Wafer Alignment Mark Positioning Without Template Re-Registration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor wafer inspection, the accuracy of alignment using pre-registered template images deteriorates when viewing conditions change, requiring frequent re-registration and increasing labor and time, and existing methods limit pattern position accuracy due to assumptions about point-symmetric patterns and low magnification alignment.
Innovation Solution
A position specification method that calculates the representative position of alignment marks on semiconductor wafers using a computer system without pre-registered template images, by setting partial image regions, dividing them into reference and searching target images, and performing template matching to determine the mark's presence and position, allowing for stable alignment even under different viewing conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If pre-registered template images are used for alignment, then alignment can be performed efficiently, but alignment accuracy deteriorates when viewing conditions change between steps
Solution Approach 1:
The patent extracts and stores reference image data (including pattern shape, size, and positional relationship information) from captured images before inspection. This preliminary extraction of essential alignment features allows the system to perform accurate alignment without requiring pre-registered template images for each step, thus maintaining both efficiency and accuracy across different viewing conditions.
2Measurement precision
If template images are re-registered for each process step, then alignment accuracy is maintained, but labor and time increase
Solution Approach 1:
The patent creates a universal reference image data structure that can be applied across multiple process steps and inspection apparatus. By extracting essential pattern information (shape, size, positional relationships) that remains consistent across different magnifications and viewing conditions, the system eliminates the need for step-specific template re-registration, reducing both time and labor while maintaining accuracy.
3Productivity
If alignment is performed at low magnification first, then high magnification alignment can be performed, but low magnification alignment accuracy limits final precision
Solution Approach 1:
The patent segments the alignment process into two distinct phases: rough alignment using captured images at any magnification level, and precise alignment using the extracted reference image data. This segmentation allows the system to first locate the target area efficiently and then achieve high-precision alignment independently of the initial magnification, overcoming the limitation of low-magnification alignment accuracy.
4Measurement precision
If the field of vision deviates due to low alignment accuracy, then the desired inspection area cannot be viewed, but increasing magnification amplifies alignment errors
Solution Approach 1:
The patent implements a feedback mechanism where the system detects the actual position and orientation of patterns in captured images, compares them with the stored reference image data, and automatically adjusts the field of vision and alignment parameters. This closed-loop feedback ensures that the desired inspection area is correctly positioned regardless of initial alignment accuracy or magnification level, eliminating the need for complex manual adjustment procedures.
Data Source
AI summary
The present invention specifies the position of a mark formed on a semiconductor wafer from an image of the semiconductor wafer. A mark 201 includes at least a first pattern 501 and a second pattern 502. The first and second patterns have an identical shape of line symmetry with respect to a first axis and a second axis that are orthogonal to each other, and are arranged so as to be point symmetric. A position specification method according to the present disclosure includes: a first procedure of setting a partial image region 301 to an image 200; a second procedure of dividing the partial image region on the basis of a division index 401, and setting a part of a region of a first divided image as a reference image 402; a third procedure of determining whether a mark exists in the partial image region; and a fourth procedure of calculating, when it is determined in the third procedure that a mark exists in the partial image region, a representative position of the mark in the partial image region.