Bayesian safe-operating map construction from multimodal reverse-bias diagnostics in perovskite/silicon tandem solar cells

Journal article, 2026

Reverse-bias stress induced by partial shading limits field reliability in perovskite and perovskite/silicon tandem photovoltaics because localized current crowding couples luminescence collapse, hotspot formation, and irreversible shunting. Existing qualification methods do not integrate early optical, thermal, electrical, and survival evidence into architecture-resolved safe-operating maps. The objective of this study is to develop a multimodal, leakage-controlled framework for predicting reverse-bias failure and optimizing resilient device architectures. Single-junction and monolithic tandem cells were fabricated with varied transport layers, electrodes, and reinforced barriers, then tested under reverse-voltage, reverse-current, partial-shading, spectral-mismatch, and endurance protocols. Synchronized electroluminescence, photoluminescence, and infrared thermography were converted into a Composite Multimodal Imaging Index and incorporated into Cox models, calibration analysis, and Bayesian optimization. The reinforced-barrier architecture raised |Vbd| from 3.2 to 20.8 V. CMII discrimination increased from an AUC of 0.78 at 30 s to 0.91 at 300 s, exceeding EL-only prediction at 0.74. Predicted single-junction failure probability increased from 0.05 at 5 mA cm−2 and 25 °C to 0.78 at 30 mA cm−2 and 60 °C. The framework supports reproducible architecture ranking, early-failure screening, and tandem operating-window definition. Future innovation extends the method to encapsulate module strings under outdoor spectral variability.