Quantitative structure-property relationship (QSPR) framework assists in rapid mining of highly Thermostable polyimides

Thermal stability is an invaluable aspect in assessing polymer properties, especially for polyimides (PIs), which are known for their excellent heat resistance. However, empirically oriented discoveries have retarded its development. Inspired by the emerging data-driven polymer informatics, we developed a framework for quan-titative structure-property relationships (QSPR) related to thermal stability (i.e., thermal decomposition tem-perature (Td)) of PIs. Given that the Td of the same polymer under different measurement atmospheres and weight loss rates cannot be generalized, we carefully sorted out the data and established four Td-related models, namely Td5(N2), Td10(N2), Td5(Air), and Td10(Air). All models passed a rigorous validation procedure (external validation, internal validation, and Y-random validation) and presented excellent predictability and stability. The reliability of the predicted values was ensured by the validation of the leverage method. For the same polymer, the calculated Td rises with increasing weight loss rate, showing a trend consistent with reality based on different Td-related models. Given that a weight loss of 10% in a nitrogen environment is commonly adopted as an evaluation criterion for Td, we selected the Td10(N2) model for high-throughput screening of nearly 3000 PIs. Thermostable candidates of interest in different fields were presented, aided by the Tg model, to inspire future new PIs design and to accelerate the polymer informatics process.