TRIDENT Neutrino Telescope

• TRIDENT is a deep-sea neutrino telescope that uses water Cherenkov detection to study all neutrino flavors with high angular and energy resolution.
• It features ~1,000 vertical detection strings with hybrid Digital Optical Modules, employing sub-nanosecond timing and real-time calibration to optimize event reconstruction.
• Innovative methods like GNN-based reconstruction and in-situ optical calibration enhance point-source localization and flavor discrimination for breakthrough astrophysical insights.

The TRIDENT (tRopIcal DEep-sea Neutrino Telescope) experiment is a planned multi-cubic-kilometer-scale water Cherenkov neutrino observatory intended for deployment in the South China Sea at depths of 2800–3500 meters. TRIDENT targets the detection and detailed study of astrophysical neutrinos across all flavors, offering a near-equatorial vantage point to complement IceCube, KM3NeT, and similar telescopes, and incorporates advanced digital optical modules, precision timing, and real-time calibration systems. The conceptual and technical advances validated through the TRIDENT Pathfinder (T-REX) have provided essential measurements and technology demonstrations crucial for the full realization of this next-generation instrument.

1. Site Selection and Scientific Motivation

The TRIDENT project emerged from systematic site surveys in the north-eastern sector of the South China Sea (centered at 17.4° N, 114.0° E), with the selection of an abyssal plain at depths near 3.5 km, characterized by clay-silt sediments, rms seabed slopes <0.01°, and minimal background from benthic activity or biological sources below 3 km. The chosen location provides a mean overburden sufficient to suppress the atmospheric muon flux to below $10^{-8}$ cm$^{-2}$ s$^{-1}$ sr$^{-1}$, with hydrodynamic models (ROMS) and in-situ LADCP confirming bottom currents $<10$ cm/s at target depths and temperature stabilities of 2–2.5 °C. The equatorial proximity ($17^\circ$ N latitude) yields nearly uniform full-sky neutrino exposure, augmenting source searches in celestial regions inaccessible to northern and polar telescopes. Submarine power and data transmission are feasible via a $180$ km cable to Yongxing Island (Ye et al., 2022).

The scientific agenda includes:

• High-statistics observation of steady and transient astrophysical neutrino sources.
• Sub-degree point-source localization and all-flavor sensitivity for flavor physics (e.g., $\nu_\tau$ appearance).
• Searches for physics beyond the Standard Model, including dark matter annihilation in the Galactic Centre, with cross-section sensitivity to $$\langle\sigma v\rangle\approx5\times10^{-27}\,\mathrm{cm}^3\,\mathrm{s}^{-1}$$ for $m_\chi=10\,$TeV, below the canonical thermal relic benchmark (Wang et al., 11 Jan 2026).
• Probing the origin of cosmic rays via flavor-specific neutrino observations spanning 1 TeV–10 PeV (Ye et al., 2022).

2. Detector Architecture and Hybrid Digital Optical Modules

Array and Geometry

The baseline TRIDENT design comprises approximately $1\,000$ vertical detection strings occupying a roughly circular instrumented volume with a radius of $\sim1.4$ km ($\sim6$ km$^3$ fiducial). Each string supports $20$ hybrid Digital Optical Modules (hDOMs) spaced $30$ m apart over a vertical reach of $570$ m. Strings are arranged in a Penrose tiling to minimize regular grid artifacts and avoid "dead zones," balancing detection efficiency against mechanical complexity (Morton-Blake et al., 28 Oct 2025).

hDOMs and Sensor Technologies

The hDOM encapsulates a pressure-resistant ($\geq5$ km) glass sphere housing:

• $31$ high-QE $3''$ PMTs (e.g., Hamamatsu R14374, QE $\geq30\%$, TTS $\sim1.4$ ns) in the reference design, or alternatively $19$ $4''$ PMTs (e.g., NNVT N2042-like) with high QE—a choice offering nearly 40% reductions in channel count, per-module cost, and power draw ($\sim$5.7 W vs. 9.4 W), while maintaining or exceeding efficiency and angular resolution (Shao et al., 14 Jul 2025).
• $24$ SiPM arrays (e.g., $4\times8$ or $8\times8$ format), enhancing photon detection around interstices and extending dynamic range.
• PMTs are coupled to the sphere with optical gel and oriented to ensure near-$4\pi$ solid angle coverage.
• Local coincidence logic among PMTs and SiPMs is implemented to suppress dark and bioluminescence background.

Auxiliary electronics provide $<1$ ns timing via White Rabbit distribution, active temperature control, and event triggering under a variety of coincidence schemes ("5L1": five hDOMs each showing $L\geq2$ PMT hits within a set window) (Shao et al., 14 Jul 2025, Wang et al., 2023).

3. Calibration, Synchronization, and the Pathfinder Campaign

T-REX Pathfinder Experiment

The 2021 T-REX deployment performed in-situ optical calibration and technology validation at 3420 m depth. The setup consisted of three modules: a Light Emitter Module (LEM) located mid-string, and two Light Receiver Modules (LRMs) at $+41.79\pm0.04$ m and $-21.73\pm0.02$ m vertical separation.

LEM: Light-Source Instrumentation

• The LEM is a 17″ glass sphere containing pulsed LEDs (405, 450, 525 nm; FWHM $\sim3$–$6.6$ ns) driven by a Kapustinsky-type fast pulser with a tunable bias voltage (0–30 V, DAC step $\sim$7 mV), supporting pulse rates up to 10 kHz (and extendable to 100 kHz) (Tang et al., 2023, Li et al., 2023).
• Steady illumination is delivered via fifteen LEDs (405, 460, 525 nm) in isotropic arrays ($\leq$20% anisotropy), with constant-current drivers ensuring stability over $\Delta T$ of $\pm10^\circ$C.
• The driver architecture features an FPGA-based Central Logic Board with remote real-time control (Ethernet $\rightarrow$ White Rabbit $\rightarrow$ SPI $\rightarrow$ DAC), enabling automated bias-voltage scans and intensity tuning (Tang et al., 2023).

LRM: PMT and Camera Systems

• Each LRM houses three 3″ PMTs (HZC XP72B22, QE $>$25%, TTS $<$4 ns, dark count $<1.25$ kHz at 2 °C) arranged $\pm$30° from vertical, with front-end electronics providing $\sim$1 ns pulse digitization and sub-nanosecond White Rabbit time-stamping (Zhang et al., 2023).
• CMOS cameras (Sony IMX265, 2448×2048 px, 3.45 µm pitch, $\sim$5 MP, quantum efficiency 40–65% over 350–600 nm) are installed for real-time optical calibration using steady LEDs (Tian et al., 2024).

Timing and Data Acquisition

• WR-based PTP and Synchronous Ethernet provide clock distribution with residual $\sigma_t\sim11$ ps skew; overall event timing resolution $\lesssim200$ ps (including ADC and PMT jitter) (Wang et al., 2023).
• The DAQ system utilizes 12-bit ADCs at 250 MSPS, FPGA ring buffers, and high-speed Ethernet for lossless streaming (individual module bandwidth $>400$ Mbps, with $\sim$90 Mbps sustained at 10 kHz global trigger) (Wang et al., 2023).

In-Situ Optical Measurements

• PMT-based time-of-flight fits (using Geant4-simulated Green's functions) yielded absorption ($\lambda_\mathrm{abs} = 27.4^{+1.1}_{-0.9}$ m), Rayleigh scattering ($\lambda_\mathrm{Ray} = 200^{+13}_{-10}$ m), Mie scattering ($\lambda_\mathrm{Mie} = 84^{+12}_{-8}$ m) at 450 nm, and mean attenuation $\lambda_\mathrm{att} = 18.7^{+3.0}_{-2.1}$ m (Ye et al., 2022, Zhang et al., 2023). CMOS-camera analysis confirmed attenuation lengths (e.g., $\lambda_\mathrm{att}(460\,\mathrm{nm})=19.3\pm1.3$ m) through image gray-value ratios (Tian et al., 2024).
• The camera system demonstrated real-time calibration (<1 min per update; $\lambda_{\mathrm{att}}$ variance $<3$\%/30 min), allowing integration of dynamic water properties into event reconstruction to reduce angular and energy scale systematics (Tian et al., 2024). These calibrations improved angular resolution at 10 TeV from 0.35° to 0.31°, and reduced energy-scale bias from 8% to 3%.

4. Performance Simulations and Event Reconstruction

Simulation and Reconstruction Chain

The full-chain TRIDENTSim framework incorporates CORSIKA8/Pythia8/Geant4/OptiX for neutrino event generation, muon transport, photon propagation, and detector response, using site-specific (T-REX-derived) water optical parameters (Morton-Blake et al., 28 Oct 2025, Ye et al., 2022).

Topologies and Resolutions

• Tracks (νμ CC): Muons traversing $O(1$–$10)\,$km, with angular resolution $\sigma_\theta \sim0.2^\circ$ at 100 TeV ($\sim$0.7° at 1 TeV), energy resolution $\Delta E/E$ $\sim50$–100%.
• Cascades (νe, ντ CC & NC): Point-like showers (length $\lesssim10$ m), energy resolution $\Delta E/E\lesssim10$–15%, angular resolution $\sim1$–2° at TeV, $\sim0.7^\circ$ at 100 TeV (Morton-Blake et al., 28 Oct 2025, Ye et al., 2022).

Graph Neural Network (GNN)-Based Reconstruction

A GNN-based event reconstruction pipeline leverages the sparse and irregular TRIDENT geometry:

• Each triggered hDOM is a graph node with attributes (arrival-time histograms for showers; earliest photon time and integrated charge for tracks).
• Edges connect k-nearest neighbors, optionally embedding time-residual features.
• EdgeConv blocks aggregate local features, propagate messages, and the global representation is used to regress event direction, energy, and, prospectively, interaction vertex (Mo et al., 2024).
• Performance: Median angular error for showers (νe CC, 100 TeV) $\sim1.3^\circ$ ($68\%$ quantile $\sim0.9^\circ$); for tracks (νμ CC, $E_\nu\gtrsim100\,$TeV) median $\sim0.1^\circ$ ($68\%$ quantile at $0.07^\circ$), surpassing or matching traditional likelihood fitters as in KM3NeT (Mo et al., 2024).

5. Detector Optimization, Deployment, and Systematics

Geometry and Layout Studies

Simulation-driven optimization indicated:

• Inter-string spacing: Reference value of 100 m (uniform Penrose tiling) optimizes discovery time across all channels, with total instrumented volume $\sim6.1\,$km$^3$ (Morton-Blake et al., 28 Oct 2025).
• Vertical hDOM spacing: 30 m as the baseline; taller strings (45 m) slightly benefit high-energy track acceptance but lose cascade sensitivity.
• Clustering: Dense string clusters degrade point-source sensitivity due to "holes" and reduced uniformity; uniform distribution avoids these pitfalls.
• Optical properties: Variations in $\lambda_\mathrm{att}$ overwhelmingly dominate sensitivity; a 20% reduction in attenuation length can increase discovery times by factors of 2–4, mandating real-time, distributed calibration (Morton-Blake et al., 28 Oct 2025).

Instrumentation Trade-offs

Transitioning to $4"$ PMT-based hDOMs (high QE, TTS $\sim1.4\,$ns) offers $\sim$40% savings in channel count, cost, and power per module with equal or better detection efficiency, especially if large PMTs maintain the quantum efficiency of the $3"$ reference sensors. Performance remains robust for tracks, cascades, and $\nu_\tau$-double pulse ID (Shao et al., 14 Jul 2025).

Calibration and Environmental Control

Real-time optical calibration is achieved with distributed OCM units on every fourth string, each hosting a LEM and two LRMs (camera + PMTs), updating absorption/scattering constants hourly. Mechanical tilt, biofouling, and data dropouts were negligible or controlled during T-REX, foreseeing routine diagnostic operations for the full array (Tian et al., 2024). Sub-nanosecond synchronization via White Rabbit is standard across all modules.

6. Physics Reach and Future Prospects

TRIDENT’s projected sensitivities include:

• Rapid identification (within one year) of several IceCube-source candidates (e.g., NGC 1068 at $5\sigma$ significance) (Ye et al., 2022).
• $\nu_\mu$ track angular resolution of $0.2^\circ$ at 100 TeV, with cascade angular resolution reaching $0.7^\circ$ and energy resolution $\leq15\%$.
• Discovery times for hard-spectrum ($\gamma=2$) point sources at declination $\delta=0$ of $\sim5$ years for baseline fluxes; for soft-spectrum ($\gamma=3$) sources, $\sim3$–$4$ years due to superior cascade sensitivity (Morton-Blake et al., 28 Oct 2025).
• Sensitivity to WIMP dark matter annihilation cross-sections to the level of $5\times10^{-27}\,$cm$^3$ s$^{-1}$ at 10 TeV, covering new regions of the parameter space, with critical all-flavor and especially cascade-channel performance (Wang et al., 11 Jan 2026).
• Real-time (sub-10 ms) event-by-event GNN-based reconstruction enabling multi-messenger transient triggers (Mo et al., 2024).

Planned enhancements include higher LED pulse repetition ($\geq100\,$kHz), extension of wavelength coverage, on-board closed-loop photon counters, autofocus-capable calibration cameras, and further refinement of dynamic graph/event reconstruction algorithms. Robustness against environmental changes and hardware variability is incorporated at all system levels.

TRIDENT, through its densely-instrumented geometry, advanced timing, and comprehensive calibration, is positioned to achieve unprecedented neutrino source sensitivity and flavor discrimination on astronomical baselines, constituting a foundational advance in neutrino astronomy and multi-messenger astrophysics.