A virtual defect metrology system utilizing photolithography scanner topography maps to quantitatively detect gross defects, perform in-line yield prediction, and identify defect generating tools

In semiconductor high-volume manufacturing (HVM) fabs, the sheer volume of daily processed wafers necessitates stringent defect metrology strategies. Current methodologies, constrained by throughput limitations, sample only a small percentage of wafers to minimize efficiency losses, thus creating large coverage gaps between process steps. This restricted coverage does not fully capture all gross defect modes generated across various process tools, hence extending the timeline for defects’ root cause analysis. Here, we present an innovative inline virtual metrology (VM) system designed for gross defect detection, leveraging the wafer topography data captured by various photolithography scanner sensors. Our system is comprised of three components. First, the wafer image generation flow, which transforms wafer topography tool sensor data into two-dimensional heat-map images using traditional mathematical modeling techniques. Second, the defect classification and yield impact estimation flow, that uses a deep learning model to detect and classify defects within these images. Defect morphologies are then quantitatively extracted using computer vision (CV) techniques which leads to inline die impact estimation by analyzing the affected pixels and their corresponding wafer coordinates. This system effectively identifies common gross defects and correlates impacted dies to end-of-line historical yield data thus enabling early inline defect detection. Third, the automated root cause analysis flow, which clusters similar defect modes and employs probabilistic analysis and ranking techniques to pinpoint the likely defect generating process tools. By utilizing virtual metrology at each photolithography step, our system ensures exhaustive gross defect monitoring for every wafer and die, hence enhancing gross defect detection coverage and bridging existing gaps. The automated root cause analysis further enhances the ability to quickly identify defect sources, thus significantly improving early detection and intervention. This comprehensive approach promises substantial advancements in early inline detection and yield improvement.