A review on the application of molecular descriptors and machine learning in polymer design

Polymers are an important class of materials with vast arrays of physical and chemical properties and have been widely used in many applications and industrial products. Although there have been many successful polymer design studies, the pace of materials discovery research can be accelerated to meet the high demand for new, functional materials. With the advanced development of artificial intelligence, the use of machine learning has shown great potential in data-driven design and the discovery of polymers to date. Several polymer datasets have been compiled, allowing robust machine learning models to be trained and provide accurate predictions of various polymer properties. Such models are useful for screening promising candidate polymers with high-performing properties prior to lab synthesis. In this review, we focus on the most critical components of polymer design using molecular descriptors and machine learning algorithms. A summary of existing polymer databases is provided, and the different categories of polymer descriptors are discussed in detail. The application of these descriptors in machine learning studies of polymer design is critically reviewed, leading to a discussion of the challenges, opportunities, and future perspectives for polymer research using these advanced computational tools.

Automated summary

Polymers are important materials with diverse properties, and machine learning has shown great potential in data-driven polymer design. Machine learning models trained on polymer datasets can accurately predict polymer properties and assist in candidate polymer screening before lab synthesis.