Accelerating Surface Passivation Material Design for High-Efficiency Perovskite LEDs via Reinforcement Learning
발표자
이동빈 (광주과학기술원)
연구책임자
김호범 (광주과학기술원)
공동저자
이동빈 (광주과학기술원), 김호범 (광주과학기술원)
초록
내용
Surface passivation is a key strategy to improve the external quantum efficiency (EQE) of perovskite light-emitting diodes (PeLEDs) by mitigating non-radiative recombination and stabilizing perovskite interfaces. However, the vast chemical space and the complex relationship between molecular structures and passivation performance present significant challenges in the rational design of effective surface passivators. Herein, we present a reinforcement learning (RL) framework combined with a pre-trained deep neural network (DNN) that utilizes molecular features learned through SMILES (Simplified Molecular Input Line Entry System) embedding. The DNN model, acting as a reward function in the DRL environment, enables the autonomous exploration and optimization of surface passivation molecules. Our approach establishes a robust and data-driven strategy for accelerating the discovery of high-efficiency passivation molecules, paving the way for next-generation, high-performance PeLEDs.