AI-driven Accelerated Material Discovery for Organocatalysis

This thesis is about chemical polymer recycling for a better circular economy assisted by AI foundation’s model predictions. The research focuses on the selection and the design of organocatalysts for selective depolymerization of polycondensation polymers with glycolysis. 9% of post-consumer plastics are recycled and most of them are mechanical (less than 1% chemically recycled) [7]. Glycolysis still faces some energy efficiency, environmental and economical issues due to the low price of virgin oil. Those issues are notably related to the quality of the product after depolymerization as mixed polymers, food and metal contamination and additives are contained in post-consumer plastics. The solution lies mostly into organocatalyst’s selectivity and activation capabilities. This kind of catalysts avoid usage of metal elements that are a source of monomers contamination. Because of the infinite number of possible organic molecule structures and compounds, and stochastic nature of a chemical reactions, studies of catalytic activation mechanisms are still needed for an efficient selection of the right organocatalyst depending on the conditions of the plastic substrate glycolysis. AI foundation models is a trendy, very efficient and state-of-the-art tool to assist such organocatalysis studies and selections. By collecting experimental reaction performances and correlate them with electronic and steric effects, thermal and chemical stability and geometry, the computational model is fine-tuned to provide better generated chemical structures and predicted performances. A combination of acid and base forming protic ionic salt (PIS) is a very promising catalyst by displaying high thermal stability and selectivity. A mechanistic study on a wide selection of combinations varying in pKa s is proposed in this thesis on dimethyl terephthalate model system for AI training purposes and selection of most effective salt organocatalysts for glycolysis.