Cross-calibration of Suzaku/XIS and XMM-Newton/EPIC using galaxy clusters

Aims. We extend a previous cross-calibration study conducted by the International Astronomical Consortium for High Energy Calibration (IACHEC) on XMM-Newton/EPIC, Chandra/ACIS and BeppoSAX/MECS X-ray instruments with galaxy clusters to Suzaku/XIS instruments. Our aim is to study the accuracy of the energy-dependent effective area calibration of the XIS instruments using observations during 2005–2008. Methods. We performed a spatially resolved X-ray spectral analysis on Suzaku/XIS and XMM-Newton/EPIC-pn data for a sample of galaxy clusters. We extracted spectra from 3–6 arcmin annular regions. By comparing the spectroscopic temperatures, fluxes, and fit residuals obtained with different instruments for the same cluster, we evaluated the systematic uncertainties of the energy dependence and the normalisation of the effective area between different detectors. Results. The temperatures measured in the hard 2.0–7.0 keV energy band with all instruments are consistent within ~5%. Thus Suzaku/XIS can be added to the previously found list of instruments with good agreement of the effective area shape in the hard band (XMM-Newton/EPIC, Chandra/ACIS and BeppoSAX/MECS). However, using the public calibration, the temperatures obtained with the XIS instruments in the soft 0.5–2.0 keV band disagree by 9–29%. We investigated residuals in the XIS soft band, which showed that if the XIS0 effective area shape is accurately calibrated, the effective areas of XIS1 and XIS3 are overestimated below ~1.0 keV (or vice versa), with the difference increasing to ~20% at 0.5 keV. Adjustments to the modelling of the column density of the XIS contaminant in the 3–6 arcmin extraction region while forcing consistent emission models in each instrument for a given cluster significantly improved the fits. Assuming the composition of the contaminant as in the public calibration, the oxygen column density in the XIS1 and XIS3 contaminants must be increased by ~1–2 × 1017 cm-2 in comparison to the values implemented in the current calibration, while the column density of the XIS0 contaminant given by the analysis is consistent with the public calibration. This indicates that the effective area shape of XIS0 in the soft band is more accurately modelled in the public calibration than that of XIS1 and XIS3. XIS soft band temperatures obtained with the modification to the column density of the contaminant agree better with temperatures obtained with the EPIC-pn instrument of XMM-Newton than with th