Improvements in calibration of GSO scintillators in the Suzaku Hard X-ray Detector

Improvements of in-orbit calibration of GSO scintillators in the Hard X-ray Detector on board Suzaku are reported. To resolve an apparent change of the energy scale of GSO which appeared across the launch for unknown reasons, consistent and thorough re-analyses of both pre-launch and in-orbit data have been performed. With laboratory experiments using spare hardware, the pulse height offset, corresponding to zero energy input, was found to change by ~0.5 of the full analog voltage scale, depending on the power supply. Furthermore, by carefully calculating all the light outputs of secondaries from activation lines used in the in-orbit gain determination, their energy deposits in GSO were found to be effectively lower, by several percent, than their nominal energies. Taking both these effects into account, the in-orbit data agrees with the on-ground measurements within ~5%, without employing the artificial correction introduced in the previous work (Kokubun et al. 2007). With this knowledge, we updated the data processing, the response, and the auxiliary files of GSO, and reproduced the HXD-PIN and HXD-GSO spectra of the Crab Nebula over 12-300 keV by a broken powerlaw with a break energy of ~110 keV.

💡 Research Summary

The paper addresses a long‑standing calibration problem of the GSO (Gd₂SiO₅:Ce) scintillators in the Hard X‑ray Detector (HXD) aboard the Suzaku satellite. After launch, the energy scale of HXD‑GSO appeared to shift relative to the pre‑launch calibration, as demonstrated by the discrepancy between the cyclotron absorption line of the pulsar A0535+26 measured with HXD‑PIN (≈45 keV) and HXD‑GSO (≈60 keV). In earlier work (Paper II) an artificial non‑linearity term was introduced to force agreement, but the physical origin of the shift remained unknown.

The authors pursued two complementary lines of investigation: (1) hardware‑level tests with spare components, and (2) a re‑examination of the in‑orbit calibration methodology. Laboratory measurements showed that the analog pulse‑height offset (the “pedestal” value when no photon is deposited) depends on the power‑supply voltage. A change of roughly 0.5 % of the full ADC scale (≈20 channels) was observed when the supply voltage differed by a few volts. This offset shift explains the ∆H_i ≈ –40 channel offset seen in the in‑orbit data relative to ground calibrations.

The second effect concerns the activation lines (511 keV, 350 keV, 153 keV) used to monitor the gain G in orbit. The authors realized that the light yield associated with these lines is not exactly proportional to the nominal photon energy because secondary particles and non‑radiative de‑excitations reduce the scintillation output. By detailed Monte‑Carlo simulations and laboratory measurements they quantified the effective light‑output reduction: several percent lower than the nominal energies for each line.

Incorporating both the pedestal shift and the corrected light‑output factors into the energy‑to‑pulse‑height conversion yields a revised gain calibration that aligns the in‑orbit data with the ground measurements within ~5 %, eliminating the need for the artificial correction previously applied. The authors updated the data‑processing pipeline, generated new response matrix files (RMFs) and ancillary response files (ARFs), and validated the new calibration using the Crab Nebula spectrum from 12 to 300 keV. The Crab spectrum is well described by a broken power‑law with a break energy of ~110 keV, photon indices of –2.07 (below) and –2.34 (above), and a reduced χ² close to unity, confirming the correctness of the revised calibration.

The paper also discusses ancillary issues that were observed after launch: a widening of the GSO branch in the slow‑fast pulse‑shape discrimination diagram, and an over‑prediction of the GSO response below 100 keV by the previous response matrix. Both anomalies are mitigated by the new calibration, which corrects the underlying analog gain and offset variations.

Finally, the authors recommend regular monitoring of the pedestal level, verification of power‑supply stability, and periodic re‑evaluation of activation‑line light yields to maintain calibration accuracy over the mission lifetime. This work demonstrates how meticulous hardware characterization combined with careful re‑analysis of calibration data can resolve apparent in‑orbit performance changes without resorting to ad‑hoc corrections, thereby improving the scientific reliability of high‑energy X‑ray observations.