Machine Learning–Based Optimization and Engineering of Rb 2 CuSbF 6 ‐Based Perovskite LEDs With Different Electron Transport Layers Using SETFOS

ABSTRACT A machine‐learning approach is developed to optimize the Rb 2 CuSbF 6 ‐based perovskite light‐emitting diodes (PeLEDs). Using density functional theory, the structure, electronic, and optical properties of these materials, Rb 2 CuSbX 6 (X = F, Cl, Br) are explored and the properties needed for device modeling are determined. In light of these results, the electron transport layer (ETL) materials SnO 2 , TiO 2 , ZnO and WO 3 are systematically studied by the SETFOS. The results of the parametric sweep of the thickness of ETL, electron mobility, donor concentration, and refractive index show the best conditions to optimize the luminous efficacy, current efficiency, luminance, and external quantum efficiency (EQE). The nonlinear interactions among these parameters are captured by a Random Forest model trained with 3000 device configurations simulated and the model is able to predict the performance with high accuracy for all performance metrics as seen through R 2 = 0.997 and low RMSE. The feature importance analysis reveals that the two important parameters that affect the performance are the electron mobility in ETL and its thickness and the donor concentration, with little contribution from the refractive index. This physics‐informed framework offers a quantitative design rule for ETL engineering and fast optimization of high‐performance, lead‐free PeLEDs.