Perovskite single crystal SCLC measurement prediction using a machine learning model

Abstract

Organic–inorganic halide perovskite (OIHP) materials are recognized for their comprehensive solution processability and pronounced carrier mobility. However, the commercialization of perovskite materials is hindered because of the complexity of achieving surface stability. In this study, we developed a machine learning (ML) model to extract the temperature-dependent space-charge-limited current (SCLC) measurements of the methylammonium lead tribromide (MA) single crystals (SCs) by varying the additives. The model is tested for the unseen data of pristine MA and additive-based MASC's conductivity and resistivity at 20 °C, 40 °C, 60 °C and 80 °C temperatures. The additives used are choline bromide (CB), phenylethylamine hydrochloride (PEA), and p-xylylenediamine (PXA). Different models have been trained on the current–voltage (I–V) curves by varying the temperature, pinpointing the additives' role in enhanced device performance using 10 837 data points. Specifically, in random forest (RF) regression, we obtained a higher R² score of 0.929 with a low root mean squared error (RMSE) of 0.00766 for a new data set, which suggests the model's generalizability. Our comprehensive methodology elucidates the intricate relationship between the functionalities of additives in optimizing device performance, facilitating the informed selection of additives for the development of stable photodetectors. This approach mitigates the necessity for extensive trial-and-error experimentation. Surface passivation using additives is a promising technique for enhancing, cultivating, and refining next-level perovskite photodetectors (PDs).