Semiconductor Defect Detection via Frequency Domain Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting defects in semiconductor patterns require large storage spaces and precise image alignment, and are prone to errors due to color differences and alignment issues when comparing multiple images.
Innovation Solution
A method involving a Fourier transform and low-pass filtering of images to transform and filter them into a frequency domain, allowing for defect detection by comparing the filtered images with a reference image without the need for precise alignment, using relation values between pixel grey levels to determine defect existence and positions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple adjacent images are compared to detect defects, then defect detection capability is improved, but storage space requirements increase
Solution Approach 1:
The patent extracts only the essential defect information from images by transforming to frequency domain and applying low-pass filtering. This removes redundant high-frequency data while preserving defect characteristics, allowing effective defect detection with reduced storage requirements for image data
Solution Approach 2:
The patent changes the representation parameters of images by transforming from spatial domain to frequency domain. This parameter transformation allows the system to work with transformed image data that requires less storage while maintaining defect detection capability through frequency domain analysis
2Measurement precision
If multiple images with different colors are compared, then defect detection is performed, but color differences cause detection errors
Solution Approach 1:
The patent applies low-pass filtering in the frequency domain to remove high-frequency color variations while preserving the structural defect information. This parameter filtering approach changes the image representation to emphasize structural features over color differences, improving detection reliability
Solution Approach 2:
The frequency domain transformation acts as an intermediary that separates structural defect information from color variation information. By working in this intermediate representation space, the system can compare images while ignoring color differences that would otherwise cause false detections
3Measurement precision
If precise alignment between images is performed, then defect detection precision is improved, but alignment complexity and time increase
Solution Approach 1:
The patent transforms images to frequency domain where periodic patterns appear as distinct frequency components. This parameter transformation makes the images inherently more robust to misalignment, reducing the complexity of alignment operations while maintaining detection precision
Solution Approach 2:
The patent performs low-pass filtering as a preliminary action before defect comparison. This preprocessing step smooths out minor misalignments and reduces the sensitivity to alignment errors, allowing simpler alignment procedures while maintaining detection accuracy
Data Source
AI summary
In a method of detecting defects in patterns and an apparatus for performing the method, a first image of a detection region on a semiconductor substrate may be acquired. A second image may be acquired from the first image by performing a Fourier transform and performing a low pass filtering. The second image may be compared with a reference image so that the defects of the detection region are detected. Existence of the defect of the second image is determined using a relation value between a grey level of each of pixels of the second image and the reference image, respectively. When a defect exists, the horizontal and the vertical positions of the pixel where the relation value is minimum are combined to determine the position of the defect.