Calibration of the INTEGRAL SPI Anti Coincidence Shield with Gamma Ray Bursts observations

The Anti Coincidence Shield (ACS) of the INTEGRAL SPI instrument provides an excellent sensitivity for the detection of Gamma Ray Bursts (GRBs) above ~ 75keV, but no directional and energy information is available. We studied the ACS response by using GRBs with known localizations and good spectral information derived by other satellites. We derived a count rate to flux conversion factor for different energy ranges and studied its dependence on the GRB direction and spectral hardness. For a typical GRB spectrum, we found that 1 ACS count corresponds on average to ~ 1E-10 erg/cm^2 in the 75keV-1MeV range, for directions orthogonal to the satellite pointing axis. This is broadly consistent with the ACS effective area derived from the Monte Carlo simulations, but there is some indication that the latter slightly overestimates the ACS sensitivity, especially for directions close to the instrument axis.

💡 Research Summary

The paper presents an in‑flight calibration of the Anti‑Coincidence Shield (ACS) of the INTEGRAL SPI instrument using Gamma‑Ray Bursts (GRBs) with well‑determined positions and spectra obtained from other satellites. The ACS, composed of 91 BGO crystals surrounding the SPI spectrometer, is highly sensitive to photons above ~75 keV but provides no directional or energy information. By exploiting GRBs detected simultaneously by ACS and by instruments such as Swift‑BAT, Fermi‑GBM, and Konus‑Wind, the authors derive a conversion factor that translates ACS count rates into physical flux units.

Data selection and sample construction
The authors examined the GCN archive and the Swift GRB catalogue for the period 1 January 2003 to 30 June 2009, identifying 764 GRBs. After filtering out events that were not visible in the ACS light curves (due to satellite perigee passages, IBIS shielding, missing spectral data, or large positional uncertainties), they retained 196 GRBs with usable ACS detections. Some multi‑peak bursts were split into separate events, yielding a final sample of 205 ACS‑detected GRB episodes.

Spectral characterization
For calibration the authors relied primarily on GRBs observed by Konus‑Wind (KW) and Fermi‑GBM, because these instruments cover a broad energy range (15 keV–10 MeV for KW, 10 keV–30 MeV for GBM) that overlaps the ACS response. After discarding a few events with incomplete ACS light curves or missing error estimates, a “spectral sample” of 133 bursts remained (62 GBM, 71 KW). Most spectra were fitted with the Band function or a Cut‑off Power‑Law (CPL). The mean low‑energy photon index is ⟨α⟩ = ‑0.86 ± 0.30, the high‑energy index ⟨β⟩ = ‑2.31 ± 0.30, and the characteristic break energy ⟨E₀⟩ ≈ 448 keV.

Deriving the count‑to‑flux conversion
For each burst the authors computed the fluence in the ACS energy band (75 keV to a chosen E_max) by extrapolating the KW/GBM spectral fits. They defined a conversion factor k = f_ACS / N_ACS, where f_ACS is the fluence in physical units (10⁻⁷ erg cm⁻²) and N_ACS is the measured ACS fluence in counts (scaled to 1000 counts). The factor k was evaluated for four upper‑energy limits (0.75, 1, 5, and 10 MeV). Because k showed a large dispersion, the authors investigated its dependence on the incident angle θ (the angle between the burst direction and the SPI pointing axis) and on spectral hardness.

Angular dependence
The sample was divided into three angular zones: top (θ < 45°), central (45° < θ < 120°, excluding the region shadowed by IBIS), and bottom (θ > 120°). The weighted mean conversion factors (⟨k⟩) decrease from top to bottom, reflecting the higher ACS effective area for directions orthogonal to the spacecraft axis. For the full sample with E_max = 10 MeV, ⟨k⟩ ≈ 1.0 × 10⁻⁷ erg cm⁻² / 1000 counts; for the bottom zone ⟨k⟩ ≈ 0.75 × 10⁻⁷, while the top zone yields ⟨k⟩ ≈ 1.37 × 10⁻⁷. The reduced chi‑square values indicate that the dispersion cannot be explained by statistical errors alone, confirming a genuine dependence on direction and spectrum.

Spectral hardness effect
Hardness was quantified by a ratio HR = (H − S)/(H + S), where S and H are the extrapolated fluences below and above a threshold energy E_T = 500 keV. The authors found that the spread of k increases with HR, especially for the hardest bursts, but no clear monotonic trend of ⟨k⟩ versus HR emerged. This suggests that uncertainties in high‑energy extrapolation dominate the scatter for hard spectra.

Comparison with Monte‑Carlo simulations
Using the MGGPOD Monte‑Carlo model of the ACS, the authors computed the effective area A_eff(E,θ,φ) for a grid of directions. By folding the mean Band spectrum through A_eff they derived a simulated conversion factor k_sim(θ). For E_max = 10 MeV, k_sim ranges from ~1.0 (orthogonal) to ~1.6 (near‑axis); for E_max = 1 MeV the range is ~0.7–1.0. These simulated values are systematically lower than the empirically derived ⟨k⟩, especially for the top and bottom angular zones, indicating that the Monte‑Carlo model slightly overestimates the ACS sensitivity, likely due to simplifications in the spacecraft geometry or shielding.

Conclusions and outlook
The study provides a practical conversion factor: for a typical GRB spectrum and a direction orthogonal to the SPI axis, 1 ACS count corresponds to ≈ 10⁻¹⁰ erg cm⁻² in the 75 keV–1 MeV band. For non‑orthogonal directions the factor is larger by a factor of 2–3. Despite the intrinsic dispersion caused by angular and spectral diversity, the calibration enables rough fluence estimates from ACS data alone, which is valuable for GRB studies when only ACS triggers are available. The authors anticipate that the rapidly growing Fermi‑GBM database will allow a larger, more spectrally homogeneous sample, reducing uncertainties and refining the directional dependence of the conversion factor. Future work may also involve improving the Monte‑Carlo mass model to reconcile simulated and observed sensitivities.