A Zero-Dimensional Hybrid Tin Halide with Efficient Near-Infrared Luminescence for Intelligent Defect Inspection

Near-infrared (NIR) luminescent materials have attracted considerable attention for biomedical imaging and industrial inspection because of their deep penetration capability and strong technical adaptability. Coupling NIR luminescence with deep-learning algorithms further offers a promising route toward intelligent inspection with enhanced automation and accuracy. Here, we report a 0D, Sn halide (MBI) 2 SnCl 6 (MBI = 2-methylbenzimidazole), composed of organic MBI + cations and isolated [SnCl 6 ] 2− octahedra, and Sb substitution yields broadband NIR emission with a highest photoluminescence quantum yield of 59.19%. On this basis, subvisible defects on silicon wafer surfaces can be clearly visualized. When integrated with a U-Net-based convolutional neural network, the NIR imaging output enables rapid and automated defect identification, allowing accurate discrimination between qualified and defective wafers as well as subclassification of defective samples, with an average accuracy of 90.92% and an average recall of 92.92%. This work establishes an effective strategy for integrating lead-free NIR-emissive halides with artificial intelligence for intelligent inspection and highlights the potential of hybrid halide luminescent materials in advanced industrial imaging applications.