Bandgap tuning strategy by cations and halide ions of lead halide perovskites learned from machine learning

Yaoyao Li; Yao Lu; Xiaomin Huo; Dong Wei; Juan Meng; Jie Dong; Bo Qiao; Suling Zhao; Zheng Xu; Dandan Song

doi:10.1039/d1ra03117a

Bandgap tuning strategy by cations and halide ions of lead halide perovskites learned from machine learning

RSC Adv. 2021 Apr 27;11(26):15688-15694. doi: 10.1039/d1ra03117a. eCollection 2021 Apr 26.

Authors

Yaoyao Li^{1

2}, Yao Lu^{1

2}, Xiaomin Huo^{1

2}, Dong Wei³, Juan Meng^{1

2}, Jie Dong^{1

2}, Bo Qiao^{1

2}, Suling Zhao^{1

2}, Zheng Xu^{1

2}, Dandan Song^{1

2}

Affiliations

¹ Key Laboratory of Luminescence and Optical Information, Beijing Jiaotong University, Ministry of Education Beijing 100044 China zhengxu@bjtu.edu.cn ddsong@bjtu.edu.cn.
² Institute of Optoelectronics Technology, Beijing Jiaotong University Beijing 100044 China.
³ College of Physics and Energy, Fujian Normal University Fuzhou 350117 China.

Abstract

Bandgap engineering of lead halide perovskite materials is critical to achieve highly efficient and stable perovskite solar cells and color tunable stable perovskite light-emitting diodes. Herein, we propose the use of machine learning as a tool to predict the bandgap of the perovskite materials from their compositions. By learning from the experimental results, machine learning algorithms present reliable performance in predicting the bandgap of the lead halide perovskites. The linear regression model can be used to manually predict the bandgap of the perovskite with the formula of Cs _a FA _b MA_(1-a-b)Pb(Cl _x Br _y I_(1-x-y))₃ (FA = formamidinium, MA = methylammonium). The neural network (NN) algorithm, which takes the interplay of cations and halide ions into account in predicting the bandgap, presents higher accuracy (with a RMSE of 0.05 eV and a Pearson coefficient larger than 0.99). Furthermore, the compositions of the mixed halide perovskites with desirable bandgaps and high iodide ratio for suppressing halide segregation are predicted by NN algorithm. These results highlight the power of machine learning in predicting the bandgap of the perovskites from their compositions and provide bandgap tuning directions for experiments.