Fermi GBM Burst Catalog Overview

• Fermi GBM Burst Catalog is a comprehensive database of gamma-ray burst events observed over more than ten years.
• It combines prompt emission measurements—durations, peak fluxes, fluences, and spectral parameters—with detailed localizations from multiple detectors.
• The catalog supports multimessenger astrophysics by providing analysis-ready data products and enabling sub-threshold event recovery.

Searching arXiv for the Fermi GBM burst-catalog papers and closely related supplements. The Fermi GBM Burst Catalog is the family of catalogs built from the Gamma-ray Burst Monitor on board the Fermi Gamma-ray Space Telescope, designed to provide the community with locations, durations, peak fluxes, fluences, spectral parameters, and classifications for gamma-ray bursts and other high-energy transients observed over the GBM energy range. Across its official gamma-ray burst catalogs, the program extends from the first two mission years to a decade of observations, while companion catalogs and supplements cover systematic spectroscopy, time-resolved spectroscopy, short gamma-ray bursts, X-ray bursts, Interplanetary Network localizations, and untriggered transient searches (Paciesas et al., 2012, Kienlin et al., 2020, Poolakkil et al., 2021).

1. Instrumental basis and catalog concept

Fermi/GBM comprises twelve NaI(Tl) detectors and two BGO scintillation detectors. The NaI detectors cover nominally 8 keV–1 MeV, or 8 keV–1 MeV in the catalog descriptions, while the two BGO detectors extend the energy range to approximately 0.2–40 MeV or 200 keV–40 MeV, depending on the formulation used in the relevant catalog paper (Kienlin et al., 2014, Kienlin et al., 2020). GBM continuously monitors detector count rates in several energy bands and on timescales from 16 ms to 8 s; when two or more detectors exceed preset significance thresholds, the flight software issues a trigger and produces burst data with sub-ms time tagging and full spectral resolution (Kienlin et al., 2020).

The catalog program is organized around a sequence of official GRB catalogs and associated higher-level products. The official GRB catalogs report the location and main characteristics of the prompt emission, the duration, peak flux, and fluence for each GRB, with peak fluxes and fluences given in both 50–300 keV and 10–1000 keV so that BATSE-comparable measurements and full-NaI-band measurements are both available (Kienlin et al., 2014, Bhat et al., 2016, Kienlin et al., 2020). Complementary products extend this baseline. The spectral catalogs provide two spectra per burst—“fluence” and “peak-flux”—fitted with four empirical models; the time-resolved catalog analyzes the brightest bursts with high temporal and spectral resolution; the short-GRB catalog re-selects bursts through a Bayesian-Block procedure; and the Interplanetary Network supplement combines arrival-time triangulation with GBM localizations to reduce sky areas (Goldstein et al., 2012, Yu et al., 2016, Lu et al., 2017, Hurley et al., 2013).

This catalog architecture reflects two distinct functions. One is uniform population-level description of prompt high-energy transients; the other is the production of analysis-ready data products—FITS tables, response matrices, light curves, and sky maps—distributed through HEASARC and related archive interfaces (Kienlin et al., 2014, Poolakkil et al., 2021).

2. Official GRB catalog sequence

The official GRB catalog sequence grows cumulatively in time span and sample size. The first two-year catalog reports 491 triggered GRBs out of 908 total GBM triggers between 12 July 2008 and 11 July 2010 (Paciesas et al., 2012). The second catalog extends coverage to 2008 July 12 through 2012 July 11 and reports 953 distinct GRBs, derived from 954 GRB-classified triggers out of 2,126 transient triggers (Kienlin et al., 2014). The third catalog reaches 2014 July 11 and lists 1,405 triggers identified as GRBs out of 3,350 GBM triggers (Bhat et al., 2016). The fourth catalog covers the decade from 2008 July 12 through 2018 July 11 and includes 2,356 events identified as cosmic GRBs out of 6,434 transient triggers (Kienlin et al., 2020).

Catalog	Coverage	Reported GRBs
First GRB catalog	2008-07-12 to 2010-07-11	491 triggered GRBs
Second GRB catalog	2008-07-12 to 2012-07-11	953 distinct GRBs
Third GRB catalog	2008-07-12 to 2014-07-11	1,405 GRB triggers
Fourth GRB catalog	2008-07-12 to 2018-07-11	2,356 cosmic GRBs

The official catalogs preserve a consistent prompt-emission parameterization. Durations are measured in 50–300 keV, and peak fluxes are reported on 64 ms, 256 ms, and 1.024 s timescales in both 50–300 keV and 10–1000 keV (Kienlin et al., 2014, Bhat et al., 2016, Kienlin et al., 2020). The data-access model is similarly stable: catalog tables and data files are available from HEASARC, with burst light curves, spectral data, and archive-specific FITS products supplied for downstream reanalysis (Kienlin et al., 2014, Kienlin et al., 2020).

The long time baseline makes the official sequence central to duration, hardness, flux-distribution, and localization studies. Over ten years, the catalog reports 395 nominally “short” bursts and 1,958 “long” bursts using the traditional $T_{90}=2$ s boundary, and notes that a two-dimensional uncertainty-weighted analysis suggests the short-burst fraction may climb to $\sim 25\%$ (Kienlin et al., 2020). Earlier installments report comparable short-burst fractions of 17% over four years and $\sim 21\%$ over six years (Kienlin et al., 2014, Bhat et al., 2016).

3. Measurement definitions and analysis workflow

The GRB catalogs use standard duration definitions based on cumulative background-subtracted counts or fluence. In the official GRB catalogs, $T_{90}$ is defined as the interval during which 90% of the background-subtracted counts are observed, from 5% to 95% of the total fluence,

$T_{90}=t_{95\%}-t_{5\%},$

and $T_{50}$ is defined analogously between 25% and 75% (Paciesas et al., 2012, Kienlin et al., 2014, Kienlin et al., 2020). Peak flux is the maximum flux over a sliding window, with standard windows of 64 ms, 256 ms, and 1.024 s, while fluence is the time-integrated model photon or energy flux over the burst interval (Kienlin et al., 2014, Bhat et al., 2016).

The core GRB-catalog workflow derives durations, peak fluxes, fluences, and hardnesses from background-subtracted, spectrally deconvolved photon-flux histories. In the first two-year catalog, the standard photon model is the COMP exponentially cut-off power law fitted in each time bin (Paciesas et al., 2012). The four-year and six-year catalogs describe a similar procedure: detector selection typically uses NaI detectors with source angles $<60^\circ$; background intervals are fitted with low-order polynomials; source intervals are chosen to encompass significant emission; and deconvolved flux histories are integrated over the required energy bands to obtain duration and fluence observables (Kienlin et al., 2014, Bhat et al., 2016).

Localization is treated separately from flux and duration measurement. Onboard locations are computed from relative count rates in the 12 NaI detectors against a 5°-grid lookup table, while ground-automated and manual refinements use a finer 1° grid and more accurate scattering and atmospheric corrections (Kienlin et al., 2014). The first two-year catalog describes localization through minimization of a detector-response $\chi^2(\theta,\phi)$, with systematic uncertainty modeled as two components, 2.6° with 72% weight and 10.4° with 28% weight (Paciesas et al., 2012). The second catalog states that a systematic uncertainty of $\sim 3^\circ$ (1$\sigma$) must be added in quadrature to the statistical error (Kienlin et al., 2014). The ten-year catalog adopts a systematic error model described as “core plus tail,” and for bursts with better localizations from Swift-BAT/XRT, LAT, IPN, INTEGRAL, or other instruments, the superior location is adopted (Kienlin et al., 2020).

The short-GRB catalog by Lu et al. modifies the temporal-selection stage by computing $\sim 25\%$0 in the GBM NaI band (8–1000 keV) via the Bayesian-Block algorithm and adopting $\sim 25\%$1 s as the typical short-burst selection criterion (Lu et al., 2017). This produces a catalog specifically optimized for short transients rather than the broader mission-wide GRB sample.

4. Spectral catalogs and time-resolved spectroscopy

The Fermi GBM Gamma-Ray Burst Spectral Catalogs are parallel products to the prompt-observable catalogs. The first two-year spectral catalog analyzes 487 GRBs and extracts two spectra per burst: a time-integrated “fluence” spectrum and a “peak-flux” spectrum from the single brightest time bin (Goldstein et al., 2012). The four-year spectral catalog expands this to 943 triggered GRBs and more than 7,500 fitted spectra (Gruber et al., 2014). The ten-year spectral catalog extends the uniform analysis to 2,297 GRBs out of 2,356 GBM triggers, giving 4,594 spectral data sets and approximately 18,376 individual fits (Poolakkil et al., 2021).

Across these spectral catalogs, four empirical models are fitted: power law, Comptonized or COMP, Band function, and smoothly broken power law. The ten-year catalog states that all models give photon flux in units of photons s$\sim 25\%$2 cm$\sim 25\%$3 keV$\sim 25\%$4 with pivot energy fixed at 100 keV (Poolakkil et al., 2021). Detector selection typically uses up to three NaI detectors with $\sim 25\%$5 plus the BGO with smallest angle, and fitting is performed in RMfit using a Poisson-likelihood statistic—Castor C-statistic in the earlier catalogs and C-Stat in the ten-year catalog (Goldstein et al., 2012, Gruber et al., 2014, Poolakkil et al., 2021).

The spectral catalogs also formalize model-quality and model-selection criteria. The ten-year catalog defines “GOOD” fits through model-specific parameter-uncertainty thresholds and defines the “BEST” model as the simplest GOOD fit unless a more complex model improves C-Stat by more than a critical $\sim 25\%$6C-Stat per extra degree of freedom (Poolakkil et al., 2021). The four-year spectral catalog describes a related procedure based on average critical values of $\sim 25\%$7 and $\sim 25\%$8 (Gruber et al., 2014).

The time-resolved spectral catalog is narrower in scope but much deeper in temporal analysis. It selects 81 bright bursts from the first four years using a fluence threshold $\sim 25\%$9 erg cm$\sim 21\%$0 and/or peak photon flux $\sim 21\%$1 ph s$\sim 21\%$2 cm$\sim 21\%$3, together with a minimum of five $\sim 21\%$4 time bins (Yu et al., 2016). This yields 1,802 individual time bins, 7,208 model fits, and a BEST sample of 1,491 spectra (Yu et al., 2016). The catalog identifies two “canonical” $\sim 21\%$5 evolution patterns—hard-to-soft and intensity-tracking—through Spearman-rank decision rules, and it reports that only 3 bursts, comprising 36 spectra in total, show evidence of a pure Planck function (Yu et al., 2016). It also states that averaged time-resolved low-energy power-law index and peak energy are slightly harder than time-integrated values, and concludes that time-resolved spectroscopic results should be used when interpreting physics from observed spectra instead of time-integrated results (Yu et al., 2016).

5. Specialized supplements and extensions

A major external supplement is the Interplanetary Network localization catalog. For the first Fermi GBM catalog, IPN data are provided for 427 of 491 bursts, and 149 of those yielded useful triangulation annuli (Hurley et al., 2013). The method uses arrival-time differences between spacecraft separated by distance $\sim 21\%$6, with annulus half-angle satisfying

$\sim 21\%$7

and annulus half-width derived from the time-delay uncertainty (Hurley et al., 2013). Among the 149 triangulated bursts, 3$\sim 21\%$8 annulus half-widths range from $\sim 21\%$9 to $T_{90}$0, with an average of 1.8°, while the resulting IPN error boxes are, on average, a factor of 180 smaller in area than the corresponding GBM localizations (Hurley et al., 2013). This supplement also provides a calibration template for combining GBM statistical and systematic localization errors.

Another extension is the dedicated short-GRB catalog of Lu et al. It analyzes 2,217 GBM triggers from August 2008 through August 2017, derives a catalog of 275 “typical” sGRBs by applying $T_{90}$1 s in the 8–1000 keV NaI band, and identifies 48 additional “GRB 170817A-like” weak events through a deeper Bayesian-Block search down to $T_{90}$2 above background (Lu et al., 2017). Each entry includes Trigger ID, $T_{90}$3, best-fit photon index $T_{90}$4 or Band parameters, $T_{90}$5, peak flux, fluence, PGSTAT/d.o.f., and light-curve pattern, with $T_{90}$6 additionally reported for GRB 170817A and deep-search candidates (Lu et al., 2017).

The Fermi-GBM 3-year X-ray Burst Catalog broadens the notion of a GBM burst catalog beyond extragalactic GRBs. It presents 1,084 X-ray bursts selected from a systematic search in continuous CTIME data and partitions them into 752 thermonuclear X-ray bursts, 267 accretion-powered flares and pulsations, and 65 untriggered GRBs (Jenke et al., 2016). The catalog states that the thermonuclear events are softer than $T_{90}$7, cluster in the Galactic plane, and yield a total photospheric radius-expansion burst rate of $T_{90}$8 d$T_{90}$9 integrated over all Galactic bursters within $T_{90}=t_{95\%}-t_{5\%},$0 kpc (Jenke et al., 2016).

Recent offline searches extend GBM burst-catalog practice below the onboard trigger threshold. The 11-year gamma-ray transient catalog conducts a blind search over 2010 July–2021 June using four search modes and three statistical methods—signal-to-noise ratio, Poisson, and Bayesian statistics—and delivers 12 machine-readable tables plus a public web interface (Kaneko et al., 16 Jan 2026). After SAA filtering, the catalog holds $T_{90}=t_{95\%}-t_{5\%},$1 million candidate events and includes known events such as GRBs, soft-gamma repeater bursts, galactic X-ray source activities, terrestrial gamma flashes, and solar flares, together with flagging results and classification probabilities (Kaneko et al., 16 Jan 2026). The 13-year coherent short-GRB search applies a fully coherent Poisson matched-filter analysis to CTTE data from 2013 to 2025, identifies 568 new GRB candidates with $T_{90}=t_{95\%}-t_{5\%},$2, and reports 1,736 temporally coincident Swift/BAT follow-ups with association probability above 90% (Perera et al., 29 May 2026). These newer products do not replace the official trigger catalogs; rather, they complement them by extending sensitivity to sub-threshold and otherwise missed short transients.

6. Population results, classification, and disputed subclass structure

Several statistical features recur across the GBM catalog family. The GRB catalogs report an isotropic sky distribution consistent with BATSE, a bimodal duration distribution with a division at 2 s, and a hardness–duration anti-correlation in which short bursts tend to be spectrally harder than long bursts (Paciesas et al., 2012, Kienlin et al., 2014, Kienlin et al., 2020). The six-year catalog states that statistical clustering in duration and hardness favors a two-component model with short-hard and long-soft bursts, and that a three-component model is disfavored by $T_{90}=t_{95\%}-t_{5\%},$3 in one dimension and by Monte Carlo tests in two dimensions (Bhat et al., 2016).

This conclusion is not the only classification result present in the literature. Horváth et al. analyzed 425 GBM bursts in the observable space $T_{90}=t_{95\%}-t_{5\%},$4 using PCA and Gaussian-mixture clustering, and found three groups: short GRBs, long-dim GRBs, and long-bright GRBs (Horvath et al., 2015). The paper explicitly states that additional analysis is needed to determine whether the bright/dim split of the long class is only a mathematical byproduct of the analysis or has some real physical meaning (Horvath et al., 2015). The coexistence of these results marks a genuine interpretive disagreement rather than a simple inconsistency: one analysis emphasizes model-based clustering in duration-hardness space, while the other incorporates peak flux and fluence and recovers a three-group structure.

The short-GRB catalog introduces a separate subclass question within the short population. Lu et al. report that 61% of their 275 typical short bursts show a single-episode light curve, designated Pattern I, while 39% show multiple episodes, designated Pattern II; only two clear precursors are found in the entire sample (Lu et al., 2017). Their duration distribution shows tentative bimodality with two Gaussian components at $T_{90}=t_{95\%}-t_{5\%},$5 and $T_{90}=t_{95\%}-t_{5\%},$6 and chance probability $T_{90}=t_{95\%}-t_{5\%},$7, and their $T_{90}=t_{95\%}-t_{5\%},$8 distribution shows a tentative bimodal distribution with peaks at $T_{90}=t_{95\%}-t_{5\%},$9 keV and $T_{50}$0 keV and $T_{50}$1 (Lu et al., 2017). No statistically significant correlation is found among $T_{50}$2, $T_{50}$3, and $T_{50}$4 (Lu et al., 2017). On that basis, the authors suspect that the current sample may include two distinct types of sGRBs from different progenitors and suggest that GRB 170817A-like events may be from NS-NS mergers while Pattern I light curves may be from another distinct type of compact binary (Lu et al., 2017). This is explicitly a proposed interpretation rather than a catalog-level classification consensus.

A further statistical theme concerns spectral phenomenology. The ten-year spectral catalog reports observer-frame median $T_{50}$5 keV in fluence spectra and $T_{50}$6 keV in peak-flux spectra, with short GRBs having systematically higher $T_{50}$7 and harder $T_{50}$8 than long GRBs (Poolakkil et al., 2021). The four-year spectral catalog states that some 17% of BEST fluence fits violate the synchrotron “line-of-death” and that $T_{50}$9 peaks at $<60^\circ$0 (Gruber et al., 2014). These catalog-level distributions do not resolve prompt-emission physics by themselves, but they define the observational parameter space against which synchrotron, photospheric, and mixed empirical interpretations are tested.

7. GRB 170817A and the catalog’s multimessenger role

GRB 170817A occupies a special position in the GBM short-burst literature because it is treated both as an individual event and as the template for a deeper weak-event search. In the short-GRB catalog, its duration is measured as $<60^\circ$1 s by Bayesian blocks, consistent with Fermi’s 0.647 s (Lu et al., 2017). Its cutoff-power-law fit gives $<60^\circ$2 keV and $<60^\circ$3, while the 8 keV–40 MeV peak flux and fluence are reported as $<60^\circ$4 erg cm$<60^\circ$5 s$<60^\circ$6 and $<60^\circ$7 erg cm$<60^\circ$8 (Lu et al., 2017). At $<60^\circ$9 Mpc, the corresponding isotropic quantities are

$\chi^2(\theta,\phi)$0

and

$\chi^2(\theta,\phi)$1

and the burst lies within the $\chi^2(\theta,\phi)$2 band of the typical short-GRB $\chi^2(\theta,\phi)$3–$\chi^2(\theta,\phi)$4 relation (Lu et al., 2017).

Within that same catalog, GRB 170817A is classified as a soft, weak sGRB with a Pattern II light curve (Lu et al., 2017). The deeper search for “GRB 170817A-like” weak events is therefore both a methodological extension and a population study. This suggests a shift in the role of GBM burst catalogs after the joint GW 170817/GRB 170817A discovery: the catalog is no longer only a record of detected prompt gamma-ray emission, but also a search space for sub-threshold counterparts to external messengers.

The broader catalog ecosystem reinforces that multimessenger function. The IPN supplement states that the average 180-fold reduction in localization area opens counterpart searches in optical, radio, gravitational waves, and neutrinos for bursts outside Swift’s field of view and for events too weak or too high in background for GBM-alone localizations (Hurley et al., 2013). The 13-year coherent short-GRB catalog explicitly frames its probabilistic ranking and joint GBM–BAT follow-up as a framework for rapidly identifying credible sub-threshold triggers for gravitational-wave, neutrino, and wide-field optical surveys (Perera et al., 29 May 2026). In that sense, the Fermi GBM Burst Catalog has evolved from a trigger-based prompt-emission archive into a layered observational infrastructure that supports localization, population studies, sub-threshold recovery, and joint transient inference across instruments and messengers.