Wafer Alignment via Image Projection Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Patch-to-design alignment (PDA) in semiconductor manufacturing is negatively impacted by process variations, leading to poor alignment accuracy, especially in low-contrast images or images difficult to align, which can result in incorrect processing decisions and reduced yield.
Innovation Solution
The method involves aligning setup and runtime images using a processor to determine normalized cross-correlation scores and image projections in perpendicular directions, adjusting the images to overlap peak locations, and determining offsets for accurate placement of care areas, thereby improving alignment stability and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional patch-to-design alignment is used, then the alignment process is simple, but alignment accuracy deteriorates due to process variation and low image contrast
Solution Approach 1:
The alignment process is segmented into multiple stages: initial patch-to-design alignment, normalized cross-correlation scoring, and conditional projection-based refinement. This segmentation allows the system to apply complex processing only when needed (when NCC score is below threshold), thereby improving alignment accuracy while controlling overall process complexity.
Solution Approach 2:
The patent performs preliminary alignment using traditional PDA methods to establish an initial alignment state before evaluating image quality metrics. This preliminary action provides a baseline that enables subsequent conditional refinement only when necessary, balancing simplicity and accuracy.
2Measurement precision
If image projection alignment is always applied, then alignment accuracy improves, but processing time increases
Solution Approach 1:
The patent applies projection-based alignment partially - only when the normalized cross-correlation score falls below a predetermined threshold. This partial application of the alignment refinement process avoids unnecessary processing time for images that already have good contrast and alignment quality, while still improving accuracy for challenging cases.
Solution Approach 2:
The alignment process is made dynamic by conditionally adjusting the processing path based on image quality assessment. The system adapts its complexity in real-time, applying full projection-based refinement only when image characteristics warrant it, thereby optimizing the balance between accuracy and processing time.
3Productivity
If traditional alignment methods are used, then processing speed is maintained, but alignment stability deteriorates leading to target failure
Solution Approach 1:
The patent implements feedback through normalized cross-correlation scoring to assess alignment quality. This feedback mechanism identifies cases where traditional alignment has failed or is insufficient, triggering projection-based refinement only when needed. This maintains high processing speed for good cases while improving reliability for problematic cases.
Solution Approach 2:
The patent replaces purely mechanical/image-based alignment with a hybrid approach that incorporates signal processing techniques (projection and correlation analysis). This substitution enhances alignment stability by using mathematical transformations that are more robust to process variation and low contrast conditions.
4Measurement precision
If projection-based alignment refinement is applied, then alignment accuracy improves for low-contrast images, but computational complexity increases
Solution Approach 1:
The patent applies computationally intensive projection-based alignment refinement only partially - specifically when the normalized cross-correlation score indicates poor image quality. This selective application reduces overall computational complexity while maintaining high alignment accuracy for the subset of images that need it.
Solution Approach 2:
The computational process is segmented into a fast initial alignment stage and a conditional refinement stage. This segmentation allows the system to maintain low computational complexity for most cases while providing high accuracy refinement when image quality metrics indicate it is necessary.
Data Source
AI summary
Image alignment or image-to-design alignment can be improved using normalized cross-correlation. A setup image to a runtime image are aligned and a normalized cross-correlation scores is determined. Image projections for the images can be determined and aligned in the perpendicular x and y directions. Alignment of the image projections can include finding projection peak locations and adjusting the projection peak locations in the x and y directions.