Machine learning-guided search for high-efficiency perovskite solar cells with doped electron transport layers

Abstract

The experimental search for high-efficiency perovskite solar cells (PSCs) is an extremely challenging task due to the vast search space comprising the materials, device structures, and preparation methods. Herein, using a two-step machine learning approach and 2006 PSC experimental data points extracted from 880 articles published between 2013 to 2020, we develop some heuristics for high-efficiency PSC and power conversion efficiency (PCE) improvement induced by doping of the electron transport layer (ETL). We show that the utilizations of SnO₂ and TiO₂ ETLs, mixed-cation perovskites, dimethyl sulfoxide and dimethylformamide perovskite precursor solvents and anti-solvent treatment are the most significant factors that lead to the high PCEs of PSCs. The PCE can be further improved by ETL doping for tuning the conduction band minimum, Fermi level, and conductivity of the ETL. Moreover, we predict that a FA–MA based PSC with a Cs-doped TiO₂ ETL and a Cs–FA–MA based PSC with S-doped SnO₂ ETL exhibit PCEs of as high as 30.47% and 28.54%, respectively. This study provides insightful guidance for the development of high-efficiency PSCs.