Polymer Informatics Method for Fast and Accurate Prediction of the Glass Transition Temperature from Chemical Structure

Abstract

We present a new polymer informatics framework that successfully predicts the glass transition temperature T_g of polymers based on their chemical structure. The framework combines ideas from group additive properties (GAP) and quantitative structure–property relationship (QSPR) methods, where GAP (or group contributions) assumes that submonomer motifs contribute additively to T_g, and QSPR links T_g to the physicochemical properties of the structure through a set of molecular descriptors. By integrating these methodologies, our combined QSPR–GAP framework overcomes limitations inherent in using either method independently. We demonstrate its application on a data set of 146 linear homo- and copolymers of the poly(aryl ether ketone) (PAEK) family, achieving a median root mean square error of 8 K for T_g, representing a significant improvement over standalone QSPR or GAP models. Moreover, using a genetic algorithm, we identify two molecular descriptors that predominantly drive T_g predictions. The QSPR–GAP framework can be readily adapted to forecast other physical properties and activity (QSAR) or transferred to other polymer families, including conjugated and biopolymers.