ML Defect Detection in Semiconductor Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor fabrication processes face challenges in accurately and efficiently detecting defects on semiconductor wafers, particularly at the submicron level, due to the need for high precision and uniformity.
Innovation Solution
A system utilizing one or more processing circuitries and a machine learning algorithm to analyze inspection images of semiconductor specimens. The system determines parameters of a model informative of pixel intensity distribution, allowing for the detection of defects by comparing measured pixel intensities with expected values in the absence of defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional defect detection methods are used, then the examination process can be performed, but the measurement precision and reliability are insufficient for submicron features
Solution Approach 1:
The patent replaces traditional mechanical/optical inspection systems with a machine learning-based automated inspection system. The ML algorithm analyzes inspection images to detect and classify defects, substituting manual or conventional automated optical inspection methods. This enables higher precision detection of submicron features through pattern recognition and statistical analysis of pixel intensity distributions.
Solution Approach 2:
The patent changes the parameters used for defect detection by analyzing pixel intensity distributions and their statistical properties rather than relying on fixed threshold values. The system calculates parameters such as mean, standard deviation, and skewness of pixel intensities to identify defects, adapting to varying noise conditions and improving detection reliability across different inspection scenarios.
2Productivity
If automated examination is implemented, then productivity increases, but the complexity of the inspection system increases
Solution Approach 1:
The patent creates a universal inspection system that performs multiple functions: defect detection, defect classification, noise differentiation, and parameter estimation all through a single machine learning framework. The same ML model processes inspection images to achieve various inspection objectives, reducing the need for multiple specialized systems and simplifying the overall inspection infrastructure.
Solution Approach 2:
The inspection system uses the inspection images themselves to train and adapt the machine learning model, enabling the system to self-improve and adjust to different inspection conditions without external intervention. The ML algorithm automatically learns optimal detection parameters from the data, reducing the need for manual calibration and system configuration.
3Measurement precision
If noise differentiation is implemented, then measurement precision improves, but the computational requirements increase
Solution Approach 1:
The patent applies partial action by focusing computational resources only on regions of interest within the inspection images where defects are likely to occur. The machine learning model processes pixel intensity data selectively, analyzing only relevant features and parameters rather than performing exhaustive computations on the entire image, thereby reducing overall computational energy consumption while maintaining high precision in defect identification.
Data Source
AI summary
There are provided systems and methods comprising obtaining a first inspection image informative of a first area of a specimen acquired by an examination tool, feeding at least the first inspection image to a machine learning algorithm configured to determine, for each given pixel of a plurality of pixels of the first inspection image, or for each given group of pixels of a plurality of groups of pixels of the first inspection image, one or more given parameters of a given model informative of pixel intensity distribution, for said each given pixel or given group of pixels, using at least some of the one or more given parameters, or the given model associated with the one or more given parameters, and measured pixel intensity of the given pixel or group of pixels, to determine whether a defect is present in the given pixel or in the given group of pixels.