A MULTIWAVELENGTH STUDY OF A SAMPLE OF 70 mu m SELECTED GALAXIES IN THE COSMOS FIELD. I. SPECTRAL ENERGY DISTRIBUTIONS AND LUMINOSITIES

International audienceWe present a large robust sample of 1503 reliable and unconfused 70 mu m selected sources from the multiwavelength data set of the Cosmic Evolution Survey. Using the Spitzer IRAC and MIPS photometry, we estimate the total infrared (IR) luminosity, L(IR) (8-1000 mu m), by finding the best-fit template from several different template libraries. The long-wavelength 70 and 160 mu m data allow us to obtain a reliable estimate of L(IR), accurate to within 0.2 and 0.05 dex, respectively. The 70 mu m data point enables a significant improvement over the luminosity estimates possible with only a 24 mu m detection. The full sample spans a wide range in IR luminosity, L(IR) approximate to 10(8)-10(14) L(circle dot), with a median luminosity of 10(11.4) L(circle dot). We identify a total of 687 luminous, 303 ultraluminous, and 31 hyperluminous infrared galaxies (LIRGs, ULIRGs, and HyLIRGs) over the redshift range 0.01 \textless z \textless 3.5 with a median redshift of 0.5. Presented here are the full spectral energy distributions (SEDs) for each of the sources compiled from the extensive multiwavelength data set from the ultraviolet (UV) to the far-infrared. A catalog of the general properties of the sample (including the photometry, redshifts, and L(IR)) is included with this paper. We find that the overall shape of the SED and trends with L(IR) (e.g., IR color temperatures and optical-IR ratios) are similar to what has been seen in studies of local objects; however, our large sample allows us to see the extreme spread in UV to near-infrared colors spanning nearly 3 orders of magnitude. In addition, using SED fits we find possible evidence for a subset of cooler ultraluminous objects than observed locally. However, until direct observations at longer wavelengths are obtained, the peak of emission and the dust temperature cannot be well constrained. We use these SEDs, along with the deep radio and X-ray coverage of the field, to identify a large sample of candidate active galactic nuclei (AGNs). We find that the fraction of AGNs increases strongly with L(IR), as it does in the local universe, and that nearly 70% of ULIRGs and all HyLIRGs likely host a powerful AGN