Electron Microscope Autofocus and Astigmatism Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current focus and astigmatism correction algorithms in electron microscopy are slow, require extensive tuning, and struggle with generalization to new settings, making them impractical for high-throughput imaging.
Innovation Solution
A data-driven, machine-learning based method for automatic focusing and astigmatism correction in scanning electron microscopy, which operates in low signal-to-noise conditions, reduces processing times significantly, and can be recalibrated for different microscopes and samples without expert knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If explicit physical models are used for focus and astigmatism correction, then generalization properties are improved, but processing time increases and parameter tuning complexity increases
Solution Approach 1:
The patent replaces explicit physical models and iterative optimization algorithms with a trained neural network model. The neural network directly predicts focus and astigmatism correction parameters from input images, eliminating the need for complex physical modeling and iterative computation, thereby achieving near-instant inference while maintaining generalization capability across different microscopes and samples
Solution Approach 2:
The patent performs preliminary training of the neural network model using synthetic training data generated from physical models before deployment. This preliminary action pre-computes the correction mappings, allowing the model to make rapid predictions during actual operation without performing time-consuming iterative optimization in real-time
2Adaptability or versatility
If explicit physical models are used for focus and astigmatism correction, then generalization properties are improved, but device complexity increases
Solution Approach 1:
The patent replaces complex physical modeling and manual parameter tuning with a trained neural network. The model learns the correction mappings during training and automatically applies them during inference, eliminating the need for users to understand or tune physical parameters, thereby reducing device complexity and ease of operation
Solution Approach 2:
The neural network model performs self-calibration by learning from synthetic training data that covers a wide range of conditions. Once trained, the model autonomously predicts correction parameters for new images without requiring manual intervention or expert knowledge, making the system self-sufficient and easy to operate
3Manufacturing precision
If manual parameter adjustments are used, then image quality is maintained, but productivity decreases
Solution Approach 1:
The system implements automated focus and astigmatism correction using a trained neural network that autonomously analyzes input images and predicts optimal correction parameters. This self-service capability eliminates the need for manual parameter adjustments by operators, maintaining high image quality while enabling high-throughput automated operation
Solution Approach 2:
The system uses the neural network to continuously predict correction parameters based on actual image content, creating a feedback loop that automatically adjusts focus and astigmatism settings. This closed-loop automation maintains image quality standards while dramatically increasing throughput by eliminating manual intervention
4Manufacturing precision
If existing correction algorithms are used, then focus and astigmatism correction is achieved, but execution time budget is exceeded
Solution Approach 1:
The patent replaces iterative optimization algorithms with a trained neural network that directly predicts correction parameters in a single forward pass. This substitution eliminates the need for repeated iterations and complex computations, achieving accurate focus and astigmatism correction in near-instant time that fits within tight execution time budgets
Data Source
AI summary
Method for automatic focusing and astigmatism correction for a specific microscope setup, in particular an electron microscope, the microscope being at least adjustable in microscope parameters a working distance, a stigmator in a x-direction and a stigmator in a y-direction, the method comprising the steps: a) capturing a first image of a sample with a first working distance perturbation and capturing a second image of the sample with a second working distance perturbation around a current working distance and current stigmator settings; b) selecting n subareas of the first image and n subareas of the second image, n≥1, wherein an i-th subarea of the first image and an i-th subarea of the second image form an i-th input patch pair, 1≤i≤n; c) processing each i-th input patch pair and receiving an i-th correction term comprising a correction to the current working distance, stigmator in x-direction and stigmator in y-direction; d) receiving an output correction term as a function of all the correction terms; e) adjusting the current working distance and stigmator settings by applying the output correction term to the current working distance and stigmator settings.