SEM Pattern Matching Using Design-Guided Edge Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pattern matching technologies struggle to accurately select edge candidates in SEM images with weak edges, leading to unstable matching and prolonged processing times, especially in multilayer patterns, due to variations in semiconductor pattern configurations and measurement conditions.
Innovation Solution
A pattern matching apparatus and system that utilizes a computer system to acquire edge candidate groups based on design data and SEM images, calculate an association evaluation value for different combinations, and select the most appropriate association, thereby determining a matching shift amount for accurate positioning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If threshold processing is used to select edge candidates, then false edges are reduced, but true weak edges may be failed to be extracted
Solution Approach 1:
The patent changes the parameter selection criterion from edge intensity (threshold-based) to a combination of edge intensity and positional coincidence with design data. This allows weak edges that coincide with design patterns to be selected even when their intensity is low, while still filtering out false edges that do not match design positions.
Solution Approach 2:
The patent introduces design data as an intermediary to mediate between the SEM image edges and the final edge candidate selection. The design data serves as a reference to verify whether detected edges are true edges, allowing weak but valid edges to be retained while filtering false detections.
2Reliability
If all edge candidates are selected without threshold processing, then no true edges are missed, but processing time becomes long
Solution Approach 1:
The patent extracts only the necessary edge candidates by using design data to identify which edges are relevant. Instead of processing all detected edges, it selectively extracts edges that coincide with design pattern positions, significantly reducing the number of candidates for subsequent association processing.
Solution Approach 2:
The patent performs preliminary filtering of edge candidates using design data before the association processing stage. By pre-identifying which edges are likely to be true edges based on design pattern positions, it reduces the computational burden of the subsequent association processing.
3Reliability
If all edge candidates are selected without threshold processing, then no true edges are missed, but matching stability decreases
Solution Approach 1:
The patent changes the selection parameter from purely intensity-based to a composite parameter that includes both intensity and positional coincidence with design data. This provides more stable and reliable edge candidate selection by anchoring to the known design pattern positions, reducing variability in matching results.
Solution Approach 2:
The patent uses design data as an intermediary to stabilize the edge candidate selection process. The design data provides a stable reference framework that reduces the impact of image quality variations and noise, leading to more consistent and stable matching results across different measurements.
Data Source
AI summary
A pattern matching apparatus includes a computer system configured to execute pattern matching processing between first pattern data based on design data and second pattern data representing a captured image of an electron microscope. The computer system acquires a first edge candidate group including one or more first edge candidates, acquires a selection-required number (the number of second edge candidates to be selected based on the second pattern data), acquires a second edge candidate group including the second edge candidates of the selection-required number, acquires an association evaluation value for each of different association combinations between the first edge candidate group and the second edge candidate group, selects one of the combinations based on the association evaluation value, and calculates a matching shift amount based on the selected combination.