Demonstration of an AI-driven workflow for autonomous high-resolution scanning microscopy

Area of Science:

• Materials Science
• Microscopy
• Data Science

Background:

• Modern scanning microscopes generate massive datasets, posing storage and analysis challenges.
• High-resolution imaging requires efficient methods for data acquisition and interpretation.

Purpose of the Study:

• To develop a self-driving experimental toolkit (FAST) for efficient data acquisition in scanning microscopy.
• To reduce the volume of data required for accurate materials analysis without prior sample information.

Main Methods:

• Integration of neural networks, route optimization, and hardware controls for autonomous data selection.
• Development of the Fast Autonomous Scanning Toolkit (FAST) for self-driving experiments.
• Testing FAST in simulations and dark-field X-ray microscopy of WSe2 films.

Main Results:

• FAST enables accurate imaging and analysis using less than 25% of the full dataset.
• The toolkit is computationally efficient and requires minimal experiment-specific configuration.
• Demonstrated successful application in X-ray microscopy of WSe2 films.

Conclusions:

• FAST significantly reduces data requirements for scanning microscopy, making advanced materials analysis more accessible.
• The toolkit's adaptability to various scanning microscopes facilitates broader studies of materials evolution.
• Autonomous data acquisition empowers efficient multi-parameter investigations of material properties.