Dortmund 2021 – wissenschaftliches Programm

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

T: Fachverband Teilchenphysik

T 21: Data analysis, Information technology I

T 21.9: Vortrag

Montag, 15. März 2021, 18:00–18:15, Tu

Muon Track Reconstruction in Liquid Scintillators with Graph Neural Networks — •Rosmarie Wirth — Hamburg University, Hamburg, Germany

Large liquid scintillation detectors are successfully used to observe the neutrino oscillation parameters by detecting reactor neutrinos. A main, hard to identify background are cosmogenics. These are ⁹Li and ⁸He atoms, which are produced in showers along cosmic muon tracks. While decaying the cosmogenics mimic the inverse β-decay, which is the detection process to identify reactor neutrinos. While muon vetos are a straight forward method to reduce this background, they create a lot of dead time. With the JUNO experiment 15.7 % reactor neutrino events are predicted to be missed, due to the muon veto. A superior muon track and shower reconstruction method, could improve the data taking of JUNO and comparable experiments tremendously. Classical and machine learning approaches are being developed for JUNO.

The here presented work studies the use of Graph Neural Networks to reconstruct muon tracks and corresponding showers. Graph Neural Networks provide the option to include the geometrical detector setup to improve the reconstruction. On TOY Monte Carlo Data showers in the detector volume could be located with an accuracy of ± 0.22 ± 0.14 m. Additionally results on a voxelwise photon emission distribution are presented.