Intelligent Weak Pattern Diagnosis for Semiconductor Yield
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional defect inspection tools in semiconductor fabrication fail to effectively identify weak defect patterns in the early stages of the process, leading to low yield rates and increased costs due to optical limitations and complex design layouts.
Innovation Solution
A computer-implemented intelligent weak pattern diagnosis system that screens and filters weak patterns using a software and hardware framework, comparing design layout data with defect pattern libraries to categorize known and unknown patterns, and utilizing image processing to measure contours, dimensions, and coordinates of weak patterns for systematic defect recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional defect inspection tools are used, then the inspection process is simple, but weak defect patterns cannot be effectively identified
Solution Approach 1:
The inspection system is segmented into multiple functional modules: defect pattern library construction module, weak pattern screening module, contour matching module, and verification module. Each module handles a specific aspect of the inspection process, enabling complex weak defect detection through coordinated simple operations.
Solution Approach 2:
Defect pattern libraries are constructed in advance containing known defect patterns and their characteristics. During inspection, the system performs preliminary screening by comparing detected defects against these pre-established patterns, enabling rapid identification of weak defects without complex real-time analysis.
2Productivity
If early defect detection is implemented, then yield rate improves, but detection complexity increases
Solution Approach 1:
The system performs preliminary defect pattern library construction and weak pattern screening at the beginning of the fabrication process. By detecting and categorizing weak defects early using automated comparison with pre-stored patterns, the system enables early intervention without requiring complex manual analysis throughout the production cycle.
Solution Approach 2:
The system creates digital copies of defect patterns from physical inspections and stores them in defect pattern libraries. These digital copies are then reused for comparing against new defects, eliminating the need to re-analyze defect characteristics manually and enabling scalable early detection across multiple production batches.
3Measurement precision
If comprehensive pattern matching is performed, then detection accuracy improves, but processing time increases
Solution Approach 1:
The pattern matching process is segmented into hierarchical stages: first screening by basic defect characteristics, then contour matching for suspected weak patterns, and finally detailed verification only for patterns matching known defect types. This segmentation enables comprehensive analysis without processing every defect at maximum detail level.
Solution Approach 2:
The system applies different levels of analysis quality to different defects based on their characteristics. Known defect patterns receive standardized quick matching, while unknown or suspicious patterns receive more detailed contour analysis. This localized quality adjustment maintains high accuracy for critical cases while reducing processing time for routine defects.
Data Source
AI summary
Disclosure herein is related to a method and a system for intelligent weak pattern diagnosis for semiconductor product, and a related non-transitory computer-readable storage medium. In the method, a weak pattern layout is firstly retrieved from a defect pattern library and a frequent failure defect pattern library; defect data is retrieved from fab defect inspection tool; a design layout is then received and weak defect pattern screen is performed to extract known and unknown weak defect patterns. In addition to updating the weak pattern library, the weak pattern contour can be made upon SEM image data, and then the true systematic weak pattern can be justified.