SEM Image Alignment via Mask Fingerprint Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In the manufacturing of integrated circuits, aligning multiple SEM images of complex patterns without landmarks is challenging due to uncertainties in pattern placement and sample positioning, which affects defect detection and inspection accuracy.
Innovation Solution
A method is developed to determine offsets between SEM images by detecting the 'fingerprint' of the mask pattern in noise, using contour extraction and iterative correlation maximization to align images of parallel lines, thereby aligning images to the mask and each other.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If landmarks are added to align images of straight line patterns, then image alignment accuracy is improved, but device complexity and manufacturing complexity increase
Solution Approach 1:
The invention extracts and removes the mask fingerprint noise from the SEM images through signal processing techniques. By identifying and eliminating this systematic noise component, the method achieves accurate image alignment without requiring additional landmarks, thus solving the contradiction between alignment accuracy and pattern complexity
Solution Approach 2:
The invention introduces an intermediary processing step that analyzes the spectral content of SEM images to identify mask fingerprint signatures. This intermediary analysis layer enables the system to distinguish between actual pattern features and mask-induced noise, allowing accurate alignment without modifying the original pattern design
2Measurement precision
If multiple SEM images are acquired to improve alignment accuracy, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The invention enables images to self-align by detecting and utilizing the inherent mask fingerprint patterns that naturally appear in SEM images. The mask fingerprint serves as an automatic reference that guides the alignment process without requiring multiple images or complex external referencing, thus maintaining high productivity while achieving accurate alignment
Solution Approach 2:
The invention transforms the alignment approach by changing from spatial referencing (using landmarks or multiple images) to spectral analysis (detecting mask fingerprint signatures in the frequency domain). This parameter change allows single-image or fewer-image alignment with high accuracy, improving inspection throughput
3Ease of manufacture
If image alignment is performed without landmarks using straight line patterns, then ease of manufacture is improved, but measurement precision deteriorates
Solution Approach 1:
The invention replaces the mechanical/physical landmark-based alignment system with a signal processing-based system. By substituting physical landmarks with digital detection and analysis of mask fingerprint patterns, the method achieves accurate alignment while maintaining simple straight line patterns that are easy to manufacture
Data Source
AI summary
A method of determining offsets between a plurality of data sets, each data set representing a sampling area of a pattern formed on a sample, wherein each sampling area derives from a predetermined portion of a mask pattern, the method comprising: detecting a fingerprint of the mask pattern in noise of the data sets; and determining offsets based on the fingerprint of the mask pattern.