Connecting polymer network mechanical properties with molecular-level behavior through data science

Polymeric materials are ubiquitous in nearly every aspect of modern society and are continuously transforming the way human beings interact with physical world. Despite many publications by the scientific community, the generated polymer data hasn’t been efficiently aggregated to provide a comprehensive picture of polymer network properties. Herein, utilizing a recently developed polymer graph data model CRIPT (Community Resource for Innovation in Polymer Technology), the mechanical properties and molecular characteristics of polymeric materials in hundreds of literature reports were aggregated and used to generate different Ashby plots. Specifically, a GPT-4 zero-shot prompt was utilized in conjunction with traditional scientific publication database search prompts for identifying useful papers related to the mechanical properties of associative polymer networks. The mechanical properties data were then manually extracted and classified based on different measurement techniques, including tensile testing and oscillatory rheology. In terms of molecular characteristics, the data related to both strands and junctions were recorded, such as backbone glass transition temperature, bond energy and dissociation rate. The generated Ashby plots are meaningful in terms of two aspects, (1) advancing fundamental understanding of the relationship of network macroscopic properties and molecular level behavior, (2) future development of polymeric materials with desired properties combination.