We report the final optical identifications of the medium-depth (∼60 ks), contiguous (2 deg 2 ) XMM-Newton survey of the COSMOS field. XMM-Newton has detected ∼1800 X-ray sources down to limiting fluxes of ∼5 × 10 −16 , ∼3 × 10 −15 , and ∼7 × 10 −15 erg cm −2 s −1 in the 0.5-2 keV, 2-10 keV, and 5-10 keV bands, respectively (∼1 × 10 −15 , ∼6 × 10 −15 , and ∼1 × 10 −14 erg cm −2 s −1 , in the three bands, respectively, over 50% of the area). The work is complemented by an extensive collection of multiwavelength data from 24 μm to UV, available from the COSMOS survey, for each of the X-ray sources, including spectroscopic redshifts for 50% of the sample, and high-quality photometric redshifts for the rest. The XMM and multiwavelength flux limits are well matched: 1760 (98%) of the X-ray sources have optical counterparts, 1711 (∼95%) have IRAC counterparts, and 1394 (∼78%) have MIPS 24 μm detections. Thanks to the redshift completeness (almost 100%) we were able to constrain the high-luminosity tail of the X-ray luminosity function confirming that the peak of the number density of log L X > 44.5 active galactic nuclei (AGNs) is at z ∼ 2. Spectroscopically identified obscured and unobscured AGNs, as well as normal and star-forming galaxies, present well-defined optical and infrared properties. We devised a robust method to identify a sample of ∼150 high-redshift (z > 1), obscured AGN candidates for which optical spectroscopy is not available. We were able to determine that the fraction of the obscured AGN population at the highest (L X > 10 44 erg s −1 ) X-ray luminosity is ∼15%-30% when selection effects are taken into account, providing an important observational constraint for X-ray background synthesis. We studied in detail the optical spectrum and the overall spectral energy 348 No. 1, 2010 THE XMM-NEWTON WIDE-FIELD SURVEY IN THE COSMOS FIELD 349distribution of a prototypical Type 2 QSO, caught in a stage transitioning from being starburst dominated to AGN dominated, which was possible to isolate only thanks to the combination of X-ray and infrared observations.

The COSMOS-Legacy survey is a 4.6 Ms Chandra program that has imaged 2.2 deg2 of the COSMOS field with an effective exposure of

The Chandra COSMOS Survey (C-COSMOS) is a large, 1.8 Ms, Chandra program that has imaged the central 0.5 sq.deg of the COSMOS field (centered at 10 h , +02 o ) with an effective exposure of ∼160 ksec, and an outer 0.4 sq.deg. area with an effective exposure of ∼80 ksec. The limiting source detection depths are 1.9×10 −16 erg cm −2 s −1 in the Soft (0.5-2 keV) band, 7.3×10 −16 erg cm −2 s −1 in the Hard (2-10 keV) band, and 5.7×10 −16 erg cm −2 s −1 in the Full (0.5-10 keV) band. Here we describe the strategy, design and execution of the C-COSMOS survey, and present the catalog of 1761 point sources detected at a probability of being spurious of <2×10 −5 (1655 in the Full, 1340 in the Soft, and 1017 in the Hard bands). By using a grid of 36 heavily (∼50%) overlapping pointing positions with the ACIS-I imager, a remarkably uniform (±12%) exposure across the inner 0.5 sq.deg field was obtained, leading to a sharply defined lower flux limit. The widely different PSFs obtained in each exposure at each point in the field required a novel source detection method, because of the overlapping tiling strategy, which is described in a companion paper. This method produced reliable sources down to a 7-12 counts, as verified by the resulting logN-logS curve, with sub-arcsecond positions, enabling optical and infrared identifications of virtually all sources, as reported in a second companion paper. The full catalog is described here in detail and is available on-line.

Context. The COSMOS survey is a multiwavelength survey aimed to study the evolution of galaxies, AGN and large scale structures. Within this survey XMM-COSMOS a powerful tool to detect AGN and galaxy clusters. The XMM-COSMOS is a deep X-ray survey over the full 2 deg 2 of the COSMOS area. It consists of 55 XMM-Newton pointings for a total exposure of ∼1.5 Ms with an average vignetting-corrected depth of 40 ks across the field of view and a sky coverage of 2.13 deg 2 . Aims. We present the catalogue of point-like X-ray sources detected with the EPIC CCD cameras, the log N − log S relations and the X-ray colour-colour diagrams. Methods. The analysis was performed using the XMM-SAS data analysis package in the 0.5-2 keV, 2-10 keV and 5-10 keV energy bands. Source detection has been performed using a maximum likelihood technique especially designed for raster scan surveys. The completeness of the catalogue as well as log N − log S and source density maps have been calibrated using Monte Carlo simulations. Results. The catalogs contains a total of 1887 unique sources detected in at least one band with likelihood parameter det_ml > 10. The survey, which shows unprecedented homogeneity, has a flux limit of ∼1.7×10 −15 erg cm −2 s −1 , ∼9.3×10 −15 erg cm −2 s −1 and ∼1.3×10 −14 erg cm −2 s −1 over 90% of the area (1.92 deg 2 ) in the 0.5-2 keV, 2-10 keV and 5-10 keV energy band, respectively. Thanks to the rather homogeneous exposure over a large area, the derived log N − log S relations are very well determined over the flux range sampled by XMM-COSMOS. These relations have been compared with XRB synthesis models, which reproduce the observations with an agreement of ∼10% in the 5-10 keV and 2-10 keV band, while in the 0.5-2 keV band the agreement is of the order of ∼20%. The hard X-ray colors confirmed that the majority of the extragalactic sources in a bright subsample are actually type I or type II AGN. About 20% of the sources have a X-ray luminosity typical of AGN (L X > 10 42 erg/s) although they do not show any clear signature of nuclear activity in the optical spectrum.

A large population of heavily obscured, Compton-thick AGN is predicted by AGN synthesis models for the cosmic X-ray background and by the "relic" super-massive black-hole mass function measured from local bulges. However, even the deepest X-ray surveys are inefficient to search for these elusive AGN. Alternative selection criteria, combining mid-infrared with near-infrared and optical photometry, have instead been successful to pin-point a large population of Compton thick AGN. We take advantage of the deep Chandra and Spitzer coverage of a large area (more than 10 times the area covered by the Chandra deep fields, CDFs) in the COSMOS field, to extend the search of highly obscured, Compton-thick active nuclei to higher luminosity. These sources have low surface density and large samples can be provided only through large area surveys, like the COSMOS survey. We analyze the X-ray properties of COSMOS MIPS sources with 24µm fluxes higher than 550µJy. For the MIPS sources not directly detected in the Chandra images we produce stacked images in soft and hard X-rays bands. To estimate the fraction of Compton-thick AGN in the MIPS source population we compare the observed stacked count rates and hardness ratios to those predicted by detailed Monte Carlo simulations including both obscured AGN and star-forming galaxies. The volume density of Compton thick QSOs (logL(2-10keV)=44-45 ergs s −1 , or logλL λ (5.8µm)=44.79-46.18 ergs s −1 for a typical infrared to X-ray luminosity ratio) evaluated in this way is (4.8 ± 1.1) × 10 −6 Mpc −3 in the redshift bin 1.2-2.2. This density is ∼ 44% of that of all X-ray selected QSOs in the same redshift and luminosity bin, and it is consistent with the expectation of most up-to-date AGN synthesis models for the Cosmic X-ray background ). The density of lower luminosity Compton-thick AGN (logL(2-10keV)=43.5-44) at z=0.7-1.2 is (3.7 ± 1.1) × 10 −5 Mpc −3 , corresponding to ∼ 67% of that of X-ray selected AGN. The comparison between the fraction of infrared selected, Compton thick AGN to the X-ray selected, unobscured and moderately obscured AGN at high and low luminosity suggests that Compton-thick AGN follow a luminosity dependence similar to that discovered for Compton-thin AGN, becoming relatively rarer at high luminosities. We estimate that the fraction of AGN (unobscured, moderately obscured and Compton thick) to the total MIPS source population is 49 ± 10%, a value significantly higher than that previously estimated at similar 24µm fluxes. We discuss how our findings can constrain AGN feedback models.