Machine-learning-assisted searching for thermally conductive polymers: A mini review

Abstract

Polymers, known for their lightweight, high strength, and ease of processing, serve as a key component in engineering materials. Polymers with high thermal conductivity (TC) present enormous potential applications in thermal management for high-performance electronic devices. However, the discovery of thermally conductive polymers is still in a time-consuming and labor-intensive trial-and-error process, which undoubtedly hinders the progress in related applications. Fortunately, machine learning (ML) enables to overcome this obstacle by building precise models to predict the TC of polymers through learning from a large volume of data and it can quickly identify polymers with high TC and provide significant insights to guide further design and innovation. In this mini review, we briefly describe the general process of using ML to predict polymers with high TC and then give guidance for the selection and utilization of three important components: database, descriptor, and algorithm. Furthermore, we summarize the predicted thermally conductive single polymer chains, amorphous polymers, and metal-organic frameworks via ML and identify the key factors that lead to high TC. Finally, we touch on the challenges faced when utilizing ML to predict the TC of polymer and provide a foresight into future research endeavors.