Statistical evaluation of the flux cross-calibration of the XMM-Newton EPIC cameras

Context. The second XMM-Newton serendipitous source catalogue, 2XMM, provides the ideal data base for performing a statistical evaluation of the flux cross-calibration of the XMM-Newton European Photon Imaging Cameras (EPIC). Aims. We aim to evaluate the status of the relative flux calibration of the EPIC cameras on board the XMM-Newton observatory (MOS1, MOS2, and pn) and investigate the dependence of the calibration on energy (from 0.2 to 12.0 keV), position of the sources in the field of view of the X-ray detectors, and lifetime of the mission. Methods. We compiled the distribution of flux percentage differences for large samples of good quality objects detected with at least two of the EPIC cameras. The mean offset of the fluxes and dispersion of the distributions was then found by Gaussian fitting. Count rate to flux conversion was performed with a fixed spectral model. The impact on the results of varying this model was investigated. Results. Excellent agreement was found between the two EPIC MOS cameras to better than 4% over the entire energy range where the EPIC cameras are best calibrated (0.2-12.0 keV). We found that MOS cameras register 7-9% higher flux than EPIC pn below 4.5 keV and a 10-13% flux excess at the highest energies (greater than or similar to 4.5 keV). No evolution of the flux ratios is seen with time, except at the lowest energies (less than or similar to 0.5 keV), where we found a strong decrease in the MOS to pn flux ratio with time. This effect is known to be due to a gradually degrading MOS redistribution function. The flux ratios show some dependence on distance from the optical axis in the sense that the MOS to pn flux excess increases with off-axis angle. Furthermore, in the 4.5-12.0 keV band there is a strong dependence of the MOS to pn excess flux on the azimuthal-angle. These results strongly suggest that the calibration of the Reflection Grating Array (RGA) blocking factors is incorrect at high energies. Finally, we recommend ways to improve the calculation of fluxes in future versions of XMM-Newton serendipitous source catalogues.