Semiconductor Inspection Template Matching with Multiple Likelihoods
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Template matching in semiconductor inspection fails when patterns are not visible, have low contrast, or are deformed, making it difficult to accurately align and measure semiconductor patterns, especially when patterns are covered by upper layers or antireflective films.
Innovation Solution
An inspection apparatus that performs template matching using multiple templates to calculate single and multiple template likelihoods, selecting the highest assimilation likelihood matching candidate to determine accurate alignment positions, even in cases where patterns are not clearly visible or have low contrast, by incorporating statistical correlation values and assimilation likelihood calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If template matching is performed using a single template, then the matching process is simple and fast, but the matching accuracy deteriorates when patterns are not visible, have low contrast, or are deformed
Solution Approach 1:
The patent divides the template matching process into multiple independent stages: first template matching, then second template matching, and finally third template matching. Each stage uses a different template (first template, second template, third template) to progressively refine the matching result. This segmentation allows the system to maintain simplicity in each individual matching step while achieving high overall accuracy through the combination of multiple matching results.
2Reliability
If multiple templates are used for template matching, then the robustness and accuracy of matching improve, but the computational complexity and processing time increase
Solution Approach 1:
The patent segments the template matching into three distinct stages, each handling specific aspects of pattern recognition. The first template matching handles initial position identification, the second template matching refines the alignment, and the third template matching confirms the final position. This segmentation reduces the computational burden of each individual stage while maintaining overall robustness.
Solution Approach 2:
The patent performs preliminary template matching using the first template before conducting more complex matching operations. This preliminary action establishes a baseline matching result that guides subsequent matching steps, reducing the search space and computational complexity of later stages while improving overall reliability.
3Loss of time
If a single template is used for alignment, then the alignment process is fast, but the alignment accuracy deteriorates when patterns are covered by upper layers or antireflective films
Solution Approach 1:
The patent performs preliminary alignment using the first template to establish an initial position estimate. This preliminary action is quickly executed to reduce time loss, while subsequent second and third template matching steps refine the alignment accuracy for patterns that may be covered or have low contrast, thereby resolving the contradiction between speed and precision.
Solution Approach 2:
The patent introduces intermediate template matching steps as mediators between the initial single-template alignment and the final precise alignment. The second and third templates act as intermediaries that progressively refine the alignment result, allowing the system to maintain fast initial alignment while achieving high final accuracy for difficult-to-detect patterns.
Data Source
AI summary
To provide means according to which template matching is performed successfully in an apparatus that performs inspection or measurement of a semiconductor pattern, which is formed on a wafer, even in a case in which a pattern for alignment, which is in design data, has been eliminated from an image in which an image of a practical pattern was captured using the apparatus, in which the brightness value contrast of a pattern is lower (more indistinct) than that of other locations, or in which a pattern is deformed and there is a discrepancy with the shape of a template (a pattern for alignment in the design data). An inspection apparatus according to the invention acquires a target retrieval image, and carries out template matching on the target retrieval image, and includes template input, means for inputting a plurality of templates, a plurality of matching candidate selection sections that select a matching candidate group by performing a matching process of the target retrieval image and the plurality of templates, a plurality of single template likelihood calculation process sections that calculate single template likelihoods for a plurality of matching candidate groups that are selected by the plurality of matching candidate selection sections, a multiple template assimilation likelihood calculation process section that calculates a multiple template assimilation likelihood for the matching candidate groups using a plurality of single template likelihoods that are calculated by the plurality of single template likelihood calculation process sections, and a highest assimilation likelihood matching candidate selection section that selects a matching candidate for which the multiple template assimilation on likelihood is the highest, from among the matching candidate groups, using the multiple template assimilation likelihoods that are calculated by the multiple template assimilation likelihood calculation process section.