INTEGRAL IBIS catalog of magnetar bursts

One of the distinctive properties of magnetars, young neutron stars powered mainly by magnetic energy, is the emission of short ($\lesssim$1 s) bursts of hard X-rays. Such bursts have been observed in nearly all the known magnetars, although at different and time-variable rates of occurrence. In the last two decades, the INTEGRAL satellite has extensively covered with good imaging capabilities the Galactic plane, where most magnetars reside. We present the results of a comprehensive search for magnetar bursts in more than twenty years of archival data of the INTEGRAL IBIS instrument (15 keV - 1 MeV). This led to the detection of 1349 bursts with 30-150 keV fluence in the $\sim2\times10^{-9} - 3\times10^{-6}$ erg cm$^{-2}$ range from 21 of the 34 examined magnetars and candidate magnetars with well known positions. The durations of the bursts, in terms of $T_{90}$, follow a lognormal distribution centered at $\sim0.1$ s. Most of the detected bursts originated from three particularly active sources: 1E 1547-5408, SGR 1806-20, and SGR 1935+2154. The integral distributions of their burst fluences follow power laws with slopes $β$= 0.76$\pm$0.04, 0.95$\pm$0.06, and 0.92$\pm$0.10, respectively. The burst spectra are generally well fit with an exponentially cut-off power law with peak energy $E_{peak}$ in the range $\sim20-60$ keV for SGR 1806-20 and SGR 1935+2154, while the bursts of 1E 1547-5408 are slightly harder ($E_{peak}\sim35-100$ keV). A significant anti-correlation between $E_{peak}$ and fluence is found for SGR 1806-20, which provided the largest number of bursts among the sources of our sample.

💡 Research Summary

The paper presents a systematic, long‑term survey of magnetar short‑burst activity using the INTEGRAL satellite’s IBIS/ISGRI detector, covering more than twenty years of public data (2002–2024). The authors selected 34 objects—34 confirmed magnetars or magnetar candidates with well‑determined positions—including two rotation‑powered pulsars that have shown magnetar‑like bursts. For each source they extracted all Science Windows (ScWs) within 14.5° of the source, removed ScWs with variable background (χ²_r ≤ 1), and applied a pixel illumination factor (PIF) filter to keep only those detector pixels actually illuminated by the source.

Burst detection was performed on eight logarithmically spaced timescales ranging from 0.01 s to 1.28 s. For each timescale a significance threshold corresponding to a false‑positive probability of 10⁻³ per ScW was set. Candidate triggers were then verified by creating sky images for the corresponding intervals using the IBAS imaging pipeline; this step eliminated spurious events caused by background fluctuations, other sources in the field of view, or telemetry saturation. Out of roughly 75 000 initial triggers, 1 349 bursts were confirmed as originating from 21 of the 34 examined objects. The three most prolific sources—1E 1547‑5408, SGR 1806‑20, and SGR 1935+2154—account for 92 % of all detections (146, 934, and 232 bursts respectively).

Timing analysis employed the Bayesian Blocks algorithm to identify statistically significant changes in the unbinned event list. From the BB segmentation, T₉₀ durations (the interval containing 90 % of the burst fluence) were derived. The distribution of T₉₀ follows a log‑normal shape centred at ∼0.1 s with a standard deviation of ∼0.3 dex, consistent with previous studies of magnetar bursts.

Spectral analysis was carried out in the 20–150 keV band using the summed counts of each burst. An exponentially cut‑off power‑law model (cut‑off power law) provided satisfactory fits. The peak energy (E_peak) lies between 20–60 keV for SGR 1806‑20 and SGR 1935+2154, while bursts from 1E 1547‑5408 are slightly harder, with E_peak ≈ 35–100 keV. Cumulative fluence distributions for the three main sources are well described by power laws, with slopes β = 0.76 ± 0.04 (1E 1547‑5408), 0.95 ± 0.06 (SGR 1806‑20), and 0.92 ± 0.10 (SGR 1935+2154). Notably, SGR 1806‑20 exhibits a statistically significant anti‑correlation between fluence and E_peak (correlation coefficient ≈ ‑0.45, p < 10⁻⁴), indicating that brighter bursts tend to have softer spectra.

Instrumental systematics were carefully addressed. Telemetry saturation during very bright events creates short gaps (< 1 s) in the light curves; these were linearly interpolated and flagged as lower limits in fluence plots. The ISGRI detector suffered a gradual gain loss of ~2.6 % yr⁻¹ and occasional solar‑flare‑induced drops, leading to a rising low‑energy threshold (≈ 25 keV in 2014, ≈ 40 keV at mission end). Dead‑time corrections, including contributions from the VETO system and calibration source, were applied using housekeeping data. All reported count rates are corrected for dead time and for the coding fraction (COD) of partially coded fields of view.

The catalog represents the most extensive imaging‑based compilation of magnetar bursts to date, providing precise source identification, uniform timing and spectral measurements, and long‑term activity histories. Compared with earlier non‑imaging surveys, it enables source‑by‑source statistical studies and cross‑comparison with contemporaneous observations from NICER, HXMT, and Fermi/GBM. The authors discuss the implications for magnetar burst mechanisms, supporting models where crustal fractures trigger rapid magnetic reconnection, and they highlight the observed fluence–spectral hardness anti‑correlation as a potential diagnostic of energy release processes. Future work will focus on multi‑wavelength correlation studies, detailed modeling of burst spectra, and extending the analysis to the full INTEGRAL data set, including the still‑operational SPI instrument, to further elucidate the physics of magnetar outbursts.