Machine Learning Accelerated Characterization of Single Adatoms Via STM Imaging
H. Sabri (TU Delft - Applied Sciences)
A.A.F.M. Artaud – Mentor (TU Delft - QN/Artaud Lab)
K.R. Rossi – Mentor (TU Delft - Team Kevin Rossi)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
We demonstrate the ability of computer vision methods and unsupervised machine learning models to analyze atom-scale images acquired by scanning tunneling microscopy (STM). The imaged system of interest consists of Fe and Ti atoms adsorbed on bilayer MgO islands grown on an Ag substrate, a system studied for its quantum magnetic properties. The purpose of this analysis is to detect and classify each atom site, which includes Ti atoms, Fe atoms, defects on the surface of the sample, and unknown contaminants. Our method identifies Fe and Ti atoms from the rest in an effort to aid laboratory researchers during single-atom probing tasks. Each acquired STM image is processed with computer vision libraries to locate each atom site and to build a dataset of atom crops that highlight their features. Atom site classification and characterization tasks are performed using Gaussian mixture models (GMM), principal component analyses (PCA), and density-based clustering to classify each atom site by species and quantify structural differences and similarities between them. Data and code are openly available at https://github.com/AudreySabri/unsupervised-ml-for-microscopy.git