Mask Inspection Using Statistic-Based EPD for True Defect Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mask inspection technologies face challenges in accurately detecting defects in semiconductor wafers due to limitations in light wavelength and diffraction orders, leading to misalignment and electrical disturbances, and struggle with filtering out false defects caused by tool noise and pattern noise.
Innovation Solution
A method and system for mask inspection using statistic-based Edge Positioning Displacement (EPD) parameters, which calculates EPD using a Print Threshold (PT) and filters defects as 'true' or 'false' based on predefined thresholds, enhancing defect detection and reducing noise through overlapping image processing and multi-reference calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional intensity-based defect detection is used, then defect detection capability is maintained, but false defects caused by tool noise and pattern noise cannot be effectively filtered
Solution Approach 1:
The patent transitions from intensity-based detection parameters to EPD (Edge Positioning Displacement) parameters that measure positional deviations of pattern edges. By calculating EPD values and comparing them against thresholds, the system effectively distinguishes true defects from false defects caused by noise, improving detection reliability without excessive complexity
Solution Approach 2:
The patent replaces traditional optical/mechanical detection methods with a computational approach that processes aerial images through algorithmic EPD calculation. This substitution enables more sophisticated defect filtering by leveraging statistical analysis of edge positions rather than relying solely on intensity thresholds
2Measurement precision
If light wavelength and diffraction orders are used for feature projection, then imaging capability is achieved, but edge positioning precision is limited
Solution Approach 1:
The patent implements a feedback mechanism where aerial images are processed to calculate EPD values, which are then used to classify defects. The system continuously refines its defect identification by comparing calculated EPD against predefined thresholds, enabling precise edge positioning measurement while filtering out artifacts that cause misalignment and electrical disturbances
3Reliability
If multiple aerial images are processed to calculate statistic-based EPD, then false defect filtering is improved, but processing time increases
Solution Approach 1:
The patent applies partial action by processing multiple aerial images selectively - specifically processing images at different focus positions only for regions where defects are detected. This approach improves false defect filtering effectiveness by using statistical EPD analysis while limiting processing time by avoiding exhaustive processing of the entire mask area
Data Source
AI summary
There is provided a mask inspection system and a method of mask inspection. The method comprises: detecting, by the inspection tool, a runtime defect at a defect location on a mask of a semiconductor specimen during runtime scan of the mask, and acquiring, by the inspection tool after runtime and based on the defect location, a plurality sets of aerial images of the runtime defect corresponding to a plurality of focus states throughout a focus process window, each set of aerial images acquired at a respective focus state. The method further comprises for each set of aerial images, calculating a statistic-based EPD value of the runtime defect, thereby giving rise to a plurality of statistic-based EPD values each corresponding to a respective focus state, and determining whether the runtime defect is a true defect based on the plurality of statistic-based EPD values.