Wafer Defect Inspection Using Wavelet Position Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect inspection devices face challenges in achieving high accuracy for defect position detection due to factors like wafer alignment errors and lack of access to design data, with previous methods either relying on insufficient reference patterns or requiring confidential design data.
Innovation Solution
The method involves setting multiple reference patterns on a wafer and using wavelet transformation to correct position errors in swath channel images, allowing for accurate defect detection without relying on design data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If one reference pattern is set for a reference die, then the device complexity is reduced, but the defect position accuracy deteriorates due to insufficient reference patterns
Solution Approach 1:
The reference die swath image is divided into multiple swath channel die images, and multiple reference patterns are extracted from different regions (e.g., center region and corner regions). This segmentation allows the system to use multiple reference patterns without proportionally increasing overall complexity, as each pattern is processed independently in its specific region.
Solution Approach 2:
The reference patterns are extracted and stored in advance from the reference die swath image before actual defect inspection. This preliminary extraction creates a library of reference patterns that can be quickly applied during inspection, reducing the computational burden during real-time operation and improving defect position accuracy without adding operational complexity.
2Measurement precision
If design data is used for position error correction, then the defect position accuracy is improved, but the ease of operation deteriorates due to restricted access to confidential design data
Solution Approach 1:
Instead of requiring access to confidential design data, the system creates a copy of the actual pattern structure by extracting reference patterns directly from the reference die swath image. These extracted reference patterns serve as proxies for the design data, enabling position error correction using only the physical wafer image data that is already available to the inspection device.
Solution Approach 2:
The system performs self-calibration by extracting reference patterns from its own captured images of the reference die. This self-service approach eliminates the need for external design data input, making the system independently operable without requiring access to confidential manufacturer data while still achieving accurate position correction.
3Productivity
If wavelet transformation is applied to reduce image sizes, then the productivity is improved by decreasing calculation amount, but the measurement precision deteriorates due to reduced image resolution
Solution Approach 1:
The wafer image is divided into multiple swath channel die images, and wavelet transformation is applied selectively to reduce the size of each segment rather than the entire image. This segmented approach allows for faster processing of individual regions while maintaining the ability to perform precise position error correction on each segment using multiple reference patterns before final defect detection.
Data Source
AI summary
The purpose of the present invention is to provide a defect inspection device with which it is possible to determine a defect candidate position more accurately than before, even when design data cannot be obtained or are difficult to be utilized sufficiently. The present invention solves the problem by: setting an appropriate reference die or reference chip over a wafer to be inspected; setting, with respect to each of swath channel die images obtained by dividing a reference die swath image into a plurality of portions and detecting the portions, one or more reference patterns; correcting a position error of a swath image obtained from another die to be inspected, using the reference pattern for each swath channel image; and performing defect detection using the corrected swath channel image.