An XMM–Newtonstudy of the hard X–ray sky

We report on the spectral properties of a sample of 90 hard X-ray selected serendipitous sources detected in 12 XMM–Newtonobservations with $1 \la{F_{2-10}}\la80 \times 10^{-14}$erg cm-2s-1. Approximately 40% of the sources are optically identified with $0.1 \la\ z\la2$and most of them are classified as broad line AGNs. A simple model consisting of power law modified by Galactic absorption offers an acceptable fit to ~65% of the source spectra. This fit yields an average photon index of $\langle\Gamma\rangle \approx 1.55$over the whole sample. We also find that the mean slope of the QSOs in our sample turns out to remain nearly constant ($\langle\Gamma\rangle \approx 1.8{-}1.9$) between $0 \la\ z\la2$, with no hints of particular trends emerging along z. An additional cold absorption component with $10^{21}\la{N_{\rm H}}\la10^{23}$cm-2is required in ~30% of the sources. Considering only subsamples that are complete in flux, we find that the observed fraction of absorbed sources (i.e. with ${N_{\rm H}}\ga\ 10^{22}$cm-2) is ~30%, with little evolution in the range $2 \la{F_{2-10}}\la\ 80 \times 10^{-14}$erg cm-2s-1. Interestingly, this value is a factor ~2 lower than predicted by the synthesis models of the CXB. This finding, detected for the first time in this survey, therefore suggests that most of the heavily obscured objects which make up the bulk of the CXB will be found at lower fluxes (${F_{2-10}}< 10^{-14}$erg cm-2s-1). This mismatch together with other recent observational evidences which contrast with CXB model predictions suggest that one (or more) of the assumptions usually included in these models need to be revised.