KM3NeT/ORCA Neutrino Telescope

Updated 1 December 2025

KM3NeT/ORCA is a deep-sea water-Cherenkov neutrino telescope designed for precise atmospheric neutrino oscillation studies and mass ordering determination.
It employs 115 detection units with multi-PMT optical modules, achieving nanosecond timing and ~10 cm positioning for accurate event reconstruction.
Advanced methods like multivariate classifiers and deep-learning enhance flavor identification and background rejection in the few-GeV energy range.

KM3NeT/ORCA (Oscillation Research with Cosmics in the Abyss) is a deep-sea, multi-megaton water-Cherenkov array optimized for precision studies of atmospheric neutrino oscillations, the determination of the neutrino mass ordering, and searches for new physics such as non-standard interactions or sterile neutrinos. ORCA, located at a depth of 2450 m in the Mediterranean Sea off Toulon, France, is one of two major KM3NeT sites—the other being ARCA, targeting high-energy neutrino astronomy. The ORCA detector’s granular, multi-PMT optical architecture provides unique sensitivity in the few-GeV regime, enabling measurements of Earth-induced matter effects, flavor appearance/disappearance channels, and tomographic probing of the Earth’s interior electron density through neutrino oscillation patterns.

1. Detector Architecture and Calibration

ORCA consists of 115 vertical Detection Units (DUs), each a flexible string anchored to the seabed and held vertical by a submerged buoy, and instrumented with 18 Digital Optical Modules (DOMs) spaced by 9 m. Each DOM is a 17″ pressure-resistant glass sphere housing 31 three-inch photomultiplier tubes (PMTs) oriented to achieve close to 4π coverage. The array spans a horizontal footprint with ~20 m spacing between DUs, yielding a total instrumented volume of 7–8 Mton of Mediterranean seawater. Deep-sea water at the site exhibits exceptional optical properties (absorption length ≳60 m, scattering length ≳200 m), supporting nanosecond-scale timing and ~10 cm-scale positional precision essential for sub-degree directional reconstruction (Coyle, 2017, Breton et al., 2021, Adrián-Martínez et al., 2016).

Precise geometrical and timing calibration is maintained by a combination of White Rabbit clock distribution, in-DOM LED nanobeacons, external laser-flash beacons, piezo-acoustic triangulation, and in-situ environmental monitoring from a dedicated Calibration Unit. The calibration system achieves timing offsets ≲1 ns and DOM positioning <10 cm, mitigating systematics in angular response and energy scale (Breton et al., 2021).

Table 1. Select Detector Parameters

Subsystem	Value/Range	Note
DUs, DOMs/DU	115 DUs, 18 DOMs	Full ORCA configuration
PMTs per DOM	31 × 3″	Multi-PMT segmentation for wide angular coverage
Vertical spacing	9 m (DOM–DOM)
Horizontal pitch	~20 m (DU–DU)
Instrumented mass	7–8 Mton	Water-equivalent
Depth	2450 m	Mediterranean, 40 km off Toulon
DOM timing/position	≲1 ns / ≲10 cm	Calibration precision

2. Detection Principle, Data Acquisition, and Background Rejection

Atmospheric neutrinos of energies 1–100 GeV interact in or near the instrumented volume via charged-current (CC) or neutral-current (NC) weak processes, producing relativistic leptons (μ, e, τ) and hadronic showers that emit Cherenkov light. The PMTs register photon arrival times (“hits”) above a threshold of ~0.3 photoelectrons with ~1 ns precision. Hit data are streamed unfiltered (“all-data-to-shore”) over optical fibers; real-time CPU farms onshore run a hierarchical software trigger and filtering system (Adriani et al., 6 Jun 2025).

Physical event topologies are categorized into:

Track-like (μCC): Elongated photon patterns from muons, providing best angular resolution (σθ ≲ 5° at Eν ≳ 5 GeV).
Shower-like (eCC, τCC, NC): Spherical/compacted light distributions with coarser angle resolution (σθ ≃ 10–25° for cascades).

Environmental backgrounds are dominated by ^40K decays and bioluminescence (baseline PMT rates ~7–10 kHz), as well as downward-going atmospheric muons. Noise is reduced online by requiring spatiotemporal coincidences (e.g., ≥k PMTs in a DOM within ~10 ns), geometric consistency (track/vertex containment), and multivariate classification (Random Decision Forests, CNNs). Ultimate cosmic-ray muon contamination in the up-going (signal) sample is suppressed to ≲3% (Adriani et al., 6 Jun 2025, Kalaczyński, 2021).

3. Neutrino Oscillation Physics and Mass Ordering Sensitivity

ORCA’s primary scientific goal is precision measurement of three-flavor atmospheric neutrino oscillations, with emphasis on the mass ordering (normal vs. inverted). The evolution of the flavor state ψ(x) in matter is described by

$i\,\frac{d\psi}{dx} = H \psi,$

where the Hamiltonian

$H = \frac{1}{2E} U\,\mathrm{diag}(0, \Delta m^2_{21}, \Delta m^2_{31})\,U^\dagger + \mathrm{diag}(V(x), 0, 0),$

with the matter potential $V(x) = \sqrt{2}\,G_F\,N_e(x)$ depending on electron density. Resonant enhancement of the νμ→νe appearance probability (the MSW effect) occurs near $E_\text{res}\sim 5$ –8 GeV for upgoing neutrinos traversing the Earth's mantle and core (Hofestädt, 2017, Zaborov, 2018).

Mass ordering determination relies on the O(10%) modulation of the (E, cos θz) event distribution for tracks and showers, exploiting both disappearance (P{μμ}) and appearance (P_{μe}) channels. Monte Carlo-based likelihood analysis, binning reconstructed (E, cos θ_z), comparing event counts under both orderings, and profiling over systematic and oscillation nuisance parameters, yields median sensitivities of:

3σ mass-ordering discrimination in 3 years (full ORCA, favorable θ23, δCP);
>5σ in optimal conditions or with joint JUNO+ORCA data (Aiello et al., 2021, Kalaczyński, 2021, Coyle, 2017).

Precision on atmospheric parameters approaches Δθ23 ≃ ±2°, Δ(Δm^2_32) ≲ 0.1×10^–3 eV² in three years (Hofestädt, 2017, Collaboration et al., 13 Aug 2024).

4. Advanced Event Reconstruction and Analysis Techniques

ORCA leverages advanced multivariate and deep-learning architectures for event reconstruction and flavor identification. Maximum-likelihood fits to hit time PDFs reconstruct the initial interaction vertex, direction, and visible energy under both track and shower hypotheses. Multivariate classifiers—Random Decision Forests and deep convolutional neural networks (CNNs)—are trained to distinguish track/shower topology, atmospheric muons, and random noise (Aiello et al., 2020). CNNs operating on the full DOM × PMT × time data volume yield:

Improved track-vs-shower separation in the critical 3–10 GeV region (∼20% better than Random Forests);
Energy resolution σE/E ≈ 25–30% at few-GeV energies (showers), and comparable or better than likelihood methods;
Per-event uncertainty regression, enabling adaptive quality selections for oscillation fits.

Machine learning tools (e.g., t-SNE, Isomap) targeting single-DOM patterns allow sub-GeV (∼100 MeV–1 GeV) signal efficiency to reach ∼50%, essential for supernova and low-energy transient studies (Wasseige, 2021).

5. Non-Standard Interactions, Sterile Neutrinos, and Earth Tomography

Non-Standard Interactions (NSI): ORCA tests for deviations from standard three-flavor oscillations arising from forward-scattering via unknown couplings (εαβ) in the effective matter Hamiltonian. The OrCA6 dataset (433 kton·yr) yields world-leading constraints, e.g., |ε_{μτ}| ≤ 5.4×10^–3 at 90% CL (Aiello et al., 28 Nov 2024), and is projected to reach sub-10^–3 bounds in the full detector.

Sterile Neutrinos: ORCA's wide L/E range provides sensitivity to active–sterile mixing in 3+1 models from Δm^{2_41 ~ 10^{–5–10 eV^2.}} After three years, ORCA will improve bounds both at eV scale and for ultra-light mixing, probing |U_{μ4}|^2, |U_{τ4}|^2, and θ34 down to previously unconstrained regions (Aiello et al., 2021, Collaboration et al., 8 Oct 2025). The ORCA6 analysis sets |U_{μ4}|^{2 < 0.138} and |U_{τ4}|^{2 < 0.076} (90% CL) for Δm^{2_41 = 1 eV²} (Collaboration et al., 8 Oct 2025).

Neutrino Oscillation Tomography: By fitting oscillation-driven flavor transitions through reconstructed (E, θ_z) distributions, ORCA directly measures the Earth's electron density profile. After 10 years, it targets a precision of 3–5% in the mantle and 7–10% in the outer core, independent of conventional seismic/geodetic ρ(r) constraints. This capability constrains Z/A and, thereby, the chemical composition and element admixture (notably iron content) of the deep Earth (Bourret et al., 2017).

6. Broader Scientific Impact and Future Directions

ORCA’s physics reach encompasses tau-neutrino appearance (∼3000 ν_τ CC/yr), unitarity tests on the leptonic mixing matrix to ±7% (1σ, three years), dark matter via solar or galactic neutrinos, and supernova burst detection through coincident low-energy photomultiplier triggers (Aiello et al., 2021, Adrián-Martínez et al., 2016, Margiotta, 2022). Prospects for CP-violation and precision PMNS studies are augmented by future upgrades (“Super-ORCA”) and potential long-baseline beams (P2O: Protvino–ORCA, 2590 km), which dramatically boost sensitivity and access to matter-enhanced resonances and CP-violating phase effects (Zaborov, 2018).

Continuous detector calibration, expansion to full DU complement, and refinement of data-driven systematics are prioritized for achieving stringent oscillation-parameter, NSI, and sterile-neutrino sensitivity goals.

7. Summary

KM3NeT/ORCA is the premier underwater neutrino-oscillation experiment in the multi-GeV regime, deploying a grid of 115 densely-instrumented multi-PMT optical modules for comprehensive studies of atmospheric neutrino physics, Earth tomography, and new-physics searches. The design leverages megaton-scale target mass, nanosecond timing, advanced reconstruction algorithms, and robust background suppression, yielding global impact in mass ordering determination, oscillation precision, BSM physics, and deep-Earth composition studies (Coyle, 2017, Aiello et al., 2021, Kalaczyński, 2021, Bourret et al., 2017, Aiello et al., 28 Nov 2024, Collaboration et al., 8 Oct 2025).