Imprint Defect Detection Using Multi-Illumination Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing imprint techniques face challenges in accurately detecting pattern formation defects such as protrusion and unfilling on substrates, which require manual inspection of numerous high-powered microscope images, making it inefficient and prone to errors due to variations in film thickness and brightness.
Innovation Solution
An information processing apparatus that generates learning models corresponding to different illumination conditions, allowing for automated detection of pattern formation defects by inputting images from substrates under various conditions to determine the presence or absence of defects using a combination of image capturing and machine learning techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual inspection of observation images is performed to detect protrusion and unfilling, then detection accuracy can be maintained, but inspection efficiency deteriorates due to the huge number of images requiring review
Solution Approach 1:
The patent replaces manual mechanical inspection with an automated image processing system that captures observation images and automatically analyzes them to detect protrusion and unfilling defects. This substitution eliminates the need for human reviewers to manually examine each image while maintaining detection accuracy through systematic automated analysis.
Solution Approach 2:
The system enables self-inspection by automatically processing observation images without human intervention. The automated analysis system performs defect detection independently, allowing the inspection process to serve itself without requiring manual review of the huge number of captured images.
2Measurement precision
If high-powered microscope is used to capture observation images for defect detection, then detection capability is improved, but the observation range becomes narrow requiring even more images to be checked
Solution Approach 1:
The patent divides the inspection task into multiple captured images that collectively cover the entire substrate area. By segmenting the large-area substrate into smaller observable regions and capturing multiple images, the system achieves both high detection capability for each region and complete coverage of the entire substrate without requiring a single wide-angle view.
Solution Approach 2:
The system transitions from a two-dimensional observation plane to a three-dimensional inspection approach by capturing multiple images at different positions and combining them. This dimensional expansion allows comprehensive coverage of the entire substrate area while maintaining high magnification for defect detection in each captured region.
3Reliability
If film thickness varies on the substrate, then the brightness of observation images changes making defect detection difficult, but this is a natural consequence of the imprint process
Solution Approach 1:
The patent changes the illumination parameters by capturing observation images under multiple different illumination conditions. This parameter variation allows the system to obtain consistent defect detection results despite variations in film thickness, as different illumination angles and intensities compensate for the brightness changes caused by thickness variations.
Solution Approach 2:
The system performs preliminary image capture under multiple illumination conditions before final defect determination. By预先 capturing images under various lighting conditions, the system prepares multiple data sets that can be analyzed to identify defects regardless of the brightness variations caused by film thickness differences.
Data Source
AI summary
The present invention provides an information processing apparatus for determining, for a substrate having undergone an imprint process that forms a pattern of an imprint material, presence/absence of a formation defect of the pattern, including a generating unit configured to generate learning models in one-to-one correspondence with a plurality of different illumination conditions by using images obtained by capturing images of a plurality of substrates on which the pattern is formed under the plurality of illumination conditions, and an obtaining unit configured to input the images obtained by capturing images of the substrates having undergone the imprint process and containing the pattern, to a plurality of learning models generated by the generating unit, and obtain a temporary determination result temporarily indicating the presence/absence of a formation defect of the pattern from each of the plurality of learning models.