Regensburg 2025 – scientific programme

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

O: Fachverband Oberflächenphysik

O 18: Poster Focus Session Ultrafast Electron Microscopy at the Space-Time Limit

O 18.3: Poster

Monday, March 17, 2025, 18:00–20:00, P2

Machine learning-based denoising and artefact removal for multidimensional photoemission data — •Joshka Laird, Tommaso Pincelli, and Laurenz Rettig — Fritz haber institute of the Max Planck Society, Berlin, Germany

Angle-Resolved Photoemission Spectroscopy (ARPES) is a powerful tool for investigating the electronic structure of materials. While modern approaches such as momentum microscopy provide rich, multidimensional photoemission data, they pose challenges for achieving high statistics data and good signal-to-noise ratios. Additionally, image distortions such as mesh artefacts often complicate the analysis. Traditional denoising techniques, while effective in specific scenarios, can fail to preserve the fine structural details essential for accurate interpretation.

Here, we present a machine learning-based denoising and artefact removal approach for multidimensional photoemission data. Based on recent results using convolutional neural networks [1], we discuss how to extend such networks to higher dimensions to cope with data e.g. from time-resolved momentum microscopy.

[1] Y. Kim et al., Rev. Sci. Instrum. 92, 073901 (2021)

Keywords: Angle-Resolved Photoemission Spectroscopy (ARPES); Machine Learning; Convolutional Neural Networks (CNNs); Denoising Techniques; Artefact Removal