Combined KM3NeT/ARCA and ANTARES search for compact neutrino sources

Technical Summary: Combined KM3NeT/ARCA and ANTARES Search for Compact Neutrino Sources

Problem and Objective
This work addresses the search for high-energy neutrinos originating from point-like astrophysical sources. The primary objective is to demonstrate the potential of combining data from two distinct neutrino detectors: the ANTARES telescope (operational off Toulon, France, from 2008–2022) and the KM3NeT/ARCA detector (currently under construction off Sicily). Both instruments offer optimal visibility of the Southern Sky and the Galactic Centre. The analysis aims to establish a joint search framework using a binned likelihood approach to improve sensitivity and discovery potential compared to standalone analyses.

Methodology
The analysis utilizes a combined dataset comprising four distinct components:

1. ANTARES Track Dataset: 4541 days of livetime, split into non-overlapping track and shower datasets.
2. ANTARES Shower Dataset: Included in the 4541-day split.
3. ARCA19 Tracks: Track-like events from KM3NeT/ARCA with 19 detection lines (48.4 days).
4. ARCA21 Tracks: Track-like events from KM3NeT/ARCA with 21 detection lines (287.4 days).

Definitions: "Tracks" refer to events involving muon production, while "Showers" are induced by neutral-current interactions of all neutrino flavors and charged-current interactions of electron neutrinos.

Statistical Framework
The analysis employs a binned likelihood approach. The observed event histogram ($N$) is compared against a scalable signal estimate ($S$) based on a reference flux ($\Phi_0 = 10^{-8} \text{ GeV cm}^{-2} \text{ s}^{-1}$) and a background estimate ($B$). The likelihood function is defined as:
$$\ln L = \sum_{i}^{N_{bins}} N_i \ln(B_i + \zeta S_i) - (B_i + \zeta S_i)$$
where $\zeta$ is the signal strength parameter. The test statistic $\lambda$ is the log-likelihood ratio between the best-fit signal strength ($\hat{\zeta}$) and the background-only hypothesis ($\zeta=0$).

• Discovery Potential: Defined as the signal strength ($\zeta_{5\sigma}$) required to achieve a $5\sigma$ excess in 50% of pseudo-experiments.
• Sensitivity: Defined as the median 90% confidence level (CL) upper limit on $\zeta$, calculated using the Neyman approach.

Background Modeling and Systematic Uncertainties
To address systematic uncertainties related to background modeling, four independent analysis cases were performed across declination bands:

• Case 1: ARCA background is entirely data-driven; ANTARES generation is data-driven, evaluation is Monte Carlo (MC).
• Case 2: Similar to Case 1, but ANTARES generation background data is randomized (uniform azimuthal distribution).
• Case 3: All datasets use the DEFT algorithm for background estimation in both generation and evaluation.
• Case 4: ARCA is data-driven; ANTARES is entirely MC-based.

The spread of results across these four cases defines the systematic uncertainty band. Histograms are constructed in terms of angular distance ($\log_{10} \alpha$), reconstructed energy ($\log_{10} E$), and, for ANTARES tracks, angular error estimation ($\log_{10} \beta$).

Key Results
The analysis assumes an $E^{-2}$ energy spectrum and a 1:1:1 flavor ratio.

• Performance Improvement: While ANTARES data significantly drives the joint analysis results, the combination with ARCA data improves performance by approximately 10% compared to the ANTARES standalone analysis.
• Candidate Source MG3 J225517+2409: This source yielded the lowest pre-trial p-value of $(1.2 \pm 0.4) \cdot 10^{-4}$, corresponding to a significance of $(3.69 \pm 0.09)\sigma$.
  • Best-fit flux: $E^2\phi = (1.39 \pm 0.02) \cdot 10^{-8} \text{ GeV cm}^{-2} \text{ s}^{-1}$.
  • Post-trial probability: Calculated via binomial formula as $0.013 \pm 0.004$ ($(2.26 \pm 0.13)\sigma$). A simulation-based approach yielded a post-trial p-value of $0.018 \pm 0.006$ ($(2.12 \pm 0.14)\sigma$).
  • A simplified post-trial estimation accounting for the initial selection from 1255 objects and subsequent catalogue reduction suggests a probability of $\approx 0.1$ for observing such a candidate.
• Other Candidates: The analysis also evaluated other known sources, including 3C403 ($p \approx 1.5 \cdot 10^{-3}$) and TXS 0506+056, providing 90% CL upper limits for a catalogue of 106 sources.
• Sky Map: The sky map near MG3 J225517+2409 shows that one ARCA21 event has an error box including the source location, with a signal likelihood 6.3 times higher than the background-only scenario.

Significance and Claims
The paper claims to demonstrate the feasibility and potential of a combined analysis between ANTARES and KM3NeT/ARCA. It establishes that while the current results are dominated by the mature ANTARES dataset, the integration of ARCA data provides a measurable improvement in sensitivity. The authors note that as KM3NeT/ARCA continues to grow in size and exposure, it is expected to reach ANTARES performance levels within a few years, making joint data analysis increasingly critical for future discoveries. The work provides a framework for handling systematic uncertainties through multiple background modeling strategies and sets upper limits for a wide range of astrophysical candidates.