Semiconductor Template Library Creation for Inspection Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for creating templates for template matching in semiconductor inspection are either time-consuming due to simulation adjustments or require labor-intensive SEM image acquisition, especially when dealing with varying magnifications and scanner variations, leading to inefficiencies and potential misrecognition.
Innovation Solution
A method and device that store pattern information as a library for each basic circuit, allowing for efficient template creation by extracting and designating positions, and using a template library for high-accuracy and flexible pattern recognition across different magnifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If simulation or image processing is used to create a template close to actual SEM images, then template accuracy is improved, but time consumption increases due to adjusting simulation conditions or performing simulation
Solution Approach 1:
The patent pre-acquires SEM images of actual semiconductor patterns and stores them as template data before they are needed for matching operations. This preliminary action eliminates the need for time-consuming simulation adjustments when templates are required, while still providing high accuracy since the templates are based on real SEM images rather than simulated ones.
Solution Approach 2:
The patent creates templates by copying actual SEM images of semiconductor patterns directly into a database. Instead of generating templates through simulation, the system makes copies of real pattern images and stores them for later retrieval and matching, thereby achieving both accuracy (from real images) and efficiency (from direct copying).
2Measurement precision
If SEM images are obtained to create templates faithfully representing actual images, then template accuracy is improved, but labor requirement increases due to needing to obtain images through SEM and derive them whenever magnification varies
Solution Approach 1:
The patent creates templates at multiple magnifications in advance and stores them in a database with metadata indicating the magnification level. When a template is needed for a specific magnification, the system retrieves the pre-prepared template rather than deriving it from an SEM image at that moment. This universal approach allows the same template database to serve multiple magnification requirements without additional labor.
Solution Approach 2:
The patent performs the labor-intensive task of obtaining and processing SEM images in advance, creating and storing templates for various magnifications before they are needed. This preliminary preparation eliminates the need to repeat the laborious process of obtaining and deriving images whenever magnification changes, significantly reducing operational labor while maintaining template fidelity.
3Measurement precision
If templates are created for each specific magnification and scanner variation, then matching accuracy is improved, but device complexity increases
Solution Approach 1:
The patent introduces a database as an intermediary between the template creation process and the template matching operation. The database stores templates with associated metadata including magnification level and scanner type information. This intermediary structure allows the system to manage multiple templates for different conditions without increasing operational complexity, as the database automatically handles the selection and retrieval of appropriate templates based on the matching requirements.
Data Source
AI summary
Disclosed is a method wherein a template for template matching is created with high accuracy and high efficiency. With respect to each individual pattern constituting a basic circuit, pattern information regarding a plurality of layers in a semiconductor device is stored in a library. On the basis of the designation of the position and the layer, pattern information regarding the designated position and layer is extracted from the pattern information stored in the library. A template is created on the basis of the extracted pattern information.