Uncovering low-level Fermi/GBM emission using orbital background subtraction

The secondary instrument onboard Fermi, the Gamma-ray Burst Monitor (GBM) is an all sky monitor consisting of 14 scintillation detectors. When analysing transient events such as Gamma-Ray Bursts (GRBs) and Solar Flares (SFs) the background is usually modelled as a polynomial (order 0-4). However, for long events the background may vary more than can be accounted for with a simple polynomial. In these cases a more accurate knowledge of GBM’s background rates is required. Here we present an alternative method of both determining the background and distinguishing low-level emission from the instrumental background.

💡 Research Summary

The paper addresses a fundamental limitation in the analysis of long‑duration, low‑level transient events observed by the Fermi Gamma‑Ray Burst Monitor (GBM). While GBM’s standard background modeling uses low‑order polynomials (0–4) that work well for the impulsive, short‑timescale emission typical of GRBs and solar flares, these models fail to capture the slowly varying background that can mask extended emission lasting from hundreds to thousands of seconds. To overcome this, the authors propose an orbital‑background subtraction technique that exploits the fact that Fermi returns to essentially the same geographic coordinates every 15 orbits (≈24 h). In principle, the background at a time T₀ could be estimated by averaging the detector rates at T₀ ± 15 orbits. However, because the spacecraft’s rocking angle repeats every two orbits in Sky Survey mode, the detectors that view a source at T₀ are pointed elsewhere at T₀ ± 15 orbits, rendering a direct comparison invalid.

Two practical solutions are explored: (1) using the rates from T₀ ± 30 orbits (≈48 h) where the pointing geometry matches that at T₀, and (2) averaging the rates from T₀ ± 14 and T₀ ± 16 orbits to approximate the T₀ ± 15‑orbit background. The authors test both approaches with four “blank‑sky” intervals (each ≈4 days long, free of triggers) selected between May 2009 and April 2011. For each interval they compute a background estimate over a 3500‑s window using continuous CSPEC data (128 energy channels, 4.096 s time resolution). Low‑energy NaI channels (<25 keV) and the overflow channel are excluded, yielding an effective NaI range of 25–1000 keV; for BGO the first and overflow channels are discarded, giving 0.1–45 MeV.

Residuals (observed minus estimated rates) are histogrammed for all NaI and BGO detectors and fitted with Gaussian functions. The fit parameters for the ±30‑orbit and the ±14/±16‑orbit methods are essentially identical: amplitudes around –0.4 to –0.9 counts s⁻¹ and standard deviations of 14–38 counts s⁻¹ for NaI, and similar values for BGO. This demonstrates that, statistically, both offset choices provide equally reliable background estimates when the spacecraft’s environment is stable.

A critical complication arises from passages through the South Atlantic Anomaly (SAA). Activation of the BGO crystals during SAA exposure produces elevated background levels that decay over time. Because the duration of SAA exposure varies with orbital precession (~52 days), the background estimated from ±30 orbits—whose SAA passage timing more closely matches that of the target interval—shows markedly smaller residuals than the ±14/±16‑orbit average in SAA‑affected periods. Consequently, the authors recommend using the ±30‑orbit offset whenever the interval of interest follows an SAA passage, while either offset may be used otherwise.

The method cannot be applied to GRBs that trigger an Autonomous Repoint Request (ARR). An ARR forces the spacecraft to slew so that the GBM‑derived source location lies within the LAT field of view, interrupting the regular rocking pattern for up to two hours (formerly five). This breaks the geometric correspondence required for the ±30 or ±14/±16‑orbit background estimation, limiting the technique’s applicability to ARR‑free events.

In summary, the study validates a simple yet powerful background‑subtraction strategy for GBM that leverages data from adjacent days when the satellite occupies the same geographic position. For the four trigger‑less intervals examined, the estimated backgrounds match the measured rates to within statistical uncertainties, and the Gaussian residual distributions confirm the absence of systematic bias. The ±30‑orbit and ±14/±16‑orbit offsets are interchangeable except in the vicinity of SAA passages, where the longer‑offset approach is superior. This technique opens the door to systematic searches for extended, low‑intensity emission in GBM data—such as GRB afterglows, long‑duration solar flares, or other faint high‑energy phenomena—provided the observations are not compromised by ARR slews or extreme SAA activation.