Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

O: Fachverband Oberflächenphysik

O 81: Nanostructures at Surfaces I

O 81.8: Vortrag

Donnerstag, 21. März 2024, 12:30–12:45, MA 042

Automating measurements on the nanoscale: Artificial Intelligence versus classical analysis of SPM data — •Tim J. Seifert¹, Ziba Akbarian^1,2, Birka Lalkens², Ingo Busch³, Harald Bosse³, and Uta Schlickum^1,2 — ¹Institut für Angewandte Physik, Technische Universität Braunschweig — ²Laboratory for Emerging Nanometrology LENA, Braunschweig — ³Physikalisch-Technische Bundesanstalt, Braunschweig

The continuous trend in research and technology towards structures on the nanometer scale drives the growing interest in imaging mechanisms using Scanning Probe Microscopes (SPM). As the experimental methods continue to evolve, the increasing output of data requires fast, reliable and accurate analysis methods avoiding the need for an experienced user. The automatization of image analysis procedures for SPM mainly consists of well-established routines using classical methods requiring laborious manual work. Artificial Intelligence (AI) based analysis techniques have recently attracted great interest hoping to provide true autonomous imaging and analysis procedures. While the accuracy of classical methods is often limited by noise, AI can overcome these challenges, albeit with the additional need for high amounts of labeled training data. Here we present a framework to analyze SPM data and extract distance information using AI-based methods trained on synthetic data, as well as classical alternatives, highlighting the benefits of each approach. The procedure is applied to measure a novel DNA-Origami based Single-Molecule length reference providing a fast, cheap and accurate way to calibrate SPMs on the nanoscale.

Keywords: Artificial Intelligence; Computer Vision; Scanning Probe Microscopy; Nanometrology; DNA Origami