Accelerated Design of High-Efficiency Lead-Free Tin Perovskite Solar Cells via Machine Learning

Tin (Sn) perovskite solar cells (PSCs) are the most promising alternatives to lead (Pb) PSCs, which pose a theoretical limitation on efficiency and an environmental threat. However, Sn PSCs are still in the early stage of development in comparison with the conventional Pb PSCs, and still require a considerable amount of time and effort to obtain an optimum structure via manual trial-and-error methods. Herein, we propose a machine learning (ML) approach to accelerate the design of the optimized structure of Sn PSCs with high efficiency. The proposed method uses K-fold cross-validation-based deep neural networks, thus maximizing the prediction and recommendation accuracy with a limited amount of experimental data recorded for the Sn PSCs. Our approach establishes a new appropriate Sn-PSC design based on an ML recommendation algorithm. The validation experiment reveals a three times higher efficiency of the ML-designed Sn PSCs (5.57%) than that of those designed through unguided fabrication trials (avg. 1.72%).

Automated summary

This study presents a machine learning approach to accelerate the design of highly efficient Sn perovskite solar cells (PSCs). By using deep neural networks and cross-validation, the method can accurately predict and recommend optimal structures with limited experimental data. The validation experiment demonstrates that the machine learning-designed Sn PSCs exhibit significantly higher efficiency.