X-ray spectra of XMM-Newton serendipitous medium flux sources

We report on the results of a detailed analysis of the X-ray spectral properties of a large sample of sources detected serendipitously with the XMM-Newton observatory in 25 selected fields, for which optical identification is in progress. The survey covers a total solid angle of ~3.5 deg[superscript 2] and contains 1137 sources with ~10[superscript -15] < S[subscript 0.5-10] < 10[superscript -12] erg cm[superscript -2] s[superscript -1] with good enough spectral quality as to perform a detailed X-ray spectral analysis of each individual object. We find evidence for hardening of the average X-ray spectra of the sources towards fainter fluxes, and we interpret this as indicating a higher degree of photoelectric absorption amongst the fainter population. Absorption is detected at 95% confidence in 20% of the sources, but it could certainly be present in many other sources below our detection capabilities. For Broad Line AGNs (BLAGNs), we detect absorption in ~10% of the sources with column densities in the range 10[superscript 21] – 10[superscript 22] cm[superscript -2]. The fraction of absorbed Narrow Emission Line galaxies (NELGs, most with intrinsic X-ray luminosities >10[superscript 43] erg s[superscript -1], and therefore classified as type 2 AGNs) is significantly higher (40%), with a hint of moderately higher columns. After correcting for absorption, we do not find evidence for a redshift evolution of the underlying power law index of BLAGNs, which stays roughly constant at Γ ~ 1.9, with intrinsic dispersion of 0.4. A small fraction (~7%) of BLAGNs and NELGs require the presence of a soft excess, that we model as a black body with temperature ranging from 0.1 to 0.3 keV. Comparing our results on absorption to popular X-ray background synthesis models, we find absorption in only ~40% of the sources expected. This is due to a deficiency of heavily absorbed sources (with N[subscript H] ~ 10[superscript 22] – 10[superscript 24] cm[superscript -2]) in our sample in comparison with the models. We therefore conclude that the synthesis models require some revision in their specific parameters.