Probing the Low-Frequency Response of Impedance Spectroscopy of Halide Perovskite Single Crystals Using Machine Learning.

Electrochemical impedance spectroscopy (EIS) has emerged as a versatile technique for characterization and analysis of metal halide perovskite solar cells (PSCs). The crucial information about ion migration and carrier accumulation in PSCs can be extracted from the low-frequency regime of the EIS spectrum. However, lengthy measurement time at low frequencies along with material degradation due to prolonged exposure to light and bias motivates the use of machine learning (ML) in predicting the low-frequency response. Here, we have developed an ML model to predict the low-frequency response of the halide perovskite single crystals. We first synthesized high-quality MAPbBr single crystals and subsequently recorded the EIS spectra at different applied bias and illumination intensities to prepare the dataset comprising 8741 datapoints. The developed supervised ML model can predict the real and imaginary parts of the low-frequency EIS response with an score of 0.981 and a root mean squared error (RMSE) of 0.0196 for the testing set. From the ground truth experimental data, it can be observed that negative capacitance prevails at a higher applied bias. Our developed model can closely predict the real and imaginary parts at a low frequency (50 Hz-300 mHz). Thus, our method makes recording of EIS more accessible and opens a new way in using the ML techniques for EIS.