Automated Scanning Probe Tip State Classification without Machine Learning.

Area of Science:

• Materials Science
• Nanotechnology
• Surface Science

Background:

• Manual scanning probe tip state identification is time-consuming in atomic-resolution scanning probe microscopy.
• Existing machine learning methods require large, labeled datasets, which are difficult and time-intensive to create for new systems.
• Automating tip classification is essential for improving efficiency and accuracy in scanning probe microscopy.

Purpose of the Study:

• To develop a more efficient method for classifying scanning probe tip states.
• To demonstrate that template matching (TM) can achieve comparable accuracy to machine learning with less data.
• To provide a viable alternative for tip classification when large labeled datasets are unavailable.

Main Methods:

• Utilized template matching (TM) for classifying scanning probe tip states from topographical images.
• Compared the accuracy and precision of TM with machine learning classifiers.
• Applied TM to prototypical silicon-based surfaces and other systems lacking sufficient training data for machine learning.

Main Results:

• Template matching (TM) achieved accuracy and precision comparable to machine learning methods.
• TM successfully classified tip states in systems where supervised machine learning training was not feasible.
• The method requires only a single surface image and minimal prior knowledge of the system.

Conclusions:

• Template matching (TM) provides an effective and data-efficient solution for scanning probe tip classification.
• TM significantly reduces the time and effort associated with manual inspection and data preparation.
• This approach enhances the automation and applicability of atomic-resolution scanning probe microscopy across diverse systems.