2015A&A...575A..30S

Robust X-ray temperature measurements of the intracluster medium (ICM) of galaxy clusters require an accurate energy-dependent effective area calibration. Since the hot gas X-ray emission of galaxy clusters does not vary on relevant timescales, they are excellent cross-calibration targets. Moreover, cosmological constraints from clusters rely on accurate gravitational mass estimates, which in X-rays strongly depend on cluster gas temperature measurements. Therefore, systematic calibration differences may result in biased, instrument-dependent cosmological constraints. This is of special interest in light of the tension between the Planck results of the primary temperature anisotropies of the cosmic microwave background (CMB) and Sunyaev-Zel'dovich-plus-X-ray cluster-count analyses. We quantify in detail the systematics and uncertainties of the cross-calibration of the effective area between five X-ray instruments, EPIC-MOS1/MOS2/PN onboard XMM-Newton and ACIS-I/S onboard Chandra, and the influence on temperature measurements. Furthermore, we assess the impact of the cross-calibration uncertainties on cosmology. Using the HIFLUGCS sample, consisting of the 64 X-ray brightest galaxy clusters, we constrain the ICM temperatures through spectral fitting in the same, mostly isothermal regions and compare the different instruments. We use the stacked residual ratio method to evaluate the cross-calibration uncertainties between the instruments as a function of energy. Our work is an extension to a previous one using X-ray clusters by the International Astronomical Consortium for High Energy Calibration (IACHEC) and is carried out in the context of IACHEC. Performing spectral fitting in the full energy band, (0.7-7)keV, as is typical of the analysis of cluster spectra, we find that best-fit temperatures determined with XMM-Newton/EPIC are significantly lower than Chandra/ACIS temperatures. This confirms the previous IACHEC results obtained with older calibrations with high precision. The difference increases with temperature, and we quantify this dependence with a fitting formula. For instance, at a cluster temperature of 10keV, EPIC temperatures are on average 23% lower than ACIS temperatures. We also find systematic differences between the three XMM-Newton/EPIC instruments, with the PN detector typically estimating the lowest temperatures. Testing the cross-calibration of the energy-dependence of the effective areas in the soft and hard energy bands, (0.7-2)keV and (2-7)keV, respectively, we confirm the previously indicated relatively good agreement between all instruments in the hard and the systematic differences in the soft band. We provide scaling relations to convert between the different instruments based on the effective area, gas temperature, and hydrostatic mass. We demonstrate that effects like multitemperature structure and different relative sensitivities of the instruments at certain energy bands cannot explain the observed differences. We conclude that using XMM-Newton/EPIC instead of Chandra/ACIS to derive full energy band temperature profiles for cluster mass determination results in an 8% shift toward lower Ω_M values and <1% change of σ₈ values in a cosmological analysis of a complete sample of galaxy clusters. Such a shift alone is insufficient to significantly alleviate the tension between Planck CMB primary anisotropies and Sunyaev-Zel'dovich-plus-XMM-Newton cosmological constraints.