Predicting Thermal Stability of Polyimides via Reactive Molecular Dynamics

Polyimides are an important class of polymer materials used in aerospace applications, particularly for their high thermal resistance properties. A large number of possible formulations of polyimide chemistry are possible; as is typically the case, they are often explored only by laborious synthesis and experimentation. In this work, we utilize reactive molecular dynamics with the ReaxFF force field to determine in-silico degradation temperatures (T_D) for a large set of polyimides for which an experimental T_D is known. Subsequently, we train a machine learning model to map from the computed T_D to the experimental T_D. Our model can thereafter serve as a predictor of unseen polyimides' experimental T_D and furthermore as a component of an inverse design process, bypassing the need for synthesis and experimentation on potentially hundreds of new chemistries. This approach has the potential to significantly accelerate the discovery of new high-temperature polyimides.