Designing Vitrimer Polymers with Molecular Dynamics and Machine Learning

Vitrimers are a new class of self-healing polymers that can heal by rearranging their molecular structure via dynamic covalent bonds. These polymers offer promise for improving the circular life-cycle and sustainability of various polymeric systems that are used in our day-to-day life ranging from fiber composites to electronics. A recently developed framework called Accelerated ReaxFF, uses the "bond boost" approach to speed up the MD simulations by providing reactive sites in the reactants with boost energy equivalent or slightly larger than the energy reaction barrier to overcome the cross-linking process barrier and form desired products. This approach avoids unwanted high-temperature side reactions while allowing for rejection of high-barrier events. This method can be employed not just for virtual characterization of vitrimer polymers but also to understand the rearrangement reactions in epoxy-acid vitrimer polymer chemistries. Further, using coupled molecular dynamics and machine learning, we design new vitrimer chemistries with targeted applications. This includes collecting a dataset of glass transition temperature of various vitrimer chemistries calculated through molecular dynamics (MD). This data is then used to train a latent space and property predictor . Finally, a search is performed within the latent space to uncover vitrimer chemistries with desired properties.