Semiconductor Defect Classification Using Multi-Label Descriptors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for examining semiconductor specimens during fabrication lack efficient automation in defect classification, particularly in distinguishing between different defect classes with high precision and uniformity, which is crucial for ultra-large scale integration and high-performance devices.
Innovation Solution
A machine learning-based approach using processing and memory circuitry (PMC) to generate multi-label descriptors from defect images, employing trained machine learning models to classify defects into multiple classes by matching generated descriptors with multi-label data sets, enabling classification and recognition of new classes based on physical attributes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine learning models are used to automate defect classification, then productivity and classification speed are improved, but device complexity increases due to the need for training data processing and model deployment infrastructure
Solution Approach 1:
The patent introduces multi-label descriptors as an intermediary representation layer between raw defect images and classification decisions. These descriptors capture physical attributes (shape, size, orientation, texture) in a structured format that simplifies the classification task while maintaining high accuracy, thereby improving productivity without proportionally increasing system complexity
Solution Approach 2:
The classification process is segmented into distinct stages: (1) generating multi-label descriptors from defect images using trained machine learning models, and (2) matching these descriptors against reference databases. This segmentation allows each stage to be optimized independently, improving overall productivity while managing complexity through modular architecture
2Measurement precision
If multiple physical attributes are used to distinguish defect classes, then measurement precision and classification accuracy are improved, but device complexity increases due to the need to process and analyze multiple attributes simultaneously
Solution Approach 1:
The patent transforms raw defect image data into a standardized multi-label descriptor format that represents multiple physical attributes (shape, size, orientation, texture) in a consistent parameter space. This parameter transformation enables accurate classification by multiple attributes without proportionally increasing processing complexity, as the descriptors provide a unified representation framework
Solution Approach 2:
The multi-label descriptor structure serves multiple functions simultaneously: it encodes multiple physical attributes, enables matching against reference databases, and provides a standardized interface for different defect types. This multi-functionality allows the system to handle diverse defect classification requirements without increasing complexity for each additional attribute
Data Source
AI summary
There is provided a method of automated defects' classification, and a system thereof. The method comprises obtaining data informative of a set of defects' physical attributes usable to distinguish between defects of different classes among the plurality of classes; training a first machine learning model to generate, for the given defect, a multi-label output vector informative of values of the physical attributes, thereby generating for the given defect a multi-label descriptor; and using the trained first machine learning model to generate multi-label descriptors of the defects in the specimen. The method can further comprise obtaining data informative of multi-label data sets, each data set being uniquely indicative of a respective class of the plurality of classes and comprising a unique set of values of the physical attributes; and classifying defects in the specimen by matching respectively generated multi-label descriptors of the defects to the multi-label data sets.