Excursion Wafer Loss Prediction by Local Density Segmentation

The most obvious characteristics of an excursion wafer is an elevated defect density, either confined within a signature or distributed across its surface. Quantifying the impact of such abnormal levels of defectivity is the focus of this paper. Three methods are described which provide increasingly precise predictions of Defect Limited Yield. The methods are also intended to accurately handle excursion wafers that include highly divergent signatures, as well as non-optimized defect inspection recipes which may contain elevated levels of nuisance defects. This is achieved by combining the defect distribution with the kill ratio, via a geometric progression. The conditions under which each method is ideally suited are presented.