Inline automated defect classification: a novel approach to defect management

In this work, we have developed a specific automated classification scheme based on defect size, pattern density, and optical polarity. According to our defectivity control plan, this scheme has been applied to brightfield inspections across multiple devices, as well as to our 120 nm and 90 nm technologies. The classifiers account for both size and location, while distinguishing between killer and non killer defects. The defects in array areas have higher kill ratios (KR) than those in logic or open areas. Similarly, extra-pattern defects (bright pixels) have higher KR than particles (dark pixels). After a period of baselining, we were able to set control limits for each class at each layer. Out-of-spec events can be triggered by both total and killer defect densities. As a result, we can now focus our scanning electron microscope (SEM) review primarily on those defects classified by inline automated defect classification (iADC) as killers, thus enabling better utilization of this expensive tool. As the population of SEM-reviewed defects became biased, we re-normalized to the individual iADC bin populations. iADC classification can ultimately be used to prioritize yield activity based on bin-to-defect correlation. Calculating kill ratios per iADC bin by layer provides statistically significant information, unlike total defect densities. We have shown that iADC can be used for the three main defectivity control activities: excursion monitoring, evaluation of experiments, and yield improvement through yield impact evaluation