Breaking the thermal–dielectric trade-off in high-temperature polymers via transfer learning

High-temperature capacitive energy storage demands dielectric polymers that integrate high thermal conductivity with excellent electrical insulation to mitigate thermal runaway induced by Joule heating. However, conventional strategies for improving thermal conductivity through increased aromatic conjugation frequently exacerbate conductive losses under elevated temperatures and high electric fields. To resolve this fundamental trade-off between thermal conductivity and electrical insulation, we introduce a conjugation-decoupling strategy. This approach incorporates aliphatic segments to disrupt the π–π conjugation networks, implemented through a machine learning-assisted co-design workflow. A transfer learning model is built to establish the structure–property relationship between glass transition temperature and thermal conductivity, and subsequently guides the synthesis of three semi-aromatic polyimides that concurrently achieve a high glass transition temperature, a wide bandgap, and high thermal conductivity. The resulting semi-alicyclic polyimide film demonstrated outstanding discharge energy density (5.26 J cm−3) and η = 90% performance at 200 °C, significantly outperforming commercial Kapton polyimide film. We report a strategy for high-temperature dielectric development using an interpretable machine learning model, demonstrating a concurrent enhancement of electrical insulation and thermal conductivity, properties typically constrained by a conventional trade-off.