GRBs and the Neutrino That Wasn't There: IceCube's Stacking Limits

When IceCube began collecting data in 2008, one of its principal astrophysical targets was the gamma-ray burst (GRB) — the briefly brightest objects in the universe, releasing $10^{51}$ to $10^{54}$ erg of energy in seconds. The theoretical motivation was elegant: GRBs are powered by relativistic jets, and internal shocks within those jets had been calculated to accelerate protons to ultra-high energies. If those protons interacted with the dense radiation field of the burst, pion production would seed a measurable high-energy neutrino flux through exactly the kind of resonant process that produces cosmogenic neutrinos in the intergalactic medium. Eli Waxman and John Bahcall’s calculation in 1997 estimated that GRBs producing the observed ultra-high-energy cosmic-ray (UHECR) flux would generate enough neutrinos to be detectable in the kind of cubic-kilometer detector IceCube was planned to be. By the time IceCube was running, the GRB neutrino flux was expected to be the first astrophysical detection.

It wasn’t. Fifteen years and several major stacking analyses later, IceCube has placed the strongest constraint on the GRB contribution to the diffuse astrophysical neutrino flux: less than 1 to 5 per cent over most of the relevant energy range. Classical GRBs are no longer the leading candidate origin of UHECRs. The null result is one of the most informative non-detections in modern astroparticle physics.

This post is about why GRBs were such a natural candidate, how stacking analyses work, what IceCube actually saw — and didn’t — and what the result implies for the still-unresolved UHECR origin problem.

The Waxman-Bahcall picture

In the late 1990s, observational evidence had converged on GRBs as the brightest cosmic-ray sources in the universe. Their isotropic distribution on the sky, with thousands of events per year, ruled out a Galactic origin. The observed energy fluence at Earth — typically $10^{-5}$ erg/cm² per burst, integrated over Swift’s lifetime — translates into an enormous total radiated energy when scaled by their cosmological distances.

The leading model of the time, due to Mészáros and Rees and refined by many others, pictured an ultrarelativistic jet launched by the collapse of a massive star (long GRBs) or by a binary neutron-star merger (short GRBs). Within the jet, internal shocks between fast and slow shells of ejecta were thought to accelerate both electrons (radiating the observed gamma rays via synchrotron emission) and protons (potentially to ultra-high energies).

The hadronic picture was attractive because it tied two unresolved astrophysical puzzles together. GRBs produce enormous energy in a small volume; UHECRs require enormous energy in a small volume. If the same internal shocks producing the observed gamma rays also produced ultra-high-energy protons, then GRBs explained UHECRs.

Waxman and Bahcall went further. They calculated that if the protons producing UHECRs in GRBs interact with the dense photon field of the burst — primarily the synchrotron photons from the radiating electrons — then the $p\gamma \to {\Delta }^{+}\to n{\pi }^{+}$ resonance would produce pions at a rate determined by the proton energy distribution, the photon density, and the photoproduction cross-section. The pions decay through the chain

${\pi }^{+}\to {\mu }^{+}+{\nu }_{\mu },{\mu }^{+}\to e^{+}+{\nu }_e+{\bar{\nu }}_{\mu }$

producing high-energy neutrinos that escape from the jet’s optically thin region. Working through the kinematics for typical GRB parameters, they predicted a per-burst neutrino flux at the level of a few events per cubic kilometer of detector. Over the cosmological GRB rate of about one detected burst per day, this added up to a diffuse flux at the few-per-cent level of what IceCube would eventually see as the all-sky astrophysical flux.

The Waxman-Bahcall bound — the upper limit on the neutrino flux from sources that also produce the observed UHECR flux — became the benchmark prediction for the late 2000s and early 2010s. IceCube’s first targeted searches were designed against it.

How stacking works

A single GRB is far too weak for IceCube to detect through its individual neutrino flux. Even a Waxman-Bahcall-strong burst produces perhaps $10^{-3}$ events in the detector during its prompt-emission window. The expected non-detection rate per burst is essentially 100%, even for sources that contribute meaningfully to a stacked signal.

A stacking analysis exploits the well-localized direction and time of each burst as separately known input. The technique is straightforward in principle:

1. For each GRB in a catalogue, define a spatial search window around the reported direction and a temporal search window around the reported start time.
2. Count IceCube events that fall within both windows.
3. Compute the expected background rate from off-source data — events in nearby but un-correlated regions of sky.
4. Compare the on-source count to the background expectation and assess the statistical significance.

When summed over a large catalogue of GRBs — Swift’s BAT triggers, Fermi-GBM’s wide-field detections — the statistical leverage adds up. A stacked sample of 1,000 GRBs has roughly $\sqrt{1000}\approx 30$ times the sensitivity of a single-burst search to a signal that produces a constant per-burst rate.

The dominant technical challenges are:

• Background estimation: Most events in any given direction at any given time are atmospheric muons or atmospheric neutrinos. The background rate per square degree per second is precisely calibrated from off-source data.
• Source weighting: Each burst contributes differently to the stacked sensitivity depending on its energy fluence, declination (which controls atmospheric background levels), and time. Optimal stacking weights each burst by its expected signal-to-background ratio.
• Window selection: The spatial window has to be large enough to contain the signal given IceCube’s angular resolution (about 1° for through-going muons) but not so large as to dilute the signal in background.
• Energy threshold: The Waxman-Bahcall spectrum is hardest at high energies, so the analysis is typically performed in multiple energy bins to optimally probe different parts of the prediction.

What IceCube saw

IceCube’s GRB stacking has been performed in several successive analyses as the burst catalogue and detector sensitivity have grown.

The first 40-string IceCube analysis in 2011 covered 117 Swift-localized bursts and reported no significant excess, setting an upper limit at the level of about half the Waxman-Bahcall prediction.

By 2017, with the full 86-string IceCube and an expanded catalogue of 1,172 GRBs from both Swift and Fermi-GBM, the limit had improved to roughly 1% of the original Waxman-Bahcall prediction over the energy range 100 TeV to 10 PeV. The result was clear: classical GRB internal-shock models predicting hadronic neutrino emission at the Waxman-Bahcall level were excluded.

The most recent 2024 stacking analysis included nearly 2,300 GRBs and pushed the prompt-emission neutrino bound below 1% of the diffuse astrophysical flux’s GRB-attributable portion across most of the relevant energy spectrum.

Throughout, no individual GRB has produced a significant single-burst detection either. The dispersed neutrinos that should have been arriving in concert with the gamma rays have not appeared at the predicted rate.

What the null result means

The implications fall into several categories.

Classical hadronic-jet GRBs are excluded as the dominant UHECR source. If GRBs produce the observed UHECR flux through internal-shock proton acceleration with photoproduction-mediated pion production, the corresponding neutrino flux must show up at the Waxman-Bahcall level. It doesn’t, so either GRBs are not producing the UHECRs through that mechanism, or some essential ingredient of the model is wrong.

Internal-shock proton acceleration efficiency must be substantially reduced. Several modifications can save GRBs as some kind of cosmic-ray source while evading the neutrino constraint: lower internal-shock baryonic content, alternative acceleration sites (external shocks at much later times), or proton escape from the jet that bypasses the dense photon field. None of these is impossible, but each requires specific tuning to avoid the neutrino prediction.

Pre-supernova hadronic emission (“choked jets”) remains viable. A small fraction of GRBs may be relativistic jets that fail to break through the stellar envelope of a collapsing massive star. The jet then dissipates inside the star, producing a strong neutrino signal but no associated gamma rays — invisible to GRB-trigger-based stacking analyses because the GRB itself does not occur. Such “choked jets” would explain part of the diffuse astrophysical neutrino flux without showing up as GRB-correlated. They are an active area of theoretical work and an attractive complement to the visible-GRB constraint.

Alternative UHECR sources are favored. Active galactic nuclei (including the steady NGC 1068 source identified by IceCube), starburst galaxies, tidal disruption events, and magnetars have all been proposed as alternatives. None has been definitively confirmed as the dominant UHECR source, but the GRB exclusion is part of what has shifted the field toward exploring these other possibilities.

Multi-messenger constraints from the merger sample

Short GRBs from compact-object mergers — the population that produced GW170817 — provide an additional class of constraint. IceCube’s targeted search for neutrinos coincident with GW170817 yielded no detection, with limits substantially below the most optimistic merger-jet predictions. The smaller sample of well-localized short GRBs limits the statistical reach of dedicated short-GRB stacking, but the constraints already disfavour short-GRB jets as a substantial UHECR source on top of the broader GRB exclusion.

The 2024 LIGO-Virgo binary-merger detections, combined with continuing IceCube monitoring, will progressively tighten the short-GRB neutrino bounds over the rest of the decade.

What might still hide

Although the constraints are strong, several configurations of GRB hadronic emission remain at the edge of current sensitivity.

Late-time external-shock emission, when the relativistic ejecta sweep up the surrounding medium hours to days after the prompt phase, could produce a neutrino flux delayed and dispersed enough that prompt-emission stacking misses it. Dedicated searches for late-time neutrinos coincident with afterglow observations have begun to address this.

High-luminosity bursts from the early universe, observed at high redshift where IceCube’s sensitivity is highest in the source frame, could contribute disproportionately to the stacked signal but produce neutrinos that arrive over delayed timescales. The energy spectrum of any such contribution would be redshifted in a calculable way.

Population sub-classes with distinct spectral or temporal characteristics may evade the catalogue-average bound. Studies focusing on bright bursts, on long-duration bursts, or on specific gamma-ray spectral types continue to refine the picture.

Summary

Gamma-ray bursts were once the leading candidate origin for ultra-high-energy cosmic rays, with Waxman and Bahcall’s 1997 calculation predicting a detectable high-energy neutrino flux from the same internal-shock proton acceleration that powers the UHECRs. IceCube’s GRB stacking analyses over the past 15 years have combined data from thousands of GRBs and consistently found no significant excess, placing limits at less than 1-5 per cent of the original Waxman-Bahcall prediction across most of the relevant energy range. The classical internal-shock hadronic emission model is therefore excluded. GRBs remain a possible UHECR source through modified mechanisms — choked jets that hide the visible burst, alternative acceleration sites, or late external-shock emission — but they have lost their position as the leading candidate. Alternative sources, including AGN, starburst galaxies and tidal disruption events, have moved up in the candidate ranking, with NGC 1068 already established as one identified contributor. The GRB null result is one of the most informative non-detections in astroparticle physics, having moved the field decisively away from its leading 1990s hypothesis.