Defect detection in patterned wafers using anisotropic kernels
Machine Vision and Applications
| Hi-index | 0.00 |
This paper reports an image-processing algorithm forrobust inspection of LSI wafer patterns using SEM. Inorder to detect defects in a regular LSI pattern, a pair oflong patterns are compared, blocked images arealigned, and defects are judged using the alignedimages. The LSI wafer pattern is defined to consist ofblank space, fine repetitive patterns, and uniquepatterns. Distortion of the SEM image is larger than therepetitive pattern pitch, requiring the system to keeptrack of the alignment in areas without patterninformation or in blank space and mitigate theindeterminacy of repetitive patterns. To satisfy theserequirements, a two-layer algorithm is proposed. Thelower layer calculates registration candidates in eachblock, and the upper layer determines the correctregistration route, i.e. the chain of the correctregistration, using candidate information in all therelated blocks. Experimental evaluations confirm thatmost pattern cases can be inspected correctly using theproposed SEM inspection system.