A GOLDEN-BLOCK BASED SCHEME FOR CONTINUOUS PATTERNED WAFER INSPECTION

Abstract

As the complexity of integrated circuits increases rapidly, there is a strong need for computer vision based inspection system to be able to efficiently detect various defects in the wafer stage. Th.is thesis focuses on a novel technique for detecting possible defects in periodic two-dimensional wafer images when there are no reference images or priori knowledge. It is capable of creating a golden block database from the wafer image itself, modifying and refining its content when used in further inspections. Spectral estimation is used in the first step to derive the periods of repeated patterns in both directions. Then a building block representing the structure of the patterns is extracted using interpolations to obtain sub-pixel resolution. After that, a new defect-free golden template is built based on the building block extracted. A pixel-to-pixel comparison is then applied to find out possible defects. The extracted building block is stored as a golden block for the detected pattern at the end of the process. When a new wafer image with the same periodical pattern arrives, we do not have to re-calculate its periods and building block. A new building block can be derived directly from the existing golden block. If the newly derived building block has better quality than the stored golden block, then the golden block is replaced with the new building block. The proposed scheme is a bridge between the existing self-reference methods and image-to-image reference methods. Our implementation has shown that the proposed algorithms work successfully with IC samples, and the processing time is greatly improved compared to that of those previously published methods.