Heidelberg 2022 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 25: Data Analysis, Information Technology and Artificial Intelligence
T 25.5: Vortrag
Montag, 21. März 2022, 17:25–17:40, T-H38
Non-parametric background models for axion haloscopes — •Johannes Diehl1, Jakob Knollmüller2, and Oliver Schulz1 — 1Max Planck Institute for Physics, Munich, Germany — 2Max Planck Institute for Astrophysics, Munich, Germany
Axions have been introduced to solve the strong CP problem of the standard model of particle physics and turned out to be an excellent candidate to explain cold dark matter. "Haloscopes" are searching world wide for axions from the galactic dark matter halo, mostly by axion conversion to photons at radio frequencies in a strong B-field. Finding an axion signal in haloscope data means finding a small peak in a vast non-uniform RF background. One crucial challenge is therefore to selectively suppress larger frequency scales while inducing as little attenuation and correlation as possible at smaller frequency scales. This has so far been tackled using filter theory, e.g. through Savitzky-Golay filters for the HAYSTAC experiment, but proof that this is the optimal filter to use is still lacking. Using simulated data from the MADMAX haloscope, I present a novel machine-learning based approach to separate scales and subtract the background without attenuating the signal which lends itself well to being incorporated into a final Bayesian analysis.