Spatial signature analysis of semiconductor defects

Automated tools for defect analysis are becoming more and more necessary as device densities and wafer sizes continue to increase. Such tools can efficiently and robustly process the increasing amounts of data, and thus quickly characterize manufacturing processes and accelerate yield learning. A newly developed image-based method analyzes process `signatures' in defect data distributions. This article describes the statistical and morphological image-processing techniques used to segment signature events into high-level, process-oriented categories. We present examples of enhanced statistical process control, automated process characterization, and intelligent subsampling of event distributions for off-line defect review.