Wafer Intensity Change Approximation for Fast Mask Defect Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mask defect inspection systems require significant time and resources for full simulations, making them inefficient for fast defect scoring, especially in high-volume mask production where less stringent accuracy is acceptable.
Innovation Solution
The implementation of approximated wafer simulations using a Gaussian filter and single convolution to compute the maximum intensity difference between defect and reference images, allowing for quick identification of nuisance defects and reducing computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full simulations are performed for defect inspection, then measurement precision is improved, but loss of time increases significantly
Solution Approach 1:
The patent applies partial action by performing only the necessary computational steps for defect detection rather than complete simulations. The method calculates wafer image intensity changes using simplified convolution operations with Gaussian filters, omitting unnecessary simulation steps while maintaining sufficient accuracy for identifying significant defects.
Solution Approach 2:
The patent changes the computational parameters by using approximated wafer simulations with Gaussian filters instead of full simulations. This parameter change reduces computational complexity from complete optical simulations to simplified intensity change calculations, achieving faster processing while maintaining defect detection capability.
2Measurement precision
If full simulations are performed for defect inspection, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts the essential information needed for defect detection from complex simulations by focusing only on wafer image intensity changes. It separates the critical defect detection function from unnecessary simulation components, retaining only the convolution operations with Gaussian filters that provide sufficient accuracy.
Solution Approach 2:
The patent simplifies computational complexity by changing simulation parameters to use approximated wafer simulations. Instead of complete optical simulations, it uses single convolution operations with Gaussian filters, reducing computational burden while maintaining measurement precision for defect detection.
3Speed
If approximated wafer simulations are used, then speed is improved, but measurement precision deteriorates
Solution Approach 1:
The patent implements feedback by comparing approximated wafer image intensity changes against threshold values to determine defect significance. This feedback mechanism allows the system to accept approximations for speed while maintaining precision through threshold-based validation, identifying only significant defects that meet the criteria.
Solution Approach 2:
The patent changes the simulation parameter from full simulations to approximated simulations with Gaussian filters, achieving speed improvement. It compensates for potential precision loss by using intensity change calculations that focus on significant defects, maintaining adequate measurement precision for inspection purposes.
4Reliability
If full simulations are performed, then reliability is improved, but loss of energy increases
Solution Approach 1:
The patent applies partial action by performing only the necessary computational operations for reliable defect detection. Instead of complete simulations that consume excessive energy, it uses approximated wafer simulations with single convolution operations, reducing energy consumption while maintaining sufficient reliability through intensity change analysis.
Data Source
AI summary
To provide fast mask defect scoring, approximated wafer simulations (e.g. using one convolution) are performed on the defect inspection image and its corresponding reference inspection image. Using the approximated defect wafer image and the approximated reference wafer image generated by these approximated wafer simulations, a defect maximum intensity difference (MID) is computed by subtracting one approximated wafer image from the other approximated wafer image to generate a difference image. After a defect region of the difference image is clearly defined, a simulation at the centroid (i.e. a single point) of the defect region is performed. After the defect MID is computed (represented by an intensity) it can be compared to a prototype MID, which can represent a generic nuisance defect.