Hybrid Machine Learning Defect Detection for Semiconductor Specimens
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect detection methods in semiconductor fabrication lack sensitivity and specificity, particularly in identifying defects at submicron scales, leading to potential issues in high-density and high-performance device production.
Innovation Solution
A method and system utilizing a combination of supervised and unsupervised machine learning models, trained on specific image datasets, to process runtime images of semiconductor specimens, generating defect maps with optimized parameters for improved defect detection, integrating both low-resolution and high-resolution data for comprehensive analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional defect detection methods are used, then the examination process is simple, but the detection sensitivity and specificity are insufficient for submicron defects
Solution Approach 1:
The patent combines supervised and unsupervised machine learning models into a hybrid examination system. The supervised model (first machine learning model) processes images to identify known defect patterns, while the unsupervised model (second machine learning model) detects anomalies and novel defect types. The outputs of both models are integrated to produce a composite defect map, achieving superior detection sensitivity and specificity compared to traditional single-method approaches.
Solution Approach 2:
The examination system is segmented into distinct functional components: an image acquisition module that captures specimen images, a supervised learning module that identifies known defect patterns, an unsupervised learning module that detects anomalies, and a fusion module that combines results. This segmentation allows each component to be optimized independently while maintaining overall system manageability.
2Measurement precision
If high-resolution examination is performed to detect submicron defects, then the detection accuracy improves, but the examination time increases
Solution Approach 1:
The system applies partial high-resolution examination by using the unsupervised learning model to first identify candidate defect regions in the image. Only these candidate regions are then subjected to detailed analysis by the supervised learning model. This selective approach achieves high detection accuracy for submicron defects while avoiding the need to process entire high-resolution images, thereby reducing examination time.
3Reliability
If multiple examination phases are used to improve defect detection, then the detection reliability improves, but the processing time increases
Solution Approach 1:
The hybrid machine learning system operates continuously and simultaneously, with both the supervised and unsupervised models processing images in parallel. The models are trained on comprehensive datasets that include various defect types and conditions, enabling them to reliably detect defects without requiring multiple sequential examination phases. This continuous operation maintains high detection reliability while maximizing examination throughput.
Data Source
AI summary
There is provided a method of defect detection on a specimen and a system thereof. The method includes: obtaining a runtime image representative of at least a portion of the specimen; processing the runtime image using a supervised model to obtain a first output indicative of the estimated presence of first defects on the runtime image; processing the runtime image using an unsupervised model component to obtain a second output indicative of the estimated presence of second defects on the runtime image; and combining the first output and the second output using one or more optimized parameters to obtain a defect detection result of the specimen.