Wafer Defect Detection Using Adaptive Threshold Division Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional defect inspection methods in semiconductor manufacturing face challenges in accurately detecting defects due to the need for proper threshold settings, which can lead to reduced sensitivity or misjudgment between real defects and noise.
Innovation Solution
A method and system that utilize a computer algorithm to analyze a reference image of a chip pattern, produce a division map, set respective thresholds for each division, and compare test images to identify defects based on these thresholds, improving inspection accuracy and reducing manual burden.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a high threshold is set for defect detection, then false positives are reduced, but sensitivity is reduced and real defects may be missed
Solution Approach 1:
The patent divides the wafer surface into multiple regions based on pattern density characteristics. Each region is assigned a different threshold value according to its specific characteristics. This segmentation allows high thresholds to be applied in low-density areas (reducing false positives) while lower thresholds are used in high-density areas (maintaining sensitivity), thus resolving the contradiction between reliability and measurement precision.
Solution Approach 2:
The patent implements local quality by setting different threshold values for different regions of the wafer based on their local pattern density characteristics. Instead of using a uniform global threshold, each region receives a customized threshold that optimizes defect detection for its specific characteristics, thereby simultaneously improving both reliability and measurement precision across different areas.
2Measurement precision
If a low threshold is set for defect detection, then sensitivity is improved, but false positives increase and real defects cannot be discriminated from noise
Solution Approach 1:
The patent segments the wafer into multiple regions with different pattern density characteristics and applies different thresholds to each segment. This allows low thresholds to be used in high-density regions (maintaining sensitivity) while high thresholds are applied in low-density regions (reducing false positives), thus resolving the contradiction between measurement precision and reliability.
Solution Approach 2:
The patent applies local quality by customizing threshold values according to the local pattern density of each region. Regions with high pattern density receive lower thresholds to maintain sensitivity, while regions with low pattern density receive higher thresholds to reduce false positives, thereby simultaneously achieving high measurement precision and reliability across the entire wafer.
3Ease of operation
If a single global threshold is used for the entire wafer, then the inspection process is simple, but inspection accuracy is reduced due to varying pattern densities across different regions
Solution Approach 1:
The patent applies preliminary action by automatically dividing the wafer into multiple regions and setting appropriate thresholds for each region before the actual defect inspection process. This preliminary segmentation and threshold assignment is performed automatically by the inspection system, maintaining ease of operation while significantly improving inspection accuracy through region-specific threshold optimization.
Solution Approach 2:
The patent implements self-service by enabling the inspection system to automatically perform the segmentation and threshold setting process without requiring manual intervention. The system autonomously analyzes pattern density characteristics, divides regions accordingly, and assigns optimal thresholds, thereby maintaining operational simplicity while achieving high inspection accuracy through adaptive region-specific thresholds.
Data Source
AI summary
A method for detecting defects on a wafer including the steps of obtaining a reference image of a chip pattern formed on a reference wafer, using a computer algorithm to analyze the reference image to produce a division map for the chip pattern; setting respective thresholds for divisions of the division map, obtaining a comparison data between a test image of the chip pattern formed on a test wafer and the reference image, using the division map and the thresholds to examine the comparison data to identify a defect in the test image.