Carrier Mobility Prediction and Device Stability Evaluation of Perovskite Optoelectronic Thin Films Using Small-Sample Deep Learning Algorithms

Perovskite optoelectronic thin films have shown great potential in solar cells, photodetectors, light-emitting devices, and other optoelectronic applications owing to their strong light absorption, tunable bandgap, and favorable carrier transport properties. However, carrier mobility is affected by multiple factors, including material composition, processing conditions, film morphology, defect states, and testing methods, while device stability remains limited by ion migration, interfacial degradation, and environmental stress. To address the challenges of limited sample size, heterogeneous data sources, and label noise in perovskite thin-film datasets, this study develops a small-sample deep learning framework for carrier mobility prediction and device stability evaluation. A multidimensional feature system is constructed by integrating material composition, fabrication parameters, film structure, optoelectronic properties, and testing conditions. A lightweight deep learning model incorporating transfer learning, multi-task learning, regularization, and cross-validation is then developed to improve prediction reliability under limited data conditions. Stability classification, ablation experiments, and interpretability analysis are performed to identify the key factors governing carrier mobility and device degradation. The results indicate that the transfer-enhanced deep learning model improves carrier mobility prediction performance, while the multi-task learning model further enhances device stability evaluation. Feature importance analysis reveals that trap-state density, grain size, carrier lifetime, annealing temperature, and film thickness are the dominant factors affecting mobility and stability, whereas encapsulation state, humidity, and transport layer materials significantly influence long-term device stability. This study provides a data-driven approach for the rapid screening and optimization of perovskite optoelectronic thin films with high mobility and enhanced stability.