Mask Weak Pattern Recognition via Image Overlay and ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor manufacturing, the effects of optical diffraction lead to unpredictable risks in mask layouts, requiring costly pilot runs to identify and modify risk patterns, necessitating a method to predict risks without actual experiments.
Innovation Solution
A mask weak pattern recognition apparatus and method using machine learning to generate a recognition model by receiving and overlapping mask layouts and inspection images, analyzing risk patterns and scores, and training a model to predict risk scores without actual experiments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a mask layout draft is piloted run to obtain risk patterns, then the risk patterns can be identified and modified, but the cost of the pilot run is quite high
Solution Approach 1:
The patent creates a virtual copy of the mask layout through inspection image generation and overlay processing. By comparing the designed mask layout with the generated inspection image, the system identifies risk patterns without requiring physical pilot runs, thus eliminating the high costs associated with actual manufacturing trials while maintaining accurate risk detection
Solution Approach 2:
The patent replaces the mechanical/physical pilot run process with an automated image processing and analysis system. Instead of physically manufacturing test masks and inspecting them, the system uses computational methods to generate inspection images, overlay them with design layouts, and automatically identify risk patterns through coordinate analysis and pattern recognition algorithms
2Reliability
If conventional methods are used to predict mask layout risks, then actual experiments are required, but this reduces efficiency and increases time consumption
Solution Approach 1:
The patent performs preliminary risk assessment by generating inspection images and identifying risk patterns before actual manufacturing occurs. The system pre-processes mask layout data, generates corresponding inspection images, and identifies potential risk patterns in advance, allowing for early modifications without delaying the manufacturing schedule
Solution Approach 2:
The patent substitutes time-consuming physical experiments with automated computational analysis. The system rapidly generates inspection images, performs overlay processing, and identifies risk patterns through algorithmic analysis, reducing the risk assessment time from days or weeks of physical piloting to minutes or hours of computational processing
Data Source
AI summary
A mask weak pattern recognition apparatus and a mask weak pattern recognition method are provided. The mask weak pattern recognition apparatus includes a receiving unit, an overlapping unit, an analyzing unit and a training unit. The receiving unit is used for receiving a mask layout and an inspection image of a mask. The overlapping unit is used for overlapping the mask layout and the inspection image to obtain an overlapped image. The analyzing unit is used for obtaining a plurality of risk patterns and a plurality of risk scores each of which corresponds one of the risk patterns according to the overlapped image. The training unit is used for training a recognition model according to the risk patterns and the risk scores.