Defect Inspection System Using Pattern-Based Probability Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect detection systems in integrated circuit manufacturing face challenges in accuracy and throughput, particularly at smaller technology nodes, leading to high rates of nuisance defects and inefficient defect review processes.
Innovation Solution
A charged particle beam system with advanced image processing methods that include first and second image analysis, pattern grouping, and a pattern library to determine defect probabilities, allowing for selective review of high-risk defects and filtering out nuisance defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If random sampling of defects is used for review, then all defects have equal chance of being reviewed, but the rate of real defect review becomes low and misleading information is provided
Solution Approach 1:
The patent applies local quality by differentiating between different types of defects and applying different review priorities. Real defects are identified and prioritized for review while nuisance defects are filtered out, ensuring that review resources are concentrated on areas with higher value and accuracy requirements.
Solution Approach 2:
The patent replaces random mechanical sampling with an intelligent selection system that uses defect classification and probability assessment. The system substitutes random selection with a data-driven approach that calculates real defect probabilities and selects samples based on these probabilities, improving both accuracy and efficiency.
2Reliability
If repeated defect review is conducted to ensure proper coverage, then real defect review coverage improves, but the process becomes time consuming
Solution Approach 1:
The patent applies preliminary action by performing defect classification and probability assessment before the review process. By pre-identifying real defects and calculating their probabilities, the system prepares a prioritized list that ensures proper coverage without requiring repeated reviews, thus saving time while maintaining reliability.
Solution Approach 2:
The system uses feedback from defect classification results to guide the review process. By analyzing defect patterns and probabilities, the system continuously improves its selection criteria, ensuring that review coverage is optimized over time without requiring excessive repeated reviews.
3Measurement precision
If inspection is performed on all identified defects, then comprehensive defect detection is achieved, but the large amount of nuisance defects reduces efficiency
Solution Approach 1:
The patent extracts and separates real defects from nuisance defects using classification algorithms and probability assessment. By taking out only the relevant real defects for detailed inspection while filtering out nuisance defects, the system reduces inspection complexity and improves efficiency without compromising detection accuracy.
Solution Approach 2:
The system changes the parameter of defect selection from random or uniform sampling to probability-based sampling. By adjusting the selection parameters based on defect characteristics and calculated probabilities, the system optimizes the balance between comprehensive detection and efficient processing.
Data Source
AI summary
Systems and methods for detecting defects are disclosed. According to certain embodiments, a method of performing image processing includes acquiring one or more images of a sample, performing first image analysis on the one or more images, identifying a plurality of first features in the one or more images, determining pattern data corresponding to the plurality of first features, selecting at least one of the plurality of first features based on the pattern data, and performing second image analysis of the at least one of the plurality of first features. Methods may also include determining defect probability of the plurality of first features based on the pattern data. Selecting the at least one of the plurality of first features may be based on the defect probability.