Wafer Image Defect Filtering Using Reference Images and SSIM
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional wafer image defect detection and characterization methods are time-consuming, subjective, and prone to errors due to manual review and critical dimension measurement reliance, which can lead to incorrect identification and propagation of defective images in optical proximity correction processes.
Innovation Solution
A computing system implementing a raw data filtering tool that aggregates wafer images into a reference image, uses a modified structural similarity index measure (SSIM) algorithm to generate image processing scores, filters out defective images based on threshold values, and generates defect maps to characterize defects in wafer images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual review and critical dimension measurement are used to identify defective wafer images, then defect detection can be performed, but the process becomes time-consuming and error-prone
Solution Approach 1:
The patent replaces manual mechanical review processes with an automated image processing system that uses structural similarity index measure (SSIM) algorithms and critical dimension (CD) measurements to automatically identify defective wafer images. This substitution eliminates human subjectivity and time constraints while maintaining or improving detection accuracy.
Solution Approach 2:
The system creates a reference image by aggregating multiple wafer images and then compares individual wafer images against this reference. This copying and comparison approach enables automated defect detection without requiring manual review of each image, significantly reducing time loss while improving consistency.
2Manufacturing precision
If critical dimension measurements are used to filter wafer images, then defective images can be identified, but incorrect filtering occurs due to cross-line dependency
Solution Approach 1:
The patent introduces a reference image as an intermediary between multiple wafer images. By comparing each wafer image against this intermediate reference, the system eliminates the cross-line dependency problem inherent in direct pairwise comparisons, thereby improving measurement precision while maintaining manufacturing precision.
Solution Approach 2:
The system segments the defect detection process into distinct stages: creating a reference image from aggregated data, comparing individual images against the reference, and filtering based on SSIM and CD measurements. This segmentation allows each stage to be optimized independently, improving overall precision.
3Reliability
If multiple wafer images are manually reviewed to eliminate defective ones, then quality control can be maintained, but the process is subjective and error-prone
Solution Approach 1:
The patent transforms subjective quality assessment into objective parameter-based evaluation by using SSIM values and critical dimension measurements as quantitative criteria. This parameter change from subjective to objective metrics improves reliability while the automated system manages the complexity that would otherwise require manual intervention.
Data Source
AI summary
A computing system implementing a raw data filtering tool can aggregate multiple wafer images depicting a portion of an electronic device into a reference image, detect one or more of the wafer images have defects based on a comparison of the reference image to the wafer images, and generate a gauge file to include a set of the wafer images selected based on the detection of defects in the wafer images. The raw data filtering tool also can iteratively build defect maps that include differences between the reference image and the wafer images, and characterize the detected defect in the wafer images with a size and a location based on the defect maps. The raw data filtering tool can provide feedback to a foundry about wafer images were excluded from the set of the wafer images based on the detection of defects in the wafer images.