| Summary: | <p>The detection of neutrinos produced from core-collapse supernovae will provide invaluable insights into constraining cosmological models, explosion dynamics, star formation, and neutrino properties. Given the low burst rate of Milky Way supernovae, there is strong interest in detecting the accumulated flux of neutrinos from more distant supernovae, known as the Diffuse Supernova Neutrino Background (DSNB). The gadolinium-loaded Super-Kamiokande (SK) experiment currently exhibits the best sensitivity for the discovery of the DSNB flux due to enhanced neutron tagging capability with 0.011% Gd₂(SO₄)₃ · 8H₂O, as per this analysis. However, the low-energy signal is dominated by cosmic muon spallation and atmospheric neutrino backgrounds. This thesis presents a novel approach to atmospheric NCQE background reduction by leveraging the spatial and temporal features of events in SK with the discriminative power of Convolutional Neural Networks (CNNs). The techniques developed in this work were applied to a full DSNB search using 552.2 days of data from the SK-VI phase and demonstrated a 40–60% reduction in the expected NCQE background spectrum compared to the previous analysis. Since no significant excess of signal events was observed in the data, the observed and expected upper limits of the DSNB flux were determined to be 0.27–16.03 and 0.31–49.53 cm⁻²s⁻¹MeV⁻¹, across the energy bins. The results highlight the potential that further developing machine learning techniques has to improve sensitivity to the DSNB flux.</p>
<p>The Hyper-Kamiokande experiment is a next-generation water Cherenkov detector that will be able to constrain neutrino oscillation parameters, measure supernova neutrinos, and measure CP violation in the lepton sector with unprecedented statistical precision. However, these analyses will rely on an effective outer detector (OD) to veto against cosmic muon backgrounds. In this thesis, the OD's photosensor arrangement has been optimised, and the performance of wavelength-shifting (WLS) plates was evaluated in two unique setups. The optical measurements made in this work improved on existing absorbance results and demonstrated a previously unknown artefact of Mie scattering present in all samples. In addition, a new water-based test facility, called BabyK, was constructed to determine the light collection efficiency of all samples in ultra-pure water. This work proposes instrumenting approximately 7,200 photosensors and using either V.A. Kargin POPOP50-PPO3000 or Kuraray B2 WLS plates in the OD design.</p>
|
|---|