Limits on Neutrino Emission from GRB 221009A from MeV to PeV Using the IceCube Neutrino Observatory

On 2022 October 9 at 13:16:59.99 UT, the Gamma-ray Burst Monitor (GBM) on board the Fermi satellite triggered and located the exceptionally bright long-duration gamma-ray burst (GRB) GRB 221009A (Lesage et al. 2022; Veres et al. 2022). This source was also observed by Swift (Kennea et al. 2022; Dichiara et al. 2022), Fermi-LAT (Bissaldi et al. 2022), INTEGRAL (SPI-ACS), and Konus-Wind and triangulated by the InterPlanetary Network (Svinkin et al. 2022). A highlight among the many multiwavelength and multimessenger follow-up searches ⁶⁸  of GRB 221009A has been the first detection of γ-rays above an energy of 10 TeV from a GRB as observed by LHAASO (Huang et al. 2022).

The GBM light curve consists of an initial 17 s long pulse after the GBM trigger (T0), followed by an extraordinarily bright episode starting 221.1 s after T0. The duration (T90) that contains the central 90% of emission from the GRB has a length of 325.8 ± 6.8 s. Given that this GRB saturated the Fermi-GBM instrument, it is possible that this is an overestimation of the T90. The time-integrated energy flux in the range 10–1000 keV is reported as (2.912 ± 0.001) × 10⁻² erg cm⁻² with a 1.024 s peak photon flux at the level of (2385 ± 3) cm⁻² s⁻¹, making this the most fluent and intense GRB detected by Fermi-GBM (Lesage et al. 2022).

The GRB was also localized using measurements from the Swift Observatory at R.A. α = 2882645 and decl. δ = + 197735. The 90% containment of this measurement, as provided by Swift, is 061 (Dichiara et al. 2022). The redshift z ≃ 0.151 of the GRB has been inferred from afterglow emission observed with X-SHOOTER of the Very Large Telescope 11.55 hr after T0 (de Ugarte Postigo et al. 2022), corresponding to a luminosity distance of about D_L ≃ 740 Mpc (Workman et al. 2022).

Long-duration GRBs are triggered by the death of massive stars. The gravitational and/or magnetic energy that is released in these events is the principal power for the observed bright γ-ray display and has long been speculated to facilitate cosmic-ray acceleration (Vietri 1995; Waxman 1995). The interaction of these cosmic rays with ambient radiation and/or matter could yield broadband neutrino emission that, depending on the mechanism, can even precede the electromagnetic emission. The IceCube Observatory has already reported upper limits on neutrino emission of GRB 221009A from a fast-response analysis (FRA; Thwaites et al. 2022) searching for TeV–PeV neutrino emission in the periods [T0 − 1 hr, T0 + 2 hr] and T0 ± 1 day. An independent search by the KM3NeT Observatory did not find significant neutrino emission in the period [T0 − 50 s, T0 + 5000 s] either (Dornic 2022).

In this Letter, we report the results of four searches for neutrino emission coincident with GRB 221009A using data collected with the IceCube Observatory. These searches cover complementary energy ranges from a few MeV up to the PeV scale, probing neutrino emission from the burst core to high-energy neutrinos produced in external shocks during the GRB.

One of the leading models for the GRB prompt emission involves a hot and dense plasma ("fireball") that is initially opaque to radiation and expands under its pressure. The kinetic energy of the resulting highly relativistic outflow is dissipated via internal shocks, which are expected to form from variations of the bulk Lorentz factor (Rees & Meszaros 1994; Kobayashi et al. 1997; Daigne & Mochkovitch 1998), or via magnetic reconnection in Poynting-flux-dominated outflows (Meszaros & Rees 1997). In both cases, electrons are accelerated and produce the bright γ-ray prompt display via synchrotron emission in magnetic fields.

Protons or heavier nuclei entrained in the outflow can also be accelerated by these mechanisms and produce high-energy TeV–PeV neutrinos on collisions with background photons (Waxman & Bahcall 1997; Becker et al. 2006; Murase & Nagataki 2006; Hummer et al. 2012; Bustamante et al. 2015; Biehl et al. 2018; Pitik et al. 2021; Ai & Gao 2023; Liu et al. 2023; Rudolph et al. 2023). Prior to the prompt emission, internal collisions of protons and neutrons below the photosphere have been considered as a source of GeV neutrino emission (Bahcall & Meszaros 2000; Bartos et al. 2013; Murase et al. 2013, 2022). Furthermore, reverse shocks that form during the afterglow phase have been considered sites for efficient cosmic-ray acceleration and would yield extremely high energy neutrinos in the EeV energy range from interactions with optical−UV photons (Waxman & Bahcall 2000; Murase 2007; Thomas et al. 2017; Murase et al. 2022; Zhang et al. 2022). Detection of these neutrinos would be a compelling indicator of (ultra−)high-energy cosmic-ray acceleration in GRBs.

GRBs could also be promising sites of production of thermal neutrino distributions through different mechanisms, for example, in the standard core-collapse processes, where neutrinos would arrive before the shock breakout (Kistler et al. 2013). Neutrino-dominated accretion flows (Liu et al. 2017; Qi et al. 2022) could form in the case of a supernova with jets, producing neutrinos in the MeV energy range that would arrive either prior to the GRB (Wang & Meszaros 2007; Liu et al. 2016; Wei et al. 2019) or during the γ-ray emission (Liu et al. 2016). Finally, the fireball could also be a potential site of neutrino emission with thermal spectra (Halzen & Jaczko 1996), where these neutrinos are predicted to arrive shortly before the photons in a millisecond burst.

Which (if any) of these mechanisms leads to detectable neutrino emission is uncertain, so it is imperative to consider different time windows and energy spectra of GRB neutrino emission in dedicated searches. Previous searches of high-energy neutrino emission carried out by IceCube have so far yielded upper limits that imposed constraints on optimistic neutrino emission models (Abbasi et al. 2009a, 2012a, 2022a; Aartsen et al. 2015, 2016a, 2017a; Abbott et al. 2017); see also Lucarelli et al. (2022).

The IceCube Observatory (Aartsen et al. 2017b) is a multicomponent facility located at the geographic South Pole. Its main component consists of an in-ice array that utilizes 1 km³ of the deep ultraclear glacial ice as its detector medium. The fiducial volume is instrumented with 5160 digital optical modules (DOMs; Abbasi et al. 2009b), each hosting one downward-facing photomultiplier tube (PMT), that register the Cherenkov light emitted by relativistic charged particles passing through the detector (Abbasi et al. 2010). The DOMs are distributed on 86 readout and support cables ("strings") and are deployed between 1.45 and 2.45 km below the surface. Most strings follow a triangular grid with a width of 125 m, evenly spaced over the volume.

Eight strings are placed in the center of the array and are instrumented with a denser DOM spacing and typical interstring separation of 55 m. They are equipped with PMTs with 35% higher quantum efficiency. These strings, along with the nearest layer of the surrounding standard strings, form the DeepCore low-energy subarray (Abbasi et al. 2012b). While the original IceCube array has a neutrino energy threshold of about 100 GeV, the addition of the denser infill lowers the energy threshold to about 1 GeV. An additional component of IceCube is the surface air shower array, IceTop (Abbasi et al. 2013), whose data were not used in the analyses presented in this Letter.

Neutrino interactions with matter in the vicinity of the in-ice array create energetic charged particles that are visible via Cherenkov emission. Several triggers are active in IceCube that are designed for the selection of these and other interesting physics events. The principal challenges of astrophysical neutrino observations with IceCube are the large background generated by cosmic-ray interactions in the atmosphere and reconstruction of neutrino direction. The better the direction of the neutrino is known, the more background from directions not compatible with the GRB can be reduced. At high energies (≳100 GeV), enough light is deposited, in particular by energetic muons that are individually reconstructable in both direction and energy, to provide half-degree typical angular resolution. At lower energies (∼10–100 GeV), only enough light is collected to reconstruct directions within uncertainties of several to tens of degrees, forcing higher reliance on time correlations with the GRB to reject background. At the lowest energies (≲10 GeV), individual neutrino interactions are not reconstructable, forcing a complete reliance on timing information.

The analyses described in the following section use different strategies to search for neutrino emission from astrophysical transients of a wide variety of energy scales and timescales. Each analysis is optimized for the detection of neutrino emission in a different energy range, where it surpasses the sensitivity of other analyses. In combination, this allows us to probe neutrino emission of GRB 221009A over nine orders of magnitude in neutrino energy, from MeV to PeV.

In the following, we summarize the methodology and results of four complementary IceCube analyses of neutrino emission of GRB 221009A. Three of these are based on neutrino event samples that have been developed for previous IceCube analyses: the Gamma-ray Follow-Up (GFU) sample has been optimized for IceCube's real-time program (Aartsen et al. 2016b), the GeV Reconstructed Events with Containment for Oscillations (GRECO) sample was originally developed for oscillation studies with atmospheric neutrinos (Aartsen et al. 2019) and has since been used for low-energy neutrino transients (GRECO Astronomy; Abbasi et al. 2022b), and the Extremely Low-Energy (ELOWEN) event sample was developed for the search of low-energy neutrino emission during solar flares (Abbasi et al. 2021a). In addition, IceCube is able to observe a significant burst of MeV neutrinos through an increased level of single-photon hit rates in DOMs across the entire detector. This search is based on a real-time data stream called the Supernova Data Acquisition (SNDAQ) that was developed to detect MeV neutrinos from supernovae (Abbasi et al. 2011).

Table 1 summarizes the test hypotheses for the time-integrated neutrino flux F(E) ≡ d² N/dE/dA (neutrinos per area A and energy E) and corresponding neutrino energy scales accessible by these different data sets. The energy-dependent effective areas from the direction of GRB 221009A can be found in the Appendix. For the GFU and GRECO data sets the energy interval $[{E}_{\min },{E}_{\max }]$ corresponds to the central 90% energy range of predicted events for a power-law emission with spectral index γ at the location of GRB 221009A; harder (softer) indices will shift this energy range to higher (lower) values. Table 1 shows the wide MeV–PeV bandwidth covered in the following analyses.

Table 1. Models for the Time-integrated Neutrino Flux F(E) and Energy Ranges Probed by Different Data Sets

Data Set	Time-integrated Neutrino Flux Model
	Power Law: F(E) ∝ E−γ for ${E}_{\min }\leqslant E\leqslant {E}_{\max }$
	γ	${E}_{\min }$	${E}_{\max }$
GFU a	1.5	6.8 TeV	9.9 PeV
	2.0	0.83 TeV	0.96 PeV
	2.5	0.23 TeV	0.086 PeV
	3.0	0.13 TeV	0.013 PeV
GRECO a	1.5	40 GeV	1.5 TeV
	2.0	26 GeV	1.2 TeV
	2.5	15 GeV	0.70 TeV
	3.0	11 GeV	0.35 TeV
ELOWEN	2.0–3.0	0.5 GeV	5.0 GeV
	Quasi-thermal : ${F}_{{\bar{\nu }}_{e}}(E)\propto {E}^{2}\exp (-3E/\langle E\rangle )$
SNDAQ	E ≃ 〈E〉 ≃ (10–20) MeV

Notes. While the analyses based on GFU, GRECO, and ELOWEN test broad power-law fluxes, the analysis based on SNDAQ probes peaked quasi-thermal fluxes motivated by core-collapse supernovae.

^a Central 90% energy range depends on spectral index γ.

Download table as:  ASCIITypeset image

4.1. Fast-response Analysis: >100 GeV

The FRA (Abbasi et al. 2021b, 2022c) is a framework established to rapidly follow up interesting transients or multimessenger events in real time with the IceCube Neutrino Observatory. The FRA requires a real-time data stream, with low-latency access to data from the South Pole, in order to follow up interesting transients in real time. This GFU sample uses events triggered by causally connected single-photon hits on eight or more neighboring IceCube DOMs and employs a faster event reconstruction at the expense of some angular resolution available for offline reconstructions (Aartsen et al. 2016b). At final filter level, the stream has an all-sky rate that varies between approximately 6 and 7 mHz owing to seasonal variations (Abbasi et al. 2021b). Figure A1 in the Appendix shows the effective area of the GFU sample for the direction of GRB 221009A.

The FRA uses an unbinned maximum likelihood method to identify significant neutrino emission following an E⁻² spectrum on top of background events. The method has been used in other searches for short-timescale neutrino emission from potential sources (Aartsen et al. 2020; Abbasi et al. 2022a, 2023). We initiated an FRA of GRB 221009A in response to the observations reported by Swift and Fermi-GBM using the time windows (a) [T0 − 1 hr, T0 + 2 hr] (3 hr in total) and (b) T0 ± 1 day (2 days in total) to cover extended emission periods before and after the Fermi-GBM trigger, as summarized in Thwaites et al. (2022).

In addition to this initial real-time FRA, we also searched for neutrino emission using the same data sample for an extended set of time windows following the method of a previous archival search for neutrinos coincident with GRBs (Abbasi et al. 2022a). This analysis used three time windows in order to search for neutrinos (c) during the T90 phase as given by Fermi-GBM, (d) during the period [T0 − 200 s, T0 + 2000 s] to include times corresponding to the high-energy photons observed by LHAASO (Huang et al. 2022) as well as potential precursor emission, and (e) during the extended period [T0 − 1 day, T0 + 14 days] covering precursor-to-afterglow emission. We scanned over a circle of radius of 1°, centered on the location of GRB 221009A in order to find the location with highest significance, which is consistent with the previous method. Both these analyses assumed an E⁻² spectrum as in the initial FRA.

None of these analyses—the initial FRA and its application to an extended set of time windows—found indications for neutrino emission above background expectations, with p-values being close to 1. These results allow us to set upper limits on the energy-scaled time-integrated neutrino flux, E² F(E), which are summarized in Table 2 at a reference energy E₀ = 100 TeV. In the case of the standard FRA time windows of 3 hr and 2 days, we provide upper limits for four different power-law indices γ based on Monte Carlo injections following F(E) ∝ E^−γ .

Table 2. Upper Limits (UL) on the Time-integrated Neutrino Flux of GRB 221009A for the Different Time Windows

Data Set	Time Window and Index a		90% C.L. Upper Limits (ULs) on the Time-integrated Neutrino Flux F(E)
			Power Law F(E) ∝ E−γ : Per-flavor ULs Show E2 F(E) [GeV cm−2] at E0
			E0	γ = 1.5	γ = 2.0	γ = 2.5	γ = 3.0
GFU	[T0 − 1 hr, T0 + 2 hr]	(a)	100 TeV	0.0359	0.0393 b	0.0143	0.00240
	T0 ± 1 day	(b)		0.0370	0.0410 b	0.0176	0.00345
	T90 phase	(c)		⋯	0.0364	⋯	⋯
	[T0 − 200 s, T0 + 2000 s]	(d)		⋯	0.0369	⋯	⋯
	[T0 − 1 day, T0 + 14 days]	(e)		⋯	0.0471	⋯	⋯
GRECO	T90 phase	(c)	1 TeV	1.052	1.015	0.561	0.174
	[T0 − 200 s, T0 + 2000 s]	(d)		1.387	1.338	0.740	0.229
ELOWEN	T0 ± 500 s	(f)	1 GeV	⋯	5.3 × 103	8.7 × 103	1.4 × 104
	[T0 − 200 s, T0 + 2000 s]	(d)		⋯	7.9 × 103	1.3 × 104	2.0 × 104
			Quasi-thermal ${F}_{{\bar{\nu }}_{e}}(E)\propto {E}^{2}\exp (-3E/\langle E\rangle )$ : ${\bar{\nu }}_{e}$ UL on total and peak flux
			〈E〉	Total ${\bar{\nu }}_{e}$ Flux [cm−2]		${E}^{2}{F}_{{\bar{\nu }}_{e}}(E)$ [GeV cm−2] at 〈E〉
SNDAQ	[T0 − 100 s, T0]	(g)	15 MeV	7.98 × 108		8.05 × 106
	[T0 − 1 s, T0]	(h)		1.81 × 109		1.82 × 107
	[T0, T0 + 17 s]	(i)		8.00 × 108		8.07 × 106
	[T0 + 18 s, T0 + 174 s]	(j)		3.08 × 108		3.11 × 106
	[T0 + 174 s, T0 + 175 s]	(k)		1.35 × 109		1.36 × 107
	[T0 + 175 s, T0 + 547 s]	(l)		4.00 × 108		4.03 × 106

Notes. The three analyses based on GFU, GRECO, and ELOWEN assume a per-flavor time-integrated power-law flux F(E) ∝ E^−γ with variable spectral index γ. The results are given for the energy-scaled time-integrated per-flavor neutrino flux, E² F(E), at a reference energy E₀ depending on the data set. The analysis based on the SNDAQ assumes an electron antineutrino spectrum following a quasi-thermal spectrum peaked at 〈E〉 = 15 MeV. Results are shown for the total and peak time-integrated ${\bar{\nu }}_{e}$ flux.

^a The different time windows are discussed in Section 4 and are referenced by their inline text indices. ^b Values corresponding to the real-time FRA results published in Thwaites et al. (2022).

Download table as:  ASCIITypeset image

Figure 1 shows the upper limits on the time-integrated neutrino flux for the T90 phase and the time window [T0 − 1 hr, T0 + 2 hr]. These upper limits can be translated into limits on the differential isotropic equivalent energy $d{{ \mathcal E }}_{\mathrm{iso}}/{dE}$ of GRB 221009A at redshift z and luminosity distance D_L via the relation ${E}^{2}F(E)={Ed}{{ \mathcal E }}_{\mathrm{iso}}/{dE}\times (1+z)/(4\pi {D}_{L}^{2})$. This quantity is indicated by the right axis in Figure 1.

Zoom In Zoom Out Reset image size

4.2. GRECO Astronomy Analysis: 10–1000 GeV

This analysis focuses on the low-energy GRECO Astronomy data set, which is optimized for neutrinos between 10 GeV and 1 TeV (Abbasi et al. 2022d). This sample is based on events with at least three coincident hits in DeepCore DOMs and has an average event rate of about 4 mHz at final filter level (Aartsen et al. 2019). The effective area for GRB 221009A is shown in Figure A1. In order to test for neutrino emission from GRB 221009A, we use an extended unbinned maximum likelihood outlined in Abbasi et al. (2022a) combined with a spatial prior to account for the source localization uncertainty. The signal hypothesis assumes an E^−2.5 neutrino spectrum with equal contributions from all flavors.

Zoom In Zoom Out Reset image size

To account for the fact that angular uncertainties for the low-energy events used in the GRECO Astronomy data set are relatively large, a Kent distribution (Kent 1982) is assumed for the spatial likelihood. The localization prior is a 1° radius top-hat distribution centered at the localization provided by Swift. This is done for computational reasons, and the size of the prior does not affect the result given that angular uncertainties used in GRECO Astronomy are relatively large.

The analysis of GRB 221009A focused on two time windows, matching those of other analyses at different energy ranges. We searched for neutrino emission (c) coincident with the central 90% emission from the GRB as reported as the T90 phase by Fermi-GBM and (d) the 2200 s window [T0 − 200 s, T0 + 2000 s] including potential precursor emission and coinciding with high-energy photons observed by LHAASO. Though even longer time windows are desirable too, this is prevented by computational constraints.

We did not find significant deviations from a background distribution in either of the two time windows, with p-values being close to 1 in this analysis. The corresponding upper limits on the energy-scaled time-integrated neutrino flux, E² F(E), at a reference energy of E₀ = 1 TeV for each time window and four different spectral indices can be found in Table 2. Similar to the previous analysis, we provide upper limits for different power-law indices γ derived from the corresponding Monte Carlo injections. Figure 1 shows the upper limits on the time-integrated neutrino flux for the T90 phase.

4.3. ELOWEN Analysis: 0.5–5 GeV

4.4. MeV Neutrino Burst Analysis: <1 GeV

Our searches did not find significant neutrino emission in an energy range from MeV to PeV, before, during, or after the electromagnetic emission, resulting in upper limits on the time-integrated neutrino flux of GRB 221009A. These upper limits allow constraints on neutrino emission models from this source, as discussed by Ai & Gao (2023), Murase et al. (2022), Liu et al. (2023), and Rudolph et al. (2023). Previous IceCube searches for the joint neutrino emission of GRBs showed that high-energy neutrino emission during the prompt phase (or during 10⁴ s) of GRBs is limited to ≤1% (or ≤24%) of the high-energy diffuse neutrino flux observed by IceCube (Abbasi et al. 2022a).

Future upgrades and proposed extensions of IceCube could provide a crucial step toward neutrino detection from GRBs. The IceCube-Upgrade (Ishihara 2021) will enhance the detection of 1–10 GeV neutrino events with respect to the GRECO Astronomy and ELOWEN samples and will improve the angular reconstruction above 10 GeV. The proposed multicomponent extension IceCube-Gen2 (Aartsen et al. 2021) would significantly increase the sensitivity to high-energy neutrino transients, e.g., increasing the observable volume within our universe by one order of magnitude for TeV–PeV neutrinos compared to IceCube.

The IceCube Collaboration acknowledges Eric Burns for valuable discussions of public Fermi-GBM results. The authors gratefully acknowledge the support from the following agencies and institutions: USA—U.S. National Science Foundation-Office of Polar Programs, U.S. National Science Foundation-Physics Division, U.S. National Science Foundation-EPSCoR, Wisconsin Alumni Research Foundation, Center for High Throughput Computing (CHTC) at the University of Wisconsin–Madison, Open Science Grid (OSG), Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS), Frontera computing project at the Texas Advanced Computing Center, U.S. Department of Energy-National Energy Research Scientific Computing Center, Particle astrophysics research computing center at the University of Maryland, Institute for Cyber-Enabled Research at Michigan State University, and Astroparticle physics computational facility at Marquette University; Belgium—Funds for Scientific Research (FRS-FNRS and FWO), FWO Odysseus and Big Science programs, and Belgian Federal Science Policy Office (Belspo); Germany—Bundesministerium für Bildung und Forschung (BMBF), Deutsche Forschungsgemeinschaft (DFG), Helmholtz Alliance for Astroparticle Physics (HAP), Initiative and Networking Fund of the Helmholtz Association, Deutsches Elektronen Synchrotron (DESY), and High Performance Computing cluster of the RWTH Aachen; Sweden—Swedish Research Council, Swedish Polar Research Secretariat, Swedish National Infrastructure for Computing (SNIC), and Knut and Alice Wallenberg Foundation; European Union—EGI Advanced Computing for research; Australia—Australian Research Council; Canada—Natural Sciences and Engineering Research Council of Canada, Calcul Québec, Compute Ontario, Canada Foundation for Innovation, WestGrid, and Compute Canada; Denmark—Villum Fonden, Carlsberg Foundation, and European Commission; New Zealand—Marsden Fund; Japan—Japan Society for Promotion of Science (JSPS) and Institute for Global Prominent Research (IGPR) of Chiba University; Korea—National Research Foundation of Korea (NRF); Switzerland—Swiss National Science Foundation (SNSF); United Kingdom—Department of Physics, University of Oxford.

The complementary energy coverage of the different data sets used in this Letter can be appreciated by a comparison of the respective neutrino effective areas. The decl.- and energy-dependent all-flavor effective area A_eff(δ, E_ν ) is defined via the relation

$$\begin{eqnarray}&&{N}_{\mathrm{events}}=\int {{dE}}_{\nu }{A}_{\mathrm{eff}}(\delta ,{E}_{\nu })F({E}_{\nu }),\end{eqnarray} \tag{ A1 }$$

where N_events is the analysis-specific number of signal events and F(E_ν ) is the per-flavor time-integrated neutrino flux (assuming identical flux for each flavor) for a source at decl. δ. Figure A1 shows these effective areas for the ELOWEN, GRECO, and GFU samples for the decl. δ ≃ + 198 of GRB 221009A. Starting at low energy, the ELOWEN sample contributes the most up to neutrino energies of 5 GeV, the GRECO selection performs best in the 10–100 GeV range, and the GFU sample dominates beyond a few hundred GeV. The GFU data sample used for our analysis is described in full by Aartsen et al. (2016b). Note that the effective area of this GFU selection is substantially larger, especially at low energies, than the small subsample of GFU events used to trigger alerts (Blaufuss et al. 2020) and is similar to those from other neutrino source searches at high energy. The SNDAQ effective area is not shown, as it is does not work with individual neutrino events and so its effective area is not comparable to the other analyses and is not a good guide to sensitivity. The SNDAQ analysis instead searches for a sudden increase of the overall background rate in the detector induced by a burst of sub-GeV neutrinos. The figure of merit is not the total number of signal events, as in Equation (A1), but rather the rate of correlated single-photon hits across the detector, which is a function of both flux and energy spectrum. More details on SNDAQ sensitivity can be found in Abbasi et al. (2011).