Regression Analysis of Low Kill Potential Defect Mechanisms (Moving Beyond Contingency Analysis)

A means of measuring the yield influence for high density, low kill potential defect mechanisms is presented. The method can be applied to wafers individually and is only, marginally more expensive to implement than a least squares fit, but has a much smaller standard error and provides a clear indicator of impact on yield. The ability to tolerate up to 50% outliers enables the method to analyze zone-based data and consequently be resilient to the presence of systematic yield profiles. This is achieved by focusing on the zone specific, kill ratio per defect. This can greatly improve the estimate of yield impact on priority wafers when affected by a defect mechanism whose density varies wildly between wafers, or confirm the validity of the standard defect classification kill ratios. For wafers with typical levels of defectivity, the method can confirm the results of traditional contingency table analysis, with the added confidence that the results are based on a curve fit to any progressive yield response, rather than a single calculation.