Automatic defect classification system for patterned semiconductor wafers

Abstract

The automatic detection of defects on patterned wafers is a well-developed technology. There are a number of commercially available defect detection systems which analyze the patterned wafers, determine the defect locations, and provide information on defect size. The next logical step would be to visually (optically or SEM) review the defects and determine the types which comprise the defect distributions obtained from the inspection tool, i.e. classify the defects according to some predetermined scheme. Of strong interest is then to use this information to isolate those problems which result in electrical failures and focus effort on solving those problems. Unfortunately, the process of classification is typically a manual process limited in its effectiveness by human factors. Replacing this manual bottleneck with an automated system is highly desirable. An automatic Defect Classification System (DCS-1) which combines image processing techniques and Fuzzy Logic Expert System (FLES) is now commercially available. The DCS-1 re-detects the defects in the field of view of a review station using image processing techniques and then classifies them using the FLES engine. In this paper we describe the DCS-1 which automates the defect classification process, offering a high degree of adaptivity, customization, and automation.