A neural network based UHE neutrino reconstruction method for the Askaryan Radio Array (ARA)

• Main Authors: Pan, Y, Allison, P, Archambault, S, Beatty, JJ, Beheler-Amass, M, Besson, DZ, Beydler, M, Chen, CH, Chen, P, Chen, YC, Clark, BA, Clay, W, Connolly, A, Cremonesi, L, Dasgupta, P, Davies, J, de Kockere, S, de Vries, KD, Deaconu, C, DuVernois, MA, Flaherty, J, Friedman, E, Gaior, R, Hanson, J, Hanson, K, Harty, N, Hendricks, B, Hoffman, KD, Hokanson-Fasig, B, Hong, E, Hsu, SY, Huang, JJ, Huang, MH, Hughes, K, Ishihara, A, Karle, A, Kelley, JL, Khandelwal, R, Kim, KC, Kim, MC, Kravchenko, I, Krebs, R, Ku, Y, Kuo, CY, Kurusu, K, Landsman, H, Latif, UA, Laundrie, A, Liu, TC, Lu, MY, Madison, B, Mase, K, Meures, T, Nam, J, Nichol, RJ, Nir, G, Novikov, A, Nozdrina, A, Oberla, E, ÓMurchadha, A, Osborn, J, Pfendner, C, Punsuebsay, N, Roth, J, Sandstrom, P, Seckel, D, Shiao, YS, Shultz, A, Smith, D, Toscano, S, Torres, J, Touart, J, van Eijndhoven, N, Varner, GS, Vieregg, A, Wang, MZ, Wang, SH, Wang, YH, Wissel, SA, Yoshida, S, Young, R
• Format: Report
• Language: English
• Published: Sissa Medialab srl Partita 2022
• Subjects: South Pole; South pole
• Online Access: https://discovery.ucl.ac.uk/id/eprint/10162675/1/ICRC2021_1157.pdf; https://discovery.ucl.ac.uk/id/eprint/10162675/

The Askaryan Radio Array (ARA) is an ultra-high energy (UHE) neutrino (Eν > 1017 eV) detector at South Pole. ARA aims to utilize radio signals detected from UHE neutrino interactions in the glacial ice to infer properties about the interaction vertex as well as the incident neutrino. To retrieve these properties from experiment data, the first step is to extract timing, amplitude and frequency information from waveforms of different antennas buried in the deep ice. These features can then be utilized in a neural network to reconstruct the neutrino interaction vertex position, incoming neutrino direction and shower energy. So far, vertex can be reconstructed through interferometry while neutrino reconstruction is still under investigation. Here I will present a solution based on multi-task deep neural networks which can perform reconstruction of both vertex and incoming neutrinos with a reasonable precision. After training, this solution is capable of rapid reconstructions (e.g. 0.1 ms/event compared to 10000 ms/event in a conventional routine) useful for trigger and filter decisions, and can be easily generalized to different station configurations for both design and analysis purposes.