Numerical and machine learning approach to optimize the p-Cu2FeSnS4/n-CuAlSe2/n-FTO thin film solar cell for optimum device performance
TL;DR
The promising outcomes from numerical and machine learning approaches will eventually provide feasible directions to fabricate high efficiency CFTS based PV cells in the near future.
Abstract
A systematic numerical assessment using a solar cell capacitance simulator is implemented to optimize the device outcome of a p-type Cu2FeSnS4 (CFTS) layer as an absorber, n-type CuAlSe2 (CASe) as an electron transport layer, and n-type fluorine-doped tin oxide (n-FTO) as a window layer with a molybdenum back metal contact of a solar cell. Later, a random forest regression approach was utilized to validate the simulation results and to identify the relative importance of the device features to the photovoltaic (PV) characteristics. A Pearson correlation matrix has been analyzed to pinpoint which layer’s properties should be adjusted carefully to obtain the optimum outcomes. A remarkable power conversion efficiency (PCE) of 27.36% and a fill factor of 85.22% with an open circuit voltage of 0.934 V and a short circuit current density of 34.38 mA cm−2 are attained. It is found that the p-CFTS/n-CASe/n-FTO cell with dimensions of 6 µm/0.1 µm/0.1 µm layer thicknesses, and doping concentrations of 5 × 1019 cm−3/1016 cm−3/1016 cm−3, with an overall defect density of less than 1014 cm−3 are required for such performance. The promising outcomes from numerical and machine learning approaches will eventually provide feasible directions to fabricate high efficiency CFTS based PV cells in the near future.