(148h) Machine Learning Enhanced Modeling of Chemically Recyclable Polymeric Materials for Sustainability

Plastic waste management remains a key sustainability challenge. Mechanical recycling often degrades polymer properties due to contamination and chain scission. On the other hand, Chemical recycling offers a promising alternative by depolymerizing plastics into monomers for repolymerization into virgin-quality materials. Recent efforts have focused on designing chemically recyclable polyethylene-like (PE-like) polymers that retain polyethylene’s toughness and melt processability. These materials incorporate cleavable functional groups separated by long-chain aliphatic spacers.

This work leverages polymer simulations to complement experimental efforts and accelerate the discovery of PE-like materials. Properly modeling recycling and degradation processes are essential for assessing the performance of PE-like materials. These processes require capturing adaptable bond topologies, which are unaccounted for in most existing molecular dynamics force fields. To overcome this, highly accurate machine learning force fields (MLFFs) are trained on single-point DFT calculations of PE-like materials. The DFT input configurations were obtained from biased molecular dynamics trajectories involving changing bond topologies. MLFF molecular dynamics simulations are then used to study how certain design choices including functional group selection and chemical environment affect mechanical properties and recyclability.