Machine learning-enabled autonomous operation for atomic force microscopes

Area of Science:

• Materials Science
• Nanotechnology
• Automation Engineering

Background:

• Scientific instrument operation often requires specialized skills, leading to variability in results.
• Autonomous operation can enhance reproducibility and reliability in scientific measurements.

Purpose of the Study:

• To demonstrate the autonomous operation of an atomic force microscope (AFM) using machine learning.
• To reduce operator-to-operator variation in AFM measurements.

Main Methods:

• Implementation of a machine learning-based object detection technique using region-based convolutional neural networks.
• Utilizing two cameras for object recognition, self-calibration, and instrument alignment.
• Development of an automated system for AFM initialization, surface imaging, and image analysis.

Main Results:

• Successful autonomous initialization, surface imaging, and image analysis of the atomic force microscope.
• Demonstrated capability of the machine learning algorithm to perform self-calibration and alignment.
• Achieved reproducible and reliable results with minimal operator intervention.

Conclusions:

• Machine learning enables autonomous operation of atomic force microscopes.
• The developed approach can be generalized to other scanning probe microscopes and scientific instruments.
• Autonomous systems offer significant benefits for scientific research by improving data consistency.