Wafer Backside Defect Detection via Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for detecting wafer backside defects, such as roughness and scratch defects, are inaccurate and rely on manual labor, failing to ensure firm bonding in wafer-to-wafer stack technology.
Innovation Solution
A method and apparatus for analyzing wafer backside images to automatically detect defects by cropping peripheral edges, connecting adjacent white pixels to identify abnormal regions, and determining defects based on area percentage and length criteria, enabling accurate identification of roughness and scratch defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual detection methods are used, then device complexity is reduced, but measurement precision and reliability deteriorate
Solution Approach 1:
The patent replaces manual mechanical inspection with an automated image processing system that uses optical imaging combined with digital image analysis algorithms. The system captures wafer backside images and automatically analyzes them to detect defects, eliminating the need for manual visual inspection while achieving higher measurement precision.
Solution Approach 2:
The patent creates a digital copy of the wafer backside surface through image capture, then analyzes the digital image data to detect defects. This copying approach allows for precise measurement and analysis without physically contacting or manipulating the wafer, thereby improving detection accuracy while maintaining system simplicity.
2Productivity
If automated image analysis is implemented, then productivity and measurement precision improve, but device complexity increases
Solution Approach 1:
The patent segments the wafer backside surface into a grid of analysis regions, processing each region independently to identify defects. This segmentation approach enables automated high-speed analysis of large wafer surfaces while keeping the processing complexity manageable through divide-and-conquer strategy.
Solution Approach 2:
The patent changes the analysis parameters dynamically based on detected abnormal regions, adjusting detection criteria and analysis depth according to the specific characteristics of each defect type. This adaptive parameter adjustment enables efficient automated detection across different defect scenarios without requiring overly complex fixed-systems.
3Measurement precision
If peripheral edge areas with notches are included in analysis, then measurement precision improves, but false positive rate increases
Solution Approach 1:
The patent extracts and removes the peripheral edge area containing notches from the analysis region, creating a cleaned-up image dataset for defect detection. By taking out the problematic edge regions that cause false positives, the system maintains measurement precision on the actual wafer surface while eliminating spurious detections from the notch areas.
Data Source
AI summary
A wafer backside defect detection method and a wafer backside defect detection apparatus are provided. The wafer backside defect detection method includes the following steps. A peripheral edge area of a wafer backside image that at least one notch is located is cropped off. Adjacent white pixels on the wafer backside image are connected to obtain a plurality of abnormal regions. If a total area of top N of the abnormal regions is more than 10% of an area of the wafer, it is deemed that the wafer has a roughness defect. N is a natural number. If the total area of the top N of the abnormal regions is less than 1% of the area of the wafer and a largest abnormal region of the abnormal regions is longer than a predetermined length, it is deemed that the wafer has a scratch defect.