The composite nature of Dust-Obscured Galaxies (DOGs) at z~2-3 in the COSMOS field: I. A Far-Infrared View
L. Riguccini, E. Le Floc'h, J.R. Mullaney, K. Menendez-Delmestre, H. Aussel, S. Berta, J. Calanog, P. Capak, A. Cooray, O. Ilbert, J. Kartaltepe, A. Koekemoer, D. Lutz, B. Magnelli, H. McCracken, S. Oliver, I. Roseboom, M. Salvato, D. Sanders, N. Scoville, Y. Taniguchi, E. Treister
MMon. Not. R. Astron. Soc. , 1–17 (2002) Printed 16 August 2018 (MN L A TEX style file v2.2)
The composite nature of Dust-Obscured Galaxies (DOGs)at z ∼ L. Riguccini , (cid:63) , E. Le Floc’h , J.R. Mullaney , K. Men´endez-Delmestre , H. Aussel ,S. Berta , J. Calanog , P. Capak , A. Cooray , O. Ilbert , J. Kartaltepe ,A. Koekemoer , D. Lutz , B. Magnelli , H. McCracken , S. Oliver , I. Roseboom ,M. Salvato , D. Sanders , N. Scoville , Y. Taniguchi , E. Treister , Observat´orio do Valongo, Universidade Federal do Rio de Janeiro, Ladeira do Pedro Antˆonio 43, Sa´ude,Rio de Janeiro, RJ 20080-090, Brazil ([email protected]) CAPES/BJT Science Without Borders Postdoctoral Fellow, Brazil Laboratoire AIM, CEA/DSM-CNRS-Universit´e Paris Diderot, IRFU/Service d’Astrophysique, Bˆat.709, CEA-Saclay,91191 Gif-sur-Yvette Cedex, France Department of Physics and Astronomy, Hicks Building, University of Sheffield, S3 7RH, U.K. Max-Planck-Institut f¨ur extraterrestrische Physik, Postfach 1312, Giessenbachstrasse 1, 85741 Garching, Germany Center for Cosmology, Department of Physics and Astronomy, University of California, Irvine, CA 92697, USA Spitzer Science Center, 314-6 Caltech, Pasadena, CA 91125, USA Laboratoire d’Astrophysique de Marseille, BP 8, Traverse du Siphon, 13376 Marseille Cedex 12, France National Optical Astronomy Observatory, 950 N. Cherry Ave., Tucson, AZ, 85719, USA Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA Institut d’Astrophysique de Paris, UMR7095 CNRS, Universit´e Pierre et Marie Curie, 98 bis Boulevard Arago, 75014 Paris, France Astronomy Centre, Department of Physics & Astronomy, University of Sussex, Brighton BN1 9QH, UK Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ, UK Max-Planck-Institute f¨ur Plasma Physics, Boltzmann Strasse 2, Garching 85748, Germany Institute for Astronomy, 2680 Woodlawn Dr., University of Hawaii, Honolulu, HI 96822, USA California Institute of Technology, MC 105-24, 1200 East California Boulevard, Pasadena, CA 91125, USA Research Center for Space and Cosmic Evolution, Ehime University, Bunkyo-cho, Matsuyama 790-8577, Japan Universidad de Concepci´on, Departamento de Astronom´ıa, Casilla 160-C, Concepci´on, Chile
Accepted 2015 June 08. Received 2015 May 15; in original form 2014 September 23.
ABSTRACT
Dust-Obscured galaxies (DOGs) are bright 24 µ m-selected sources with extreme obscu-ration at optical wavelengths. They are typically characterized by a rising power-lawcontinuum of hot dust (T D ∼ µ m flux display a stellar bump in the near-IR and their mid-IR luminosity appearsto be mainly powered by dusty star formation. Alternatively, it may be that the mid-IR emission arising from AGN activity is dominant but the torus is sufficiently opaqueto make the near-IR emission from the AGN negligible with respect to the emissionfrom the host component. In an effort to characterize the astrophysical nature of theprocesses responsible for the IR emission in DOGs, this paper exploits Herschel data(PACS + SPIRE) on a sample of 95 DOGs within the COSMOS field. We derivea wealth of far-IR properties (e.g., total IR luminosities; mid-to-far IR colors; dusttemperatures and masses) based on SED fitting. Of particular interest are the 24 µ m-bright DOGs (F µm > µ m flux increases as a function of the rest-frame 8 µ m-luminosityirrespective of the redshift. This confirms that faint DOGs (L µm < L (cid:12) ) aredominated by star-formation while brighter DOGs show a larger contribution from anAGN. Key words:
Galaxies: high-redshift - Infrared: galaxies - Cosmology: observations (cid:63)
E-mail: [email protected] (cid:13) a r X i v : . [ a s t r o - ph . GA ] J un Riguccini et al.
The unprecedented sensitivity and angular resolution of the
Spitzer
Space Telescope at infrared (IR) wavelengths ledto the discovery of a new type of galaxy that is extremelyfaint in the optical ( ∼ < R < f ν (24 µm ) /f ν ( R ) > µ m detected sourcesare DOGs, while ∼
40 % of the sources in the 2 deg COS-MOS field optical catalog have similar R-band magnitude[22.4-26.4]. However, their contribution to the total IR out-put of the Universe at z ∼ L IR > L (cid:12) ; e.g., Riguccini et al. 2011). DOGshave IR luminosities > L (cid:12) placing them in the LIRGand ULIRG class of galaxies (e.g., Dey et al. 2008; Buss-mann et al. 2009; Riguccini et al. 2011). Such luminositiesrequire significant amounts of dust-heating, most probablyarising from star-formation and/or high levels of nuclear ac-tivity (i.e., active galactic nucleus or AGN). A number ofrecent studies have split the DOG population along theselines: i.e., DOGs showing a “bump” at 1.6 µ m indicative ofstar-formation (Farrah et al. 2008; Desai et al. 2009, here-after bump DOGs) and DOGs displaying a rising power-lawSED in the near- to mid-IR bands, suggesting a dominantAGN (Houck et al. 2005; Weedman et al. 2006, hereafterPL-DOGs). Estimating the star-formation rate of the latterhas proved extremely difficult due to the dominant AGNcomponent washing out any host galaxy signatures.The faintness of DOGs at optical wavelengths has madethe characterization of their physical properties particularlychallenging. The launch of the Herschel Space Telescope in 2009 provided a new window onto these galaxies thatis largely independent of dust obscuration, thereby givingus the clearest view yet of these galaxies. The wavelengthsprobed by Herschel cover the peak of the spectral energy dis-tribution (hereafter, SED) of DOGs at the redshifts wheretheir numbers are highest (i.e., 1 . (cid:46) z (cid:46) Spitzer and
Herschel observations to determinehow these properties relate to the dominant source of energyin these galaxies be it AGN, intense star formation or a com-bination of both. For this we use a sample of Spitzer/MIPS24 µ m-selected DOGs (satisfying F µm > .
08 mJy) se-lected from the COSMOS field (Scoville et al. 2007), anddetected in all 5
Herschel bands. We calculate the contri-bution from AGN and/or star-formation to the total energyoutput of these galaxies via SED fitting and relate this totheir dust temperature and masses.The paper is organized as follows. Our data are de- Luminous Infra-red Galaxies with 10 L (cid:12) < L IR < L (cid:12) and Ultra-Luminous Infra-red Galaxies with L IR > L (cid:12) (e.g.,Sanders et al. 1988a,b) scribed in Sect. 2, the far- to mid-IR colors of DOGs sourcesare detailed in Sect. 3. The SED-fitting procedure used andthe results obtained are described in Sect. 4. In Sect. 5 wepresent the model and results on the dust temperature andmass of our DOG sample and discuss if the presence of AGNsignatures induce a particular trend in the T dust distribu-tion. We discuss our results and present our conclusions inSect. 6. Throughout this paper we assume a ΛCDM cosmol-ogy with H =70 km s − , Ω m = 0.3, and Ω Λ = 0.7. Unlessotherwise specified, magnitudes are given in the AB system. The sample of DOG sources is selected from the deep
Spitzer /MIPS observations of the 2 deg COSMOS field(Sanders et al. 2007). Our starting point are the 24 µ m de-tected sources from the catalogue described in Le Floc’het al. (2009) (see also Riguccini et al. 2011). We note thatother studies further require a source to satisfy f µm >
300 mJy in order to classify it as a DOG (e.g., Deyet al. 2008). In this study we consider all sources satisfying f ν (24 µm ) /f ν ( R ) >
982 as DOGs. Furthermore, DOG stud-ies focussing on heavily-obscured AGNs (e.g., Fiore et al.2008, 2009; Treister et al. 2009) also impose an additionalR-K > COSMOS is a wide-area equatorial field with deep cover-age at all wavelengths spanning radio to X-rays (Hasingeret al. 2007; Schinnerer et al. 2007; Elvis et al. 2009). Crucialfor this study is the deep IR coverage of this field, partic-ularly at mid- to far-IR wavelengths by the
Spitzer SpaceTelescope with the MIPS instrument (Le Floc’h et al. 2009)and, more recently, with PACS (Poglitsch et al. 2010) andSPIRE (Griffin et al. 2010) onboard
Hershel (Pilbratt et al.2010).The extensive UV to near-IR coverage of COSMOS(e.g., Taniguchi et al. 2007; Capak et al. 2007) allows forprecise photometric redshifts (hereafter, photo-z) to be de-rived for extragalactic sources within this field. For thephoto-zs used in this work, we use an updated version ofthe photometric redshift catalog of Ilbert et al. (2009) thatprovides photo-zs for 1 400 237 i + -detected sources amongthe 2 017 800 sources of the COSMOS photometric catalog.These redshifts are obtained with an unprecedented accu-racy, with a dispersion of σ ∆ z/ (1+ z ) = 0 .
012 for sources sat-isfying i + AB <
24 and z < − where we focus on dusty 24 µm -selected sources that are veryfaint at optical wavelengths − is their comparison with theoptically-faint spectroscopic sample from the z-COSMOSsurvey (Lilly et al. 2007) where Ilbert et al. (2009) re-port a dispersion of only σ ∆ z/ (1+ z ) = 0.06 for sources with23 < i + AB <
25 at 1.5 < z <
3. Given their accuracy for faintsources, we use these photo-zs for our 24 µ m sources, match-ing their optical counterparts following the procedure out-lined in Le Floc’h et al. (2009) and Riguccini et al. (2011). version 1.8: the main improvements compared to Ilbert et al.(2009) reside in relying on the median of the PDF to define the“best” photo-z, instead of the minimum χ c (cid:13) , 1–17 omposite nature of Dust-Obscured Galaxies (DOGs) at z ∼ We briefly describe this approach in the following subsec-tion. µ m-selectedsources Our 24 µ m parent sample (from Le Floc’h et al. (2009) andRiguccini et al. (2011)) contains 29 395 sources detected at24 µ m with F µm > µ Jy over a total area of 1.68 deg ,which excludes regions contaminated by bright, saturatedobjects. In the interest of focussing on the sources’ star-formation histories, we exclude X-ray detected AGNs downto a flux limit of S . − kev = 5 × − erg cm − s − basedon AGN catalogs from Brusa et al. (2007, 2010) and Salvatoet al. (2009).We limit our counterpart identification to 24 µ m sourceswith a 3- σ PACS detection at 100 µ m and 160 µ m. SPIREfluxes will be used for a subset of our sample. The COSMOSfield was observed as part of the PACS Evolutionary Probe(PEP, Lutz et al. 2011) and the Herschel Multi-tiered Ex-tragalactic Survey (HerMES, Oliver et al. 2012) campaigns(i.e., PACS 100 & 160 µ m and SPIRE 250, 350 & 500 µ m re-spectively). The catalogs provided by PEP and HerMES cal-culate source fluxes in each of these 5 bands by performingPSF fitting at the positions of the 24 µ m-detected sourcesfrom Le Floc’h et al. (2009). One of the key benefits of us-ing such 24 µ m “priors” as opposed to generating blind cat-alogues, is that it helps with deblending, which is particu-larly problematic at the longer Herschel wavelengths. TheHerMES catalog was built following the method presentedin Roseboom et al. (2010), based on the 24 µ m position pri-ors from Le Floc’h et al. (2009). The PEP catalog was ob-tained using the same µ m priors (Berta et al. 2011). Thereliability and the completeness of the PACS and SPIRECOSMOS catalogs are detailed in Lutz et al. (2011) andOliver et al. (2012), respectively. We identify a total of 6 02924 µ m-detected sources with a > σ detection in the PACS-bands with F µm > µm > µ m-selected catalog and the optical observa-tions could lead to a large number of spurious associationswith optically-detected galaxies randomly-aligned close tothe line of sight of the MIPS sources. To minimize this,we first matched our 24 µ m catalogue to the K-band cat-alogue of McCracken et al. (2010), employing a matchingradius of 2” and following the same procedure described inLe Floc’h et al. (2009) and Riguccini et al. (2011). Of the6 029 sources in our sample, 5858 were found to have a K-band counterpart. In an attempt to reduce the number ofnon-matches, we also matched our 24 µ m catalogue to theIRAC-3.6 µ m catalog from Capak et al. (2007), adopting thesame 2” matching radius. This led to 34 additional matches,increasing to 5892 the number of MIPS-24 µ m+Herschelsources with either a K-band or an IRAC-3.6 µ m counter-part. These 5892 sources were then matched to the updatedversion of the i + -band selected catalog of photometric red-shifts from Ilbert et al. (2009) using a matching radius of Figure 1.
Photometric redshift distribution of the 24 µ m sourcesfrom Le Floc’h et al. (2009) in black and of the PACS-DOGs fromthis work in purple. + -band counterparts andassociated photometric redshifts, leaving 261 sources amongthe 6 029 24 µ m+Herschel sources (i.e., ≈ Our DOGs sample was selected from the 5 892 sources se-lected at 24 µ m with a 3- σ detection in at least one ofthe PACS bands (100 or 160 µ m). The DOGs criterion in-troduced by Dey et al. (2008) is based on the following: F µm /F R >
982 and F µm > µ Jy, where the latter isa direct consequence of the depth of the MIPS imaging inthe Bootes field. Considering that the source-extraction per-formed by Le Floc’h et al. (2009) reaches a completeness of ∼
90% with F µm > µ Jy, we extend the DOGs 24 µ m-flux cut down to 80 µ Jy. The R-band magnitudes used inthis work are from Ilbert et al. (2009) based on observationswith the Subaru telescope by Capak et al. (2007); these in-clude a correction for Galactic extinction – not applied inCapak et al. (2007) – and reach a limiting magnitude ofM R > .The sample contains 57 and 138 DOGs detected inonly one of the PACS bands at 100 µ m and 160 µ m, re-spectively, while 119 DOGs are detected in both PACSbands. This amounts to a total of 314 PACS-detected DOGsources (cf Table 1) with M R > µm > µm > < z < c (cid:13)000
90% with F µm > µ Jy, we extend the DOGs 24 µ m-flux cut down to 80 µ Jy. The R-band magnitudes used inthis work are from Ilbert et al. (2009) based on observationswith the Subaru telescope by Capak et al. (2007); these in-clude a correction for Galactic extinction – not applied inCapak et al. (2007) – and reach a limiting magnitude ofM R > .The sample contains 57 and 138 DOGs detected inonly one of the PACS bands at 100 µ m and 160 µ m, re-spectively, while 119 DOGs are detected in both PACSbands. This amounts to a total of 314 PACS-detected DOGsources (cf Table 1) with M R > µm > µm > < z < c (cid:13)000 , 1–17 Riguccini et al.
Figure 2.
Distribution of 24 µ m flux for the DOG parent sam-ple ( ∼ Herschel -DOGs from this work (red).The
Herschel -DOGs distribution peaks at higher 24 µ m fluxes( ∼ ∼ ∼ µ m bright sources. SeeSect. 2.4 for details. Figure 3.
Distribution of 24 µ m flux of DOGs detected in atleast one PACS-band (314 sources) in blue and the same Herschel -DOGs distribution than in Fig. 2 in red.
Herschel -DOGs sample
Considering that we seek to undertake SED-fitting acrossthe mid- to far-IR wavelength range, we further define asub-sample of PACS-DOGs detected in all 5 Herschel bands; this allows for a better constraint on the peak of the DOGs’SEDs. To achieve this we matched the PACS-DOGs sample We note that 24 µ m-selected sources with no R-band detectionare also considered DOG sources. with the SPIRE catalog (Roseboom et al. 2010). This re-sults in 95 Herschel-detected DOGs , i.e. detected in the 5 Herschel bands (see Table 1).The fluxes in the PACS bands are obtained for allsources with a > σ detection. Although the SPIRE catalogreaches a 3- σ limit of ∼
10 mJy, ∼
12 mJy and ∼
15 mJy at250, 350 and 500 µ m respectively, the corresponding 3- σ ex-tragalactic confusion limits are 14.4 mJy, 16.5 mJy and 18.3mJy (Nguyen et al. 2010). In the SED-fitting procedure weare cautious (Magnelli et al. 2012a) when including fluxesthat are lower than the 3- σ extragalactic confusion limitsand use them merely as upper limits.Among our sample of 95 Herschel-detected DOGs, 40have their fluxes above the 3- σ threshold only for the 250 and350 µ m bands, 20 merely for the 250 µ m band, one DOG forthe 350 and 500 µ m bands and another DOG solely for the350 µ m; 9 of the Herschel-detected DOGs have all SPIREfluxes below the 3- σ threshold. We quote upper limits forall of these cases. Only 24 sources have fluxes above the 3- σ limit in the three SPIRE bands.We acknowledge that imposing a detection in the 5 Her-schel bands will impart a bias towards the brightest and red-dest IR sources in our sample. This is shown in Fig. 2, wherethe distribution of
Herschel -DOGs peaks at higher 24 µ mfluxes than that of the DOGs parent sample from Rigucciniet al. (2011). Of particular interest is to note that althoughthe faintest DOGs (F µm < Herschel -selection, beyond F µm > µ m-bright sources: not only they have a 24 µ m flux distributionthat peaks slightly higher than the whole DOGs populationdistribution, but their selection represents 60% of the DOGpopulation with F µm > µm =0.3 mJy Bussmann et al. (2009).In this paper, two samples of DOGs are used. To studythe IR colors of DOGs (Section 3) we use only PACS dataand thus base our analysis on the 314 PACS-DOGs, in an ef-fort to improve our statistics. For the remainder of our studywe restrict our analysis to the 95 Herschel-detected DOGs,noting that both samples probe the same DOG population,as illustrated by their 24 µ m flux distributions on Fig. 3. Wefocus our study on the differences observed between mid-IRbright DOGs (F µm > ∼ < F µm < The high accuracy of the photometric redshifts for sourcesof the COSMOS catalogue (see Sect. 2.1) make them highlyreliable for statistical studies on large (i.e., > All 95 Herschel-detected have R-band detectionsc (cid:13) , 1–17 omposite nature of Dust-Obscured Galaxies (DOGs) at z ∼ Table 1.
Number of sources6 029 24 µ m-sources with a 3- σ PACS detection (at 100 µ mand 160 µ m)5 892 sources from the previous sample with a photo-z314 DOGs (i.e., F µm /F R > F µm > µ Jyand 1.5 < z < σ detection in one orthe two PACS-bandsSample used for the far-IR/mid-IR color analysis[Sect. 3]95 DOGs with F µm > µ Jy and 1.5 < z < σ detection in the 2 PACS bands and with a detec-tion (potentially > σ ) in the 3 SPIRE bandsSample used for the remainder of the paper each source and divided our sample into three categories,according to the photometric-redshift reliability. The cat-egories are the following: (1) sources have a single, securephoto-z, for which the PDF has a gaussian shape with asingle peak (39 sources); (2) those with multiple potentialphoto-zs, for which either the PDF’s peak is spread over awider range of redshifts (∆ z ∼ The DOG sources are not only an extreme sub-sample ofULIRGs but also represent a mix between sources domi-nated by star-formation and those dominated by AGN ac-tivity (e.g., Houck et al. 2005; Fiore et al. 2008, 2009; Buss-mann et al. 2009; Melbourne et al. 2012). In this paper weseek to quantify the AGN contribution of these sources andstudy the evolution of this contribution with respect to othergalaxy properties, including redshift, the 8 µ m rest-frameluminosity, total IR luminosity, dust temperature and dustmass.Studies in the past years have explored the PL- andbump-DOGs population (e.g., Pope et al. 2008; Melbourneet al. 2009). It has been well established that PL-DOGs havean AGN contribution to their near-IR emission and thattheir far-IR emission is most likely dominated by star for-mation(Calanog et al. 2013). In this paper we aim to gaugethe AGN contribution of these DOGs using Herschel far-IRdata. As an initial, crude assessment of the dominant pro-cess responsible for producing most of the IR output inDOGs (i.e., AGN vs. star formation), we first consider thefar- to mid-IR colors (hereafter FIR/MIR) of our sample(e.g., Mullaney et al. 2012). Fig. 4 shows the 100 µ m/24 µ mand 160 µ m/24 µ m color distributions for our sample ofPACS-detected DOGs as a function of redshift; we includeall 24 µ m -detected COSMOS sources for comparison. Wesee no noticeable trend for the DOGs sample at z (cid:62)
2; thecurves shown for the 100/24 and 160/24 median color evo-lution with redshift seem to follow the same evolution thanthat of the whole 24 µ m-detected sample. However, at lowerredshifts the DOGs display a steeper evolution, with bluer100/24 colors than the non-DOG 24 µ m-detected sources.We find that the FIR/MIR distribution of both the bulkof the 24 µ m comparison sample as well as the majority ofour DOGs are well represented by the star forming tem-plates from (Chary & Elbaz 2001, hereafter CE01) with IRluminosities L IR = 10 − . L (cid:12) . This is to be expectedgiven that IR-selected galaxies at z > ∼ χ < F µm > µ m colors than the general 24 µ m-detected population (i.e, with F µm > µ Jy). We discardthe possibility that a variation in the Photodissociation re-gions (PDR) component and/or variation in the intensityof the field is responsible for the bluer color of these brightDOGs, by comparing to the 100 µ m/24 µ m colors derivedfrom the templates of Magdis et al. (2012) (see Fig. 4). TheseSED templates are based on stacked ensembles at differentredshift intervals, considering the varying radiation field andPDR contribution to ULIRGs as a function of redshift; forour study we rely on their starburst-dominated templatesat the two relevant redshift intervals: 1 . < z < .
25 and2 . < z < .
0. We compare to the FIR/MIR colors ofAGN/galaxy composites – using the intrinsic AGN SED ofMullaney et al. (2011) and assuming different AGN contri-butions (25, 50, 100%) at 100 µ m and 160 µ m – and findsignificant similarities, suggesting that the brightest DOGshave a significant AGN contribution. As we consider higherredshifts, the median FIR/MIR color of these bright DOGspoint towards a lower fraction of the AGN contribution, con-sistent with the SFRs of galaxies of similar mass increasingwith redshift (e.g., Brinchmann et al. 2004; Daddi et al.2007; Pannella et al. 2009; Magdis et al. 2010). We checkthe validity of these trends in the following section by look-ing at the AGN contribution (based on SED-fitting) as afunction of the 24 µ m flux. c (cid:13)000
0. We compare to the FIR/MIR colors ofAGN/galaxy composites – using the intrinsic AGN SED ofMullaney et al. (2011) and assuming different AGN contri-butions (25, 50, 100%) at 100 µ m and 160 µ m – and findsignificant similarities, suggesting that the brightest DOGshave a significant AGN contribution. As we consider higherredshifts, the median FIR/MIR color of these bright DOGspoint towards a lower fraction of the AGN contribution, con-sistent with the SFRs of galaxies of similar mass increasingwith redshift (e.g., Brinchmann et al. 2004; Daddi et al.2007; Pannella et al. 2009; Magdis et al. 2010). We checkthe validity of these trends in the following section by look-ing at the AGN contribution (based on SED-fitting) as afunction of the 24 µ m flux. c (cid:13)000 , 1–17 Riguccini et al.
Figure 4.
Left:
Distribution of 100 µ m/24 µ m color as a function of redshift for all the 24 µ m-selected sources detected at 100 µ m (blackopen circles), the 176 DOGs detected at 100 µ m with F µm > µ mJy (grey open squares) and with F µm > µ m-selected sources with F µm > µ Jy, the median of the 24 µ m-sources selected as DOGs with F µm > µ mJy and the median of the brightest DOGs sources (i.e.,with F µm > µ m band result from an AGNcomponent (Mullaney et al. 2011, 2012). We also include for comparison the star-forming ULIRG CE01 templates for an IR luminosityof 10 L (cid:12) (bottom blue-dashed line) and 10 . L (cid:12) (top blue-dashed line), as well as the template from Magdis et al. (2012) (lightblue dotted-dashed line). The observed PACS/24 colors of the bulk of the 24 µ m sources and that of the DOG sources are consistentwith these ULIRG templates. Right:
Distribution of 160 µm/ µm color as a function of the redshift for all the 24 µ m-selected sourcesdetected at 160 µ m (black open circles) and the 257 DOGs detected at 160 µ m. The colors and tracks are the same as on the left panel.For clarity, we do not overplot the CE01 templates on the right panel as they would give the same results than for the 100/24 color. µ M LUMINOSITY INDOGS
Based on FIR/MIR colors, bright DOGs likely contain anAGN component, contributing partly or even dominatingtheir IR luminosity. To have a better understanding of thesesources and to have a global view of their stage in the evolu-tion of the galaxies, it becomes important to know the exactcontribution of a potential AGN to their total 8-1000 µm restframe luminosity. In this section we present our method todetermine the potential contribution of an AGN componentto these DOG sources and show our results on the variationin AGN contribution with the 24 µ m flux. Studying the SED of a galaxy provides insights to the physi-cal nature of the underlying continuum source and can unveilthe presence of an AGN. The impact that an AGN contribu-tion has on the shape of the SED is distinct from that of dustheated by star-forming activity. However, deriving the SEDof a galaxy is not an easy exercise especially in the case of anAGN where the imprints of the host galaxy is always present.In our study we use the IDL-based SED-fitting procedure
DecompIR , detailed in Mullaney et al. (2011). Combininga set of five starburst templates and an average AGN tem-plate, this approach is aimed at fitting the IR photometry ofcomposite galaxies and to measure the AGN contribution totheir total IR output. A χ method is used to know which combination of these templates best fits the data; i.e., thecombination with the lowest associated χ value is adoptedas the best fit. The validity of this procedure as an accurateway to determine the AGN contribution to the total IR out-put of composite galaxies has been verified by several testslead by Mullaney et al. (2011), including a comparison withalternative measures of the AGN contribution (e.g., emissionline diagnostics). Although there are significant uncertain-ties associated to the precise AGN contribution to an indi-vidual galaxy, this approach is adequate from a statisticalpoint of view (i.e., large samples, average SEDs).We apply the DecompIR procedure to our sampleof DOGs sources. Nonetheless, considering that DOGshave ULIRG-class luminosities (e.g., Bussmann et al. 2009;Riguccini et al. 2011), we add two ULIRGs templates with L IR = 10 L (cid:12) and L IR = 10 . L (cid:12) to the set of starbursttemplates from Mullaney et al. (2011) to fully cover the lu-minosity range of our sample. The ULIRG templates aretaken from CE01 as they build their library from SEDs andmodels that only take into account star formation activity.The median PACS/MIPS colors of the bulk of the DOGssample are well represented by these ULIRGs templates (seeFig. 4), motivating their use as part of our SED-fitting pro-cedure. To determine the AGN contribution to the IR lu-minosity of our sources, we first derive a best SED-fit withtemplates based only on star-forming sources. If the star-forming templates do not provide a satisfactory result, anAGN component is added and the SED-fitting proceeds witha composite spectra. We consider the AGN component as a c (cid:13) , 1–17 omposite nature of Dust-Obscured Galaxies (DOGs) at z ∼ reasonable option only if it improves the χ of the fit by atleast 50%.We implement our SED-fitting procedure to each DOGsource in our sample. Each of the 95 sources are first fitby a star-forming component only and then by a compositespectra when the χ from the star-forming fit is >
20. Inthe case of sources with a less-accurate photo-z, the sourceis fit with a star-forming template for all possible photo-zs obtained from the PDF (see Sect. 2.5). If none of thesefits are suitable, we add an AGN component and the fits foreach possible photo-z are performed once again. Our methodis robust in fitting most of our sources (90%). Out of 95DOGs, 71 are fit with a star-forming galaxy template only,15 require an AGN component to the fit, and in 9 cases,no reliable fit was obtained either using a star-forming-onlytemplate nor a composite spectra. The failure of successfullyfitting these sources could be due to the fact that we cannotreproduce the SED of these sources; this is most likely due toa wrong redshift, even after probing the different possibilitiesindicated by the PDF.Due to the expected uncertainties on the AGN frac-tion for an individual galaxy, we implement the SED-fittingprocedure to average DOG SEDs. We divide our DOG sam-ple in 3 different redshift bins (1.5 < z < < z < < z < µm -flux bins (0.09 < F µm < < F µm < < F µm < < F µm < µm -flux bin, an averaged SED is calculated at 8 µm , 24 µm ,100 µm , 160 µm , 250 µm , 350 and 500 µm , leading to a to-tal of 12 average SEDs. The results of the fits are shown inFig 5 for the average SEDs and in Fig 6 & 7 for the individualsources fitted with an AGN.The results from the SED-fitting procedure on the av-erage DOG SEDs are presented on Table 2, including theAGN contribution to the total IR flux (f AGN ), the best-fit template and corresponding χ . Our method to get theAGN contribution does not appear to be biased toward onespecific template. Independently of the AGN contribution,the two templates that were the most successful at fittingthe average SEDs were the CE01 templates for IR lumi-nosities of 10 L (cid:12) (CE12) and 10 . L (cid:12) (CE12.5). This isas expected, since these templates have PACS/MIR colorsconsistent with that our DOG sources. We list the mid-to-far IR photometry and the total IR lumi-nosity of the 95 Herschel-DOGs on Table 3, indicating alsowhether an AGN component is included as part of the SEDfit. The redshift distribution of our DOG sample peaks atz ∼
2, allowing us to use the 24 µ m band as a probe of themid-IR emission close to the rest frame 8 µ m. This allowsus to derive a L µm that minimizes the dependence on thechoice of SED template used to perform the k -corrections.To determine the rest-frame L µm we interpolate the CE01library at the redshift and flux of each average SED. We arethen in a position to study the fractional AGN contributionto the total 8-1000 µ m output as a function of the 8 µ m rest-frame luminosity (L µm ); we do this for the average SEDs atthe three redshift bins: z = 1 . − .
0, 2 . − . . − . µm for all our redshift bins (see Fig. 8).This confirms the findings of Pope et al. (2008), where theyreport – based on mid-IR colors of 79 sources within theGOODS field and with 24 µ m fluxes down to 100 µ Jy –that low-luminosity DOGs are primarily powered by star-formation activity. However, only MIPS-70 µ m observationswere available for their analysis and the inability to sam-ple properly the peak of the SED lead to large uncertaintieson the derivation of the dust temperature and the AGNcontribution; Penner et al. (2012) extend the study out tofar-IR wavebands but miss the faintest DOGs by focussingon GOODS DOGs with luminosities 10 L (cid:12) < L IR < L (cid:12) . Furthermore, Fiore et al. (2008) claimed that even faintDOGs show evidence of hard X-ray emission, suggesting thepresence of an underlying AGN contribution. Within thiscontext our work – based on the large 2 deg area of theCOSMOS field and the good sensitivity of the MIPS-24 µ mobservations that allows us to sample a wider range of 24 µ mfluxes (i.e, 80 µ Jy < F µm < Herschel – allows us to conclude that faint DOGsare mainly star-forming systems while brighter sources be-come dominated by an AGN.By separating our sample in redshift bins, Fig. 8 alsoshows that the relation between AGN fraction and L µm evolves with redshift: the slope of the AGN contributionwith respect to the L µm is steeper at low redshifts, whileat higher redshifts the AGN contribution is less important.This is consistent with the results from Merloni & Heinz(2008), where they find that although the accretion ratedensity onto supermassive black holes (SMBH) and star-formation rate (SFR) densities increase from z ∼ ∼
2, the decrease in SMBH activity issharper than that of the SFR. We note that the uncertaintieson the derived AGN contributions are calculated from theformal error output resulting from the χ in the SED-fittingprocedure.In the interest of studying the star formation activityin our sources, we use the results from our SED decom-position to extract the AGN contribution and calculate IRluminosities only due to star formation. The resulting valuesspan the range of 10 L (cid:12) < L IR < L (cid:12) , correspond-ing to one order of magnitude fainter than the analysis byPenner et al. (2012). We study these IR luminosities as afunction of the rest-frame 8 µ m luminosity and find that fora given 8 µ m luminosity DOG sources whose SEDs are bestfit with the addition of an AGN component exhibit signifi-cantly lower IR luminosities than DOGs fit with a host-onlycomponent (see Fig. 9, top panel). In fact, for a given 8 µ mluminosity the majority ( ∼ µ m luminosity ratios (IR8 = L IR /L µm ;see Fig. 9, bottom panel). We observe an anti-correlationbetween the IR8 ratio and the L µm with the AGN-DOGspopulating the brightest L µm -end.For each DOG source in our sample we convert theIR luminosity into SFR according to Kennicutt (1998) andadopt the stellar mass from Ilbert et al. (2009). We considerthe redshift evolution of the specific SFR (sSFR = SFR/M ∗ )of our DOG sample and find that the majority of DOGswith no AGN component display sSFRs that place them at c (cid:13)000
2, the decrease in SMBH activity issharper than that of the SFR. We note that the uncertaintieson the derived AGN contributions are calculated from theformal error output resulting from the χ in the SED-fittingprocedure.In the interest of studying the star formation activityin our sources, we use the results from our SED decom-position to extract the AGN contribution and calculate IRluminosities only due to star formation. The resulting valuesspan the range of 10 L (cid:12) < L IR < L (cid:12) , correspond-ing to one order of magnitude fainter than the analysis byPenner et al. (2012). We study these IR luminosities as afunction of the rest-frame 8 µ m luminosity and find that fora given 8 µ m luminosity DOG sources whose SEDs are bestfit with the addition of an AGN component exhibit signifi-cantly lower IR luminosities than DOGs fit with a host-onlycomponent (see Fig. 9, top panel). In fact, for a given 8 µ mluminosity the majority ( ∼ µ m luminosity ratios (IR8 = L IR /L µm ;see Fig. 9, bottom panel). We observe an anti-correlationbetween the IR8 ratio and the L µm with the AGN-DOGspopulating the brightest L µm -end.For each DOG source in our sample we convert theIR luminosity into SFR according to Kennicutt (1998) andadopt the stellar mass from Ilbert et al. (2009). We considerthe redshift evolution of the specific SFR (sSFR = SFR/M ∗ )of our DOG sample and find that the majority of DOGswith no AGN component display sSFRs that place them at c (cid:13)000 , 1–17 Riguccini et al. . < F < . m Jy . < F < . m Jy . < F < . m Jy . < F < . m Jy Figure 5.
SEDs of the 12 average templates of DOGs galaxies from our sample. The solid line corresponds to the total SED fit, thedashed line is the host template and the dotted line is the AGN component. The name of the template used is written on each panels.The flux bins are specified on the left side and the redshift bins are written on the top. c (cid:13) , 1–17 omposite nature of Dust-Obscured Galaxies (DOGs) at z ∼ Table 2.
Results from the SED-fitting for the average SEDs per redshift bins and per flux bins. f AGN is the percentage of total 8-1000 µ m flux that comes from AGN. We specify the star-forming template best fitting the host galaxy (SB for the starburst templatesfrom (Mullaney et al. 2011) and CE 12.5 and CE 12 for the CE01 ULIRG templates.le Redshift Flux f AGN error Template χ L ir (mJy) % % (Host galaxy) L (cid:12) < z < < f < < f < < f < < f < < z < < f < < f < < f < < f < < z < < f < < f < < f < < f < Figure 8.
Evolution of the contribution of the AGN componentto the total rest-frame 8-1000 µ m flux of the sources as a functionof the 8 µ m rest-frame luminosity (L µm ). The AGN fraction isgiven for 3 different redshift bins: 1.5 < z < < z < < z < µm of the source irrespective of theredshift range. or above the main sequence (MS) from Elbaz et al. (2011),while 50% of the AGN-DOGs show significantly lower sSFRvalues (i.e., they lie below the MS)(see Fig. 10). Sources thatlie a factor of 2 above the MS are considered as “starbursts”by Elbaz et al. (2011). All but three host-component galaxieslie within a factor of 2 around the MS or in the starburst’szone. The distribution in sSFRs shown in Fig. 10 highlightsthe composite nature of the DOG population: some DOGsare dominated by starburst activity, the majority is undergo-ing star-formation as part of the MS, while others are domi- nated by an AGN. This prompts the idea that DOGs are atthe crossroads of the ULIRG-quasar scenario proposed by(Sanders et al. 1988a,b; Bussmann et al. 2012), with AGN-DOGs being closer to a quasar phase, where the AGN hasalready started to quench the star formation (explaining thelower sSFR observed on Fig. 10). Our SED-fitting analysis identifies 15 Herschel-detectedDOGs with an important AGN contribution to the totalIR output. We compare our AGN classification of DOGs toprior approaches relying on an IRAC-color selection. Fig. 11shows the IRAC-color selection of AGNs by Lacy et al.(2004), as well as the refined IRAC-color selection of Donleyet al. (2012), which also includes a power-law criteria in themid-IR: S . < S . and S . < S . and S . < S . .The majority of our sources display IRAC colors con-sistent with the criteria of Lacy et al. (2004), which wouldsuggest that 90% of our DOGs are AGNs. However, ourSED-fitting analysis indicates that only ∼
15% of our sourceshave a large AGN contribution. Based on this we concludethat relying on the AGN criteria of Lacy et al. (2004) wouldlead to a lack of precision in selecting AGNs versus galaxiesdominated by star formation. On the other hand, more than50% of our AGN-DOGs lie within the Donley et al. (2012)criterion, suggesting that it is a more reliable way of select-ing AGNs in DOGs when considering merely IRAC colors.However, from the 19 DOGs that lie within the AGN-criteriaof Donley et al. (2012) – and excluding the 3 that do notfollow the power-law criteria required by the authors – only9 are classified as AGNs following our SED-fitting analysis.That is, 40% of the Herschel-DOGs with IRAC colors con-sistent with the criterion of Donley et al. (2012) do not havea significant AGN contribution according to our analysis.Of particular interest is that out of all our AGN-DOGs, six(i.e., ∼ c (cid:13)000
15% of our sourceshave a large AGN contribution. Based on this we concludethat relying on the AGN criteria of Lacy et al. (2004) wouldlead to a lack of precision in selecting AGNs versus galaxiesdominated by star formation. On the other hand, more than50% of our AGN-DOGs lie within the Donley et al. (2012)criterion, suggesting that it is a more reliable way of select-ing AGNs in DOGs when considering merely IRAC colors.However, from the 19 DOGs that lie within the AGN-criteriaof Donley et al. (2012) – and excluding the 3 that do notfollow the power-law criteria required by the authors – only9 are classified as AGNs following our SED-fitting analysis.That is, 40% of the Herschel-DOGs with IRAC colors con-sistent with the criterion of Donley et al. (2012) do not havea significant AGN contribution according to our analysis.Of particular interest is that out of all our AGN-DOGs, six(i.e., ∼ c (cid:13)000 , 1–17 Riguccini et al.
With AGN
Figure 6.
Following the same format as Fig. 5, this figure shows the results of the SED-fitting procedure for DOG80, one of the 15DOGs which require an AGN component: the first 7 panels are the results of the SED-fitting with a host component only and the lastpanel (bottom middle panel) is the acceptable fit, with a contribution of an AGN component (20% < f AGN < ria by Donley et al. (2012), four of which do not follow thepower-law criterion required.On the one hand, our AGN classification is based onthe availability of far-IR data for obscured sources such asDOGs. On the other, Lacy et al. (2004) and Donley et al.(2012) classify sources as AGN-dominated based on IRAC-color selections. When considering these selections side byside, we draw two main conclusions: (1) non PL-DOGs po-tentially host an AGN that may dominate the far-IR regimeeven when missed by the IRAC-color selection criteria ofLacy et al. (2004) and Donley et al. (2012); and (2) PL-DOGs with an AGN according to our SED-fitting procedurecan be missed by IRAC colors criteria. We conclude thatour method provides an alternate means of determining thecomposite nature of DOGs. It has been well established that interstellar dust absorbs alarge fraction of the UV/optical radiation from DOGs andreemits it in the IR (Penner et al. 2012) As such, it is essen-tial that we understand the dust properties of these galaxiesif we are to understand this potentially-important popula-tion of galaxies. In this section, we derive the dust temper-atures and masses for our sample of DOGs. The availabilityof far-IR data from
Herschel is crucial to obtain these prop-erties. We are now able to extend previous studies on DOGsthat did not have access to such high-quality far-IR data(e.g., Dey et al. 2008; Bussmann et al. 2009, 2012). We arealso in a position to compare results with other recent stud-ies using (limited)
Herschel data on DOGs, including thatof SPIRE-detected sources (down to only F µm > c (cid:13) , 1–17 omposite nature of Dust-Obscured Galaxies (DOGs) at z ∼ DOG7 DOG43 DOG44 DOG58 DOG71 DOG73 DOG74 DOG80 DOG88 DOG91 DOG4 DOG16
DOG83 DOG86 DOG90
Figure 7.
Similar to Fig. 5, SEDs of the 15 DOGs which require the contribution of an AGN component.c (cid:13)000
Similar to Fig. 5, SEDs of the 15 DOGs which require the contribution of an AGN component.c (cid:13)000 , 1–17 Riguccini et al.
Figure 9.
Upper panel:
Comparison of L IR with L (rest-frame8 µ m) for DOGs from our sample (galaxies with host componentonly are marked with black open circles and AGNs are markedwith red filled circles). For comparison, we show the median loca-tion of star-forming galaxies from Elbaz et al. (2011) (solid line),with the dashed lines showing the 68% dispersion. Lower panel:
Variation of the IR8 (=L IR /L ) ratio with the 8 µ m luminosityfor our DOG sample, following the same color code as in theupper panel. For comparison we also plot galaxies from Elbazet al. (2011), including local galaxies (blue crosses), star-forminggalaxies at z > > The availability of
Herschel far-IR data allows us to con-strain the peak of the SED in the far-IR regime and to cal-culate the dust masses and temperatures of our galaxies witha higher accuracy than previous studies. DecompIR does notprovide information on the dust amount of our sources; wefit a blackbody spectrum B ν of temperature T using far-IRdata (see Amblard et al. 2014). To summarize, we performa single temperature fit (hereafter 1T model) with an emis-sivity, β , of 1.5 to fluxes long ward of λ rest − frame > µm .The luminosity is then expressed as L ( ν ) ∝ B ( ν, T d ) ν β . (1)Considering λ rest − frame > µm , we avoid emission fromthe AGN that can boost the dust temperature and bias ourresults (Netzer et al. 2007; Mullaney et al. 2011). For thisreason we restrict ourselves to using only the PACS 100, 160 µm and SPIRE 250, 350, 500 µm bands for galaxies with Riguccini+15a
Starburst Main Sequence
Figure 10.
Redshift evolution of the specific SFR (sSFR =SFR/M ∗ ) of DOGs; we distinguish between DOG sources whoseSEDs are best fit with a host-only component (black open circles)and with the addition of an AGN component (red filled circles).The solid line represents the star-forming main sequence from El-baz et al. (2011) and the dashed lines are a factor 2 above andbelow this fit. See text for details. Figure 11.
IRAC colors of host-component DOGs (black opencircles) and of the AGN-DOGs (red filled circles) from this work; we also indicate the IRAC colors of the 9 sources for which noSED fit was possible (blue crosses; see Sect. 4.3 for details). Thesolid line box is the AGN selection criterion from Donley et al.(2012) and the wider dashed box is from Lacy et al. (2004). z < µm bands (whenavailable) for z > < < > σ limit, 22 have 3 data points.We enforce the dust temperature to be constrainedbetween 10 and 95 K, the luminosity between 10 and10 L (cid:12) , acknowledging the high luminosities of our sample c (cid:13) , 1–17 omposite nature of Dust-Obscured Galaxies (DOGs) at z ∼ Figure 12.
Distribution of the dust temperature of the 24 DOGswith a detection in the 3 SPIRE bands with the 1T model (22sources at z > . d ∼
40 K and is comprised within the range24 < T dust <
65 K. (see Fig. 9). We observed the same definition as Amblardet al. (2014) for the dust luminosity and for the dust mass : L d ( λ ) = 4 πM d κ ( λ ) B ( λ, T d ) (2) M d = L d / (cid:90) πκ ( λ ) B ( λ, T ) dλ (3)where κ is taken at 850 µ m (Dunne et al. 2000) andequal to 0.077 kg − /m (Draine & Lee 1984; Hughes et al.1993).We find a median dust temperature ofT d ∼ (40.6 ± d > ∼ µ m: only 4 of their 12 DOGs have 350 µ m fluxes, therest of the sample has only upper limits). Melbourne et al.(2012) find lower dust temperatures for their Herschel -detected DOG sources (i.e., 20 < T d <
40 K). They splittheir sample into bump DOGs and PL-DOGs and findthat the PL-DOGs are less likely to be detected at far-IRwavelengths using SPIRE than the bump DOGs. They alsoclaim that SPIRE detections are biased towards very coldsources. We note that our range of temperature ( ∼ d = (37 ±
6) K fordetected PL-DOGs and T d = (35 ±
7) K for detected bumpsources. As raised in the literature (e.g., Melbourne et al.2012), using only SPIRE data tends to underestimate thedust temperature. Therefore, we need to be cautious in ouranalysis since more than half of our dust temperatures areobtained using only SPIRE data.
Figure 13.
Evolution of the dust temperature obtained with the1T model as a function of the AGN fraction. The results are ob-tained for the 12 average SEDs presented in Table 2 and detailedin Sec. 4.1. To improve clarity, we have slightly changed the xvalues for the 6 points with no AGN fraction in order to exhibitmore clearly the error bars on the figure. The blue line is the best χ fit with a slope of 0.33 ± σ error. With the goal of improving our accuracy in deriving dusttemperature and mass, we use the average SEDs obtainedper bin of 24 µ m fluxes and per bin of redshift (see Sec. 4.1and in Table 2 for details) instead of the SEDs of individ-ual DOG sources. The dust temperatures have been derivedfor the average SEDs following the procedure described inSec. 5.1. These are presented in Fig. 13 as a function of AGNfractional contribution. Roughly half of the average SEDshave no AGN contribution and show a wide range of T d as seen on the left side of Fig. 13. The two sources with noAGN fraction and with the highest dust temperatures showextremely large errors bars on the dust temperature (i.e. ±
15 K and ±
16 K) while the average dust temperature er-ror for the sample is around 9 K. The rest of the averageSEDs have an AGN contribution ranging from 20% to al-most 90% and are within the same T d range as the sourceswith no AGN component. The DOGs’ average SEDs withan AGN contribution display a correlation between the AGNcontribution and the dust temperature. We perform a best- χ fit on the 12 data points, taking into account the errorsboth on the x and y axes and find a slope of 0.33 ± χ of 0.94. To insure the validity of our fittingmethod, we also perform a fit T = cste with cste = Σ T /σ T Σ 1 /σ T = 42 . K. The reduced χ for this flat fit being 2.53, this gives us astrong indication that our previous fit is valid. We also es-timate the Spearman’s (rho) rank correlation of T dust and c (cid:13)000
16 K) while the average dust temperature er-ror for the sample is around 9 K. The rest of the averageSEDs have an AGN contribution ranging from 20% to al-most 90% and are within the same T d range as the sourceswith no AGN component. The DOGs’ average SEDs withan AGN contribution display a correlation between the AGNcontribution and the dust temperature. We perform a best- χ fit on the 12 data points, taking into account the errorsboth on the x and y axes and find a slope of 0.33 ± χ of 0.94. To insure the validity of our fittingmethod, we also perform a fit T = cste with cste = Σ T /σ T Σ 1 /σ T = 42 . K. The reduced χ for this flat fit being 2.53, this gives us astrong indication that our previous fit is valid. We also es-timate the Spearman’s (rho) rank correlation of T dust and c (cid:13)000 , 1–17 Riguccini et al. the AGN percentage. The significance is low (0.04), whichindicates a significant correlation. The correlation betweenthe dust temperature and the AGN percentage for the aver-age SEDs with AGN contribution is thus real, even thoughthe slope is small. However, we do not confirm the presenceof a general trend between the AGN fraction and the dusttemperature of the sources since half of the data points withno AGN activity present similar dust temperatures than theaverage SEDs with a large contribution from an AGN.
In addition to dust temperatures, the 1T model fitting pro-cedure also provides us with the dust masses of our sam-ple sources. We obtain a range for the entire sample of7 × < M dust < M (cid:12) and a median dust mass of ∼ (3 ± × M (cid:12) . Our results are in very good agreementwith Bussmann et al. (2009) who found a median dust massof 3 × M (cid:12) for their sample of 31 of the brightest DOGs(F µm > ∼ κ which could be as much as a factor of three,the median dust mass of our sample is not different fromthose estimated for high-redshift SMGs by Magnelli et al.(2012b) (M d ∼ M (cid:12) ). Pope et al. (2008) found that 30 %of the SMGs from their sample also satisfy the DOG criteria,and of those SMG-DOGs, 30 % are AGN dominated. DOGscould then be the descendants of these SMGs with similardust content, but representing a more advanced AGN-phasethan could later quench the star-formation and lead to el-liptical galaxies. We carry out a study that aims to understand the compos-ite nature of 24 µm -bright Dust-Obscured Galaxies (DOGs).These sources are a subset of ULIRGs at high redshift( z ∼
2) with F µm /F R >
982 . ULIRGs are consideredto represent an important phase in the evolution of galax-ies as they are linked to the formation of massive galax-ies via gas-rich starbursting mergers followed by an AGN-driven quenching of the star-formation (e.g., Sanders et al.1988a,b). Recent studies (Dey et al. 2008; Bussmann et al.2009, 2012) have suggested a similar evolutionary sequencewhere DOGs are an important intermediate phase betweengas-rich major mergers (traced by submillimeter galaxies,SMGs) and quasars at z ∼
2. These studies describe an evo-lutionary scenario in which the starbursting nature of SMGsevolves into the composite nature of DOGs as an underlyingAGN grows; this is followed by a quasar phase that termi-nates star formation, leading to the formation of a passive,massive elliptical galaxy. Within this context, DOGs couldprovide a key insight to an extremely dusty stage in theevolution of galaxies at z ∼
2, where both AGN and starformation activity coexist. Their composite nature was un- til relatively recently inaccessible prior to the availability ofsensitive mid- to far-infrared data.We base our work on a sample of 95
Herschel -detectedDOG sources. We perform SED-fitting on our sources usingcomposite spectra to obtain AGN contributions, dust tem-peratures and dust masses. We summarize below our resultsand our conclusions:(i) DOGs with the brightest 24 µm fluxes (F µm > µm colors thanother 24 µm -selected sources. These bluer colors may be ex-plained by templates containing an AGN contribution of atleast 25%.(ii) Among our sample of 95 sources, 74% are fit by a hostgalaxy template while for 16% require an additional AGNcomponent. The remaining 10% of the sample could not beproperly fit, likely due to inaccurate photometric redshifts.(iii) Faint DOG sources with L µm < L (cid:12) are domi-nated by star-formation at all redshifts, while DOGs brighterthan L µm > × L (cid:12) display a high contribution ( > IR /L8) ratio of our sampleand 50% of them lie below this sequence, with significantlylower specific star-formation rates. This results support theevolutionary scenario where DOGs may represent a transi-tion phase between high-redshift starburst-dominated SMGsand red-dead ellipticals, passing through an AGN-phase thatwould quench star formation.(v) The dust temperature of DOGs peaks at (40 ±
9) Kand our range of temperatures (24 < T d <
65 K) is overall ingood agreement with the literature (Bussmann et al. 2009,2012; Melbourne et al. 2012; Calanog et al. 2013). DOGswith a contribution from an AGN in the far-IR of at least60 % have dust temperatures >
50 K, suggesting that theAGN heats the dust of its host galaxy. We find a mediandust mass of ∼ (3 ± × M (cid:12) for our sample consistentprevious analysis in the literature (Bussmann et al. 2012).This work sheds light on DOG sources and their un-derlying composite nature, bringing unequivocally to lightthat mid-IR bright DOGs are powered by an AGN. Thesubmillimeters facilities in the near future, such as CCATand ALMA will provide critical insight to study the AGNproperties of these obscured ULIRGs at z ∼ Acknowledgments:
COSMOS is based on observations with theNASA/ESA Hubble Space Telescope, obtained at theSpace Telescope Science Institute, which is operated byAURA, Inc., under NASA contract NAS 5-26555; alsobased on data collected at: the Subaru Telescope, whichis operated by the National Astronomical Observatory c (cid:13) , 1–17 omposite nature of Dust-Obscured Galaxies (DOGs) at z ∼ of Japan; XMM-Newton, an ESA science mission withinstruments and contributions directly funded by ESAMember States and NASA; the European Southern Ob-servatory, Chile; Kitt Peak National Observatory, CerroTololo Inter-American Observatory, and the NationalOptical Astronomy Observatory, which are operated bythe Association of Universities for Research in Astronomy(AURA), Inc., under cooperative agreement with the Na-tional Science Foundation; the National Radio AstronomyObservatory, which is a facility of the National ScienceFoundation operated under cooperative agreement byAssociated Universities,Inc; and the Canada-France-HawaiiTelescope, operated by the National Research Council ofCanada, the Centre National de la Recherche Scientifiquede France, and the University of Hawaii.PACS has been developed by a consortium of institutesled by MPE (Germany) and including UVIE (Austria);KU Leuven, CSL, IMEC (Belgium); CEA, LAM (France);MPIA (Germany); INAF-IFSI/OAA/OAP/OAT, LENS,SISSA (Italy); IAC (Spain). This development has beensupported by the funding agencies BMVIT (Austria),ESA-PRODEX (Belgium), CEA/CNES (France), DLR(Germany), ASI/INAF (Italy), and CICYT/MCYT(Spain).SPIRE has been developed by a consortium of institutesled by Cardiff University (UK) and including Universityof Lethbridge (Canada), NAOC (China), CEA, LAM(France), IFSI, University of Padua (Italy), IAC (Spain),Stockholm Observatory (Sweden), Imperial College London,RAL, UCL-MSSL, UKATC, University of Sussex (UK),Caltech, JPL, NHSC, University of Colorado (USA). Thisdevelopment has been supported by national funding agen-cies: CSA (Canada); NAOC (China); CEA, CNES, CNRS(France); ASI (Italy); MCINN (Spain); SNSB (Sweden);STFC, UKSA (UK) and NASA (USA). SPIRE has beendeveloped by a consortium of institutes led by CardiffUniv. (UK) and including Univ. Lethbridge (Canada);NAOC (China); CEA, LAM (France); IFSI, Univ. Padua(Italy); IAC (Spain); Stockholm Observatory (Sweden);Imperial College London, RAL, UCL-MSSL, UKATC,Univ. Sussex (UK); Caltech, JPL, NHSC, Univ. Colorado(USA). This development has been supported by nationalfunding agencies: CSA (Canada); NAOC (China); CEA,CNES, CNRS (France); ASI (Italy); MCINN (Spain); SNSB(Sweden); STFC, UKSA (UK); and NASA (USA).This paper has been typeset from a TEX/ L A TEX file preparedby the author.
REFERENCES
Amblard A., Riguccini L., Temi P., Im S., Fanelli M., SerraP., 2014, ApJ, 783, 135Berta S. et al., 2011, A&A, 532, A49Brinchmann J., Charlot S., White S. D. M., Tremonti C.,Kauffmann G., Heckman T., Brinkmann J., 2004, MN-RAS, 351, 1151Brusa M. et al., 2010, ApJ, 716, 348Brusa M. et al., 2007, ApJs, 172, 353Bussmann R. S. et al., 2012, ApJ, 744, 150Bussmann R. S. et al., 2009, ApJ, 705, 184 Calanog J. A. et al., 2013, ApJ, 775, 61Capak P. et al., 2007, ApJs, 172, 99Chary R., Elbaz D., 2001, ApJ, 556, 562Daddi E. et al., 2007, ApJ, 670, 156Desai V. et al., 2009, ApJ, 700, 1190Dey A. et al., 2008, ApJ, 677, 943Donley J. L. et al., 2012, ApJ, 748, 142Donley J. L., Rieke G. H., P´erez-Gonz´alez P. G., RigbyJ. R., Alonso-Herrero A., 2007, ApJ, 660, 167Draine B. T., Lee H. M., 1984, ApJ, 285, 89Dunne L., Eales S., Edmunds M., Ivison R., Alexander P.,Clements D. L., 2000, MNRAS, 315, 115Elbaz D. et al., 2011, A&A, 533, A119Elbaz D. et al., 2010, A&A, 518, L29Elvis M. et al., 2009, ApJs, 184, 158Farrah D. et al., 2008, ApJ, 677, 957Fiore F. et al., 2008, ApJ, 672, 94Fiore F. et al., 2009, ApJ, 693, 447Griffin M. J. et al., 2010, A&A, 518, L3Hasinger G. et al., 2007, ApJs, 172, 29Houck J. R. et al., 2005, ApJl, 622, L105Hughes D. H., Robson E. I., Dunlop J. S., Gear W. K.,1993, MNRAS, 263, 607Ilbert O. et al., 2009, ApJ, 690, 1236Kennicutt R. C., 1998, ARA&A, 36, 189Lacy M. et al., 2004, ApJs, 154, 166Le Floc’h E. et al., 2009, ApJ, 703, 222Le Floc’h E. et al., 2005, ApJ, 632, 169Lilly S. J. et al., 2007, ApJs, 172, 70Lutz D. et al., 2011, A&A, 532, A90Magdis G. E. et al., 2012, ApJ, 760, 6Magdis G. E., Rigopoulou D., Huang J.-S., Fazio G. G.,2010, MNRAS, 401, 1521Magnelli B., Elbaz D., Chary R. R., Dickinson M., LeBorgne D., Frayer D. T., Willmer C. N. A., 2009, A&A,496, 57Magnelli B. et al., 2012a, A&A, 539, A155Magnelli B. et al., 2012b, A&A, 548, A22McCracken H. J. et al., 2010, ApJ, 708, 202Melbourne J. et al., 2009, AJ, 137, 4854Melbourne J. et al., 2012, AJ, 143, 125Merloni A., Heinz S., 2008, MNRAS, 388, 1011Mullaney J. R., Alexander D. M., Goulding A. D., HickoxR. C., 2011, MNRAS, 414, 1082Mullaney J. R. et al., 2012, MNRAS, 419, 95Netzer H. et al., 2007, ApJ, 666, 806Nguyen H. T. et al., 2010, A&A, 518, L5Nordon R. et al., 2012, ApJ, 745, 182Nordon R. et al., 2010, A&A, 518, L24Oliver S. J. et al., 2012, MNRAS, 424, 1614Pannella M. et al., 2009, ApJl, 698, L116Penner K. et al., 2012, ApJ, 759, 28Pilbratt G. L. et al., 2010, A&A, 518, L1Poglitsch A. et al., 2010, A&A, 518, L2Pope A. et al., 2008, ApJ, 689, 127Riguccini L. et al., 2011, A&A, 534, A81Roseboom I. G. et al., 2010, MNRAS, 409, 48Salvato M. et al., 2009, ApJ, 690, 1250Sanders D. B. et al., 2007, ApJs, 172, 86Sanders D. B., Soifer B. T., Elias J. H., Madore B. F.,Matthews K., Neugebauer G., Scoville N. Z., 1988a, ApJ,325, 74 c (cid:13)000
Amblard A., Riguccini L., Temi P., Im S., Fanelli M., SerraP., 2014, ApJ, 783, 135Berta S. et al., 2011, A&A, 532, A49Brinchmann J., Charlot S., White S. D. M., Tremonti C.,Kauffmann G., Heckman T., Brinkmann J., 2004, MN-RAS, 351, 1151Brusa M. et al., 2010, ApJ, 716, 348Brusa M. et al., 2007, ApJs, 172, 353Bussmann R. S. et al., 2012, ApJ, 744, 150Bussmann R. S. et al., 2009, ApJ, 705, 184 Calanog J. A. et al., 2013, ApJ, 775, 61Capak P. et al., 2007, ApJs, 172, 99Chary R., Elbaz D., 2001, ApJ, 556, 562Daddi E. et al., 2007, ApJ, 670, 156Desai V. et al., 2009, ApJ, 700, 1190Dey A. et al., 2008, ApJ, 677, 943Donley J. L. et al., 2012, ApJ, 748, 142Donley J. L., Rieke G. H., P´erez-Gonz´alez P. G., RigbyJ. R., Alonso-Herrero A., 2007, ApJ, 660, 167Draine B. T., Lee H. M., 1984, ApJ, 285, 89Dunne L., Eales S., Edmunds M., Ivison R., Alexander P.,Clements D. L., 2000, MNRAS, 315, 115Elbaz D. et al., 2011, A&A, 533, A119Elbaz D. et al., 2010, A&A, 518, L29Elvis M. et al., 2009, ApJs, 184, 158Farrah D. et al., 2008, ApJ, 677, 957Fiore F. et al., 2008, ApJ, 672, 94Fiore F. et al., 2009, ApJ, 693, 447Griffin M. J. et al., 2010, A&A, 518, L3Hasinger G. et al., 2007, ApJs, 172, 29Houck J. R. et al., 2005, ApJl, 622, L105Hughes D. H., Robson E. I., Dunlop J. S., Gear W. K.,1993, MNRAS, 263, 607Ilbert O. et al., 2009, ApJ, 690, 1236Kennicutt R. C., 1998, ARA&A, 36, 189Lacy M. et al., 2004, ApJs, 154, 166Le Floc’h E. et al., 2009, ApJ, 703, 222Le Floc’h E. et al., 2005, ApJ, 632, 169Lilly S. J. et al., 2007, ApJs, 172, 70Lutz D. et al., 2011, A&A, 532, A90Magdis G. E. et al., 2012, ApJ, 760, 6Magdis G. E., Rigopoulou D., Huang J.-S., Fazio G. G.,2010, MNRAS, 401, 1521Magnelli B., Elbaz D., Chary R. R., Dickinson M., LeBorgne D., Frayer D. T., Willmer C. N. A., 2009, A&A,496, 57Magnelli B. et al., 2012a, A&A, 539, A155Magnelli B. et al., 2012b, A&A, 548, A22McCracken H. J. et al., 2010, ApJ, 708, 202Melbourne J. et al., 2009, AJ, 137, 4854Melbourne J. et al., 2012, AJ, 143, 125Merloni A., Heinz S., 2008, MNRAS, 388, 1011Mullaney J. R., Alexander D. M., Goulding A. D., HickoxR. C., 2011, MNRAS, 414, 1082Mullaney J. R. et al., 2012, MNRAS, 419, 95Netzer H. et al., 2007, ApJ, 666, 806Nguyen H. T. et al., 2010, A&A, 518, L5Nordon R. et al., 2012, ApJ, 745, 182Nordon R. et al., 2010, A&A, 518, L24Oliver S. J. et al., 2012, MNRAS, 424, 1614Pannella M. et al., 2009, ApJl, 698, L116Penner K. et al., 2012, ApJ, 759, 28Pilbratt G. L. et al., 2010, A&A, 518, L1Poglitsch A. et al., 2010, A&A, 518, L2Pope A. et al., 2008, ApJ, 689, 127Riguccini L. et al., 2011, A&A, 534, A81Roseboom I. G. et al., 2010, MNRAS, 409, 48Salvato M. et al., 2009, ApJ, 690, 1250Sanders D. B. et al., 2007, ApJs, 172, 86Sanders D. B., Soifer B. T., Elias J. H., Madore B. F.,Matthews K., Neugebauer G., Scoville N. Z., 1988a, ApJ,325, 74 c (cid:13)000 , 1–17 Riguccini et al.
Table 3.
DOGs sampleDOG ID redshift F F F F F F flag AGN χ L ir (mJy) (mJy) (mJy) (mJy) (mJy) (mJy) (L (cid:12) )0 1.87 0.386 9.798 14.33 35.05 17.40 14.67 0 2.90 1.02e+121 1.60 0.178 11.48 10.53 23.54 16.14 10.17 0 9.59 8.41e+112 2.65 0.176 9.251 11.81 21.27 19.14 14.26 0 6.22 3.41e+123 2.34 0.491 16.73 26.67 27.36 20.72 7.077 0 3.00 2.48e+124 3.00 0.097 9.961 27.31 25.93 14.49 3.826 3 0.409 3.67e+125 1.00 0.404 9.088 20.96 12.49 8.223 6.096 0 37.8 1.30e+126 1.40 0.440 7.806 16.88 15.89 13.25 9.133 0 20.5 6.84e+117 2.85 0.326 8.145 13.69 22.02 18.26 5.261 2 2.19 1.20e+128 2.34 0.402 8.467 24.79 22.32 20.23 9.081 0 3.00 2.48e+129 1.89 0.312 15.11 10.92 10.72 10.60 8.063 0 7.78 1.49e+1210 1.14 0.103 5.576 16.27 15.17 17.43 5.545 0 3.56 4.15e+1111 1.98 0.156 14.28 18.67 25.49 17.19 2.676 -99 -99 -9912 1.88 0.236 11.65 29.26 19.52 7.441 3.868 0 24.5 1.587e+1213 1.81 0.444 14.62 43.96 31.01 17.75 7.619 0 0.0739 9.502e+1114 1.79 0.201 6.918 32.55 30.33 19.84 9.915 0 31.7 2.460e+1215 2.30 0.326 8.989 18.93 30.08 21.45 20.49 0 8.01 2.068e+1216 2.04 0.656 9.732 20.43 30.88 34.16 15.04 4 0.668 1.753e+1217 2.34 0.716 11.18 20.34 20.45 16.57 8.984 0 0.311 1.760e+1218 2.88 0.330 17.99 54.18 43.51 35.79 32.05 0 12.6 4.012e+1219 2.74 0.223 12.73 22.93 32.10 20.40 5.487 0 0.509 4.354e+1220 2.41 0.665 16.44 27.88 17.78 14.48 0.769 0 2.57 1.842e+1221 1.90 0.363 6.506 9.178 21.98 19.80 13.77 0 4.31 1.538e+1222 2.15 0.319 9.573 31.35 50.37 43.09 29.44 0 29.9 1.798e+1223 2.11 0.400 9.302 26.70 34.57 26.41 13.75 0 15.8 2.122e+1224 2.14 0.208 5.767 15.81 18.79 12.01 5.824 0 32.7 2.412e+1225 2.50 0.414 4.650 14.08 35.78 28.10 10.19 -99 -99 -9926 2.29 0.485 7.434 15.55 26.60 25.71 25.31 0 9.11 3.31e+1227 1.84 0.285 7.054 23.17 24.96 21.93 9.473 0 5.71 9.19e+1128 1.79 0.400 7.824 16.42 12.27 14.91 6.783 0 3.61 1.59e+1229 1.80 0.532 7.399 21.88 28.97 24.07 13.26 -99 -99 -9930 1.91 0.287 8.802 23.61 27.99 26.50 17.52 0 2.37 1.14e+1231 1.80 0.601 10.28 32.24 50.22 57.26 37.58 0 0.503 1.67e+1232 1.43 0.271 27.75 50.52 54.32 22.67 22.84 0 47.8 1.42e+1233 1.24 0.405 7.274 12.76 24.40 25.07 12.53 0 0.0981 6.33e+1134 1.73 0.090 3.589 13.36 18.04 8.671 1.258 0 1.04 1.04e+1235 1.83 0.669 7.372 18.38 29.18 21.97 15.27 0 3.06 1.73e+1236 1.88 0.388 12.93 17.25 11.61 5.563 3.736 0 13.3 1.63e+1237 2.18 0.185 11.57 16.36 20.27 11.07 7.944 0 4.96 2.85e+1238 1.68 0.846 7.959 10.58 43.39 42.47 40.47 0 2.16 9.40e+1139 1.56 0.280 9.949 15.29 16.03 18.52 12.97 0 10.0 1.25e+1240 1.71 0.412 17.15 40.69 38.81 23.32 9.754 0 4.66 1.02e+1241 2.00 0.737 10.41 21.83 31.49 28.50 17.59 0 9.79 1.17e+1242 1.79 0.282 5.335 14.58 22.06 15.15 3.131 0 8.24 1.37e+1243 2.75 0.277 7.221 21.95 14.15 16.22 13.70 3 15.7 8.33e+1244 2.03 1.038 29.37 43.28 34.96 28.66 7.363 2 13.4 4.74e+1145 1.91 0.248 5.540 14.59 26.78 19.12 7.714 0 5.65 1.43e+1246 1.81 0.551 16.28 28.14 17.47 10.42 6.205 0 4.42 1.72e+1247 2.12 0.179 5.090 8.722 25.19 24.05 19.96 -99 -99 -9948 2.35 0.386 9.840 30.80 33.42 44.45 28.70 0 11.8 4.28e+1249 1.96 0.130 6.263 12.26 20.61 23.57 12.48 0 1.40 1.48e+1250 2.92 0.296 9.159 25.80 40.72 28.25 21.91 0 8.62 3.72e+1251 2.55 0.559 9.291 20.90 20.95 4.990 16.31 0 1.26 1.82e+1252 1.94 0.208 6.285 11.61 19.85 18.14 7.037 0 2.37 1.72e+1253 2.26 0.379 9.616 28.90 38.44 34.29 20.98 0 8.56 1.51e+1254 1.88 0.704 4.882 29.71 30.02 36.79 7.011 0 0.672 1.41e+1255 1.61 0.187 7.526 18.13 19.77 23.50 15.58 0 25.1 2.08e+1256 1.61 0.284 7.408 19.38 21.56 17.52 9.390 0 2.67 1.00e+1257 1.73 0.537 5.251 16.61 29.93 25.76 9.697 0 24.1 1.73e+1258 2.77 0.165 7.150 14.23 13.25 8.367 8.701 2 20.9 2.33e+1259 1.94 0.330 14.65 24.56 26.93 21.13 10.50 0 0.369 1.94e+1260 1.27 0.433 10.52 27.31 44.89 44.77 34.52 0 9.25 7.57e+11 flag AGN is the contribution from an AGN to the host galaxy obtained from DecompIR: (0): only host galaxy, (1): % AGN (cid:54) < % AGN (cid:54) < % AGN (cid:54) >
70% c (cid:13) , 1–17 omposite nature of Dust-Obscured Galaxies (DOGs) at z ∼ Table 3 – continued DOG ID redshift F F F F F F flag AGN χ L ir (mJy) (mJy) (mJy) (mJy) (mJy) (mJy) (L (cid:12) )61 2.00 0.745 15.31 25.86 38.35 47.96 29.10 0 50.1 3.18e+1262 1.58 0.505 6.231 31.96 8.530 45.61 8.272 0 7.96 1.22e+1263 1.88 0.621 17.10 39.68 41.01 29.16 21.66 0 6.60 6.77e+1164 2.55 0.366 8.821 15.62 8.706 3.102 6.735 0 2.24 2.61e+1265 1.76 0.419 9.343 19.55 28.59 27.63 20.37 -99 -99 -9966 2.03 0.407 14.68 28.11 22.86 15.96 10.10 0 2.44 2.69e+1267 2.85 0.497 16.95 27.73 10.42 8.087 11.99 0 14.7 4.00e+1268 1.94 0.943 13.47 26.86 15.43 7.120 2.860 -99 -99 -9969 1.62 0.925 43.88 66.54 54.21 32.34 6.816 -99 -99 -9970 1.93 0.437 7.177 19.51 21.00 8.710 8.849 0 1.69 1.39e+1271 2.53 0.132 16.52 44.25 51.89 34.34 23.83 3 1.20 2.98e+1272 2.91 0.555 7.702 23.22 19.01 19.22 10.53 -99 -99 -9973 1.61 0.395 13.92 23.40 31.45 13.30 11.82 4 1.44 2.49e+1274 1.61 0.266 7.749 20.32 20.94 10.23 1.243 3 0.200 3.54e+1175 2.70 0.256 6.372 13.61 29.14 24.31 23.73 0 1.67 3.36e+1276 2.92 0.465 13.72 41.37 39.27 37.63 28.69 0 1.60 2.94e+1277 2.00 1.487 10.00 17.00 29.15 27.47 5.519 0 14.1 1.48e+1278 1.98 0.554 22.49 71.89 74.66 50.33 50.14 -99 -99 -9979 2.11 0.359 14.87 21.01 25.52 18.83 15.39 0 11.0 2.55e+1280 1.89 4.742 20.84 26.75 20.47 16.42 8.412 2 0.761 6.38e+1181 2.33 1.385 30.01 59.63 62.33 50.04 24.83 0 3.97 1.31e+1282 1.97 0.625 12.46 20.73 26.62 27.96 11.84 0 5.23 1.26e+1283 1.98 1.059 28.67 36.87 32.63 40.06 27.67 3 15.6 2.07e+1284 1.91 0.578 7.835 13.90 48.09 51.41 32.16 0 3.45 1.74e+1285 2.58 1.917 14.49 18.95 13.09 9.595 2.781 0 15.9 2.26e+1286 2.91 0.759 11.60 21.18 35.11 49.91 24.43 2 0.021 1.00e+1287 2.34 0.637 7.290 17.49 20.34 23.81 13.98 0 6.19 2.14e+1288 2.64 3.744 55.74 102.9 100.3 59.92 55.24 1 8.16 3.29e+1289 2.20 0.860 12.87 22.32 12.75 19.16 21.22 0 19.6 5.25e+1290 2.89 0.671 16.75 27.59 41.20 36.35 4.689 1 4.33 5.54e+1191 2.88 0.931 5.011 11.63 15.30 13.39 16.10 1 5.19 8.52e+1192 2.60 2.392 17.72 34.43 49.46 32.61 10.89 0 8.08 2.23e+1293 1.73 3.131 8.783 10.84 18.91 13.82 2.563 0 12.8 1.76e+1294 1.75 1.559 20.80 20.48 16.66 8.258 4.248 0 6.89 2.05e+12 Sanders D. B., Soifer B. T., Elias J. H., Neugebauer G.,Matthews K., 1988b, ApJl, 328, L35Schinnerer E. et al., 2007, ApJs, 172, 46Scoville N. et al., 2007, ApJs, 172, 38Smolˇci´c V. et al., 2014, MNRAS, 443, 2590Taniguchi Y. et al., 2007, ApJs, 172, 9Treister E. et al., 2009, ApJ, 706, 535Weedman D. W., Le Floc’h E., Higdon S. J. U., HigdonJ. L., Houck J. R., 2006, ApJ, 638, 613 c (cid:13)000