La résonance magnétique nucléaire est une technique de choix pour étudier la dynamique de l'échange dans des composés à portée thérapeutique. Cependant, les spectres acquis avec les méthodes 1D et 2D classiques présentent souvent des recouvrements importants, ce qui rend l'attribution de chaque espèce en échange difficile. Le développement méthodologique d'une approche pure shift permettrait donc d'améliorer la résolution de ces données. Au cours de cette thèse, nous avons effectué une série de développements méthodologiques des expériences de type EXSY dans lesquelles nous avons mis en œuvre différentes méthodes de découplage homonucléaire. En particulier, deux approches ont été testées: la méthode PSYCHE appliqué aux dimensions F1 et F2 des cartes EXSY, et la méthode Zangger-Sterk en F2. Ces approches ont toutes mené à une amélioration significative de la résolution qui nous a permis de caractériser les deux conformères s-cis/s-trans du Trandolapril. Nous avons également étudié l’intérêt de la technique d'échantillonnage non-uniforme (NUS) du signal pour réduire le temps d'analyse. Cette approche nous a permis de gagner en temps d'analyse, mais la présence d'artefacts à certaines températures suggère que d'autres développements seront encore nécessaires. Ces résultats ouvrent la voie vers une analyse plus fine du processus d’échange dans des composés présentant un spectre RMN complexe. / Nuclear magnetic resonance is a technique of choice for studying chemical exchange in therapeutic compounds. However, spectra acquired with standard 1D and 2D methods often show spectra with overlapping signals, which makes the assignment of each species difficult. Methodological development of a pure shift approach would therefore make it possible to improve the resolution of these data. During this thesis, we carried out a series of developments of the EXSY type experiments in which we implemented different homonuclear decoupling methods. Two approaches were tested: the PSYCHE method applied to the F1 and F2 dimensions of EXSY maps, and the Zangger-Sterk method in F2. These approaches all led to a significant resolution improvements allowing us to characterize the s-cis/ s-trans conformers in Trandolapril. We have also studied the interest of the non-uniform sampling (NUS) technique to reduce the analysis time. This approach allowed us to accelerate the experiment, but the presence of artifacts at various temperatures suggests that further developments will still be needed. These results pave the way for more analysis of the exchange process in compounds with complex NMR spectrum.

A unique characteristic of NMR is that, unlike other spectroscopic techniques, it separates the excitation of signals from their detection. By manipulating the type of signal excitation used, the chemical information content of a spectrum can be controlled. This versatility has made NMR a powerful and flexible weapon in the analytical arsenal of chemists, not only for the determination of structural, chemical, dynamic, and physical properties of molecules, but also for the analysis of mixtures, since NMR has the ability to study these intact without the need for physical separation. Chapter 1 contains an introduction to the theoretical NMR background necessary for this thesis. Chapter 2, 3 and 6 detail the development of new methods that suppress 13C satellites not only in conventional 1D 1H and 19F spectra, but also in 1H DOSY spectra, and can facilitate the analysis of minor components in high dynamic range mixtures (i.e. those with a wide range of concentrations). Chapter 4 introduces a new experiment which suppresses low-level artefacts in pure shift NMR, and gives clean pure shift spectra that can be used for the detection of minor components in the presence of strong signals. Chapter 5 and 7 illustrate how 19F NMR can be exploited for the acquisition of simplified proton spectra associated with a given 19F chemical shift, or for the virtual separation of mixture components using broadband 19F DOSY. Chapter 8 summarises the conclusions extracted from the research introduced in the main body of this thesis, and gives suggestions for future developments. Chapters 2, 3, 4, 5, and 7 contain published research articles and their Supporting Information and are presented without modification. Chapter 6 is presented as a manuscript intended for publication.

Developments in Multivariate DOSY processing and Pure Shift NMR, authored by Adam Colbourne and submitted for the degree of Doctor of Philosophy in the Faculty of Engineering and Physical Sciences at the University of Manchester, 26th February 2014. The theme of this thesis is resolution; the separation of overlapping, entangled information in NMR spectroscopy data. The ability to resolve the features of a dataset is important because it greatly simplifies, or even makes possible, the interpretation of those features to yield information. Here, methods developed to increase resolving power in two different areas of NMR spectroscopy are described; these areas are so-called 'pure shift' or δ-resolved NMR and diffusion-ordered spectroscopy (DOSY). Pure shift NMR aims to reduce the overlap of the signals present in an NMR spectrum by collapsing the multiplet structure caused by spin-spin coupling. There are a variety of methods for achieving this, each of which has its pros and cons. A homo-nuclear decoupling scheme originated by K. Zangger and H. Sterk is implemented in its most recent form to decouple the F1 and F2 dimensions of the 2D NOESY experiment individually. The application of covariance processing to allow the removal of all the multiplet structure from data produced by these singly decoupled experiments is demonstrated and the results discussed. Full experimental homo-nuclear decoupling of 2D NMR is discussed and demonstrated with the TOCSY experiment using a combination of Bax's constant time decoupling scheme in F1 and Zangger-Sterk decoupling in F2. DOSY is strictly a catch-all term for the data processing applied to pulsed field gradient NMR data to extract information on the diffusion of chemical species, but is widely accepted as referring to the combination of the two. Applied to mixtures, DOSY is a powerful tool that can allow the separation of the spectra of the mixture components; this greatly simplifies the process of interpreting mixture NMR data. However, DOSY processing struggles where signals from different, but similarly diffusing chemical species overlap; one is faced with the problem of separating similar, overlapping exponentials in noisy data. Standard DOSY processing schemes can be described as univariate or multivariate with respect to the way in which they handle DOSY data; the former analyses the data a single frequency at a time, the latter tries to untangle the whole dataset at once. Multivariate processing schemes are better suited to resolving overlap in DOSY data, because they use all of the information available, the counter point being that too much information causes them to break down. SCORE is one such algorithm. Research into constraining and augmenting SCORE is presented, leading into a discussion of the potential application of prior knowledge of the DOSY dataset. While exploring the application of prior knowledge, it was realised that the differences between the spectra extracted by SCORE could be used to separate mixture components in a general manner. The presented OUTSCORE algorithm uses information from both the spectra and diffusion dimensions of DOSY data to separate components almost an order of magnitude more similarly diffusing than was previously possible. Finally, a hybrid processing scheme termed LOCODOSY is reported, that breaks a dataset down into smaller sections for individual multivariate analysis before recombination of the results; circumventing the problem of having too much or too little data in any one analysis. The LOCODOSY processing scheme is demonstrated on both the SCORE and OUTSCORE algorithms.