Machine learning-assisted design of wide bandgap perovskite materials for high-efficiency indoor photovoltaic applications

Abstract

Recent growing research interests in Indoor Perovskite Solar Cells are being observed due to the combined effect of the progress in highly efficient Perovskite Solar Cells (PSCs) and exponential growth in the Internet of Things (IoTs). Indoor PSCs (IPSCs) present immense potential as they can suitably power the IoTs due to their solution-processability, lightweight nature, and their ability to harvest indoor light energy effectively. Proper choice of absorber material is crucial for developing high-efficiency IPSCs. In this study, a Machine Learning (ML) model has been constructed to predict bandgaps of diverse perovskite material compositions. This bandgap prediction model (BPM) consistently predicts perovskite bandgaps that align closely with their true values. We have also analysed the importance of the key features impacting the bandgap by Correlation matrix and assessed their relative importance using SHAPley analysis. Thereafter, the BPM was employed to predict bandgaps of varying organic-inorganic lead halide perovskite types and compositions. The suitable bandgap materials for IPSCs are then identified based on the predicted values. A representative perovskite material with a suitable bandgap is selected, and the indoor lighting performance of the related device is simulated and analysed thereafter. Notably, the device utilising the selected perovskite material presented excellent performances under indoor illumination with efficiency exceeding 35%.