Role of defect size distribution in yield modeling

Most publications on yield modeling for large-scale integrated circuits have concerned themselves only with the role of the spatial distribution of defects. While it was recognized that the size of the defects affected the yield, the role of the size distribution does not appear to have been fully investigated. This paper examines the effect of different assumptions concerning the defect size distribution and shows that they exert a strong effect on the projected yield. It is also shown that attempts to determine the defect size distribution from yield data are subject to fundamental limitations that need to be addressed.