Charged Particle Beam Apparatus Machine Learning Position Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current charged particle beam apparatuses for TEM observation are limited by slow automated micro-sampling speeds due to the need for high-definition template images, which are time-consuming to obtain and prone to precision issues caused by sample changes and individual differences.
Innovation Solution
A charged particle beam apparatus utilizing machine learning to detect target positions, eliminating the need for high-definition images and enabling faster automated micro-sampling by using a machine learning model for position control, allowing for improved throughput and adaptability to sample changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If template matching is used for position control, then measurement precision is improved, but loss of time increases due to needing to obtain high-definition images
Solution Approach 1:
The system performs preliminary action by obtaining and storing template images in advance at multiple positions before actual position control is needed. This allows the system to quickly retrieve pre-prepared templates during automated micro-sampling without spending time acquiring high-definition images in real-time, thus resolving the contradiction between measurement precision and time loss.
2Measurement precision
If scanning speed is slowed to obtain high-definition images, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The system prepares template images in advance at multiple positions before the actual sampling process. By having templates pre-acquired and stored, the system can perform position control using these pre-prepared images rather than acquiring high-definition images during the sampling process itself, thereby maintaining image quality while improving productivity.
Solution Approach 2:
The system dynamically adjusts the scanning speed based on the operational phase: slower scanning speeds are used when acquiring template images in advance to ensure high definition, while faster scanning speeds are used during the actual automated micro-sampling process. This dynamic adjustment allows the system to maintain measurement precision while improving overall productivity.
3Measurement precision
If template images are obtained each time for position control, then measurement precision is maintained, but productivity decreases due to repeated image acquisition
Solution Approach 1:
The system obtains template images in advance at multiple positions before the actual sampling process and stores them for reuse. This preliminary preparation eliminates the need to acquire template images repeatedly during each sampling operation, thereby maintaining position control accuracy while significantly improving automated micro-sampling throughput.
Solution Approach 2:
The system creates and stores copies of template images at multiple positions in advance. These copied templates can be quickly retrieved and used for position control during automated micro-sampling without requiring repeated image acquisition, thus maintaining measurement precision while enhancing productivity.
Data Source
AI summary
To accomplish fast automated micro-sampling, provided is a charged particle beam apparatus, which is configured to automatically fabricate a sample piece from a sample, the charged particle beam apparatus including: a charged particle beam irradiation optical system configured to radiate a charged particle beam; a sample stage configured to move the sample that is placed on the sample stage; a sample piece transportation unit configured to hold and convey the sample piece separated and extracted from the sample; a holder fixing base configured to hold a sample piece holder to which the sample piece is transported; and a computer configured to perform position control with respect to a second target, based on a machine learning model in which first information including a first image of a first target is learned, and on second information including a second image, which is obtained by irradiation with the charged particle beam.