Principle of TEM alignment using convolutional neural networks: Case study on condenser aperture alignment.

Area of Science:

• Materials Science
• Instrumentation
• Computational Science

Background:

• Transmission electron microscopy (TEM) requires precise alignment for optimal performance.
• Manual alignment is time-consuming and requires extensive user training.
• Automating alignment can significantly improve efficiency and accessibility of TEM techniques.

Purpose of the Study:

• To explore the feasibility of using artificial intelligence (AI) for automated TEM alignment.
• To develop and test an AI-based method for centering the condenser aperture, a critical alignment step.
• To assess the potential for AI to reduce the skill and time required for TEM operation.

Main Methods:

• A convolutional neural network (CNN) was developed to predict necessary shifts for aperture realignment.
• A simplified digital twin was used for automated data acquisition for training the CNN.
• Various CNN models were evaluated to determine the optimal design for performance.

Main Results:

• The developed AI method achieved human-level performance in centering the condenser aperture.
• The system demonstrated the ability to predict the required x and y-shifts accurately in a single step.
• The AI approach proved compatible with continuous drift correction and uniformity of illumination.

Conclusions:

• AI, specifically CNNs, offers a viable solution for automating TEM alignment tasks.
• This method can substantially decrease the learning curve and operational time for TEM users.
• The AI-driven alignment is adaptable for various alignment steps and continuous correction during experiments.