Wafer Inspection Filtering Nuisance Defects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing semiconductor wafer defect detection methods struggle to differentiate between yield-limiting and non-yield-limiting defects, often classifying harmless anomalies as problematic, leading to unnecessary filtration and increased false alarm rates.
Innovation Solution
A wafer inspection system that performs initial filtration to generate candidate defects, calculates image attributes, and uses iterative threshold refinement based on operator classification to achieve a desired false alarm rate, distinguishing between true defects and nuisances by analyzing attribute distributions and correlations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing defect detection methods classify all identified defects as problematic, then the detection sensitivity is high, but the false alarm rate increases and productivity decreases due to unnecessary filtration of harmless anomalies
Solution Approach 1:
The patent segments defects into different categories (yield-limiting defects versus nuisance defects) based on their impact on wafer performance. The system classifies defects into distinct groups such as critical defects, minor defects, and nuisance defects, allowing selective attention to only those defects that truly affect yield, thereby reducing false alarms and improving classification efficiency
Solution Approach 2:
The patent changes the parameters used for defect evaluation by introducing multiple defect attributes (size, shape, location, pattern matching characteristics) and using statistical analysis of these parameters to distinguish between harmful and harmless defects. This multi-parameter approach enables accurate differentiation without requiring excessive filtration of all detected defects
2Measurement precision
If existing defect detection methods perform comprehensive analysis of all defects, then the detection precision is high, but the complexity of the detection system increases
Solution Approach 1:
The patent performs preliminary classification of defects based on easily obtainable attributes such as size, shape, and location before conducting more detailed analysis. By pre-filtering and pre-categorizing defects using simple criteria, the system avoids complex analysis of all defects and focuses computational resources only on defects that require detailed characterization
Solution Approach 2:
The system uses self-service mechanisms by automatically learning from historical defect data and operator feedback to refine its classification algorithms. The system autonomously improves its defect characterization capabilities through iterative learning without requiring increasing system complexity, as the intelligence is embedded in software rather than hardware
3Quantity of substance
If existing defect detection methods classify all defects without differentiation, then the detection coverage is complete, but the loss of information occurs by treating harmless anomalies as problematic
Solution Approach 1:
The patent extracts and isolates the critical information about defect relevance by separating yield-limiting defects from nuisance defects. The system extracts key attributes and patterns that indicate true yield impact, removing the noise of harmless anomalies from the critical defect list, thereby preserving information about which defects truly matter while maintaining complete detection coverage
Data Source
AI summary
A method for inspecting a sample, consisting of receiving a definition of image attributes that are characteristic of defects, and processing an image of the sample so as to identify candidate defects on the sample. The method further includes forming distributions of values of the respective attributes from the candidate defects, and selecting a set of the candidate defects that are characterized by respective candidate attribute values that fall in one or more tails of the distributions. The selected set is presented to a human operator, and respective classifications of the candidate defects in the selected set are received from the operator. A definition of the one or more tails of the distributions is refined responsively to the classifications. The method may be used as a filter to remove false alarms, or nuisances. The method may also be used to categorize the candidate defects into two or more classes.