Data-driven approaches for structure-property relationships in polymer science for prediction and understanding

• Published in: Polymer journal 2022-08, Vol.54 (8), p.957-967
• Main Author: Amamoto, Yoshifumi
• Format: Article
• Language: English
• Subjects: 119/118; 639/301/923/1028; 639/301/930; Addition polymerization; Biomaterials; Bioorganic Chemistry; Chains (polymeric); Chemistry; Chemistry and Materials Science; Chemistry/Food Science; Crystal structure; Crystallinity; Focus Review; Informatics; Machine learning; Measurement techniques; Molecular chains; Polymer Sciences; Polymers; Science; Statistical methods; Surfaces and Interfaces; Thin Films
• ISSN: 0032-3896; 1349-0540
• DOI: 10.1038/s41428-022-00648-6

In this review, recent developments in data-driven approaches for structure-property relationships in polymer science are introduced. Understanding the structure-property relationship in polymeric materials is a significant challenge. This is because long molecular chains generate unique structures and properties over a wide range of spatial and temporal scales, which are often difficult to address using theoretical models or single simulation/measurement techniques. Recently, the data-driven modeling of structure-property relationships based on statistical/informatics methods has been employed in polymer science to obtain the desired properties and understand the mechanisms. This review summarizes the reports from this domain in the previous three years. A concept and some methods in data-driven science are first explained to readers unfamiliar with this area. Additionally, various examples, such as the description of a single chain, phase separations, network polymers, and crystalline polymers, are introduced. A topic for dealing with chemically specified coarse-grained simulations is also included. Finally, future perspectives in this area are presented. In this review, recent developments in data-driven approaches for structure-property relationships in polymer science based on statistical/informatics methods are introduced. A concept and some methods in data-driven science to obtain the desired properties and understand the mechanisms in polymeric materials are first explained. Additionally, various examples, such as the description of a single chain, phase separations, network polymers, crystalline polymers, and machine learning potential are introduced.