Designing polymer nanocomposites with high energy density using machine learning

Addressing microstructure-property relations of polymer nanocomposites is vital for designing advanced dielectrics for electrostatic energy storage. Here, we develop an integrated phase-field model to simulate the dielectric response, charge transport, and breakdown process of polymer nanocomposites. Subsequently, based on 6615 high-throughput calculation results, a machine learning strategy is schemed to evaluate the capability of energy storage. We find that parallel perovskite nanosheets prefer to block and then drive charges to migrate along with the interfaces in x - y plane, which could significantly improve the breakdown strength of polymer nanocomposites. To verify our predictions, we fabricate a polymer nanocomposite P(VDF-HFP)/Ca 2 Nb 3 O 10 , whose highest discharged energy density almost doubles to 35.9 J cm −3 compared with the pristine polymer, mainly benefit from the improved breakdown strength of 853 MV m −1 . This work opens a horizon to exploit the great potential of 2D perovskite nanosheets for a wide range of applications of flexible dielectrics with the requirement of high voltage endurance.