Design of Elastomers with Tailored Thermal Properties Using Molecular Modeling and Machine Learning

Speaker:
Manav Bhati, Senior Scientist II, Schrödinger

Abstract:
Elastomers are integral to industrial applications because of their useful material properties, such as elasticity, durability, and versatility. Innovating new elastomers can lead to the development of superior products that are more durable and stable. A key thermal property of elastomers is the glass transition temperature (Tg), which indicates the temperature at which an elastomer transitions from a glassy/hard state to a soft/rubbery state. Tg is a critical parameter because it determines an elastomer’s operability and performance at various temperatures. Traditional experimental techniques for determining Tg are time-consuming and expensive, necessitating computational approaches to accelerate the design of new elastomers and minimize the failure rate of resource-intensive experimentation. This study focuses on using machine learning (ML) and classical molecular dynamics (MD) simulations to predict the Tg of elastomers. Using curated datasets of experimental Tg values for homopolymers and copolymers from literature, we developed predictive ML models that can accurately predict Tg for new elastomers that are outside of the dataset used to train the model. We then use these ML models to efficiently screen the design space of elastomers by enumerating a large library of copolymer systems. The Tg of the top-performing elastomers were then validated using MD simulations, which have been shown previously to accurately capture the experimental Tg trends of polymer systems. This computational modeling approach not only accelerates the development of new elastomers, but it also provides insights into the relationship between chemical structure and composition to thermal properties.