Dresden 2026 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

O: Fachverband Oberflächenphysik

O 44: Scanning probe techniques: Method development – Poster

O 44.5: Poster

Tuesday, March 10, 2026, 14:00–16:00, P2

Automatic tip functionalization with CO on Ag(111) using machine learning — •Johanna Schnorrenberg, Jonas Heggemann, Paul Laubrock, and Philipp Rahe — Universität Osnabrück, Institut für Physik, Barbarastraße 7, 49076 Osnabrück, Germany

Functionalization of scanning probe microscopy tips with CO molecules enables high-resolution imaging, sub-molecular contrast and access to charge distributions of adsorbed molecules. [1, 2]

However, the manual functionalization procedure is time-consuming and limits experimental throughput. To address this, tip functionalization can be automated. [3]

Here, we implement a full pipeline for CO tip functionalization on Ag(111). The process includes localizing a CO molecule during scanning the surface, performing the pick-up of this CO molecule and evaluating the resulting tip condition. Tip quality is assessed by imaging CO molecules on the substrate and classifying the achieved resolution as "good" or "bad" for a CO-tip using a convolutional neural network (CNN). The CNN performance is compared to a conventional image analysis algorithm, in this case a template matching algorithm.

We expect this automated approach to significantly improve the efficiency of AFM experiments and thereby enabling faster and more systematic data acquisition on molecular systems.

[1] B. Schulze Lammers et al., Nanoscale, 13, 13617 (2021)

[2] L. Gross et al., Angew. Chemie Int. Ed., 57, 3888 (2018)

[3] B. Alldritt et al., Comp. Phys. Commun., 273, 108258 (2022)

Keywords: Scanning probe microscopy; Scanning tunneling microscope; High resolution atomic force microscopy; Tip functionalization; Machine learning