Wafer Image Templates for Robust Target Segmentation Metrology
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing techniques for semiconductor metrology are prone to noise and defects, leading to inaccurate measurement of fine features due to missing or merged features in repetitive patterns, and unguided sampling fails to provide comprehensive metrology data.
Innovation Solution
The method involves creating algorithmically generated image templates that mimic the repetitive pattern of device features on a wafer, anchoring these templates to original images, and segmenting the images into target areas for robust metrology, using techniques like K-means clustering to identify and isolate features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If direct measurement from raw images is performed, then measurement process is simple, but measurement precision deteriorates due to noise and artifacts
Solution Approach 1:
The patent segments the wafer image into multiple tiles and further divides each tile into superpixels based on color similarity. This hierarchical segmentation allows independent processing of small regions, reducing noise impact while maintaining overall measurement accuracy through aggregated results from multiple segments.
Solution Approach 2:
The patent introduces intermediate processing steps including superpixel formation, template matching, and feature point detection as mediators between raw image acquisition and final measurement. These intermediate representations filter out noise and artifacts while preserving critical dimensional information.
2Area of stationary object
If unguided sampling is used on repetitive patterns, then measurement coverage is limited, but measurement precision deteriorates due to noise and missing features
Solution Approach 1:
The patent performs preliminary actions by detecting the repetitive pattern structure before sampling, creating a template that guides subsequent measurements. This preliminary pattern recognition enables systematic sampling across the entire wafer area while avoiding regions with defects or artifacts, thereby maintaining high measurement precision across large coverage areas.
Solution Approach 2:
The patent creates a universal template that can be applied across the entire wafer surface, allowing the same measurement methodology to be used consistently over large areas. This template-based approach provides comprehensive coverage while maintaining precision through the reusable, validated measurement procedure.
3Ease of manufacture
If traditional edge detection or threshold methods are used, then feature extraction is straightforward, but reliability deteriorates when features are missing or merged
Solution Approach 1:
The patent creates idealized templates that represent the expected perfect pattern of device features. By comparing actual measurements against these copied ideal templates, the system can reliably detect deviations, missing features, or merged features even when traditional edge detection would fail, thereby maintaining high reliability without significantly increasing complexity.
Data Source
AI summary
One or more images of a portion of a wafer with fabricated devices are acquired using an imaging tool. A pattern of repeating features in an input image of a wafer is identified using various methods, such as correlation and clustering of neighboring vectors. A template is generated based on the found pattern of repeating features. The template is aligned with the acquired image to identify target locations. The target locations are then isolated from the original image for performing detailed metrology.