Spectrum Image Defect Detection Model for Semiconductor Substrates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As the integration of semiconductor processes increases and line widths become finer, the detection and prediction of defects in semiconductor devices become more complex, requiring effective techniques to identify and classify defects in a substrate based on spectrum images.
Innovation Solution
A defect detection method and device that utilize a spectral microscopic inspection system to radiate light onto a substrate, generate spectrum images, perform an EDS test, create a defect map, and train a defect detection model to predict defect grades, allowing for the classification and targeting of defects in various blocks of the substrate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the degree of integration of the semiconductor process increases and line width becomes finer, then the semiconductor device performance is improved, but the types and amounts of defects increase exponentially
Solution Approach 1:
The substrate is divided into multiple blocks, and spectrum images are obtained for each block separately. This segmentation allows for localized defect analysis and grading, enabling the system to handle the increased defect complexity in fine-line width devices by processing each block independently with the trained defect detection model.
Solution Approach 2:
A defect detection model is trained in advance using spectrum image information and defect grades obtained from EDS tests. This preliminary training enables the model to predict defects in new substrates before final inspection, allowing for early defect identification and reducing the impact of exponential defect increases in high-integration devices.
2Device complexity
If traditional defect detection methods are used, then the detection process is simple, but the ability to predict and classify defects in fine-line width devices is insufficient
Solution Approach 1:
A defect detection model serves as an intermediary between spectrum image acquisition and defect classification. The model is trained using spectrum image information and EDS test results, then used to predict defect grades in target blocks. This intermediary enables accurate defect prediction and classification without requiring complex manual inspection processes.
Solution Approach 2:
The traditional mechanical EDS testing process is supplemented with an optical-based spectrum image analysis system. The defect detection model processes spectrum images to predict defect grades, replacing or augmenting the need for extensive physical testing and improving detection accuracy for fine-line width devices.
3Area of stationary object
If spectrum images are obtained for the whole substrate, then the inspection coverage is complete, but the detailed defect analysis for each block is insufficient
Solution Approach 1:
The substrate is divided into multiple blocks, and spectrum images are obtained for each block separately rather than treating the whole substrate as a single unit. This segmentation enables detailed defect analysis at the block level while maintaining complete substrate coverage, as each block's spectrum image is processed individually through the defect detection model.
Solution Approach 2:
Different blocks of the substrate are analyzed with localized defect detection. The defect detection model is applied to each block's spectrum image to determine specific defect grades for each block, allowing for localized quality assessment rather than a single overall assessment of the entire substrate.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables accurate prediction and classification of defects, improving the reliability and stability of semiconductor devices by effectively identifying and addressing defects during the manufacturing process.
Implementation Method 1
obtaining a spectrum image indicating an amount of the light according to a wavelength from reflected light reflected from the substrate
Data Source
AI summary
A defect detection method includes radiating light onto a substrate, obtaining a spectrum image, performing an electrical die sorting (EDS) test on the substrate, inspecting defects of each of a plurality of blocks of the substrate based on a result of the EDS test, generating a defect map, generating spectrum image information by matching the spectrum image with the defect map, training a defect detection model by using the defect grade as an output value and the spectrum image information as an input value, obtaining a target spectrum image with respect to a target substrate, extracting a feature vector from the target spectrum image by using the defect detection model, and detecting a target defect grade of the target spectrum image based on the feature vector, and generating a target defect map based on the target defect grade.