Semiconductor Wafer Defect Detection Using Relative Pixel Intensity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical inspection tools for semiconductor wafers face challenges in accurately detecting defects while minimizing false alarms, particularly due to gray level differences caused by process variations or image conditions, which can lead to incorrect defect identification.
Innovation Solution
The method involves comparing the similarity of pattern shapes between inspection and reference images by assigning relative values to pixels based on their neighboring pixels, using comparison metrics such as ordinal ranks and directional vectors to determine the degree of similarity, thereby reducing false alarms and improving defect detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If gray level comparison is used for defect detection, then defect detection capability is provided, but false alarm rate increases due to process variation and image condition changes
Solution Approach 1:
The patent transforms the comparison parameter from absolute gray levels to relative intensity relationships (ratios and differences between neighboring pixels). This parameter transformation makes the comparison invariant to uniform illumination changes and process variations, thereby reducing false alarms while maintaining defect detection capability
Solution Approach 2:
The patent performs preliminary processing to compute relative intensity values (ratios and differences) for all pixels before the actual defect detection comparison. This preliminary transformation of the image data into relative intensity space prepares the images for robust comparison that is insensitive to gray level variations caused by process changes
Data Source
AI summary
Inspection of objects such as semiconductor wafers can include comparisons of shapes between inspection and reference images. As part of the inspection process, relative values may be assigned to pixels within each image based on comparison of such pixels to neighboring pixels. For instance, the pixels may be ranked by relative brightness in each image. Alternatively, directional vectors may be defined based on slopes between pixels and their neighbors. Various comparison metrics may be utilized to determine the degree of correlation between the relative values for pixels in the inspection image and corresponding pixels in the reference image. Relative values may be combined with conventional techniques as part of an inspection process. The inspection may be performed using an optical inspection tool that uses conventional techniques to identify defect candidates, with relative value analysis performed on areas containing defect candidates to confirm or deny the existence of a defect.