Parametric quantification of silicon-based heterojunctions via equivalent circuit and deep learning model

Silicon-based heterojunction solar cells have the highest efficiency among single-junction silicon solar cells. A comprehensive understanding of the current-voltage characteristics of silicon-based heterojunctions is essential for determining the performance of relative devices. In this study, we propose a lumped-parameter equivalent circuit that incorporates a reverse diode and an additional resistance to depict carrier tunneling quality. This circuit accurately describes the S-type character and shows the difference between dark and light current-voltage curves. The key elements in the circuit possess specific physical meanings, allowing the microscopic characteristics of the device to be obtained once the parameters of the circuit are inferred. Furthermore, we develop an analytical model utilizing deep learning techniques to realize automatic parameter extraction from measured current-voltage curves. This methodology enables the parametric quantification of microscopic characteristics for silicon-based heterojunctions with greater convenience and accuracy.