Detection of defects on the surface of microelectronic structures

Many types of defects on photolithographic masks can be detected by using only local pattern properties specified by general design rules. No comparison between adjacent chip images or an external data base is required, eliminating the need for high metrological precision. A number of detection criteria such as local deviations from straightness or admissible contour angles were tested on simulated images containing various defect types. Usable algorithms that could be implemented as high-speed circuits permitting data throughputs at TV rates are discussed