Main

On 15 April 2020, the extremely bright, short γ-ray burst GRB 200415A occurred at 08:48:06 utc at Earth, and was detected by five space-based missions of the Interplanetary Network of γ-ray detectors (IPN, Methods). Here we report the final localization of the burst by the IPN to a roughly 20-arcmin2 region (Methods) that overlaps with the central part of the nearby galaxy NGC 253, at a distance of DNGC253 ≈ 3.5 Mpc (ref. 14) (Fig. 1). The chance occurrence for GRB 200415A to be spatially consistent with a nearby galaxy likely to produce detectable giant flares is approximately 1 in 200,000 (ref. 15).

Fig. 1: The final IPN localization of GRB 200415A.
figure 1

The localization is superimposed on an image of the NGC 253 galaxy from the GALEX survey (1,750–2,800 Å; Methods). It is defined by the 4.73-arcmin-wide Wind (Konus)–Odyssey (HEND) and 3.58-arcmin-wide Wind (Konus)–Fermi (GBM) annuli. The IPN error box (grey parallelogram) is shown along with the 20 arcmin2 3σ error ellipse for the position. The coordinates are J2000.

Full size image

GRB 200415A triggered Konus–Wind at T0 = 08:48:01.403 utc. As observed by Konus–Wind, the lightcurve of the burst starts with the fast (around 2 ms) rise of a narrow (around 4 ms) initial spike, which is followed by an exponentially decaying phase with a count-rate e-folding time of τcr ≈ 50 ms (Fig. 2a). The total duration of the burst is 0.138 s, and T90 (the duration of the time interval that contains the central 90% of the total count fluence of the burst) is 0.100 ± 0.014 s (hereafter, all the quoted uncertainties are at the 68% confidence level).

Fig. 2: Time histories of GRB 051103 (open symbols) and GRB 200415A (filled symbols) and evolution of their spectral parameters.
figure 2

All times are given relative to the Konus–Wind trigger time T0. a, Burst time histories as recorded by Konus–Wind. Both events start with a sharp rise of an exceptionally bright, narrow (4 ms) initial spike (grey shaded area), followed by an exponential decay with τcr ≈ 50 ms (dotted lines). b, d, The temporal evolution of the emission spectra is illustrated by the behaviour of the best-fitting parameters of the cutoff power-law model: the peak energy Ep (b) and the photon power-law index α (d). Both bursts are characterized by Ep ≈ 1.2 MeV in the initial pulse, which is the hardest part of GRB 200415A, whereas the hardest emission in GRB 051103 (with Ep ≈ 3 MeV) was detected during the subsequent approximately 30 ms. A non-thermal cutoff power-law model adequately describes burst spectra up to about T0 + 100 ms; afterwards, the hard power-law photon index α becomes poorly constrained and, simultaneously, the emission spectrum can be described by a blackbody function with kT ≈ 70–100 keV. c, Evolution of the 20 keV–10 MeV energy flux, which, in both cases, peaks in the initial spike (grey shaded area) and, starting from about T0 + 50 ms, decays with τflux ≈ 30 ms (dotted lines). Vertical error bars indicate 68% confidence intervals; horizontal error bars indicate the duration of the interval.

Source data

Full size image

The hardness of the burst (the ratio between the 390–1,600-keV and 90–390-keV count rates) increases rapidly during the initial spike, peaks during the following approximately 8 ms and then decays gradually with the count rate of the burst. Our spectral analysis (Methods) shows that, starting from the rise of the spike and up to about T0 + 100 ms, the energy spectrum of the burst is well described by a cutoff power-law function (proportional to Eαexp[−E(α + 2)/Ep]). The temporal evolution of the spectrum is illustrated in Fig. 2, which shows the behaviour of the cutoff power-law model parameters: the peak energy Ep, which corresponds to the maximum of the spectral energy distribution νFν, where Fν is the energy flux per unit frequency interval at frequency ν (Fig. 2b) and the photon power-law index α (Fig. 2d). The initial spike is characterized by Ep ≈ 1.2 MeV, with α ≈ −0.6. This Ep was the highest reached in the entire event. A non-thermal cutoff power-law model, with Ep decaying nearly exponentially, adequately describes burst spectra up to about T0 + 100 ms. Afterwards, the very hard photon index α becomes poorly constrained; and, simultaneously, the emission spectrum can be described by a blackbody function (with a temperature kT ≈ 70 keV), which is excluded by our analysis at the initial stage of the burst.

In Fig. 2c we show the temporal evolution of the 20 keV–10 MeV energy flux. It peaks in the initial spike with a 4-ms peak flux of \({0.96}_{-0.16}^{+0.32}\times {10}^{-3}\,{\rm{erg}}\,{{\rm{cm}}}^{-2}\,{{\rm{s}}}^{-1}\) and, starting from around T0 + 50 ms, decays with τflux ≈ 30 ms. The time-integrated spectrum, measured from T0 to T0 + 0.192 s, is best described by a sum of non-thermal (cutoff power-law) and thermal (blackbody) components (see Extended Data Table 1 for the model parameters). The total burst fluence is \({8.5}_{-1.0}^{+1.2}\times {10}^{-6}\,{\rm{erg}}\,{{\rm{cm}}}^{-2}\) in the 20 keV–10 MeV band.

A preliminary analysis of the Konus–Wind detection16 of GRB 200415A revealed its remarkable similarity to GRB 051103, historically the first extragalactic giant-flare candidate outside the Local Group, associated with the M81/M82 group of galaxies8,9,10 at a distance of DM81/M82 ≈ 3.6 Mpc (ref. 17). To explore this similarity further, we performed a detailed comparison between the temporal and spectral properties of GRB 200415A and GRB 051103. The bursts have similar lightcurve and spectral evolution patterns (Fig. 2; Extended Data Tables 1, 5, 6). Although the peak count rates, reached in the first 2 ms of the initial spikes, are very similar (about (1.5–1.7) × 105 s−1), the photon flux over the entire extent of the decaying phase is about twice as high in GRB 051103 as in GRB 200415A. The initial pulses of both bursts are best described by the cutoff power-law model, with Ep ≈ 1.2 MeV, but GRB 051103 has a much harder α ≈ −0.1. In contrast to GRB 200415A, the hardest emission in GRB 051103 (Ep ≈ 3 MeV, α ≈ 0.2) was observed during the roughly 30 ms immediately after the initial spike. In accordance with the similarities of the bursts in peak count rate and Ep, measured in the initial spikes, their 4-ms peak flux estimates also agree within uncertainties. The blackbody components in the time-integrated spectra of GRB 200415A and GRB 051103 have similar temperatures (kT ≈ 100 keV), with blackbody contributions to the total fluence of about 14% and 9%, respectively. The contribution of the initial short spike to the total fluence is about 45% for GRB 200415A and 13% for GRB 051103.

Thus, the extremely bright, short GRB 200415A, which strong evidence suggests is associated with the NGC 253 galaxy, is remarkably similar to GRB 051103, which presumably originated from the M81/M82 group of galaxies at nearly the same distance, in terms of light curve morphology, spectral behaviour and observed peak energy flux. A lightcurve with a bright, millisecond-scale initial pulse followed by an exponentially decaying emission is unusual for short cosmological γ-ray bursts (GRBs); none of more than about 500 short bursts detected by Konus–Wind in more than 25 years of observations displays such a shape15,18. On the other hand, this pattern was observed in two Galactic giant flares, from SGR 1900+1419,20 and SGR 1806−2020,21. Furthermore, higher-time-resolution lightcurves of GRB 200415A from Swift–Burst Alert Telescope (BAT) and Fermi–Gamma-ray Burst Monitor (GBM)22 have an initial short (less than 1 ms) subpeak, followed by a sharp decrease for approximately 1 ms, before the main part of the peak. This pattern is also seen in SGR 1806−2012 and may be a general property of giant flares that can be used to identify them within the short-GRB sample. Thus, the interpretation of GRB 200415A and GRB 051103 as magnetar giant flares is strongly suggested, with additional support provided by the non-detection of an accompanying gravitational-wave signal for GRB 05110323 (there is no sensitive coverage by a gravitational-wave detector for GRB 200415A).

At source distances of DNGC253 = 3.5 Mpc and DM81 = 3.6 Mpc, the characteristic radius of the emission region, estimated from the blackbody spectral fits, is R ≈ 20−40 km, the same order of magnitude as the radius of a neutron star or its magnetosphere. The implied isotropic-equivalent energy release in γ-rays for GRB 200415A (GRB 051103) is Eiso ≈ 1.3 × 1046 erg (Eiso ≈ 5.3 × 1046 erg) and the isotropic-equivalent peak luminosity is Liso ≈ 1.4 × 1048 erg s−1 (Liso ≈ 1.8 × 1048 erg s−1). Therefore, the total energies released in both flares are comparable with that estimated for the most energetic flare from a Galactic magnetar5,12,13. Taken together, these results make GRB 200415A and GRB 051103 the most substantial candidates for extragalactic magnetar giant flares: both are at least about five times more luminous than any Galactic magnetar flare observed previously2. Such high luminosities may indicate that the magnetar sources of GRB 200415A and GRB 051103 are younger than a few hundred years, so their magnetic fields are strong enough to power such flares5,24. Assuming the same spectra and energetics, similar events could be detected with Konus–Wind from distances up to around 16 Mpc.

Despite the strong evidence in favour of the giant-flare nature of GRB 200415A and GRB 051103, it cannot completely be ruled out that they might belong to an as-yet-undiscovered branch of the cosmological short-GRB population. For the observed energy fluence, and assuming a cosmological redshift of z = 0.05–1, GRB 200415A is consistent with the Konus–Wind sample of short GRBs with known redshifts25,26 in terms of a hardness–intensity relation in the cosmological rest frame (Extended Data Fig. 1). In the case of GRB 051103, the implied short-GRB redshift is z≈ 1, with intrinsic Ep ≈ 5 MeV.

The detection of extragalactic giant flares facilitates the study of emission processes on millisecond and sub-millisecond timescales22; such study is not possible for galactic events because they saturate almost all γ-ray detectors. On the timescale relevant to Konus–Wind (more than about 2 ms), the single-peaked GRB 200425A and GRB 051103 are clearly different from the third known extragalactic giant-flare candidate, GRB 0702012, which has highly variable emission during the first roughly 50 ms. This suggests that the physical processes behind the emission in the initial pulses of giant flares may develop on timescales that span more than an order of magnitude.

The IPN box of GRB 200415A is projected partially into the nuclear region, the bar, a ring-like structure enclosing the bar and a spiral arm of NGC 253, which contain many young star groups27. This is consistent with the Galactic magnetars, which are associated with the young stellar population24. The current sample of giant-flare host galaxies includes five massive galaxies (the Milky Way, M31, M81/M82 group and NGC 253) and the Large Magellanic Cloud, which resemble the host galaxies of non-repeating fast radio bursts28. This provides evidence for a connection between soft γ-ray repeaters and fast radio bursts.

Methods

Observations

GRB 200415A occurred at 08:48:06 utc at Earth, and was detected by the GBM onboard the Fermi Gamma-ray Space Telescope Mission29 and the BAT (outside the coded field of view) onboard the Neil Gehrels Swift Observatory30 in low-Earth orbits; the SPI telescope anticoincidence system31 (SPI-ACS) and Pixellated Imaging Caesium Iodide Telescope32,33 (IBIS-PICsIT) instruments onboard the International Gamma-Ray Astrophysics Laboratory (INTEGRAL) in an eccentric Earth orbit 0.44 light seconds from Earth; the Mars–Odyssey High Energy Neutron Detector34,35 (HEND) in orbit around Mars 672 light seconds from Earth; the Konus–Wind36 γ-ray spectrometer on the Wind mission37 in a Lissajous orbit at the L1 libration point of the Sun–Earth system at a distance of 4.8 light seconds; and the Modular X- and Gamma-Ray Sensor (MXGS) of The Atmosphere-Space Interactions Monitor ASIM38 onboard the International Space Station (not a part of the IPN).

Two independent and consistent Fermi–GBM localizations39,40, RoboBA41 and BALROG42,43, were announced in GRB Coordinates Network circulars at 08:58 utc and 09:11 utc, respectively, each with a 3σ probability region covering more than 1,000 deg2. By exploiting the difference in the arrival time of the γ-ray signals at the four IPN (see next section) instruments (GBM, BAT, HEND and SPI-ACS), a preliminary 1.5-deg2 IPN error box was announced at 16:51 utc44, and it was pointed out that this event might be a magnetar giant flare in NGC 253 (the Sculptor galaxy) located at DNGC253 ≈ 3.5 Mpc. The Fermi–LAT localization, announced at 19:30 utc45, was consistent with the box.

As soon as Konus–Wind data arrived, the improved 274-arcmin2 box (about 20 times smaller than the initial box) was published, on 16 April 16:16 utc46, which strengthened the association of the burst with the galaxy. The box was close to the 68% confidence contours of the updated Fermi–LAT localization47, published on 16 April 20:48 utc, and inside the 2σ and 1σ contours of the GBM RoboBA and BALROG localizations, respectively. The IPN localization was within about 37° of the Sun, making X-ray and optical follow-up observations challenging. The only optical observation of NGC 253, which resulted in an upper limit, was reported by the MASTER telescope network on 17 April48.

IPN

The IPN (http://www.ssl.berkeley.edu/ipn3/) is a group of spacecraft orbiting Earth and Mars equipped with GRB detectors used to localize GRBs49. When a GRB arrives at two spacecraft, it may be localized (triangulated) to an annulus on the sky, determined by the measured propagation time delay and spacecraft positions. Three spacecraft produce two possible locations (IPN error boxes). The ambiguity can be eliminated by the addition of a fourth, non-coplanar spacecraft, by the anisotropic response of Konus–Wind or by the GBM localization50.

The propagation time delay and its uncertainty are calculated by cross-correlation51. The systematic uncertainties are estimated using the comparison of IPN triangulations with precise GRB positions50.

The declared onboard clock accuracy of the spacecraft are: down to 1 μs for Fermi; about 200 μs for Swift; less than about 1 ms for Wind; and about 100 μs for INTEGRAL. For Mars–Odyssey, an overall 3σ systematic uncertainty, which includes timing and other effects derived from IPN observations of precisely localized GRBs, is better than 360 ms. The Wind clock drift information is provided at ftps://pwgdata.sci.gsfc.nasa.gov/pub/wind_clock/.

Near-Earth spacecraft ephemerides are derived from two-line elements available at https://www.space-track.org using the SGP8 model. The Wind predicted ephemerides and their description are available at https://spdf.gsfc.nasa.gov/pub/data/wind/orbit/pre_or and https://cdaweb.gsfc.nasa.gov/misc/NotesW.html#WI_OR_PRE, respectively. Mars–Odyssey ephemerides were taken from the Jet Propulsion Laboratory’s HORIZONS system (https://ssd.jpl.nasa.gov/horizons.cgi).

For the near-Earth spacecraft and Wind, ephemeris uncertainties contribute less than 1 ms to the propagation time delay, so we conservatively assume a systematic error in Konus–GBM and Konus–BAT triangulations to be 1 ms. For the Konus–HEND and GBM–HEND triangulations, we take 360 ms as the 3σ systematic uncertainty.

For the GRB 200415A triangulation, we used the following lightcurves: 2-ms Konus (see Methods section ‘Konus–Wind’), 390–1,600 keV; 0.1-ms GBM, 360–1,000 keV, constructed from the time-tagged event data of triggered detectors (0, 1, 2, 3, 4, 5, 9 and a; only NaI data were used); 0.1-ms BAT, 25–350 keV, constructed from the time-tagged event data from the GUANO system52; 250-ms HEND, 50–3,000 keV; and 7.8-ms INTEGRAL–PICsIT, 250–2,000 keV.

Using these data, we derived five annuli (Extended Data Table 2). The final IPN 3σ box was constructed from Konus–GBM and Konus–HEND annuli (Extended Data Table 3). We used the Konus–GBM annulus instead of the narrower Konus–BAT one owing to the similarity between the energy bands of the two instruments. GBM saturation occurred near the burst peak, but does not significantly affect the cross-correlation with the Konus 2-ms-resolution lightcurve.

The annuli were combined to yield an error ellipse53 with a major axis corresponding to the Konus–GBM annulus and a minor axis corresponding to the Konus–HEND annulus. We obtain a 3σ error ellipse centred at a right ascension (J2000) of 11.885° and declination (J2000) of −25.263°, with major and minor axes of 6.25 arcmin and 4.07 arcmin, respectively, and a position angle of 61.135°. The area of the ellipse is 20 arcmin2.

The ellipse contains the central part of NGC 253 (Fig. 1). The image was obtained with the Galaxy Evolution Explorer (GALEX)54 during the GALEX Nearby Galaxies Survey55 (observation ID, 2482083865531777024). The NGC 253 image was obtained via the MAST portal (https://mast.stsci.edu).

Konus–Wind

Konus–Wind36 consists of two identical NaI(Tl) scintillation detectors, each with a 2π-sr field of view, mounted on opposite faces of the rotationally stabilized Wind spacecraft37, such that one detector (S1) points towards the south ecliptic pole, thereby observing the south ecliptic hemisphere, while the other (S2) observes the north ecliptic hemisphere.

Each Konus–Wind detector is a cylinder 5 inches in diameter and 3 inches in height, placed into an aluminium container with a beryllium entrance window. The crystal scintillator is viewed by a photomultiplier tube through a 20-mm-thick lead glass, which provides effective detector shielding from the spacecraft’s background in the soft spectral range. The detector effective area is about 80–160 cm2, depending on the photon energy and incident angle. The energy range of γ-ray measurements covers the incident photon energy interval from 20 keV to 20 MeV.

The instrument has two operational modes: waiting and triggered. While in the waiting mode, the count rates (lightcurve) are recorded in three energy bands (G1, G2 and G3) covering roughly 20–1,500 keV (Extended Data Table 3), with 2.944-s time resolution. When the count rate in the roughly 80–350-keV band exceeds an approximately 9σ threshold above the background on one of two fixed timescales (1 s or 140 ms), the instrument switches into the triggered mode.

In the triggered mode, lightcurves are recorded in the same bands, starting from 0.512 s before the trigger time T0, with time resolution varying from 2 ms to 256 ms. For the bursts of interest here, the whole time history is available with 2-ms resolution.

Multichannel spectral measurements are carried out, starting from T0 (no multichannel spectra are available before T0) in two overlapping energy intervals, PHA1 and PHA2 (Extended Data Table 4), with 64 spectra being recorded for each interval over a 63-channel, pseudo-logarithmic energy scale. The first four spectra are measured with a fixed accumulation time of 64 ms to study short bursts.

For this analysis, we use a standard Konus–Wind dead-time correction procedure for lightcurves (with a dead time of a few microseconds) and multichannel spectra (with a dead time of about 42 μs).

Temporal analysis

For the temporal analysis, we used time histories from T0 − 0.512 s to T0 + 0.512 s in three energy bands (G1, G2 and G3), with a time resolution of 2 ms. The total burst duration T100, and the T90 and T50 durations (the time intervals that contain 5%–95% and 25%–75% of the total burst count fluence, respectively56), were calculated using the lightcurve in the roughly 80–1,500-keV energy band (G2 + G3). Burst start and end times in each band were calculated at the 5σ level using a method similar to that developed for BATSE57. The background count rates, estimated using the data from about T0 − 2,500 s to about T0 − 150 s, are 958.7 s−1 (G1), 349.5 s−1 (G2) and 223.0 s−1 (G3) for GRB 200415A, and 1,080.3 s−1 (G1), 394.0 s−1 (G2) and 135.5 s−1 (G3) for GRB 051103.

Spectral analysis

For the bursts of interest, we analysed multichannel and three-channel Konus–Wind energy spectra. The multichannel spectra accumulation intervals are presented in Extended Data Tables 5 and 6. The background multichannel spectra were extracted in the intervals from T0 + 8.448 s to T0 + 491.776 s and from T0 + 98.560 s to T0 + 491.776 s for GRB 200415A and GRB 051103, respectively. The emission evolution at a finer timescale can be explored using three-channel spectra, constructed from the counts in the G1, G2 and G3 energy bands in the six intervals (Extended Data Tables 5, 6). Details on Konus–Wind three-channel spectral analysis can be found elsewhere18.

We performed the spectral analysis in XSPEC, version 12.10.158, using the following spectral models: a simple power law, a custom exponential cutoff power-law (CPL) parameterized by the peak of νFν spectrum and with the energy flux as the model normalization, the Band GRB function59, a single blackbody (BB) function with the normalization proportional to the surface area, and a sum of the CPL and BB functions (CPL + BB). The details of each model are as follows: the power law model is described by

$${f}_{{\rm{PL}}}=A{\left(\frac{E}{{E}_{n}}\right)}^{\alpha },$$

the custom exponential CPL model by

$$\begin{array}{c}n(E)={\left(\frac{E}{{E}_{n}}\right)}^{\alpha }\exp \left[-\frac{E(2+\alpha )}{{E}_{{\rm{p}}}}\right],\\ {f}_{{\rm{CPL}}}=\frac{Fn(E)}{{\int }_{{E}_{{\rm{\min }}}}^{{E}_{{\rm{\max }}}}n(E)E{\rm{d}}E},\end{array}$$

and the Band function by

$${f}_{{\rm{Band}}}=A\,\{\begin{array}{ll}{\left(\frac{E}{{E}_{n}}\right)}^{\alpha }\exp \left[-\frac{E(2+\alpha )}{{E}_{{\rm{p}}}}\right], & E < \frac{(\alpha -\beta ){E}_{{\rm{p}}}}{2+\alpha },\\ {\left(\frac{E}{{E}_{n}}\right)}^{\beta }\,{\left[\frac{{E}_{{\rm{p}}}(\alpha -\beta )}{{E}_{n}(2+\alpha )}\right]}^{(\alpha -\beta )}\exp (\beta -\alpha ), & E\ge \frac{(\alpha -\beta ){E}_{{\rm{p}}}}{2+\alpha },\end{array}$$

where fPL,CPL,Band is the relevant photon spectrum (measured in photons cm−2 s−1 keV−1), A is the model normalization, En = 100 keV is the pivot energy, Ep is the peak energy of the νFν spectrum and F is the model energy flux in the Emin–Emax energy band; α and β are the low-energy and high-energy photon indices, respectively, and n(E) is the unnormalized photon spectrum. The single BB function is the bbodyrad XSPEC model.

The Poisson data with Gaussian background statistic (PG-stat) was used in the model-fitting process as a figure of merit to be minimized. The spectral channels were grouped to have a minimum of one count per channel to ensure the validity of the fit statistic. Because the CPL fit to a three-channel spectrum has zero degrees of freedom (and, in the case of convergence, PG-stat = 0), we do not report the statistic for such fits. The 68% confidence intervals of the parameters were calculated using the command steppar in XSPEC.

A summary of constrained spectral fits with the CPL, BB and CPL + BB models is presented in Extended Data Tables 5 and 6. For GRB 200415A and GRB 051103, the power-law model failed to describe the spectra, with PG-stat/dof > 10 in all cases. Use of the Band GRB function does not constrain the high-energy photon index β for GRB 200415A spectra, and only marginally improves the CPL fit to the time-integrated spectrum of GRB 051103, with similar (within errors) Ep and α, and β ≈ −3.

Burst energetics

For both bursts, the total energy fluence S was derived using the 20 keV–10 MeV energy flux of the best-fitting (CPL + BB) spectral model. Because the time-integrated spectrum accumulation interval differs from the T100 interval, a correction that accounts for the emission outside the time-integrated spectrum was introduced when calculating S.

The peak flux Fpeak was calculated on the 4-ms scale using the energy flux of the best fit with the CPL model to the three-channel spectrum at the peak count rate interval (T0 − 0.002 s to T0 + 0.002 s). The peak flux of GRB 051103 estimated here is a factor of about 2.5 lower than that reported from previous analyses of Konus–Wind and RHESSI data9,10, which used wider spectral intervals and did not separate the relatively soft spectrum in the huge 4-ms spike (Ep ≈ 1.2 MeV) and the considerably harder emission observed immediately after its falling edge (Ep ≈ 3 MeV).