This study presents a comprehensive simulation-based investigation into the detection and directional localization of reactor antineutrinos (\bar\nu_e) in deep-sea environments using a water Cherenkov detector. The work addresses the significant challenges posed by the low energy of reactor antineutrinos (2–10 MeV), their tiny weak-interaction cross-sections, and the presence of intense natural radioactivity in seawater. A cylindrical detector (10 m diameter \times 10 m height) filled with pure or gadolinium-doped water and instrumented with 8-inch photomultiplier tubes (PMTs) spaced 40 cm apart is modeled within the WCSim/Geant4 framework. The simulation fully reconstructs both elastic scattering (\bar\nu_e e^-) and inverse-beta-decay (IBD, \bar\nu_e p \rightarrow e^+ n) events, accounting for Cherenkov-photon production, propagation, and PMT response.
Vertex reconstruction is performed by maximizing a likelihood based on time residuals, while the direction of the recoil electron is extracted via a Hough-transform technique that identifies the Cherenkov cone axis. To suppress the dominant IBD background, which lacks directional correlation with the incident neutrino, a coincidence-based identification strategy is developed: a primary signal satisfying a PMT-multiplicity threshold (M_\textpmt>20) is followed by a secondary neutron-capture signal within a 0.5 ms (pure water) or 0.2 ms (Gd-doped water) window. The neutron-capture signal is discriminated from the ubiquitous 40K background by applying an optimal PMT-multiplicity cut (M_\textpmt\approx5). With 0.06% natural Gd doping, the IBD identification efficiency rises to ~72%, nearly doubling that of pure water. Furthermore, a novel correlation-based purification method is introduced to enhance the signal-to-background ratio during source reconstruction. By pairing events and assigning positive weights to small-angle pairs (cosine >0.75) and negative weights to large-angle pairs, the background is effectively suppressed without sacrificing signal statistics, reducing the required number of detector modules by about one-sixth for the same localization confidence.
Simulation results show that the reconstruction efficiency for reactor-spectrum antineutrinos (2–10 MeV) is 2.65%, with vertex and angular resolutions suitable for directional analysis. Under realistic seawater-radioactivity conditions, the expected elastic-scattering rate from a 1 GW reactor at 1 km distance is 0.35 counts/day per module, while the total radioactive background (dominated by 208Tl and 214Bi) amounts to ~5 counts/day per module. With an array of 100 identical modules, the cumulative signal reaches 35 counts/day, yielding a background-to-signal ratio of 13. Applying the correlation-based purification and Hough-transform source reconstruction, the array can localize the reactor direction with >90% confidence (angular cosine >0.8) within one day. The study demonstrates the feasibility of deep-sea water Cherenkov detectors for kilometer-scale monitoring of reactor antineutrinos and outlines a path toward extending the monitoring range to 10 km through further optimization of detector parameters and array size.