AI-Driven Discovery of Amorphous Fluorinated Polymer Electret with Improved Charge Stability for Energy Harvesting

Abstract

Electret materials are promising dielectric materials with trapped charges for various applications such as vibration energy harvesters and acoustic transducers. In the present work, ionization potential is discovered as the descriptor to quantify the charging performance for amorphous fluorinated polymer electrets. Using this descriptor, high-throughput computations, and graph neural network models, 1 176 591 functional groups are screened on the cyclic transparent optical polymers (CYTOP), and 3 promising electrets are identified. The electrets are synthesized experimentally as 15 µm-thick films. The films are able to keep their both bipolar surface potentials above ±3.1 kV for over 1500 h and are estimated to have longevity of 146 years under 80 °C, achieving significant improvements on charging stability among CYTOP-based polymer electrets. The excellent bipolar charging performance can greatly enhance power generation capacity of electret-based vibration energy harvesters. This work also demonstrates the use of deep learning as a new paradigm for accelerating practical materials discovery.