Exploring zeolite-water interactions using a combined experimental-theoretical NMR approach

The aim of the work in this thesis is to use a combined experimental and theoretical approach to investigate the behaviour of water in zeolites. Recently, zeolite-water interactions have gained interest in the literature, as aluminosilicate zeolites show facile lability in water. This phenomenon affects how we view zeolites, however this mechanism is not fully understood. Experimental ¹⁷O NMR has allowed for the observation of zeolite-water interactions and ab initio molecular dynamics studies have highlighted the surprising and complex behaviour of water in zeolites. A combined experimental-theoretical approach enables a better understanding of zeolite-water interactions, but new theoretical methods must first be developed to model zeolite-water structures realistically and these methods must be applied to simpler zeolite models such as siliceous zeolites. This is what was performed over the course of this PhD project. This thesis has shown that experimental ¹⁷O NMR indicates clear ¹⁷O-enrichment of selected siliceous zeolites (AFI, FER and MFI) regardless of ¹⁷O-enrichment method, implying that this lability is a general feature of high silica and siliceous zeolites. The extent of ¹⁷O-enrichment of the framework oxygens varied due to the presence of silanols and differences in the zeolite topologies. A theoretical NMR approach was developed, using two machine learning techniques known as a “double machine learning approach” which showed good correlations with published experimental ¹⁷O and ²⁹Si NMR parameters, and the ability to enhance the prediction of tensorial NMR parameters, under near-operando conditions. Using this approach, defective and pristine siliceous zeolites were selected, then the effects of external environmental factors, such as water loading and temperature on the ¹⁷O/²⁹Si NMR signals were investigated and enumerated. This approach was proven valid, allowing the modelling of dynamical zeolite-water behaviour which would be difficult to observe using DFT-MD. This idea was extended to aluminium-containing zeolites, in order to investigate the Al distributions via double machine learning, giving rise to a new approach for estimating Al siting, which is of paramount importance in controlling the reactive properties of zeolites. In future, combining this approach with experimental NMR could help determine the possible zeolite-water interactions in more complex systems such as the role of zeolites as catalysts in biomass conversion which uses water as a solvent.