Dortmund 2021 – wissenschaftliches Programm

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

T: Fachverband Teilchenphysik

T 21: Data analysis, Information technology I

T 21.7: Vortrag

Montag, 15. März 2021, 17:30–17:45, Tu

Applications of Graph Neural Networks in Liquid Scintillator Neutrino Detectors — •Alexandros Tsagkarakis, Markus Bachlechner, Thilo Birkenfeld, Philipp Soldin, Achim Stahl, and Christopher Wiebusch — III. Physikalisches Institut B, RWTH Aachen University

In neutrino physics, liquid scintillator detectors like Double Chooz and JUNO are utilized to measure the elements of the Pontecorvo-Maki-Nakagawa-Sakata matrix or to determine the sign of the mass difference Δ m₃₁² of the neutrino mass hierarchy. The main channel for the detection of neutrinos is the Inverse Beta Decay with protons in the detector medium, resulting in a positron and a neutron. Those initiate a prompt signal from the positron and a delayed signal from the neutron. On the other hand electrons from various sources, such as the decay of 9Li and 8He atoms, produce similar signatures, which cause a significant amount of background and hence challenges to achieve the above goals. Therefore, we apply machine learning algorithms for energy and vertex reconstruction or direct electron-positron discrimination to reduce this background. The geometry of the experiments can be well mapped in a Graph Neural Network. In this talk, we present the implementation and the first results of the aforementioned tasks.