Copolymer Sequence Regulation Enabled by Reactivity Ratio Fingerprints via Machine Learning

Abstract Sequence control is critical for tuning polymer properties in high‐end applications, where reactivity ratios serve as key parameters for analyzing and regulating sequences. Nevertheless, traditional determination methods exhibit low experimental efficiency and are typically confined to two‐component copolymerization. Here, we develop a machine learning platform, which leverages a novel design of “reactivity ratio fingerprints” ( r FPs) to determine reactivity ratios in binary and ternary copolymerizations. Deep learning models trained on millions of r FPs enable highly efficient (millisecond‐level) determination from sparse experimental data (random monomer structures, arbitrary reaction design). This approach demonstrates outstanding versatility to analyze reactivity ratios under diverse conditions (e.g., temperature, solvent). Notably, r FP‐guided reaction design promotes on‐demand sequence tailoring, compatible with a wide range of binary and ternary monomer combinations. Kinetic investigations and glass transition characterizations support the formation of varied sequence structures, facilitating the identification of binary and ternary azeotropic copolymerizations. This work not only unveils an attractive strategy for determining reactivity ratios but also offers a generalizable framework for sequencing complicated chain structures toward property engineering.