Defect Pattern Grouping Using Polygon Similarity in Wafer Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Identifying pattern defects in semiconductor manufacturing processes is time-consuming, hindering throughput and yield, as existing methods rely heavily on manual review and grayscale image information, which can lead to incorrect classifications for certain defect types.
Innovation Solution
A defect pattern grouping method that uses graphical information to classify defects by comparing polygons from images with a set of representative polygons, allowing for automatic identification and grouping of defects based on similarity, including topological characteristics, to facilitate faster and more accurate defect identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual review of grayscale images is used for defect identification, then operators can review patterns, but the process takes substantial time and reduces throughput
Solution Approach 1:
The patent replaces manual mechanical review processes with an automated defect classification system that uses graphical pattern recognition. The system automatically compares defect polygons with representative polygons from a database, eliminating the need for operators to manually review grayscale images while maintaining or improving classification accuracy.
Solution Approach 2:
The patent creates a database of representative polygons that are copies or templates of known defect patterns. The classification system compares newly detected defects against these stored representative patterns, enabling rapid automated classification without manual intervention. This copying approach allows the system to leverage historical defect data for efficient real-time classification.
2Productivity
If grayscale image information is used for defect classification, then existing systems can operate, but classification accuracy deteriorates for certain defect types
Solution Approach 1:
The patent fundamentally changes the parameter used for classification from grayscale intensity values to graphical polygon representations. By converting defect regions into geometric shapes with defined vertices and edges, the system captures topological and morphological features that are invariant to lighting and imaging conditions, thereby improving classification accuracy for diverse defect types.
Solution Approach 2:
The patent transitions from two-dimensional grayscale pixel data to a different dimensional representation using polygon vertices and topological relationships. This dimensional transformation allows the system to capture essential defect characteristics in a more compact and discriminative form, improving classification performance while reducing data complexity.
3Productivity
If automatic defect classification is implemented, then throughput is improved, but system complexity increases
Solution Approach 1:
The patent segments the defect classification task into distinct modular components: defect detection, polygon extraction, feature characterization, database comparison, and classification. Each module performs a specific function and can be independently optimized or replaced. This segmentation manages system complexity by breaking down the overall classification process into manageable, well-defined stages.
Solution Approach 2:
The patent creates a universal classification system that handles multiple defect types through a single unified approach. The representative polygon database and comparison algorithms serve as a multi-functional core that can classify various defect patterns without requiring separate specialized systems for each defect type, thereby managing complexity through generalization.
Data Source
AI summary
A defect pattern grouping method is disclosed. The defect pattern grouping method comprises obtaining a first polygon that represents a first defect from an image of a sample, comparing the first polygon with a set of one or more representative polygons of a defect-pattern collection, and grouping the first polygon with any one or more representative polygons identified based on the comparison.