Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
O: Fachverband Oberflächenphysik
O 85: Vacuum Science & Technology: Theory and Applications II
O 85.6: Talk
Thursday, March 12, 2026, 16:30–16:45, HSZ/0401
ESEM Automation - Agent-Oriented Autonomous Microscopy — •Maurits Vuijk1, Johannes Zeininger2, Karsten Reuter1, Thomas Lunkenbein3, and Christoph Scheurer1 — 1Fritz-Haber-Institut der MPG, Berlin — 2TU Wien — 3Universität Bayreuth
Are the datasets captured in usual Environmental Scanning Electron Microscopy (ESEM) experiments truly representative? While focusing on a single view or feature, critical events may occur elsewhere on the accessible surface. In addition, effects such as beam damage and inhomogeneous gas concentration introduce further need for more representative surface sampling. Starting from an automation interface to the ESEM, we have built an agent-oriented framework that enables fully autonomous ESEM experiments. To achieve this, the actions normally carried out by the human operator are instead performed by various agents. To this end, a set of agents keep up image quality by controlling brightness and contrast, focus and astigmatism without interrupting the imaging flow. A set of computer vision agents analyze captured images on the fly for interesting features and activity. Using a dynamic priority queue system, high-resolution frames are automatically taken of identified active zones. Each overview image is processed and scored based on activity, feature presence, or other desired qualities. The queue system then assigns an acquisition frequency based on the score, giving priority to more active zones. This gives a full set of overview images and multiple smaller image sets of dynamically identified active areas.
Keywords: Electron microscopy; Lab automation; Computer vision