Speaker
Description
In water Cherenkov detectors, particle direction is reflected in the spatial and temporal patterns of photomultiplier tube (PMT) signals. Individual PMT hits provide several low-level observables, such as hit position, charge, and timing, while additional directional information may also arise from correlations among these quantities. Understanding how such information is distributed across PMT observables is important for determining which detector information is most relevant for direction reconstruction.
In this study, simulated PMT hit data are used to investigate the directional information contained in low-level detector observables. Basic quantities including PMT position, charge, and hit timing are first examined individually and in combination using machine-learning-based direction reconstruction. Derived spatial, temporal, and charge-related quantities are then considered to explore whether additional directional information can be extracted from hit patterns and correlations between observables.
The contribution of different groups of observables will be evaluated through feature ablation and importance analyses, with particular attention to complementary information and correlations between spatial and timing measurements. Comparisons across particle types and detector conditions will also be considered to examine the robustness of the observed feature dependencies. This study aims to clarify how particle-direction information is encoded in low-level PMT responses and to identify the observables most relevant for direction reconstruction in water Cherenkov detectors.