Semiconductor Wafer Inspection With Sub-Pixel Die Image Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is a growing need for a reliable method and system to detect defects in fine patterns on semiconductor devices, as the integration density of these devices increases, and existing inspection methods face challenges in accuracy due to misalignment and noise.
Innovation Solution
The proposed solution involves a semiconductor wafer inspection method that includes providing a wafer with a target die and a reference die, obtaining candidate and reference images, performing super resolution imaging to enhance image resolution, shifting the reference image in sub-pixel units, and generating a difference image to detect defects based on threshold values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional inspection methods are used to detect defects in fine patterns, then the inspection process can be performed, but the measurement precision deteriorates due to misalignment and noise
Solution Approach 1:
The patent performs preliminary alignment by obtaining a reference image from a reference die and using it to align the candidate image from the target die before defect detection. This preliminary alignment action eliminates misalignment issues during the actual inspection, thereby improving measurement precision without compromising reliability
Solution Approach 2:
The patent creates a reference image by copying the pattern from a reference die that is known to be defect-free. This reference copy is then used for comparison with the target die image, enabling accurate defect detection by eliminating the need for direct comparison with ideal patterns and reducing noise interference
2Manufacturing precision
If conventional image comparison is used without super resolution imaging, then the inspection process is faster, but the manufacturing precision deteriorates due to insufficient resolution for fine patterns
Solution Approach 1:
The patent transitions from conventional pixel-level image comparison to sub-pixel level comparison through super resolution imaging. By dividing each pixel into multiple sub-pixels and performing comparisons at this finer dimensional level, the system achieves higher manufacturing precision for fine pattern inspection while maintaining computational efficiency through optimized processing algorithms
3Reliability
If threshold value-based defect detection is used, then the inspection process is simple, but the reliability deteriorates due to strong dependence on threshold value settings
Solution Approach 1:
The patent replaces the conventional threshold-value-based defect detection mechanism with a reference image comparison mechanism. Instead of relying on manually set threshold values to determine defects, the system uses differential comparison with a reference image, automatically identifying defects as deviations from the reference pattern. This substitution eliminates the need for threshold tuning while improving reliability
Solution Approach 2:
The reference image serves as a self-generated standard from the same wafer, eliminating the need for external reference standards or manual threshold setting. The system uses its own reference die to create the comparison baseline, making the inspection process self-sufficient and reducing dependence on external parameters
Data Source
AI summary
A semiconductor wafer inspection method is provided. The semiconductor wafer inspection method includes: providing a wafer with target and reference dies; obtaining a candidate image of a first region of the target die and a reference image of a second region of the reference die; performing an imaging process on the candidate image to obtain a high resolution candidate image including sub-pixels for each pixel of the candidate image; performing the imaging process on the reference image to obtain a high resolution reference image including sub-pixels for each pixel of the candidate image; shifting the high resolution reference image in units of the sub-pixels; obtaining a difference image based on a difference between the high resolution candidate image and the high resolution reference image; and detecting whether a defect signal generated based on the difference image exceeds a threshold value.