ANN-integrated modeling of HTL-free Cs₂SnI₆ perovskite solar cells under indoor and outdoor light spectra

To address the cost and technical challenges in the commercial scalability of perovskite solar cells (PSCs), we propose an HTL-free PSC device with the structure FTO/TiO2/Cs2SnI6/Ni. In this study, we employ affinity engineering methodology to select the most appropriate ETL materials for this device. A range of ETL materials, including SnO2, CdS, GO, TiO2, MZO, and WO3, are evaluated. Among these, TiO2 shows the best performance and is selected as the optimal ETL material. The HTL-free device is investigated under two light spectra: solar light (outdoor) and LED light (indoor). Key parameters, such as absorber thickness, absorber defect density, electron affinity, and work function of back contact metal, are optimized through simulations. The final optimized device structure—FTO/TiO2/Cs2SnI6/Ni—achieves an impressive efficiency of 27.01 % under solar light and 38.75 % under LED light. Additionally, an Artificial Neural Network (ANN) model is used to predict the power conversion efficiency (PCE). The input data are obtained from SCAPS simulations, and the ANN model is designed with a 4–12–4 architecture. The predictive model is able to estimate the efficiency with a mean squared error (MSE) of 8.127 × 10⁻⁴, a root mean squared error (RMSE) of 0.00944, and a R² value of 0.968 under the solar light spectrum. For the LED light spectrum, the MSE is 3.6585 × 10⁻⁴, the RMSE is 0.00614, and the R² is 0.986. The sensitivity analysis reveals that the absorber thickness is the most significant parameter influencing the device performance under present conditions. The hybrid model demonstrates a strong predictive ability, providing an efficient technique for optimizing the device investigated.