Hybrid Machine Learning to Classify STM Tip States (infocus #81 March 2026)

Scanning tunneling microscopy (STM) is the premier technique for atomic-scale imaging.

DOI: 10.22443/rms.inf.1.310

By raster scanning a surface with an atomically sharp tip, an image topography can be deduced via the amount of tunneling current induced by quantum tunneling electrons. A serious drawback is the maintenance of the tip itself. The state that it is in can be inferred only by the images taken, a process that is not only time consuming, but also unreliable.

A solution to the problem of classifying these STM tip states is to train a neural network on years of already-collected data. Once trained, the programme aims to identify the tip state of the STM based on the image and spectroscopy data.

Return to listing