Multi-Level Semiconductor Wafer Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing semiconductor wafer defect detection systems face challenges in accurately detecting small defects on complex lithography patterns due to noise in images and heterogeneity in wafer nodes, leading to reduced defect detectability and increased production costs.
Innovation Solution
A multi-level defect detection method involving AI-based models for image enhancement and defect detection, where input images are processed in multiple stages: initial defect detection, image parameter extraction, enhancement profile generation, image modification, and repeated defect detection to identify previously undetected defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional defect detection methods are used on noisy wafer images, then processing speed is maintained, but defect detection accuracy deteriorates due to noise and heterogeneity
Solution Approach 1:
The patent applies preliminary image enhancement and denoising processing before defect detection. The system performs initial defect detection, then selectively enhances images that may contain defects but were not detected, improving the quality of input data for subsequent detection passes and enabling better detection accuracy on noisy images
Solution Approach 2:
The patent introduces an intermediary image enhancement module between the noisy input images and the defect detection model. This intermediary process includes denoising, sharpening, and contrast enhancement operations that mediate the harmful effect of noise while preserving defect information, allowing the detection model to work with improved image quality
2Measurement precision
If multiple defect detection passes are performed on all images, then defect detection accuracy improves, but computational resources and processing time increase
Solution Approach 1:
The patent applies local quality by performing image enhancement and multiple detection passes only on specific subsets of images rather than all images uniformly. The system identifies images that require enhanced processing based on initial detection results and image quality metrics, applying computational resources selectively to cases where they will most improve detection accuracy
Solution Approach 2:
The patent segments the image processing workflow into multiple stages: initial defect detection, image quality assessment, selective enhancement, and re-detection. This segmentation allows the system to process different image subsets through different processing pipelines, reducing overall computational burden while maintaining high detection accuracy for critical cases
3Object-affected harmful factors
If uniform noise correction is applied to all images, then noise is reduced, but small defects are smoothed and lost
Solution Approach 1:
The patent applies partial action by performing selective denoising rather than uniform noise correction on all images. The system applies enhancement operations only to specific image regions and only when necessary, using adaptive thresholds that prevent over-smoothing. This partial application of noise correction reduces the risk of eliminating small defects while still improving overall image quality
Solution Approach 2:
The patent introduces dynamic adaptability in the image enhancement process, where denoising strength and enhancement parameters are adjusted based on local image characteristics, defect size, and noise levels. The system dynamically balances noise reduction and defect preservation by modifying processing intensity according to the specific conditions of each image or image region
Data Source
AI summary
Disclosed is a multi-level defect detection method and system for detecting one or more defects in semiconductor wafers, including forming, among input images of the semiconductor wafers, a first image set of input images in which defects were not detected and a second image set of input images in which defects were detected; extracting image parameters from the first image set and determining whether the first image set can be enhanced; generating an image enhancement profile for the first image set and modifying the first image set based on the image enhancement profile and detecting one or more defects in the modified image set by performing a defect detection process thereon.