Skip to content
Open access

Machine Learning-Assisted Optimization of Lead-Free Cs₂AgBiBr₆ Perovskite Solar Cells via ETL/HTL Selection: A Combined SCAPS-1D and Random Forest Approach

2026 · IEEE Access · Vol 14, pp. 161871-161888 · 0 citations · 54 references

Abstract

The performance of lead-free Cs2AgBiBr6 double perovskite solar cells (PSCs) was investigated using SCAPS-1D simulation, followed by machine learning (ML) assisted modelling of the optimised device. Four device architectures employing ZnO and WS2 as electron transport layers (ETLs) and Cu2O and CNTs as hole transport layers (HTLs) were systematically evaluated to identify the optimum configuration. The influence of absorber thickness, bulk defect density, HTL/absorber interface defect density, ETL/absorber interface defect density, acceptor density, donor density, electron capture cross-section area, and hole capture cross-section area was comprehensively analysed. The investigated device architectures achieved power conversion efficiencies (PCEs) of 26.64%, 26.66%, and 27.89%, while the optimised FTO/ZnO/Cs2AgBiBr6/CNTs/Au configuration achieved the efficiency of 27.91%, with a $J_{sc}$ of 23.93 mA/cm2, a $V_{oc}$ of 1.30 V, and an FF of 89.35%. Based on its superior photovoltaic (PV) performance and favourable behaviour under variations in interface defect density and hole capture cross-section, the FTO/ZnO/Cs2AgBiBr6/CNTs/Au configuration was selected for further ML-based analysis. An ML based prediction framework was developed using 20,736 feature entries generated from 2,592 SCAPS-1D simulated device configurations by systematically varying eight device parameters. Four regression algorithms were evaluated, and the Random Forest Regressor demonstrated the highest predictive capability, achieving $R^{2}$ values greater than 0.99 with minimal prediction errors for all PV parameters. Furthermore, the correlation matrix and SHAP analysis identified the Cs2AgBiBr6 bulk defect density as a major influential parameter governing the predicted device performance within the investigated simulation space. These findings demonstrate that integrating SCAPS-1D simulation with ML-assisted modelling provides an efficient framework for the systematic design, optimisation, and performance prediction of lead-free Cs2AgBiBr6 PSCs.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.