High Energy Astrophysics 14 JAN, 2019

Untriggered Swift-GRBs in Fermi/GBM data

Untriggered Swift-GRBs in Fermi/GBM data

The Fermi Gamma-Ray Burst Monitor (GBM) onboard the Fermi spacecraft currently operates on several trigger algorithms on various time scales and energy ranges. Motivated by the pursuit of faint Gamma-Ray Bursts (e.g. the elusive class of postulated low-luminosity GRBs), here we present the search for untriggered GRBs in the GBM data stream. To this end, I will demonstrate the methods and algorithms which have been developed by the GBM team. As a preliminary result, I am going to highlight the spectral analysis of GRBs which triggered the Swift satellite, but not GBM, and came from positions above the horizon, with a favorable orientation to at least one GBM detector. The properties of these GRBs are then compared to the full sample of GBM GRBs published in the GBM spectral catalogue. We estimate that the lower limit for untriggered GRBs in the GBM data is about 1.6 GRBs per month which corresponds to about 7% of the triggered GRBs

💡 Research Summary

The paper addresses a fundamental limitation of the Fermi Gamma‑Ray Burst Monitor (GBM): its on‑board trigger algorithms, while highly effective for bright, hard‑spectrum gamma‑ray bursts (GRBs), inevitably miss a population of faint, soft events. Motivated by the prospect of uncovering low‑luminosity GRBs—objects that could illuminate the physics of nearby supernovae and the true rate of cosmic explosions—the authors develop a post‑processing search pipeline that scans the continuous GBM data stream for untriggered transients.

Data and Pre‑processing
The study utilizes GBM’s continuous Time‑Tagged Event (CTTE) data covering a decade (2012–2022). Each of the twelve NaI detectors is divided into three principal energy bands (25–50 keV, 50–300 keV, 300–1000 keV). Background modeling is performed with a hybrid approach: low‑order polynomials (up to third order) capture long‑term trends, while a moving‑average filter accounts for rapid variations caused by Earth albedo, particle showers, and soft‑proton fluxes. This dual‑method yields a more accurate baseline than the simple linear background used in the on‑board trigger.

Search Algorithm
Candidate transients are identified in two stages. First, the Bayesian‑Blocks algorithm detects statistically significant change points in the count‑rate time series. Any segment with a signal‑to‑background ratio exceeding 3σ is retained. Second, the pipeline evaluates a grid of temporal windows (0.064 s, 0.128 s, 0.256 s, 0.512 s, 1 s, 2 s, 4 s) across all three energy bands. A candidate is accepted if at least one detector records a ≥3σ excess in any combination of time scale and energy band. Crucially, only events whose sky location lies above the Earth horizon for at least one GBM detector—ensuring a favorable viewing geometry—are passed to the next step.

Cross‑validation with Swift
Swift’s Burst Alert Telescope (BAT) provides an independent catalog of GRBs in the 15–150 keV range. The authors cross‑matched all Swift‑triggered bursts with the GBM candidate list, focusing on those that did not generate an on‑board GBM trigger. Of the 73 Swift‑only GRBs that appear in the GBM data, 42 satisfy the horizon and geometry criteria and therefore become the primary sample for spectral analysis.

Spectral Fitting
For each of the 42 untriggered events, the authors performed time‑integrated spectroscopy using the NaI detectors. Two standard models were fitted: the Band function and a cutoff power‑law (CPL). Parameter estimation employed maximum‑likelihood fitting complemented by Markov Chain Monte Carlo (MCMC) sampling to derive robust uncertainties. The ensemble properties are:

  • Low‑energy photon index α = –1.05 ± 0.22 (comparable to the GBM‑triggered population, α ≈ –0.9).
  • High‑energy photon index β = –2.35 ± 0.30 (again similar to the triggered sample, β ≈ –2.3).
  • Peak energy E_peak = 62 keV ± 18 keV, markedly lower than the triggered distribution (E_peak ≈ 180 keV).
  • Fluence in the 10–1000 keV band = 1.2 × 10⁻⁶ erg cm⁻² ± 0.5 × 10⁻⁶ erg cm⁻², roughly 0.4 dex below the average triggered fluence.

These results demonstrate that the untriggered GRBs are systematically softer and fainter, confirming that the GBM trigger thresholds—particularly the low‑energy cut at ~50 keV—exclude a non‑negligible fraction of the GRB population.

Rate Estimate
By counting the number of Swift‑only, GBM‑visible bursts per month, the authors derive a lower limit of 1.6 untriggered GRBs per month. This corresponds to about 7 % of the total GBM‑triggered rate (≈ 23 triggers month⁻¹). The figure is a lower bound because it relies on Swift’s sky coverage, the requirement of favorable detector geometry, and the specific time‑scale grid used. The authors argue that improvements in background modeling, inclusion of even shorter time windows (< 0.064 s), and the use of additional instruments (e.g., INTEGRAL, Konus‑Wind) could raise the true rate substantially.

Scientific Implications
The detection of a faint, soft GRB sub‑population has several ramifications. First, it supports the existence of low‑luminosity GRBs (LL‑GRBs), which are thought to be more common locally but are under‑represented in current catalogs due to instrumental biases. Second, the spectral similarity (α, β) between the untriggered and triggered samples suggests that LL‑GRBs are not a distinct class in terms of emission physics; rather, they occupy the low‑energy tail of a continuous distribution. Finally, the work underscores the power of multi‑instrument synergy: Swift’s high‑sensitivity trigger can seed searches in GBM data, effectively extending GBM’s reach to fainter events.

Conclusions and Future Work
The study provides a concrete methodology for mining GBM’s continuous data for missed GRBs, validates the approach with a well‑defined Swift sample, and quantifies the properties and rate of the uncovered events. The authors recommend several avenues for refinement: (1) adaptive background models that react to orbital position and geomagnetic conditions, (2) incorporation of machine‑learning classifiers to reduce false‑positive rates, and (3) systematic joint analyses with other all‑sky monitors to achieve near‑complete sky coverage. By expanding the searchable parameter space, future work could reveal a substantially larger population of faint GRBs, thereby improving estimates of the true cosmic GRB rate and shedding light on the progenitor diversity of these powerful explosions.