SNEWS: The Network That Will Tell Us When the Next Star Goes Off

When the next core-collapse supernova occurs in our Galaxy, the first signal won’t be visible light. It will be a few-second burst of neutrinos arriving at underground detectors worldwide, hours before the supernova’s electromagnetic emission punches through the stellar envelope and becomes visible to optical telescopes.

That neutrino burst is short — under ten seconds. It will deliver tens of thousands of events to Super-Kamiokande, hundreds of thousands to Hyper-Kamiokande when it comes online, and tens of thousands across the global network of underground detectors. The detection, by itself, will be a stunning physics result: the most data-rich supernova event ever observed, by orders of magnitude, compared to SN 1987A.

But there’s a coordination problem. The detectors are scattered across multiple continents and operated by independent collaborations. Each sees the burst slightly differently — different angular coverage, different energy thresholds, different latencies. To produce a single coherent alert and to enable optical observatories to point at the supernova before it becomes visible, the detectors have to talk to each other in real time.

The solution is the Supernova Early Warning System — SNEWS. Established in the early 2000s and operated since 2005, SNEWS is a software-and-protocol layer connecting the world’s underground neutrino detectors. When two or more independent detectors report supernova-like signals within a brief coincidence window, SNEWS issues a global alert. Optical observatories receive the alert within minutes. The next galactic supernova, whenever it occurs, will be the first that humans saw coming.

This is what SNEWS is, how it works, and why it matters.

Why we get neutrino warning at all

The mechanism is straightforward. A core-collapse supernova begins when a massive star’s iron core, no longer able to extract energy from fusion, collapses gravitationally. Within milliseconds, the core compresses to nuclear densities. The collapse produces an enormous burst of neutrinos through electron capture (turning protons into neutrons and emitting electron neutrinos), reaching luminosities of about $10^{53}$ erg/s — orders of magnitude greater than the entire electromagnetic luminosity of the Sun integrated over all of cosmic history.

The neutrinos escape the collapsing star almost immediately because they interact so weakly. Estimates put the escape time at a few seconds — roughly the light-crossing time of the proto-neutron-star surface (~50 km).

The visible explosion takes much longer. The shock wave that produces the supernova has to traverse the stellar envelope — typically several solar radii thick for a red-supergiant progenitor. Even at the shock speed of ~10,000 km/s, this takes about an hour to traverse a 10-solar-radius envelope. For a more massive progenitor with thicker envelope, several hours are typical.

So the neutrinos arrive at Earth several hours before the optical light. This time gap is the warning.

For SN 1987A, the gap was about three hours. The neutrinos were detected at Kamiokande, IMB, and Baksan at 7:35 UT on 23 February 1987. The optical brightening, when it became visible the following morning, traced back to the same time but accounting for the intervening atmosphere — about three hours later than the neutrino arrival.

The detector network

SNEWS coordinates a global network of underground neutrino detectors. Membership has evolved as new detectors come online and old ones decommission. The current (2026) participants:

Super-Kamiokande (Japan, Kamioka mine, 50 kt water Cherenkov). Sensitive primarily to ${\bar{\nu }}_e$ via inverse beta decay. Expected event count for galactic supernova at 10 kpc: ~7,500.

IceCube (Antarctica, South Pole, 1 km³ ice Cherenkov). Detects burst-induced PMT count-rate increases — does not reconstruct individual events but accumulates statistical excess. Expected response at 10 kpc: ~10⁶ excess hits across the array, about 100 sigma significance.

KamLAND (Japan, Kamioka mine, 1 kt liquid scintillator). Sensitive to ${\bar{\nu }}_e$ inverse beta decay and ν-e elastic scattering. Expected events at 10 kpc: ~330.

LVD (Italy, Gran Sasso, 1 kt liquid scintillator). Similar to KamLAND but optimized for short-burst response. Expected events: ~120.

HALO (Canada, Sudbury, lead target). Sensitive to electron neutrinos via charged-current scattering on lead. Distinctive complement to water and scintillator detectors. Expected events: ~45.

Daya Bay (China, reactor neutrino experiment). Sensitive to ${\bar{\nu }}_e$. Expected events: ~110.

NOvA (USA, Fermilab beam experiment). Sensitive to multiple flavors. Expected events: ~50.

KM3NeT (Mediterranean, partial). Sensitivity scales with deployment status; eventually 200+ events.

Borexino (Italy, decommissioned 2021). Was a key SNEWS member through its operating period.

The detectors collectively cover the full sky (some always have the supernova above the horizon; others will see it through the Earth) and span multiple flavor channels (water Cherenkov sees mostly ${\bar{\nu }}_e$; HALO sees ${\nu }_e$; argon detectors see both). This complementarity will be crucial for multi-flavor analysis of an actual burst.

How the alert works

Each detector runs a “supernova trigger” algorithm in real time. The trigger looks for an unusually high event rate within a short time window (typically 10 seconds). If the rate exceeds a threshold corresponding to a few sigma above background, the detector generates an alert message containing:

• Detector identity
• Approximate event count
• Timestamp (synchronized to UTC)
• Confidence level

The alert message is automatically transmitted to the SNEWS coordinator (a central server, currently at Brookhaven). The coordinator looks for coincidences — two or more independent detectors triggering within the same time window (currently set to about 10 seconds, accommodating relative latencies and clock synchronisation).

If a coincidence is found, the coordinator issues a global alert. The alert goes to:

• Optical observatories worldwide
• Gamma-ray observatories
• The amateur astronomy community via standard astronomical alert networks
• Internal collaborations of all SNEWS member experiments

The alert latency is typically a few minutes from the actual neutrino burst — set primarily by detector trigger latencies and message propagation. For a galactic supernova, this leaves several hours before the optical light arrives, more than enough for telescopes to slew to the predicted position.

Direction reconstruction

The trickier question is: where in the sky is the supernova? Most underground detectors are not directional. Inverse beta decay events have only weak forward asymmetry. Cherenkov detectors can do somewhat better through electron scattering on neutrinos, but the directional information is limited.

The combined approach: use multiple detectors and triangulate from their relative timing. A neutrino burst from a galactic source arrives at different detectors at slightly different times — the difference depends on the detectors’ relative positions and the source direction. With enough detectors at sufficient geographical separation, the source position can be triangulated to roughly 5-10 degrees precision.

For SN 1987A, the timing difference between Kamiokande (Japan), IMB (Ohio), and Baksan (Caucasus) was consistent with the LMC source position — though by then the optical counterpart was already known.

For a future galactic supernova, the SNEWS triangulation will provide a few-degree-precision direction even if the optical counterpart is still hidden inside the stellar envelope. Combined with the multiple-flavor cross-check from multiple detector technologies, the resulting picture will be vastly richer than SN 1987A.

Failure modes

SNEWS has been running since 2005 and has issued approximately a dozen “alerts” over that period, all of them either calibration tests or false positives from individual detectors. None has been a real coincidence at the level needed for a real galactic supernova.

The system has had failures. In 2008, a NOvA-like detector calibration accidentally generated a coincidence with another detector’s background fluctuation, producing a false alert. The protocol was updated to require higher significance from each detector and to add more sophisticated cross-checks.

In 2018, IceCube’s data acquisition glitch triggered an internal supernova alert that was caught before being transmitted to the global system. The detector’s internal monitoring was upgraded as a result.

These failures are not embarrassing in the operational sense; they are reminders that the system is in a constant test mode, and that the “real” alert will not look exactly like any of the test alerts. The current system is conservative — designed to minimise false positives, accepting that some real signals might be missed if they don’t trigger multiple detectors simultaneously.

What a real burst would look like

When the next galactic supernova occurs, the SNEWS sequence will play out approximately as follows:

T+0: Neutrinos arrive at all SNEWS detectors essentially simultaneously (within ~30 milliseconds for typical galactic distances).

T+5 to T+30 seconds: Each detector’s trigger algorithm processes the burst and generates a candidate alert. Latency varies by detector but typically falls in this range.

T+1 to T+5 minutes: Alerts arrive at the SNEWS coordinator. Coincidence is detected. Global alert is issued.

T+5 to T+30 minutes: Alert reaches optical observatories. Astronomers begin pointing telescopes toward the predicted region.

T+1 to T+10 hours: Optical light from the supernova reaches Earth. The supernova becomes visible.

This sequence has not yet occurred for a real galactic supernova. The coordinated alert will be the first event in human history where a star’s death is announced before it is seen.

Looking ahead

Hyper-Kamiokande’s commissioning in 2027 will join SNEWS. With ~75,000 expected events at 10 kpc, Hyper-K alone will provide enormous statistical leverage for the burst. JUNO’s commissioning in 2026 adds another ~15,000 expected events at 10 kpc with excellent energy resolution. DUNE, with its ${\nu }_e$-channel sensitivity through ⁴⁰Ar, will provide unique flavor information not available from water Cherenkov detectors.

By 2030, the total expected event count for a galactic supernova at 10 kpc will be well over 100,000 across the SNEWS network. The fidelity of the resulting science — neutrino spectrum vs. time, mass ordering through matter effects, neutrino-driven nucleosynthesis constraints, neutron-star equation of state — will be vastly beyond anything possible with SN 1987A’s 24 events.

A galactic supernova is, statistically, overdue. Whether the next one occurs tomorrow or in a century, SNEWS is ready. The protocol has been tested. The detectors are calibrated. The astronomical community knows how to receive and act on the alert.

Whoever is watching SNEWS when it triggers — the first real burst — will be witness to a uniquely ordered moment: a star, dying somewhere in our galaxy, sending its first message ahead of itself, in particles whose existence Pauli postulated in 1930 and we have learned since to read.