Machine learning-accelerated screening of all-inorganic perovskites for tandem solar cells

Machine learning-accelerated screening of all-inorganic perovskites for tandem solar cells

Materials Today Energy

10.1016/j.mtener.2025.102179

https://doi.org/10.1016/j.mtener.2025.102179

Chong Guo, Kun Guo, Ziliang Deng, Huacheng Zhang, Kang Li, Zhiyong Liu, Bing Xie, Pu Mao, Chee Kiang Ivan Tan, Kui Yao, Francis Eng Hock Tay

All-inorganic halide perovskites, Machine Learning, DFT calculations, Tandem solar cells

19 December 2025

Guo, C., Guo, K., Deng, Z., Zhang, H., Li, K., Liu, Z., Xie, B., Mao, P., Tan, C. K. I., Yao, K., & Tay, F. E. H. (2026). Machine learning-accelerated screening of all-inorganic perovskites for tandem solar cells. Materials Today Energy, 55, 102179. https://doi.org/10.1016/j.mtener.2025.102179

The vast compositional space of all-inorganic halide perovskites (AIHPs) poses a significant challenge to developing stable, high-performance photovoltaic materials. To address this, we established a machine learning (ML)-accelerated workflow for screening ABX3 AIHPs for tandem solar cells. A dataset of 2,115 compounds from the Open Quantum Materials Database (OQMD) and 535 from the Materials Project (MP) was constructed, encompassing formation energy, energy above the convex hull (Ehull), bandgap, and bandgap nature. Crystal-structural descriptors affecting optoelectronic performance and stability were incorporated. Among the six algorithms evaluated, CatBoost exhibited superior performance in predicting formation energy (MAE = 0.058 eV/atom), Ehull (MAE = 0.034 eV/atom), and bandgap nature classification (88% accuracy), whereas LightGBM demonstrated the highest accuracy for bandgap prediction (MAE = 0.281 eV). SHapley Additive exPlanations (SHAP) analysis revealed that B–X orbital interactions and crystal symmetry dominate bandgap behavior, whereas A-site ionic size and bond strength govern stability. High-throughput screening of 32,256 candidates identified 23 promising absorbers with optimal bandgaps and thermodynamic stability. The potential of representative candidates was further validated through Density Functional Theory (DFT) calculations. This study provides a theoretical basis and a generalizable design strategy for rapid screening of ABX3 AIHPs in high-efficiency tandem solar cells.

Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)

This research / project is supported by the A*STAR - Industry Alignment Fund – Industry Collaboration Projects Grant
Grant Reference no. : I2301E0027

https://oar.a-star.edu.sg/communities-collections/articles/22795

2468-6069

Institute of Materials Research and Engineering