Charged Particle Microscopy with ML Atom Prediction at Low Dose
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Charged particle microscopy, particularly STEM, faces challenges in obtaining atomic resolution images efficiently due to limited electron dose budgets and the need for lengthy navigation to find interesting areas, which can result in sample damage.
Innovation Solution
The use of a trained machine learning model to predict atom structure probability from low signal-to-noise ratio (SNR) images acquired with charged particle microscopy, enabling live-assisted navigation and imaging with reduced electron dose, faster scanning, and sparse scanning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional navigation with high radiation dose is used to find interesting areas, then atomic resolution images can be obtained, but sample damage occurs
Solution Approach 1:
The system performs preliminary low-dose imaging and machine learning-based atom structure prediction before final high-resolution imaging. This preliminary action identifies regions of interest and predicts atomic structures without exposing the sample to high radiation doses, thereby preventing sample damage while maintaining measurement precision.
Solution Approach 2:
Instead of applying full high-dose imaging across the entire sample during navigation, the system uses partial low-dose imaging only in regions being actively explored, combined with machine learning predictions for areas not yet imaged. This reduces overall radiation exposure while still enabling atomic resolution measurements where needed.
2Measurement precision
If lengthy navigation is performed to find interesting areas, then atomic alignment can be achieved, but scan time increases
Solution Approach 1:
The machine learning model acts as an intermediary that predicts atomic structures from low-resolution or low-dose images. This intermediary enables the system to identify regions of interest and achieve atomic alignment without performing lengthy traditional navigation, significantly reducing scan time while maintaining measurement precision.
Solution Approach 2:
The system performs preliminary low-dose scanning combined with machine learning prediction to pre-identify atomic structures and regions of interest before final high-resolution imaging. This preliminary action eliminates the need for lengthy navigation during actual imaging, reducing scan time while ensuring proper atomic alignment.
3Object-affected harmful factors
If low electron dose is used during navigation, then sample damage is reduced, but image quality decreases
Solution Approach 1:
The machine learning model serves as an intermediary that enhances low-dose images by predicting atomic structures from noisy, low-signal data. This allows the system to maintain low electron doses during navigation (reducing sample damage) while still achieving atomic-level measurement precision through the predictive modeling capability.
Solution Approach 2:
The system changes the parameter of image processing by applying machine learning-based denoising and structure prediction to low-dose images. This transformation allows low-dose images to provide sufficient information for atomic structure identification, maintaining both sample integrity and measurement precision.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach allows for the generation of enhanced images indicating atomic structure, facilitating live assistance during navigation and reducing sample damage by minimizing radiation exposure and scan time.
Implementation Method 1
The electrons interact with the sample, resulting in elastically scattered electrons exiting the sample
Implementation Method 2
In a transmission imaging mode, the electrons transmitted through the sample are detected and used to form a microscopic image of the sample
Data Source
AI summary
A method of imaging a sample includes acquiring one or more first images of a region of the sample at a first imaging condition with a charged particle microscope system. The one or more first images are applied to an input of a trained machine learning model to obtain a predicted image indicating atom structure probability in the region of the sample. An enhanced image indicating atom locations in the region of the sample based on the atom structure probability in the predicted image is caused to be displayed in response to obtaining the predicted image.