Status of the KM3NeT real-time analysis framework

The paper describes the advanced commissioning status of the KM3NeT real-time analysis framework, which enables low-latency event reconstruction, multi-messenger alert follow-up, supernova burst monitoring, and autonomous cosmic neutrino alerts for its ARCA and ORCA detectors.

The Gist

Imagine the universe as a giant, dark ocean. Most of us look at the stars with telescopes that catch light (photons), like looking at the surface of the water. But some of the most violent events in the universe—like exploding stars or colliding black holes—send out invisible messengers called neutrinos. These particles are like ghostly fish that swim right through the Earth without stopping.

To catch these "ghost fish," scientists built KM3NeT, a massive underwater telescope in the Mediterranean Sea. It's not made of glass lenses, but of thousands of light-sensors (called Digital Optical Modules) hanging on long cables deep underwater. When a neutrino hits a water molecule, it creates a tiny flash of blue light (Cherenkov light), which the sensors catch.

This paper is a status report on the "brain" of this telescope: a real-time computer system designed to react instantly to what it sees. Here is how that system works, broken down into simple parts:

1. The Two Eyes of the Telescope

KM3NeT has two main "eyes" (detectors) looking at different things:

• ARCA: This eye is tuned to spot very energetic, high-speed neutrinos (like a shark spotting a fast-moving tuna).
• ORCA: This eye is tuned to spot slightly slower, lower-energy neutrinos (like a dolphin spotting a school of smaller fish).
• Bonus: Both eyes can also spot a specific type of "faint glow" from exploding stars (supernovae) happening right now in our galaxy.

2. The "Speedy Secretary" (The Real-Time Framework)

The data from these underwater sensors is huge. The paper describes a new software system that acts like a super-fast secretary.

• The Job: As soon as the sensors see a flash, this secretary grabs the data, figures out where it came from, and decides if it's important.
• The Speed: It does this incredibly fast—taking less than 10 to 15 seconds to process a single event.
• The Filter: The ocean is full of "noise" (like regular waves or fish swimming by). The secretary uses smart computer tricks (machine learning) to ignore the noise and only flag the interesting "ghost fish."

3. Checking the Guest List (Following External Alerts)

Sometimes, other observatories (like those watching for gravitational waves or gamma-ray bursts) shout, "Hey, something big just happened over there!"

• The Process: Since June 2023, the KM3NeT secretary has been checking over 3,500 of these shouts. It looks at the data from that specific time and place to see if KM3NeT saw a matching neutrino.
• The Result: So far, the secretary has checked the guest list and found no matches. The number of "almost matches" is exactly what you'd expect by pure luck (background noise). This is normal; it means the system is working correctly and isn't crying wolf.

4. The Supernova Alarm Clock

If a star explodes in our galaxy, it sends out a massive burst of neutrinos in less than half a second.

• The System: The KM3NeT system has a special "alarm clock" that watches for a sudden spike in tiny flashes.
• The Goal: If it sees a spike, it immediately sends a warning to the global SNEWS network (the Supernova Early Warning System), telling astronomers, "Look up! A star is dying right now!"
• Status: This alarm is ready and running, with a more detailed version currently being tested.

5. The "Alert Sender" (Finding New Neutrinos)

This is the most exciting part. The system doesn't just wait for others to shout; it can shout on its own.

• The Hunt: It constantly scans for high-energy neutrinos that look like they came from deep space (not from our atmosphere).
• The "False Alarm" Check: To make sure it doesn't send a fake alert, the system uses a clever math trick. It compares the event to millions of simulated "fake" events. It calculates a "Hyper FAR" (a fancy way of saying: "How likely is it that this is just a coincidence?"). If the chance of it being a coincidence is less than once a month, it's a real alert.
• The Map: Once an alert is confirmed, the system instantly draws a map of the sky showing where the neutrino came from. It then cross-references this map with a "phone book" of known cosmic objects (like black holes or active stars) to suggest, "Hey, look at this specific object."
• The Timeline: This whole process—from seeing the flash to sending a map to other telescopes—takes about 3 minutes.

The Bottom Line

The KM3NeT real-time system is currently in the "advanced dress rehearsal" phase. It is successfully processing data, checking other astronomers' alerts, and monitoring for exploding stars.

The team expects to start sending official alerts to the rest of the world's observatories by summer 2026. Once fully operational, KM3NeT will be a key player in the global team that watches the universe in real-time, helping us understand the most energetic events in the cosmos.

Technical Summary

Based on the provided paper, here is a detailed technical summary of the status of the KM3NeT real-time analysis framework.

1. Problem Statement

Multi-messenger astronomy requires rapid, coordinated responses to transient cosmic events (e.g., gravitational waves, gamma-ray bursts, supernovae) to maximize scientific return. Traditional neutrino telescopes often suffer from high latency or lack the capability to autonomously issue alerts. The challenge is to develop a system capable of:

• Processing massive data streams from deep-sea detectors with low latency.
• Distinguishing rare cosmic neutrino signals from the dominant background of atmospheric muons.
• Autonomously identifying significant events and distributing alerts to partner observatories within minutes.
• Following up on external alerts from other multi-messenger channels to search for neutrino counterparts.

2. Methodology and System Architecture

The KM3NeT Collaboration has developed a dedicated real-time analysis framework that processes data from two deep-sea detectors in the Mediterranean Sea: ARCA (optimized for TeV–PeV neutrinos) and ORCA (optimized for GeV–TeV neutrinos). Both detectors utilize Digital Optical Modules (DOMs) containing 31 photomultiplier tubes (PMTs) arranged in Detection Units (DUs).

The framework operates through four primary pipelines:

A. Real-Time Event Reconstruction and Classification

• Data Mirroring: Data from ARCA and ORCA are continuously mirrored to dedicated dispatchers.
• Reconstruction: Events are reconstructed under both "track-like" (muons) and "shower-like" (cascades) hypotheses.
• Background Suppression: Machine-learning algorithms are applied to suppress atmospheric muon backgrounds.
• Latency: The total processing latency is kept below ~10 seconds for ARCA and ~15 seconds for ORCA.

B. External Alert Follow-up

• Trigger Handling: Since June 2023, the system has processed over 3,550 external alerts (GRBs, GWs, FRBs, etc.).
• Analysis Technique: A binned ON/OFF technique is used. The "ON" region is a space-time window around the alert, while "OFF" regions are control bands following the Earth's rotation to estimate background.
• Optimization: Selection cuts are optimized based on elevation to balance background suppression and statistical uncertainty. Results are reported as pre-trial p-values.

C. Core-Collapse Supernova (CCSN) Detection

• Signal: Detection of MeV-scale electron anti-neutrinos via inverse beta decay.
• Method: Since individual MeV tracks are too short for reconstruction, the system monitors excesses of hit coincidences within single DOMs.
• Pipelines:
  1. Real-time: Monitors coincidence levels and alerts the SNEWS 2.0 network upon significant excess.
  2. Quasi-online: Performs detailed analysis on accumulated statistics to reconstruct light curves and arrival times (currently in testing).
  3. Triggered Follow-up: Re-analyzes archival data upon external CCSN triggers.

D. Autonomous Neutrino Alert System

• Selection Strategies:
  1. High-Energy Selection: Targets well-reconstructed, upgoing, high-energy track-like events.
  2. Multiplet Selection: Targets pairs of events from compatible sky directions within short time windows (indicating flaring sources).
• False Alert Rate (FAR) Calculation:
  • Integral FAR: Counts background events with more extreme parameters in multi-dimensional space.
  • Hyper FAR: Introduced to solve degeneracies in the selection space. It counts background events with an Integral FAR equal to or lower than the candidate, effectively reducing the problem to one dimension to control the physical alert rate.
• Sky Localization: Uses on-the-fly simulations (re-sampling detected events) to generate HEALPix probability sky maps and containment radii (50%, 68%, 90%) with a latency of 1.5–2 minutes.
• Counterpart Identification: An "astro module" cross-correlates sky maps with astrophysical catalogs (e.g., 2RXS, 4FGL-DR4) to rank candidate sources based on probability density, brightness, and variability.

3. Key Contributions

• Low-Latency Framework: Successfully implemented a system with sub-15-second processing latency for event reconstruction and classification.
• Advanced Statistical Metrics: Developed the Hyper FAR metric to ensure a controlled, predictable alert emission rate, overcoming limitations of multi-dimensional background estimation.
• Automated Multi-Messenger Integration: Created a fully automated pipeline that not only follows up external alerts but also autonomously generates and distributes GCN notices for internal discoveries.
• Real-Time CCSN Monitoring: Established a multi-tiered pipeline for detecting supernova bursts, including integration with the SNEWS 2.0 network.
• Astrophysical Prioritization: Integrated a module that automatically ranks potential electromagnetic counterparts to facilitate rapid follow-up observations.

4. Results

• External Alerts: As of the report, 3,550 external alerts have been analyzed. No significant neutrino counterparts have been observed. The number of analyses exceeding $2\sigma$, $2.5\sigma$, and $3\sigma$ thresholds is consistent with background expectations.
• CCSN Sensitivity: Current configurations are sensitive to the vast majority of potential CCSN candidates in the Milky Way. Full configuration (ARCA230/ORCA115) will achieve complete Galactic sensitivity.
• Alert System Status: The system is in advanced commissioning. The expected alert rate is 1–2 events per month.
• Timeline: Public distribution of GCN notices is scheduled to begin by Summer 2026.

5. Significance

The KM3NeT real-time analysis framework represents a major milestone in the global multi-messenger network. By combining a wide field of view, nearly continuous duty cycles, and autonomous low-latency processing, KM3NeT is transitioning from a data-taking instrument to an active participant in real-time astronomy.

• Coordination: It enables rapid, coordinated observations with partner observatories (gamma-ray, gravitational wave, and radio telescopes).
• Discovery Potential: The ability to autonomously identify and distribute alerts for high-energy cosmic neutrinos and supernova bursts significantly increases the probability of capturing transient, fast-fading astrophysical phenomena.
• Future Outlook: As the detector expands toward its full configuration, KM3NeT is poised to become a leading facility for real-time neutrino astronomy and multi-messenger discovery.