Design of polymers for energy storage capacitors using machine learning and evolutionary algorithms

To meet the demands of emerging electrification technologies, polymers that are capable of withstanding high electric fields at high temperatures are needed. Given the staggeringly large search space of polymers, traditional, intuition- and experience-based Edisonian approaches are too slow at discovering new polymers that can meet these demands. In this work, a genetic algorithm was combined with five machine learning-based property predictors to design over 50,000 hypothetical polymers that achieve target properties. Additionally, a polymer synthesiz,ability-based criterion was used to narrow these polymers down to 23 candidates likely to be synthesizable and 3665 that may be synthesizable. A version of the genetic algorithm code is also made available for public use on GitHub. [GRAPHICS] .