Wafer Image Alignment Using Deep Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect characterization techniques for crystalline wafers, such as silicon carbide (SiC), face challenges in accurately aligning images from different imaging sources, leading to difficulties in combining information for high-resolution analysis and spatial coordination, which is crucial for improving crystal growth and device yield.
Innovation Solution
The method involves aligning wafer images by identifying common crystalline features, determining spatial differences, and applying transformation matrices to achieve precise alignment, allowing for the formation of combined high-resolution images and enabling nondestructive defect characterization using deep convolutional neural networks (DCNNs).
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If images from different imaging sources are aligned using conventional techniques, then spatial coordination between images is achieved, but alignment accuracy and resolution are insufficient
Solution Approach 1:
The patent uses deep convolutional neural networks (DCNNs) as an intermediary to automatically identify and match crystalline features across different wafer images. The DCNN learns to detect common crystalline features and establish correspondences between images, replacing manual alignment methods and achieving sub-pixel alignment accuracy while reducing process complexity through automation.
Solution Approach 2:
The patent transforms wafer images through coordinate transformations and spatial adjustments to achieve precise alignment. By changing the spatial parameters of images (position, orientation, scale) based on feature correspondences identified by the DCNN, the system achieves high-precision alignment that preserves crystalline feature integrity across different imaging sources.
2Loss of information
If multiple imaging sources are used to characterize defects, then comprehensive defect information is obtained, but spatial coordination and image integration become difficult
Solution Approach 1:
The patent merges multiple wafer images from different imaging sources into a single integrated representation by aligning them through common crystalline features. The DCNN identifies corresponding features across images and combines the defect information spatially, creating a comprehensive defect map that preserves all characterization data while eliminating the complexity of managing separate images.
Solution Approach 2:
The system creates transformed copies of original wafer images that are spatially aligned through the DCNN-based feature matching. These copied and transformed images maintain the defect information from different imaging sources while being coordinated in a common reference frame, enabling easy integration and comprehensive analysis.
3Manufacturing precision
If wafer images are aligned with high precision, then defect characterization accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary feature extraction and matching using the DCNN before detailed defect characterization. The neural network pre-identifies common crystalline features and establishes initial alignment, which significantly reduces the computational burden and processing time for subsequent high-precision defect analysis while maintaining overall accuracy.
Solution Approach 2:
The patent replaces traditional mechanical and manual alignment methods with DCNN-based automated feature recognition and matching. This substitution of intelligent algorithms for conventional processing techniques achieves high-precision alignment faster and with fewer computational resources than traditional image registration methods.
Data Source
AI summary
Wafer images and related alignment methods for crystalline wafers are disclosed. Certain aspects relate to accessing and aligning images of a same or similar crystalline wafer captured from different imaging sources. Alignment may include determining spatial differences between shared crystalline features in various wafer images of the same or similar crystalline wafer and transforming at least one of the images according to the determined spatial differences. With sufficient alignment, information may be associated and/or transferred between the various images, thereby providing the capability of forming a combined wafer image and sub-images thereof with high resolution and spatial coordination between different image sources. Certain aspects relate to development of nondestructive, high fidelity defect characterization and/or dislocation counting methods in crystalline materials based on modern deep convolutional neural networks (DCNN).