The benefi ts of automatic classifi cation of microlithography defects include fast and reliable rework decisions, improved root-cause analysis, and more consistent SPC data that signifi cantly enhances yield in the lithography cell. An adaptive knowledge-based system has demonstrated the ability to accurately classify defects more than 85% of the time and is suffi ciently versatile to classify new defect modes that will accompany advanced lithography processes. The knowledgebased system defi nes each class of defects with mathematical descriptors that include categories such as size, intensity, edge sharpness, color, etc. New defect classes can be defi ned with as few as three to fi ve images of the specifi c defect. All defect classes are stored in the knowledge-base as rule vectors consisting of values for each descriptor. Different defect classes can share many common descriptors. However, as long as there is at least one descriptor that differentiates them, the defect class can be deemed unique. This method provides manufacturers the ability to defi ne defects according to their existing rules and to defi ne new defect types as they occur.
The IBM 300 mm wafer manufacturing line provides a case study for the optimization of an automated macro defect inspection system to accurately flag global wafer color variation. The IBM inspection system was falsely flagging a large number of wafers primarily for global wafer color variation, leading to unacceptable amounts of production volume being placed on hold. A review of the macro inspection system identified several areas for improvement. An investigation into the installed hardware base found a panel behind the beam splitter was introducing noise through reflected light. This panel was replaced with a less reflective material. A review of the failed wafers found that maximum light levels were not achieved across all tools and an improved diffuser plate for the fiber optic output was designed to improve light transmittance. Global wafer color is determined by comparing the scanned wafer image to a "golden" data set, referred to as a "color baselist," which is composed of data from 30 wafers. A review of the recipe baselists revealed that some of the wafer samples did not accurately represent process conditions, and new wafer samples were collected. Finally, a tool-to-tool matching test revealed that the set of weightings given to each of the color parameters in the baselists was not optimized. After implementing the above changes, false global wafer color failures were virtually eliminated.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.