Charge trap depth prediction of grafted polypropylene system using machine learning

Carbon neutrality is a state of dynamic equilibrium in carbon dioxide emissions. In insulation material aspect, polypropylene (PP) with grafted side chain segments possesses recyclability and may have better electrical property under high temperature compared with traditional norecyclable insulation insulation material like cross-linked polyethylene, which will reduce the carbon emission and help carbon neutrality processing. However, focused on the ideal grafted PP chain, the bandgap (Eg) parameter is no longer suitable for describing the electrical properties of this system for the small grafting ratio. The induced grafted monomer might generate new band energy level and reduce the bandgap but with improvement on the electrical properties, which is not consistent with the positive correlation of the bandgap and the electrical properties. Alternatively, the trap depth, as one of the important electrical performance parameters, can reflect its influence on conductivity and breakdown. In this paper, the trap depth of the grafted PP chain is predicted by machine learning method with recursive feature elimination-based fingerprints. The paper analyzes the influence of different component groups and the ratioes of PP monomers on the charge trap depth. The grafted monomers with strong positive correlations on the electron/hole trap depth are given, and the electrical conductivity of PP-g-VK polymer has been tested for validation from the side. This paper provides a predictable parameter for the analysis of grafted systems and accelerates future predictions of performance parameters such as conductivity and breakdown strength.