Thesis (M.Sc.) - Cairo University - Faculty of Engineering - Department of Computer Engineering

Due to the continuous scaling down in the feature size of semiconductor process technology nodes, the industry is greatly challenged in terms of manufacturing and design. The challenge is due to the widening gap between the lithography capabilities that is still limited by the continued scaling of the feature size and the 193nm wavelength. To cope up with these challenges, many manufacturing-friendly design techniques are proposed and developed to detect the process hotspots in the early stages and to ensure high production yield at post silicon stage. The feature extraction stage is one of the most critical steps in machine learning and pattern matching techniques that are recently used for hotspot detection. This is because feature extraction is representing the patterns using simple but comprehensive feature information so as to distinguish all the selected features. This thesis proposes a new feature representation for fuzzy matching for lithography hotspot detection. This technique enhances the Modified-Transitive-Closure-Graph (MTCG) by adding specific Do-not-Care (DC) regions to filter out unwanted polygons. Our technique is capable of reaching 88% success rates by 10% increase with no impact on total run-time, compared to the conventional MTCG. Building on the uniqueness of MTCGs for any tiled pattern, this thesis also introduces a new similarity-detection technique that detects hotspots of similar shapes with acceptable defined tolerance. This technique, besides MTCG, is able to reach 97.044% success rate with only 1.0287% increase in total run-time

Issued also as CD