Atomic-level design and device modeling of lead-free Cs2AgSbCl6 double perovskite solar cells through DFT, SCAPS-1D, and predictive algorithms

Lead-free double perovskite solar cells (DPSCs) have emerged as attractive alternatives to conventional lead-based perovskites because of their superior environmental compatibility, enhanced chemical stability, and reduced toxicity. In this research study, the performance of a solar cell architecture FTO/IGZO/Cs 2 AgSbCl 6 /Cu 2 O are systematically studied by the simulation software known as SCAPS-1D. DFT calculations were performed using WIEN2k package to investigate structural and optoelectronic properties of Cs 2 AgSbCl 6 material to determine its suitability for photovoltaic applications. Afterwards, the photovoltaic performance is comprehensively studied with respect to the absorber thickness (t active ), bulk defect density (N t ), donor and acceptor concentration (N A and N D ), carrier mobility (u n and u p ), operating temperature (T), illumination intensity (I sun ), series and shunt resistance (R s and R sh ) and contact work functions (WFs). The optimized device shows a higher power conversion efficiency (PCE) of 28.53%, open circuit voltage (V oc ) of ~1.068 V, short circuit current density (J sc ) of 32.81 mA cm −2 and fill factor (FF) of 81.38%, which is found better than the initial device having PCE of 24.77%. The improvement in performance mainly has been attributed to the improvement in energy-band alignment (EBA), efficient extraction of the carriers, suppression of the interfacial recombination, and optimization of the electrical transport through the IGZO/Cs 2 AgSbCl 6 /Cu 2 O heterostructure. In addition, the data set created from the above-mentioned SCAPS-1D simulations was used to train six supervised machine learning algorithms including Linear Regression (LR), Support Vector Regression (SVR), K-Nearest Neighbours (KNN), Random Forest (RF), Artificial Neural Network (ANN), and Extreme Gradient Boosting (xGB), to further accelerate the optimization of the device. The machine-learning analysis accurately estimated the photovoltaic (PV) parameters and defined the major factors that affect the PV performance, such as the carrier mobility, the density of traps, the thickness of the absorber and the illumination intensity.