Discovery of All-Inorganic Lead-Free Perovskites with High Photovoltaic Performance via Ensemble Machine Learning

Xia Cai,Yan Li,Jianfei Liu,Hao Zhang,Jianguo Pan,Yiqiang Zhan
DOI: https://doi.org/10.1039/d3mh00967j
IF: 13.3
2023-09-17
Materials Horizons
Abstract:Growing evidence shows that all-inorganic lead-free perovskites hold promise for solving stability and toxicity problems for perovskite solar cells. However, the power conversion efficiency of all-inorganic perovskite cannot rival that of hybrid organic-inorganic perovskite. To face the challenges of efficiency, stability and toxicity simultaneously for the application in perovskite solar cell, this study conducts high-throughput materials search via ensemble machine learning for nearly 12 million AA'BB'X 3 X' 3 all-inorganic perovskites to obtain candidates with non-toxic and excellent photovoltaic performance. Based on experimental data, the models of structure discrimination and band gap classification are established for AA'BB'X 3 X' 3 , and physics-inspired multi-component neural network is proposed under the model exploration of logical structure. It is found that extracting key features to input into the model and treating non-key features as supplements are easier for model learning and more effective to reduce model parameters. Then, based on established ensemble models as well as the new criteria of ion radius difference and the optimization rules of toxicity and cost, over 80000 candidates are screened. Among 34 lead-free AA'BB'X 3 X' 3 identified with suitable band gap and negative formation energy by first principle calculation, 17 candidates have the theoretical power conversion efficiency over 20%. Debye temperature of 10 lead-free AA'BB'X 3 X' 3 , basically Bi-based compounds, are larger than 350 K, which is advantageous for suppressing nonradiative recombination and thermally induced degradation.
materials science, multidisciplinary,chemistry
What problem does this paper attempt to address?