(169i) Studying Depolymerization of Polyurethanes Using Reaction-Aware Deep-Learning Potentials

Polyurethanes are one of the most widely employed classes of polymers. Polymerization of polyurethanes (PUs) is a simple process facilitated by the formation of strong urethane bonds between the constituent moieties. This strong urethane linkage contributes to wide-ranging applications across a diverse spectrum, from hard insulating materials to soft foams. However, the high stability of the urethane bond makes it non-degradable in nature, making its disposal an environmental concern. This makes chemical depolymerization an important recovery method to mitigate the waste accumulation and reduce the environmental impact of this polymer. Our work focuses on developing a framework for efficiently training reaction-aware deep-learning potentials (DPs) that can be used to study many possible reactions and reaction pathways for PU depolymerization. Our aim is to implement the “reactive active-learning" formalism that will efficiently train reaction-aware machine learning potentials to study PU depolymerization. This implementation will also help us identify the reaction conditions that will improve the efficiency and selectivity of the process. Our initial work involved training bootstrap DPs using density functional theory data. Next steps include sampling possible bond breaking and bond making events to form products using the DPs and adding them back into the training data set to retrain the DPs.