SN1987A as a Scale: Pinning Down the Neutrino Mass From 19 Events

Three independent laboratory probes of the absolute neutrino mass dominate the modern programme — KATRIN’s tritium beta-decay endpoint at 0.45 eV, the cosmological sum of masses at $\sum m_{\nu }<0.12$ eV from CMB plus large-scale structure, and the half-life limits from neutrinoless double beta decay at $|m_{\beta \beta }|<0.08$ eV depending on the nuclear matrix elements. Each comes with its own assumptions and its own systematic budget.

A fourth probe, less heralded but conceptually clean, operates entirely outside the laboratory. A neutrino with rest mass $m$ travels at a velocity slightly below the speed of light, with the deficit scaling as $m^2/2E^2$. Over a cosmological propagation distance $L$, this produces a small but calculable arrival-time delay: a heavier neutrino arrives later than a lighter one of the same energy, and a low-energy neutrino arrives later than a high-energy one of the same mass. The energy-dependent spread of an astrophysical neutrino burst encodes the mass directly.

The 1987 detection of a neutrino burst from supernova SN1987A in the Large Magellanic Cloud — 19 events spread across approximately 13 seconds at Kamiokande, IMB, and Baksan — provided the first and so far only opportunity to apply this measurement to a single astronomical event. The resulting bound, approximately $m_{{\nu }_e}<5.7$ eV, was the strongest direct mass limit for years afterward and remains a clean reference for the timing technique.

This post is about how the timing extraction works, what the SN1987A data actually said, and what a future galactic supernova will deliver when one happens.

The relativistic delay

A particle with rest mass $m$ and energy $E\gg m$ travels at velocity $v=c\sqrt{1-m^2/E^2}\approx c(1-m^2/2E^2)$. The time it takes to traverse a distance $L$ exceeds the light-travel time by

$\Delta t=\frac{L}{c}\left(\frac{1}{v/c}-1\right)\approx \frac{L}{c}\cdot \frac{m^2}{2E^2}$

Plugging in convenient units — distance in kiloparsecs, energy in MeV, mass in electron-volts — the delay becomes

$\Delta t\approx 5.15\mathrm{ms}\cdot \left(\frac{L}{10\mathrm{kpc}}\right)\cdot {\left(\frac{m}{1\mathrm{eV}}\right)}^2\cdot {\left(\frac{1\mathrm{MeV}}{E}\right)}^2$

For SN1987A at $L\approx 50$ kpc and neutrinos at $E\approx 10\text{–}40$ MeV, a mass of 10 eV produces a delay of order 1-30 seconds — directly comparable to the observed burst duration. The energy dependence is crucial: the delay scales as $1/E^2$, so low-energy events arrive systematically later than high-energy events of the same mass. The energy-time correlation across the detected events is the signature one searches for.

What SN1987A actually delivered

The neutrino burst from SN1987A was detected at three sites within seconds of each other on 23 February 1987 at approximately 07:35 UT, about 3 hours before the optical brightening was first noted:

• Kamiokande II in Japan: 11 events between 0 and 13.4 seconds, energies 7.5 to 35.4 MeV
• IMB in Ohio: 8 events between 0 and 6.0 seconds, energies 19 to 39 MeV (with higher threshold than Kamiokande)
• Baksan in the Caucasus: 5 events, with weaker timing precision

Both Kamiokande and IMB were water-Cherenkov detectors whose dominant interaction at these energies is inverse beta decay, ${\bar{\nu }}_e+p\to e^{+}+n$. The detected events are therefore predominantly electron antineutrinos. The energy of each event is reconstructed from the Cherenkov ring it produced.

The 19 events at Kamiokande + IMB are the dataset from which the absolute mass bound is extracted. The exact reconstruction is technical: one fits the observed event distribution in the energy-time plane to a model that combines the predicted intrinsic burst spectrum, the predicted intrinsic burst time profile, and the propagation-time delay $L{m}^2/2E^2$. The mass enters as a single parameter that distorts the time-vs-energy correlation.

Different analyses of the same data have used slightly different burst-time-profile models — single-exponential decay, parametrized two-phase models, full hydrodynamic simulations — and arrived at slightly different mass limits. The range of published bounds typically spans $m_{{\nu }_e}<5.7$ eV (Loredo and Lamb 2002, the most careful analysis with the most thorough handling of the intrinsic burst-time uncertainty) up to roughly 30 eV (more conservative early estimates).

Why the SN1987A bound has the form it does

The dominant systematic in the SN1987A analysis is the intrinsic duration of the burst. The observed event spread is the convolution of the propagation-time delay (which depends on mass and energy) with the intrinsic burst duration (which depends on supernova physics). To extract the mass cleanly one needs an independent prediction of the intrinsic duration, and supernova-modelling uncertainties propagate directly into the mass bound.

Standard simulations of core-collapse supernovae predict an intrinsic neutrino burst lasting about 10 seconds, with the initial deleptonization burst (a few hundred milliseconds), the accretion phase (~1 second), and the cooling phase (a few seconds). The Kamiokande and IMB observations are consistent with this picture, but the precise mapping between the intrinsic supernova-emission profile and the detected event time profile carries non-trivial uncertainty.

The most careful analyses (Loredo and Lamb 2002 in particular) marginalize over the intrinsic-duration uncertainty, producing a posterior probability distribution for $m_{{\nu }_e}$ that has a long tail extending to about 5-6 eV at 95% confidence. Simpler analyses that assume a specific intrinsic-duration model give tighter limits, sometimes as low as 2-3 eV, but these depend more strongly on supernova-modelling assumptions.

The robust statement is that SN1987A’s data are consistent with massless neutrinos and require the mass to be no larger than approximately 6 eV. This is weaker than the laboratory limit from tritium beta-decay endpoint measurements (originally 5-10 eV at the time, now 0.45 eV from KATRIN), but it has the advantages of being independent of any laboratory systematic and probing the absolute mass over an astronomical baseline.

What a galactic supernova would do

The 19 events at Kamiokande and IMB were what the technology of 1987 could deliver from the LMC. A core-collapse supernova in the Milky Way, at distances of 1 to 10 kpc rather than 50 kpc, would produce hundreds to thousands of times more events thanks both to the closer distance and to the much larger detector volumes now operating. Specifically:

• Super-Kamiokande would detect about $5,000$ events at 10 kpc, dominantly through inverse beta decay.
• IceCube would detect approximately $10^6$ events as a diffuse rise in the photomultiplier rates (no individual event reconstruction, but exquisite time profile of the burst).
• KamLAND, Borexino, JUNO would each detect several hundred to several thousand inverse-beta-decay events.
• DUNE would detect about $3,000$ electron-neutrino charged-current events on argon — the first large sample sensitive to ${\nu }_e$ rather than ${\bar{\nu }}_e$.
• HALO would record approximately 30 events with two-neutron discrimination.

The improved statistics translate directly into the mass bound. For a 10-kpc supernova with 5,000 events at Super-Kamiokande, the statistical power on the energy-time correlation should reach the level of approximately $m_{\nu }\lesssim 0.5$ to $1.0$ eV at 90% confidence — comparable to KATRIN’s projected final sensitivity and to the cosmological mass-sum bound at $\sum m_{\nu }\approx 0.06$ eV (after dividing by three for the heaviest eigenstate).

The IceCube time profile, while not energy-resolved per event, provides the most precise determination of the burst onset and the early-time rise — directly constraining the intrinsic duration model that dominates the SN1987A systematic. Combined with the energy-resolved samples from the other detectors, the systematic budget should shrink to the few-percent level on the intrinsic duration.

A galactic supernova in 2030 would deliver a kinematic neutrino-mass measurement comparable to or better than the laboratory and cosmological probes — and through completely independent physics, providing a powerful cross-check on the absolute mass scale.

Why this matters

The absolute neutrino mass is one of the half-dozen most consequential numbers in fundamental physics. It enters the predictions of cosmological structure formation, the limits on neutrinoless double beta decay, the seesaw scale, the leptogenesis efficiency, and the unitarity of the PMNS matrix at high precision. Three independent probes — KATRIN’s beta-decay endpoint, the cosmological mass sum, and the limits from $0\nu \beta \beta$ — converge on a value below 0.2 eV, but each carries assumptions: KATRIN depends on molecular final-state modelling, cosmology depends on the underlying cosmological model, and $0\nu \beta \beta$ depends on neutrinos being Majorana.

A supernova-timing measurement adds a fourth probe that depends on none of those assumptions. It depends instead on supernova-emission modelling and on detector timing precision — both areas where the systematic budget is being actively reduced. The next galactic supernova, expected somewhere in the range from “any moment now” to “another century or two,” will be one of the cleanest absolute-mass measurements available — and it will happen as a single event of enormous astrophysical and particle-physics value.

Summary

A neutrino with rest mass $m$ and energy $E$ arrives later than a massless particle of the same energy by $L{m}^2/2E^2$ over a propagation distance $L$. The energy-dependent delay of the SN1987A neutrino burst, spread across 13 seconds at Kamiokande and IMB after travelling 50 kpc, allowed an absolute-mass bound of approximately $m_{{\nu }_e}<5.7$ eV at 95% confidence — the strongest direct astrophysical limit at the time, independent of any laboratory systematic. The dominant remaining uncertainty was the intrinsic supernova burst duration. A future galactic supernova at 1-10 kpc, detected across Super-K, IceCube, JUNO, DUNE, KamLAND, Borexino and HALO, will deliver thousands to tens of thousands of events with the IceCube time profile pinning down the intrinsic duration directly. The projected mass sensitivity from timing alone reaches below 1 eV, comparable to or better than KATRIN’s final reach. The next supernova in our galaxy will provide one of the cleanest absolute-mass measurements in the field — through physics that is entirely complementary to laboratory and cosmological approaches.