Numerical study of chalcopyrite n-CuAlSe 2 /p-CuGaSe 2 solar cell with an additional CuTiTe 2 back absorber layer using a SCAPS-1D simulator and machine learning approach

Abstract The chalcopyrite compound CuTiTe 2 can be a potential absorber layer along with the CuGaSe 2 absorber layer of a thin-film photovoltaic solar cell due to its low-cost fabrication process, non-toxic nature, optimum bandgap energy, electron affinity and optical absorption coefficient. Using numerical analysis, we propose a novel solar cell structure of n-CuAlSe 2 /p-CuGaSe 2 /p-CuTiTe 2 with optimum device parameters of layer thicknesses 20 µ m/1 µ m/1 µ m along with doping concentrations of 10 15 c m − 3 / 10 18 c m − 3 / 10 18 c m − 3 , respectively, with a defect density of 10 12 c m − 3 in each layer. A remarkable power conversion efficiency (PCE) of 30.6% with an open circuit voltage of 1.54 V, short circuit current density of 21.85 mA cm −2 and fill factor (FF) of 91% are achieved with the proposed solar cell structure. An improvement of more than 7% higher PCE, 0.40 V greater open circuit voltage and an almost 2.15% increase in FF value with much better quantum efficiency and temperature stability are achieved when incorporating p-CuTiTe 2 as the back absorber layer to an n-CuAlSe 2 /p-CuGaSe 2 solar cell. A machine learning tool, namely the random forest regression (RFR) model, is used to predict photovoltaic performance and to estimate the relative impact of key device parameters on overall device performance. The RFR model shows an impressive predictive accuracy ( R 2 > 0.96) along with low mean absolute error and root mean square error values across all output parameters. Correlation matrix analysis supports these findings and provides a data-driven pathway for optimization. However, experimental validation is the next crucial step for further development.