Effective Downsampling Techniques for SEM Defect Inspection Using Design Insights in Machine Learning: Topic/category: DI: Defect Inspection and Reduction

During wafer manufacturing, Bright Field Inspection (BFI) followed by SEM review is a widely used method for inline defect capture. Given the large number of defects reported by BFI, especially during R&D stage where the defectivity is often high, users need a way to effectively down sample the defects to be sent for SEM review. In this paper, we demonstrate various BFI to SEM down sampling techniques including, pattern-based clustering, and machine learning (ML) based down sampling that effectively find trends and make prediction to identify systematic weak patterns with increased SEM defect hit rate and systematic defect discovery.