材料科学
计算机科学
带隙
密度泛函理论
遗传算法
吞吐量
化学
计算化学
机器学习
光电子学
电信
无线
作者
Byung‐Do Lee,Woon Bae Park,Jin Woong Lee,Minseuk Kim,Myoungho Pyo,Kee‐Sun Sohn
标识
DOI:10.1021/acs.chemmater.0c04499
摘要
Hybrid organic/inorganic halide perovskites are considered to be a key material for high-end applications such as photovoltaic and light-emitting devices. Despite the phase stability and toxicity issues, the future potential of these materials is promising. A computational approach to discover novel perovskites based on density functional theory (DFT) calculations is booming since it is more favorable in terms of cost savings compared with the experimental synthesis approach. High-throughput DFT calculations associated with machine learning (ML) algorithms have recently attracted a great deal of attention in materials research. Rather than typical ML modeling and high-throughput DFT calculations, we suggest a direct discovery of novel perovskites using metaheuristic optimization algorithms in association with conventional lab-scale DFT calculations. Both an elitism-reinforced non-dominated sorting genetic algorithm (NSGA-II) and a reference point-involved NSGA-II (NSGA-III) were employed to nominate 25 novel perovskites (or their variants) that would be free from any toxic elements including Pb. The formation energy, band gap, and effective mass for these novel materials are all adequate for photovoltaic and light-emitting applications. While the ML-based prediction has an inverse prediction complication and even requires DFT calculation-based revalidation, the suggested strategy provides a process for direct discovery with no increase in computational cost.
科研通智能强力驱动
Strongly Powered by AbleSci AI