Field-test results of an image retrieval system for semiconductor yield learning

Images of semiconductor defects are maintained in semiconductor yield management systems to diagnose problems that arise during the manufacturing process. A semiconductor -specific content-based image retrieval system was developed by Oak Ridge National Laboratory (ORNL) under the auspices of International SEMATECH (ISMT) during 1998-1999. The system uses commercial databases to store image information and uses a customized indexing technology to rapidly retrieve similar images. Additional defect information (position, wafer ID, lot, etc) has now been incorporated into the system through the use of additional database tables. During Fall 2000, the system was deployed in two ISMT member company fabs to demonstrate the utility of this approach in managing large databases of images and to show causal relationships between image appearance and wafer information such as processing layer, wafer lot, analysis dates, etc. This paper summarizes the results of these field tests and shows the utility of this approach through data analysis conducted on approximately one month of historical defect data.