Polymers for Resolution of Enantiomers. J. Polym. Sci. Part A Polym. Chem. 47, 1731–1739 (2009). 111. Gil-Av, E., Feibush, B. & Charles-Sigler, R. Separation of enantiomers by gas liquid chromatography with an optically active stationary phase. Tetrahedron Lett. 7, 1009– 1015 (1966). 112. Patil, R. A., Weatherly, C. A. & Armstrong, D. W. Chiral Gas Chromatography. Chiral Analysis (Elsevier B.V., 2018). doi:10.1016/B978-0-444-64027-7/00012-4. 113. Srisanga, S. & ter Horst, J. H. Racemic Compound , Conglomerate , or Solid Solution : Phase Diagram Screening of Chiral Compounds. Cryst. Growth Des. 10, 1808–1812 (2010). 114. Coquerel, G. Preferential Crystallization. in Novel Optical Resolution Technologies 1– 51 (Springer-Verlag, 2007). 115. Jacques, J., Collet, A. & Wilen, S. Enantiomers, Racemates and Resolutions. (Krieger Publishing Company, 1994). 116. Tobe, Y. The reexamination of Pasteur’s experiment in Japan. Mendeleev Commun. 13, 93–94 (2003). 117. Addadi, L., Mil, J. Van & Lahav, M. Useful impurities for optical resolutions. 2. Generality and mechanism of the rule of reversal. J. Am. Chem. Soc. 103, 1249–1251 (1981). 118. Tadashi, S., Masahiro, Y., Yumi, Y. & Hidemoto, K. Optical Resolution by Replacing Crystallization of DL-Threonine. Bull. Chem. Soc. Jpn. 63, 3296–3299 (1990). 119. Levilain, G. & Coquerel, G. Pitfalls and rewards of preferential crystallization. CrystEngComm 12, 1983–1992 (2010). 120. Secor, R. M. Resolution of Optical Isomers by Crystallization Procedures. Chem. Rev. 63, 297–309 (1963). 121. Coquerel, G., Tauvel, G. & Petit, M.-N. Method for splitting Omeprazole salts. WO2009027614 (2009). 122. Yamada, S., Yamada, S., Hongo, C. & Chibata, I. Preparation of D-p-Hydroxypheny ] glycine : Optical Resolution of DL-p- Hydroxyphenylglycine by Preferential Crystallization Procedure. Agric. Biol. Chem. 42, 1521–1526 (1978). 71 General Introduction

123. Polenske, D., Lorenz, H. & Seidel-morgenstern, A. Potential of Different Techniques of Preferential Crystallization for Enantioseparation of Racemic Compound Forming Systems. Chirality 21, 728–737 (2009). 124. Coquerel, G. & Levilain, G. Process for the resolution of enantiomers by preferential evaporative crystallization. US8907106B2 (2010). 125. Rougeot, C. & Hein, J. E. Application of Continuous Preferential Crystallization to Efficiently Access Enantiopure Chemicals. Org. Process Res. Dev 19, 1809–1819 (2015). 126. Ito, K., Akashi, T. & Tatsumi, S. Method of optically resolving racemic amino acids. US3260744A (1966). 127. Waagepetersen, H. S., Sonnewald, U. & Schousboe, A. Glutamine, Glutamate, and GABA: Metabolic Aspects. in Handbook of Neurochemistry and Molecular Neurobiology (eds. Lajtha, A., Oja, S. S., Schousboe, A. & Saransaari, P.) 19–20 (Springer, 2007). 128. Brands, K. M. J. & Davies, A. J. Crystallization-Induced Diastereomer Transformations. Chem. Rev. 106, 2711−2733 (2006). 129. Kozma, D. Basic concepts and nomenclature of stereochemistry. in CRC Handbook of Optical Resolutions via Diastereomeric Salt Formation 3–8 (CRC Press, 2002). 130. Whitesell, J. K. & Reynolds, D. Resolution of chiral alcohols with mandelic acid. J. Org. Chem. 43, 3548–3551 (1983). 131. Kozma, D. Alternative methods of resolution by diastereomeric salt formation. in CRC Handbook of Optical Resolutions via Diastereomeric Salt Formation 151–168 (CRC Press, 2002). 132. Kozma, D. Resolution by formation and fractional crystallization. in CRC Handbook of Optical Resolutions via Diastereomeric Salt Formation 9–50 (CRC Press, 2002). 133. Yee, T. T. & Cahill, J. A. Resolution of DL-diacetyllysine. US3651138 (1972). 134. Process for the resolution of DL-2-(6-methoxy)-2-naphthyl-propionic acid. GB1426186A (1976). 135. Marso, K. No Title. (Technical university of Budapest, 1988). 136. Sasaki, T., Kanematsu, K., Minamoto, K. & Fujimura, H. Researches on morphine- like analgesics. I. Syntheses and analgesic activity of desylamine derivatives. Chem. Pharm. Bull. (Tokyo) 12, 191–195 (1964). 137. McKenzie, A. & Walker, N. Optically active derivatives of phenylaminoacetic acid. J. Chem. Soc. 646–652 (1926). 138. Shiraiwa, T., Sado, Y., Fujii, S., Nakamura, M. & Kurosawa, H. Optical Resolution of (±)-Phenylsuccinic Acid by Using (−)-Proline as Resolving Agent. Bull. Chem. Soc. Jpn. 60, 824–826 (1987). 139. Shiraiwa, T., Morita, M., Iwafuji, K. & Kurokawa, H. Optical Resolution of (±)-α- Methylbenzylamine Salt of (±)-Malic Acid by Preferential Crystallization. Nippon Kagaku Kaishi 1983, 1743–1746 (1983). 140. Desguin, B. et al. Biosynthesis of the nickel-pincer nucleotide cofactor of lactate racemase requires a CTP-dependent cyclometallase. J. Biol. Chem. 293, 12303–12317 (2018). 141. Katagiri, H. & Kitahara, K. Racemiase, an enzyme which catalyses racemization of lactic acids. Biochem. J. 31, 909–914 (1937). 142. Koszelewski, D., Grischek, B., Glueck, S. M., Kroutil, W. & Faber, K. Enzymatic Racemization of Amines Catalyzed by Enantiocomplementary w-Transaminases. Chem. Eur. J. 17, 378–383 (2011). 143. Gallo, K. A., Tanner, M. E. & Knowles, J. R. Mechanism of the Reaction Catalyzed by Glutamate Racemase. Biochemistry 32, 3991–3997 (1993). 144. Solntsev, K. M. et al. Excited-State Proton Transfer in Chiral Environments: Photoracemization of BINOLs. Isr J Chem. 49, 227–233 (2009). 145. Cagle Jr., F. W. & Eyring, H. Application of Absolute Reaction Rate Theory to the 72 General Introduction

Racemization of Certain Sterically Hindered Compounds. J. Am. Chem. Soc. 73, 5628–5630 (1951). 146. Christian Wolf. Racemization, enantiomerization and diastereoisomerization. in Dynamic Stereochemistry of Chiral Compounds: Principles and Applications 29–34 (Royal Society of Chemistry, 2008). 147. Fuguitt, R. E. & Hawkins, J. E. Rate of the Thermal Isomerization of α-Pinene in the Liquid Phase. J. Am. Chem. Soc. 69, 319–322 (1947). 148. Chen, S., Wu, S. & Wang, K. Rapid racemization of optically active amino acids by oven‐based heating treatment. Chem. Biol. Drug Des. 33, 73–75 (1989). 149. Rougeot, C., Guillen, F., Plaquevent, J. C. & Coquerel, G. Ultrasound-Enhanced Deracemization: Toward the Existence of Agonist Effects in the Interpretation of Spontaneous Symmetry Breaking. Cryst. Growth Des. 15, 2151–2155 (2015). 150. Cai, Y. et al. Organocatalytic Aza-Michael / Retro-Aza-Michael Reaction : Pronounced Chirality Amplification in Aza-Michael Reaction and Racemization via Retro-Aza-Michael Reaction. Chirality 23, 397–403 (2011). 151. Mislow, K., Simmons, T., Melillo, J. T. & Ternay, A. L. The Hydrogen Chloride- Catalyzed Racemization of Sulfoxides. J. Am. Chem. Soc. 86, 1452–1453 (1964). 152. Koh, J. H., Jeong, H. M. & Park, J. Efficient Catalytic Racemization of Secondary Alcohols. Tetrahedron Lett. 39, 5545–5548 (1998). 153. Wuyts, S. et al. A heterogeneous Ru–hydroxyapatite catalyst for mild racemization of alcohols. J. Catal. 219, 417–424 (2003). 154. Ahn, Y., Ko, S., Kim, M. & Park, J. Racemization catalysts for the dynamic kinetic resolution of alcohols and amines. Coord. Chem. Rev. 252, 647–658 (2008). 155. Wuyts, S., Wahlen, J., Jacobs, P. A. & Vos, D. E. De. Heterogeneous vanadium catalysts for racemization and chemoenzymatic dynamic kinetic resolution of benzylic alcohols. Green Chem. 9, 1104–1108 (2007). 156. Schroeter, L. C. & Higuchi, T. A kinetic study of acid‐catalyzed racemization of epinephrine. J. Am. Pharm. Assoc. 47, 426–430 (1958). 157. Prasad, A. V. G. S., Venkateswara Rao, P. & Prasad, P. S. S. Economical, One-Pot Synthesis and Application of Schiff Base in Resolution-Racemization. Int. J. Pharm. Pharm. Sci. 2, 133–138 (2015). 158. Burwell Jr., R. L. & Brigg, W. S. Reaction of Alkanes with Hydrogen and Deuterium. J. Am. Chem. Soc. 74, 5096–5102 (1952). 159. Noyori, R. et al. Stereoselective Hydrogenation via Dynamic Kinetic Resolution. J. Am. Chem. Soc. 111, 9134–9135 (1989). 160. Ward, R. S. Dynamic Kinetic Resolution. Tetrahedron: Asymmetry 6, 1475–1490 (1995). 161. Nunami, K., Kubota, H. & Kubo, A. Dynamic kinetic resolution by stereospecific amination utilizing 2-oxoimidazolidine-4-carboxylate as a novel chiral auxiliary. Tetrahedron Lett. 35, 8639–8642 (1994). 162. Qamar, S., Galan, K., Elsner, M. P., Hussain, I. & Seidel- Morgenstern, A. Theoretical investigation of simultaneous continuous preferential crystallization in a coupled mode. Chem. Eng. Sci. 98, 25–39 (2013). 163. Sakamoto, M. et al. Breaking the Symmetry of Axially Chiral N -Aryl- 2(1H)- pyrimidinones by Spontaneous Crystallization. Angew. Chem. Int. Ed. 42, 4360–4363 (2003). 164. Yagishita, F. et al. Total Spontaneous Resolution by Deracemization of Isoindolinones. Angew. Chem. Int. Ed. 51, 13023–13025 (2012). 165. Gonella, G., Levilain, G. & Coquerel, S. Racemizable systems crystallizing as conglomerate and spontaneous symmetry breaking. J. Therm. Anal. Calorim. 103, 125–129 (2011). 166. Berova, M. & Zlatev, Z. Physiological response and yield of paclobutrazol treated 73 General Introduction

tomato plants (Lycopersicon esculentum Mill.). Plant Growth Regul. 30, 117–123 (2000). 167. Levilain, G., Rougeot, C., Guillen, F., Plaquevent, J. & Coquerel, G. Attrition- enhanced preferential crystallization combined with racemization leading to redissolution of the antipode nuclei. Tetrahedron: Asymmetry 20, 2769–2771 (2009). 168. Shieh, W.-C. & Carlson, J. A. Asymmetric Transformation of Either Enantiomer of Narwedine via Total Spontaneous Resolution Process, a Concise Solution to the Synthesis of (-)-Galanthamine. J. Org. Chem. 59, 5463–5465 (1994). 169. Köllges, T. & Vetter, T. Model-Based Analysis of Continuous Crystallization/Reaction Processes Separating Conglomerate Forming Enantiomers. Cryst. Growth Des. 17, 233–247 (2017). 170. Marco-Contelles, J., do Carmo Carreiras, M., Rodríguez, C., Villarroya, M. & García, A. G. Synthesis and Pharmacology of Galantamine. Chem. Rev. 2006, 106, 1, 116-133 106, 116–133 (2006). 171. Viedma, C. Experimental evidence of chiral symmetry breaking in crystallization from primary nucleation. J. Cryst. Growth 261, 118–121 (2004). 172. Van Der Meijden, M. W. et al. Attrition-Enhanced Deracemization in the Synthesis of Clopidogrel - A Practical Application of a New Discovery Abstract : Org. Process Res. Dev. 13, 1195–1198 (2009). 173. Viedma, C., Ortiz, J. E., de Torres, T., Izumi, T. & Blackmond, D. G. Evolution of Solid Phase Homochirality for a Proteinogenic Amino Acid. J. Am. Chem. Soc. 130, 15274–15275 (2008). 174. Kaptein, B. et al. Attrition-Enhanced Deracemization of an Amino Acid Derivative That Forms an Epitaxial Racemic Conglomerate. Angew. Chem. Int. Ed. 47, 7226– 7229 (2008). 175. Noorduin, W. L. et al. Emergence of a Single Solid Chiral State from a Nearly Racemic Amino Acid Derivative. J. Am. Chem. Soc. 130, 1158–1159 (2008). 176. Noorduin, W. L. et al. Fast Attrition-Enhanced Deracemization of Naproxen by a Gradual In Situ Feed. Angew. Chem. Int. Ed. 48, 4581–4583 (2009). 177. Noorduin, W. L. et al. The Driving Mechanism Behind Attrition-Enhanced Deracemization. Angew. Chem. Int. Ed. 49, 8435–8438 (2010). 178. Noorduin, W. L. et al. Complete Deracemization by Attrition-Enhanced Ostwald Ripening Elucidated. Angew. Chem. Int. Ed. 47, 6445–6447 (2008). 179. Iggland, M. & Mazzotti, M. Solid state deracemisation through growth, dissolution and solution-phase racemization. CrystEngComm 15, 2319–2328 (2013). 180. Iggland, M. & Mazzotti, M. Population Balance Modeling with Size-Dependent Solubility: Ostwald Ripening. Cryst. Growth Des. 12, 1489–1500 (2012). 181. Noorduin, W. L. et al. Explanation for the Emergence of a Single Chiral Solid State during Attrition-Enhanced Ostwald Ripening : Survival of the Fittest. Cryst. Growth Des. 8, 1675–1681 (2008). 182. Suwannasang, K., Flood, A. E. & Coquerel, G. A Novel Design Approach To Scale Up the Temperature Cycle Enhanced Deracemization Process : Coupled Mixed- Suspension Vessels. Cryst. Growth Des. 16, 6461–6467 (2016). 183. Belletti, G., Meekes, H., Rutjes, F. P. J. T. & Vlieg, E. Role of Additives during Deracemization Using Temperature Cycling. Cryst. Growth Des. 18, 6617–6620 (2018). 184. Baglai, I., Leeman, M., Kellogg, M. & Noorduin, W. L. Biomolecular Chemistry A Viedma ripening route to an enantiopure building block for Levetiracetam and. Org. Biomol. Chem. 17, 35–38 (2019). 185. Hassan, N. A., Bayer, E. & Jochims, J. C. Syntheses of optically active -amino nitriles by asymmetric transformation of the second kind using a principle of O . Dimroth. J. Chem. Soc., Perkin Trans. 1 3747–3757 (1998). 186. Leuchs, H. & Wutke, J. . Ber. Dtsch. Chem. Ges. 46, 2420 (1913). 74 General Introduction

187. Konoike, T. et al. Practical Enantioselective Synthesis of Endothelin Antagonist S- 1255 by Dynamic Resolution of 4-Methoxychromene-3-carboxylic Acid Intermediate. J. Org. Chem. 67, 7741–7749 (2002). 188. Mueller-Boetticher, H., Bressler, G.-R. & Kreye, P. Enantiomer Separation of Piperidone Derivatives with Simultaneous In-situ Racemization of the Unwanted Enantiomer. WO 2001096306 A1 (2004). 189. Shiraiwa, T. et al. Asymmetric Transformation of (RS)-1,2,3,4-Tetrahydro-3- isoquinolinecarboxylic Acid via Salt Formation with (1S)-10-Camphorsulfonic Acid. Bull. Chem. Soc. Jpn. 64, 3729 (1991). 190. Toda, F. & Tanaka, K. Conversion of racemic cyanohydrin into one optically active isomer in the presence of brucine. Chem. Lett. 12, 661–664 (1983). 191. Soloshonok, V. A. et al. General method for the synthesis of enantiomerically pure β- hydroxy-α-amino acids, containing fluorine atoms in the side chains. Case of stereochemical distinction between methyl and trifluoromethyl groups. X-Ray crystal and molecular structure of the nic. J. Chem. Soc., Perkin Trans. 1 24, 3143–3155 (1993). 192. Lopez, F. J., Ferrino, S. A., Reyes, M. S. & Roman, R. Asymmetric transformation of the second kind of racemic naproxen. Tetrahedron: Asymmetry 8, 2497–2500 (1997). 193. Clark, J. C., Phillipps, G. H. & Steer, M. R. J. A new asymmetric transformation of α- amino-acid esters with (+)-tartaric acid. Chem. Soc., Perkin Trans. 1 475 (1976). 194. Meltz, C. N., Post, R. J. & Rose, P. R. Process for Preparation of Pyrazolopyridine Tartrates. EP1449842A1 (2004). 195. Busch, F. R., Chiu, C. K.-F., Meltz, C. N., Post, R. J. & Rose, P. R. Process for Preparing Growth Hormone Secretagogues. EP1031575A1 (2000). 196. Brussee, J., Groenendijk, J. L. G., te Koppele, J. M. & Jansen, A. C. A. On the mechanism of the formation of s(−)-(1, 1’-binaphthalene)-2,2’-diol via copper(II)amine complexes. Tetrahedron 41, 3313–3319 (1985). 197. Kumar, A., Kumar, S. & Nanda, A. A Review about Regulatory Pharmaceutical Co- Crystals Status and Recent Patents of. Adv. Pharm. Bull. 8, 355–363 (2018). 198. Desiraju, G. R. Pharmaceutical Salts and Co-crystals : Retrospect and Prospects. in Pharmaceutical salts and co-crystals (eds. Wouters, J. & Quéré, L.) 1–8 (R, 2012). 199. Pfeiffer, P. Organische Molekulverbindungen. (1922). 200. Herbstein, F. H. Crystalline Molecular Complexes and Compounds. (Oxford University Press, 2005). 201. Bernstein, J. Introduction and historical background. in Polymorphism in Molecular Crystals 1–40 (Oxford University Press, 2020). 202. Mukherjee, A. & Desiraju, G. R. Synthon polymorphism and pseudopolymorphism in co-crystals. The 4,40-bipyridine–4-hydroxybenzoic acid structural landscape. Chem. Commun. 47, 4090–4092 (2011). 203. Purohit, R. & Venugopalan, P. Polymorphism : An Overview. Resonance 882–893 (2009). 204. Hilfiker, R., Blatter, F. & Von Raumer, M. Relevance of solid-state properties for pharmaceutical products. in Polymorphism: In the Pharmaceutical Industry (ed. Hilfiker, R.) 1–20 (Wiley-VCH, 2006). 205. Bond, A. D. Fundamental Aspects of Salts and Co-crystals. in Pharmaceutical salts and co-crystals (eds. Wouters, J. & Quéré, L.) 9–28 (Royal Society of Chemistry, 2012). 206. Aitipamula, S. et al. Polymorphs, salts, and cocrystals: What’s in a name? Cryst. Growth Des. 12, 2147–2152 (2012). 207. Bond, A. D. What is a co-crystal? CrystEngComm 9, 833–834 (2007). 208. Sekhon, B. S. Pharmaceutical co-crystals - a review. Ars Pharm 50, 99–117 (2009). 209. Braga, D. et al. Combining piracetam and lithium salts: ionic co-crystals and co-drugs? 75 General Introduction

Chem. Commun. 48, 8219–8221 (2012). 210. Madusanka, N., Eddleston, M. D., Arhangelskis, M. & Jones, W. Polymorphs, hydrates and solvates of a co-crystal of caffeine with anthranilic acid. Acta Crystallogr. Sect. B 70, 72–80 (2014). 211. Reddy, D. S., Craig, D. C. & Desiraju, G. R. Supramolecular Synthons in Crystal Engineering . 4 . Structure Simplification and Synthon Interchangeability in Some Organic Diamondoid Solids. J. Am. Chem. Soc. 118, 4090–4093 (1996). 212. Vishweshwar, P., Mcmahon, J. A., Bis, J. A. & Zaworotko, M. J. Pharmaceutical Co- Crystals. J. Pharm. Sci. 95, 499–516 (2006). 213. Vangala, V. R., Chow, P. S. & Tan, R. B. H. Co-Crystals and Co-Crystal Hydrates of the Antibiotic Nitrofurantoin: Structural Studies and Physicochemical Properties. Cryst. Growth Des. 12, 5925–5938 (2012). 214. Bhatt, P. M., Azim, Y., Thakur, T. S. & Desiraju, G. R. Co-Crystals of the Anti-HIV Drugs Lamivudine and Zidovudine. Cryst. Growth Des. 9, 951–957 (2009). 215. Demeulemeester, J., De Maeyer, M. & Debyser, Z. Therapy of Viral Infections - HIV- 1 integrase drug discovery comes of age. Top. Med. Chem. 15, 1–52 (2015). 216. Shemchuk, O. et al. Ionic Co-Crystal Formation as a Path Towards Chiral Resolution in the Solid State. Chem. Eur. J. 24, 1–11 (2018). 217. Eccles, K. S., Deasy, R. E., Fábián, L., Maguire, A. R. & Lawrence, S. E. The Use of Co-crystals for the Determination of Absolute Stereochemistry: An Alternative to Salt Formation. J. Org. Chem. 76, 1159–1162 (2011). 218. Harmsen, B. & Leyssens, T. Enabling Enantiopurity: Combining Racemization and Dual-Drug Co-crystal Resolution. Cryst. Growth Des. 18, 3654−3660 (2018). 219. Springuel, G. & Leyssens, T. Innovative Chiral Resolution Using Enantiospecific Co- Crystallization in Solution. Cryst. Growth Des. 12, 3374−3378 (2012). 220. He, L., Chen, X., Li, X., Zhou, Z. & Ren, Z. Chiral co-selector induced chirality switching in the enantioseparation of by forming co-crystal. New J. Chem. (2019) doi:10.1039/C9NJ02921D. 221. Sánchez-Guadarrama, O. et al. Chiral Resolution of RS-Praziquantel via Diastereomeric Co-Crystal Pair Formation with l-Malic Acid. Cryst. Growth Des. 16, 307–314 (2016). 222. Ratajczak-sitarz, M., Katrusiak, A., Gawronska, K. & Gawronski, J. Racemate resolution via diastereomeric helicates in hydrogen-bonded co-crystals : the case of BINOL – diamine complexes. Tetrahedron: Asymmetry 18, 765–773 (2007). 76 Goal of this work

Goal of this work This work was carried out in the laboratories of Professor Tom Leyssens and Professor Olivier Riant, thus combining the expertise in crystal engineering and process development with the expertise in . Part of this work was also carried out in the laboratory of Professor Joop ter Hoost as part of a scientific exchange, taking advantage of his expertise in online process analysis.

This thesis had as principal objective the development of an original and innovative deracemization method based on the co-crystallization resolution process previously developed in the laboratory of Professor Leyssens. The idea was to start from a new molecule based on a model, for which a suitable co-crystal had to be found and to develop the process on this co-crystal. Once the process proven, its development and optimization were the secondary objectives. For this, two major studies were conducted, one on the process itself and one on the racemization reaction. Finally, a final objective was to pave the way into development of the process on other analog systems and to valorize the compounds under study.

Starting from scratch, this work began with the synthesis of several analogs, based on the racemic precursor of Paclobutrazol, which is known for its easy racemization in solution. The analogs were designed by changing the structure of the precursor in three chosen zones. The carbonyl part with the α-chiral-carbon possessing one hydrogen that renders the molecule racemizable in solution was kept unchanged for all the synthetized analogs.

Following the synthesis and purification of the analogs, each of them was put through a co-crystal screening where each analog was ground co-jointly with a co-former from a library of chiral compounds. Grinding of those mixtures followed by a Powder X- ray Diffraction analysis allowed highlighting possible hits. Confirmation of these hits was carried out by growing single crystals and their analysis by Single-crystal X-ray Diffraction. The results of this study allowed for identification of a suitable co-former while providing data on possible other systems to develop the process on. Those data also allowed for comparison between analogs and identification of possible hit structures and effects of specific synthons on the capacity of co-crystallization of an analog molecule.

With the system in hand, both resolution and racemization were studied in parallel in order to verify the viability of the system. Second, the aim was to identify a suitable solvent to carry out the resolution and racemization, to eventually combine both in a deracemization process. Regarding racemization, a suitable racemizing agent was identified to induce racemization of the molecule in the solvent used for resolution. 77 Goal of this work

Kinetics of the racemization reaction with and without the co-former were studied. This study allowed for better understanding of the kinetic model behind the racemization of ketones though their keto-enol equilibrium. Furthermore, this study produced important data for the optimization of the process e.g. whether alternative solvents were worth trying and how the co-former influences the kinetics of racemization.

Then, both processes were combined to yield the deracemization process. As combining such processes required compromise between the resolution and the racemization part, the deracemization process was successfully carried out in a two- pot continuous set-up. Once the concept proven, the deracemization process was optimized by changing various parameters including concentration of racemizing agent, temperature of cooling, solvent, presence of water,… to eventually yield a robust and efficient process with optimized productivity.

Finally, the deracemized molecule was valorized by the study of the diastereoselective reduction of its ketone group to yield an analog of Paclobutrazol. Diastereoselectivity of the reaction was studied with correlation of the favored diastereomer to the mechanism of the reaction and the stability of each diastereomer.

The following chapters will detail the completion of each objective starting with the investigation of the process, detailing its development and optimization while highlighting its contribution to the chiral toolbox. The first chapter will display the search for a suitable co-crystal through synthesis of target molecules and screening with possible co-formers, revealing at the end a suitable co-crystal. The following chapter will show the development of the co-crystallization induced spontaneous deracemization process on the suitable system from the parallel studies of resolution and racemization to their combination into the deracemization process. Then, the third chapter will detail the optimization of the developed system to yield an efficient deracemization process by kinetic study of the racemization reaction and further study of the process. Finally, deracemized molecule was valorized into a possible interesting fungicide or growth retardant. 78

Chapter 1- Identifying a suitable system 79 Chapter 1- Identifying a suitable system

1. Overview To develop a new deracemization method, using co-crystallization, the first step was to identify a suitable system. This system should be composed of a target compound, able to racemize in solution, and a chiral co-former, co-crystallizing with the target compound in an enantiospecific manner, or through formation of a diastereomeric pair. This chapter is divided in two parts, a first one treating with the synthesis of 10 potential target compounds and a second focusing on the co-crystal screening carried out on these using 35 potential co-formers.

2. Target compounds: from design to synthesis 2.1. Introduction When one wishes to develop a new methodology meant to be applicable to a series of molecules, one has to carefully select a first model compound prior to expanding to a larger panel. Recently, this approach was applied in the case of Viedma Ripening that was first conceptually proven for the case of sodium chlorate.1 Then, it was extended to a series of similar organic molecules, and finally applied to a broader scope of compounds, now even being used on organic salts.2,3 The general key features the model compound should possess are the following:

Synthons allowing for co-crystallization A chiral feature that is racemizable under certain conditions

The first feature is a requirement for the resolution while the second is a requirement for the resolution as well as the racemization. 4,4-dimethyl-1-(4-chlorophenyl)-2- (1H-1,2,4-triazol-1-yl)-Pentan-3-one (figure 1-1) fits both criteria and was recently shown to easily racemize in solution under basic conditions.4

Figure 1-1 Skeletal formula of 4,4-dimethyl-1-(4-chlorophenyl)-2-(1H-1,2,4-triazol-1-yl)- Pentan-3-one, with in red the synthons for hydrogen bonding and halogen bonding and in blue for π-π interactions. 80 Chapter 1- Identifying a suitable system

In addition to being a suitable model compound, this compound is also a precursor of Paclobutrazol that has recently been used in a Viedma ripening scheme (cf. General introduction section 2.4.3 figure 57). Finally, this compound is easily synthesized through relatively simple and efficient reactions. Indeed, it can be synthetized in two steps (scheme 1-1). The first reaction is a SN2 of 1-H-1,2,4-triazol on chloropinacolone. The second reaction is a SN2 of the product enolate on p- chlorobenzyl chloride.

Scheme 1-1 Reaction scheme for the synthesis of Paclobutrazol’s precursor.5,6

Figure 1-2 Skeletal structure of all analogues synthetized with in red the model compound. The grey arrows display lines in which each compound, situated on the line, only differs from one group change with the other compounds on the same line.

Except for the necessity to work under anhydrous conditions for the second reaction, this set of reactions does not require any further specific precautions, and allows 81 Chapter 1- Identifying a suitable system producing high yields with gram-scaled reactions.5,7,6 Further, this reaction allows easy access to various analogues, similar to drug design where one wishes to find a potent API starting from a model structure.8–10 Starting from the reaction scheme shown above, nine analogues were synthetized using equivalent reactions complemented with some other reactions if needed and paying attention to keeping the synthesis scheme shorter than four reactions (figure 1-2).

The analogues were designed in such a way that each analogue can be linked to another analogue changing only one . This allows direct comparison when processing the results of the co-crystal screening.

2.2. Material & methods Column chromatography was performed using silica gel 60 Å (40-63 μm). Chloropinacolone, 1,2,4-triazol, anhydrous DMF, benzyl chloride, p-fluorobenzyl chloride, p-chlorobenzylchloride, NaH, pinacolone, DL-phenylalanine, chloroacetamide, sodium nitrite, benzyl alcohol, allyl alcohol, 2,2-dimethyloxirane, thionyl chloride, N-benzyl-N,N-diethylethanaminium and bromoacetophenone were purchased from commercial sources and used as received.

NMR spectra were obtained on a 300 MHz spectrometer. Chemical shifts are reported in parts per million (ppm) and were normalized regarding the chemical shift of the peak of the deuterated solvent used. For the 1H NMR spectra, the value of the different solvent used are the following: CDCl3 7.26ppm; CD3OD 3.31ppm; D2O 4.79ppm; 13 (CD3)2SO 2.50ppm. For the C NMR spectra, the value of the different solvent used are the following: CDCl3 77.36ppm; CD3OD 49.00ppm; (CD3)2SO 39.52ppm. Multiplicities are abbreviated as follows: singlet (s), doublet (d), triplet (t), quartet (q) and multiplet (m). For the 19F NMR, the peak of the probe was automatically removed from the spectra.

Single Crystal-X-ray Diffraction (SC-XRD) was performed on a Gemini Ultra R system (4-circle kappa platform, Ruby CCD detector) using Cu Kα radiation (λ = 1.54056 Å)/ or on a MAR345 detector using monochromated Mo Kα radiation (λ = 0.71073 Å) (Xenocs Fox3D mirror) produced by a Rigaku UltraX 18 generator. The structures were solved by dual-space direct methods with SHELXT and then refined on |F2| using SHELXL2014 or SHELXL2018. Non-hydrogen atoms were anisotropically refined and the hydrogen atoms (not implicated in H-bonds) in the riding mode with isotropic temperature factors fixed at 1.2 times U(eq) of the parent atoms (1.5 times for methyl groups). Hydrogen atoms implicated in H-bonds were localized in the Fourier difference maps (ΔF). Data collection and refinement details are shown in Appendix A.

Differential Scanning Calorimetry measurements were performed on either a Mettler Toledo DSC821e or TA instrument DSC2500. Each sample was put in an aluminum pan (roughly 5-10mg of sample added) pierced with a hole and heated over time at a 82 Chapter 1- Identifying a suitable system

rate of 10°C per minute. Melting points were determined by measuring the onset of the peak of fusion. DSC thermograms are given in Appendix A (figures A-1 to A-16).

2.2.1. Model compound: BnClTP 3,3-dimethyl-1-(1H-1,2,4-triazol-1-yl)-butan-2-one (TP). 5.13g of 1,2,4-triazol, 11g of chloropinacolone (CP)

(1.1eq) and 10.27g of K2CO3 (1.5 eq) were added to 100mL of acetonitrile. The mixture was stirred under reflux for 8h30.6 The mixture was filtered and the solvent was removed under vacuum. Then, water was added and the residue was extracted with ethyl acetate, 3 times. The organic layers were combined, dried over MgSO4, filtered and concentrated under vacuum. The crude was purified adding 10mL of diethyl ether and the mixture was stirred at room temperature for 5-6h in a closed round-bottom flask. Then, the mixture was filtered and a white solid obtained. The solid was dried at room temperature for 8h then weighed. 10.404g (84%) were 1 obtained. mp = 67.7°C, Rf= 0.1 (1:1 EA/PE) Vanillin revelation (white spot), H NMR

(300 MHz, CDCl3) ẟ (ppm): 1.24 (s, 9H, C(CH3)3), 5.17 (s, 2H, N-CH2-C=O), 7.93 13 (s, 1H, N=CH-N) 8.12 (s, 1H, N=CH-N). C NMR (75 MHz, CDCl3) ẟ (ppm): 26.23

(C(CH3)3), 43.68 (C(CH3)3), 53.32 (N-CH2-C=O), 144.86 & 151.90 (N=CH-N), 206.31 (C=O).

4,4-dimethyl-1-(4-chlorophenyl)-2-(1H-1,2,4- triazol-1-yl)-pentan-3-one (BnClTP).5 1g of TP dissolved in 3mL of DMF are added dropwise at 0°C to a stirred suspension of 0.240g of NaH (1eq) in 3mL of DMF. Once all the TP added, 966mg of chlorobenzyl chloride (2eq) were added at 0°C. Then, the mixture was warmed up to room temperature and stirred for 3h30. Water was added to quench the reaction. Then, ethyl acetate was added to perform a liquid-liquid extraction (3x). The organic layers were combined and dried over MgSO4, filtered and concentrated under strong vacuum to remove the remaining DMF. The crude was purified by crystallization in hexane for at least 24h. 1.381g (79%) of a white solid were obtained. mp= 122.0°C, Rf= 0.45 (1:1 EA/PE) UV revelation + Vanillin (whitish 1 spot). H NMR (300 MHz, CDCl3) ẟ (ppm): 1.02 (s, 9H, C(CH3)3), 3.21 (dd, 1H, J =

13.7, 7.8 Hz, CH-CHaHb- CAr), 3.33 (dd, 1H, J = 13.7, 7.6 Hz, CH-CHaHb- Car), 5.67

(t, 1H, J = 7.7 Hz, CH2-CH-C=O), 6.99 (m, 2H, CAr=CArH-CArH), 7.23 (m, 2H, 13 CAr=CArH-CArCl), 7.87 (s, 1H, N=CH-N), 8.21 (s, 1H, N=CH-N). C NMR (75 MHz,

CDCl3) ẟ (ppm): 25.59 (C(CH3)3), 38.67 (CH-CH2-CAr) 44.95 (C(CH3)3), 62.60

(CH2-CH-C=O), 129.10 & 130.58 (CArH), 133.52 (CAr), 133.70 (CArCl), 142.36 & 151.42 (N=CH-N), 208.98 (C=O). 83 Chapter 1- Identifying a suitable system

By-products BnClTP 4,4-dimethyl-1-(4-chlorophenyl)-2-(4- chlorobenzyl)-2-(1H-1,2,4-triazol-1-yl)-pentan-3- 1 one ((BnCl)2TP). H NMR (300 MHz, CDCl3) ẟ

(ppm): 0.95 (s, 9H, C(CH3)3), 3.45 (d, 2H, J = 15.0

Hz, 2 x CH-CHaHb-CAr), 3.58 (d, 2H, J = 14.9 Hz, 2 x

CH-CHaHb-CAr), 6.68 (m, 4H, CAr=CArH-CArH), 7.20

(m, 4H, CAr=CArH-CArCl), 7.93 (s, 1H, N=CH-N), 13 8.12 (s, 1H, N=CH-N). C NMR (75 MHz, CDCl3) ẟ

(ppm): 28.68 (C(CH3)3), 39.98 (C-CH2-CAr) 46.25

(C(CH3)3), 74.65 ((CH2)2C-C=O), 128.92 & 131.28 (CArH), 132.58 (CAr), 133.66 (CArCl), 142.23 & 151.92 (N=CH-N), 210.22 (C=O).

(Z)-1-(3-((4-chlorobenzyl)oxy)-1-(4-chlorophenyl)- 4,4-dimethylpent-2-en-2-yl)-1H-1,2,4-triazole (Z- 1 BnClTPBnCl). H NMR (300 MHz, CDCl3) ẟ (ppm):

1.38 (s, 9H, C(CH3)3), 4.10 (s, 2H, C-CH2-CAr), 4.12 (s,

2H, O-CH2-CAr), 6.97-7.01 (m, 4H, CArH), 7.15-7.20 (m,

2H, CArH), 7.23-7.27(m, 2H, CArH), 7.94 (s, 1H, N=CH- 13 N), 8.02 (s, 1H, N=CH-N). C NMR (75 MHz, CDCl3) ẟ

(ppm): 30.20 (C(CH3)3), 35.64 (C-CH2-CAr) 39.00

(C(CH3)3), 74.69 (O-CH2-CAr), 123.18 (CH2-C=C-O), 128.78 & 128.89 (CArH), 128.93 (CArCl), 129.34 & 129.46 (CArH), 131.28 (CArCl), 134.79 (CAr), 135.35 (CAr), 146.17 & 151.54 (N=CH- N), 161.00 (=C-O).

2.2.2. BnFTP analog 4,4-dimethyl-1-(4-fluorophenyl)-2-(1H-1,2,4- triazol-1-yl)-pentan-3-one (BnFTP). 2g of TP, dissolved in 10mL of anhydrous DMF, were added dropwise at 0°C to a stirred suspension of 0.480g of NaH (60% dispersion in mineral oil, 1eq) in 5mL of anhydrous DMF. Once, all the TP was added, 2826µL of p-fluorobenzyl chloride (2eq) were added at 0°C. Then, the mixture was warmed up to room temperature and let to react for 3h30. Water was added to quench the reaction. Then, ethyl acetate was added to perform a liquid-liquid extraction (3x). The organic layers were combined and dried over MgSO4, filtered and concentrated under a strong vaccum to remove the remaining DMF. The crude was purified by column chromatography (Packing: PE pure, Eluent: 20:80 (EA:PE) until the compound starts coming out, then 50:50). 2.347g (71%) of a white solid was obtained after removal of 1 solvent. mp = 41°C, Rf= 0.58 (1:1 EA/PE) UV revelation + Vanillin (whitish spot), H 84 Chapter 1- Identifying a suitable system

NMR (300 MHz, CDCl3) ẟ (ppm): 1.01 (s, 9H, C(CH3)3), 3.20 (dd, 1H, J = 13.7, 7.7

Hz, CH-CHaHb- CAr), 3.34 (dd, 1H, J = 13.7, 7.7 Hz, CH-CHaHb- Car), 5.68 (dd, 1H,

J = 7.7, 7.7 Hz, CH2-CH-C=O), 6.95 (m, 2H, CAr=CArH-CArF), 7.03 (m, 2H, 13 CAr=CArH-CArH), 7.88 (s, 1H, N=CH-N), 8.26 (s, 1H, N=CH-N). C NMR (75 MHz,

CDCl3) ẟ (ppm): 25.57 (C(CH3)3), 38.62 (CH-CH2-CAr) 44.99 (C(CH3)3), 62.84 (CH2- CH-C=O), 115.75-116.04 (d, J = 21.44 Hz, CArH), 130.81-130.92 (d, J = 8.22 Hz, CArH), 130.92-130.97 (d, J = 3.79 Hz, CAr), 142.32 & 151.33 (N=CH-N), 160.58- 19 163.85 (d, J = 246.46 Hz, CArF), 209.17 (C=O). F NMR (282 MHz, CDCl3) ẟ (ppm): -114.58.

By-products BnFTP 4,4-dimethyl-1-(4-fluorophenyl)-2-(4-fluorobenzyl)- 2-(1H-1,2,4-triazol-1-yl)-pentan-3-one ((BnF)2TP). 1 H NMR (300 MHz, CDCl3) ẟ (ppm): 0.93 (s, 9H,

C(CH3)3), 3.45 (d, 2H, J = 14.9 Hz, 2 x CH-CHaHb-

CAr), 3.58 (d, 2H, J = 14.9 Hz, 2 x CH-CHaHb-CAr), 6.73

(m, 4H, CAr=CArH-CArF), 6.91 (m, 4H, CAr=CArH-

CArH), 7.92 (s, 1H, N=CH-N), 8.10 (s, 1H, N=CH-N). 13 C NMR (75 MHz, CDCl3) ẟ (ppm): 28.58 (C(CH3)3),

38.85 (C-CH2-CAr) 46.15 (C(CH3)3), 74.92 ((CH2-)2C- C=O), 115.45-115.74 (d, J = 21.21 Hz, CArH), 129.80- 129.85 (d, J = 3.37 Hz, CAr), 131.48-131.58 (d, J = 7.92 Hz, CArH), 142.22 & 151.76 (N=CH-N), 160.48- 19 163.76 (d, J = 246.98 Hz, CArF), 210.34 (C=O). F NMR (282 MHz, CDCl3) ẟ (ppm): -114.58.

(Z)-1-(3-((4-fluorobenzyl)oxy)-1-(4- fluorophenyl)-4,4-dimethylpent-2-en-2-yl)-1H- 1,2,4-triazole (Z-BnFTPBnF). 1H NMR (300

MHz, CDCl3) ẟ (ppm): 1.39 (s, 9H, C(CH3)3), 4.11

(s, 2H, C-CH2-CAr), 4.12 (s, 2H, O-CH2-CAr), 6.97

(m, 8H, CArH), 7.94 (s, 1H, N=CH-N), 8.00 (s, 1H, 13 N=CH-N). C NMR (75 MHz, CDCl3) ẟ (ppm):

30.23 (C(CH3)3), 35.49 (C-CH2-CAr) 38.93

(C(CH3)3), 74.79 (O-CH2-CAr), 115.36-115.64 (d, J = 21.48 Hz, CArH), 115.44-115.73 (d, J = 21.48 Hz,

CArH), 123.56 (CH2-C=C-O), 129.55-129.65 (d, J = 7.95 Hz, CAr), 129.93-130.04 (d, J = 8.29 Hz, CArH), 132.19-132.23 (d, J = 3.24 Hz, CAr), 132.46- 132.50 (d, J = 3.18 Hz, CAr), 146.21 & 151.50 (N=CH-N), 160.11-163.36 (d, J = 85 Chapter 1- Identifying a suitable system

19 245.31 Hz, CArF), 160.79 (=C-O), 161.07-164.34 (d, J = 246.65 Hz, CArF). F NMR

(282 MHz, CDCl3) ẟ (ppm): -116.06, -113.64.

(E)-1-(3-((4-fluorobenzyl)oxy)-1-(4- fluorophenyl)-4,4-dimethylpent-2-en-2-yl)- 1H-1,2,4-triazole (E-BnFTPBnF). 1H

NMR (300 MHz, CDCl3) ẟ (ppm): 0.93 (s, 9H,

C(CH3)3), 3.69 (s, 2H, C-CH2-CAr), 4.91 (s, 2H,

O-CH2-CAr), 6.99 (m, 4H, CArH), 7.11 (m, 2H,

CArH), 7.34 (m, 2H, CArH), 7.35 (s, 1H, N=CH- N), 7.88 (s, 1H, N=CH-N).

2.2.3. BnTP analog 4,4-dimethyl-1-phenyl-2-(1H-1,2,4-triazol-1-yl)- pentan-3-one (BnTP). 2g of TP dissolved in 10mL of DMF are added dropwise at 0°C to a stirred suspension of 0.480g of NaH (1eq) in 5mL of DMF. Once all the TP added, 2760µL of benzyl chloride (2eq) were added at 0°C. Then, the mixture was warmed up to room temperature and stirred for 3h30. Water was added to quench the reaction. Then, ethyl acetate was added to perform a liquid-liquid extraction (3x). The organic layers were combined and dried over MgSO4, filtered and concentrated under a strong vacuum to remove the remaining DMF. The crude was purified by column chromatography (Packing: PE pure, Eluent: 20:80 (EA:PE) until the compound starts coming out, then 50:50).

1.016g (66%) of a white solid was obtained. mp= 70.4°C, Rf= 0.63 (1:1 EA/PE) UV 1 revelation + Vanillin (whitish spot). H NMR (300 MHz, CDCl3) ẟ (ppm): 0.98 (s,

9H, C(CH3)3), 3.21 (dd, 1H, J = 13.6, 7.7 Hz, CH-CHaHb- CAr), 3.35 (dd, 1H, J = 13.6,

7.7 Hz, CH-CHaHb- CAr), 5.70 (t, 1H, J = 7.7 Hz, CH2-CH-C=O), 7.05 (m, 2H,

CAr=CArH-CArH), 7.23 (m, 3H, CArH=CArH-CArH), 7.85 (s, 1H, N=CH-N), 8.21 (s, 13 1H, N=CH-N). C NMR (75 MHz, CDCl3) ẟ (ppm): 25.61 (C(CH3)3), 39.47 (CH- p CH2-CAr) 45.00 (C(CH3)3), 62.92 (CH2-CH-C=O), 127.61 (CAr H), 128.99 & 129.30 (CArH), 135.23 (CAr), 142.41 & 151.38 (N=CH-N), 209.43 (C=O). 86 Chapter 1- Identifying a suitable system

By-products BnTP 4,4-dimethyl-1-phenyl-2-benzyl-2-(1H-1,2,4- 1 triazol-1-yl)-pentan-3-one (Bn2TP). H NMR (300

MHz, CDCl3) ẟ (ppm): 0.93 (s, 9H, C(CH3)3), 3.53 (d,

2H, J = 14.8 Hz, 2 x CH-CHaHb-CAr), 3.64 (d, 2H, J =

14.8 Hz, 2 x CH-CHaHb-CAr), 6.80 (m, 4H, CAr=CArH-

CArH), 7.22 (m, 6H, CArH=CArH-CArH), 7.90 (s, 1H, N=CH-N), 8.11 (s, 1H, N=CH-N). 13C NMR (75 MHz,

CDCl3) ẟ (ppm): 29.28 (C(CH3)3), 39.77 (C-CH2-CAr)

46.26 (C(CH3)3), 75.09 ((CH2-)2C-C=O), 127.45 p (CAr H), 128.63 & 130.01 (CArH), 134.33 (CAr), 142.22 & 151.64 (N=CH-N), 210.60 (C=O).

(Z)-1-(3-(benzyloxy)-4,4-dimethyl-1-phenylpent-2-en-2- yl)-1H-1,2,4-triazole (Z-BnTPBn). 1H NMR (300 MHz,

CDCl3) ẟ (ppm): 1.40 (s, 9H, C(CH3)3), 4.15 (s, 2H, O-CH2-

CAr), 4.17 (s, 2H, C-CH2-CAr), 7.07 (m, 4H, CAr=CArH-CArH),

7.23 (m, 6H, CArH=CArH-CArH), 7.93 (1H, N=CH-N), 8.03 13 (s, 1H, N=CH-N). C NMR (75 MHz, CDCl3) ẟ (ppm):

30.24 (C(CH3)3), 36.27 (C-CH2-CAr) 37.50 (C(CH3)3), p 75.52 (O-CH2-CAr), 123.52 (CH2-C=C-O), 126,79 (CAr H), p 128.11 & 128.16 (CArH), 128.29 (CAr H), 128.57 & 128.70 (CArH), 136.46 & 136.96 (CAr), 146.29 & 151.41 (N=CH- N), 160.91 (=C-O).

(E)-1-(3-(benzyloxy)-4,4-dimethyl-1- phenylpent-2-en-2-yl)-1H-1,2,4-triazole (E- 1 BnTPBn). H NMR (300 MHz, CDCl3) ẟ (ppm):

1.02 (s, 9H, C(CH3)3), 3.81 (s, 2H, C-CH2-CAr),

5.04 (s, 2H, O-CH2-CAr), 6.95 (m, 2H,

CAr=CArH-CArH), 7.19-7.23 (m, 3H,

CArH=CArH-CArH), 7.36 (1H, N=CH-N), 7.40

(m, 5H, CArH=CArH-CArH), 8.01 (s, 1H, N=CH- 13 N). C NMR (75 MHz, CDCl3) ẟ (ppm): 28.57

(C(CH3)3), 38.24 (C-CH2-CAr) 38.43 (C(CH3)3), 76.24 (O-CH2-CAr), 125.83 (CH2- p p C=C-O), 127.03 (CAr H), 127.15 (CArH), 128.19 (CAr H), 128.74 & 128.79 (CArH), p 128.92 (CAr H), 136.99 & 137.11 (CAr), 145.64 & 151.66 (N=CH-N), 163.93 (=C- O). 87 Chapter 1- Identifying a suitable system

2.2.4. PgTP analog 2,2-dimethyl-4-(1H-1,2,4-triazol-1-yl)hept-6-yn-3- one (PgTP).5 2g of TP dissolved in 5.5mL of DMF was added dropwise at 0°C to a stirred suspension of 0.48g of NaH (1eq) in 1.5mL of DMF. Once, all the TP added, 1.5mL (1.1eq) of a 80% propargyl bromide solution in toluene were added at 0°C. Then, the mixture was warmed up to room temperature and left to react for 3h30. Water was added to quench the reaction followed by HCl 1M. Then, ethyl acetate was added to perform a liquid-liquid extraction (3x). The organic layers were combined and dried over MgSO4, filtered and concentrated under a strong vacuum to remove the remaining DMF. The mixture was purified by crystallization, in toluene/hexane (1:1). 2.16g (88%) of a white solid was obtained after solvent 1 removal. mp= 85.7°C. Rf= 0.29 (1:1 EA/PE) p-anisaldehyde (light orange spot). H

NMR (300 MHz, CDCl3) ẟ (ppm): 1.16 (s, 9H, C(CH3)3), 2.04 (td, 1H, J = 2.7, 0.6 Hz, C≡CH), 2.90 (ddd, 1H, J = 2.7, 0.6, 7.3 Hz, CH-CH2-C≡CH), 5.65 (t, 1H, J = 7.3 13 Hz, CH2-CH-C=O), 7.92 (s, 1H, N=CH-N), 8.27 (s, 1H, N=CH-N). C NMR (75

MHz, CDCl3) ẟ (ppm): 23.03 (CH-CH2), 26.00 (C(CH3)3), 44.84 (C(CH3)3), 60.07

(C≡CH), 72.45 (CH-CH2), 77.99 (C≡CH), 142.73 & 151.80 (N=CH-N), 207.51 (C=O).

By-product of PgTP 2,2-dimethyl-4-(prop-2-yn-1-yl)-4-(1H-1,2,4-triazol-1- 1 yl)hept-6-yn-3-one (Pg2TP). H NMR (300 MHz, CDCl3) ẟ

(ppm): 1.00 (s, 9H, C(CH3)3), 2.07 (dd, 2H, J = 2.6 Hz, 2 x

CH2-C≡CH), 3.31 (t, 4H, J = 2.2 Hz, 2 x CH2-C≡CH), 8.00 (s, 1H, N=CH-N), 8.26 (s, 1H, N=CH-N).

2.2.5. BnBnOP analog 1-(benzyloxy)-3,3-dimethylbutan-2-one (BnOP). 1.16mL (1.5eq, 11.1mmol) of benzylic alcohol is added dropwise at 0°C to a stirred suspension of 325.6mg of NaH (1.1eq) in 8mL of THF. Once all the alcohol added, 1g of CP (0.97mL) is added and the reaction is warmed up to RT and left to react for 8h. Water was added to quench the reaction and the mixture was extracted with Et2O. The organic phase was dried over

MgSO4, filtered and concentrated under vacuum. The crude was purified by column chromatography (Packing: PE, First eluent: PE, Second eluent: 5:95 (Et2O/PE).

955mg (63%) of a yellowish liquid was obtained after removal of the solvent. Rf= 0.9 88 Chapter 1- Identifying a suitable system

1 (1:1 Et2O/PE) UV revelation + p-anisaldehyde (blue spot). H NMR (300 MHz,

CDCl3) ẟ (ppm): 1.11 (s, 9H, C(CH3)3), 4.29 (s, 2H, O-CH2-CAr), 4.55 (s, 2H, O-CH2- C=O), 7.30 (m, 5H, CHAr).

2-(benzyloxy)-4,4-dimethyl-1- phenylpentan-3-one (BnBnOP). 1g of BnOP (4.8mmol) dissolved in 5mL of anhydrous THF is added dropwise at 0°C to a stirred suspension of 0.192g of NaH (1eq) in 5mL of anhydrous THF. After 30min of reaction at 0°C, 1mL of benzyl bromide (2eq) was added at 0°C followed by 36mg of NaI (5%). Then, the mixture was heated to reflux overnight. Water was added to quench the reaction. Then, Et2O was added to perform a liquid-liquid extraction (3x). The organic layers were combined and dried over

MgSO4, filtered and concentrated under vacuum. The crude was purified by column chromatography (Packing: PE pure, first eluent: 10:90 (Toluene/PE), second eluent:

25:75, Third eluent: 1:1, Fourth eluent: 10:90 (Et2O/PE)). 383mg (27%) of a yellow liquid was obtained after removal of the solvent. Rf= 0.72 (1:1 EA/PE) UV revelation 1 + p-anisaldehyde (black pot). mp= 40.0°C. H NMR (300 MHz, CDCl3) ẟ (ppm): 1.01

(s, 9H, C(CH3)3), 2.83 (dd, 1H, J = 13.8, 7.9 Hz, CH-CHaHb- CAr), 2.93 (dd, 1H, J =

13.8, 4.8 Hz, CH-CHaHb- CAr), 4.21 (d, 1H, J = 11.8 Hz, O-CHaHb-CAr), 4.38 (d, 1H,

J = 11.8 Hz, O-CHaHb-CAr), 4.4 (dd, 1H, J = 4.9, 7.9 Hz, CH2-CH-C=O), 7.02-7.07 13 (m, 2H, CAr=CArH-CArH), 7.12-7.21 (m, 8H, CArH=CArH-CArH). C NMR (75 MHz,

CDCl3) ẟ (ppm): 26.16 (C(CH3)3), 38.32 (CH-CH2-CAr) 43.70 (C(CH3)3), 71.49 (O- p p o CH-CAr), 79.63 (O-CH-C=O), 126.58 (CAr H), 127.64 (CAr H),127.69 (CAr H), 128.26, 128.35 & 129.61 (CArH), 137.57 & 137.61 (CAr), 213.57 (C=O).

By-product of BnBnOP 2-benzyl-2-(benzyloxy)-4,4-dimethyl-1- 1 phenylpentan-3-one (Bn2BnOP). H NMR (300

MHz, CDCl3) ẟ (ppm): 0.52 (s, 9H, C(CH3)3),

3.28 (d, 2H, J = 14.00 Hz, 2 x C-CHaHb-CAr),

3.37 (d, 2H, J = 14.00 Hz, 2 x C-CHaHb-CAr),

5.14 (s, 2H, O-CH2-CAr), 7.11 (m, 4H, CArH), 13 7.19 (m, 6H, CArH), 7.40 (m, 5H, CArH). C

NMR (75 MHz, CDCl3) ẟ (ppm): 25.75

((C(CH3)3), 41.90 (C-CH2-CAr) 45.44 (C(CH3)3), p 64.97 (O-CH2-CAr), 94.27 (O=C-C-O), 126.79 (CAr H), 127.45 (CArH), 127.64 p (CAr H), 128.19 (CArH), 128.58 (CArH), 131.16 (CArH), 136.61 (CAr), 137.88 (CAr), 215.30 (C=O). 89 Chapter 1- Identifying a suitable system

2.2.6. BnTAP analog 1-phenyl-2-(1H-1,2,4-triazol-1-yl)ethan-1-one (TAP). 1.83g of 1,2,4-triazol (5mmol), 5.78g of bromoacetophenone (BAP) (1.1eq) and 5.183g of K2CO3 (1.5eq) were added in 10mL of acetonitrile. The mixture was stirred under reflux for 8h30. The mixture was filtered and the solvent was removed under vacuum. Then, water was added and the residue was extracted with ethyl acetate, 3 times. The organic layers were combined, dried over MgSO4, filtered and concentrated under vaccum. The crude was purified by suspension in diethylether. 3.4g (63%) of an orange solid was obtained. Rf= 0.35 (1:1 EA/PE) UV visible + Vanillin revelation (white spot). mp= 1 115.8°C. H NMR (300 MHz, CDCl3) ẟ (ppm): 5.68 (s, 2H, N-CH2-C=O), 7.54 (m, m p o o 2H, C ArH), 7.66 (m, 1H, C ArH), 7.97 (m, 1H, C ArH), 7.99 (m, 1H, C ArH), 8.00 (s, 13 1H, N=CH-N), 8.24 (s, 1H, N=CH-N). C NMR (75 MHz, CDCl3) ẟ (ppm): 55.14 m o p (N-CH2-C=O), 128.22 (CAr H), 129.29 (CAr H), 134.10 (CAr), 134.69 (CAr H), 144.98 & 152.04 (N=CH-N), 190.66 (C=O).

1,3-diphenyl-2-(1H-1,2,4-triazol-1- yl)propan-1-one (BnTAP). 532.5mg of TAP dissolved in 5mL of DMF was added dropwise at 0°C to a stirred suspension of 0.114g of NaH (1eq) in 2mL of DMF. Once, all the TAP added, 644µL of benzyl chloride (2eq) were added at 0°C. Then, the mixture was warmed up to room temperature and left to react for 3h30. Water was added to quench the reaction. Then, ethyl acetate was added to perform a liquid-liquid extraction (3x). The organic layers were combined and dried over MgSO4, filtered and concentrated under a strong vacuum to remove the remaining DMF. The crude was purified by filtration on silica with 1:1 (EA/PE) followed by recrystallization in Et2O.

665mg (86%) of a white solid was obtained by filtration. mp=145.0°C. Rf= (1:1 1 EA/PE) UV revelation + Vanillin (whitish spot). H NMR (300 MHz, CDCl3) ẟ

(ppm): 3.42 (dd, 1H, J = 14.2, 8.8 Hz, CH-CHaHb- CAr), 3.55 (dd, 1H, J = 14.2, 5.8

Hz, CH-CHaHb- CAr), 6.24 (dd, 1H, J = 5.8, 8.8 Hz, CH2-CH-C=O), 7.01 (m, 2H,

CAr=CArH-CArH), 7.22 (m, 3H, CArH=CArH-CArH), 7.47 (m, 2H, CArH=CArH-CArH),

7.59 (m, 1H, CArH=CArH-CArH), 7.91 (s, 1H, N=CH-N), 7.94 (m, 2H, CAr=CArH-

CArH), 8.27 (s, 1H, N=CH-N). NMR (75 MHz, CDCl3) ẟ (ppm): 38.73 (CH-CH2- p CAr), 65.01 (N-CH-C=O), 127.56 (CAr H) 128.75 (CArH), 128.97 (CArH), 129.04 p (CArH), 129.17 (CArH), 134.34 (CAr H), 134.58 (CAr), 135.36 (CAr), 143.40 & 151.63 (N=CH-N), 193.59 (C=O). 90 Chapter 1- Identifying a suitable system

By-product of BnTAP (E)-1-(1-(benzyloxy)-1,3-diphenylprop-1- en-2-yl)-1H-1,2,4-triazole (E-BnTAPBn). 1 H NMR (300 MHz, CDCl3) ẟ (ppm): 3.83 (s,

2H, C-CH2-CAr), 4.52 (s, 2H, O-CH2-CAr), o o 6.85 (m, 2H, CAr H), 6.97 (m, 2H, CAr H),

7.12 (m, 3H, CArH), 7.27 (m, 3H, CArH), 7.50

(s, 5H, CArH), 7.96 (s, 1H, N=CH-N), 8.20 (s, 13 1H, N=CH-N). C NMR (75 MHz, CDCl3) ẟ

(ppm): 36.03 (C-CH2-CAr), 71.91 (O-CH2- p CAr), 120.24 (C-C=C-O), 126.62 (CAr H), p 128.11 & 128.13 (CArH), 128.44 (CAr H), p 128.53 & 128.62 (CArH), 128.98 (CArH), 129.90 (CAr H), 129.99 (CArH), 132.22 (CAr), 136.07 (CAr), 137.84 (CAr), 146.17 (N=CH-N), 148.97 (O-C=C), 151.54 (N=CH-N).

2.2.7. BnClTAmBn2 analog 1H-1,2,4-triazole-1-acetamide (TAm). 739mg of triazol, 3g of chloroacetamide (CAm) (3eq), 2.96g of

K2CO3 (2eq) and 121mg of BnEt3NCl (5%) were added in a round bottom flask to 20mL of acetone. The mixture was stirred at reflux for 4h. The mixture was left to cool down to room temperature and was then filtered. The solid was then suspended in MeOH and stirred for 1h. The mixture was filtered and both filtrates were combined and concentrated under vacuum. The crude was purified by filtration on silica with MeOH10:DCM90/PE (ratio going from 1/2 to 1/0) . 842mg (62%) of a white solid is obtained after filtration. mp = 138.3°C, Rf= 0 (1:1 EA/PE); 0.44 (1:9 MeOH/DCM) Vanillin (white spot). 1H NMR (300 MHz, MeOD) ẟ (ppm): 5.00 (s, 13 2H, N-CH2-C=O), 8.00 (s, 1H, N=CH-N) 8.50 (s, 1H, N=CH-N). C NMR (75 MHz,

MeOD) ẟ (ppm): 52.15 (N-CH2-C=O), 146.59 & 152.27 (N=CH-N), 170.47 (C=O). 91 Chapter 1- Identifying a suitable system

N,N-bis(phenylmethyl)-1H-1,2,4-triazole-1- acetamide (TAmBn2). 945mg of TAm dissolved in 10mL of DMF was added dropwise at 0°C to a stirred suspension of 0.635g of NaH (2eq) in 5mL of DMF. Once, all the TAm added, 2.74mL of benzyl chloride (3eq) were added at 0°C. Then, the mixture was warmed up to room temperature and left to react for 4h. Water was added to quench the reaction and ethyl acetate was added to perform a liquid-liquid extraction. The organic layers were combined and dried over MgSO4, filtered and concentrated under a strong vacuum to remove the DMF. The mixture was purified by filtration over silica, Eluent: 20:80 then 50:50. Two products are obtained, the desired one and the α-benzylated version (BnTAmBn2).

670mg (29%) of a white solid was obtained. mp= 72.4°C, Rf= 0.12 (1:1 EA/PE) UV 1 revelation + Vanillin (white spot). H NMR (300 MHz, CDCl3) ẟ (ppm): 4.46 (s, 2H,

N-CH2-CAr), 4.57 (s, 2H, N-CH2-CAr), 4.98 (s, 2H, N-CH2-C=O), 7.13 (m, 4H,

CAr=CArH-CArH), 7.27 (m, 6H, CArH=CArH-CArH), 7.87 (s, 1H, N=CH-N), 8.16 (s, 13 1H, N=CH-N). C NMR (75 MHz, CDCl3) ẟ (ppm): 49.57 (N-CH2-CAr) 49.79 (N- N N p CH2-CAr), 50.60 (CH2-CH-C=O), 126.28 ( CArH), 127.97 ( CAr H), 128.26 N p N N N N ( CAr H), 128.56 ( CArH), 128.88 ( CArH), 129.41 ( CArH), 135.29 ( CAr), 136.21 N ( CAr), 144.94 & 151.85 (N=CH-N), 166.05 (N-C=O).

α-[4-chlorophenylmethyl]-N,N- bis(phenylmethyl)-1H-1,2,4-triazole-1- acetamide (BnClTAmBn2). 1g of

TAmBn2 dissolved in 10mL of DMF was added dropwise at 0°C to a stirred suspension of 0.131g of NaH (1eq) in

5mL of DMF. Once, all the TAmBn2 was added, 1.05g of 4-chlorobenzyl chloride (2eq) were added at 0°C. Then, the mixture was warmed up to room temperature and stirred for 3h30. Water was added to quench the reaction and ethyl acetate was added to perform a liquid- liquid extraction. The organic layers were combined and dried over MgSO4, filtered and concentrated under a strong vacuum to remove the DMF. The mixture was purified by column chromatography (Packing: pure PE, Eluent: 20:80 (EA/PE) until the product comes out then 50:50). 1.27g (73%) of a white solid was obtained after removal of the solvent. mp= 139.5°C, Rf= 0.45 (1:1 EA/PE) UV revelation + Vanillin 1 (white spot). H NMR (300 MHz, CDCl3) ẟ (ppm): 3.25 (dd, 1H, J = 13.1, 5.9 Hz,

CH-CHaHb- CAr), 3.58 (dd, 1H, J = 13.1, 9.4 Hz, CH-CHaHb- CAr), 4.26 (d, 1H, J 92 Chapter 1- Identifying a suitable system

=17.3 Hz, N-CHaHb-CAr), 4.39 (d, 1H, J = 14.6 Hz, N-CHaHb-CAr), 4.51 (d, 1H, J =

17.3 Hz, N-CHaHb- CAr), 4.69 (d, 1H, J = 14.6 Hz, N-CHaHb-CAr), 5.50 (dd, 1H, J =

9.4, 5.9 Hz, CH2-CH-C=O), 6.79 (m, 2H, CAr=CArH-CArH=CArCl), 7.00 (m, 4H,

CAr=CArH-CArH), 7.17 (m, 2H, CAr-CArH=CArCl), 7.25 (m, 6H, CArH=CArH-CArH), 13 7.89 (s, 1H, N=CH-N), 8.38 (s, 1H, N=CH-N). C NMR ( 75 MHz, CDCl3) ẟ (ppm):

38.87 (CH-CH2-CAr), 49.55 (N-CH2-CAr) 50.23 (N-CH2-CAr), 60.91 (CH2-CH- N N p N p N C=O), 126.24 ( CArH), 127.93 ( CAr H), 128.07 ( CAr H), 128.25 ( CArH), 128.86 N N ( CArH), 129.12 (CArH), 129.29 ( CArH), 130.89 (CArH), 133.47 (CAr), 133.70 N N (CArCl), 135.40 ( CAr), 136.06 ( CAr), 142.36 & 151.41 (N=CH-N), 167.96 (N- C=O).

2.2.8. BnFTAmBn2 analog α-[4-fluorophenylmethyl]-N,N- bis(phenylmethyl)-1H-1,2,4-triazole-1-

acetamide (BnFTAmBn2). 1g of TAmBn2 dissolved in 10mL of DMF was added dropwise at 0°C to a stirred suspension of 0.131g of NaH (1eq) in 5mL of DMF. Once,

all the TAmBn2 was added, 782µL of 4- fluorobenzyl chloride (2eq) were added at 0°C. Then, the mixture was warmed up to room temperature and stirred for 3h30. Water was added to quench the reaction and ethyl acetate was added to perform a liquid-liquid extraction. The organic layers were combined and dried over MgSO4, filtered and concentrated under a strong vacuum to remove the DMF. The mixture was purified by column chromatography (Packing: pure PE, Eluent: 20:80 (EA/PE) until the product comes out then 50:50). 1.16g (82%) of a white solid was obtained after removal of the solvent. mp= 122.3°C, Rf= 0.34 (1:1 1 EA/PE) UV revelation + Vanillin (whitish spot). H NMR (300 MHz, CDCl3) ẟ

(ppm): 3.26 (dd, 1H, J = 13.2, 5.9 Hz, CH-CHaHb- CAr), 3.58 (dd, 1H, J = 13.2, 9.3

Hz, CH-CHaHb- CAr), 4.25 (d, 1H, J = 17.3 Hz, N-CHaHb-CAr), 4.43 (d, 1H, J = 14.6

Hz, N-CHaHb-CAr), 4.49 (d, 1H, J = 17.4 Hz, N-CHaHb- CAr), 4.65 (d, 1H, J = 14.6

Hz, N-CHaHb-CAr), 5.51 (dd, 1H, J = 9.3, 6.0 Hz, CH2-CH-C=O), 6.83 (m, 2H,

CAr=CArH-CArH-CArF), 6.89 (m, 2H, CArH-CArH=CArF), 7.02 (m, 4H, CAr=CArH-

CArH), 7.26 (m, 6H, CArH=CArH-CArH), 7.88 (s, 1H, N=CH-N), 8.37 (s, 1H, N=CH- 13 N). C NMR (75 MHz, CDCl3) ẟ (ppm): 38.70 (CH-CH2-CAr), 49.47 (N-CH2-CAr)

50.13 (N-CH2-CAr), 61.05 (CH2-CH-C=O), 115.64-115-92 (d, J= 21.4 Hz, CArH- N N p N p N CArF), 126.27 ( CArH), 127.88 ( CAr H), 128.02 ( CAr H), 128.28 ( CArH), 128.81 N N ( CArH), 129.15 ( CArH), 130.93-130.97 (d, J = 3.2 Hz, CAr), 131.04-131.15 (d, J = N N 8.1 Hz, CArH-CAr), 135.42 ( CAr), 136.09 ( CAr), 142.37 & 151.41 (N=CH-N), 93 Chapter 1- Identifying a suitable system

19 160.54-163.80 (d, J = 246.1 Hz, CArF) 168.06 (N-C=O). F NMR (282 MHz, CDCl3) ẟ (ppm): -114.88.

2.2.9. BnTAmBn2 analog α-[phenylmethyl]-N,N-bis(phenylmethyl)- 1H-1,2,4-triazole-1-acetamide (BnTAmBn2). 1g of TAm dissolved in 10mL of DMF was added dropwise at 0°C to a stirred suspension of 0.952g of NaH (3eq) in 5mL of DMF. Once, all the TAm added, 5.46mL of benzyl chloride (6eq) were added at 0°C. Then, the mixture was warmed up to room temperature and left to react for 6h30. Water was added to quench the reaction and ethyl acetate was added to perform a liquid-liquid extraction. The organic layers were combined and dried over MgSO4, filtered and concentrated under a strong vacuum to remove the DMF. The mixture was purified by column chromatography (Packing: 50:50 (EA:PE), Eluent: 50:50 ; 75:25 ; 1:0. Two products are obtained, the desired one (comes out between 50:50 and 75:25) and the diprotected amide (N,N- bis(phenylmethyl)-1H-1,2,4-Triazole-1-acetamide (TAmBn2). 1.14g (36%) of a white solid was obtained. mp = 113.5°C, Rf = 0.46 (1:1 EA/PE) UV revelation + Vanillin 1 (white spot). H NMR (300 MHz, CDCl3) ẟ (ppm): 3.32 (dd, 1H, J = 13.1, 6.1 Hz,

CH-CHaHb- CAr), 3.61 (dd, 1H, J = 13.1, 9.1 Hz, CH-CHaHb- CAr), 4.24 (d, 1H, J =

17.2 Hz, N-CHaHb- CAr), 4.43 (d, 1H, J = 17.7 Hz, N-CHaHb-CAr), 4.50 (d, 1H, J =

15.1 Hz, N-CHaHb- CAr), 4.57 (d, 1H, J = 14.7 Hz, N-CHaHb-CAr), 5.58 (dd, 1H, J =

9.1, 6.1 Hz, CH2-CH-C=O), 6.84 (dd, 2H, J = 6.9, 2.5 Hz, CAr=CArH-CArH), 7.03 (dd,

2H, J = 6.7, 3.0 Hz, CAr=CArH-CArH), 7.07-7.10 (m, 2H, CAr=CArH-CArH), 7.19-7.34 13 (m, 9H, CArH=CArH-CArH), 7.89 (s, 1H, N=CH-N), 8.40 (s, 1H, N=CH-N). C NMR

(75 MHz, CDCl3) ẟ (ppm): 39.44 (CH-CH2-CAr), 49.13 (N-CH2-CAr) 49.92 (N-CH2- N p N p CAr), 60.96 (CH2-CH-C=O), 126.32 ( CArH), 127.39 (CAr H), 127.68 ( CAr H), N p N N N 127.88 ( CAr H), 128.22 ( CArH), 128.68 ( CArH), 128.85 (CArH), 129.05 ( CArH), N N 129.37 (CArH), 135.09 (CAr), 135.33 ( CAr), 136.05 ( CAr), 142.41 & 151.25 (N=CH- N), 168.16 (N-C=O).

2.2.10. MBnTA analog

94 Chapter 1- Identifying a suitable system

α-Bromo-benzenepropanoic acid (BnBAA) & α-(1H-1,2,4-triazol-1-yl)- benzenepropanoic acid (BnTAA). Their synthesis and characterization by NMR are described in Appendix B. Their racemic DSCs are displayed in Appendix A (Figures

A-14 & A-15). For BnBAA, mp = 48.1°C. For BnTAA, mp = 222.6°C.

Methyl-(1H-1,2,4-triazole-phenylmethyl)-acetate (MBnTA). Thionyl chloride (368 µL, 1.1eq) is added

to a solution of BnTAA (1g, 4.6 mmol) and Et3N (933 µL, 1.5eq) in methanol (10 mL) at 0 °C. The solution is stirred for 20 minutes at 0 °C and then for 12 hours at 40 °C. The reaction was quenched with 1N HCl. The

resultant mixture was extracted with CH2Cl2. The combined organic layers were washed with a saturated

NaHCO3 solution and brine successively. The resulting organic layers were dried over Na2SO4. The solvent was removed under reduced pressure.11 The crude was purified by column chromatography (Packing (25:75

EA90MeOH10/PE) First eluent 25:75 second 1:2 when product starts coming out).

751mg (71%) of a white solid is obtained after removal of solvent. mp= 89.4°C. Rf= 1 0.35 (1:1 EA/PE) UV revelation + Vanillin (white spot). H NMR (300 MHz, CDCl3)

ẟ (ppm): 3.48 (d, 1H, J = 8.3 Hz, CH-CHaHb- CAr), 3.48 (d, 1H, J= 6.5 Hz, CH-CHaHb-

CAr), 3.74 (s, 3H, O-CH3), 5.20 (dd, 1H, J = 8.3, 6.5 Hz, CH2-CH-C=O), 6.95 (m, 2H,

CArH=CArH-CArH), 7.20 (m, 3H, CAr=CArH-CArH), 7.93 (s, 1H, N=CH-N), 7.95 (s, 13 1H, N=CH-N). C NMR (75 MHz, CDCl3) ẟ (ppm): 38.10 (CAr-CH2-CH), 53.23 (O- p CH3), 63.94 (CH2-CH-C=O), 127.62 (CAr H), 128.91 & 128.97 (CArH), 135.28 (CAr),

143.93 & 152.02 (N=CH-N), 168.66 (CH3O-C=O).

Original precursor and by-products of MBnTA Methyl-(1H-1,2,4-Triazole)-acetate (MTA). 1H NMR (300 MHz, CDCl3) ẟ (ppm): 3.58 (s, 3H, CH3-O), 4.84 (s, 2H, N-CH2-C=O), 7.76 (s, 1H, N=CH-N), 8.06 (s, 1H, N=CH-N). 13C NMR (75 MHz, CDCl3) ẟ (ppm): 50.58 (CH3-O), 53.31 (N-CH2- C=O), 144.70 & 152.44 (N=CH-N), 167.23 (O-C=O).

1 Methyl 2-benzyl-3-phenyl-2-(1H-1,2,4-triazol-1-yl)propanoate (MBn2TA). H

NMR (300 MHz, CDCl3) ẟ (ppm): 3.56 (d, 2H, J = 13.6 Hz, 2 x CH-CHaHb-CAr), 3.76

(s, 3H, CH3-O), 3.81 (d, 2H, J = 13.6 Hz, 2 x CH-CHaHb-CAr), 6.70-6.74 (m, 4H,

CAr=CArH-CArH), 7.12-7.19 (m, 6H, CArH=CArH-CArH), 8.11 (s, 1H, N=CH-N), 8.11 13 (s, 1H, N=CH-N). C NMR (75 MHz, CDCl3) ẟ (ppm): 45.14 (C-CH2-CAr), 52.76 p (CH3-O), 75.12 ((CH2)2-C-C=O), 127.85 (CAr H), 128.81 & 129.61 (CArH), 134.78 (CAr), 146.09 & 151.08 (N=CH-N), 169.43 (O-C=O).

1-(1-(benzyloxy)-1-methoxy-3-phenylprop-1-en-2-yl)-1H-1,2,4-triazole 1 (MBnTABn). H NMR (300 MHz, CDCl3) ẟ (ppm): 2.90 (s, 2H, C-CH2-CAr), 3.78 95 Chapter 1- Identifying a suitable system

(s, 3H, O-CH3), 4.96 (s, 2H, O-CH2-CAr), 6.96-7.37 (m, 10H, CArH=CArH-CArH), 7.96 13 (s, 1H, N=CH-N), 8.17 (s, 1H, N=CH-N). C NMR (75 MHz, CDCl3) ẟ (ppm): 46.58

(C-CH2-CAr), 53.13 (CH3-O), 64.24 (O-CH2-CAr), 126.00 (CArH), 126.92 (C-C=C), 127.23 (CArH), 128.42 & 128.54 (CArH), 128.63 (CArH), 129.22 (CArH), 129.51(CArH), 137.51 (CAr), 141.86 (CAr), 144.53 & 152.26 (N=CH-N), 167.03 (O- C=O).

2.3. Results & discussion The synthesis of the model compound was a success following literature-based procedures. Additionally, this set of reactions allowed for the synthesis of the majority of the analogues by changing the substrate, the reagent and/or the solvent when necessary. Out of all 9 analogues, only MBnTA was not synthetized in a similar fashion. We did indeed first try to synthesize MBnTA in a two-reaction synthesis starting from Methyl chloroacetate, but this proved to be impossible due to the phenomenon of dialkylation.

2.3.1. Dialkylation: A competition between O- and C- alkylation By-product study Indeed, even if the reaction of alkylation of a carbonyl derivative through enolization and reaction on a halogenoalkane derivative may be very straightforward; in theory this can lead to by-products when more than one equivalent of NaH is used (Scheme 1-2). In the case of BnFTP, BnTP, BnClTP, BnBnOP, PgTP and BnTAPa, the use of 1 equivalent of NaH also led to the formation of by-products, albeit in relatively low amounts (the total ratio of by-products ranged between 5 and 10% depending on the reactions) compared to the desired product. Up to three by-products could be identified by NMR in some cases:b

BnFTP & BnTP : three by-products identified BnClTP: two by-products identified PgTP, BnBnOP & BnTAP: One by-product identified

The general structure of each by-product is given in Scheme 1-2.

a For the TAm derivatives, no formation of dibenzylated by-products was detected over all carried out reaction. b When not all three by-products were identified, it does not mean they don’t exist but if they do their formation remains very low compared to the other one(s) 96 Chapter 1- Identifying a suitable system

Scheme 1-2 The reaction of alkylation of a carbonyl derivative by enolization with NaH. First alkylation gives the major product in red. Second alkylation can give the following dibenzylated by-products: the C-dibenzylated by-product (top right), the (Z)-O-dibenzylated by-product (bottom left) and the (E)-O-dibenzylated by-product (bottom right).

These by-products form once the monobenzylated product is present. The enolate of this latter is more stable than the reactant enolate. Hence, the proton of the product is more acidic. Using only one equivalent of base, the side products can be kept in low concentration, as initially no product is present. Considering the halogenated reactant is present in excess, carrying out the reaction with two equivalents of base, the by- products would be the major end-products.

Indeed, in the case of an enolization, both the oxygen and the carbon can form the new bond. Whether it is one or the other depends on the hard/soft nature of the , on the size of the counter (since small counter ion binds tightly to the oxygen, promoting C-alkylation) and on the solvent. This latter, if protic, will do hydrogen bond interaction with the oxygen, attenuating its reactivity while polar aprotic solvents are poor anion solvators but good cation solvator, leaving the oxygen free to attack.12-15 In the case of this synthesis, the electrophile (R-Cl) is soft, though less soft than R-Br or R-I equivalent and the counter ion is small (Na+). Those two criteria favor C-alkylation. However, the solvent, DMF is polar aprotic, favoring O- Alkylation. Experimentally we can observe that C-alkylation is majorly favored since at the first stage, O-alkylation was not observed. Experimentally, things get more complicated for the second stage where both alkylation are observed for BnFTP, BnTP & BnClTP. Since, the conditions of solvent, counter ion and electrophile remained the same, one could reason O-alkylation competing with C-Alkylation (helped by DMF favoring O-Alkylation) at the second stage resulted from another parameter. Indeed, in this specific case, the carbon where the second benzyl group attaches is sterically hindered with the triazole and the other benzyl group. This would decrease the difference in the activation energy between both alkylation, making O- Alkylation a viable option. It must be noted that the dibenzylated compound remained more favorable as it was the major by-product of the three. In addition, PgTP by- 97 Chapter 1- Identifying a suitable system products follow the steric hindrance principle with the exact same reaction conditions. Indeed, for PgTP, only the C-dialkylated product was observed, due to reduced steric hindrance of the propargyl group. For BnTAP, replacing the tertbutyl by a phenyl provides the negatively charged oxygen atom with a +M mesomeric effect, making it more reactive toward alkylation and explaining why it is the major by-product of the three. Finally, in the case of BnBnOP, NaI was used as catalyst, favoring even more C-alkylation, this can be experimentally observed by the C-alkylated by-product being the major by-product despite a non-negligible bulk in the molecule.

Furthermore, alongside the NMR analysis, in case of BnFTP synthesis, two of the three products were analyzed by SC-XRD. The results of the analysis confirmed the existence of the following two compounds (figure 1-3, crystallographic data in Appendix A, tables A-3 & A-4):

C-dibenzylated compound: 4,4-dimethyl-1-(4-fluorophenyl)-2-(4- fluorobenzyl)-2-(1H-1,2,4-triazol-1-yl)-pentan-3-one O-dibenzylated compound: (Z)-1-(3-((4-fluorobenzyl)oxy)-1-(4- fluorophenyl)-4,4-dimethylpent-2-en-2-yl)-1H-1,2,4-triazole.

Figure 1-3 On the left the molecular structure of the C-dibenzylated by-product of BnFTP is displayed. On the right, the molecular structure of the (Z)-O-dibenzylated by-product of BnFTP is shown.

The crystalline structure of the O-alkylation product showed a (Z) configuration of the double bond. Thanks to this information, by analogy, for BnTP and BnClTP, the O-alkylation products showing more or less the same chemical shifts for the direct substituents of the double bond were attributed with the (Z) configuration. Finally, the third by-product was attributed as the (E)-O-dibenzylated compound, because its carbon NMR showed the enol carbon peaks comparable to those of the (Z)-O- dibenzylated compound while the proton NMR showed the exact same type of peaks though shifted. This by-product was the least present each time with BnTP and BnFTP. Thus, the (E)-enol ether appeared to be less favored than its (Z)-diastereomer. This could either be because of a chelate effect of Na+ with the oxygen of the ketone and the nitrogen 2 of the triazole, or because the (Z)-diastereomer is the one with the 98 Chapter 1- Identifying a suitable system

less steric hindrance or a combination of both. In the case of the chelation, (Z)- selectivity was already observed in the synthesis of silyl enol ether with a small base like NaH. The authors hypothesized a 6-center transition state with chelation of the oxygen atom and the nitrogen atom present in β position.16 In our case, the nitrogen 2 of the triazole could also yield a 6-center transition state (figure 1-4).

Figure 1-4 Hypothized 6-center transition state explaining (Z) configuration. On the left, the one described in reference 16 and on the right, the equivalent one obtained with our set of molecules. The 6-center ring is drawn in grey and R1 & R2 refer to the groups described in scheme 1-2.

Regarding hindrance, generally (E)-alkenes are more stable than their (Z)-version because the two high priority groups are often the bulkier ones. In this case, the priority numbers do not match the relative bulkiness of the substituents, making the

(Z)-version more stable. The R1 and triazole groups were bulky at short range while the benzylated groups were bulky at long range. Consequently, having one short-range bulky group next to a large-range bulky group leads to less hindrance than two groups of the same type of bulkiness next to each other. Since the highest priority group was the triazole on one side and the O-benzyl group on the other, the (Z)-enol ether was less hindered and more stable. Interestingly, by HNMR analogy, it appeared that in the case of BnTAP by-products, the (E) form was the more stable. This would be explained by the fact that both the triazole and benzyl group are planar and can position themselves in either a parallel plane or perpendicular to avoid hindrance.

MBnTA synthesis: the issue of low steric hindrance The first intermediate, methyl-(1H-1,2,4-triazole)-acetate (MTA, figure 1-5) was synthetized using a similar reaction as for synthesizing TP with in addition the presence of a phase transfer agent, triethyl benzyl amine chloride (BTEAC, + – Et3BnN Cl ) in order to facilitate the reaction between the triazole and the carbonate ion. This reaction was based on literature17 and was extended in this work to the synthesis of TAm. Once MTA was synthetized, MBnTA was attempted in the same way as all other analogues. However, with the same conditions as for the other reactions, no product was formed. Several rates of MTA addition were tried but either no reaction occurred and MTA was retrieved at the end of the reaction or degradation 99 Chapter 1- Identifying a suitable system was observed. In consequence, stronger bases were tried: n-BuLi, LDA & KHMDS. The solvent used with both was THF.

With n-BuLi, when MTA and Benzyl chloride were added togetherc slowly to the solution of n-BuLi at 0°C, a dibenzylated product was obtained with many impurities (degradation must also have occurred) and no trace of MBnTA: (E)-1-(1-(benzyloxy)- 1-methoxy-3-phenylprop-1-en-2-yl)-1H-1,2,4-triazoled (figure 1-5). The dibenzylated product has the second benzyl added on the O-position. The NMR suggests the (E) form by analogy with the BnFTP by-product. This result highlights that the proton of the monobenzylated product is more acidic than the protons of the reagent as stated before. Indeed, there was only one equivalent of base present and no monobenzylated product was observed.

Figure 1-5 From left to right, the structures of: MTA, (E)-1-(1-(benzyloxy)-1-methoxy-3- phenylprop-1-en-2-yl)-1H-1,2,4-triazole, and methyl 2-benzyl-3-phenyl-2-(1H-1,2,4-triazol-1- yl)propanoate.

Attempts with LDA & KHMDS led either to no reaction or only dibenzylation resulting from the C-alkylation of MBnTA: methyl 2-benzyl-3-phenyl-2-(1H-1,2,4- triazol-1-yl)propanoate (figure 1-5).

Consequently, it appeared that the steric hindrance of the methyl ester group was too low to stop the reaction at the first step. Thus, an alternative synthetic path was thought of starting with a compound that already possessed the benzyl ring, the naturally occurring phenylalanine amino acid. This path successfully led to the synthesis of MBnTA but also to the development of a green synthesis of the precursor of MBnTA, BnTAA in its enantiopure form, which is detailed in Appendix B.

c When n-BuLi was added on MTA alone, being too nucleophile, there was competition between deprotonation and the nucleophilic attack of n-BuLi on the ester. d E form was attributed by comparison of the NMR with the ones obtained in the case of the synthesis of BnFTP. 100 Chapter 1- Identifying a suitable system

Figure 1-6 Asymmetric unit of BnBAA (left) and BnTAA (right). Hydrogen bonds are displayed in red.

In the case of the synthesis of racemic BnTAA, both its crystal structure and that of its precursor BnBAA were determined (figure 1-6, crystallographic data in Appendix A, tables A-1 & A-2), as ultimate proof of their synthesis. The crystal structure of BnTAA shows an intermolecular hydrogen bond between the nitrogen in position 4 of the triazole and the proton of the acid that potentially explains its higher melting point compared to the rest of the analogues.

2.3.2. Crystallization developments Except for the synthesis of BnTAA, whose product precipitates after post-workup, all carried-out reactions were first purified by column chromatography. In the case of TP,

BnClTP, BnFTP and BnTP, PgTP, BnFTAmBn2, BnClTAmBn2 and TAP crystallization upgrades were performed.

For TP and TAP, post-workup, the 4-H-triazole by-product that is hydrophilic was already removed by liquid-liquid separation. In the organic phase remained the product, the excess of starting material (respectively chloropinacolone and bromoacetophenone) and the polar orange colored impurities. Upon 3 consecutive suspensions in diethylether at room temperature, respectively 84% of a slightly yellow up to white product and 63% of an orange solid was obtained. Coloration of the products is due to traces of a strongly colored impurity, undetectable in NMR.

For the remaining compounds, they were synthetized with analogue reactions making their purification by crystallization similar:

For BnClTP, the by-products, the excess of chlorobenzyl chloride and the yellow impurities were removed by suspension of the crude in hexane at room temperature for at least one day with a yield of 79%. For BnTP, the by-products, the excess of benzyl chloride and the yellow impurities were removed by suspension of the crude in a 50/50 mixture of toluene/cyclohexane at room temperature for at least one day with a yield of 78%. 101 Chapter 1- Identifying a suitable system

For BnFTAmBn2 and BnClTAmBn2, the crude was purified by crystallization in a 50/50 mixture of hexane/toluene, yielding for both a yield of 62%. For BnFTP, The crude was purified by crystallization in toluene at T<20°C using formation of a solvate with a yield of 82%. BnFTP is difficult to crystallize and cannot even be crystallized from pure hexane. However, its toluene solvate crystallizes well but shows a low melting peak (around 20°C), making the crystallization impossible at room temperature.

2.4. Conclusion Nine analogs of BnClTP were successfully synthetized with some synthesis scaled- up to several grams, allowing production of the necessary quantities of analogs for the co-crystal screening. The second step of the synthesis though straightforward revealed to favor dialkylation, limiting the use of base to a maximum of one equivalent. Furthermore, the by-products formed showed competition between C-alkylation and O-alkylation. Finally, in case of MBnTA, the original synthesis was not viable due to low steric hindrance of the methyl ester group compared to other groups, yielding only dialkylation with one equivalent of base. The use of an alternative synthesis led to the development of a green synthesis.

3. Co-crystal screening 3.1. Introduction Forming a co-crystal is not as straightforward as making a salt. One usually follows the principles of crystal engineering, which can be seen as a synonym to supramolecular synthesis.18 Indeed, where a salt relies on an ionic bonding between the two partners, a co-crystal is based on weaker interactions linking the two partners, generally hydrogen bonding interactions.19 In theory, a co-crystal of a target molecule can be formed with any co-former of any shape or size as long as it possesses complementary hydrogen bonding functionalities.20 In reality, the ability of a target molecule to co-crystallize with a co-former is much more complex as the overall solid state has to be more stable than any other solid states that can be formed.19 Thus, two molecules that can hydrogen bond will therefore not always form a co-crystal. Overall, the free energy of the following reaction must be negative for co- crystallization to occur, where A and B are the two components:

As+Bs  ABs

Furthermore, the free energy of this reaction depends on the stoichiometry of the co- crystal, the temperature and pressure. On top, even if the cocrystal is the favored outcome, kinetic aspects also play and some techniques might lead to cocrystal 102 Chapter 1- Identifying a suitable system

formation while others do not.20 Consequently, experimental screening for a co- crystal is not trivial and typically consists of 4 stages (figure 1-7):

Selection of co-formers Co-crystal synthesis Solid-state Analysis Co-crystal proof

Figure 1-7 The four stages of co-crystal screening.

The first step is the selection of the co-formers. There are several approaches that exist with the most common being the use of the Cambridge Structural Database (CSD). Looking for already existing co-crystals of the target molecule, effective synthons for co-crystallization can be identified.18,21,22 As the reaction shown above has a stabilizing free energy of a couple of kcal.mol-1, coming to a methodology that is truly predictive remains a challenge. Even though some computer-guided tools exist, co- crystal screening remains a trial and error approach, where one can still be guided by some common principles.20-22 For this reason, we proceeded through an experimental screening. The general idea behind screening implies mixing the solid compound of interest and the co-former and analyzing the solid outcome by a solid-state characterization method e.g. Powder X-Ray Diffraction (PXRD), Single-Crystal X- ray Diffraction (SCXRD), Differential Scanning Calorimetry (DSC).22,23

Typically used screening methods are based on grinding, crystallization from solution or from the melt.18,24-26 Grinding remains the most attractive method for large screening of co-crystals, being also eco-friendly due to no or low quantities of solvent used in respectively neat-grinding and liquid-assisted grinding.25 In the case of liquid- assisted grinding also called wet-grinding, a small quantity of solvent, is used to catalyze the process through local dissolution and/or increased orientational and 103 Chapter 1- Identifying a suitable system conformational freedom at the interface; it is usually preferred to neat grinding.25 In the case, of neat grinding, there are three possible mechanism explaining co-crystal formation:27,28

Through molecular diffusion: the co-crystal is formed by vapor and/or surface diffusion upon grinding. Via a liquid phase: the co-crystal can be formed from the melt of the eutectic upon grinding. Via an amorphous phase: the co-crystal recrystallizes from the amorphous phase formed upon grinding.

The latter is generally the most accepted hypothesis and likely occurs for most cases as grinding is known to induce amorphization. The first generally occurs for solids with relatively high vapor pressure while the second occurs for low eutectic melting combinations. In the case of solution-based screening, suspension mediated transformation (slurrying)29 or slow evaporation from an undersaturated solution are the most widely used.18

Regarding the analysis of the obtained solid phases, PXRD is the most practical method for analyzing a large number of samples. DSC is generally used complementarily to PXRD. When comparing PXRD patterns of parent compounds and the outcome, co-crystal formation is suspected by the appearance of new diffraction peaks, not belonging to either of the parent compounds. Furthermore, in general this is accompanied by complete disappearance of the peaks corresponding to both starting compounds.25,30

Finally, when a possible co-crystal is identified by PXRD, single crystals are usually grown to ultimately confirm co-crystallization occurred.26,28 When no single crystal can be grown, construction of a phase diagram alternatively can prove the existence of a co-crystal albeit being a more cumbersome technique.31,32

In this part, the 10 synthetized compounds (BnClTP and its nine analogues) were screened using a library of different co-formers, selected according to three criteria. A second screening was performed with a library selected based on the positive hits of the first screen. Several hits were suspected with some being confirmed by SC- XRD. From all the confirmed co-crystals, the BnFTP and 3-Phenylbutyric acid, system was selected for the development of the CoISD process. This system is ideal, as a pair of diastereomeric co-crystals is formed showing strong difference in solubility, racemization conditions were easily identified, and furthermore the co- former is liquid at room temperature, avoiding its crystallization during the process. 104 Chapter 1- Identifying a suitable system

3.2. Material & methods 3.2.1. Co-crystal screening For co-crystal screening, a 1:1 mixture of the target compound and the co-former, were ground in a RETSCH MM 400 mixer mill for 90 min with a beating frequency of 30 Hz. Typically, between 20 and 25mg of the target were weighed in a 2mL Eppendorf. One molar equivalent of the co-former was then added together with 3- mm stainless balls.

3.2.2. Solid-state analysis The resulting solid-phases were all analyzed by PXRD. Some powder X-ray Diffraction measurements were performed with a Siemens D5000 diffractometer equipped with a Cu X-ray source operating at 40 kV and 40 mA and a secondary monochromator allowing the selection of the Kα radiation of Cu (λ = 1.5418 Å). A scanning range of 2θ values from 5° to 50° at a scan rate of 0.6° min−1 was applied. The others were done with a PANalytical Bragg-Brentano diffractometer equipped with a Ni-filtered CuKα (λ = 1.54179 Å) at 45 kV and 30 mA with an X’Celerator detector. A scanning range of 2θ values from 4 to 40° was applied for a total scan time of 6,7min.

3.2.3. Single crystal Growth For Single crystal growth, the 2 following methods were used as detailed in the results and discussion part:

 Slow evaporation (with and without seeding): Solvents tried are typically = Methanol, Ethanol, Acetone, Toluene, Ethyl Acetate, Acetonitrile, Tert- butanol and 1:1 mixtures of Ethyl Acetate with toluene, THF or acetonitrile.

 Cooling (9°C or 25°C) (with and without seeding): Solvents tried are typically = Methanol, Ethanol, Acetone, Toluene

SC-XRD measurement: cf. section 2.2.

3.2.4. Co-crystal study The (S)-BnFTP and (S)-3-Phenylbutyric acid co-crystal was produced upon crystallization in toluene from a mixture of (R,S)-BnFTP and (S)-3-Phenylbutyric acid. The (S)-BnTP and (S)-3-Phenylbutyric acid co-crystal was obtained in a similar way. Enantiopure co-crystal was obtained after 2-3 recrystallizations in toluene. Enantiopurity was verified by chiral HPLC, with the following set-up: the pump is a Waters 600, the auto-sampler is a Waters 717 and the detector is a Waters 996. The column used for the BnFTP co-crystal was a Chiralpak 1B chiral column with the following dimensions 250x4.6mm and a particle diameter of 5μm. The column used for the BnTP co-crystal was a Chiral Pak IA chiral column with the same dimensions. 105 Chapter 1- Identifying a suitable system

The mobile phase was 95% isohexane and 5% ethanol at a flow rate of 1 mL/min for the BnFTP co-crystal and 90% isohexane and 10% isopropanol at the same flow rate.

DSCs were measured and melting points were measured following the same methodology as detailed in section 2.2.

Solubility curves were measured using a Crystal16 device from Technobis. Using turbidity, the device measures the opacity of the solution over time. Each point was obtained from a certain concentration in the solvent to which a ramp of temperature was applied. A point results from the couple (Concentration, temperature) when the solution reaches a transmittance of 100%.

3.3. Results and discussion 3.3.1. Solid state analysis of the target library Prior to starting the screen, the solid state of all 10 analogues was investigated. Single crystals could be grown for BnFTP, BnTP, BnClTAmBn2, BnFTAmBn2, BnTAmBn2 and MBnTA (figures 1-8 to 1-10). Structural information for all is given in Appendix A (tables A-5 to A-12). For all compounds, the PXRD reference pattern was measured beforehand. Doing so, we obtained a reference PXRD for comparison in the co-crystal screen.

Two single-crystal forms of BnFTP were obtained. In one case, a racemic toluene solvate was obtained, and in the other an enantiopure crystal, showing the likely existence of a conglomerate. However, neither of both simulated patterns corresponds with the pattern of the bulk (ground) material, nor material recrystallized from the melt. This implies the existence of another solid form that could be the racemic compound.

Figure 1-8 Asymmetric unit of BnTP, BnFTP and BnFTP-Toluene from left to right.

BnTP was crystallized as a conglomerate with twinning and its simulated pattern matches the experimental pattern. MBnTA, BnClTAmBn2 and BnFTAmBn2 all crystallized as racemic compounds and the simulated patterns match the experimental ones. 106 Chapter 1- Identifying a suitable system

Figure 1-9 Asymmetric unit of MBnTA, BnFTAmBn2 and BnClTAmBn2 from left to right.

From evaporation experiments, BnTAmBn2 was crystallized either as a racemic compound or as a conglomerate. Both were obtained by slow evaporation and both were encountered during the co-crystal screening (when no co-crystallization occurred). Most of the time, the racemic pattern was the one obtained, which is therefore likely the most stable phase, although a thorough study would be required to confirm this.

Figure 1-10 Asymmetric unit of BnTAmBn2 as a conglomerate (left) and as a racemic compound (right).

With this data in hand, the co-crystal screening was initiated.

3.3.2. Co-crystal screening Initial screen A list of 18 co-formers was chosen based on two mandatory criteria:

The molecule must be chiral (enantiopure or racemic) The molecule must have at least one structural part allowing H-bond donation

We then selected synthons using representatives of different groups of compounds:

Primary and secondary amides: (R,S)-Oxiracetam, (S)-Oxiracetam, Carphedon, Primidone, Etiracetam and Levetiracetam (S). Alcohols (benzylic or aliphatic): (+)-Catechine, (−)-Epicatechine, Dyphilline, Sucrose. 107 Chapter 1- Identifying a suitable system

Carboxylic acids (with alcohol functions: *): D-Tartaric acid*, DL-Tartaric acid*, (R,S)-3-Phenylbutyric acid, R-mandelic acid*, (R)-Phenyllactic acid* and (R,S)-Phenyllactic acid*. Amines (with alcohol functions: *): (R,S)-4-(1-Aminoethyl) and (1S,2R)-(+)-2-Amino-1,2-diphenylethanol*.

The grinding results of the 18 co-formers with the 10 target compounds is given in table 1-1. In case a positive result was observed, single crystal growth was attempted. 5 co-crystal structures were accordingly confirmed by SC-XRD. The asymmetric units of these are given in figures 1-11 to 1-13. Their crystallographic data are given in Appendix A (tables A-13 to A-17).

Table 1-1 Results of the first co-crystal screening: Only those co-formers giving hits are presented. CC means co-crystal was proven by single crystal; SCC means that no proof was obtained by single crystal but the PXRD pattern showed the appearance of a new crystalline phase different from all present compounds, including known polymorphs. A grey cell means the experiment showed no new crystalline phase; or a liquid or an amorphous phase was obtained.

D- DL- RS-3- R- R-Phenyl (+)-2-Amino- RS-Phenyl RS-4-(1-Amino Tartaric Tartaric Phenyl- Mandelic lactic 1,2-diphenyl lactic acid ethyl)pyridine acid acid butyric acid acid acid ethanol BnClTP SCC SCC SCC BnFTP CC CC SCC SCC SCC SCC SCC BnTP CC CC SCC SCC PgTP SCC SCC BnBnOP SCC BnTAP

BnClTAmBn2 SCC

BnFTAmBn2 SCC

BnTAmBn2 CC SCC MBnTA SCC SCC (RS)-Oxiracetam, (S)-Oxiracetam, Carphedon, Primidone, Etiracetam, Levetiracetam, (+)-Catechine, (-)-Epicatechine, Dyphilline & Sucrose did not give any positive result.

These results show that amides and alcohols are not ideal co-formers. Furthermore, only one hit was obtained using amines. However, in the case of acids more positive hits occur. Indeed, BnFTP and BnTP gave hits for every acid. Moreover, out of the eight remaining target compounds, 7 gave at least one positive hit with an acidic co- former. 108 Chapter 1- Identifying a suitable system

Figure 1-11 Asymmetric unit of (R,S)-BnTP + D-tartaric acid co-crystal (left). Asymmetric unit of (R,S)-BnFTP + D-tartaric acid co-crystal (right). Hydrogen bonds are displayed in red.

To understand the propensity of acidic co-formers towards co-crystallization of the target compounds, we had a closer look at the 5 co-crystal structures we were able to obtain. As can be observed for 4 of them the interaction between the target molecule and the acid occurs via a hydrogen bond between the H of the carboxylic acid and the N4 of the triazole. This type of hydrogen bond can be considered strong since the best donor is combined with the best acceptor. This is confirmed by looking at the distance between the oxygen and the nitrogen and the N---H-O angle. For all four co-crystals, the angle is higher than 170°, which corresponds to a strong hydrogen bond (bond between 170° and 180°)33 while the distance between the nitrogen and the oxygen varies between 2.5 and 2.7 Å, which corresponds to a medium-strong hydrogen bond (strong hydrogen bonds are between 2.2 and 2.5 Å while medium hydrogen bond are between 2.5 and 3.2 Å).34 All values are given in Appendix A, table A-20).

Figure 1-12 Asymmetric unit of (R)-BnTP + (R)-3-Phenylbutyric acid co-crystal (left). Asymmetric unit of (S)-BnFTP + (S)-3-Phenylbutyric acid co-crystal (right). Hydrogen bonds are displayed in red.

The only exception with respect to this interaction is found for the co-crystal between

BnTAmBn2 and phenyl lactic acid, where the hydrogen bond occurs between the 109 Chapter 1- Identifying a suitable system oxygen of the amide and the H of the carboxylic acid. Though the triazole is present in BnTAmBn2, it does not form the hydrogen bond observed for the other 4, probably because of too bulky benzyl groups on the amide. The resulting hydrogen bond is less strong, with an angle of around 160° and N---(H)--O distance in the medium-strong range (2.542 Å). Additionally, table 1-1 highlights that for the bulkier molecules fewer positive hits are identified, as well as for those molecules having a higher melting point. The exception to this rule being BnBnOP, which has the lowest melting point but only yielded a single positive hit. Its difficulty to co-crystallize can however be understood by the absence of the triazole function.

Figure 1-13 Asymmetric unit of (R,S)-BnTAmBn2 + (R,S)-Phenyl lactic acid co-crystal. Hydrogen bonds are displayed in red.

3.3.2.1. Second screen Based on these results, a second, more targeted screening was performed focusing on co-formers that contain at least one carboxylic acid function. 17 co-formers were selected for this second phase of screening: (S)-methyl succinic acid, D-4-chloro mandelic acid, DL-4-chloro mandelic acid, D-2-chloro mandelic acid, DL-2-chloro mandelic acid, (R)-3-chloro mandelic acid, (S)-2-phenyl butyric acid, 2-pyrrolidone- 5-carboxylic acid, L-aspartic acid, (1R,3S)-camphoric acid, (S)-Ibuprofen, L-malic acid, (R,S)-Phenylsuccinic acid, DL-tropic acid, (R)-2-(benzyloxy) propionic acid, (R)-2-(4-hydroxy phenoxy propionic acid and (S)-Naproxen. The results of the screen are given in tables 1-2 to 1-4. In case a positive result was observed, single crystal growth was attempted. 2 co-crystal structures were accordingly obtained. The asymmetric units of those co-crystals are given in figure 1-14. Their crystallographic data are given in Appendix A (tables A-18 & A-19).

110 Chapter 1- Identifying a suitable system

Table 1-2 Results of the second co-crystal screening part 1: CC means co-crystal was proven by single crystal, SCC means that no proof was obtained by single crystal but the PXRD pattern showed the appearance of a new crystalline phase different from all present compounds, including known polymorphs. A grey cells means the experiment showed no new crystalline phase, whether either both compounds or only one were present; or a liquid phase or an amorphous phase was obtained.

R-Methyl D-4-Chloro DL-4-Chloro RS-Phenyl DL-3-Chloro R-2-Phenyl succinic acid mandelic acid mandelic acid succinic acid mandelic acid butyric acid BnClTP CC SCC CC SCC SCC SCC BnFTP SCC SCC SCC SCC BnTP SCC SCC SCC SCC SCC SCC PgTP SCC SCC SCC SCC BnBnOP SCC BnTAP SCC SCC SCC SCC SCC

BnClTAmBn2

BnFTAmBn2 SCC

BnTAmBn2 SCC SCC MBnTA SCC SCC SCC

Table 1-3 Results of the second co-crystal screening part 2.

(+)-Camphoric S-Ibuprofen S-Naproxen D-2-Chloro DL-2-Chloro R-2-(Benzyloxy) acid mandelic acid mandelic acid propionic acid BnClTP SCC BnFTP SCC SCC SCC SCC SCC SCC BnTP SCC SCC SCC SCC PgTP SCC SCC SCC BnBnOP SCC BnTAP

BnClTAmBn2

BnFTAmBn2

BnTAmBn2 MBnTA

111 Chapter 1- Identifying a suitable system

Table 1-4 Results of the second co-crystal screening part 3.

R-2-(4-Hydroxy phenoxy L-Malic RS-2-Pyrrolidone- L-Aspartic DL-Tropic propionic acid acid 5-carboxylic acid acid acid BnClTP SCC BnFTP SCC SCC SCC SCC SCC BnTP SCC SCC SCC PgTP BnBnOP SCC SCC BnTAP SCC

BnClTAmBn2

BnFTAmBn2

BnTAmBn2 MBnTA SCC

Working in such a two-step approach is indeed fruitful, as a hit rate of 36% compared to the 15% hit rate of the first screen was achieved. Second, analogues with very close structural features display similar co-crystallization propensity, which was previously demonstrated35 and can be confirmed once more with this study. Indeed, BnClTP,

BnFTP and BnTP show all a high hit rates while BnClTAmBn2, BnFTAmBn2 and

BnTAmBn2 all show low hit rates. Out of the 10 analogues, only one was identified

for which no positive hit could be given for the second screen, BnClTAmBn2. However, it got one hit during the first screen, likely leading to a racemic co-crystal. Its low success rate is probably due to its high bulkiness, rendering hydrogen bond formation more restricted, or due to the high melting point of the parent compound (illustrative of an already good interaction system in the solid state). Indeed, of the

three amide analogues, the one with the lowest melting point, BnTAmBn2, yields most hits, while all three are very similar structure-wise. A similar observation can be made for the BnTP, BnFTP, BnClTP series. Here again, BnFTP has the lowest melting point and the highest hit rate.

When looking at the two new co-crystal structures determined by SC-XRD (figure 1- 14), the importance of the triazole in the co-crystallization process is again highlighted with the same N---H-O hydrogen bond described above. This can be again correlated with the low success rate of BnBnOP, which lacks the triazole. Finally, BnTAP, the only analog not to give positive results after the first screen gave six possible hits after this screening, showing again the efficiency of carboxylic acids for the analogues considered here.

112 Chapter 1- Identifying a suitable system

Figure 1-14 Asymmetric unit of (R,S)-BnClTP + (R,S)-methyl succinic acid co-crystal (left). Asymmetric unit of (R,S)-BnClTP + (R,S)-4-chloromandelic acid co-crystal (right). Hydrogen bonds are displayed in red.

To conclude, the best options for co-crystallization with this series of analogs is to use carboxylic acid co-formers, combined with an analogue that has a low melting point. The presence of a triazole should improve significantly the success rate of the screening, especially for non-bulky molecules.

3.3.3. Selecting a suitable co-crystal for the development of the CoISD process To develop a co-crystallization induced spontaneous deracemization process, it is necessary to obtain a system in which both enantiomers of the target compound are present in different crystals when an enantiopure co-former is added. Following situations (figure 1-15) can technically occur when adding an enantiopure co-former (S-2) to a racemic target compound (RS-1).

Figure 1-15 The three possible outcomes when formation of a co-crystal occurs between a racemic compound of interest and an enantiopure co-former. 113 Chapter 1- Identifying a suitable system

Our system requires either to be enantiospecific (situation 1) or to form a diastereomeric pair of co-crystals (situation 2). Systems where both target enantiomers link to the co-former in a single crystal (situation 3) are to be discarded. To identify an ideal case, we looked for a single crystal with a chiral space group and only one of the enantiomers in the asymmetric unit. According to table 1-5, only 2 out of the 7 discovered co-crystals were matching these criteria: BnTP + 3-Phenylbutyric acid and BnFTP + 3-phenylbutyric acid.

It appears from this table that the use of dicarboxylic acid co-formers is detrimental as the enantiopure co-former crystallizes with both enantiomers of the target compound in the same crystal lattice, but more data would be required to confirm this behavior.

Table 1-5 Enantiomeric composition in target compound of the asymmetric unit and the nature of the space group for the single crystals obtained from a racemic mixture of target compound and an enantiopure co-former.

Co-crystal Asymmetric unit Space group BnTP + tartaric acid Racemic BnTP Chiral BnClTP + tartaric acid Racemic BnClTP Chiral BnTP + 3-Phenylbutyric acid Enantiopure Chiral BnFTP + 3-Phenylbutyric acid Enantiopure Chiral

BnTAmBn2 + Phenyl lactic acid Enantiopure Centrosymmetric BnClTP + methyl succinic acid Enantiopure Centrosymmetric BnClTP + 4-chloromandelic acid Racemic BnClTP Chiral

To choose between both systems mentioned above, we looked at their melting points as well as the solubility of the enantiopure co-crystal in toluene (Appendix A, figures A-17 & A-18). Table 1-6 and figure 1-16 show a higher melting point for the co- crystal with BnFTP together with a much lower solubility in toluene. For this reason, it was decided to move to the next phase of the development using the BnFTP and 3- Phenylbutyric acid system.

Table 1-6 Melting points of enantiopure co-crystals of BnTP and BnFTP with 3-Phenylbutyric acid.

Co-crystal Melting point (°C) Enthalpy of fusion (J/g) (S)-BnTP + (S)-3-Phenylbutyric acid 97.1 154.8 (S)-BnFTP + (S)-3-Phenylbutyric acid 111.5 111.1

114 Chapter 1- Identifying a suitable system

300

250

200 s [g/L] 150 100

50

0 -5 5 15 25 35 45 T [°C] (S)-BnFTP +( S)-3-Phenylbutyric acid (S)-BnTP + (S)-3-Phenylbutyric acid

Figure 1-16 Solubility curve of the enantiopure co-crystals of S-BnFTP and S-BnTP with S-3- Phenylbutyric acid in toluene.

3.4. Conclusion The synthesis of BnClTP and 9 analogues was performed. These compounds were screened with 35 co-formers yielding two suitable systems for the further development of the CoISD process. Thanks to the development of an appropriate purification process for several reactions, the synthesis of BnFTP could be upscaled to several gram, allowing constitution of a stock of target compound. The screening of the analogues was performed in two steps. A first step showed carboxylic acids to be good co-formers for this class of compounds and the second step confirmed this observation. Of these, a suitable system was identified for further development. Moving on to the next stage, the diastereomeric co-crystal between BnFTP and 3- Phenylbutyric acid will be used as the model system for the development of the CoISD process through chiral resolution and racemization.

4. Bibliography 1. Viedma, C. Experimental evidence of chiral symmetry breaking in crystallization from primary nucleation. J. Cryst. Growth 261, 118–121 (2004). 2. Steendam, R. E., Meekes, H., Vlieg, E. & Rutjes, F. P. J. T. Viedma ripening: a reliable crystallisation method to reach single chirality. Chem. Soc. Rev. 44, 6723–6732 (2015). 3. Spix, L., Alfring, A., Meekes, H., Enckevort, W. J. P. Van & Vlieg, E. Formation of a Salt Enables Complete Deracemization of a Racemic Compound through Viedma Ripening. Cryst. Growth Des. 14, 1744−1748 (2014). 4. Levilain, G., Rougeot, C., Guillen, F., Plaquevent, J. & Coquerel, G. Attrition- enhanced preferential crystallization combined with racemization leading to redissolution of the antipode nuclei. Tetrahedron: Asymmetry. 20, 2769–2771 (2009). 115 Chapter 1- Identifying a suitable system

5. Sugavanam, B. & Shepard, M. C. Fungicical Compounds. US4243405 (1981). 6. Ding, M. W. & Yuan, D. Triazole acetylene compound and application thereof. CN105820129A (2016). 7. Sugavanam, B. Diastereoisomers and Enantiomers of Paclobutrazol: Their Preparation and Biological Activity. Pestic. Sci. 15, 296–302 (1984). 8. Dholwani, K., Saluja, A., Gupta, A. & Shah, D. A review on plant-derived natural products and their analogs with anti-tumor activity. Indian J. Pharmacol. 40, 49–58 (2008). 9. Lange, J. H. M. & Kruse, C. G. Medicinal chemistry strategies to CB1 cannabinoid receptor antagonists. Drug Discov. Today. 10, 693–702 (2005). 10. Yogeeswari, P., Ragavendran, J. V. & Sriram, D. An update on GABA analogs for CNS drug discovery. Recent Pat. CNS Drug Discov. 1, 113–118 (2006). 11. Horikawa, R., Fujimoto, C., Yazaki, R. & Ohshima, T. µ-Oxo-Dinuclear-(III)- Catalyzed O-Selective Acylation of Aliphatic and Aromatic Amino Alcohols and Transesterification of Tertiary Alcohols. Chem. Eur. J. 22, 12278–12281 (2016). 12. Craig, D. 2.O1 Organic Synthesis Lecture 3. https://www.ch.ic.ac.uk/local/organic/tutorial/OrganicSynthesis3.pdf (2004). 13. Rathke, M. W. & Sullivan, D. F. O-Silylation and Attempted O-Alkylation of Lithium. Synth. Commun. 3, 67–72 (1973). 14. Heiszwolf, G. J. & Kloosterziel, H. Alkylation of enolate anions formation of enol ethers. Recl. Trav. Chim. Pays-Bas. 89, 1153–1169 (1970). 15. Damoun, S. et al. Influence of Alkylating Reagent Softness on the Regioselectivity in Enolate Ion Alkylation : A Theoretical Local Hard and Soft Acids and Bases Study. J. Phys. Chem. A. 103, 7861–7866 (1999). 16. In, J.-K., Lee, M.-S., Lee, M.-W. et al. Stereoselective synthesis of (E)- and (Z)-enol ethers from β-amino aldehydes. Arch Pharm Res 30, 695–700 (2007). 17. Danagulyan, G. G. et al. Synthesis of N - and C -azolyl-substituted pyrazolo [1 ,5-a] pyrimidines by recyclization of pyrimidinium salts. Chem. Heterocycl. Compd. 51, 483–490 (2015). 18. Vishweshwar, P., Mcmahon, J. A., Bis, J. A. & Zaworotko, M. J. Pharmaceutical Co- Crystals. J. Pharm. Sci. 95, 499–516 (2006). 19. Lehn, J.-M. Perspectives in supramolecular chemistry: From molecular recognition towards self-organisation. Pure Appl. Chem. 66, 1961–1966 (1994). 20. Sekhon, B. S. Pharmaceutical co-crystals - a review. Ars Pharm. 50, 99–117 (2009). 21. Wood, P. A. et al. Knowledge-based approaches to co-crystal design. CrystEngComm. 16, 5839–5848 (2014). 22. Norberg, B., Robeyns, K., Wouters, J. & Leyssens, T. Advances in Pharmaceutical Co-crystal Screening : Effective Co- crystal Screening through Structural Resemblance. Cryst. Growth Des. 12, 475–484 (2012). 23. Mirza, S., Miroshnyk, I. & Yliruusi, J. Co-crystals: An emerging approach for enhancing properties of pharmaceutical solids. Dosis. 24, 90–96 (2008). 24. Chadwick, K. & Cross, W. How does grinding produce co-crystals ? Insights from the case of benzophenone and diphenylamine. CrystEngComm. 9, 732–734 (2007). 25. Shan, N., Toda, F. & Jones, W. Mechanochemistry and co-crystal formation : effect of solvent on reaction kinetics. Chem. Commun. 2372–2373 (2002). 116 Chapter 1- Identifying a suitable system

26. Fucke, K., Myz, S. A., Shakhtshneider, T. P., Boldyreva, E. V & Griesser, U. J. How good are the crystallisation methods for co-crystals ? A comparative study of piroxicam. New J. Chem. 36, 1969–1977 (2012). 27. Fris, T. & Jones, W. Recent Advances in Understanding the Mechanism of Cocrystal Formation via Grinding. Cryst. Growth Des. 9, 1621–1637 (2009). 28. Braga, D., Maini, L. & Grepioni, F. Mechanochemical preparation of co-crystals. Chem. Soc. Rev. 42, 7638–7648 (2013). 29. Leyssens, T., Springuel, G., Candoni, N. & Veesler, S. A thermodynamically guided approach to co-crystal screening: application to maleic acid / caffeine system. Int. Symp. Ind. Cryst. (2011). 30. Hickey, M. B. et al. Performance comparison of a co-crystal of carbamazepine with marketed product. Eur. J. Pharm. Biopharm. 67, 112–119 (2007). 31. Editors, G., James, S., Queen, T. & Nangia, U. K. A. Mechanochemistry and cocrystals forms. CrystEngComm. 11, 412–414 (2009). 32. Chiarella, R. A., Davey, R. J. & Peterson, M. L. Making Co-CrystalssThe Utility of Ternary Phase Diagrams. Cryst. Growth Des. 7, 1223–1226 (2007). 33. Fahrney, D. Hydrogen Bond Lengths and Angles. in The Fundamentals of Biochemistry: Interactive Tutorials (eds. Robinson, B. & Hansen, J.) (Colorado State University). 34. Jeffrey, G. A. An introduction to hydrogen bonding. (Oxford University Press, 1997). 35. Norberg, B., Robeyns, K., Wouters, J. & Leyssens, T. Advances in Pharmaceutical Co-crystal Screening : Effective Co-crystal Screening through Structural Resemblance Ge r. Cryst. Growth Des. 12, 475–484 (2012). 117

Chapter 2- The road to the development of the CoISD process 118 Chapter 2- Development of the CoISD process: A proof of concept

1. Overview This chapter will detail the stages that led to the development of the CoISD process. In the first section, an ideal solvent needs to be identified in order to have good crystallization and efficient selectivity between the diastereomers. Then, the ternary phase diagram of the enantiopure co-crystal needs to be constructed. This led to the development of the first step of the CoISD process, the chiral resolution. To conclude this first section, the separation of the co-crystal was successfully carried out in order to retrieve enantiopure BnFTP after resolution, for the second step in the development of the CoSID: the racemization. Thus, the second section focuses on the identification of a suitable racemizing agent for the system followed by the study of the racemization kinetics of BnFTP in presence of the chosen racemizing agent. The racemization rates in solution as a function of solvent and racemizing agent concentration were determined by measuring the decreasing enrichment of the chiral ketone due to racemization over time, using a polarimeter set-up with a continuous recycling loop through the polarimeter cell. The established racemization kinetics model aligns with the experimental data giving access to the intrinsic racemization rate constant. The proposed mechanism is first order with respect to the enantiomeric excess of the target compound and the racemizing agent concentration. The solvent is shown to strongly affect the racemization constant, with protic solvents increasing this rate substantially due to hydrogen-bond stabilization of the enolate. Then, the kinetics were studied with the additional presence of the co-former. It was observed that the presence of the chiral acid co-former alters the reaction mechanism albeit remaining first order with respect to the enantiomeric excess. Though more complex, the mechanism still followed Arrhenius law, providing key information on the impact of temperature.

2. Chiral resolution: Finding a suitable solvent and method to break down the co-crystal 2.1. Introduction The following step in the development of the CoISD process was to choose which solvent to use with our system. Indeed, as any crystallization-based process, the choice of the solvent can determine the very success or failure of the CoISD process. As a result, choosing a solvent is critical for the crystallization part of the process. The solvent has an impact on the shape of the produced crystals and the solubility of the compound to crystallize (and hence the yield of the process).1 Because of that, when developing batch crystallization, finding an optimal solvent is one of the main 119 Chapter 2- Development of the CoISD process: A proof of concept challenges.2 Furthermore, when considering co-crystals, another factor to take into account is congruency, the capacity of a co-crystal to be the only thermodynamically stable phase in suspension when starting from the co-crystal’s stoichiometry.3 Finally, in the case of a chiral resolution by crystallization, the solvent also impacts the difference of solubility of the two diastereomers as it was shown in the case of a pair of diastereomeric salts:

푐p ∆∆퐺solv = 2푅푇푙푛( ) 푐n

With ∆∆퐺solv the difference in Gibbs free energy of solvation for a pair of diastereomeric salts, cp the solubility of the p-diastereomeric salt in the solvent and cn the solubility of the n-diastereomeric salt in the solvent. By extension to diastereomeric co-crystals, the solvent is expected to affect the solubility difference.

Once a suitable solvent is identified, a co-crystal system is typically characterized by drawing its ternary phase diagram in that solvent. Then, for any chiral resolution, the right conditions of concentration in the solvent must be identified in order to have only one diastereomer precipitating out. Finally, the right conditions for breaking the co-crystal must be identified in order to readily retrieve the resolved target compound.

2.2. Materials & methods (R,S)-BnFTP was synthesized from commercially available chloropinacolone (purchased from VWR, ≥95%), 1H-1,2,4-triazole (purchased from Carl Roth, ≥99%), 4-fluorobenzyl chloride (purchased from Fluorochem) and NaH (60% suspension in mineral oil, purchased from Acros) with a 2-reaction synthesis as specified in chapter 1 section 2.2.2. All the solvents used were commercially available and of analytical grade. Finally, the acid added to the system, (S)-3-phenylbutyric acid, was commercially purchased as racemic from Sigma Aldrich (98%) and then resolved by using either (S)- or (R)-phenylethylamine,5 purchased from VWR (≥98%).

2.2.1. Preparation of (S)-3-phenylbutyric acid (S)-3-phenylbutyric acid was obtained from (R,S)-3-phenylbutyric acid (Sigma- Aldrich) carrying out a diastereomeric salt resolution with the following protocol:

A 175g/L solution of (R,S)-3-phenylbutyric acid and (S)-1-phenylethylamine (1 equivalent of each, mPhenylbutyric acid = 20.13g; mPhenylethylamine = 14.86g) was stirred in 200mL of a 1:3 ethanol/toluene solvent mixture overnight at room temperature and yielded post filtration 13.24g of (S-S)-salt with an enantiomeric excess of 0.41. This salt is then stirred overnight in 90mL of a 1:3 ethanol/toluene solvent mixture at room temperature and filtrated to yield 7.64g of (S-S)-salt with an enantiomeric excess of 0.91. This salt is again stirred overnight at room temperature in 35mL of a 1:3 ethanol/toluene solvent mixture and filtrated to yield 6.23g of (S-S)-salt with an enantiomeric excess of 0.98 (yield = 18%). 120 Chapter 2- Development of the CoISD process: A proof of concept

At each crystallization, the cake enantiopurity was analyzed by polarimetry with an Anton Paar polarimeter at 589nm and 20°C in order to obtain the enantiomeric excess of the salt. Each calculation was done according to the following reasoning:

Phenylbutyric acid was shortened as PBA and Phenylethylamine as PEA. The “+” stand for the protonated PEA while the “−“ stands for the deprotonated PBA.

PEA+ MPEA+ = 122,18g/mol xm = 0.428

PBA− MPBA− = 163,2g/mol xm = 0.572

100훼 = [훼]SPEA+푙푐m푥mPEA+ + [훼]SPBA-푙푐m푥mSPBA-퐸

For an equimolar mixture of PEA+ et PBA− with a mass concentration of cm ∈ [10 ;10,07g/L], [α]SPEA+ was determinated in ethanol with an equimolar mixture of (S)- PEA and racemic PBA (E=0):

SPEA+ α = -0,068 for cm = 10,045g/L gives [α] = -15.81°

[α]SPBA- was determinated in ethanol with an equimolar mixture of (S)-PEA and (S)- PBA (E=1):

SPBA- α = 0,091 for cm = 10,07g/L gives [α] = 27.64°

The enantiomeric excess of each (S)-PEA sample was calculated with the following formula:

100훼 − [훼]SPEA+푙푐m푥mPEA+ 퐸 = [훼]SPBA-푙푐m푥mSPBA-

The enantiomeric excess of each (R)-PEA sample was calculated with the following formula:

100훼 + [훼]SPEA+푙푐m푥mPEA+ 퐸 = − [훼]SPBA-푙푐m푥mSPBA-

All samples concentrations were comprised between 10g/L and 10,07g/L and all samples were an equimolar mixture of PEA and PBA.

2.2.2. Preparation of cocrystal (S)-BnFTP-(S)-PBA (S)-BnFTP was resolved from (R,S)-BnFTP by co-crystallization with (S)-3- phenylbutyric acid yielding the (S)-BnFTP-(S)-PBA co-crystal. The protocol is the following: In a vial, to 1g of (R,S)-BnFTP, 8 mL of Toluene was added. Then, 0.6g of (S)-3-Phenylbutyric acid (1 eq) was added and the mixture was stirred overnight with a magnetic stirrer at room temperature. A white solid in suspension was obtained, filtered over Buchner and washed with toluene. The solid was dried at 50°C for 2h or 121 Chapter 2- Development of the CoISD process: A proof of concept at room temperature overnight. The cake was analyzed by chiral HPLC to determine its enantiomeric ratio. As a typical example, an amount of 516.1mg of a white solid (yield of 32%) can be obtained with an enantiomeric excess of 99.3%.

2.2.3. Preparation of (S)-BnFTP (S)-BnFTP was obtained by breaking the enantiopure co-crystal of (S)-BnFTP and (S)-PBA. The protocol was the following: For 1g of co-crystal, 15mL of ethyl acetate were added to achieve full dissolution at room temperature. In parallel, 1.1 equivalent of NaOH was dissolved in 20mL of water. Two successive extractions were performed using 10mL of the NaOH solution. The organic phase was then dried over

MgSO4, and evaporated under vacuum. Enantiopure (S)-BnFTP was obtained with a 90% yield.

2.2.4. Choosing the solvent Protocol for solubility curves: cf. Chapter 1 section 3.2.4.

Protocol congruency test: About 100mg of the (S-S)-co-crystal (ee>99%) were weighed in a 2mL vial. Then, each vial was equipped with a magnetic stirrer. Then, solvent was added until a suspension was obtained. The vials were left to equilibrate under magnetic stirring over 2 days. The powder obtained was analyzed by PXRD and compared to the pattern of the co-crystal and both compounds alone.

Protocol for the study of counter solvent for the chiral resolution: About 500mg of (R,S)-BnFTP and 300mg of (S)-phenylbutyric acid were weighed in three different vials. For each vial a different solvent mixture and a different volume were used. The slurries were stirred overnight, then filtered and washed with 500μL of toluene. The solids were dried in the oven at 50°C for 2 hours. The solids were then analyzed by polarimetry to get the ratio of the (S)-enantiomer. The enantiomeric excess was calculated with this value.

2.2.5. Ternary phase diagram The phase diagram curve was determined using a reverse phase HPLC. The set-up is the following: the pump and injection system is a Waters 2695 and the detector is a Waters PDA 2998. The column is situated in an oven at a temperature of 40°C. The column used is a Waters Sunfire C18 column with the following dimensions 150x4.6mm and a particle diameter of 3.5μm. The chromatograms were obtained from a detection at 210 nm. The mobile phase was starting from a 1:1 mixture of water and acetonitrile with 0.1% of formic acid until gradually reaching a 10:90 water/acetonitrile ratio with always 0.1% of formic acid in 6.5min. The flow rate was 1.5 mL/min.

Protocol for the ternary phase diagram: 122 Chapter 2- Development of the CoISD process: A proof of concept

With HPLC: Different ratio of (S)-BnFTP and (S)-3-phenylbutyric acid were made by weighing with precision both compounds in a 2mL vial. For most of the ratios, 750μL of toluene was added except for the 0.4; 0.5 and 0.6 ratio where 1mL of toluene was added. A cap was put on every vial and each of them were put in a shaker at 25°C and 700rpm for 72h. After 3h30, each vial was seeded with a mixture of the (S-S)-co- crystal and (S)-BnFTP. (S)-3-phenylbutyric acid is liquid at room temperature, thus there was no need for seeding with it. After 72h, each vial was left to rest and then the liquid was taken with a syringe, filtrated with a micro filter to remove the traces of solid still in suspension while the solid was rapidly dried on a filtration paper. Each collected liquid was weighed and then 50μL were taken using a micropipette and the remaining liquid was weighed again to know the mass of those 50μL. The 50μL were diluted in 25mL volumetric flask (1/500 dilution). Those solutions were analyzed by HPLC while the dried solids were analyzed by PXRD.

With mass weighing: A specific mass of (S)-BnFTP and (S)-PBA is weighed in a tared vial, whose mass is known with the lid. Each mass is noted down. A magnetic stirrer is added and the mass of the total is noted down (starting mass). Then, toluene is added and the mixture is stirred at 25°C at 700rpm for at least 1h. This is repeated until all solid has dissolved after the last addition of toluene. Then, the mass of toluene added is obtained by weighing the mass of the vial and subtracting the starting mass.

2.2.6. Breaking the co-crystal Enantiomeric purities were again determined using chiral HPLC (cf. Chapter 1 section 3.2.4). Column chromatography was performed using neutral alumina silica gel 60 Å (40-63 μm). Optical rotations were obtained at a wavelength of 589 nm with an Anton Paar Polarimeter. All measurements were done at 20°C. The length of the cell is 100mm. The concentration “c” has units of g/100 mL. Some enantiomeric purities were determined with this polarimeter by measuring the optical rotation and calculating the enantiomeric ratio of the (S)-enantiomer.

Protocol of the base study for the liquid-liquid separation of the co-crystal: The (S-S)-co-crystal of BnFTP and 3-phenylbutyric acid is dissolved in solvent 2 in a vial. Then water with an excess of base was added. The vial is closed and shaken for 1 minute, three times, while, in-between opening the vial to prevent high pressure build- up. Then, the vial was let to rest. Once both phases had separated, the upper phase (organic) was removed and let to dry. The solid was analyzed by chiral HPLC.

2.3. Results & discussion 2.3.1. Choosing the solvent To efficiently screen for a suitable solvent for the chiral resolution of (R,S)-BnFTP with (S)-PBA, the solubility curves of 8 different solvents, chosen for their different 123 Chapter 2- Development of the CoISD process: A proof of concept relative polarity and their different functional groups were measured by turbidity with the Crystal16 device: acetonitrile, ethyl acetate, tetrahydrofuran (THF), acetone, ethanol, 1-propanol, 1-butanol and toluene. Congruency of the co-crystal in all those solvents was verified prior to the solubility curve collection (Appendix C, figure C- 1). Solubility curves are given in figure 2-1. Van’t Hoff plots (Appendix C, figure C- 2) allowed extracting enthalpy (H°) and entropy of crystallization (S°) shown in table 2-1.

600

500

400

s [g/L]300

200

100

0 0 10 20 T [°C] 30 40 50 Toluene THF Butan-1-ol Acetone Acetonitrile Ethanol Propanol Ethyl Acetate Figure 2-1 Evolution of (S-S)-co-crystal solubility with temperature in 8 different solvents.

Table 2-1 Experimental values of Hdiss° and Sdiss° of the (S-S)-cocrystal for each studied solvent alongside its solubility at 10°C.

ΔHdiss° ΔSdiss° s(10°C) Solvent [KJ/mol] [J/mol.K] [g/L] Toluene 90.8 259 10.9 THF 44.8 145 200.4 Ethyl Acetate 62.5 186 54,1 Butanol 80.2 233 21.3 Acetone 59.8 194 97.3 Acetonitrile 82.5 250 36.6 Ethanol 71.4 211 37.3 Propanol 78.9 232 26.6 124 Chapter 2- Development of the CoISD process: A proof of concept

Entropy of dissolution is comparable for all solvents, with difference in solubility mainly due to the enthalpic part (representative of intermolecular interactions). As expected, dissolution of the cocrystal is endothermic, with acetone and THF showing the lowest enthalpy of dissolution, which is possibly explained by a stabilizing interaction of these solvents for both BnFTP as well as PBA, resulting in a higher solubility at 10°C. Alcohols mainly act on the acid through hydrogen bonding interaction, all showing a similar dissolution enthalpy. Finally, toluene showed the highest enthalpy of dissolution and lowest associated solubility. Using high solubility solvents will be detrimental for the yield, and hence, toluene and alternatively an alcohol solvent are advisable for the resolution. Finally, toluene was chosen as the most suitable solvent for the next steps in the development of the process since it gave the lowest solubility.

Mixtures of toluene and hexane The amount of solvent used for the process is critical for the whole process. If not sufficient, filtration becomes problematic and if too much, the yield is significantly reduced. Using a counter-solvent became an interesting idea to allow larger volumes of solvent keeping a similar yield. Hexane was chosen as a counter solvent because it is miscible with toluene and is one of the most apolar solvents. The co-crystal was expected to be almost insoluble in this solvent. The results and conditions of this experiment are given in table 2-2.

Table 2-2 The enantiomeric excess of the co-crystal in suspension and its mass recovered for varying rations of toluene and hexane starting from 500mg of (R,S)-BnFTP and 300mg of (S)- PBA.

Tol/Hex ratio VToluene VHexane mCo-crystal E (polarimetry) 1 : 0 2mL 0mL 294mg 0.9317 2 : 1 4mL 2mL 277mg 0.8748 4 : 3 4mL 3mL 289mg 0.8658 As expected, hexane decreased the solubility of the co-crystal. Indeed, 7mL of a 4:3 mixture of toluene/hexane gave a similar mass of cake than 2mL of pure toluene. However, both diastereomers crystallized out as shown by the reduced E. It can be expected that the addition of hexane lead to a reduction of solubility of both diastereomers. Consequently, the use of hexane as a counter solvent was efficient for increasing the volume of solvent without decreasing the yield. However, the price to pay was the decrease in selectivity of the chiral resolution process. Regardless, for the deracemization process, this is irrelevant, as in the end the undesired enantiomer in solution transforms into the desired one, the quantity of undesired enantiomer will decrease and induce the dissolution of the excess that was in suspension. It was however not be used at the end because of incompatibility between hexane and the 125 Chapter 2- Development of the CoISD process: A proof of concept developeda process but using a co-solvent with low polarity remains an interesting option to reduce the solubility, especially when the co-crystal is quite soluble.

2.3.2. Ternary phase diagram The ternary phase diagram of the (S-S)-co-crystal of BnFTP and 3-phenylbutyric acid in toluene was constructed using reverse phase HPLC for some points and mass weighing for others (Appendix C, calibration curves for HPLC in figure C-3, Chromatogram in figure C-4). Powder X-rays Diffraction was used for identification of the different zones by analysis of the solid-phase (figure 2-2).

Figure 2-2 (left) Isothermal cut (25°C) of a ternary phase diagram between (S)-BnFTP, S-3- phenylbutyric acid ((S)-PBA) and toluene. (Right) Zoom on the upper part of the isothermal cut of a ternary diagram. The table of points from which this diagram is drawn is given in appendix C, table C-3.

This diagram is congruent as expected, showing a large zone for pure co-crystal. This is partly due to the fact that (S)-3-phenylbutyric acid is liquid at room temperature. The zone where (S)-BnFTP precipitates alone is very narrow due to the low solubility of the co-crystal in toluene. Thus, as soon as there is enough (S)-3-phenylbutyric acid, the co-crystal precipitates with (S)-BnFTP. This makes this system very suitable for an efficient crystallization and resolution process.

The composition stability zones of the diagram were determined by PXRD of the solid in suspension, with the composition of the supernatant determined by HPLC. The key diffractograms are given in figure 2-3. Though difficult to see, for the blue diffractogram, peaks of enantiopure BnFTP can be found with relative low intensity. Low intensity of the peaks is suspected to be linked to bad crystallinity of BnFTP compared to the co-crystal.

a Latter, heating will need to be implemented and hexane was too low-boiling. 126 Chapter 2- Development of the CoISD process: A proof of concept

Figure 2-3 Diffractograms of the solid phase for the starting mixtures (10% PBA / 90% BnFTP) in blue and (40%/60%) in red. In green is shown the diffractogram of pure co-crystal and in pink of enantiopure BnFTP.

2.3.3. Diastereomeric system The system composed of racemic BnFTP and (S)-PBA was shown to be diastereomeric with existence of the other diastereomer, the (R-S)-co-crystal, made of (R)-BnFTP and (S)-PBA.

300

250

200 s [g/L] 150 100

50

0 -5 5 15 25 35 45 T [°C] (S)-BnFTP + (S)-3-Phenylbutyric acid (R)-BnFTP + (S)-3-Phenylbutyric acid

Figure 2-4 Solubility curves of (S)-BnFTP + (S)-PBA co-crystal and (R)-BnFTP + (S)-PBA co- crystal in toluene.

As a result, toluene’s capacity to separate both diastereomers by crystallization was assessed by comparing the solubility curves of each diastereomer (the (S-S)-co-crystal 127 Chapter 2- Development of the CoISD process: A proof of concept and the (R-S)-co-crystal) in toluene. As shown by figure 2-4, the difference in solubility is quite high (Δs = 256g/L at 20°C) and increases with temperature. At low temperature, this difference is consequently decreased but remains good enough for separation (Δs = 77g/L at 0°C).

2.3.4. Resolution in toluene Starting from racemic BnFTP and (S)-PBA, resolution was easily achieved by crystallization in toluene. The condition of starting concentration in BnFTP were fine- tuned in order to get enantiopure co-crystal. These conditions are described in section 2.2.2. Knowing the (S-S)-co-crystal is the one precipitating in the conditions of the resolution when (S)-PBA is used, we were able to attribute the HPLC peak belonging to the (S)-enantiomer of the target compound as the second one.

2.3.5. Breaking the co-crystal Finally, the aim of a deracemization process is to obtain enantiopure compound of interest. However, when the process was finished, the enantiopure compound is still under the form of a co-crystal. Thus, the last step left for us was to break the co-crystal in order to separate the compound of interest, (S)-BnFTP here.

The difficulty in doing so lies in the fact that both compounds have similar behavior in various solvents. The acid is liquid and soluble/miscible in almost every organic solvent, in which BnFTP is also soluble. The acid as well as BnFTP are insoluble in water. Moreover, the co-crystal behaves congruently in almost all solvents. However, the co-crystal relies on one feature that can be toyed with, the hydrogen bond between the hydrogen of the acid and the nitrogen of the triazole. Indeed, by removing the acidic hydrogen of the acid, the co-crystal would fall apart and the salt of the acid would become very soluble in water (and almost insoluble in organic solvents) while BnFTP would keep the same behavior as before. The only issue is that under basic condition, BnFTP racemizes. As a result, the amount of base used to deprotonate the acid should be carefully controlled. Alternatively, chromatography can be performed with no risk of racemization. Both methods were tried and the results are given below. All chromatograms from chiral HPLC for this part are given in appendix C figure C- 5.

By chromatography column: The co-crystal obtained from the chiral resolution experiment was used to do this experiment. 700mg of (S-S)-co-crystal was put inside the column with a starting enantiomeric ratio of 5.7/94.3 (R/S). The first step, before starting the column, was to find the most suitable eluent to do the separation. At best, this eluent should be able to give completely separated spots in thin layer chromatography (TLC). For this purpose, three different eluents were tried: 128 Chapter 2- Development of the CoISD process: A proof of concept

b A. 50:50 EA/ PE  RfBnFTP = 0.65 / Rfco-former = 0.74 / ΔRf = 0.09 B. 40:60 EA/Toluene  RfBnFTP = 0.44 / Rfco-former = 0.58 / ΔRf = 0.14 C. 30:70 EA/Toluene  RfBnFTP = 0.34 / Rfco-former = 0.54 / ΔRf = 0.20  C Chosen as final gradient

The first eluent tried was a combination of ethyl acetate and petroleum ether with a 1:1 volume ratio, which is a basic eluent, also known to elute BnFTP alone. The acid was less retained but the spots were too close to each other. Then, petroleum ether was switched with toluene in order to help the separation since both compounds showed aromatic features. Moreover, the polarity was decreased because of the high values of retardation factors obtained with the first eluent. The separation was better but still not satisfactory. Consequently, the polarity was again decreased. The separation between the two compounds with the third solvent was high enough to do a column without risking a mixture of both compounds coming out.

Now that the eluent was chosen, it was still too polar to use it as such. So to start the column it was diluted with petroleum ether. The different ratio of the third eluent and petroleum ether used were the following:

Packing of the column + first eluent till the first one starts coming out = 2:1 PE/C Second eluent till the first compound stop coming out = 1:2 PE/C Third and last eluent = C

This method enabled the total separation of both compounds without traces of one mixed with the other. The enantiomeric ratio of BnFTP in the (S-S)-co-crystal used for this experiment was 87.2%. The masses obtained after evaporation of the solvent are:

20 M(S)-BnFTP = 370 mg (n = 1.34 mmol) / HPLC: eeBnFTP = 91.8% / [α]D = - 13.00° / Yield = 84% M(S)-3-phenylbutyric acid= 195 mg (n = 1.19 mmol) / Yield = 75%

The yield was acceptable and the enantiomeric excess obtained was even higher than before. This is not surprising since in the case of an enantio-enriched mixture, non- linear interactions can occur yielding fractions with different E.6 Since some compound got lost in the column, because the yield is not 100%, there must have been more (R)-BnFTP than (S)-BnFTP in that fraction.

By liquid-liquid separation: Before doing the actual separation, the condition of basicity needed to be studied. The solvents used for the study are the following:

b Ethyl acetate = EA; Petroleum ether = PE 129 Chapter 2- Development of the CoISD process: A proof of concept

Solvent 1 = Water with a base (NaOH) Solvent 2 = EA or Toluene

- The point was to investigate whether HO (aq) could deprotonate the acid of the co- former while avoiding racemization.

Starting mixture: 6.4/93.6 EA/HO- = HPLC: 15/85 + No more co-former Toluene/HO- = HPLC: 10.8/89.2 + No more co-former

The separation was effective, showing that basic water could deprotonate the acid. However, racemization occurred because of the basic conditions. By carefully controlling the base quantity to a 1:1 ratio, racemization could be prevented from occurring. Furthermore, toluene seemed to perform better than ethyl acetate because it induced less racemization, probably because of its lower miscibility with water. The next test was done at two different pH values, but always one equivalent of base, to see if the volume of the aqueous phase impacted racemization. The protocol remained the same and the same starting mixture was used (6.4/93.6).

Toluene / HO- (pH = 13 / ratio 1:1) = 6.2/93.8 + no more co-former Toluene / HO- (pH = 14 / / ratio 1:1) = 7.9/92.1 + no more co-former

The separation was still effective but this time no racemization occurred for a pH of 13. A pH of 14 induced little racemization. The more concentrated the aqueous phase is, the higher the chance to induce racemization. Therefore, to ensure that no racemization occurs during the separation process, the base needs to be diluted. A pH of around 10 would ensure that no racemization occurs.

In conclusion, for the rest of the work, splitting the co-crystal will be done by liquid- liquid extraction because the process is efficient and quite fast compared to chromatography. Moreover, scaling up this process would be less expensive and more practical. This process was used with bigger quantities of co-crystal and no racemization was observed. However, using exactly one equivalent, there was still some acid left after separation. Consequently, to ensure total separation 1.1 equivalent is recommended. Doing so, we managed to remove all the acid and at the same time avoid racemization.

Conclusion These studies are carried out upstream to the development of the CoISD process but are of utmost importance as they lay the basis for the model process. For any other system, those steps have to be repeated in order to determine the solvent of crystallization and ensure retrieval of the wanted enantiopure compound. However, this crystallization solvent has to also be fitting with the racemization reaction that has to be studied co-jointly and is detailed below. 130 Chapter 2- Development of the CoISD process: A proof of concept

3. Racemization: Finding a suitable racemizing agent and kinetic study of the racemization reaction 3.1. Introduction In this section, we focus specifically on the racemization reaction. Indeed, in order to have the deracemization occurring, the chiral resolution must be combined to a racemization reaction in solution. For the compound BnFTP, the chiral carbon can be racemized through enolization, which produces the instable enol(ate) function. This reaction needs to be either base or acid catalyzed. Consequently, the first step of this stage was to screen for the type of catalyst to use. Once a suitable racemizing agent was found, the kinetics of racemization were then studied since knowledge of the racemization rate and kinetics is crucial in developing an efficient deracemization process. Indeed, it was shown by R. Oketani et al.7 that, for their system, productivity was directly correlated to the product of racemization rate k and solubility; with k governing the deracemization time. Over the years, there have been an extensive number of kinetic studies on the keto-enol equilibrium mostly on acetophenone or acetyl heterocycle derivatives in water8-15 with only a few studies on pinacolone derivatives like BnFTP16,17 and an even more limited number in organic solvents.18 Every kinetic study carried out on the enolization of carbonyls was assessed by either deuteration, racemization or halogenation,10 the latter one being the most used.8-15,19- 21 Here we use the racemization approach, conducted with a polarimeter set-up that continuously samples the racemizing mixture to the polarimeter cell, using a recycling loop connected to the reactor in order to measure directly the enantiomeric excess of the compound during the racemization reaction. Such a set-up allows for collection of a data point every 5 seconds, leading to a substantial amount of data, particularly useful for rapid racemization processes. Such a method is quite uncommon for these types of studies but was recently shown to produce easily accessible and high-quality data.22,23

This section will first show how diazabicyclo[5.4.0]undec-7-ene (DBU) was chosen as the racemizing agent. Then, the kinetics of racemization of BnFTP in toluene and in presence of DBU are studied. The experimental data are compared to a kinetic model in order to determine the racemization rate. Then, the influence of the solvent on the racemization rate is studied. Finally, the impact of the presence of an acid, the co-former, to the system is assessed. 131 Chapter 2- Development of the CoISD process: A proof of concept

3.2. Material and methods

3.2.1. Material The base 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU), was purchased from Sigma Aldrich (≥99%) and used as such. All co-crystal and enantiopure (or enantioenriched) BnFTP were obtained as described in sections 2.2.2 & 2.2.3.

3.2.2. Racemizing agent screening Enantiomeric purities were determined as enantiomeric ratios (er) using chiral HPLC (cf. Chapter 1 section 3.2.4). Chromatograms are given in appendix C, figures C-6 and C-7.

Protocol for the acid and base screening: 10mg of the (S-S)-co-crystal with an enantiomeric ratio of 6.4/93.6c were weighed in vial, five times. All catalysts (except DMAP) were used as solvent (100 eq mol). DMAP, which is solid, was used in toluene at a ratio of 1 equivalent compared to the quantity of BnFTP contained in the co- crystal. The mixtures were agitated at 30°C for 24h. The quenching was performed by adding water (acidic water if a base was used). Then, toluene was added and the mixture was transferred to a bigger vial. Then, the vial was closed and agitated strongly for 1 minute, three times, while opening the vial in between to avoid high pressure. Then, the vial was let to rest until both phases were completely separated. The organic phase was removed and left to dry overnight. They were then analyzed by chiral HPLC.

Protocol for the study of DBU capacity to racemize BnFTP: About 26mg of enantiopure (S)-BnFTP (4.06/95.94) were weighed in a vial. About 5mL of toluene was. For the DBU, 14.25μL were taken with a micropipette for the one equivalent vial and 0.72μL for the 5% vial. The vials were stirred with a magnetic stirrer and samples were taken after 30 minutes, one hour and six hours of reaction. Each sample was treated with a slightly acidic water solution, to remove DBU, and extracted with ethyl acetate. The solvent was removed and the sample was analyzed by chiral HPLC.

3.2.3. Polarimetric analysis The polarimeter used for all the measurement was a Rudolph Autopol IV automatic polarimeter, used in its continuous mode with a 5-second measurement interval. All the measurements were carried out at a wavelength of 589nm –since this wavelength gave the highest rotary power out of the 4 available. The temperature of the cell was measured continuously using a temperature probe. During every experiment, the evolution of the optical rotation α was measured and linked to the evolution of the enantiomeric excess E using Biot’s law.

c This co-crystal mixture was one obtained . 132 Chapter 2- Development of the CoISD process: A proof of concept

Racemization in absence of the co-former Biot’s law gives the expression of the optical rotation α. In our case, we consider two enantiomers of molecule whose concentration is the product of the molecule’s concentration cm by the weight ratio of the enantiomer with respect to the other one xS and xR:

100훼 = [훼]푙푆0(푥S − 푥R) (1)

With α the optical rotation of the sample in °, [α] the specific rotation of (S)-BnFTP in the solvent of the experiment, l the path length in dm, S0 the concentration of BnFTP in g/100mL xs and xR the enantiomer weight fraction of respectively (S)- and (R)-

BnFTP so that 푥S + 푥R = 1

By definition, the enantiomeric excess E of a substance of (S)-configuration is the difference between the weight fraction of the (S)-enantiomer and that of the (R)- enantiomer, divided by their sum.

푥S − 푥R 퐸 = (2) 푥S + 푥R Equation (3) can then be obtained from (1) and (2), putting in relation the enantiomeric excess of a solution of two enantiomers and the optical rotation observed.

100훼 퐸 = (3) [훼]푙푆0

For each experiment, the value of the specific rotation of (S)-BnFTP was estimated using the starting value of the optical rotation taken right upon the addition of DBU. Doing so, the change of specific rotation due to the presence of DBU is implicitly taken into account. As the starting E, l and cm are known, the specific rotation can be estimated through equation (3). All subsequent optical rotation values were then converted to an equivalent enantiomeric excess according to this equation.

Racemization in presence of the co-former When 1eq of a chiral acid is added to the system, a change both in the overall optical rotation as well as in the specific rotation of (S)-BnFTP will occur, this latter due to intermolecular interactions. Taking this effect into account, the new specific rotation of (S)-BnFTP is called [α]’. Furthermore, when DBU is added to the mixture, a carboxylate species is created. Since this carboxylate has a different specific rotation from the acid, this once more impacts the overall optical rotation of the system. This latter was experimentally observed as the optical rotation went from a negative to positive value upon DBU addition to a mixture of (R)-BnFTP and (R)-PBA. The overall expression of the optical rotation is a sum of the optical rotations caused by 133 Chapter 2- Development of the CoISD process: A proof of concept

BnFTP, the acid and the carboxylate with the specific rotation value taking into account the intermolecular interactions:

′ 퐴 퐴 퐴− 퐴− 100훼 = [훼] 푙푆0푥m(푥S − 푥R) + [훼] 푙푐m푥m + [훼] 푙푐m푥m (4)

With [α]’ the specific rotation of (S)-BnFTP (in presence of the acid) in the solvent of the experiment, cm the overall mass concentration of the 1:1 BnFTP/acid mixture in g/100mL, xm the weight fraction of BnFTP compared to the overall BnFTP/acid A A amount, [α] the specific rotation of the acid in the considered system, xm the mass fraction of the acid, [α]A- the specific rotation of the carboxylate in the considered A- system and xm the mass fraction of the carboxylate.

With equation (4), a new relation between the enantiomeric excess and the observed optical rotation can be drawn:

퐴 퐴 퐴− 퐴− 100훼 − [훼] 푙푆0푥m − [훼] 푙푆0푥m 퐸 = ′ [훼] 푙푆0푥m

퐴 퐴 퐴− 퐴− [훼] 푙푆0푥m + [훼] 푙푆0푥m For E = 0, 훼f = With αf the optical rotation at the end of the 100 experiment. This value can be determined by either waiting for full racemization and extrapolating the value of αf from the experimental data or by experimentally creating a mixture corresponding to the fully racemized mixture containing (R,S)-BnFTP, the acid and the DBU with the exact same concentrations. We chose this latter option for all experiments, as reaction times were very long in some case. This yields a final equation linking the enantiomeric excess E to the in situ measured optical rotation α.

100(훼 − 훼f) 퐸 = (5) [훼]′푙푆0푥m

For each experiment, the enantiomeric excess was extracted from the optical rotation d using eq. (5), after [α]’ was estimated using the corrected onset value of the optical rotation right upon DBU addition, when the enantiomeric excess is still the same (before racemization occurs).

Racemization set-up The in-situ set-up consisted of a polarimeter with a flow through cell. This cell was connected to a 100mL-glass vessel (VWR, 215-1592) containing the racemization solution. A gear pump (ISMATEC, ISM405A-230V) was used for a continuous flow of the solution. The tubing used was made of Nylon (RS PRO Air Hose Blue Nylon, OD: 8mm, ID: 5.5mm), which is toluene resistant, and resistant to all used solvents. The reaction vessel was located on a stirring plate (IKA RH digital) and the environment stirred with a stirring bar at a speed of 400 rpm. The solution was pumped

d The measured onset value was subtracted by 훼f to yield the corrected onset value. 134 Chapter 2- Development of the CoISD process: A proof of concept

at a rate of 43.32mL/min, which allowed a stable optical measurement. All experiments were conducted at room temperature with continuous monitoring of the temperature in the cell showing a variation between 22-24°C at maximum. Real-life picture and 3D-model of the set-up are given in Appendix C, figures C-8 and C-9.

Protocol for racemization measurements For each racemization experiment, whether it is with enantiopure BnFTP, enantiopure co-crystal, the volume of the solvent used was 50mL.Then, the BnFTP or co-crystal was dissolved in the solvent, the quantity varied with the experiment depending on the concentration specified. Then, the pump was started and the clear solution spread all over the circuit. When the optical rotation value was stable, the base was added with a micropipette, the volume depended on the needed concentration for the experiment.

3.2.4. Protocol for study of the impact of temperature This study was carried out in a POLAR Bear Plus Crystal device. 200mg of R- BnFTP/R-PBA co-crystal were weighed in an 8mL glass vial with a screw gap. 3mL of toluene were added. Then, the mixture was stirred at a certain temperature in the Polar Bear device for 10minutes. Then, 20µL of DBU were added and samples were taken over time. When a sample was taken, three drops of the liquid were taken out from the vial and immediately quenched in a 1M HCl solution. Then, ethyl acetate was added and the two phases mixture was shaken. The organic phase was isolated and left to evaporate. The obtained solid was analyzed by chiral HPLC to assess the enantiomeric excess.

3.3. Results and Discussion 3.3.1. Racemizing agent screening To screen for a suitable racemizing agent, several acids and bases were assessed in their capacity to induce racemization when put in excess in presence of (S-S)-co- crystal. In the following conditions, the co-former cannot be racemized. The list of acids and bases tried are the following:

 B = Et3N; DBU; DMAP  AH = Formic acid; Acetic acid

The results of these experiments are given in table 2-3.

Table 2-3 Results of the acid/base screening. The enantiomeric ratio were determined by chiral HPLC.

Starting Formic Acetic Et3N DBU DMAP mixture acid acid Ratio “R/S” 6.4/93.6 28.6/71.4 50.3/49.7 17.8/82.2 8.7/91.3 6.4/93.6 135 Chapter 2- Development of the CoISD process: A proof of concept

DBU led to total racemization after 24h while triethylamine and DMAP only led to partial racemization. It must be noted that only one equivalent of DMAP was used, which was in presence of one equivalent of acid. Formic acid and acetic acid did not seem to induce racemization. In general, BnFTP did not seem to racemize under weak acidic conditions but did under basic conditions. To further study the racemization reaction, DBU was chosen as it gave the best results.

The next step required using lower amounts of DBU to see if catalytic amounts of DBU would still induce racemization. For that part, BnFTP was put alone with DBU in toluene. Samples were taken at different times, quenched with diluted HCl (1M). The organic phase was kept and a sample was diluted and injected in chiral HPLC. Two cases were compared: one equivalent of DBU and 0.05eq of DBU. The results are given in table 2-4.

Table 2-4 Enantiomeric ratio of the BnFTP treated with DBU in relation with time, depending on the equivalent used.

0.05eq DBU 1eq DBU %R-BnFTP 4.06 t = 0 %S-BnFTP 95.94 %R-BnFTP 28,62 50.46 30min %S-BnFTP 71.38 49.54 %R-BnFTP 37.2 49.38 1h %S-BnFTP 62.8 50.62 %R-BnFTP 49.75 6h / %S-BnFTP 50.25 With one equivalent of DBU, the racemization of (S)-BnFTP is very fast. After 30 minutes, the compound is completely racemized. With 5%, the racemization is logically slowed down but still occurs relatively fast since after 6 hours at 25°C, the compound was completely racemized. This clearly shows that DBU is an efficient base to racemize BnFTP and that we could move on to the study of its kinetics.

3.3.2. Kinetic study of DBU Kinetic model of racemization In order to study the kinetics of the racemization reaction, we assume an underlying racemization mechanism corresponding to a typical enolization. The overall mechanism can be cut in elementary steps as shown in figure 2-5. 136 Chapter 2- Development of the CoISD process: A proof of concept

Figure 2-5 Acid-basic equilibrium between BnFTP and DBU is in favor of the keto-form of BnFTP due to the lower pKA value of DBU compared to BnFTP. S and R stands for respectively the (S)-enantiomer of BnFTP and (R)-enantiomer in the mathematical model. B stands for the base, A- for the enolate species and BH+ the protonated base in the mathematical model.

The deprotonation and re-protonation rate constants k1 and k2 are identical for both enantiomers, as the base used is non-selective to the enantiomers. The reaction rate of each species is then given by:

푑[푆] = 푘 [퐴-][퐵퐻+] − 푘 [푆][퐵] (6) 푑푡 2 1 푑[푅] = 푘 [퐴-][퐵퐻+] − 푘 [푅][퐵] (7) 푑푡 2 1 푑[퐴-] 푑[퐵퐻+] = = 푘 [푆][퐵] − 2푘 [퐴-][퐵퐻+] + 푘 [푅][퐵] (8) 푑푡 푑푡 1 2 1 At any moment, the sum of the concentration of (R)-BnFTP, (S)-BnFTP and enolate is equal to the starting concentration of BnFTP, S0.

- [퐴 ] + [푅] + [푆] = 푆0 (9)

Similarly, at all times the overall concentration of base and protonated base equals the starting concentration B0,

+ [퐵] + [퐵퐻 ] = 퐵0 (10)

In our model system, formation of the BnFTP enolate (A−) is thermodynamically not 24 favored as the pKA of DBU (pKA ≈ 12) is lower than that of the acid in the α position 25 of the ketone (pKA < 20) (figure 2-5). Consequently, at all times during the reaction, we only expect small concentrations of A− and BH+ to be present. Consequently, we can assume the steady state principle26 for the enolate and the conjugated base:

푑[퐴-] 푑[퐵퐻+] = ≅ 0 (11) 푑푡 푑푡 Combining the steady state principle (11) with eqs. (7), (6) and (8) the rate equation for both the (S)- and (R)-enantiomer can be expressed as: 137 Chapter 2- Development of the CoISD process: A proof of concept

푑[푆] 1 = 푘 [퐵]([푅] − [푆]) (12) 푑푡 2 1

푑[푅] 1 푑[푆] = 푘 [퐵]([푆] − [푅]) = − (13) 푑푡 2 1 푑푡 Subtracting both results in:

푑[푆] 푑[푅] − = 푘 [퐵]([푅] − [푆]) 푑푡 푑푡 1

[푅] + [푆] is constant and approximately equal to the starting concentration S0 of BnFTP as the concentration of the enolate species A− can be neglected compared to that of the other species present. Dividing the previous equation by [R] + [S] leads to:

푑(퐸) = −푘 [퐵]퐸 (14) 푑푡 1 [푆] − [푅] with 퐸 = 푆0 The rate kinetics are thus expected to be first order with respect to the enantiomeric excess E and first order with respect to the concentration [B] of free base, which remains constant during the time of the experiment. The concentration in free base st [퐵] ≅ 퐵0 and can be incorporated in an apparent rate constant yielding an overall 1 order rate equation:

푑(퐸) = −푘′퐸 (15) 푑푡

′ with 푘 = 푘1퐵0

Which integrates to

ln(퐸) = −푘′푡 + ln(퐸0) (16)

With E0 the initial enantiomeric excess.

First order verification Taking a standard experiment (500.2mg of (S)-BnFTP, 50mL of toluene, 100µL of DBU), an excellent linearization of the datae is obtained when the natural logarithm of the enantiomeric excess is plotted against time (Figure 2-6). This agrees with a first

e The end of each experiment (generally, when the enantiomeric excess goes under 0.2) was not considered as the polarimeter is not accurate enough at lower E values. 138 Chapter 2- Development of the CoISD process: A proof of concept

order equation with respect to the enantiomeric excess E as shown by equation 16. All experiments involving BnFTP and DBU show such typical first order behavior.

1 0 0,5 1 0 0,8 -0,5 0,6 E ln(E) [-] [-] 0,4 -1

0,2 -1,5

0 -2 0 1 t [h] 2 3 t [h]

Figure 2-6 Left: the evolution of the enantiomeric excess E versus time t for a racemization with 10 mg/mL of (S)-BnFTP in toluene using 100µL of DBU (2.04g/L). Right: the linearization obtained through equation 11 for the same experiment gave a R² of 0,9997. The slope (with the linear model error) is -2.129 ± 0.003h-1.

In the suggested model, the observed rate constant k’ is independent of the BnFTP starting concentration S0. This was verified by a series of experiments, where only the concentration in BnFTP was varied (Figure 2-7). Value table and linearization for each experiment are given in appendix C table C-4 and figures C-10 to C-16.

14 12 10 k' 8 [h-1] 6 4 2 0 4 6 8 10 12 14

S0 [g/L] Figure 2-7 The observed rate constant k’=5.01±0.48 h-1 does not significantly vary with the starting concentration S0 of BnFTP using 250µL of DBU in toluene. The horizontal dashed line represents the mean value 5.01 h-1. Standard deviation is 0.48 h-1.

As an excellent linearization is obtained for each individual experiment, the error on each k’ resulting from the modeling is very small. There is however a stronger variability between different experiments (as shown by figure 2-7). The average value 139 Chapter 2- Development of the CoISD process: A proof of concept of k’ is 5.01 h-1 with a standard deviation of 10%. As a result, for any experiment, the standard deviation is expected to be around 10%. This latter represents the experimental variability between different samples and one should therefore consider an experimental error of about 0.48 h-1 on the values of the rate constants presented here. This error may come from the error on the exact concentration of DBU (error on the volume of added DBU and on the volume of solvent).

Influence of DBU concentration on racemization rate 1

0,75

E 0,5 [-]

0,25

0 0 0,2 0,4 t [h] 0,6 0,8 1

1.02g/L 2.04g/L 5.1g/L 7g/L 9.3g/L 12g/L 15.6g/L

Figure 2-8 Enantiomeric excess E over time of (S)-BnFTP in the reaction mixture for different concentrations B0 of base catalyst, which is indicated by the labels. The higher the concentration of the base, the faster the racemization. For all experiments, the concentration of BnFTP is 10 mg/mL.

We then investigated the order with respect to DBU. Since DBU is a catalyst and not consumed during the process, its concentration remains constant over time with [퐵] ≅ 퐵0 as mentioned above. To study the rate order with respect to DBU, we therefore determined the rate constant k’=k1B0 dependence on DBU concentration. The evolution of the enantiomeric excess of these experiments over time is given in figure 2-8. This figure shows an increase in the racemization rate as the concentration of DBU increases. Figure 2-9 shows the k’ values as a function of base concentration

B0. Value table & Linearized data are shown in appendix C, table C-5 & figures C-17 to C-20. The almost perfect linear relation obtained between k’ and B0 confirms first order rate in DBU as predicted by our model and the value of the intrinsic -1 -1 -1 -1 deracemization rate constant k1 in toluene is 157.7L.mol .h or 2.63L.mol .min . Brevegleri et al. measured a rate constant of the same order (9.44L.mol-1.min-1 at 25°C in acetonitrile) for a similar compound, N-(2-methyl-benzylidene)-phenylglycine amide (NMPA) in a more polar solvent.27 140 Chapter 2- Development of the CoISD process: A proof of concept

Thus, the kinetic model we propose for the racemization reaction of BnFTP in presence of DBU is confirmed by the experimental data. It is a first-order kinetic model similar to the one developed and validated by Brevegleri et al. and in accordance with literature.28

16 14 12 10 k' 8 [h-1] 6 4 2 0 0 0,05 0,1 B0 [mol/L]

Figure 2-9 Linear evolution of the observed rate constant k’ with the base concentration B0 (error bars show the standard deviation). The linear regression gave the following equation: -1 -1 k'= 157.7B0 with a R² of 0.9992. The linear model error on the slope is ±1.1L.mol h .

Impact of the solvent nature 1

0,75

E 0,5 [-]

0,25

0 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 t [h] Toluene Ethanol Propan-1-ol THF Butan-1-ol Acetone Acetonitrile

Figure 2-10 Time evolution of the enantiomeric excess E of BnFTP in different solvents (grey for ethanol, orange for propan-1-ol, green for butan1-ol, red for acetonitrile, purple for acetone, blue for toluene and yellow for THF) using the same base concentration (5.1g/L). 141 Chapter 2- Development of the CoISD process: A proof of concept

Seven solvents were then selected to study the impact of the solvent nature has on the racemization rate. Figure 2-10 shows a strong impact of the solvent on the racemization rate for experiments performed with 9.7g/L BnFTP and a catalyst concentration of 5.1g/L. All the experiments result in the typical behavior of the enantiomeric excess E following eq. 16. Ethanol and propanol yield the fastest racemization rate by far. Moreover, a 10-fold difference can be observed between the lowest and highest rate constant (respectively for THF and ethanol). At first sight, the rate constant shows no clear relation with the dielectric constant (table 2-5). For instance, acetonitrile, the solvent with the highest dielectric constant, only gave the fourth highest rate constant k’. Nevertheless, out of these solvents, three are protic (Ethanol, Propan-1-ol and Butan-1-ol) and four are not (THF, Toluene, Acetone and Acetonitrile). Among the protic solvents, the most polar ones lead to higher racemization rate (figure 2-11). Similarly, solvent polarity seems to increase racemization rate also for non-protic solvents, even though the effect is less pronounced (figure 2-11). Overall racemization kinetics of our system seem therefore enhanced when using protic solvents with dielectric constant  >18, contrary to the system studied by Breveglieri et al.27 that showed a strong enhancement of the kinetics when using acetonitrile compared to isopropanol (which has a similar dielectric constant as propan-1-ol). In comparison, in our case the rate constant in propan-1-ol is significantly higher than in acetonitrile. This highlights the importance of the nature of the compound for the enolization kinetics and illustrates that solvent selection in this case will have to be performed on a case dependent basis.

Table 2-5 Value of the observed rate constant k’ with its experimental error for every studied solvent, compared with their dielectric constant. Linearization for each curve is given in appendix C, figures C-21 to C-26. Linearization error is discussed in appendix C with figure C-27.

Dielectric Protic k’ Linear model k Linear model error Solvent Constant29,30 solvent [h-1] error on k’ [h-1] [L.mol-1h-1] on k [L.mol-1h-1] THF 7.58 No 3.88 ±0.01 116 ±1 Toluene 2.38 No 5.01 ±0.01 150 ±1 Acetone 20.7 No 5.89 ±0.01 176 ±1 Acetonitrile 37.5 No 7.07 ±0.02 211 ±3 Butan-1-ol 17.8 Yes 8.03 ±0.01 240 ±1 Propan-1-ol 20.3 Yes 19.50 ±0.05 582 ±8 Ethanol 24.6 Yes 37.02 ±0.46 1105 ±70 142 Chapter 2- Development of the CoISD process: A proof of concept

40 35 30 25 k' 20 [h-1] 15 10 5 0 0 20 40 ε [-] Figure 2-11 Observed rate constant k’ as a function of the dielectric constant of the solvent. Aprotic solvents are plotted in blue while protic ones in orange. For protic solvents, a small increase in ε induces a high increase in k’ while for aprotic solvents, a high increase in ε only slightly increases k’.

The impact of proticity on racemization kinetics can be explained by the stabilization of the enolate. Protic solvents can stabilize the negatively charged enolate through assisted hydrogen bonds. We therefore expect a lowering of the activation energy for the formation of these species as illustrated in figure 2-12. A computational study previously showed the importance of the solvents’ ability to interact through hydrogen bonding with carbonylsf in stabilizing the enol form. They also showed that if two solvents are able to hydrogen bond with the solute, then the difference in stability would be due to the difference in the dipolar moment between the enol and keto-form. Hence, the dielectric constant of the solvent is the determining criterion explaining the difference in stability.31 Similarly, we observed that a more polar solvent increases the stability of the intermediate enolate, when comparing solvents with similar hydrogen bonding capacity (figure 2-11). Furthermore, out of these two effects, the formation of hydrogen bonding seems to have a stronger effect, as the rate constants for isopropanol compared to acetone, though similar in polarity, show a four-fold increase while for propanol a three-fold increase is observed compared to acetonitrile. In the case of Breveglieri et al., the non-stabilizing effect of hydrogen bonding when using NMPA,27 could be explained by a disruption in the mesomeric effect of the imine group, in the β position of the carbonyl, caused by hydrogen bonding of its nitrogen atom with the alcohol group of the solvent. Resulting in its lone pair no longer being available in the resonance. This likely explains their seemingly contradicting results.

f Those results were drawn for isomer pairs in which the enol cannot form an internal hydrogen bond. 143 Chapter 2- Development of the CoISD process: A proof of concept

Figure 2-12 Hypothetical energetic diagram for BnFTP racemization in acetone, acetonitrile and isopropanol. On the left the molecule of (S)-BnFTP, on the right that of (R)-BnFTP and in the middle the enolate species. Each curve is color-coded according to the solvent.

3.3.3. Kinetic study with DBU in presence of (R)-PBA Influence of (R)-3-Phenylbutyric acid on racemization The deracemization process we aimed at developing occurs in presence of (R)-3- Phenylbutyric acid as a co-former. The presence of this acid is expected to significantly impact racemization kinetics as it interacts with the base DBU. Indeed, when an additional acid with a pKa ≈ 4-5 is present in the environment, the previous model (eq. 16) no longer holds as DBU is expected to be fully protonated (ΔpKa > 4, quantitative reaction). The carboxylate is not expected to be strong enough to deprotonate BnFTP, leading us to suggest a coupled DBUH+/carboxylate action (figure 2-13), provided the acid is in excess compared to DBU.

Figure 2-13 Assumed mechanism explaining the enolization of BnFTP in the presence of an excess of additional acid. 144 Chapter 2- Development of the CoISD process: A proof of concept

This mechanism can be seen as a base-assisted enolization in an acidic medium. A similar process was observed for the self-condensation of acetaldehyde when water (a weak base) deprotonates the enol of acetaldehyde to trigger its attack on acetaldehyde.32 No kinetic studies, to best of our knowledge, were carried out on such systems.

To study the kinetics of DBU in presence of PBA, 6 experiments were carried out in toluene with one equivalent of (R)-3-Phenylbutyric acid (compared to (R)-BnFTP) varying the concentration of the racemization catalyst DBU. For 3 experiments, the acid is in excess compared to DBU and for the remaining 3, DBU is in excess. All experiments were carried out in toluene, the solvent previously chosen for the chiral resolution. The evolution of the enantiomeric excess over time is given in figure 2-14. All linearizations are given in appendix C, figures C-28 to C-33.

1 0,9 0,8 0,7 E 0,6 [-] 0,5 0,4 0,3 0,2 0,1 0 0 2 4 t [h] 6 8 10 5g/L - 0.74eq 5.5g/L - 0.81eq 6.5g/L - 0.96eq 7g/L - 1.03eq 8g/L - 1.16eq 9g/L - 1.31eq

Figure 2-14 Time evolution of the enantiomeric excess E of BnFTP in presence of 1 equivalent of 3-Phenylbutyric acid, with different base concentration. In the caption, next to the concentration of base is given its equivalent compared to the acid.

A significant drop in racemization rate is observed compared to the experiment without the acid. However, linearization of the data according to first order kinetics remains possible for all experiments, suggesting at first a similar mechanism. Regarding the evolution of the observed rate constant as a function of DBU concentration, the first observation is that the experiment with 7g/L of DBU gave faster kinetics than the one with 8g/L. Indeed, two cases are to be considered:

Excess of acid Excess of base. 145 Chapter 2- Development of the CoISD process: A proof of concept

When DBU is in sufficient excess, the observed rate constants are similar to the ones expected for an experiment without acid taking into account the concentration of unprotonated DBU. Indeed, the observed rate constant k’ obtained for these experiments yields a value of 2.30h-1 for 1.31eq and 0.64h-1 for 1.16eq. Using the intrinsic rate constant k of 157.7L.mol-1.h-1 from the experiment without the acid, this would imply a DBU concentration of respectively 2.22 (0.32eq.) and 0.55 (0.09eq) g.L-1. This is approximately the excess amount of DBU present (respectively 0.31eq and 0.16eq), suggesting that for these experiments, the reaction occurs via the proposed racemization mechanism, but with an apparent reduced concentration of DBU available for the racemization, therefore reducing the racemization rate.

0,03 0,035 0,04 0,045 0,05

0 -1 ln(k') [-] -2 -3 -4 -5 B0 [mol/L]

Figure 2-15 Linear evolution of the natural logarithm of the observed rate constant ln(k’) with the base concentration B0. Linear regression gave the following equation: ln(k') = 329.7B0 – 15.4 with a R² of 0.999. The linear model error on the slope and intercept is respectively ±7.3L.mol-1h-1 and ±0.3h-1. The intercept of -15.4 suggest the kinetics of racemization of BnFTP in toluene in presence of the acid but without DBU are very low (2.1,10-7h-1) and can be neglected.

When the acid is in excess, the value of k’ can no longer be correlated using the same relation. Indeed, the mechanism appears to be different regarding the relation between DBU’s concentration and the observed rate constant. In figure 2-15, it can be seen that the evolution can be linearly fitting by expressing the natural logarithm of k’ as a function of the concentration of DBU. Moreover, it can be seen that the value obtained with a very small excess of DBU fits the same correlation. As a result, the mechanism is more complex when the acid is in excess. The explanation for the switch between these two cases is simply a matter of dominant mechanism. When the acid is in excess, all DBU is protonated and the mechanism cannot be the same. When DBU is in excess, the mechanism appears to follow that without the acid considering only the concentration of free DBU. In the case of a small excess of base, the mechanism with the acid appears to be kinetically deciding, probably because the concentration of free DBU is too low compared to the concentration of carboxylate/ammonium. 146 Chapter 2- Development of the CoISD process: A proof of concept

Though first order to the enantiomeric excess, the mechanism when the acid is in excess, is complex as shown by varying concentrations of DBU. An exponential relationship between the observed rate constant and the concentration of DBU suggests a rather complicated system since no simple mechanism can explain this evolution of the observed constant rate with the base. Consequently, the mechanism is more complex with possible complexation equilibriums in play. Within the scope of this chapter, the evolution of the observed rate constant with base concentration was a fruitful information for developing and optimizing the CoISD process. However, as the kinetics are too slow when base concentration is low, for the viability of the process, heating was thought of in order to ensure fast enough kinetics when the acid is present.

For this reason, the relationship of the observed constant rate with temperature was preferentially studied as opposed to the establishment of a full kinetic model explaining empirical observations that would require collecting data on every possible parameter like concentration of acid, concentration of BnFTP… To do so, a different set-up, adapted to heating was used as detailed in the material and methods (section 3.2.4). Five experiments were carried out at different temperatures with a base concentration of 6.8g/L (0.29eq) to obtain the observed rate constant (curves, linearizations and value table can be found in appendix C figures C-34 to C-36 and table C-6). 1 0,5 ln(k') 0 [-] -0,5 -1 -1,5 0,0026 0,0027 0,0028 0,0029 0,003 -1 1/T [K ] Figure 2-16 Linear evolution of the natural logarithm of the observed rate constant ln(k’) with the inverse of the temperature 1/T. The linear regression gave the following equation: ln(k') = -8225/T + 22,89 with a R² = 0,9979. The linear model error on the slope is respectively ±219L.mol-1h-1.

The evolution of the logarithm of the observed rate constant k’ for each experiment with the inverse of the temperature gave a straight line (figure 2-16) showing that despite its complexity, the mechanism of racemization with an excess amount of acid still responds to an Arrhenius law and the activation energy of the reaction is about 68.4KJ.mol-1. Extrapolating from this data, at 20°C, the observed rate constant is of 8.9x10-3h-1. The activation energy is in the range of other similar racemization 147 Chapter 2- Development of the CoISD process: A proof of concept reactions like that of L-glutamic acid, catalyzed by salicylaldehyde derivatives, ranging from 54 to 67KJ/mol.33 When compared to thermal racemization of axial or planar chiral compounds, the activation energy remains of the same order (75 to 116KJ/mol).34-37 In those cases, the molecules are generally stable at room temperature and racemize upon heating. Similarly, with our system, we have the same result with low kinetics at room temperature, increasing sufficiently upon heating to induce racemization. 3.4. Conclusion DBU was found as the best racemizing agent for BnFTP out of all tried and more generally, only bases are efficient to racemize BnFTP. Then, the kinetics of DBU induced BnFTP racemization were studied and the racemization reaction was shown to be first order with respect to the base as well as BnFTP, confirming the proposed enolization racemization mechanism. Protic-polar solvents such as ethanol and 1- propanol lead to fast racemization showing a relatively high racemization rate constant. Addition of a chiral acid strongly decreases the kinetics of the racemization due to the reaction of the base with the acid, yielding an ammonium salt species. Racemization with this ammonium salt as a base displays a seemingly more complex mechanism, with an apparent exponential evolution of the rate constant with respect to the base concentration. Consequently, a full understanding of this system demands further studies. However, the mechanism does remain first order with respect to the enantiomeric excess and was shown to follow the Arrhenius law. This section allowed finding a suitable racemizing agent system for BnFTP in presence of PBA. BnFTP racemization kinetics provides key information for the development of the CoISD process and its optimization, like the interest of using protic solvents to increase racemization kinetics, as well as the importance of the base concentration and temperature on racemization kinetics.

4. Bibliography 1. Modarresi, H., Conte, E., Abildskov, J., Gani, R., Crafts, P. Model-Based Calculation of Solid Solubility for Solvent Selections-A Review. Ind. Eng. Chem. Res. 47, 5234– 5242 (2008). 2. Myerson, A. S., Decker, S. E., Fan, W. Solvent selection and batch crystallization. Ind. Eng. Chem. Process Des. Dev. 25, 4, 925–929 (1986). 3. Chiarella, R. A., Davey, R. J., Peterson, M. L. Making Co-Crystals The Utility of Ternary Phase Diagrams. Cryst. Growth & Des. 7, 7, 1223-1126 (2007). 4. Leusen, F. J. J., Noordik, J. H., Karfunkel, H. R. Racemate Resolution via Crystallization of Diastereomeric Salts: Thermodynamic Considerations and Molecular Mechanics Calculations. Tetrahedron. 49. 24, 5377-5396 (1993). 5. Smith, H. E., Padilla, B. G., Neergaard, J. R. Chen, F.-M. J. Am. Chem. Soc. 100, 19, 6035-6039 (1978). 6. Allen, J. R. F. Crop Prot. 17, 207–212 (1998). 148 Chapter 2- Development of the CoISD process: A proof of concept

7. Oketani, R., Hoquante, M., Brandel, C., Cardinael, P., Coquerel, G. Cryst. Growth Des. 18 (11), 6417-6420 (2018). 8. De Maria, P., Fontana, A., Siani, G., Spinelli, D. Eur. J. Org. Chem. 1867-1872 (1998). 9. De Maria, P., Fontana, A., Spinelli, D. J. Chem. Soc., Perkin Trans. 2. 1067-1070 (1991). 10. Malhotra, S., Jaspal, D. Bulletin of Chemical Reaction Engineering & Catalysis. 8 (2), 106 (2013). 11. Malhotra, S., Jaspal, D. Bulletin of Chemical Reaction Engineering & Catalysis., 9 (1), 17 (2014). 12. De Maria, P., Fontana, A., Frascari, S., Ferroni, F., Spinelli, D. J. Chem. Soc. Perkin Trans. 2. 825-828 (1992). 13. De Maria, P., Fontana, A., Arlotta, M., Chimichi, S., Spinelli, D. J. Chem. Soc. Perkin Trans. 2. 415-419 (1994). 14. Khirwar, S. S. Orient. J. Chem. 26(3), 1037-1042 (2010). 15. Cox, R. A., Smith, C. R., Yates, K. Can. J. Chem. 57, 2952-2959 (1979). 16. Cox, R. A., Warkentin, J. Can. J. Chem. 50, 3242-3247 (1972). 17. Warkentin, J., Barnett, C. J. Am. Chem. Soc. 90,17, 4629-4633 (1968). 18. He, X., Morris, J. J., Noll, B. C., Brown, S. N., Henderson, K. W. J. Am. Chem. Soc. 128, 13599-13610 (2006). 19. Bel, R. P., Hillier, G. R., Mansfield, J. W., Street, D. G. J. Chem. Soc. B, 827-832 (1967). 20. Eberlin, A. R., Williams, D. L. H. J. Chem. Soc., Perkin Trans. 2. 883-887 (1996). 21. Hynes, M. J., Clarke, E. M. J. Chem. Soc., Perkin Trans. 2. 901-904 (1994). 22. Lorenz, H., Seidel-Morgenstern, A. Angew. Chem. Int. Ed. 53, 5, 1218–1250 (2014). 23. Dunn, A. S., Svoboda, V., Sefcik, J., ter Horst, J. H. Org. Process Res. Dev. 23, 9, 2031–41 (2019). 24. Khurana, J. M., Nand, B., Saluja, P. Tetrahedron. 66, 5637–5641 (2010). 25. Babler, J. H., Liptak, V. P., Phan, N. J. Org. Chem. 61,1, 416–417 (1996). 26. Espenson, J. H. Consecutive Reactions: the Steady-State and Other Approximations. in Chemical Kinetics and Reaction Mechanisms (ed. McGraw-Hill) 77–82 (1995). 27. Breveglieri, F., Mazzotti, M. Cryst. Growth Des. 19, 3551−3558 (2019). 28. Wolf, C. Racemization, enantiomerization and diastereoisomerization. in Dynamic Stereochemistry of Chiral Compounds: Principles and Applications, Royal Society of Chemistry, 29–34 (2008). 29. https://people.chem.umass.edu/xray/solvent.html 30. Wohlfarth, C. Dielectric constant of butan-1-ol. in Landolt-Börnstein IV/17: Static Dielectric Constants of Pure Liquids and Binary Liquid Mixtures (Eds. Lechner, M. D.) (2008). 31. Mills, S. G., Beak, P. J. Org. Chem. 50, 1216-1224 (1985). 32. Baigrie, L. M., Cox, R. A., Slebocka-Tilk, H., Tencer, M., Tidwell, T. T. J. Am. Chem. Soc. 107, 3640-3645 (1985). 33. Ando, M., Emoto, S. Bull. Chem. Soc. Jpn. 42, 2628–2631 (1969). 34. Oguz, S. F., Dogan, I., Tetrahedron: Asymmetry. 14, 1857–1864 (2003). 35. Tajiri, A., Morita, N., Asao, T., Hatano, M. Angew. Chem. Int. Ed. 24, 329 330 (1985). 36. Westheimer, F. H. J. Chem. Phys., 15, 252 (1947). 37. Anet, F. A. L., Jochims, J. C., Bradley, C. H. J. Am. Chem. Soc. 92, 2557–2558 (1970). 149

Chapter 3- Development & optimization of the CoISD process

The section 2. of this chapter comes from the following communication:

150 Chapter 3- Optimization of the CoISD process

1. Overview Processes leading to enantiopure compounds are of utmost importance, in particular for the pharmaceutical industry. Starting from a racemic mixture, Crystallization Induced Diastereomeric Transformation allows for a theoretical 100% transformation of the desired enantiomer. However, this method has the inherent limiting requirement for the organic compound to form a salt. In this chapter, this limitation is lifted by introducing co-crystallization in the context of thermodynamic deracemization, with the process applied to a model chiral fungicide. This chapter first presents the new general single thermodynamic deracemization process based on co-crystallization for the deracemization of (R,S)-4,4-dimethyl-1-(4-fluorophenyl)-2-(1H-1,2,4-triazol-1- yl)-Pentan-3-one (BnFTP). Its feasibility is proven paving the way to further development of such processes, which was secondly achieved by detailing the optimization of the CoISD process. Considering both kinetic and thermodynamic parameters obtained from the previous chapter; we investigated the impact of water, nature of the solvent, concentration of base, equivalent of acid co-former and temperature of the crystallization cell. The evaluation of each parameter allowed for improvement of the yield and/or deracemization ultimately providing an optimized process with a yield of 73% of pure co-crystal for an overall deracemization of 80%, furthermore, highlighting the viability of recycling the remaining mother liquors.

2. CoISD: A general innovative thermodynamic approach to deracemization: Proof of concept. 2.1. Introduction With the increasing number of enantiopure, chiral drugs developed every year1 and regulatory instances encouraging the development of enantiopure compounds2, processes allowing access to these, are of utmost importance. In spite of significant advances in asymmetric synthesis (in particular asymmetric catalysis), the most prominent way to enantiopure drugs nowadays still involves formation of a racemic compound3 and separation of the unwanted enantiomer through a resolution process4-8, or its transformation into the desired enantiomer, in a so-called deracemization process. Crystallization-based resolution processes are less costly than e.g. chromatographically based techniques, and therefore industrially wide- spread. Typical crystallization-based resolution processes are preferential crystallization9-11 and diastereomeric resolution.12-14 151 Chapter 3- Optimization of the CoISD process

Going beyond separation, crystallization based deracemization processes aim at transforming the unwanted enantiomer (distomer) into the desired one (eutomer). Over the recent years, different deracemization tools were developed. The kinetic process of Viedma Ripening (VR)15,16 and Dynamic Preferential Crystallization (DPC)17 require a conglomerate forming racemate and are therefore inherently limited to 5-10% of all compounds. Crystallization Induced Diastereomeric Transformation--CIDT18,19, on the other hand, is a thermodynamical approach, based on the differences in solubility between two diastereomeric salts and does therefore not require the formation of such a conglomerate.

As highlighted by a 2006 literature revue CIDT can only be performed on salt-forming compounds with the vast majority of studied systems combining a carboxylic acid with an amine.20 For non-salifiable compounds, to the best of our knowledge, no thermodynamically based deracemization method has been reported and thus many racemizable compounds are left with no viable option for deracemization. We are the first, to introduce here such a method, based on co- crystallization, expanding the scope of thermodynamically based deracemization processes to all racemizable compounds (scheme 3-1).

Scheme 3-1 State of the art regarding deracemization and how co-crystallization induced spontaneous deracemization (CoISD) redistributes the cards and opens new possibilities in the world of deracemization.

Co-crystallization typically relies on strong intermolecular interactions like hydrogen or halogen bonding21, which are more universal. Co-crystallization was recently explored by others and us in the context of chiral resolution, targeting several racemic drug systems.22-25 Based on these methods, and drawing a parallel to CIDT, 152 Chapter 3- Optimization of the CoISD process

we set out to go beyond chiral resolution targeting a Co-crystallization Induced Spontaneous Deracemization (CoISD) process. The process developed here is innovative, industrially friendly and scope-expanding. It is a thermodynamic process applicable to all, non-salt as well as salt forming compounds, and both to conglomerate or racemic compound forming systems, hereby making it a general process compared to all the other crystallization based deracemization processes.

2.2. Materials and methods 2.2.1. The studied system DBU (Sigma-Aldrich) & toluene (VWR) were used as bought.

(S)/(R)-3-phenylbutyric acid The same protocol as described in chapter 2 section 2.2.1 for (S)-3-phenylbutyric acid was applied to obtain the (R)-3-phenylbutyric acid.

(R,S)-BnFTP (R,S)-BnFTP was synthetized in 2 steps as detailed in chapter 1 section 2.2.2.

2.2.2. The co-crystals The structure of both co-crystals was determined by single crystal measurement (cf. chapter 1 section 2.2). For single crystal growth, the (S-S)-co-crystal was obtained from a cooling experiment in acetone (9°C) of a 1:1 mixture of (R,S)-PBA and (R,S)- BnFTP while the (R-S)-co-crystal was obtained from solvent evaporation in ethyl acetate of a 1:1 mixture of a R-BnFTP and S-PBA. Structural information for the (R- S)-co-crystal can be found in appendix D, table D-1.

2.2.3. Deracemization process Technical details of the set-up The developed system (figure 3-1) uses two double-jacket vessels, one with a 10°C cooling liquid circulating through and the other with a 90°C heating liquid. Two gear pumps are used, one transferring liquid from the crystallization vessel to the racemization vessel, and one pump doing the opposite. A filter was used to avoid solid transfer occurring between the crystallization vessel and the racemization vessel. PTFE tubing used are toluene resistant. Stirring was carried out magnetically with a bottom stirrer in each vessel. The processes were carried out in a semi-continuous fashion: Over-night transfer of liquid between vessels was stopped. For the first two experiments, heating was maintained during the night while this was not the case for the third experiment. 153 Chapter 3- Optimization of the CoISD process

Figure 3-1 Real-view image of the deracemization set-up. Red tubing are for the heating or cooling liquids while white PTFE tubing is used for the solution transfer. On the left, one can see the crystallization vessel with solid in suspension and on the right the racemization vessel with a clear solution inside.

General procedure The speed of the pumps is variable over the experiment. The speed is normally taken as the lowest possible (4.2mL/min) but due to a difference in the strain between the two pumps, the amount of liquid being pumped over time from the crystallization vessel differs from the one pumped to this vessel and speed had to be changed over time to keep each vessel volume constant.

All experiments were run with 90mL of toluene. Stirring was achieved magnetically at the specified speed in table 3-1. The system (BnFTP, PBA and toluene) is equilibrated before addition of DBU (tequ, table 3-1). Then, the pumps were turned off and DBU was added in the racemization vessel. Both vessels were stirred for a specified time (trac, table 3-1) then the pumps were turned on again and the process run for n days (table 3-1). Each night, the pumps are stopped until the next morning.

The cooler’s temperature is as specified (Tcryst, table 3-1) and is left unchanged at night. The heater’s temperature during the day is 90°C and at night as specified (Night heating, table 3-1). The last day, the racemization vessel was pumped back to the crystallization vessel and the pumps were turned off. Heating was turned off. The crystallization vessel was kept stirring at Tf (table 3-1) for tf and the suspension was filtered over Büchner.

After filtration, the filtrate and the cake were analyzed by chiral HPLC (appendix D). For experiment 1 to 3, chromatograms are given respectively in figure D-3, figure D- 10 and figure D-11.

154 Chapter 3- Optimization of the CoISD process

Table 3-1 Varying parameters in the general procedure for each experiment.

mPBA Stirring Exp. mBnFTP VDBU Stirring Cryst. n (R/S) Racem. 1 5g (S) 2.98g 203 µL 1000 rpm 400 rpm 4 2 6.25g (R) 3.73g 254 µL 600 rpm 400 rpm 5 3 7.5g (R) 4.47g 305 µL 600 rpm 400 rpm 5

Exp. tequ trac Tcryst Night heating Tf tf 15°C (Day 1-2) 60°C (day 1) 1 1h30 4h 10°C 1h30 10°C (Day 3-4) 90°C (day 2-4) 2 1h 3h30 10°C 75°C 5°C 1h 3 1h 3h30 10°C Off 0°C 1h Follow-up Exp. 1 Samples were taken at different moments during the four days of the run (table 3-2). All chromatograms can be found in appendix D, figures D-1 to D-3.

Day 1: One sample was taken from the racemization vessel prior to DBU addition (sample A0) and after 3h45 (sample A1). Day 2: At 11h08, the pumps were turned off for 3h45. Two samples were taken, one at 11h08 (sample B0) and one 3h45 later (sample B1). Day 3: A sample was taken at 10 a.m. (sample B2) before turning the system on again. At 14h20, a sample was taken from the racemization vessel (sample C0). Day 4: A sample was taken at 9h40 (sample C1) before the pumps restarted.

Table 3-2 Enantiomeric excess of the racemization vessel over the course of the deracemization process 1.

eeracemization vessel Sample A0 0,32[R] Sample A1 0,00 Sample B0 0,22[R] Sample B1 0,14[R] Sample B2 0,01[R] Sample C0 0,24[R] Sample C1 0,06[S] Follow-up Exp. 2 Samples were taken at different moments during the two first days of the run (table 3- 3). It must be noted that the sample from the crystallization vessel was taken out of the pump that brings the liquid from the crystallization vessel to the racemization vessel (to avoid having solid in suspension). All chromatograms can be found in appendix D, figures D-4 to D-10. 155 Chapter 3- Optimization of the CoISD process

Table 3-3 Enantiomeric excess of both the crystallization and racemization vessel over the first two days of the deracemization process 2.

Sample Sample eeCrystallization eeRacemization Number Name vessel vessel D1.1 Day 1 11h15 0.54 [S] 0.20 [R] D1.2 Day 1 14h45 0.08 [R] 0 D1.3 Day 1 15h45 0.02 [S] 0.05 [S] D1.4 Day 1 16h45 0.34 [R] 0.30 [R] D1.5 Day 1 17h45 0.25 [R] 0.04 [S] D1.6 Day 1 18h45 0.33 [R] 0.04 [R] D2.1 Day 2 9h30 0.25 [S] 0.13 [R] D2.2 Day 2 10h30 0 0.02 [R] D2.3 Day 2 11h30 0.34 [R] 0.02 [R] D2.4 Day 2 14h30 0.04 [R] 0.04 [S] 2.3. Results and discussion We used a model system to develop the CoISD process. The racemic target compound (R,S)-4,4-dimethyl-1-(4-fluorophenyl)-2-(1H-1,2,4-triazol-1-yl)-Pentan-3-one [(R,S)-BnFTP] belongs to a family of fungicidal compounds26, for which a conglomerate forming system has already successfully been deracemized through the kinetic Viedma ripening procedure.27,28 Combining BnFTP with the chiral co-former, (S)-3-Phenylbutyric acid [(S)-PBA], a diastereomeric pair of co-crystals can be obtained. Each diastereomer crystallizes in a chiral space group with the asymmetric unit only containing one enantiomer of the target compound alongside (S)-PBA.23,25,29

A. B.

Figure 3-2 A. Asymmetric unit of the (S)-4,4-dimethyl-1-(4-fluorophenyl)-2-(1H-1,2,4-triazol- 1-yl)-Pentan-3-one-(S)-3-Phenylbutyric acid-co-crystal. B. Asymmetric unit of the (R)-4,4- dimethyl-1-(4-fluorophenyl)-2-(1H-1,2,4-triazol-1-yl)-Pentan-3-one-(S)-3-Phenylbutyric acid- co-crystal. Displacement ellipsoids are drawn at the 50% probability level. Hydrogen bonds are shown as black dashed lines. Disorder is left out for clarity.

The diastereomers crystallize in the P212121 and P21 space groups for [(S)- BnFTP-(S)-3-Phenylbutyric acid] (figure 3-2A) and [(R)-BnFTP-(S)-3-Phenylbutyric acid] (Figure 3-2B) respectively. The former will be referred to as the (S-S)-co-crystal 156 Chapter 3- Optimization of the CoISD process

and is the energetically favored diastereomera. As a consequence, this diastereomer has a lower solubility compared to the (R-S)-co-crystal.

The principle behind CoISD (Scheme 3-2) taps into this solubility difference.30 Given the right conditions, addition of (S)-PBA to a racemic mixture of BnFTP will selectively lead to crystallization of only the (S-S)-cocrystal (purple cubes). This induces a solution enantiomeric excess towards (R)-BnFTP (orange squares). Addition of a racemizing agent will pull the solution imbalance towards the racemic equilibrium once more, implying a net transformation in solution of (R)- BnFTP to (S)-BnFTP. The associated concentration increase in (S)-BnFTP will lead to a solution that is supersaturated with respect to the (S-S)-cocrystalb, which continues to crystallize as long as a sufficient amount of co-former is present in solution. This process is purely thermodynamic and eventually leads to spontaneous full deracemizationc.

Scheme 3-2 Principle of the Co-crystallization Induced Spontaneous Deracemization process.

Toluene was selected as crystallization solvent, as the (S-S)-cocrystal behaves congruently in this solvent and furthermore shows low solubility. On top, this solvent allows the BnFTP racemization reaction to run without major difficulty. Moreover,

a When mixing both racemic (R,S)-BnFTP and (R,S)-PBA a mixture of the (R-R) and (S-S) co- crystals are formed instead of the (R-S) – (S-R) mixture, showing a higher stability of the (S- S) with respect to the (R-S) diastereomer. b The co-crystal solubility product depends on the concentration of both components in solution Ksp= [(S)-BnFTP]*[ (S)-PBA] c Full deracemization does not imply a 100% transformation of R into S. The final solution (in equilibrium with the enantiopure solid state) still contains a mixture of (R)- and (S)-BnFTP. The lower the solubility of the co-crystal, the higher the overall deracemization. 157 Chapter 3- Optimization of the CoISD process there is a substantial solubility difference between both diastereomers. Chiral resolution conditions in toluene (chapter 2 section 2.3.4), allow to crystallize the (S- S)-cocrystal with a 32% yieldd and an ee of 98.6% (figure 3-3) starting from the (R,S)- BnFTP racemate. This process leaves a solution imbalance in favor of (R)-BnFTP (ee=58.6%).

Figure 3-3 (left) Chiral chromatography of the cake (right) and the filtrate obtained from the (R,S)-BnFTP chiral resolution process in toluene.

Besides induction of a solution enantiomeric imbalance, a racemization reaction is also a prerequisite for the development of a deracemization process. In our case, racemization is based on the keto-enol equilibrium of BnFTP using either a Brønsted acid or base.31 BnFTP racemizes freely in the presence of weak bases but does not in presence of weak or strong acids. 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) was chosen as the racemizing agent since the use of one equivalent of DBU at room temperature led to full racemization in 30 minutes, whereas a 5 mol% catalytic amount of DBU was shown to induce complete racemization in 6h (chapter 2 section 3.3.1). Unfortunately, addition of the base to a solution containing both BnFTP and co-former no longer led to racemization at this temperature. This can easily be understood, as DBU (less than 1eq with respect to BnFTP and co-former) will deprotonate the carboxylic acid of the co-former, producing a much weaker carboxylate base. This latter is not strong enough to induce racemization under the initial conditions studied. A similar situation is often encountered in CIDT-processes for which a temperature increase is typically required for the racemization to occur.32,33 Keeping this in mind, we performed racemization in presence of the co-former (and sub-stoichiometric amounts of DBU) at higher temperatures. After 12 hours at 110°C, the filtrate obtained from the resolution fully racemized while 2h at 90°C partially racemized it (appendix D, figure D-12). Temperature increases are usually counterproductive with respect to crystallization processes. To be able to access high yields, we decided to physically separate both processes working with a crystallization vessel at 10°C and a

d It must be noted that the maximum yield for a resolution is 50%. 158 Chapter 3- Optimization of the CoISD process

racemization vessel at 90°C.e The liquid from the crystallization vessel with an enantiomeric imbalance in favor of the distomer is continuously transferred to the racemization vessel, and the racemic solution from the racemization to the crystallization vessel, as shown in figure 3-4. A 3D representation of the system is given in figure 3-5.

Figure 3-4 Sketch of the single process deracemization process set-up.

Figure 3-5 3D representation of the deracemization system set-up. Liquids are continuously pumped between the crystallization (blue) and racemization (red) vessel.

e For a process, which does not require high temperatures, a one-pot method combining crystallization and racemization is fully achievable. 159 Chapter 3- Optimization of the CoISD process

In this chapter, we pioneer in showing the success of the CoISD process through 3 experiments. Each experiment is performed using a 1 equivalent mixture of (R,S)- BnFTP and co-former and 7.5% mol of DBU. After each experiment, the solid in suspension was filtered and the cake washed with toluene. Tubing and reactor vessels were flushed with acetone, and solutions added to the filtrate. This latter was evaporated and the weight of the cake and solid recovered from the filtrate were determined. Both were further analyzed by chiral HPLC to check the ratio of (R)- vs. (S)-BnFTP. Combining these measurements allowed for a full mass balance. Results are given in table 3-4. In all cases, the cake was found to be enantiopure. Expectedly, starting with the (S)-co-former leads to crystallization of (S)-BnFTP while the (R)- enantiomer can be crystallized using the co-former of opposite handedness. Experiment 1, led to a recovered yield of over 50%, inherently implying that we went beyond mere resolution (max. yield of 50%). The full mass balance, showed a deracemization to have occurred (bold numbers) for all experiments as can be observed by looking at the total R/S ratio at the end of the experiments, with in the case of experiment 2, the 50/50 R/S mixture being thermodynamically transformed into a 87/13 mixture. From the follow-up of experiment 2 (Materials and methods), racemization kinetics were shown faster than crystallization kinetics. For this reason, heating was decreased for the third experiment with racemization becoming the limiting factor, yielding less deracemization over the same period of time. With these 3 experiments, we are the first to demonstrate that deracemization can be thermodynamically induced to yield enantiopure co-crystalline solid with high purity. Depending on the co-former’s handedness used, one can select the desired enantiomer. As both racemization and crystallization kinetics interplay, developing an optimized process requires future optimization of all process parameters.

Table 3-4 Key parameters and results for each of the three deracemization experiments. For each experiment, the same volume of toluene was used, 90mL. For the enantiomeric excess (ee), the enantiomer in excess is given in brackets. Yield is calculated with respect to the total mass retrieved at the end of the experiment. Run time is given as trun.

f Exp. C(BnFTP-PBA) VDBU / %mol Yield (cake) eecake eefiltrate Ratio R/S total Trac. trun 1 (S-PBA) 0.20 mol/L 203µL / 7.5% 50.7% 0.999 [S] 0.276 [S] 18/82 90°C 4 days 2 (R-PBA) 0.25 mol/L 254µL / 7.5% 44.6% 1.00 [R] 0.54 [R] 87/13 90°C 5 days 3 (R-PBA) 0.30 mol/L 305µL / 7.5% 38.7% 0.97 [R] 0.246 [S] 38/62 75°C 5 days

f It must be noted that both experiment 1 and 2 did not reach crystallization equilibrium before filtration and thermodynamic yields are expected to be higher. Before filtration, it would be advisable to leave the process enough time to equilibrate. 160 Chapter 3- Optimization of the CoISD process

2.4. Conclusion In conclusion, this section shows the feasibility of an innovative thermodynamic deracemization process coupling selective co-crystallization to a racemization reaction. We successfully deracemized (R,S)-BnFTP targeting either the (S)- or (R)- enantiomer based on choice of the co-former handedness. Unlike kinetic processes such as Viedma Ripening or dynamic preferential crystallization, CoISD can target conglomerates as well as racemic compounds, and contrary to CIDT, CoISD can be used for those compounds that do not form salts. Overall, this makes CoISD a general deracemization process, which in the future we will likely see applied to a multitude of compounds. For that purpose, the following section will detail the optimization of the process.

3. Toward an optimal CoISD process: the impact of operational parameters 3.1. Introduction Single-process deracemization processes offer considerable advantages over resolutions as they allow access to a theoretical 100% yield in the desired enantiomer, through conversion of the distomer in the eutomer.34-38 However, to reach high levels, deracemization processes require optimization. In the previous section, we demonstrated the feasibility of this Co-crystallization Induced Spontaneous Deracemization (CoISD) process, deracemizing 4,4-dimethyl-1-(4-fluorophenyl)-2- (1H-1,2,4-triazol-1-yl)-pentan-3-one (BnFTP) using 3-phenylbutyric acid (PBA) as the co-former and 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) as the base. The racemization reaction kinetics of this reaction were shown in the previous chapter to respond to first order reactivity in agreement with a keto-enol racemization scheme.39,40 Addition of the co-former, nevertheless had a significant impact on the overall process kinetics, with racemization no longer occurring at room temperature, imposing a racemization temperature of 90°C and leading to an overall yield of 51% and deracemization of 64%. The goal of this part is to further optimize this system aiming at reaching yields similar to Crystallization-Induced Diastereomer Transformation (CIDT) in the range from 70 to 90% with high enantiopurity.20,41 Reaching these numbers requires optimal combination of crystallization42 and racemization43 conditions. To do so, we studied the impact of operational parameters such as catalyst concentration, solvent nature and crystallization temperature on the overall process outcome,44-46 ultimately yielding an optimal process with an overall 80% deracemization and 73% enantiopure co-crystal yield. CoISD is therefore shown to display yields comparable to other deracemization methods. Furthermore, we are 161 Chapter 3- Optimization of the CoISD process paving the way toward further development of this methodology for other systems using the results of this optimization to give general optimization advice on CoISD optimization.

3.2. Materials and method 3.2.1. General procedure for all runs A specified equivalent of PBA (Eq PBA, table 3-5) was added to a stirring solution of specified concentration of (R,S)-BnFTP (CBnFTP, table 3-5) in 90mL of solvent (table 3-5) in the crystallization-cell, at 10°C. The pumps were turned on. The system was stirred for 1h to reach equilibrium. After that, the pumps were turned off and the racemization-cell was heated to 90°C and DBU (VDBU, table 3-5) was added to the racemization-cell with or without 150µL of water (table 3-5). When water was added, at day 1, the same volume was added again at day 3. Racemization alone ran for 2h. Then, the pumps were turned on for the rest of the day. Each evening, the pumps were turned off and the heater is either tuned off. The cooler was left as is except when specified otherwise (Tcryst, table 3-5). Each morning, the pumps were turned on and the heater put back at 90°C, after which the system was left to run for 8h. This was repeated each day until the end of the last day (Run time, table 3-5) where the pumps were turned off, the cooler put at a specific temperature (End Tcryst, table 3-5), the heater turned off and all the liquid phase transferred to the crystallization cell and let equilibrated over 2 days. Then, the mixture was filtered and washed with toluene at the end temperature of the crystallization cell.

The specific conditions for each run are given in table 3-5. Run 9 & 10 give the optimized procedure.

Table 3-5 The varying parameters from the general procedure for all runs.

Experiment CBnFTP Eq VDBU (µL) VWater Solvent [co-former] (mol/L) PBA Eq (mol%) (µL) Run 1 305 0.30 1 Toluene Unknown (S)-3-PBA 7.5 Run 2 305 0.30 1 Toluene / (S)-3-PBA 7.5 Run 3 305 150 0.30 1 Toluene (S)-3-PBA 7.5 Run 4 305 0.30 1 Propanol / (S)-3-PBA 7.5 Run 5 305 0.30 1 Butanol / (S)-3-PBA 7.5 Run 6 203 0.20 1 Toluene 150 (S)-3-PBA 7.5 162 Chapter 3- Optimization of the CoISD process

Run 7 305 0.30 1.2 Toluene 150 (S)-3-PBA 7.5 Run 8 610 0.30 1.25 Toluene 150 (R)-3-PBA 15 Run 9 1017 Run 10 0.30 1.25 Toluene 150 25 (R)-3-PBA Experiment End Ratio R/S Tcryst Run time E(cake) [co-former] Tcryst in filtrate (Run 1 10°C / 2 days / 28/72 (S)-3-PBA Run 2 10°C 0°C 5 days 100% 57/43 (S)-3-PBA Run 3 10°C 0°C 5 days 100% 43.5/56.5 (S)-3-PBA Run 4 10°C 0°C 5 days 95% 43/57 (S)-3-PBA Run 5 10°C 0°C 5 days 100% 51.5/48.5 (S)-3-PBA Run 6 10°C 0°C 5 days 98% 61/39 (S)-3-PBA Run 7 5 days 10°C 0°C 100% 60/40 (S)-3-PBA Run 8 10°C 0°C 5 days 100% 54/46 (R)-3-PBA Day 1: 10°C Run 9 Day 2: 5°C 100% 61.5/38.5 Run 10 -10°C 4 days Day 3: 0°C 97% 40/60 (R)-3-PBA Day 4: -5°C

3.2.2. Kinetic measurements All table values are given in appendix D.

Full curves with Toluene, Toluene-water, Propanol and Butanol 0.200g of (R-R)-BnFTP and PBA co-crystal was added to 3mL of solvent heated at 90°C in a POLAR Bear Plus Crystal device. Then, 20µL of DBU was added to the solution. Samples were taken overtime and directly quenched with HCl (1M). Ethyl acetate was added and the biphasic mixture was shaken two times. Then, the organic 163 Chapter 3- Optimization of the CoISD process layer was taken out and put in a vial to evaporate. The dried samples were then analysed by chiral HPLC with the same set-up as specified above.

Study of racemization kinetics of BnFTP and PBA co- crystal in Toluene-water mixtures The (S)-enantio-enriched mixture was obtained with the following protocol: 1.99g of (R)-PBA (0.0115mol) and 3.14g of (R,S)-BnFTP (0.0114mol) were added to 20mL of toluene. The resulting mixture was stirred overnight at RT, and then filtered. The cake and filtrate were recovered and dried. The filtrate was then analyzed by chiral HPLC. The filtrate is used as the (S)-enantio-enriched mixture.

The protocol for the kinetic measurements is the following: 0.3924g of this (S)- enantio-enriched mixture was added to a mixture of 3mL of toluene and 0 – 1.21 – 2.43 – 4.82 – 7.23 – 9.65 – 16.88 – 24.12 – 48.24 – 96.48 and 192.96L of water. The solution was heated to 90°C and 20L of DBU (15mol%) were added. Samples were taken overtime, directly quenched with HCl (1M) and treated the same way as for the full-curve measurements.

3.3. Results & discussion 3.3.1. The system

Figure 3-6 On top, from left to right, BnFTP, PBA, DBU and toluene. Below, 3D representation of the two-vessel set-up with on the left the co-crystal in suspension in the crystallization vessel and on the right the racemization vessel.

Figure 3-6 summarizes the principles of the deracemization process. Starting from a racemic BnFTP solution in toluene, addition of (R)-PBA, will lead to the selective crystallization of the less soluble (R-R)-co-crystal diastereomer, leaving the solution enriched in (S)-BnFTP. In principle, addition of a racemizing agent (DBU) would pull 164 Chapter 3- Optimization of the CoISD process

the solution enantiomeric imbalance back to equilibrium, decreasing the concentration of (S)-BnFTP in favor of (R)-BnFTP. This latter continues to co-crystallize, reinstalling the solution enantiomeric imbalance, and hence racemization continues. Even though this process could in principle run in a one-pot set-up, we encountered difficulties in combining crystallization and racemization conditions. For this reason, a two-vessel set-up was devised, with crystallization running in a vessel kept at 10°C and racemization running in a vessel placed at 90°C. The continuous transfer of solution between both vessels assures the continuity of the process.

Table 3-6 Results obtained during the process optimization, showing the overall yield Y and the overall deracemization Etot. The overall deracemization stands for the percentage of the unwanted enantiomer that was deracemized. It is calculated as the enantiomeric excess of the overall composition of (R) and (S)-BnFTP obtained at the end, combining both cake and filtrate. Ratio of each filtrate and cake are given as chromatograms in the order of the runs in appendix D, figures D-13 to D-22.

CBnFTP Eq CDBU (g/L) / Experiment Solvent Y Etot (mol/L) PBA %mol Toluene + Run 1 0.30 1 3.5 / 7.5 // 44%g water Toluene Run 2 0.30 1 3.5 / 7.5 33% 20% (dry) Toluene + Run 3 0.30 1 3.5 / 7.5 33% 41% water Run 4 0.30 1 Propanol 3.5 / 7.5 15% 26% Run 5 0.30 1 Butanol 3.5 / 7.5 30% 28% Toluene + Run 6 0.20 1 2.3 / 7.5 30% 14% water Toluene + Run 7 0.30 1.2 6.9/15 56% 47% water Run 8 55.5% 59% Toluene + Run 9 0.30 1.25 11.5 / 25 73% 80% water Run 10 80% 74%

Using this set-up, we previously observed (section 2.3), a negative impact of a decrease in racemization temperature, which will therefore remain fixed at 90°C for this section. The goal here being to evaluate how other parameters impact the process

g With no filtration done, the filtrate ratio represents the overall enantiomeric ratio. 165 Chapter 3- Optimization of the CoISD process performance, evaluated by the overall deracemization ratio (overall enantiomeric excess - cake and filtrate combined-, Etot), as well as the yield Y (amount of co-crystal recovered form the process, vs the total amount that can be recovered should there be a full deracemization and complete crystallization). Solvent nature, equivalents of base and acid and temperature of the crystallization cell were evaluated for their impact on the overall process. A first reference experiment (Run 1, table 3-6) was performed using a BnFTP (and PBA) concentration of 0.30M and a 7.5mol% concentration in DBU using toluene as a solvent, based on conditions used in the previous section. After having run the process for two days, a phase separation was observed, with a water phase appearing in the bottom of the racemization reactor.h The process was stopped, but to our surprise gave a very good overall deracemization (44%). This finding incited us to study the impact of small amounts of water on the overall process performance prior to studying all other parameters.

3.3.2. Water impact on the process To study the water impact, reactants were dried and the process run avoiding all traces of water (Run 2, table 3-6). This reference experiment unambiguously confirmed, a lower overall deracemization level (20%), even upon a prolonged run time (5 days instead of 2).

As a small amount of water seemingly has a positive impact on the overall process, we looked for its optimal value. To do so, the racemization rate constant was evaluated for various quantities of water added. As observed previously, the racemization reaction responds to first order kinetics even in presence of water (figure 3-7).

0 1 2 3 1 0 -0,5 0,8 -1 0,6 E -1,5 ln(E) [-] -2 0,4 [-] -2,5 0,2 -3 -3,5 0 0 1 2 3 -4 t [h] t [h] Figure 3-7 (left) Time evolution of the enantiomeric excess E in BnFTP of a (R)-BnFTP: (R)- PBA toluene solution with 0.00016%vol of water and 6.8g/L of DBU. (right) Time evolution of

h The water was present in the PBA reactant, and originated from the phase separation of PBA and Phenylethylamine during the resolution of PBA by diastereomeric salt formation. 166 Chapter 3- Optimization of the CoISD process

the natural logarithm of the enantiomeric excess E in BnFTP of a (R)-BnFTP: (R)-PBA toluene solution with 0.00016%vol of water and 6.8g/L of DBU (29mol%).

Figure 3-8 shows the impact of the amount of water on the obtained rate constants (original curves and linearization are given in appendix D, figures D-23 to D-26). Results show that for a water/toluene volume ratio of less than 0.16%, water has no significant impact on the racemization kinetics. However, a positive impact is seen as soon as phase separation occurs. This positive effect remains present up to about 0.32% of water present. The positive effect of water is therefore inherently linked to the phase separation, and can potentially be due to the presence of hydroxide in the aqueous phase, with a phase transfer phenomenon enhancing the reaction. However, this enhancement is no longer present when the amount of water becomes too important, which could be due to a too important amount of DBU-PBA pair moving to the water phase or dilution effects on hydroxide ions concentration.

0,9 0,8 0,7 0,6 k' 0,5 [h-1] 0,4 0,3 0,2 0,1 0 0 0.04 0.08 0.16 0.24 0.32 0.56 0.80 1.60 3.20 6.40 %vol water/toluene [-]

Figure 3-8 Value of the observed rate constant according to the %vol of water in toluene for the racemization of a (R)-BnFTP enriched solution in presence of 1 eq. of (R)-PBA and 6.8g/L of DBU (15mol%). The red line shows when phase separation occurs at 90°C. The error bar represents the experimental error (10%), previously determined.

Run 3 (table 3-6) was run under similar conditions as Run 2 but with a controlled amount of water added to the system in order to confirm its positive impact. After 2 days, de-mixing was no longer visible in the system, likely due to crystallization of the co-crystal, and another 0.16% of water added. Upon filtration, a similar yield of around 30% of enantiopure co-crystal was obtained with an overall deracemization of 41%, double the value obtained with respect to the run without water. For further process development we decided to set the added water to 0.16% of toluene volume, 167 Chapter 3- Optimization of the CoISD process value at which initial de-mixing occurs with an additional 0.16% added two days later, going up to an overall 0.32% of ratio water/toluene.

3.3.3. Alternative solvents In chapter 2 section 3.3.2, we showed protic polar solvents to increase the racemization rate. Furthermore, with water having an impact on the process, toluene is likely not the most optimal solvent for the racemization reaction. However, for our CoISD process, one should also take the impact of the solvent on the crystallization process (co-crystal solubility) into account. Too high solubility leads to a reduced yield, and associated overall deracemization. We therefore used the solubility curves determined in chapter 2 section 2.3.1 for the same 7 solvents used in the kinetic study and chose Butanol and Propanol as interesting alternatives.

Propanol (Run 4, table 3-6) and butanol (Run 5, table 3-6) were therefore tested on the system as alternative solvents to toluene, using similar conditions of overall concentration. 1-butanol and, to a lesser extent, 1-propanol led to more viscous solutions than toluene, which results in increased difficulties exchanging the solution between both vessels. The use of both solvents did not impact the enantiopurity of the resulting outcome, showing enantiopure cocrystal material. However, the yield (15 and 30% respectively) compared to the toluene/water mixture (33%), as well as the overall deracemization (26 and 28% respectively compared to 41%) are reduced. We therefore decided to continue with this latter solvent system, even though butanol remains an interesting alternative, likely requiring a longer running time.

3.3.4. DBU concentration We then turned to optimizing the concentration of DBU, as the kinetic study showed an exponential impact of DBU on the apparent rate constanti when working with substoichiometric amounts of DBU (chapter 2 section 3.3.3). This impact reflects in Run 6 (table 3-6), where a lower concentration of DBU was used.j The overall deracemization strongly decreased, from 41% to 14%. The reduced overall deracemization can be a reflection of the reduced racemization kinetics, or alternatively a shift in the racemization thermodynamics as will be discussed later on.

One can however, not increase the DBU equivalents (and hence concentration) freely, as this would imply a deprotonation of the acid available for co-crystallization and hence a strong impact on process yield. For this reason, runs 7 & 8 (table 3-6) were run increasing both DBU and PBA concentration in such a manner that at least one equivalent of unreacted acid is present to co-crystallize with BnFTP. The results showed a significant increase on the overall deracemization with respectively 47% and 59%, compared to run 3 (Etot = 41%). Regarding yield, both runs produced

i The apparent rate constant k’ includes the effect of the concentration in DBU. j Keeping a similar compound/DBU ratio, the overall concentration was reduced. 168 Chapter 3- Optimization of the CoISD process

roughly the same quantity of enantiopure co-crystal, overcoming the 50% maximum yield of resolution (Y = 56%). Yield was strongly increased compared to run 3 (Y = 33%).

These results highlight not only the kinetics but likely also the thermodynamics of the system is impacted by the overall BnFTP/PBA/DBU proportion. However, it is clear, that working with at least ‘one equivalent of unprotonated PBA seems beneficial to the yield and overall deracemization ratio while increasing DBU concentration increases the overall deracemization obtained for the same run time.

Running time of the process Additionally, for run 8, the enantiomeric excess of the overall liquid phase was followed from the addition of DBU to the last day (table 3-7). The evaluation was done sampling the liquid phase obtained from mixing both racemization and crystallization vessels for all points. Chromatograms are given in appendix D, figures D-27 & D-28.

Table 3-7 Follow-up of the deracemization run 8, (R)-PBA was used as co-former and (R-R)- co-crystal precipitated. The enantiomeric excess of the liquid phase (EϕL) is given regarding the major enantiomer, detailed in brackets [ ] and associated with a color code: blue for the (S)- enantiomer and orange for the (R)-one..

Before 2h after Start of End of Start of End of addition of addition of Day 2 day 2 day 3 day 3 DBU DBU

EϕL 0.56 [S] 0.13 [R] 0.29 [S] 0.13 [S] 0.13 [S] 0.01 [S] Start of day Start of End of End of day 4 Filtrate 4 day 5 day 5

EϕL 0.06 [R] 0.07 [R] 0.04 [R] 0.11 [R] 0.08 [R]

From table 3-7, it is clear that deracemization occurs throughout the three first days. Indeed, during those three days, the ratio of the liquid phase at the beginning of the day and at the end decreases in the unwanted (S)-enantiomer. Past the third day, the evolution of the enantiomeric excess did not clearly show further deracemization; suggesting the process had reached its end or was close to it. Furthermore, the strongest evolution in the enantiomeric excess are observed at the beginning (day 1 in

2h: ΔEϕL = 0.69; day 2 in 8h: ΔEϕL = 0.16; day 3 in 8h: ΔEϕL = 0.12). This is consistent with the theory behind the racemization. Indeed, the closer to the racemic mixture, the longer it takes to racemize. Finally, it can be seen from the point “2h after the addition of DBU” that deracemization occurred in the liquid phase since the enantiomeric excess goes the other way. 169 Chapter 3- Optimization of the CoISD process

Overall thermodynamics of racemization In order to confirm the presence of deracemization in the liquid phase, the same experiment as for the study of the impact of water was done but this time, the mixture was let to 90°C for several days, in order to investigate thermodynamics. After 3 days an enantiomeric excess of 0.08 [R] was obtained, showing an inversion in the major enantiomer. After 5 days, the enantiomeric excess did not change (0.07 [R]), showing deracemization/racemization should have reached the equilibrium. The enantiomeric excess at the equilibrium goes beyond the racemic mixture though not strongly higher than 0. Nevertheless, there was still a change in the major enantiomer compared to the starting enantiomeric excess (0.51 [S]). Interestingly, thermodynamically, the crystallizing enantiomer is also favored in solution as the major enantiomer when the system reaches its equilibrium ((R)-PBA was used for this experiment). This deracemization occurring in solution is a nice addition but is not significant enough to be used without the principle of this deracemization process. This just a little bonus, helping reaching a higher total deracemization.

3.3.5. Temperature of the crystallization cell Intuitively, decreasing the temperature of the crystallization cell is expected to lead to an increased yield, as the reduced solubility will lead to an increased amount of solid crystallizing out. As shown by figure 2-4 (Chapter 2 section 2.3.3), both diastereomers have a strong difference in solubility. Furthermore, the less stable co-crystal displays a sharper evolution of its solubility with temperature leading to an increased solubility difference between both diastereomers at high temperatures. To avoid crystallization of the undesired diastereomer at the onset of our process, we therefore decided to start the overall process keeping the crystallization cell at 10°C (run 9, table 3-6). We then gradually decreased the temperature by 5°C increments per day, reaching -10°C after 4 days. Compared to run 8, which with exception of the crystallization temperature, was run under similar conditions, an impressive increase in both yield (+17%) and overall deracemization (+21%) can be observed. Hence, excellent yield and deracemization were attained by resetting filtrate’s ratio each day upon decreasing temperature. This led to an ultimate process showing a 73% yield and 80% overall deracemization ratio.

Next, productivity was even further improved recycling the compound present in the filtrate. Indeed, the filtrate of run 9 was recovered and evaporated to dryness. Then, a racemic amount of BnFTP and an equivalent of (S)-PBA were added, corresponding to the amount of enantiopure co-crystal that was crystallized from run 9. Run 10 (table 3-6) was then run with the same conditions of run 9, without adding an additional amount of base (as DBU is not supposed to have been removed). The results of this run follow those of run 9, with a similar deracemization obtained for a higher yield, reaching 80%. The cherry on the cake would be that since the yield increased, if only considering the addition of racemic BnFTP and (R)-PBA, the yield of recycling 170 Chapter 3- Optimization of the CoISD process

overcomes 100%. Indeed, 8.73g of racemic BnFTP and (S)-PBA was added while 9.59g of (S-S)-co-crystal was crystallized. This result demonstrates the recyclability of the process, very important in industrial processes.

3.3.6. Productivity of the process Decreasing process time while not decreasing yield allowed for an efficient increase in productivity as it is the mass of co-crystal obtained per unit of time. For run 9 and k -1 -1 10, productivities could be calculated to be respectively 0.12KgKg RSCCday and -1 -1 0.18KgKg RSCCday when considering running time and crystallization time overnight. If only considering running time productivity increases respectively to -1 -1 -1 -1 0.55KgKg RSCCday and 0.82KgKg RSCCday . As a point of comparison, the best -1 -1 result of productivity we had before optimization was: 0.38KgKg RSCCday when only comparing the running time. Productivity was more than doubled when looking at run 10, highlighting the success of the optimization.

3.3.7. Summary Table 3-7 Table summing up the results of the optimization with general guidance rules.

Solvent CBnFTP CPBA Our Toluene with 2x0.16-% of 0.30mol/L 0.375mol/L system water The higher the concentration, Increasing concentration of co- Compromise between low co- the higher the yield. former compared to the target crystal solubility and fast However, too much solid in compound, increases yield. General racemization kinetics. The impact suspension can lead to A minimum of a 1:1 equivalent advice of water for acid/base catalysis difficulties with stirring is advised. Using higher ratios should be studied. (limitation, especially when is feasible, but leads to higher scaling-up the process) costs. [DBU] Trac Tcryst Our Decreased over time from 11.5g/L 90°C system 10°C to -10°C The higher the concentration of When not necessary, heating the racemizing agent, the faster should not be used. However, deracemization occurs. The lower this temperature, the if needed, it must be adapted General If, the racemizing agent reacts higher the expected yield. We to the kinetics of racemization guidance with the co-former, an excess of advise not to start too low but to of the system taking solvent co-former must be added to decrease temperature over time. and compound stability into maintain 1 equivalent of unreacted account. co-former in solution.

k 푚o×(퐸+1) Pr = with mo the mass of the output (Kg), E the enantiomeric excess of the output, 2푚i푡 mi the mass of the input (KgRSCC, RSCC meaning RS-co-crystal) and t the duration of the process (day).47,48 171 Chapter 3- Optimization of the CoISD process

The above table sums up the optimized values for all studied parameters, combining the results of all previous sections. In addition, general guidance is given for other future systems.

3.4. Conclusion Concluding, the aims to optimize the process conditions and improve productivity were successfully achieved after careful evaluation of the different critical parameters of the process. The importance of water inside the process was highlighted showing the importance of controlling the volume of water inside, since it can increase or decrease racemization rate depending on the ratio. Toluene was proven to be the most suitable solvent combining good racemization kinetics with excellent crystallization of the co-crystal while protic solvents were shown to be less attractive regarding racemization as one could have thought. The importance of a minimum of 1 equivalent of acid, unreacted with the base, was demonstrated upon an important increase of the yield while increasing base concentration improved overall deracemization in correlation with kinetic results. Temperature of crystallization cell was also shown to improve yield when decreased, especially over the duration of the process instead of keeping it constant. Regarding duration, we showed that under optimized conditions, the process does not need to run for 5 days. After optimization, the process provided pure co-crystal with a yield of 73% and a deracemization of 80% after 4 days of run. Finally, the use of recycling is recommended to improve productivity while decreasing the quantity of co-former and base needed in total.

4. Bibliography 1. Murakami, H. From Racemates to Single Enantiomers – Chiral Synthetic Drugs over the last 20 Years. In: Novel Optical Resolution Technologies. Heidelberg: Springer, 274-278 (2006). Available at: https://link.springer.com/content/pdf/10.1007%2F978- 3-540-46320-7.pdf. Accessed September 5, 2018. 2. Administration F.D. (1992). Available at: www.fda.gov/Drugs/GuidanceComplianceRegulatorylnformation/Guidances/ucm12 2883.htm. Accessed September 5, 2018. 3. Calcaterra, A., D’Acquarica, I. J Pharm Biomed Anal. 147, 323–340 (2018). 4. Xie, R., Chu, L.-Y., Deng, J.-G. Chem. Soc. Rev. 37, 1243-1263 (2008). 5. Whitesell, J. K., Reynolds, D. J. Org. Chem. 48, 20, 3548-3551 (1983). 6. Xie, S.-M., Zhang, Z.-J., Wang, Z.-Y., Yuan, L.-M. J. Am. Chem. Soc. 133, 31, 11892-11895 (2011). 7. Noorduin, W. L., van der Asdonk, P., Meekes, H., van Enckevort, W. J. P., Kaptein, B., Leeman, M., Kellogg, R. M., Vlieg, E. Angew. Chem. Int. Ed. 48, 3278 –3280 (2009). 8. Okamoto, Y., Ikai, T. Chem. Soc. Rev. 37, 2593-2608 (2008). 9. Coquerel G. Preferential Crystallization. In: Novel Optical Resolution Technologies. Heidelberg: Springer, 4-18 (2006). Available at: 172 Chapter 3- Optimization of the CoISD process

https://link.springer.com/content/pdf/10.1007%2F978-3-540-46320-7.pdf. Accessed September 5, 2018. 10. Levilain, G., Coquerel, G. CrystEngComm. 12, 7, 1983-1992 (2010). 11. Rougeot, C., Hein, J. E. Org. Process Res. Dev. 19, 12, 1809-1819 (2015). 12. Kozma, D. CRC handbook of optical resolutions via diastereomeric salt formation. CRC Press (2001). 13. Marchand, P., Querniard, F., Cardinaël, P., Perez, G., Counioux, J. J., Coquerel, G. Tetrahedron-Asymmetry. 15, 16, 2455-2465 (2004). 14. He, Q., Peng, Y. F., & Rohani, S. Chirality., 22, 1, 16-23 (2010). 15. Sögütoglu, L.-C., Steendam, R.R. E., Meekes, H., Vlieg, E., Rutjes, F. P. J. T. Chem. Soc. Rev. 44, 6723-6732 (2015). 16. Viedma, C. Phys. Rev. Lett. 94, 3−6 (2005). 17. Sakamoto, M., Mino, T. Total Resolution of Racemates by Dynamic Preferential Crystallization. In: Tamura R., Miyata M. (eds) Advances in Organic Crystal Chemistry. Tokyo: Springer (2015). 18. Toda, F., Tanaka, K. Chem. Lett. 661 (1983). 19. Hassan, N. A., Bayer, E., Jochims, J. C. J. Chem. Soc. 3747 (1998). 20. Brands, K. M. J., Davies, A. J. Chem. Rev. 106, 7, 2711–2733 (2006). 21. Andrew D. Bond. Cryst. Eng. Comm. 9, 833–834 (2007). 22. Thorey, P., Bombicz, P., Szilágyi, I. M., Molnár, P., Bánsághi, G., Székely, E., Simándi, B., Párkányi, L., Pokol, G., Madarász, J. Thermochim. Acta. 497, 129–136 (2010). 23. Sanchez-Guadarrama, O., Mendoza-Navarro, F., Cedillo-Cruz, A., Jung-Cook, H., Arenas-García, J. I., Delgado-Díaz, A., Herrera-Ruiz, D., Morales-Rojas, H., Hopfl. H. Cryst. Growth Des. 16, 307−314 (2016). 24. Springuel, G., Norberg, B., Robeyns, K., Wouters, J., Leyssens, T. Cryst. Growth. Des. 12, 1, 475-484 (2012). 25. Springuel, G., Leyssens, T. Cryst. Growth Des. 12, 7, 3374–3378 (2012). 26. Balasubramanyan, S. (1981). Fungicidal compounds. US4243405A. 27. Suwannasang, K., Flood, A.E., Rougeot, C., Coquerel, G. Cryst. Growth Des. 13, 3498-3504 (2013). 28. Rougeot, C., Guillen, F., Plaquevent, J.C., Coquerel, G. Cryst. Growth Des. 15, 2151- 2155 (2015). 29. Habgood, M. Cryst. Growth Des. 13, 10, 4549–4558 (2013). 30. Lam, A. W. H. and Ng, K. M. AlChE J. 53, 429-437 (2007). 31. Reusch W., Farmer S. (2015). Available at: https://chem.libretexts.org/Textbook_Maps/Organic_Chemistry/Supplemental_Mod ules_(Organic_Chemistry)/Reactions/Rearrangement_Reactions/Keto- Enol_Tautomerism. Accessed September 10, 2018. 32. Davidson, E. M., Turner, E. E. J. Chem. Soc. 843 (1945). 33. Shiraiwa, T., Furukawa, T., Tsuchida, S., Sakata, M., Sunami, M., Kurokawa, H. Bull. Chem. Soc. Jpn. 64, 3729 (1991). 34. Oketani, R., Hoquante, M., Brandel, C., Cardinael, P., Coquerel, G. Cryst. Growth Des. 18, 11, 6417-6420 (2018). 35. Gruber, C. C. Adv. Synth. Catal. 348, 1789–1805 (2006). 36. Noorduin, W. L. et al. Angew. Chem. Int. Ed. 48, 4581–4583 (2009). 173 Chapter 3- Optimization of the CoISD process

37. Rubin, A. E., Tummala, S., Both, D. A., Wang, C., Delaney, E. J. Chem. Rev. 106, 2794–2810 (2006). 38. Wood, W. M. L. Crystal science techniques in the manufacture of chiral compounds. in Chirality in industry II: Developments in the commercial manufacture and applications of optically active compounds, (eds. Collins, A. N., Sheldrake, G. N. & Crosby, J.) 119–156 (1997) (John Wiley and Sons, Inc). 39. Breveglieri, F., Mazzotti, M. Cryst. Growth Des. 19, 3551−3558 (2019). 40. Wolf, C. Racemization, enantiomerization and diastereoisomerization. in Dynamic Stereochemistry of Chiral Compounds: Principles and Applications, Royal Society of Chemistry, 29–34 (2008). 41. Anderson, N. G. Org. Process Res. Dev. 9, 800–813 (2005). 42. Gagnière, E. et al. J. Cryst. Growth. 311, 2689–269 (2009). 43. Harmsen, B., Leyssens, T. Cryst. Growth Des. 18, 3654–3660 (2018). 44. Ou, L., Xu, Y., Ludwig, D., Pan, J., Xu, J. H. Chemoenzymatic Deracemization of Chiral Secondary Alcohols : Process Optimization for Production of (R)-1-Indanol and (R)-1-Phenylethanol. 12, 192–195 (2008). 45. Ji, Y., Shi, L., Chen, M., Feng, G., Zhou, Y. J. Am. Chem. Soc. 137, 2–5 (2015). 46. Li, W. W. et al. Cryst. Growth Des. 16, 5563−5570 (2016). 47. Dunn, A. S., Svoboda, V., Sefcik, J. & Horst, J. H. Resolution Control in a Continuous Preferential Crystallization Process. Org. Process Res. Dev. 23, 2031–2041 (2019). 48. Lorenz, H. & Seidel-morgenstern, A. Processes To Separate Enantiomers. Angew. Chem. Int. Ed. 53, 1218–1250 (2014). 174 Chapter 4- Valorization of BnFTP

Chapter 4 - Valorization of BnFTP

175 Chapter 4- Valorization of BnFTP 1. Overview The goal of this thesis has finally been reached, however, there still remains one aspect to treat. We used a model compound to develop the process as a proof of concept: BnFTP. This compound is an analog of BnClTP, which was already stated to be the precursor of Paclobutrazol, a fungicide and plant grow inhibitor. Because of this, being able to reduce the ketone into an alcohol while keeping the enantiopurity of the chiral carbon is of interest. This chapter will cover this part showing the possibility to reduce the compound without racemizing the deracemized carbon and inducing a preferred configuration of the so-formed new chiral center.

2. Introduction BnFTP, is only varying from BnClTP in the nature of the atom in the para-position to the phenyl group, a chlorine for BnClTP and a fluorine for BnFTP. BnClTP is the oxidized precursor of Paclobutrazol.1 The racemic mixture of Paclobutrazol possesses two interesting biological properties: it is a plant growth regulator and a demethylation inhibitor (DMI) fungicidal agent.1-3 As Paclobutrazol is composed of two asymmetric centres and is used in racemic form, two different enantiomers (2RS,3RS) – as well as two diastereomers (2RS,3SR) – exist (Figure 4-1).

Figure 4-1 Enantiomeric and diastereomeric relations between the different stereoisomers of Paclobutrazol. 176 Chapter 4- Valorization of BnFTP

Analyses on the resolved compound reveal that the (2S,3S)-enantiomer is characterized by the plant growth regulator activity while the (2R,3R)-form by the DMI fungicidal activity. As a result, their racemic mixture showed high activity for both. In the case of the racemic mixture of the two other diastereomers, low activity is observed towards both aspects.2 This confirmed the importance of chirality and the interest in having enantiopure material. Regarding BnFTP, it can be expected to possess similar activity since the activity of Paclobutrazol can be directly linked to its triazole function that BnFTP also possesses. Indeed, triazole compounds are known to display DMI fungicidal activity. There activity occurs via inhibition of the biosynthesis of ergosterol in fungi, which is a major component of their fungal cell membranes and necessary for their growth. By disrupting their membrane, their permeability increases which leads to their death.4-7 Triazole compounds act on the enzyme 14 alpha-demethylase or CYP51, which is responsible for the oxidation of the C-14 methyl group in Lanosterol (figure 4-2).8

Figure 4-2 Lanosterol being transformed after several biological steps into ergosterol. The first step involve CYP450 oxidation of methyl C14 (in red) which is replaced by a hydrogen in ergosterol.

Regarding the plant growth regulatory features of Paclobutrazol, it is believed to result from a blockage in the conversion of ent-kaurene to ent-kaurenol in the biosynthesis of gibberellins (figure 4-3).9 Gibberellins are generally biosynthesized through the methylerythritol phosphate (MEP) pathway. Conversion of ent-kaurene to ent- kaurenol is step 3.10

Figure 4-3 Biosynthesis of ent-kaurenol from ent-kaurene under the action of ent-kaurene oxidase. ent-Kaurenol is then transformed into gibberellin A12 (Ga12) after several steps. Ga12 is the precursor of all other gibberellins.

High activities of the (2R,3R)-enantiomer regarding fungicidal activity and of the (2S, 3S)-enantiomer regarding plant growth inhibition were studied using computer modelling. The results showed that each enantiomer can be superimposed on 177 Chapter 4- Valorization of BnFTP respectively Lanosterol for the (2R,3R)-enantiomer and Kaurene for the (2S,3S)- enantiomer when they bind to the iron atom of P450 phorphyrin (figure 4-4).2 The interaction of Paclobutrazol enantiomers with the metal occurs via the N-4 atom of the triazole, the same atom giving the hydrogen bond for the co-crystallization of BnFTP and BnClTP with co-formers.

Figure 4-4 (A) Binding of lanosterol (full line) and the (2R,3R)-enantiomer of Paclobutrazol (doted line) to cytochrome P450. (B) Binding of kaurene (full line) and the (2S,3S)-enantiomer of Paclobutrazol (doted line) to cytochrome P450. Figure originally from reference 2.

As both compounds are closely related and especially looking at the interaction of

Paclobutrazol, the reduced version of BnFTP (BnFTH2P) could also be expected to show similar activities. Consequently, we expect the reduced BnFTP to offer potential for several applications. Indeed, once the ketone is reduced, an alcohol group is formed with a second asymmetric carbon. First, the chirality of this newly formed chiral center would probably depend on the chirality of the other carbon that is directly connected (asymmetric induction).11 Second, an alcohol function can be transformed into an ester,12-14 ether15-17 or halogen group.18 This latter offers multiple possibilities 19 to do SN2 reactions that are stereospecific or carbon-carbon coupling giving access to more complicated derivatives, turning BnFTH2P into an enantiopure building block for synthesis. For all these reasons, BnFTP reduction was first investigated in order to find a suitable reaction reducing the ketone without racemizing the enantiopure carbon. Then, its diastereoselectivity toward different types of reductive agent was studied.

3. Materials and methods Column chromatography was performed using silica gel 60 Å (40-63 μm). (R,S)- BnFTP was synthetized according to the protocol described in chapter 1. (R)-BnFTP was resolved from (R,S)-BnFTP by the CoISD process and then isolated using the protocol described in chapter 2. Cu(II)-2-ethylhexanoate was purchased from commercial sources and used as received. NMR details can be found in chapter 1 section 2.2. 178 Chapter 4- Valorization of BnFTP

4.1. Reduction of BnFTP 4.1.1. By copper catalysis with BDP 7.0mg of Cu(II)-2-ethylhexanoate (0.020mmol, 0.02 equivalent) and 8.9mg of BDP (0.020mmol, 0.02 equivalent) were added in a dry schlenk, filled with Ar. 2mL of distilled toluene was added and the mixture was stirred at RT for 15 minutes. The mixture took a green-blue colour. 2mL of methyldiethoxysilane (1.248mmol, 1.25 equivalent) were added dropwise under stirring. The mixture took a yellow colour.

275mg of (R,S)-BnFTP were added in 2mL of distilled toluene in a dried flask filled with Ar. Then the mixture was added to the Schlenk. The mixture was stirred for 22h at RT. The reaction was quenched with 1.5mL of NaOH (5% in MeOH) and the resulting mixture was stirred for 30 minutes. The mixture was filtered on a silica pad (eluent: EP/EtOAc, 7/3) then concentrated. Three different fractions were harvested.

The product was found in fractions two and three. mproduct = 0.1930g; Yield = 70%.

1H NMR (300 MHz, Chloroform-d) anti product δ (ppm): 0.68 (s, 9H), 3.11 (dd, 1H,

J = 14.0, 5.4 Hz, CH-CHAHB-CAr), 3.36 (dd, 1H, J = 13.9, 9.8 Hz, CH-CHAHB-CAr), 3.50 (d, 1H, J = 1.7 Hz, HO-CH-CH), 4.43 (ddd, 1H, J = 9.7, 5.6, 1.7 Hz, CH-CH-

CH2), 6.86 (s, 2H, CArH), 6.88 (m, 2H, CArH), 7.80 (s, 1H, N-CH-N), 7.93 (s, 1H, N- CH-N)

1H NMR (300 MHz, Chloroform-d) syn product δ (ppm): 1.02 (s, 9H), 3.20 (dd, 1H,

J = 14.6, 10.9 Hz, CH-CHAHB-CAr), 3.28 (dd, 1H, J = 14.7, 4.0 Hz, CH-CHAHB-CAr),

3.73 (d, J = 3.0 Hz, 1H), .97 (ddd, 1H, J = 10.7, 4.0, 3.0 Hz, CH-CH-CH2), 6.73 (m,

2H, CArH), 6.82 (m, 2H, CArH), 7.55 (s, 1H, N-CH-N), 7.92 (s, 1H, N-CH-N).

The same protocol was carried out for (R)-BnFTP, yielding 75% of product for a mass of 0.104mg starting with 137.6mg of BnFTP.

4.1.2. By copper catalysis with BINAP 7.3mg of Cu(II)-2-ethylhexanoate (0.021mmol, 0.02 equivalent) and 12.7mg of enantiopure BINAP (0.020mmol, 0.02 equivalent) were added in a dry schlenk, filled with Ar. 2mL of distilled toluene was added and the mixture was stirred at RT for 15 minutes. The mixture took a green-blue colour. 2mL of methyldiethoxysilane (1.25mmol, 1.25 equivalent) were added dropwise under stirring. The mixture took a yellow colour.

275mg of BnFTP (1mmol, 1 equivalent) were added in 2mL of distilled toluene in a dried flask filled with Ar. Then the mixture was added to the Schlenk. The mixture was stirred for 4 days at RT, then at 50°C for 2 days. The reaction was quenched with 1.5mL of NaOH (5% in MeOH) and the resulting mixture was stirred for 30 minutes. The mixture was filtered on a silica pad (eluent: EP/EtOAc, 75/25) then concentrated. A mixture of both product and starting material was retrieved and analysed by NMR. 179 Chapter 4- Valorization of BnFTP

4.1.3. With NaBH4

275mg of BnFTP (1mmol, 1 equivalent) and 63mg of NaBH4 (1.7mmol, 1.7 equivalent) were added in 11mL of MeOH in a dry flask, filled with Ar. The mixture was stirred for 2h30 at RT and then quenched with distilled water. The residue was extracted 3 times with ethyl acetate. The organic layers were combined, dried over

MgSO4, filtered and concentrated. White crystals were obtained. mproduct = 0.207g; Yield = 75%.

4.1.4. With LiAlH4

65mg of LiAlH4 (1.7mmol, 1.7 equivalent) were added in 8mL of anhydrous THF in a dry flask, filled with Ar. The stirring solution was then cooled to 0°C and a solution of 275mg of BnFTP (1mmol, 1 equivalent) in 3mL of anhydrous THF added dropwise. At the end of the addition, the mixture was allowed to warm up to RT.

The mixture was stirred for 2 hours at RT. At the end of the reaction, the resulting mixture was cooled to 0°C and quenched with distilled water. The residue was extracted 3 times with DCM. The organic layers were combined, dried over MgSO4, filtered and concentrated. White crystals were obtained. mproduct = 0.184g; Yield = 67%.

4.1.5. With Me4NBH4 275mg of BnFTP dissolved in 2mL of distilled dichloromethane (DCM) were added at 0°C to a stirred suspension of 44.5mg (0.5eq) of Me4NBH4 in 3mL of distilled DCM. Then, 2mL of acetonitrile were added and the reaction ran for 3 days at room temperature. At the end, the reaction was quenched with 2M HCl and the residue was extracted 3 times with DCM. The organic layers were combined, dried over MgSO4, filtered and concentrated. A mixture of both products and starting material was retrieved and analyzed by NMR.

4.2. Chiral HPLC The column used was a Chiral Pak IB chiral column with the same dimensions. The mobile phase was 98% isohexane and 2% ethanol at a flow rate of 1 mL/min.

4. Results and discussion 4.1. Finding a suitable reduction Reducing enantiopure BnFTP was not trivial as the most commonly used reductive agents for ketone like NaBH4 and LiAlH4 are basic. Indeed, they are known to react more or less strongly with water, alcohol,… Unfortunately, BnFTP was shown in the second chapter to be quite sensitive to basic conditions when it comes to racemization.

Consequently, though probably well working as reductive agents, NaBH4 and LiAlH4 are not suitable for this typical reaction. Then, the idea came to look for a catalyzed 180 Chapter 4- Valorization of BnFTP reaction, where conditions can be more neutral. The following reaction was found as promising (Scheme 4-1) :20

Scheme 4-1 Reaction of catalyzed hydrosilylation of a ketone, yielding its reduction to the alcohol after hydrolysis.

O R O Cu O R O

R3Si-H

activation

O O R3Si O R R

Scheme 4-2 Catalytic cycle of the reduction of (R,S)-BnFTP into BnFTH2P by hydrosilylation with Cu, adapted from reference 20. The Cu(II) pre-catalyst (green-blue) is converted via a σ- bond metathesis with methyldiethoxysilane and a ligand exchange with diphosphine forming an active catalyst: the Cu(I) hydride species - 4 (yellow). After that, the first step of the catalytic cycle lies in the reaction between the active catalyst with ketone - 1 - to form the copper alkoxide - 5. A second σ-bond metathesis occurs between the copper alkoxide and the hydrosilylating 181 Chapter 4- Valorization of BnFTP reagent - 2 - to produce the desired silyl ether - 3 - and regenerating the active Cu(I) hydride catalyst – 4.

This reaction was seen promising for two reasons:

First, the reaction uses a catalytic cycle where no racemization should occur (Scheme 4-2). The only base is used for hydrolyzing the silane (NaOH in methanol), once the reaction is considered finished. Second, the reaction was well working for a compound, benzyl acetone, similar in structure as BnFTP. Though more complex and bulkier, BnFTP fully possesses the skeleton of the benzyl acetone. This elevated the chances of the reaction to be working for BnFTP.

First of all, the reaction was tried on (R,S)-BnFTP to see if it could effectively reduce the ketone. The reaction was carried out with 1,2-Bis(diphenylphosphino)benzene (BDP) as the ligand and the protocol was followed as described (experiment 1). This reaction was complete after 1 day of reaction and yielded both diastereomers of 1 BnFTH2P with a yield of 70%. Via the analysis of the H NMR spectrum of both diastereomer combined, it appeared that one diastereomer was preferentially formed, leading to a de (diastereomeric excess) of 45%. The major diastereomer was later shown to be the anti diastereomer (Scheme 4-3) i.e. the diastereomer preferentially forming from the Felking-Ann model (c.f. General introduction 2.2.1 figure 24 & here figure 4-7). Similarly, the syn diastereomer is the one forming preferentially from the Cram Chelate model (c.f. General introduction 2.2.1 figure 24 and here figure 4-6).

Scheme 4-3 Reduction of (R,S)-BnFTP possibly leading to the formation of 4 different products: two diastereomers (anti and syn) with each 2 enantiomers.

Then, the same reaction was carried out on (R)-BnFTP, with the exact same protocol (experiment 2). The reaction was complete after 8h and gave a yield of 75%. The 1H NMR spectrum of the product, confirmed preferential formation of a diastereomer 182 Chapter 4- Valorization of BnFTP over the other with the exact same de of 45%. Chiral HPLC analysis was carried on the product of both experiments (figure 4-5). The first, clearly shows 4 peaks, corresponding to 4 stereoisomers while on the second, only one peak of each diastereomer is present, clearly proving no racemization occured while reducing the ketone. As a result, this reduction of BnFTP can be used on (R)-BnFTP to access

(2R,3S)- BnFTH2P [2] and (2R,3R)-BnFTH2P [4] . By analogy, (2S,3R)- BnFTH2P

[1] and (2S,3S)-BnFTH2P [3] can be accessed by the same reaction carried out on S- BnFTP.

Figure 4-5 On the left chromatogram of the 4 peaks of each of BnFTH2P with the enantiomers of the syn (1 & 2) and the enantiomers of the anti (3 & 4), obtained from reduction of (R,S)- BnFTP. The identity of each peak was attributed using the diastereomeric ratio of the reaction and the result of reduction of (R)-BnFTP. Indeed, in the right chromatogram obtained after reduction of (R)-BnFTP, showing only one peak of each enantiomer can be seen compared to the chromatogram obtained from reduction of (R,S)-BnFTP. Then, the bigger peak should correspond to the anti diastereomer, being enantiopure it can only be (2R,3R)-BnFTH2P. The other is the syn diastereomer with the (R)-configuration of the carbon in position 2, (2R,3S)- BnFTH2P.

4.2. Diastereoselectivity of the reduction

Once BnFTH2P was successfully synthesized by reducing the carbonyl moiety of a racemate and enantiopure BnFTP was then reduced into the corresponding alcohol (Scheme 4-3) retaining the absolute configuration of the first stereogenic centre, the diastereoselectivity of the reaction was studied. The reduction reactions achieved in this part were all done on racemic BnFTP as only diastereoselectivity was of matter.

Firstly, copper-catalyzed reductions of (R,S)-BnFTP was carried out by hydrosilylation with the second ligand, 2,2'-bis(diphenylphosphino)-1,1'-binaphtyle (BINAP), to see if the ligand influences the diastereoselectivity of the reaction. Both (R)- and (S)-BINAP were tried (experiment 3 & 4) to also study the influence of the absolute configuration of the ligand on the diastereomeric selectivity. Both reactions were carried out in the exact same conditions but were not finished after one day, as they appeared to run slower. Indeed, after stirring both experiments at RT for 4 or 3 days respectively, they were still not finished. The reaction temperature was then 183 Chapter 4- Valorization of BnFTP increased to 50°C for two more days for each reaction. Even with this, the yields of the reduction were low. This could be explained by the higher steric hindrance of the BINAP ligand compared to the BDP ligand in the complex. As BnFTP already contains bulky triazole and fluoro-benzyl functions, it is likely less easily attached to the catalyst, decreasing the kinetics of the reaction. Regarding diastereoselectivity, both reactions produced preferentially the same diastereomers as experiments 1&2, moreover the de, determined by 1H NMR, were found to be similar to that of experiments 1&2. Consequently, the ratio of the diastereomers formed did not seem to depend on the chirality of the ligand or on the nature of it. However, the nature did influence the kinetics of reactions, severely.

Figure 4-6 Newman projections of the Cram chelate model applied to the reduction of (R)- BnFTP yielding the anti (2R,3S) product.

Regarding the nature of the preferred diastereomer, the envisaged mechanism of the hydrosilylation reaction was originally thought to be the Cram chelate model21 because of the chelating nature of cupper. This mechanism would lead to the syn product. Indeed, the copper catalyst may chelate with BnFTP by the oxygen of its carbonyl and the N-2 of its triazole. The nucleophile attacks from the less sterically hindered side following the Burgi-Dunitz trajectory22 and this way forms the syn product, (2R,3S)- or (2S,3R)-BnFTH2P, as the major compound (figure 4-6).

In a second time, to see if common sources of hydride nucleophile would lead to opposite results than with the copper-catalyzed reactions, the reduction of racemic

BnFTP was carried out with NaBH4 which should favour the anti product, via the Felkin-Ahn model (figure 4-7).21 184 Chapter 4- Valorization of BnFTP

Figure 4-7 Newman projections of the Felkin-Ahn model applied to the reduction of (R)-BnFTP yielding to the anti (2R,3R) product.

The reaction led to the formation of the same preferred diastereomer, but with a better diastereoselectivity (de = 75%). This result was unexpected and led to think that the mechanism with copper could also be Felkin-Ahn. Another experiment, with LiAlH4 this time, was carried out. Indeed, LiAlH4 is a strong reducing agent, despite the possibility of Li+ to chelate, the kinetics should outweigh the thermodynamics. Again, the reaction gave the same preferred diastereomer with a de of 67%. As a result, the diastereoselectivity of the reaction is greater with hydride . These reductions occur in basic conditions. It can therefore be expected that racemization of BnFTP occurs during the reactions and they were not appropriate for the reduction of enantiopure BnFTP.

Table 4-1 Summary of the different conditions tried to reduce (R,S)-BnFTP. The same diastereomer was always formed. De were determined by 1H NMR. For (R)-BINAP, (S)-BINAP and Me4NBH4 reactions were not complete, hence only their ratio product/starting material was calculated using the NMR ratio.

Ratio Experiment Reducing agent Catalyst Ligand Yield de P/SM

1 MeOEt2Si Cu(I) BDP 70% / 45% anti

3 MeOEt2Si Cu(I) (R)-BINAP / 0.47 45% anti

4 MeOEt2Si Cu(I) (S)-BINAP / 0.22 46% anti

5 NaBH4 / / 75% / 82% anti

6 LiAlH4 / / 67% / 69% anti

7 Me4NBH4 / / / 0.70 90% anti

Finally, a more hindered borane was used in order to see if this would shift the ratio one way or another. It resulted in a slower reactivity compared to NaBH4 (reaction not complete after 3 days), just like with BINAP. Furthermore, it resulted in a higher 185 Chapter 4- Valorization of BnFTP de in anti product (90%). All the reduction conditions attempted on racemic BnFTP are summarised in Table 4-1.

In order to know with certainty which diastereomer was the most favourably formed, a single crystal X-ray diffraction (SC-XRD) was performed on the large crystals obtained from experiment 5. The product of experiment 5 was chosen for this analysis as, from all the experiments; it gave the largest crystals and the second best de. The asymmetric unit can be seen in figure 4-8.

Figure 4-8 Asymmetric unit of (2S,3S)-BnFTH2P, the anti-product.

As shown by the SC-XRD, the diastereomer preferentially formed in every reaction is the anti, a racemic mixture (2S,2S) and (2R,2R) when (R,S)-BnFTP is used. To explain the diastereoselectivity of all these reactions, the Cram chelate and the Felkin- Ahn models will be further discussed.

+ The high stereo selectivity of NaBH4 is explained because Na is not a good chelating metal. The reaction cannot follow the Cram chelate model, there are no competitions between the Cram chelate and Felkin-Ahn models, the formation of the anti product is thus strongly favoured. This result is in accord with literature where NaBH4 is generally known to favour the anti-product following the Felkin-Ahn model, for 23 instance, in the reduction of α-hydroxyketone. In the case of amino ketones, NaBH4 again followed the Felkin-Ahn model, though this model produces the syn product in this case where connotation are reversed.24

Then, LiAlH4 could undergo a competition between the Cram chelate and Felkin-Ahn + models. Indeed, Li may act as a chelating metal. However, LiAlH4 is a strong reducing agent, the kinetics outweighing the thermodynamics. Moreover, due to the steric hindrance of BnFTP, the chelation between BnFTP and Li+ is not so favourable. Thus, the Felkin-Ahn model is likely the favoured mechanism. It leads to the formation of the anti product. Small competition with Cram-chelate would explain the lower diastereoselectivity of the reduction with LiAlH4. This result also concords 186 Chapter 4- Valorization of BnFTP with literature since DFT calculation showed that the syn product should be favoured in the reduction of a similarly hindered molecule, 2-o-tert-butylphenoxylpropanone, though experiments showed high selectivity in the anti product.25 On α- 23 hydroxyketone, LiAlH4 also gave similar results as NaBH4 as observed here. Furthermore, the same anti selectivity was observed on closely related molecules26, though the diastereoselectivity was low.

Finally, the hydrosilylation reactions should also promote the Cram chelate model, leading to the formation of the syn product, as copper is a good chelating metal. Once again, due to the steric hindrance of BnFTP, the chelation between BnFTP and copper is less favourable as expected but still occurring as purification showed. Indeed, removing the cupper by column chromatography did not fully work when the mixture was deposited on the column.a A second run was necessary to remove all cupper (light blue colour was visible and NMR displayed large peaks as expected in the presence of Cu(II)). Furthermore, by looking at the NMR obtained with cupper, and when compared with that without cupper, most of the peaks are at the same place, with only resolution being affected except for the triazole peaks, which indicates an interaction with cupper. Because of the strong interaction between cupper and the triazole, competition between the Cram chelate and Felkin-Anh models was stronger, inducing more syn product, thus decreasing the diastereoselectivity of the reaction.

Interestingly, the (2S,3S) and (2R,3R)-Paclobutrazol are the more active diastereomers and enantiomers and by analogy those of BnFTH2P are expected to show similar activities. However, if the two other forms were to be interesting, obtaining the syn product as major product could be carried out by trying the reaction with a catalyst that is less sterically hindered to facilitate the insertion of BnFTP on the catalyst and the chelation between BnFTP and metal. If enantiopurity is not important, borohydrides with a larger counter cation could be used as suggested by Gernot Reiϐenweber and Al.26 Indeed, using tetra-n-butylammonium as a counter cation allowed furnishing the syn product with a de of 80% for closely related compounds, including BnTP (figure 4-9). This only worked without the use of TiCl4, since its presence reverses the diastereoselectivity of the reaction. They also tried tetramethylammonium borohydride, though with TiCl4, and it gave a high diastereoselectivity in the anti product.

a Doing a silica pad allows for complete separation in one iteration. 187 Chapter 4- Valorization of BnFTP

Figure 4-9 Structure of the related compounds reduced with tetra-n-butyl ammonium borohydride, yielding the syn diastereomer as the major product.

5. Conclusion To conclude, in this chapter, the reduction of BnFTP was successfully carried out using a catalytic hydrosilylation reaction. Furthermore, no racemization occurred, allowing the formation of two enantiopure diastereomer with the anti being the preferred one. Then, the diastereoselectivity of the reduction was further studied, showing the preference in formation of the anti product no matter the conditions while highlighting competition between Felkin-Ann and Cram-Chelate in the hydrosilylation reaction mechanism, explaining the lower diastereoselectivity observed compared to the other reductions. In order to favour syn formation, Cram- chelate should be favoured using smaller catalysts for the hydrosilylation or using as suggested by another study, tetra-n-butylammonium borohydride as the reductive agent.

6. Bibliography 1. Black, S. N., Williams, L. J. & Davey, R. J. The preparation of enantiomers of paclobutrazol: A crystal chemistry approach. Tetrahedron. 45, 2677–2682 (1989). 2. Sugavanam, B. Diastereoisomers and Enantiomers of Paclobutrazol: Their Preparation and Biological Activity. Pestic. Sci. 15, 296–302 (1984). 3. Berova, M. & Zlatev, Z. Physiological response and yield of paclobutrazol treated tomato plants (Lycopersicon esculentum Mill.). Plant Growth Regul. 30, 117–123 (2000). 4. Ishii, H. & Holloman, D. W. Sterol biosynthesisinhibitors: C14 demethylation (DMIs), Fungicide resistance. in Plant Pathogens 199–200 (Springer, 2015). 5. Wyenandt, A. Growers Guide to Understanding the DMI or SBI Fungicides (FRAC Code 3). Plant & pest advisory https://plant-pest-advisory.rutgers.edu/growers-guide-to- understanding-the-dmi-or-sbi-sterol-biosynthesis-inhibitor-fungicides-frac-code-3/ (2013). 6. BASF. Efficacité des triazoles dans les fongicides. https://www.agro.basf.fr/fr/cultures/ble/protection_fongicide_ble/efficacite_des_triazoles/. 7. Nowak, A., Christensen, J. R. & Mabry, T. R. Antimicrobials in pediatric dentistry. in Pediatric Dentistry-E-Book: Infancy through Adolescence (Elsevier Health Sciences, 2018). 8. Buchenauer, H. Proc. Br. Crop Prot. Conf. 2, 699 (1977). 9. Lever, B. G., Shearing, S. J. & Batch, J. J. Proc. Br. Crop Prot. Conf. 1, 3 (1982). 188 Chapter 4- Valorization of BnFTP

10. Yamaguchi, S. Gibberellin metabolism and its regulation. Annu. Rev. Plant Biol. 59, 225–251 (2008). 11. Gawley, R. E. & Aube, J. Introduction, General Principales, and Glossary of Stereochemical Terms. in Principles of asymmetric synthesis 1–62 (Elsevier, 2012). 12. Hughes, D. L., Reamer, R. A., Bergan, J. J. & Grabowski, E. J. J. A Mechanistic Study of the Mitsunobu Esterification Reaction. 6487–6491 (2000). 13. Stergiou, P. et al. Advances in Lipase-Catalyzed Esterification Reactions Advances. in lipase-catalyzed esterification reactions. Biotechnol. Adv. 31, 1846–1859 (2013). 14. Otera, J. & Nishikido, J. Reaction of Alcohols with Carboxylic Acids. in Esterification: Methods, Reactions, and Applications 5–156 (Wiley-VCH, 2010). 15. Bender, J., Jepkens, D. & Hu, H. Ionic Liquids as Phase-Transfer Catalysts : Etherification Reaction of 1-Octanol with 1-Chlorobutane. Org. Process Res. Dev. 14, 716–721 (2010). 16. Williamson, A. W. XXII.—On etherification. Q. J. Chem. Soc. London. 4, 229–239 (1852). 17. Aspinall, H. C., Greeves, N., Lee, W., Mciver, E. G. & Smith, P. M. An Improved Williamson Etherification of Hindered Alcohols Promoted by M-Crown-5 and Sodium Hydride. Tetrahedron Lett. 38, 4679–4682 (1997). 18. Tomie, M., Sugimoto, H. & Yoneda, N. Synthesis of optically active prenylamine from (-)-norephedrine. Chem. Pharm. Bull. 24, 1033–9 (1976). 19. Yin, L. & Liebscher, J. Carbon − Carbon Coupling Reactions Catalyzed by Heterogeneous Palladium Catalysts. Chem. Rev. 107, 133–173 (2007). 20. Vergote, T., Gharbi, S., Billard, F., Riant, O. & Leyssens, T. Ketone hydrosilylation by Cu (I) diphosphine complexes : A kinetic study. J. Organomet. Chem. 745–746, 133–139 (2013). 21. Clayden, J., Greeves, N., Warren, S. & Wothers, P. Stereochemistry. in Organic Chemistry (ed. Oxford University Press) 381–406 (Oxford University Press, 2008). 22. Light, S. H., Minasov, G. & Duban, M.-E. Adherence to Bürgi–Dunitz stereochemical principles requires significant structural rearrangements in Schiff-base formation: insights from transaldolase complexes. Acta Crystallogr. Sect. B. 70, 544–552 (2014). 23. Nakata, T., Tanaka, T. & Oishi, T. Stereoselective reduction of a-hydroxy ketones. Tetrahedron Lett. 24, 2653–2656 (1983). 24. Biomol, O., Karjalainen, O. K. & Koskinen, A. M. P. Organic & Biomolecular Chemistry Diastereoselective synthesis of vicinal amino alcohols. Org. Biomol. Chem. 10, 4311–4326 (2012). 25. Suzuki, Y., Kaneno, D., Miura, M. & Tomoda, S. Solvent effects on the diastereoselection in LiAlH 4 reduction of a -substituted ketones. Tetrahedron Lett. 49, 4223– 4226 (2008). 26. Thieme, P. C., Sauter, H. & ReiBenweber, G. Diastereoselektivitat bei der Boranatreduktion von a-Triazolylketonen. Chem. Ber. 121, 1059–1062 (1988). 189 Conclusion & Perspectives

Conclusion & Perspectives

190 Conclusion & Perspectives Conclusion In this thesis, we aimed at developing a new deracemization process, Co- crystallization Induced Spontaneous Deracemization (CoISD), using co- crystallization to add a new tool to the chiral toolbox for industries and researchers wanting to obtain enantiopure material with an excellent purity and yield. Co- crystallization was used in order to create an imbalance in solution from precipitation of one enantiomer with another chiral and enantiopure molecule, called co-former. The imbalance in solution is then racemized by addition of a racemizing agent. By running both processes simultaneously, deracemization of the target compound could be achieved (figure 1).

Figure 1 Explanatory scheme of CoISD.

First of all, we chose a model compound, BnClTP, for its capacity to easily racemize1 and its structure that possesses synthons known to facilitate co-crystallization. Nine analogs of BnClTP were successfully synthetized, characterized and purified for the next step: the co-crystal screen. In this step, all nine analogs and BnClTP were ground with 35 chiral co-formers, in a two-step screening. 7 co-crystals were confirmed by single crystal analysis. Out of those seven co-crystals, 2 were suitable for developing the process. The most ideal system was chosen for the next stage after a study of the properties of all systems. The rest of the screening highlighted the importance of carboxylic acids for co-crystallization to occur with the series of synthetized compounds, with generally a very favorable interaction between the N4 of the 1,2,4- triazole with the proton of the acid.

Second, the selected system of (R,S)-BnFTP and enantiopure PBA was studied toward chiral resolution and racemization. Resolution gave excellent results, due to a strong difference in solubility between both diastereomeric co-crystals. Racemization occurred without problems for BnFTP alone but was more challenging when PBA was present. For more knowledge, the kinetics of racemization of BnFTP, with the racemizing agent, DBU, were studied with and without PBA highlighting a first order mechanism with respect to the enantiomeric excess and the base when running the racemization without PBA. A more complex mechanism is obtained, when PBA is 191 Conclusion & Perspectives present. We propose involvement of a hydrogen bond stabilization of the ketone by protonated DBU while the carboxylate of PBA deprotonates the H in the α position of the ketone (figure 2).

Figure 2 Supposed mechanism under racemization of BnFTP with DBU in presence of PBA (case of (R)-PBA).

Heating was shown to strongly increase racemization rate (Arrhenius law) and was eventually introduced to have racemization occurring in presence of PBA. Because of that, the whole idea of one pot deracemization first envisioned was revised and transformed into a two-pot one-step continuous process (figure 3) with a racemization vessel at 90°C and a crystallization vessel at 10°C. Two gear pumps, continuously transferred liquid from one vessel to the other. With this set-up, the CoISD process was successfully developed on the system with first encouraging results leading to a 51% yield.

Figure 3 Explanatory scheme of the two-pot one-step continuous process

The process was then optimized in two steps. Then, the process was optimized incorporating these results for the study of the impact of different operational parameters on the overall process. After assessing the impact of water, protic solvents, 192 Conclusion & Perspectives compound concentration, base concentration and temperature of the crystallization cell, an optimized process was obtained with a 73% yield in enantiopure co-crystal for a total deracemization of 80%, with furthermore the possibility to efficiently recycle the filtrate.

Last but not least, BnFTP was valorized both as an analog of Paclobutrazol and a chiral building block upon reduction of the ketone into an alcohol. Reduction under catalyzed hydrosilylation of enantiopure BnFTP afforded two-enantiopure diastereomers, with the anti product being the major product.

To make a long story short, after 4 years of research, starting from synthesis up to valorization, a model system was discovered and CoISD successively developed and optimized to furnish an efficient deracemization tool, applicable to almost all chiral racemizable molecules yielding enantiopure material in a single process. The enantiopure material was shown to also be of interest as a fungicide or a chiral building block, closing the circle by valorizing the compound (figure 4).

Figure 4 The cycle of BnFTP’s life from its use in developing the CoISD process leading to its deracemization, to its valorization.

Going further, figure 5 is a chart diagram detailing the steps and recommendations when developing a CoISD process for a chiral API. The different steps detailed in the different chapters of this work are summed-up and generalized in order to guide any new adventurers in the CoSID process development. 193 Conclusion & Perspectives

Figure 5 Chart diagram of the CoISD process development.

Perspectives The CoISD process was successfully carried out. However, there remains plenty of work as was also the case when Crystallization Induced Diastereomer Transformation or more recently Viedma Ripening were introduced. Indeed, CIDT is now being used for all types of chirality,2 has several variants existing,3 which offer a large range of possibilities depending on each compound case. Similarly, when Viedma Ripening was first introduced,4 it was carried out on an achiral molecule whose chirality was only expressed in the solid state, rendering the “racemization” part inherent to the re- dissolution step. Nevertheless, over the years, Viedma Ripening was shown to efficiently deracemize chiral organic molecules.5,6 Furthermore, the original abrasive grinding was improved with sonication1 and replaced with the alternative of temperature cycling.7 All those further developments of the initial process give them their breadth in term of applicability and versatility. 194 Conclusion & Perspectives

The same goes for CoISD. If its applicability regarding a large range of molecules is inherent to its co-crystallization part, the process was only applied to one compound. Hence, there should be further development of CoISD processes on other chiral molecules in order to experimentally confirm what theory dictates. One easy way would be to use the co-crystal between BnTP and PBA, which was also suitable. Another approach would involve to pursue other co-crystals of the same system identified during the co-crystal screening in order to get more single crystal structures out of all hits that were obtained. A third approach would be to look for a new type of racemizable molecule and find a suitable co-crystal. Finally, another approach could be reversed by taking PBA and looking for a good racemizable molecule to make it co-crystallize with. Regardless of the method, the idea remains to develop CoISD processes on more molecules in order to provide more data on the matter.

Similarly to CIDT, CoISD should be applicable on every type of chirality, providing that the molecule can be racemized. Then, a natural perspective of this work would be to use a different type of chiral molecule (e.g. BINOL) and find a suitable co- crystal with, to develop a CoISD process on. Using other types of chirality can be quite interesting as the means of racemization are different. For instance, for BINOL, it is known to be racemizable under UV light.8 In this case, there is no need of heating and a one-pot process could be envisioned. For other planar or axial chiral molecules, racemization can be achieved by only heating. Those would easily fit the developed process with one less variable to the mix, the racemizing agent. Other racemization techniques can also be considered. For instance in the case of chiral sulfoxides, they can be racemized by the presence of an acid9 or simply by heating.10 In the first case, the co-former could then be used as the racemizing agent or if necessity of a stronger one, if the co-former remains an acid, it would not interfere with the racemization agent and a one-pot process would be possible. In the second case, the developed process would fit perfectly.

Figure 5 Racemization of (S)-BINOL as an example of axial chiral compound and of (S)-Methyl p-tolyl sulfoxide as an example of a chiral sulfoxide.

As for the reduction of BnFTP, a further study could involve more trials in order to favor the syn diastereomer. For example, this can be done by using n-Bu4NBH4 as a reducing agent since a study on similar compounds showed this reductive agent to favor the syn diastereomer while all other reducing agents favored the anti, like in the case of BnFTP. Furthermore, a patent used the same reductive agent in combination with TiCl4 to also favor the syn product in the reduction of structurally related 195 Conclusion & Perspectives compound. Interestingly those conditions favored the anti product in the previously mentioned study. Finally, on reduced BnFTP, several modifications can be done. For instance, in literature, on a similar structure, the alcohol was esterified,11-13 etherified14 or transformed into a halogen15 (figure 6).

Figure 6 Examples of from top to bottom etherification,14 esterification13 and bromination15 reaction on BnFTP closely related compounds.

Bibliography 1. Rougeot, C., Guillen, F., Plaquevent, J. C. & Coquerel, G. Ultrasound-Enhanced Deracemization: Toward the Existence of Agonist Effects in the Interpretation of Spontaneous Symmetry Breaking. Cryst. Growth Des. 15, 2151–2155 (2015). 2. Brands, K. M. J. & Davies, A. J. Crystallization-Induced Diastereomer Transformations. Chem. Rev. 106, 2711−2733 (2006). 3. Kozma, D. Alternative methods of resolution by diastereomeric salt formation. in CRC Handbook of Optical Resolutions via Diastereomeric Salt Formation 151–168 (CRC Press, 2002). 4. Viedma, C. Experimental evidence of chiral symmetry breaking in crystallization from primary nucleation. J. Cryst. Growth. 261, 118–121 (2004). 5. Viedma, C., Ortiz, J. E., de Torres, T., Izumi, T. & Blackmond, D. G. Evolution of Solid Phase Homochirality for a Proteinogenic Amino Acid. J. Am. Chem. Soc. 130, 15274– 15275 (2008). 6. Van Der Meijden, M. W. et al. Attrition-Enhanced Deracemization in the Synthesis of Clopidogrel - A Practical Application of a New Discovery. Org. Process Res. Dev. 13, 1195– 1198 (2009). 196 Conclusion & Perspectives

7. Suwannasang, K., Flood, A. E. & Coquerel, G. A Novel Design Approach To Scale Up the Temperature Cycle Enhanced Deracemization Process : Coupled Mixed-Suspension Vessels. Cryst. Growth Des. 16, 6461–6467 (2016). 8. Solntsev, K. M. et al. Excited-State Proton Transfer in Chiral Environments: Photoracemization of BINOLs. Isr J Chem. 49, 227–233 (2009). 9. Mislow, K., Simmons, T., Melillo, J. T. & Ternay, A. L. The Hydrogen Chloride- Catalyzed Racemization of Sulfoxides. J. Am. Chem. Soc. 86, 1452–1453 (1964). 10. Aurisicchio, C., Baciocchi, E., Gerini, M. F. & Lanzalunga, O. Thermal and Photochemical Racemization of Chiral Aromatic Sulfoxides via the Intermediacy of Sulfoxide Radical Cations. Org. Lett. 9, 1939–1942 (2007). 11. Lv, C. & Zhou, Z. Chiral HPLC separation and absolute configuration assignment of a series of new triazole compounds. J. Sep. Sci. 34, 363–370 (2011). 12. Ríos‐Lombardía, N. et al. Enantiopure Triazolium Salts: Chemoenzymatic Synthesis and Applications in . ChemCatChem. 3, 1921–1928 (2011). 13. Borowiecki, P., Poterała, M., Maurin, J., Wielechowska, M. & Plenkiewicz, J. Preparation and thermal stability of optically active 1,2,4-triazolium-based ionic liquids. ARKIVOC. 8, 262–281 (2012). 14. Zeeh, B., Ammermann, E., Sauter, H. & Pommer, E. H. Fungicidal β-triazolyl ethers. DE2926096A1 (1981). 15. Todoroki, Y. et al. Structure–activity relationship of uniconazole, a potent inhibitor of ABA 8′-hydroxylase, with a focus on hydrophilic functional groups and conformation. Bioorg. Med. Chem. 16, 3141–3152 (2008).

Appendices 198 Appendix A Appendix A: SI for Chapter 1

Material & methods for part 2. TP: DSC

Figure A-1 DSC of TP. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset.

BnClTP: DSC

Figure A-2 DSC of BnClTP. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset. 199 Appendix A

BnFTP: DSC

Figure A-3 DSC of BnFTP obtained from crystallization from the melt. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset.

Figure A-4 Second DSC of BnFTP. This sample was obtained from crystallization in toluene as its solvate, followed by desolvation at RT. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset. 200 Appendix A

Discussion on BnFTP solid state: As can be seen from figure A-3 and A-4, BnFTP appears to possess at least one polymorph. In figure A-3, there is possibility of fusion of two polymorphs with close melting points or one polymorph with presence of impurities (possibly traces remaining after purification by column chromatography). Regarding the second figure, it shows the or one of the polymorph(s) present in figure A-3 with another bigger peak at a higher melting point (52.9°C) which probably correspond to the thermodynamically stable polymorph (providing there is not another higher melting polymorph).

BnTP: DSC

Figure A-5 DSC of BnTP. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset. 201 Appendix A

PgTP: DSC

Figure A-6 DSC of PgTP. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset.

BnBnOP: DSC

Figure A-7 DSC of BnBnOP. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset. 202 Appendix A

TAP: DSC

Figure A-8 DSC of TAP. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset.

BnTAP: DSC

Figure A-9 DSC of BnTAP. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset. 203 Appendix A

TAmBn2: DSC

Figure A-10 DSC of TAmBn2. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset.

BnClTAmBn2: DSC

Figure A-11 DSC of BnClTAmBn2. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset. 204 Appendix A

BnFTAmBn2: DSC

Figure A-12 DSC of BnFTAmBn2. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset.

BnTAmBn2: DSC

Figure A-13 DSC of BnTAmBn2. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset. 205 Appendix A

BnBAA: DSC + Structural information

Figure A-14 DSC of BnBAA (racemic). Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset.

Table A-1 Crystal data and structure refinement for BnBAA.

Identification code mg_brBAA Empirical formula C9 H9 Br O2 Formula weight 229.07 Temperature 150(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21/c Unit cell dimensions a = 7.5352(6) Å. b = 6.0615(3) Å; = 95.828(7)°. c = 20.4516(15) Å. Volume 929.29(11) Å3 Z 4 Density (calculated) 1.637 Mg/m3 Absorption coefficient 4.379 mm-1 F(000) 456 Crystal size 0.35 x 0.13 x 0.12 mm3 Theta range for data collection 3.208 to 25.241°. Index ranges -9<=h<=7, -7<=k<=7, -24<=l<=24 Reflections collected 4578 Independent reflections 1664 [R(int) = 0.0391] Completeness to theta = 25.241° 99.0 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.63740 Refinement method Full-matrix least-squares on F2 206 Appendix A

Data / restraints / parameters 1664 / 0 / 110 Goodness-of-fit on F2 1.131 Final R indices [I>2sigma(I)] R1 = 0.0549, wR2 = 0.1445 R indices (all data) R1 = 0.0680, wR2 = 0.1516 Largest diff. peak and hole 0.925 and -0.331 e.Å-3 BnTAA: DSC + Structural information

Figure A-15 DSC of BnTAA (racemic). Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset.

Table A-2 Crystal data and structure refinement for BnTAA.

Identification code mg_brTAA Empirical formula C11 H10 N3 O2 Formula weight 216.22 Temperature 296(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21/n Unit cell dimensions a = 5.5151(5) Å. b = 16.1549(16) Å; = 92.925(8)°. c = 12.2970(11) Å. Volume 1094.18(18) Å3 Z 4 Density (calculated) 1.313 Mg/m3 Absorption coefficient 0.094 mm-1 F(000) 452 Crystal size 0.40 x 0.35 x 0.20 mm3 Theta range for data collection 3.550 to 26.016°. Index ranges -6<=h<=6, -18<=k<=19, -15<=l<=15 Reflections collected 6993 Independent reflections 2126 [R(int) = 0.0512] Completeness to theta = 25.242° 98.8 % 207 Appendix A

Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.80567 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 2126 / 0 / 145 Goodness-of-fit on F2 1.076 Final R indices [I>2sigma(I)] R1 = 0.0649, wR2 = 0.1744 R indices (all data) R1 = 0.0783, wR2 = 0.1850 Largest diff. peak and hole 0.330 and -0.169 e.Å-3 MBnTA: DSC

Figure A-16 DSC of MBnTA. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset.

Results & discussion for part 2. By-products BnFTP

Table A-3 Crystal data and structure refinement for (BnF)2TP.

Identification code mg_52p2 Empirical formula C22 H23 F2 N3 O Formula weight 383.43 Temperature 297(2) K Wavelength 0.71073 Å Crystal system Orthorhombic Space group P212121 Unit cell dimensions a = 10.0297(10) Å. b = 12.635(2) Å. c = 15.4977(17) Å. Volume 1963.9(4) Å3 Z 4 Density (calculated) 1.297 Mg/m3 208 Appendix A

Absorption coefficient 0.094 mm-1 F(000) 808 Crystal size 0.30 x 0.25 x 0.11 mm3 Theta range for data collection 3.225 to 25.329°. Index ranges -12<=h<=12, -15<=k<=14, -18<=l<=16 Reflections collected 12613 Independent reflections 3502 [R(int) = 0.0725] Completeness to theta = 25.242° 98.2 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.82602 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 3502 / 0 / 256 Goodness-of-fit on F2 1.019 Final R indices [I>2sigma(I)] R1 = 0.0522, wR2 = 0.1211 R indices (all data) R1 = 0.0763, wR2 = 0.1370 Absolute structure parameter 0.5 Largest diff. peak and hole 0.124 and -0.122 e.Å-3

Table A-4 Crystal data and structure refinement for Z-BnFTPBnF.

Identification code mg_s2_p1 Empirical formula C22 H23 F2 N3 O Formula weight 383.43 Temperature 150(2) K Wavelength 0.71073 Å Crystal system Orthorhombic Space group P212121 Unit cell dimensions a = 5.83427(15) Å. b = 8.28435(17) Å. c = 39.6795(8) Å. Volume 1917.83(7) Å3 Z 4 Density (calculated) 1.328 Mg/m3 Absorption coefficient 0.096 mm-1 F(000) 808 Theta range for data collection 2.901 to 26.252°. Index ranges -7<=h<=6, -10<=k<=10, -45<=l<=48 Reflections collected 9683 Independent reflections 3750 [R(int) = 0.0253] Completeness to theta = 25.242° 97.9 % Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 3750 / 0 / 256 Goodness-of-fit on F2 1.097 Final R indices [I>2sigma(I)] R1 = 0.0360, wR2 = 0.0913 R indices (all data) R1 = 0.0370, wR2 = 0.0919 Absolute structure parameter 0.5 Largest diff. peak and hole 0.254 and -0.178 e.Å-3

209 Appendix A Results and discussion for part 3. Structural Information of analogs

Table A-5 Crystal data and structure refinement for BnFTAmBn2.

Identification code nt0127_TL_Michael_AmFOO Formula weight 414.47 Temperature 100(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21/c Unit cell dimensions a = 12.2591(3) Å. b = 10.1395(3) Å; = 99.610(3)°. c = 16.9773(5) Å. Volume 2080.68(11) Å3 Z 4 Density (calculated) 1.323 Mg/m3 Absorption coefficient 0.089 mm-1 F(000) 872 Crystal size 0.429 x 0.330 x 0.305 mm3 Theta range for data collection 3.007 to 32.864°. Index ranges -16<=h<=17, -9<=k<=15, -25<=l<=20 Reflections collected 16285 Independent reflections 7009 [R(int) = 0.0306] Completeness to theta = 25.242° 99.8 % Absorption correction Analytical Max. and min. transmission 0.980 and 0.972 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 7009 / 0 / 349 Goodness-of-fit on F2 1.026 Final R indices [I>2sigma(I)] R1 = 0.0490, wR2 = 0.1052 R indices (all data) R1 = 0.0758, wR2 = 0.1219 Largest diff. peak and hole 0.339 and -0.273 e.Å-3

Table A-6 Crystal data and structure refinement for BnFTP (conglomerate).

Identification code mg_Bnftp_tst-rac Empirical formula C H F N O Formula weight 62.03 Temperature 297(2) K Wavelength 0.71073 Å Crystal system Orthorhombic Space group P212121 Unit cell dimensions a = 5.8272(9) Å. b = 13.571(3) Å. c = 19.548(4) Å. Volume 1545.9(5) Å3 Z 4 Density (calculated) 0.267 Mg/m3 Absorption coefficient 0.030 mm-1 F(000) 124 Theta range for data collection 3.468 to 20.568°. Index ranges -5<=h<=5, -13<=k<=9, -19<=l<=15 Reflections collected 2007 Independent reflections 1428 [R(int) = 0.0251] 210 Appendix A

Completeness to theta = 20.568° 95.0 % Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 1428 / 154 / 216 Goodness-of-fit on F2 1.047 Final R indices [I>2sigma(I)] R1 = 0.0492, wR2 = 0.1133 R indices (all data) R1 = 0.0635, wR2 = 0.1232 Absolute structure parameter 0.5 Largest diff. peak and hole 0.101 and -0.143 e.Å-3

Table A-7 Crystal data and structure refinement for BnFTP Toluene Solvate.

Identification code mg_cpf_3 Empirical formula C37 H44 F2 N6 O2 Formula weight 642.78 Temperature 297(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21/c Unit cell dimensions a = 5.7209(6) Å. b = 14.1838(16) Å; = 93.813(10)°. c = 22.269(2) Å. Volume 1803.0(3) Å3 Z 2 Density (calculated) 1.184 Mg/m3 Absorption coefficient 0.082 mm-1 F(000) 684 Crystal size 0.50 x 0.20 x 0.20 mm3 Theta range for data collection 3.103 to 25.241°. Index ranges -6<=h<=6, -17<=k<=17, -26<=l<=26 Reflections collected 14144 Independent reflections 3199 [R(int) = 0.0551] Completeness to theta = 25.241° 98.4 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.93328 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 3199 / 106 / 287 Goodness-of-fit on F2 1.083 Final R indices [I>2sigma(I)] R1 = 0.0917, wR2 = 0.1912 R indices (all data) R1 = 0.1080, wR2 = 0.1988 Largest diff. peak and hole 0.401 and -0.192 e.Å-3

Table A-8 Crystal data and structure refinement for BnTP.

Identification code mg_cp19 Empirical formula C15 H19 N3 O Formula weight 257.33 Temperature 150(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21 Unit cell dimensions a = 5.7006(2) Å. b = 19.1073(8) Å; = 97.424(4)°. 211 Appendix A

c = 13.2911(6) Å. Volume 1435.57(10) Å3 Z 4 Density (calculated) 1.191 Mg/m3 Absorption coefficient 0.077 mm-1 F(000) 552 Crystal size 0.30 x 0.10 x 0.10 mm3 Theta range for data collection 3.091 to 25.241°. Index ranges -6<=h<=6, -22<=k<=22, -15<=l<=15 Reflections collected 8670 Independent reflections 5063 [R(int) = 0.0717] Completeness to theta = 25.241° 99.7 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.75928 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 5063 / 1 / 350 Goodness-of-fit on F2 1.042 Final R indices [I>2sigma(I)] R1 = 0.0679, wR2 = 0.1501 R indices (all data) R1 = 0.1036, wR2 = 0.1736 Absolute structure parameter 0(3) Largest diff. peak and hole 0.275 and -0.252 e.Å-3

Table A-9 Crystal data and structure refinement for BnTAmBn2.

Identification code mg_BrTAmBnz_cocr Empirical formula C25 H24 N4 O Formula weight 396.48 Temperature 297(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21/n Unit cell dimensions a = 19.6347(12) Å. b = 10.3222(5) Å; = 110.333(7)°. c = 22.4601(13) Å. Volume 4268.4(4) Å3 Z 8 Density (calculated) 1.234 Mg/m3 Absorption coefficient 0.077 mm-1 F(000) 1680 Crystal size 0.35 x 0.25 x 0.15 mm3 Theta range for data collection 2.965 to 25.681°. Index ranges -23<=h<=23, -12<=k<=12, -27<=l<=27 Reflections collected 38188 Independent reflections 8011 [R(int) = 0.0436] Completeness to theta = 25.242° 99.0 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.69625 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 8011 / 0 / 541 Goodness-of-fit on F2 1.066 Final R indices [I>2sigma(I)] R1 = 0.0496, wR2 = 0.1219 R indices (all data) R1 = 0.0612, wR2 = 0.1282 Largest diff. peak and hole 0.169 and -0.149 e.Å-3

212 Appendix A

Table A-10 Crystal data and structure refinement for BnTAmBn2 (Conglomerate).

Identification code mg_BrTAmBnz Empirical formula C25 H24 N4 O Formula weight 396.48 Temperature 297(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group I2 Unit cell dimensions a = 18.3275(18) Å. b = 5.4779(4) Å; = 106.011(10)°. c = 21.8306(19) Å. Volume 2106.7(3) Å3 Z 4 Density (calculated) 1.250 Mg/m3 Absorption coefficient 0.078 mm-1 F(000) 840 Crystal size 0.50 x 0.10 x 0.07 mm3 Theta range for data collection 3.335 to 25.366°. Index ranges -21<=h<=22, -6<=k<=6, -26<=l<=26 Reflections collected 7414 Independent reflections 3703 [R(int) = 0.0504] Completeness to theta = 25.242° 99.1 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.97290 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 3703 / 1 / 271 Goodness-of-fit on F2 1.062 Final R indices [I>2sigma(I)] R1 = 0.0521, wR2 = 0.1060 R indices (all data) R1 = 0.0674, wR2 = 0.1132 Absolute structure parameter -0.5(10) Largest diff. peak and hole 0.139 and -0.130 e.Å-3

Table A-11 Crystal data and structure refinement for BnClTAmBn2.

Identification code mg_amCl Empirical formula C25 H23 Cl N4 O Formula weight 430.92 Temperature 297(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21/c Unit cell dimensions a = 12.9347(6) Å. b = 10.3383(5) Å; = 100.457(5)°. c = 16.8752(8) Å. Volume 2219.10(19) Å3 Z 4 Density (calculated) 1.290 Mg/m3 Absorption coefficient 0.197 mm-1 F(000) 904 Crystal size 0.300 x 0.250 x 0.150 mm3 Theta range for data collection 3.148 to 25.518°. Index ranges -15<=h<=15, -12<=k<=12, -20<=l<=20 Reflections collected 19770 213 Appendix A

Independent reflections 4115 [R(int) = 0.0377] Completeness to theta = 25.242° 99.3 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.96590 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 4115 / 0 / 280 Goodness-of-fit on F2 1.046 Final R indices [I>2sigma(I)] R1 = 0.0450, wR2 = 0.1135 R indices (all data) R1 = 0.0556, wR2 = 0.1205 Largest diff. peak and hole 0.175 and -0.240 e.Å-3

Table A-12 Crystal data and structure refinement for MBnTA.

Identification code mg_MBn_TA Empirical formula C12 H13 N3 O2 Formula weight 231.25 Temperature 297(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21/n Unit cell dimensions a = 7.7725(3) Å. b = 9.8380(4) Å; = 91.391(4)°. c = 15.5448(7) Å. Volume 1188.29(9) Å3 Z 4 Density (calculated) 1.293 Mg/m3 Absorption coefficient 0.091 mm-1 F(000) 488 Crystal size 0.50 x 0.40 x 0.30 mm3 Theta range for data collection 3.341 to 26.214°. Index ranges -9<=h<=9, -12<=k<=12, -19<=l<=19 Reflections collected 13339 Independent reflections 2348 [R(int) = 0.0312] Completeness to theta = 25.242° 98.1 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.90856 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 2348 / 0 / 155 Goodness-of-fit on F2 1.070 Final R indices [I>2sigma(I)] R1 = 0.0449, wR2 = 0.1155 R indices (all data) R1 = 0.0480, wR2 = 0.1174 Largest diff. peak and hole 0.205 and -0.145 e.Å-3 Structural Information of co-crystals Table A-13 Crystal data and structure refinement for (R,S)-BnTP+D-Tartaric Acid.

Identification code mg_cp08 Empirical formula C34 H44 N6 O8 Formula weight 664.75 Temperature 293(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21 Unit cell dimensions a = 5.59158(14) Å. 214 Appendix A

b = 17.2839(5) Å; = 93.470(3)°. c = 18.9337(5) Å. Volume 1826.48(8) Å3 Z 2 Density (calculated) 1.209 Mg/m3 Absorption coefficient 0.087 mm-1 F(000) 708 Crystal size 0.45 x 0.03 x 0.02 mm3 Theta range for data collection 3.194 to 25.240°. Index ranges -6<=h<=6, -20<=k<=20, -22<=l<=22 Reflections collected 13158 Independent reflections 6350 [R(int) = 0.0513] Completeness to theta = 25.240° 99.1 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.81188 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 6350 / 1 / 443 Goodness-of-fit on F2 1.010 Final R indices [I>2sigma(I)] R1 = 0.0591, wR2 = 0.1553 R indices (all data) R1 = 0.0757, wR2 = 0.1679 Absolute structure parameter 1.3(10) Largest diff. peak and hole 0.437 and -0.189 e.Å-3

Table A-14 Crystal data and structure refinement for (R,S)-BnFTP+D-Tartaric Acid.

Identification code mg_cpf08 Empirical formula C34 H42 F2 N6 O8 Formula weight 700.73 Temperature 150(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21 Unit cell dimensions a = 5.5008(2) Å. b = 17.1791(7) Å; = 90.960(4)°. c = 18.8899(9) Å. Volume 1784.83(13) Å3 Z 2 Density (calculated) 1.304 Mg/m3 Absorption coefficient 0.101 mm-1 F(000) 740 Crystal size 0.62 x 0.20 x 0.03 mm3 Theta range for data collection 3.206 to 25.242°. Index ranges -6<=h<=6, -20<=k<=20, -20<=l<=22 Reflections collected 15091 Independent reflections 6320 [R(int) = 0.0708] Completeness to theta = 25.242° 98.9 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.82550 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 6320 / 1 / 461 Goodness-of-fit on F2 1.080 Final R indices [I>2sigma(I)] R1 = 0.0670, wR2 = 0.1745 R indices (all data) R1 = 0.0830, wR2 = 0.1865 Absolute structure parameter -1.7(10) Largest diff. peak and hole 0.265 and -0.344 e.Å-3 215 Appendix A

Table A-15 Crystal data and structure refinement for (R)-BnTP+ (R)-3-Phenylbutyric Acid.

Identification code mg_cp12 Empirical formula C25 H31 N3 O3 Formula weight 421.53 Temperature 150(2) K Wavelength 0.71073 Å Crystal system Orthorhombic Space group P212121 Unit cell dimensions a = 5.5560(3) Å. b = 15.8908(8) Å. c = 26.2729(13) Å. Volume 2319.62(19) Å3 Z 4 Density (calculated) 1.207 Mg/m3 Absorption coefficient 0.080 mm-1 F(000) 904 Crystal size 0.60 x 0.06 x 0.02 mm3 Theta range for data collection 3.101 to 25.684°. Index ranges -6<=h<=6, -19<=k<=19, -32<=l<=32 Reflections collected 20443 Independent reflections 4396 [R(int) = 0.0561] Completeness to theta = 25.242° 99.5 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.95412 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 4396 / 0 / 285 Goodness-of-fit on F2 1.060 Final R indices [I>2sigma(I)] R1 = 0.0390, wR2 = 0.0774 R indices (all data) R1 = 0.0465, wR2 = 0.0801 Absolute structure parameter -0.8(7) Largest diff. peak and hole 0.153 and -0.133 e.Å-3 Table A-16 Crystal data and structure refinement for (R)-BnFTP + (R)-PBA co-crystal.

Identification code mg_cpf12 Empirical formula C25 H30 F N3 O3 Formula weight 439.52 Temperature 150(2) K Wavelength 0.71073 Å Crystal system Orthorhombic Space group P212121 Unit cell dimensions a = 5.5481(4) Å. b = 16.4771(12) Å. c = 25.6791(19) Å. Volume 2347.5(3) Å3 Z 4 Density (calculated) 1.244 Mg/m3 Absorption coefficient 0.088 mm-1 F(000) 936 Crystal size 0.42 x 0.03 x 0.02 mm3 Theta range for data collection 2.938 to 23.248°. Index ranges -6<=h<=6, -18<=k<=18, -28<=l<=28 Reflections collected 13690 Independent reflections 3340 [R(int) = 0.1121] 216 Appendix A

Completeness to theta = 23.248° 99.5 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.58890 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 3340 / 0 / 294 Goodness-of-fit on F2 1.098 Final R indices [I>2sigma(I)] R1 = 0.0837, wR2 = 0.1152 R indices (all data) R1 = 0.1115, wR2 = 0.1236 Absolute structure parameter 2.1(10) Largest diff. peak and hole 0.198 and -0.188 e.Å-3

Table A-17 Crystal data and structure refinement for (R,S)-BnTAmBn2 + (R,S)-Phenyl lactic acid co-crystal.

Identification code mg_Am19 Empirical formula C11 H10 N O Formula weight 172.20 Temperature 293(2) K Wavelength 0.798 Å Crystal system Triclinic Space group P-1 Unit cell dimensions a = 5.4281(5) Å; = 81.341(8)°. b = 13.2938(15) Å; = 89.876(7)°. c = 39.986(3) Å;  = 78.281(9)°. Volume 2791.8(5) Å3 Z 12 Density (calculated) 1.229 Mg/m3 Absorption coefficient 0.079 mm-1 F(000) 1092 Theta range for data collection 1.778 to 28.623°. Index ranges -6<=h<=6, -15<=k<=15, -45<=l<=44 Reflections collected 14070 Independent reflections 14070 [R(int) = ?] Completeness to theta = 28.607° 89.6 % Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 14070 / 0 / 758 Goodness-of-fit on F2 0.918 Final R indices [I>2sigma(I)] R1 = 0.1074, wR2 = 0.2416 R indices (all data) R1 = 0.2446, wR2 = 0.2776 Largest diff. peak and hole 0.539 and -0.414 e.Å-3

Table A-18 Crystal data and structure refinement for (RS)-BnClTP + (RS)-methyl succinic acid co-crystal.

Identification code mg_cpc21 Empirical formula C35 H44 Cl2 N6 O6 Formula weight 715.66 Temperature 150(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21 Unit cell dimensions a = 5.5830(5) Å. b = 17.7529(16) Å; = 90.882(8)°. 217 Appendix A

c = 19.0976(17) Å. Volume 1892.6(3) Å3 Z 2 Density (calculated) 1.256 Mg/m3 Absorption coefficient 0.222 mm-1 F(000) 756 Crystal size 0.50 x 0.10 x 0.10 mm3 Theta range for data collection 3.134 to 25.271°. Index ranges -6<=h<=6, -21<=k<=21, -22<=l<=22 Reflections collected 12463 Independent reflections 6791 [R(int) = 0.0453] Completeness to theta = 25.242° 99.4 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.84144 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 6791 / 282 / 505 Goodness-of-fit on F2 1.048 Final R indices [I>2sigma(I)] R1 = 0.0563, wR2 = 0.1288 R indices (all data) R1 = 0.0660, wR2 = 0.1349 Absolute structure parameter 0.03(5) Largest diff. peak and hole 0.573 and -0.308 e.Å-3

Table A-19 Crystal data and structure refinement for (RS)-BnClTP + (RS)-4-chloromandelic acid co-crystal.

Identification code mg_cpc24 Empirical formula C27.50 H25 Cl N2.50 O2.50 Formula weight 465.95 Temperature 150(2) K Wavelength 0.71073 Å Crystal system Triclinic Space group P-1 Unit cell dimensions a = 5.6346(7) Å; = 83.182(13)°. b = 16.710(3) Å; = 85.545(11)°. c = 25.491(4) Å;  = 81.016(12)°. Volume 2349.6(6) Å3 Z 4 Density (calculated) 1.317 Mg/m3 Absorption coefficient 0.194 mm-1 F(000) 978 Theta range for data collection 2.810 to 21.953°. Index ranges -5<=h<=5, -17<=k<=17, -26<=l<=26 Reflections collected 10034 Independent reflections 10034 [R(int) = ?] Completeness to theta = 21.953° 99.7 % Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 10034 / 0 / 258 Goodness-of-fit on F2 1.440 Final R indices [I>2sigma(I)] R1 = 0.1368, wR2 = 0.2747 R indices (all data) R1 = 0.2360, wR2 = 0.3164 Largest diff. peak and hole 1.098 and -1.102 e.Å-3

218 Appendix A

Angle and distance for hydrogen bonds Table A-20 Angles (X---H-X) and distances (X-----X) for all co-crystals obtained by SC-XRD.

H-bond length Co-crystal H bond Angle (°) (Å) (R,S)-BnTP + D-Tartaric Acid 2.614 171.7 (R,S)-BnFTP + D-Tartaric Acid 2.588 177.1 (R)-BnTP+ (R)-3-Phenylbutyric Acid 2.681 175.4 (R)-BnFTP + (R)-PBA co-crystal 2.702 170.7

(R,S)-BnTAmBn2 + (R,S)-Phenyl lactic acid 2.542 160.7 RS)-BnClTP + (RS)-methyl succinic acid 2.650 176.8 (RS)-BnClTP + (RS)-4-chloromandelic acid 2.643 168.7 Suitable Co-crystals for the development of the CoISD process

Figure A-17 DSC of (S)-BnTP + (S)-3-Phenylbutyric acid. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset.

Figure A-18 DSC of (S)-BnFTP + (S)-3-Phenylbutyric acid. Exothermic peaks are represented positively while endothermic ones are negative. The melting point corresponds to the onset. 219 Appendix A

Phase diagram Calibration curves

4000000

3000000 A = 7800.2C R² = 0,9981 A [2000000-] BnFTP PBA 1000000 A = 5709.9C R² = 0,9964 0 0 200 400 C [mg/L]

Figure C-3 Calibration curves for reverse HPLC for BnFTP and PBA.

Results

Figure C-4 Reverse HPLC results for the different molar ratio of PBA of the original suspension: from 0.1 to 0.9. The peak at roughly 1.56min is the peak of PBA and at roughly 2.47, it is the peak of BnFTP. 220 Appendix A

Table C-3 Each point of the ternary diagram composed of the molar of BnFTP, PBA and toluene. In light blue are the points obtained by weighing and in yellow the one from HPLC.

x(S)-BnFTP x(S)-PBA xToluene 0.0267 0 0.9733 0.0356 0.0001 0.9643 0.0724 0.0037 0.9239 0.0522 0.0031 0.9447 0.0396 0.0030 0.9573 0.0158 0.0028 0.9814 0.0127 0.0034 0.9839 0.0094 0.0040 0.9866 0.0060 0.0052 0.9888 0.0034 0.0056 0.9910 0.0021 0.0102 0.9877 0.0013 0.0147 0.9841 0.0032 0.0591 0.9377 0.0029 0.0705 0.9266

221 Appendix B

Appendix B: Article on BnTAA synthesis

Main Article

Development and optimization of a green 2- step stereospecific triazol synthesis

Michael Guillot*‡, Olivier Riant*, Tom Leyssens*.

*Institute of Condensed Matter and Nanosciences, Université Catholique de Louvain, Louvain-La-Neuve, Belgium.

ABSTRACT: The development of green chemical processes has gained increasing importance since the introduction of the concept 21 years ago. Green alternatives are even more important when speaking of enantiopure synthesis. However, developing an efficient green process comes with obstacles, especially when one wants to carry out SN2 reactions in water. In this paper, we developed and optimized a 2-step stereospecific synthesis starting from L-Phenylalanine and yielding (R)-2-(1H-1,2,4-triazol-1-yl)-3-phenylpropanoic acid ((R)-BnTAA), involving a total of 3 SN2 reactions. The reaction was designed to minimize waste production and match other green chemistry criteria. The reaction was studied by H1NMR in order to maximize efficiency. In the end, we present a green process with a minimized waste production and a total yield of 56%.

INTRODUCTION

During the last 20 years, ecological awareness has become widespread. Chemists can take responsibility toward the planet by assuring developed processes are ‘green’[1,2]. Green chemistry revolves around twelve principles, which if all applied would lead to safe and sustainable processes[1,3]. Among those twelve principles, waste prevention is a main pillar when considering chemical production. Indeed, efficient chemical processes often produce a large amount of waste[4]. The treatment of waste does not only imply important costs[3], but also carries an intrinsic toxicity due to the use of associated chemicals. That is why, the use of less hazardous chemicals, safer products and solvents are core principles of green chemistry[5]. Also, using renewable feedstock and developing energy-efficient processes are important concepts 222 Appendix B for the development of sustainable processes. For instance, an efficient process carried out at high temperature or at sub-zero temperatures requires a lot of energy generally produced from nonrenewable resources[6]. In the world of 1,2,4-triazoles, their derivatives are often studied as effective fungicides, due to the faculty of triazols to inhibit sterol 14R-demethylase, hereby disrupting the cell-wall formation of fungi[7]. Consequently, they have been subject of numerous patents[8- 13] and research contributions[14-15]. Their application is however not limited to this particular effect, as possible applications are also studied in the medical field for various uses from chronic renal desease[16], passing by bladder hyperactivity[17], metabolic syndrome[18], anxiety disorder[19] and immunodeficiency desease[20] to anti-cancer drugs[21] to only cite a few. Formation of a 1,2,4 triazol compound is either carried out by substitution of a good leaving group by a 1H-1,2,4-triazole or by forming the triazole on the compound, through condensation reactions[22]. But, despite a strong presence in the literature, those reactions lack a green development. Indeed, in the case of the substitution reactions, they are almost all carried out in organic solvents like acetonitrile[11,13,14,23-25], dimethylformamide[8,9,16,17,18], [19,20,26] [12,27] [10] tetrahydrofuran , acetone or even ethanol . There are some occurrence of SN2 reaction in water but only on primary carbons[28-33], and most of the time, the yields are poor[28,29] and/or the reaction requires the use of additives[29], strong heating[28,29,31,32] or long [28,29,32] reaction times . When it comes to secondary carbons, SN2 reactions are in general slow because of steric hindrance. Furthermore, water-based SN2 reactions on secondary carbons are even less favored because water is a protic solvent and can compete with the nucleophile for substitution. That is why water has not yet been explored in this context, despite its general compatibility with the reactants. Consequently, there is a need for development of green approaches on SN2 reactions focusing on secondary carbons, even more so as they are stereospecific with alternative asymmetric synthesis rarely being green due to the use of toxic catalysts. Therefore, in this contribution we present an optimized 2-step stereospecific triazol reaction in water, yielding the desired product with high selectivity, acceptable overall yield and excellent enantiopurity while managing to decrease waste production and develop the reaction in the greenest way possible. By showing the possibility of an efficient green SN2 reaction on a secondary carbon, we hope to lead the way into the development of many more alike reactions.

THE REACTION & MECHANISM The developed reaction yields the compound 2-(1H-1,2,4-triazol-1-yl)-3- phenylpropanoic acid (BnTAA) which shows antifungal activity but also has application as intermediate in the synthesis of new fungicides[34] and as ligand for metal catalysed oxidation of hydrocarbons[35]. Alongside these applications, enantiopure BnTAA can serve as a co-former for co-crystal formation due to its hydrogen bond donor and acceptor capacity[36]. In addition, it can also serve as a salt former for diastereomeric resolution of chiral amines due to its carboxylic acid[37]. Since triazol compounds are interesting synthetic building blocks[38], BnTAA is commercially available as a racemic compound from at least 3 different companiesa. However, no litterature record is available for the formation of enantiopure or racemic BnTAA.

a Providers: AKos Consulting & Solutions (AKOS022818342), Ambinter (Amb18255300), Aurorafinechemical (A07.531.465) 223 Appendix B

Scheme B-1 Schematic of a green 2-step synthesis of BnTAA.

The protocol is the following (Scheme B-1):

Step 1[39].To a stirred solution of 2g of DL-PhenylAlanine and 5.04g of potassium bromide (3.5 eq) in 16mL of 2.5M H2SO4 was added dropwise at 0°C 1.040g of sodium nitrite (1.25 eq) in H2O. Once all the sodium nitrite was added, the mixture was stirred for 1h at 0°C and then for 6h at room temperature (RT). The mixture was extracted three times with ethyl acetate. The combined organic solution was washed with brine, dried (MgSO4) and concentrated in vacuum to give a crude called M.

Step 2. 3.34g of triazol (4 eq) and 1.45g of NaOH (3 eq) are added to 20 mL of water. The mixture is stirred at 50°C for 2h30. Then, the crude M is added and the mixture is stirred for 14h at 50°C. HCl (37%) is added in a dropwise manner, slowly bringing the pH down to about 2. A white product starts precipitating around pH 4. The mixture is left to equilibrate overnight, filtered and washed with water. The solid is dried on vacuum and recrystallized in diethylether. 1.47g (56%b) of a white solid is obtained.

This synthesis starts from an enantiopure essential amino acid, L-phenylalanine, hence from the chiral pool. This amino acid will undergo two water-based reactions, involving three SN2 reactions on its asymmetric carbon. The overall result is the inversion of the stereochemistry of (S)-Phenylalanine to yield the (R)-enantiomer of the desired product with an enantiomeric excess higher than 99%.

The underlying mechanism of this reaction (Scheme B-2) is assumed to involve an intramolecular SN2 of the hydroxyl of the acid on the newly formed and unstable diazonium compound (A) to yield a α-lactone (B). Then, the lactone will undergo a slower intermolecular SN2 of the nucleophile, here the bromide ion on the same carbon, opening the lactone and yielding the intermediate, (S)-2-Bromo-3-phenylpropanoic acid ((S)-BnBAA) (C). This compound undergoes again a SN2 reaction involving the deprotonated 1,2,4-triazol on the bromine, inducing a Walden inversion leading to the formation of the (R)-enantiomer of the product, (R)-2-(1H-1,2,4-triazol-1-yl)-3-phenylpropanoic acid ((R)-BnTAA) (D). The first part of the overall reaction is already described in the literature[39]. It derives from the Sandmeyer reaction on aryl diazonium salts[40] and is already considered to be environmentally friendly since its reactants and products possess relatively low hazard potential[41]. The second reaction is unpreceded in the literature, and was developed to be carried out in the greenest condition possible while yielding an efficient process.

b With respect to starting amino acid 224 Appendix B

Scheme B-2 Mechanism of the reaction, explaining the inversion of the stereochemistry of the final product compared to the starting material.

HOW IS THIS SYNTHESIS GREEN? First, both reactions were run in water, the green solvent by definition. Moreover, the other solvents used respectively for extraction and purification are ethyl acetate and diethylether which are both recognized as generally safe (GRAS) by the FDA.

Secondly, both reactions were designed to minimize waste production (figure B-1). After the first reaction, the organic phase extracted with ethyl acetate is concentrated and dried under vacuum and directly used as such. Ethyl acetate was not considered in the waste since after concentration, it was reusable and did not need disposal. In total, this synthesis produces two acidic aqueous waste and one non-chlorinated organic waste. The production of waste for a reaction can be evaluated using the E-factor, created by Sheldon, which is calculated by dividing the mass of waste produced by the mass of product made[42,43]. Consequently, the E- factor of the process was calculated. As both reactions were run in water, the mass of solvent does not enter in the calculation of the E factor as decided by its inventor[29]. Consequently, the E-factor of the whole process can be calculated by taking as mass of waste, the mass of the reagents that were not used to make the product of the first reactionc, the quantity of solid in the aqueous waste from the second reaction and the mass of the organic waste from the recrystallization. The E-factor of this reaction is 12.5 which is in the fine chemicals range. The value is quite low compared to that of the synthesis of other pharmaceutical compounds, which is comprised between 25 and 100 but also in the lower part of fine chemicals whose E-factor typically reaches up to 50[42,44].

c The excess of reactants since the formation of the diazonium only create water and its consumption liberates a gas, N2 225 Appendix B

Figure B-1 Schematic view of waste production for the whole process. The filtration is symbolized by the grey line. The cake is given by the doted arrow.

Thirdly, the hazard of every used reagent or formed product was assessed to conclude that the reaction is globally non-hazardous. For the first reaction, the starting material is a non- toxic amino acid and it is reacted with potassium bromide, which is only irritating when put in contact with the eyes, and sulfuric acid, which is corrosive. Sodium nitrite is the only reagent, which possesses a real toxicity if ingested and can harm the environment. However, it is reacted and by the end of the reaction, most of it has been transformed into nitrogen gas and water. Regarding BnBAA, it can be considered of low hazard since it completely transforms in the second reaction and its manipulation at room temperature does not require more precaution than for the other compounds. For the second reaction, 1,2,4-triazole is considered as quite safe and sodium hydroxide is only corrosive. As for the products of the reaction, the two triazolated products are fungicides and can be considered relatively safe. Cinnamic acid and phenyllactic acid can both cause skin and eye irritation but can be considered relatively safe. Regarding, the auxiliaries, 37% HCl is used to precipitate the product after the second reaction. Though corrosive and toxic for the respiratory system, its toxicity remains manageable with the use of gloves, labcoats and an aspiration unit for the fumes. Finally, both reactions have a relatively low energy consumption as the first one is mostly carried out at room temperature and ambient pressure, except for the first part at 0°C, and the second one is heated at 50°C, still at ambient pressure. In addition, the product is made from renewable feedstocks (amino acids) and the product is enantiopure. No further resolution is thus needed. 226 Appendix B

Figure B-2 NMR evolution in D2O of the starting material, the product and the by-products over time and at different temperatures. For each compound, one characteristic peak was chosen to be followed. All the characteristic peaks integrate for 1H of the molecule they refer to. From left to right, the peaks belong to cinnamic acid (1H from the double bond), BnTAA (1H from the chiral carbon), the 4-triazol by-product (1H from the chiral carbon), the starting material BnBAA (1H from the chiral carbon) and the by-product with water, 3-Phenyllacticacid (1H from the chiral carbon). Each NMR spectra was cut and the base line was smoothened for visual purposes without affecting the integrations. Each letter from the top left corner of each spectra corresponds to a certain time and temperature: a) 2h / RT, b) 1d / RT, c) 2d / RT, d) 4d / RT, e) 5d / 40°C, f) 6d / 50°C, g) 7d / 80°C & h) 8d / 65°C

DEVELOPMENT AND OPTIMIZATION Developing and optimizing an efficient process for SN2 in water is challenging. First, sodium hydroxide was chosen as a base instead of sodium or potassium carbonate to increase reactivity with the triazole (higher difference of pKA) while giving the same waste at the end. Then, HCl was preferred to sulfuric acid for the reprotonation of the acid because it would only produce table salt and water after reaction with the base.

The next parameters to tune were the selectivity and the speed of the reaction. Indeed, SN2 reactions are always in competition with second order elimination (E2) reactions. In addition, because water was used as solvent, hydroxyl ions compete with the triazole as nucleophile. Regarding the kinetics, the rate of the reaction is slowed down due to the use of water, which is protic and decreases the reactivity of the nucleophile. Heating is not always preferred to increase yield, as selectivity will be affected considering rates of both SN2 and E2 reaction are impacted. To identify ideal reaction time and temperature, the reaction behavior was studied using H1NMR (figure B-2). The results of this experiment clearly showed that no 227 Appendix B matter the time of reaction, the use of temperatures lower than 80°C did not lead to an increase in the rate of elimination, maintaining it slow compared to the others. Moreover, at room temperature the competition reaction with water as a nucleophile was almost inexistent. However, when increasing the temperature, the rate of this reaction increased and eventually overcame the SN2 reaction involving the triazole, as observed at 40°C. Past 50°C, the reaction with water became so fast, it was almost the only one occurring. Even though in principle one could therefore perform the entire reaction at room temperature, after 2 days, the desired substitution slowed down strongly and after two more days of reaction, the increase of product was trivial. Consequently, if a higher yield is aimed for as well as consumption of all the starting material, heating is required. We chose to split the apple in two in order to have a good compromise between selectivity and reaction time, selecting a reaction temperature of 50°C.

Figure B-3 NMR in D2O of the mixture obtained post reaction with the optimized condition, before addition of HCl. The characteristic peaks shown are the same as in figure 3, with the same order. The NMR was cut and the base line smoothened for visual purposes.

In order to decrease the reactivity of water, the concentration of triazole in water was increased. To increase selectivity and rate of the reaction, the quantity of base was set to 0.75 equivalent of triazol (Le Chatelier’s principle). Finally, to be sure the base reacts first with the triazole, they were stirred in water for 2h30 at 50°C prior to the addition of BnBAA. This optimized process was ran for 14 hours and the final mixture was analyzed by NMR (figure B-3). After 14h there is no more reactant present. Furthermore, the selectivity in the right product is higher than 50%, and if the triazole by-product is included, the selectivity of the triazole over the elimination reaction and the hydroxyl substitution is 73% for 27%, meaning at leastd 73% of the reagent reacted with the triazole.

CONCLUSION A 2-step green SN2 based synthesis was developed and optimized to yield an efficient and environmentally friendly process leading to the formation of an enantiopure product. The optimized process overcame two major obstacles of a secondary carbon-SN2 reaction in water:

d The BnOHA by-product is also a by product of the first reaction yielding the BnBA, consenquently, its final ratio includes the ratio formed after the first reaction and the ratio forned after the second reaction 228 Appendix B the selectivity towards the right product and the efficiency of the reaction (yield versus time). The control of the temperature and the pre-reaction of the triazole with the base were key to getting this process to work. It must be noted that this process can also be applied on the other enantiomer of phenylalanine, leading to a product of opposite chirality than the one presented here. The reaction would still remain green, even though the starting material is no longer a natural amino acid. Furthermore, the process could be applied to other α-amino acids to yield new building blocks with a 1,2,4-triazol moiety or even other types of nucleophile.

EXPERIMENTAL NMR spectra were obtained on a 300 MHz spectrometer. Chemical shifts are reported in parts per million (ppm) and were normalized regarding the chemical shift of the peak of the deuterated solvent used. For the 1H NMR spectra, the value of the different solvents used are 13 the following: CDCl3 7.26ppm; D2O 4.79ppm; (CD3)2SO 2.50ppm. For the C NMR spectra, the value of the different solvent used are the following: CDCl3 77.36ppm; (CD3)2SO 39.52ppm. Multiplicities are abbreviated as follows: singlet (s), doublet (d), triplet (t), quartet (q) and multiplet (m). Optical rotations were obtained at a wavelength of 589 nm with an Anton Paar Polarimeter All measurements were done at 20°C. The length of the cell is 100.00 mm. The concentration “c” has units of g/100 mL. For the value of R-BnTAA, the DMSO used for the measurement was distilled just before. The distillation was done by taking an excess of DMSO and distillating it with the water under vacuum to get in the remaining DMSO, less than 20ppm of water[45]. The enantiomeric purity of R-BnTAA was determined as enantiomeric ratios (er) using reverse Chiral HPLC on the MBnTA obtained from the methyl esterification of R-BnTAAe. The set-up is the following: the separation module is an Alliance Waters 2695 and the detector is a Waters 2998. The column used is a Lux® 5µm Amylose-1 chiral column with the following dimensions 250x4.6mm and a particule diameter of 5μm. The chromatograms were obtained from a detection at 254 nm. The mobile phase was 100% methanol at a flow rate of 1 mL/min. The temperature of the oven was 20°C.

Condition for the NMR study. The reaction was started with BnBAA, 6.5 equivalent of triazole and 2.25 equivalent of sodium hydroxide in water at room temperature. Then, a sample of the reaction mixture was taken from the mixture at different time and temperature, freezedried and analysed by NMR in deuterated water.

(S)-2-Bromo-3-phenylpropanoic acid ((S)-BnBAA). M1 was purified by column chromatography (Packing: 10:89:1 (EA/PE/AcOH), first eluent: 10:89:1, second eluent: 20:79:1, third eluent: 30:69:1 and fourth eluent: 40:59:1)[39]. 1.942g of an oil (70%) was obtained after removal of the traces of acetic acid by lyophilization. Rf= 0.77 (1:1 EA/PE) UV −1 −1 1 revelation. [α]= -10.38 deg·mL·g ·dm in MeOH. H NMR (300 MHz, CDCl3) ẟ (ppm): 3.27 (dd, 1H, J = 14.2, 7.3 Hz, CH-CHaHb- CAr), 3.49 (dd, 1H, J = 14.2, 8.1 Hz, CH-CHaHb- Car)), 4.44 (dd, 1H, J = 8.1, 7.3 Hz, CH2-CH-C=O), 7.23-7.26 (m, 2H, CAr=CArH-CArH), 7.30-7.38 13 (m, 3H, CArH=CArH-CArH). C NMR (75 MHz, CDCl3) ẟ (ppm): 41.06 (CAr-CH2-CHBr), p 45.08 (CH2-CHBr-C=O), 127.83 (CAr H), 129.10 & 129.50 (CArH), 136.64 (CAr), 175.56 (HO- C=O).

e BnTA is too polar to do normal phase Chiral HPLC (it would stuck to the column) and when doing reverse chiral HPLC it is not retained at all, thus the two enantiomers cannot be separated. That’s why, BnTA was esterified to have a less polar compound to use in reverse Chiral HPLC. 229 Appendix B

(R)-2-(1H-1,2,4-triazol-1-yl)-3-phenylpropanoic acid ((R)-BnTAA). [α]= 128.7 deg·mL·g−1·dm−1in DMSO. Mp = 189.7°C. 1H NMR (300 MHz, DMSO) ẟ (ppm): 3.37 (dd, 1H, J = 14.3, 10.8 Hz, CH-CHaHb- CAr), 3.48 (dd, 1H, J = 14.3, 4.8 Hz, CH-CHaHb- Car)), 5.41 (dd, 1H, J = 10.8, 4.8 Hz, CH2-CH-C=O), 7.06-7.10 (m, 2H, CAr=CArH-CArH), 7.14-7.24 (m, 13 3H, CArH=CArH-CArH), 7.90 (s, 1H, N=CH-N), 8.42 (s, 1H, N=CH-N). C NMR (75 MHz, p DMSO) ẟ (ppm): 36.43 (CAr-CH2-CH), 63.08 (CH2-CH-C=O), 126.58 (CAr H), 128.27 & 128.77 (CArH), 136.95 (CAr), 144.65 & 150.99 (N=CH-N), 169.81 (HO-C=O).

Methyl-(1H-1,2,4-Triazole-phenylmethyl)-acetate (MBnTA) for Chiral HPLC analysis. 95% sulfuric acid (5 µL, 0.2 eq) is added to a solution of BnTAA (100mg, 0.46 mmol) and activated 4A molecular sieves (80mg) in distillated methanol (5 mL) at room temperature. The solution is then heated up to reflux and stirred for 24h. The mixture is filtered and the solvent is removed under reduced pressure[46]. The solid is dried under vacuum and was analysed by NMR and Chiral HPLC. The conversion rate obtained by NMR is 37%. 1H NMR (300 MHz, DMSO) ẟ (ppm): 3.40 (dd, 1H, J = 14.1, 10.7 Hz, CH-CHaHb- CAr), 3.51 (dd, 1H, J = 14.1, 5.1 Hz, CH-CHaHb-CAr)), 3.68 (s, 3H, O-CH3), 5.66 (dd, 1H, J = 10.6, 5.1 Hz, CH2-CH-C=O), 7.05-7.11 (m, 2H, CArH), 7.13-7.24 (m, 3H, CArH), 7.97 (s, 1H, N=CH-N), 8.46 (s, 1H, N=CH- N).

ASSOCIATED CONTENT

Supporting Information This material is available free of charge via the Internet at http://pubs.acs.org/.

AUTHOR INFORMATION

Corresponding Author ‡E-mail: [email protected]. Web: https://uclouvain.be/fr/repertoires/michael.guillot.

Notes The authors declare no competing financial interests.

ACKNOWLEDGMENT The authors would like to thank the UCL and FNRS for financial support. M. Guillot is a FRIA scholar (33849008) and this work is carried out in the context of the PDR T.0149.19.

REFERENCES

[1] Anastas, P. T.; Warner, J. C. Green Chemistry: Theory and Practice; Oxford University Press, Inc.: New York, 1998. [2] Clark, J. H. Green Chem. 1999, 1(1), 1-8. [3] Anastas, P. T.; Eghbali, N. Chem. Soc. Rev. 2010, 39, 301–312. [4] Li, C. J., Trost, B. M. Proceedings of the National Academy of Sciences. 2008, 105(36), 13197-13202. [5] Geiser, K. Making safer chemicals. In Sustainability Science and Engineering; edition n°1; Abraham, M. A. Elsevier Science: Amsterdam, The Netherlands, 2006; Vol. 1, 161-175. [6] Gronnow, M. J.; White, R. J.; Clark, J. H.; Macquarrie, D. J. Org. Process Res. Dev. 2005, 9, 516−518. 230 Appendix B

[7] Xiong, P.; Li, J.; Bu, H.; Wei, Q.; Zhang, R.; Chen, S. J. Solid State Chem. 2014, 215, 292–299. [8] Balasubramanyan, S.; Shephard, M. C. (1978) U.S. Patent US4079143A. [9] Boyle, F. T. (1986) U.S. Patent US4625036A. [10] Stubenrauch, G.; Ammermann, E.; Hamprecht, G.; Pommer, E.-H. (1985) U.S. Patent US4496738A. [11] Kramer, W.; Elbe, H.-L.; Buchel, K. H.; Brandes, W. ; Frohberger, P.-E. (1983) U.S. Patent US4406909A. [12] Rogers, R. B.; Carson, C. M.; Ehr, R. J. (1980) U.S. Patent US4215127A. [13] Funaki, Y.; Tanaka, S.; Matsuo, N. (1983) US4379921A. [14] Itoh, H.; Yoneda, R.; Tobitsuka, J.; Matsuhisa, T.; Kajino, H.; Ohta, H.; Hayashi, N.; Takahi, Y.; Tsuda, M.; Takeshiba, H.; Chem. Pharm. Bull. 2000, 48, 8, 1148 – 1153. [15] Belai, I.; Darvas, B.; Bauer, K.; Tag El-Din, M. H.; Pesticide Science. 1995, 44, 3, 225 – 232. [16] Kato, Y.; Hayashi, S.; Yamazaki, A.; Yamamoto, M.; Asaoka, Y.; Yamada, M.; Yamada, N. (2015) European Patent EP2832724A1. [17] Nagabukuro, H.; Edmondson, S. D.; Sinharoy, M. S.; Denney, W. S.; Frenkl, T. L. (2011) China Patent WO201143942A1. [18] Barton, P. J.; Clarke, D. S.; Davies, C. D.; Hargreaves, R. B.; Pease, J. E.; Rankine, M. T. (2004) WO200411410A1. [19] Devereux, N. J.; McKenna, J. (2011) US2011190360A1. [20] Cristalli, G.; Eleuteri, A.; Volpini, R.; Vittori, S.; Camaioni, E.; Lupidi, G. J. Med. Chem. 1994, 37, 1, 201-205. [21] Marzano, C.; Pellei, M.; Colavito, D.; Alidori, S.; Lobbia, Giancarlo G.; Gandin, V.; Tisato, F.; Santini, C.; J. Med. Chem. 2006, 49, 25, 7317-7324. [22] Holm, S. C.; Straub, B. F. Org. Prep. Proced. Int. 2011, 43, 319-347. [23] Zhang, P. Z.; Zhou, S. F.; Li, T. R.; Jiang, L. Chinese Chemical Letters. 2012, Vol. 23, 12, p. 1381- 1384. [24] Ding, M. W., Yuan, D. (2016). China patent CN105820129A. [25] Zhao, X.-X.; Ma, J.-P.; Dong, Y.-B.; Huang, R.-Q. Cryst. Growth Des. 2007, 7, 1058. [26] Katritzky, A. R.; Wang, J.; Leeming, P.; Steel, P. J. Tetrahedron-Asymmetry. 1996, 7, 6, 1631-1640. [27] Artico, M.; Ragno, R.; Porretta, G. C.; Massa, S.; Musiu, C.; Spiga, M. G.; Corrias, S.; La Colla, P. Med. Chem. Res. 1996, 6, 3, 137-147. [28] Bachmann, S.; Flohr, A.; Groebke Zbinden, K.; Koerner, M.; Bernd, K.; Peters, J.-U.; Rudolph, M. (2013). European patent EP2753604B1. [29] Buckler, R. T.; Kurchacova, E. (1978). US patent US4109001A. [30] Zakaryan, G. B.; Hayotsyan, S. S.; Ayvazyan, A. G.; Tamazyan, R. A.; Panosyan, H. A.; Danagulyan, G. G.; Attaryan, H. S. Chem. Heterocycl. Compd. 2016, 52(4), 253–256. [31] Meti, G.Y.; Kamble, A. A.; Kamble, R. R.; Somagond, S. M.; Devarajegowda, H.C.; Kumari, S.; Kalthur, G.; Adigad, S. K. Eur. J. Med. Chem. 2016, 121, 221-231. [32] Kofman, T. P.; Trubitsyn, A. E.; Dmitrienko, I. V.; Glazkova, E. Y.; Tselinskii, I. V. Russ. J. Organ. Chem. 2007, Vol. 43, 5, 758-764. [33] Qin, J.; Ma, J.-P.; Liu, L.-L.; Huang, R.-Q.; Dong, Y.-B. Acta Crystallogr. 2009, 65, 66. [34] Ishikawa, H.; Taniguchi, M.; Ono, T.; Hirayama, K.; Kajikawa, K. (1992). Japan patent JP04154770. [35] Elgammal, R. A.; Foister, S. (2011). WIPO patent WO2011035064. [36] Desiraju, G. R. Angew. Chem. Int. Ed. 1995, 34 (21), 2311-2327. [37] Kozma, D. Handbook of optical resolutions via diastereomeric salt formation; CRC Press: Boca Raton, 2001, 3, 9-10. [38] Sathish Kumar, S.; Kavitha, H. P. Mini-Rev. Org. Chem. 2013,10(1), 40-65. [39] Tka, N.; Kraïem, J.; Hassine, B. B. Synth. Commun. 2013, 1, 43, 735–743. [40] Hodgson, H. H. Chem. Rev. 1947, 40, 2, 251-277. [41] Van Draanen, N. A.; Hengst, S. J. Chem. Educ. 2010, 87, 6, 623-624. [42] Sheldon, R. A. Green Chem. 2007, 9, 1273-1283. [43] Constable, D. J. C.; Curzons, A. D.; Cunningham, V. L. Green Chem. 2002, 4, 521-527. [44] Sheldon, R. A. The E Factor, (available at https://www.sheldon.nl/roger/efactor.html). 231 Appendix B

[45] Gaylord Chemical Company, LLC. Drying of dimethyl sulfoxide. 2015, Bulletin 109, (available at https://www.gaylordchemical.com/wp-content/uploads/2015/07/GC-Literature-109B.pdf?x86863) [46] Shahnaz Perveen*,1, Arfa Yasmeen1, Muhammad Aitmaud Khan2, Ahsana Dar2, Rehana Jafri1 and Amir Ahmed1, Letters in Drug Design & Discovery, 2010, Vol. 7, No. 1, 14-17

Supporting information Waste production data & E-factor

For a reaction starting with 1g of D-Phenylalanine, the first reaction produces 3.39g of waste from the aqueous phase (water not included). Then the second reaction produces 3.77g of aqueous waste (water not included) and 2.04g of organic waste (solvent included).

With a yield of 56%, there is 0.736g of product formed.

For the E-factor, the calculation is the following: 3.39 + 3.77 + 2.04 E-factor = = 12.5 0.736 DSC of (R)-BnTAA

Figure B-4 DSC of (R)-BnTAA showing a fusion peak followed by recrystallization and fusion again.

The DSC measurement of (R)-BnTA clearly shows that two polymorph of (R)-BnTA exists, one that melts at 189.7°C and one that melts at 225.7°C. According to the value of ΔH°, the 232 Appendix B stable one is the one melting at 225.7°C. The metastable one once melted recrystallize into the more stable one, which is classic for 2 polymorphs.

Chiral Reverse HPLC

Figure B-5 Chiral reverse HPLC of racemic MBnTA, showing both enantiomers.

Figure B-6 Chiral reverse HPLC of the crude of the methyl esterification of BnTAA, only showing one enantiomer of MBnTA, proving enantiopurity of BnTAA. The large peak before accounts for solvent and BnTAA (very polar it is not retained in reverse HPLC). 233 Appendix C Appendix C: SI for chapter 2

Part 2. XRPD for congruence test All compounds gave a similar diffractogram, showing only (S,S)-co-crystal as the solid phase. Figure C-1 displays the cases of Toluene and THF for example.

Figure C-1 Diffractograms obtained for THF (blue) and Toluene (green), compared to the diffractograms of the (S,S)-co-crystal (red).

Solubility curves Table C-1 Values of temperature/concentration obtained from crystal16 experiments for each solvent.

Toluene Acetone Acetonitrile Ethanol Butanol Propanol THF C T C T C T C T C T C T C T mg/ml ºC mg/ml ºC mg/ml ºC mg/ml ºC mg/ml ºC mg/ml ºC mg/ml ºC 20.07 18.7 75.07 3.88 87.65 25 50.46 16.2 25.04 12.2 49.82 22.1 149.99 1.4 40.17 30.8 87.63 8.3 100.52 27.3 62.58 20.3 50.36 25.7 62.54 26.9 199.9 10.4 50.07 34.7 100.31 10.3 112.27 29.7 74.83 24 62.96 30 75.13 28.3 250.7 16.4 60.07 37.7 113.07 13 125.49 31.9 87.74 27.8 75.46 33.6 87.59 31.2 300.01 21.9 69.97 40.8 124.7 16.4 137.33 33.5 100.12 29.5 87.61 36.1 100.09 33.5 350.28 27.7 80.04 42.8 150.13 19.7 149.41 35 112.35 32.1 99.92 38.4 112.53 35.4 399.69 32.2 90.04 44.5 175.91 23.5 174.81 37.5 124.65 34.8 125.2 42.5 125.09 38.9 450.63 36 234 Appendix C

99.84 46.5 200.22 25.5 200.48 40.5 150.15 37.5 150.41 45.7 150.12 42.9 601.08 47 25 22.3 225.06 28.6 250.79 43.7 25.15 1.9 37.4 19 24.75 8.1 29.98 25.5 249.66 30.6 50.06 14.5 174.78 41.2 112.85 40 37.7 15.9 35.1 28.3 275.15 35.3 75.17 22 44.93 32.5 65.1 39 75.12 41.3 109.95 47.6 120.11 49.7

0,003 0,0031 0,0032 0,0033 0,0034 0,0035 0,0036 0,0037 0,0038 1 ln(Ks) = -5384,4/T + 17,454 R² = 0,9994 0 ln(Ks) = -7463,2/T + 23,351 -1 R² = 0,9949

-2 ln(Ks) = -9919,9/T + 30,069 ln(Ks) -3 R² = 0,9967 [-] -4 ln(Ks) = -8582,6/T + 25,388 R² = 0,9967 -5 ln(Ks) = -9646,6/T + 28,019 R² = 0,996 -6 ln(Ks) = -10918/T + 31,177 ln(Ks) = -9492,5/T + 27,921 R² = 0,9983 R² = 0,9931 -7 1/T [K-1] Toluene THF Butan-1-ol Acetone Acetonitrile Ethanol Propanol

Figure C-2 Linear regression of Van’t Hoff relation for (S,S)-BnFTP-PBA co-crystal in different solvents.

Table C-2 Values of temperature of each concentration of (R)-BnFTP + (S)-PBA co-crystal in toluene.

T (°C) C (mg/mL) -1.1 75.81 3.2 100.12 6.4 126.29 9.3 150.44 11.9 175.68 14 199.16 15.9 225.57 235 Appendix C

24.3 349.66 17.5 251.6 26.9 401.74 20.2 274.82 23.1 325.44

Calculation for H°, S° and the solubility of the (S,S) co- crystal at 20°C For each solvent, Van’t Hoff’s law was verified and the value of H° and S° could be extracted with the following reasoning:

The integrated form of Van’t Hoff’s law is the following:

∆퐻° 푙푛퐾 = − + 퐶 (1) 푅푇 With the following two thermodynamic relation, equation (1) can be obtained:

∆퐺° = ∆퐻° − 푇∆푆°

∆퐺° = −푅푇푙푛퐾

∆퐻° ∆푆° 푙푛퐾 = − + (2) 푅푇 푅 By equalizing equation (1) and (2), C can be determined equal to:

∆푆° 퐶 = 푅 For a dissolution reaction with a (1:1) co-crystal of two compounds, K = s2 (with s in mol/L), and by injecting this in equation (2), the following expression can be obtained:

∆푆° ∆퐻° − 푠 = 푒 2푅 푒 2푅푇

With this equation, the solubility at 20°C can be extracted for each solvent. 236 Appendix C

HPLC for breaking the co-crystal

Figure C-5 Chromatograms for the tests of liquid-liquid separation of the co-crystal with NaOH: from top to bottom: with toluene, with ethyl acetate, for pH 13 and for pH 14. 237 Appendix C Part 3. Results of the acid/base screening

Figure C-6 From top to bottom, left to right, chromatograms of the starting mixture for the experiment, of racemization test with triethylamine, of racemization test with DBU, of racemization test with DMAP, of racemization test with formic acid and of racemization test with acetic acid. 238 Appendix C

Further study of racemization of (S)-BnFTP with DBU

Figure C-7 From top to bottom, left to right, chromatograms for the starting mixture for the experiment, after 30min with 5%mol of DBU, after 1h with 5%mol of DBU, after 6h with 5%mol of DBU, after 30min with 1eq of DBU and after 1h with 1eq of DBU.

Racemization study by polarimetry Measurement set-up

Figure C-8 Visual of the system with on the left the 100ml reaction vessel on the stirring plate and on the right the gear pump, the polarimeter and the measurement cell. The blue tubing is the recycle loop connecting the reaction vessel and the measurement cell. 239 Appendix C

Figure C-9 3-view 3D-Schematic of the system: On the top left, one can see the process from above with polarimeter represented as a transparent grey parallelepiped containing the cell inside shown as a black cylinder. The tubing (thin long white cylinder) are connecting the cell to the pump (metallic grey cube) upstream and to the vessel (transparent blue cylinder) downstream, as indicated by the black arrow

Study of the influence of (S)-BnFTP concentration The values plotted in figure 2-7 of the manuscript are detailed in table C-4.

Table C-4 Values of the observed rate constant with linearization error for different values of BnFTP concentration, S0.

Linearization -1 -1 S0 (g.L ) k' (h ) error on k' (h-1) 5 5.369 0.013 6.25 5.158 0.009 7.5 5.731 0.009 8.75 4.476 0.006 10 4.329 0.005 11.25 5.086 0.005 12.5 4.939 0.005 Average Standard k’ (h-1) deviation 5.01 0.48 Linearization for the study of the influence of (S)-BnFTP concentration

240 Appendix C

0 0,1 0,2 0,3 0,4 0 0,2 0,4 0,6 0 0

-0,5 ln(E) = -5.1585t - 0.014 -0,5 ln(E) = -5.369t - 0.005 R² = 0.9991 R² = 0.9987 -1 ln(E) ln(E) -1 -1,5 [-] [-] -2 -1,5 -2,5

-2 -3 t [h] t [h] Figure C-10 (S)-BnFTP 5g/L - DBU 5.1g/L Figure C-11 (S)-BnFTP 6.25g/L – DBU 5.1g/L

0 0,2 0,4 0 0,2 0,4 0,6 0 0 ln(E) = -4.476t - 0.05 -0,5 ln(E) = -5.731t - 0.019 -0,5 R² = 0.9993 R² = 0.9993 -1 -1 ln(E) ln(E) -1,5 [-] -1,5 [-] -2

-2 -2,5

-2,5 -3 t [h] t [h] Figure C-12 (S)-BnFTP 7.5g/L – DBU 5.1g/L Figure C-13 (S)-BnFTP 8.75g/L – DBU 5.1g/L

0 0,2 0,4 0,6 0 0,2 0,4 0,6 0,8 0 0 ln(E) = -5.086t - 0.032 ln(E) = -4.329t - 0.079 -0,5 -0,5 R² = 0.9997 R² = 0.9995 -1 -1 ln(E) ln(E) -1,5 [-] [-]-1,5 -2 -2,5 -2 -3 -2,5 -3,5 t [h] t [h] Figure C-14 (S)-BnFTP 10g/L – DBU 5.1g/L Figure C-15 (S)-BnFTP 11.25g/L – DBU 5.1g/L 241 Appendix C

0 0,2 0,4 0,6 0

-0,5 ln(E) = -4.939t - 0.046 R² = 0.9997 -1 ln(E) -1,5 [-] -2

-2,5

-3 t [h] Figure C-16 (S)-BnFTP 12.5g/L – DBU 5.1g/L

Value of k’ for the influence of DBU concentrations Table C-5 Value of the observed rate constant k’ with its experimental error for every concentration of base added.

-1 -1 -1 -1 B0 (g/L) B0 (mol/L) k’ (h ) Linearization error on k’ (h ) k (L.mol .h )

1.02 0.00667 0.987 ±0.001 147.3 2.04 0.0134 2.138 ±0.003 159.6 5.1 0.0335 5.013 ±0.007 149.6 7.14 0.0469 7.421 ±0.020 158.2 9.52 0.0625 9.93 ±0.075 158.8 12.24 0.0804 12.729 ±0.031 158.3

Linearization for study of the influence of DBU concentration 0 1 2 3 0 0,1 0,2 0,3 0,4 0 0

-0,5 ln(E) = -0.987t - 0.01 -0,5 ln(E) = -7.421t - 0.008 R² = 0.9994 R² = 0.9982 -1 -1 ln(E) ln(E) -1,5 -1,5 [-] [-] -2 -2

-2,5 -2,5

-3 -3 t [h] t [h] Figure C-17 DBU 1.02g/L Figure C-18 DBU 7.1g/L 242 Appendix C

0 0,1 0,2 0,3 0 0,1 0,2 0,3 0 0 -0,25 ln(E) = -9.93t + 0.075 -0,5 -0,5 ln(E) = -12.729t - 0.023 R² = 0.9906 -0,75 -1 R² = 0.9991 -1 ln(E)-1,25 ln(E)-1,5 [-] -1,5 [-] -2 -1,75 -2 -2,5 -2,25 -2,5 -3 -2,75 -3,5 t [h] t [h] Figure C-19 DBU 9.3g/L Figure C-20 DBU 12.24g/L Linearization for study of the influence of the solvent 0 0,2 0,4 0,6 0 0,05 0,1 0,15 0,2 0 0 ln(E)= -3.883t - 0.055 -0,5 ln(E) = -19.503t - 0.018 -0,5 R² = 0.9994 R² = 0.9993 -1 -1 ln(E) ln(E) -1,5 [-]-1,5 [-] -2 -2,5 -2 -3 -2,5 -3,5 t [h] t [h] Figure C-21 THF Figure C-22 Propan-1-ol

0 0,05 0,1 0 0,2 0,4 0,6 0 0 ln(E) = -37.015t - 0.09 -0,5 -0,5 ln(E) = -5.886t - 0.067 R² = 0.9913 R² = 0.9994 -1 -1 ln(E)-1,5 ln(E) -1,5 [-] -2 [-] -2 -2,5 -3 -2,5 -3,5 -3 t [h] t [h] Figure C-23 Ethanol Figure C-24 Acetone 243 Appendix C

0 0,2 0,4 0,6 0 0,2 0,4 0 0 -0,5 ln(E) = -8.03t - 0.063 ln(E) = -7.073t - 0.075 R² = 0.9997 -0,5 -1 R² = 0.9986 -1 ln(E) -1,5 ln(E) [-] -2 [-]-1,5 -2,5 -2 -3 -3,5 -2,5 t [h] t [h] Figure C-25 Butan-1-ol Figure C-26 Acetonitrile

Linear model error Regarding the linear model error, one can notice that it tends to increase with the quantity of base used. This is due to the increase in the speed of reaction. Indeed, the faster the reaction, the higher the error since at the beginning of each measurement, when DBU is added to the system, there are a couple of minutes before the DBU is homogenously mixed in the solution. During this period, the optical rotation fluctuates around the optical rotation that would be measured if DBU was homogenously present in the mixture as observed experimentally (figure D-20). In addition, the faster the reaction, the more predominant this effect becomes for the linearization.

A) B)

Figure C-27 A) The fluctuation effect at the beginning of one measurement (optical rotation as a function of time). B) The fluctuation effect when linearization is applied on the same measurement, showing that the fluctuation revolves around the same rate as shown by the linear fit. 244 Appendix C

Linearization for study of influence of (R)-3- phenylbutyric acid on racemization 0 5 10 0 5 10 0 0 -0,02 ln(E) = -0.0123t - 0.0086 -0,05 ln(E) = -0.0428t - 0.0357 R² = 0.9936 -0,1 R² = 0.9947 -0,04 -0,15 ln(E)-0,06 ln(E) -0,2 [-] -0,08 [-] -0,25 -0,1 -0,3 -0,35 -0,12 -0,4 -0,14 -0,45 t [h] t [h] Figure C-28 5g/L – 0.74eq Figure C-29 5.5g/L – 0.81eq

0 1 2 3 4 0 1 2 0 0 -0,2 ln(E) = -0.3449t - 0.1257 -0,2 ln(E) = -1.0847t - 0.0819 R² = 0.9925 -0,4 -0,4 R² = 0.9947 -0,6 ln(E) -0,6 ln(E) -0,8 [-] -0,8 [-] -1 -1,2 -1 -1,4 -1,2 -1,6 -1,4 -1,8 t [h] t [h]

Figure C-30 6.5g/L – 0.96eq Figure C-31 7g/L – 1.03eq

0 1 2 0 0,5 1 0 0 ln(E) = -2.298t - 0.1362 ln(E) = -0.6359t + 0.0022 -0,2 -0,5 R² = 0.9958 R² = 0.9997 -0,4 -1 ln(E) ln(E) -0,6 [-] [-]-1,5 -0,8 -2 -1

-1,2 -2,5 t [h] t [h]

Figure C-32 8g/L – 1.16eq Figure C-33 9g/L – 1.31eq 245 Appendix C

Racemization study by HPLC Linearization for study of influence of temperature on racemization in presence of (R)-3-phenylbutyric acid 5 experiments performed at different temperature followed by their linearization. Then, the table with the values of observed rate constant with temperature is displayed, followed by the graph of the evolution of the observed rate constants with temperature.

1

0,8

0,6 E [-] 0,4

0,2

0 0 1 2 3 4 5 t [h] 100°C 90°C 80°C 75°C 70°C Figure C-34 Time evolution of the enantiomeric excess E at different temperatures for the same base concentration (6.8g/L).

0 1 2 3 4 5 0 -0,5 ln(E) = -0.3237t + 0.0309 R² = 0.9918 -1 -1,5 ln(E) = -0.4827t + 0.0475 ln(E) R² = 0.9822 -2 [-] -2,5 ln(E) = -0.6682t + 0.0664 R² = 0.9964 -3 ln(E) = -2.2546t + 0.081 R² = 0.9767 ln(E) = -1,2774t - 0.0183 -3,5 R² = 0.9975 -4 t [h] 100°C 90°C 80°C 75°C 70°C Figure C-35 Time evolution of the natural logarithm of the enantiomeric excess ln(E) at different temperatures for the same base concentration (6.8g/L). 246 Appendix C

Tableau C-6 Value of the observed rate constant k’ and its natural logarithm ln(k’) for every temperature T.

T (°C) T (K) 1/T (K-1) k' (h-1) ln(k) 70 343.15 0.00291 0.324 -1.13 75 348.15 0.00287 0.483 -0.728 80 353.15 0.00283 0.668 -0.403 90 363.15 0.00275 1.28 0.245 100 373.15 0.00268 2.26 0.813

2,5

2

k' 1,5 -1 [h ] 1

0,5

0 340 350 360 370 380 T [K] Figure C-36 Evolution of the natural logarithm of the observed constant ln(k’) with temperature T for the same base concentration (6.8g/L), displaying an exponential trend. Appendix D: SI for chapter 3

Part 2. Structural Information of (R)-BnFTP + (S)-PBA co- crystal Table D-1 Crystal data and structure refinement for (R)-BnFTP + (S)-PBA co-crystal (UCL1010_mg_compB).

Identification code mg_compB Empirical formula C25 H30 F N3 O3 Formula weight 439.52 247 Appendix D

Temperature 297(2) K Wavelength 0.71073 Å Crystal system Monoclinic Space group P21 Unit cell dimensions a = 5.7163(7) Å. b = 15.9831(18) Å; β = 98.507(12)°. c = 13.7113(15) Å. Volume 1238.9(2) Å3 Z 2 Density (calculated) 1.178 Mg/m3 Absorption coefficient 0.083 mm-1 F(000) 468 Crystal size 0.50 x 0.50 x 0.07 mm3 Theta range for data collection 3.264 to 25.242°. Index ranges -6<=h<=6, -19<=k<=19, -16<=l<=16 Reflections collected 6625 Independent reflections 4294 [R(int) = 0.0489] Completeness to theta = 25.242° 99.0 % Absorption correction Semi-empirical from equivalents Max. and min. transmission 1.00000 and 0.62605 Refinement method Full-matrix least-squares on F2 Data / restraints / parameters 4294 / 106 / 379 Goodness-of-fit on F2 1.070 Final R indices [I>2sigma(I)] R1 = 0.0510, wR2 = 0.1215 R indices (all data) R1 = 0.0701, wR2 = 0.1331 Absolute structure parameter 0.1(9) Largest diff. peak and hole 0.120 and -0.127 e.Å-3 248 Appendix D

Deracemization n°1

Figure D-1 Chromatograms of from top to bottom sample A0, sample A1 and sample B0. 249 Appendix D

Figure D-2 Chromatograms of from top to bottom sample B1, sample B2 and sample C0. 250 Appendix D

FigureD-3 Chromatograms of from top to bottom sample C1, the filtrate and the cake. 251 Appendix D

Deracemization n°2

Figure D-4 Chromatograms of, from top to bottom, sample D1.1 (crystallization vessel), sample D1.1 (Racemization vessel) and sample D1.2 (Crystallization vessel). 252 Appendix D

Figure D-5 Chromatograms of, from top to bottom, sample D1.2 (Racemization vessel), sample D1.3 (Crystallization vessel) and sample D1.3 (Racemization vessel). 253 Appendix D

Figure D-6 Chromatograms of, from top to bottom, sample D1.4 (Crystallization vessel), sample D1.4 (Racemization vessel) and sample D1.5 (Crystallization vessel). 254 Appendix D

Figure D-7 Chromatograms of, from top to bottom, sample D1.5 (Racemization vessel), sample D1.6 (Crystallization vessel) and sample D1.6 (Racemization vessel). 255 Appendix D

Figure D-8 Chromatograms of sample D2.1 (Crystallization vessel), sample D2.1 (Racemization vessel) and sample D2.2 (Crystallization vessel). 256 Appendix D

Figure D-9 Chromatograms of, from top to bottom, sample D2.2 (Racemization vessel), sample D2.3 (Crystallization vessel) and sample D2.3 (Racemization vessel). 257 Appendix D 258 Appendix D

Figure D-10 Chromatograms of, from top to bottom, sample D2.4 (Crystallization vessel), sample D2.4 (Racemization vessel), Chromatogram of the filtrate (experiment 2) and Chromatogram of the cake (experiment 2).

Deracemization n°3

Figure D-11 Chromatograms of the filtrate and the cake. 259 Appendix D

Racemization of (R)-BnFTP with the 1eq of (S)- PBA when temperature is applied

Figure D-12 From top to bottom, left to right, chromatograms of the starting mixture to racemize at 110°C in toluene, the mixture obtained after 12h at 110°C in toluene, the starting mixture to racemize at 90°C in toluene and the mixture after 2h at 90°C in toluene.

Part 3. HPLC results for all runs Run 1

Figure D-13 Chromatogram of Run 1 filtrate 260 Appendix D

Run 2

Figure D-14 Chromatograms of Run 2 cake (top) and filtrate (bottom).

Run 3

Figure D-15 Chromatograms of Run 3 cake (top) and filtrate (bottom). 261 Appendix D

Run 4

Figure D-16 Chromatograms of Run 4 cake (top) and filtrate (bottom).

Run 5

Figure D-17 Chromatograms of Run 5 cake (top) and filtrate (bottom). 262 Appendix D

Run 6

Figure D-18 Chromatograms of Run 6 cake (top) and filtrate (bottom).

Run 7

Figure D-19 Chromatograms of Run 7 cake (top) and filtrate (bottom). 263 Appendix D

Run 8

Figure D-20 Chromatograms of Run 8 cake (top) and filtrate (bottom).

Run 9

Figure D-21 Chromatograms of Run 9 cake (top) and filtrate (bottom). 264 Appendix D

Run 10

Figure D-22 Chromatograms of Run 10 cake (top) and filtrate (bottom).

Kinetics measurements Study of racemization kinetics of BnFTP and PBA co- crystal in Toluene-water mixtures 0,7

0,6

0,5

E 0,4 [-] 0,3

0,2

0,1

0 0 1 2 3 4 t [h] 0 0.04 0.08 0.16 0.24 Figure D-23 Time evolution of the enantiomeric excess E for the experiment with 0; 0.04; 0.08; 0.16 and 0.32% of water in toluene. 265 Appendix D

0,7

0,6

0,5

0,4 E [-] 0,3

0,2

0,1

0 0 1 2 3 4 t [h] 0.32 0.56 0.8 1.6 3.2 6.4

Figure D-24 Time evolution of the enantiomeric excess E for the experiment with 0.32; 0.56; 0.8; 1.6; 3.2 and 6.4% of water in toluene.

0 0,5 1 1,5 2 2,5 3 3,5 0

-0,5 ln(E) = -0.6771t - 0.6135 R² = 0.956 -1 ln(E) = -0.7752t - 0.3041 ln(E) R² = 0.8544 -1,5 [-] ln(E) = -0.5255t - 0.8646 R² = 0.9877 ln(E) = -0.6288t - 0.9291 -2 R² = 0.9798

-2,5 ln(E) = -0.6217t - 0.8852 R² = 0.9772

-3 t [h] 0 0.04 0.08 0.16 0.32 Figure D-25 Time evolution of the natural logarithm of the enantiomeric excess ln(E) for the experiment with 0; 0.04; 0.08; 0.16 and 0.32% of water in toluene. 266 Appendix D

0 0,5 1 1,5 2 2,5 3 3,5 0 ln(E) = -0.7602t - 0.268 ln(E) = -0.5632t - 0.4104 -0,5 R² = 0.8158 R² = 0.9568 ln(E) = -0.5892t - 0.3321 -1 R² = 0.9858

ln(E) -1,5 ln(E) = -0.4628t - 0.6013 [-] R² = 0.9684

-2 ln(E) = -0.2739t - 0.6656 R² = 0.9921 -2,5 ln(E) = -0.1787t - 0.6692 R² = 0.9854 -3 t [h] 0.32 0.56 0.8 1.6 3.2 6.4

Figure D-26 Time evolution of the natural logarithm of the enantiomeric excess ln(E) for the experiment with 0.32; 0.56; 0.8; 1.6; 3.2 and 6.4% of water in toluene. 267 Appendix D

Follow-up of run 8

Figure D-27 Chromatograms of Run 8 follow-up: from top to bottom; at day 1, before addition of DBU; at day 1, 2h after addition of DBU; at day 2, at the start; at day 2, at the end and at day 3, at the start. 268 Appendix D

Figure D-28 Chromatograms of Run 8 follow-up: from top to bottom; at day 3, at the end; at day 4, at the start; at day 4, at the end; at day 5, at the start and at day 5, at the end. 269 Appendix D

HPLC for overall thermodynamics of racemization

Figure D-29 Chromatograms of the ratio between R and S BnFTP at 90°C, with DBU in presence of 1equivalent of R-PBA after 2 days (top) and after 5 days (bottom).

Top View