The SCUBA-2 Cosmology Legacy Survey: blank-field number counts of 450um-selected galaxies and their contribution to the cosmic infrared background
J. E. Geach, E. L. Chapin, K. E. K. Coppin, J. S. Dunlop, M. Halpern, Ian Smail, P. van der Werf, S. Serjeant, D. Farrah, I. Roseboom, T. Targett, V. Arumugam, V. Asboth, A. Blain, A. Chrysostomou, C. Clarke, R. J. Ivison, S. L. Jones, A. Karim, T. Mackenzie, R. Meijerink, M. J. Michalowski, D. Scott, J. Simpson, A. M. Swinbank, D. Alexander, O. Almaini, I. Aretxaga, P. Best, S. Chapman, D. L. Clements, C. Conselice, A. L. R. Danielson, S. Eales, A. C. Edge, A. Gibb, D. Hughes, T. Jenness, K. K. Knudsen, C. Lacey, G. Marsden, R. McMahon, S. Oliver, M. J. Page, J. A. Peacock, D. Rigopoulou, E. I. Robson, M. Spaans, J. Stevens, T. M. A. Webb, C. Willott, C. D. Wilson, M. Zemcov
aa r X i v : . [ a s t r o - ph . C O ] N ov Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 23 July 2018 (MN L A TEX style file v2.2)
The SCUBA–2 Cosmology Legacy Survey: blank-field numbercounts of 450 µ m-selected galaxies and their contribution to thecosmic infrared background J. E. Geach ⋆ , E. L. Chapin , , , K. E. K. Coppin , J. S. Dunlop , M. Halpern , IanSmail , P. van der Werf , S. Serjeant , D. Farrah , I. Roseboom , T. Targett , V.Arumugam , V. Asboth , A. Blain , A. Chrysostomou , , C. Clarke , R. J. Ivison , ,S. L. Jones , A. Karim , T. Mackenzie , R. Meijerink , , M. J. Michałowski , D. Scott ,J. Simpson , A. M. Swinbank , D. Alexander , O. Almaini , I. Aretxaga , P. Best , S.Chapman , D. L. Clements , C. Conselice , A. L. R. Danielson , S. Eales , A. C.Edge , A. Gibb , D. Hughes , T. Jenness , K. K. Knudsen , C. Lacey , G. Marsden ,R. McMahon , S. Oliver , M. J. Page , J. A. Peacock , D. Rigopoulou , , E. I.Robson , M. Spaans , J. Stevens , T. M. A. Webb , C. Willott , C. D. Wilson ,M. Zemcov Department of Physics, Ernest Rutherford Building, 3600 rue University, McGill University, Montr´eal, QC, H3A 2T8, Canada Department of Physics & Astronomy, University of British Columbia, 6224 Agricultural Road, Vancouver, BC, V6T 1Z1, Canada Joint Astronomy Centre 660 N. A’ohoku Place University Park Hilo, Hawaii 96720, USA XMM SOC, ESAC, Apartado 78, 28691 Villanueva de la Canada, Madrid, Spain Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ Institute for Computational Cosmology, Department of Physics, Durham University, South Road, Durham, DH1 3LE Leiden Observatory, Leiden University, P.O. box 9513, 2300 RA Leiden, The Netherlands Robert Hooke Building, Department of Physical Sciences, The Open University, Milton Keynes, MK7 6AA Virginia Polytechnic Institute & State University Department of Physics, MC 0435, 910 Drillfield Drive, Blacksburg, VA 24061, USA Department of Physics & Astronomy, University of Leicester, University Road, Leicester, LE1 7RH Centre for Astrophysics Research, Science & Technology Research Institute, University of Hertfordshire, Hatfield, AL10 9AB Astronomy Centre, Department of Physics and Astronomy, University of Sussex, Brighton BN1 9QH UK Astronomy Technology Centre, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ Kapteyn Institute, University of Groningen, P.O. Box 800, 9700 AV Groningen, The Netherlands School of Physics and Astronomy, University of Nottingham, University Park, Nottingham, NG9 2RD Instituto Nacional de Astrof´ısica ´Optica y Electr´onica, Calle Luis Enrique Erro No. 1, Sta. Ma. Tonantzintla, Puebla, M´exico Department of Physics and Atmospheric Science, Dalhousie University Halifax, NS, B3H 3J5, Canada Astrophysics Group, Imperial College London, Blackett Laboratory, Prince Consort Road, London, SW7 2AZ Cardiff School of Physics and Astronomy, Cardiff University, Queens Buildings, The Parade, Cardiff, CF24 3AA Department of Earth and Space Science, Chalmers University of Technology, Onsala Space Observatory, SE-43992 Onsala, Sweden Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge, CB3 OHA Mullard Space Science Laboratory, University College London, Holmbury St Mary Dorking, Surrey RH5 6NT Department of Physics, University of Oxford, Keble Road, Oxford, OX1 3RH Space Science & Technology Department, Rutherford Appleton Laboratory, Chilton, Didcot, Oxfordshire, OX11 0QX Canadian Astronomy Data Centre, National Research Council Canada, 5071 West Saanich Road, Victoria, BC, V9E 2E7, Canada Department of Physics and Astronomy, McMaster University Hamilton, ON, L8S 4M1, Canada Astronomy Department, California Institute of Technology, MC 367-17 1200 East California Blvd., Pasadena, CA 91125, USA
23 July 2018c (cid:13)
J. E. Geach et al.
ABSTRACT
The first deep blank-field 450 µ m map ( σ ≈ . mJy) from the SCUBA–2 CosmologyLegacy Survey (S2CLS), conducted with the James Clerk Maxwell Telescope (JCMT) ispresented. Our map covers 140 arcmin of the Cosmological Evolution Survey (COSMOS)field, in the footprint of the Hubble Space Telescope ( HST ) Cosmic Assembly Near-InfraredDeep Extragalactic Legacy Survey (CANDELS). Using 60 submillimetre galaxies (SMGs)detected at > σ , we evaluate the number counts of 450 µ m-selected galaxies with fluxdensities S > mJy. The 8 ′′ JCMT beam and high sensitivity of SCUBA–2 now make itpossible to directly resolve a larger fraction of the cosmic infrared background (CIB, peakingat λ ∼ µ m) into the individual galaxies responsible for its emission than has previouslybeen possible at this wavelength. At S > mJy we resolve (7 . ± . × − MJy sr − of the CIB at 450 µ m (equivalent to ± % of the absolute brightness measured by the Cosmic Background Explorer at this wavelength) into point sources. A further ∼
40% of theCIB can be recovered through a statistical stack of 24 µ m emitters in this field, indicating thatthe majority ( ≈ µ m is emitted by galaxies with S > mJy. Theaverage redshift of 450 µ m emitters identified with an optical/near-infrared counterpart is es-timated to be h z i = 1 . , implying that the galaxies in the sample are in the ultraluminousclass ( L IR ≈ . × L ⊙ ). If the galaxies contributing to the statistical stack lie at similarredshifts, then the majority of the CIB at 450 µ m is emitted by galaxies in the LIRG class with L IR > . × L ⊙ . Key words: galaxies: high-redshift, active, evolution, cosmology: observations, submillime-tre: galaxies
Fifteen years have passed since the first ‘submillimetre galaxies’(SMGs) were discovered (Smail, Ivison & Blain 1997; Barger et al.1998; Hughes et al. 1998), a high-redshift population ( z ∼ − ,Chapman et al. 2005; Aretxaga et al. 2007; Wardlow et al. 2011)with ultraluminous ( L ⊙ ) levels of bolometric emission, thebulk of which is emitted in the far-infrared (FIR) and redshifted tosubmillimetre wavelengths at z > . The power of submillimetresurveys for exploring the formation phase of massive galaxies wasrecognised before their discovery (e.g. Blain & Longair 1993; Dun-lop et al. 1994), and since their discovery, their importance as a cos-mologically significant population has been established by manystudies (e.g. Smail et al. 2002; Dunlop et al. 2004; Ivison et al.2000, 2005, 2010; Coppin et al. 2008; Michałowski et al. 2010;Hainline et al. 2011; Hickox et al. 2012; and see Dunlop et al. 2011for a review). As such, SMGs provide challenging tests for modelsof galaxy formation, both in detailed ‘zoomed’ simulations as wellas in cosmological theatres (Baugh et al. 2005; Dav´e et al. 2010).However, our view of the SMG population remains incomplete.In ground-based work, the majority of SMGs have – so-far –mainly been selected in the 850 µ m or 1 mm atmospheric windows(e.g. Coppin et al. 2006; Weiss et al. 2009; Austermann et al. 2010;Scott et al. 2010), but this is far removed from the peak of the Cos-mic infrared background (CIB), which is at λ ∼ µ m (Fixsenet al. 1998). The next available window closer to the CIB peak isat 450 µ m, but the transmission of this window is just at best 50%of the 850 µ m window, making 450 µ m SMG surveys challengingfrom ground-based sites. Submillimetre surveys working closer tothe CIB peak are essential if we are to identify the galaxies respon-sible for its emission; the S /S colours of sources identifiedin the very deepest (lensing assisted) submillimetre surveys (Blainet al. 1999; Knudsen et al. 2008) suggests that these sources con- ⋆ E-mail: [email protected] tribute less than half of the CIB at 450 µ m (and therefore even lessat the actual peak).The Balloon-borne Large Aperture submillimetre Telescope(BLAST, Pascale et al. 2008) made progress by conducting a lowresolution submillimetre survey from the stratosphere at 250, 350and 500 µ m (Pascale et al. 2008; Devlin et al. 2009; Glenn et 2010).This work was taken forward by the Herschel Space Observatory ,which carries an instrument that images in the same wavelengthranges as BLAST (the Spectral and Photometric Imaging Receiver;SPIRE), and has mapped hundreds of square degrees of the sky at250–500 µ m in a combination of panoramic and deep cosmologicalsurveys (Eales et al. 2010, Oliver et al. 2010, 2012). However, thelow resolution and high confusion limits of Herschel ( FWHM ∼ ′ at 500 µ m, σ con ≈ mJy beam − , Nguyen et al. 2010) limit thefraction of the CIB that can be directly resolved, with 15% resolvedinto individual galaxies at 250 µ m and 6% at 500 µ m (Clements etal. 2010; Glenn et al. 2010; B´ethermin et al. 2012a). Thus, thereremains work to be done in identifying the galaxies that emit theCIB, and thus finally complete the census of dust-obscured activityin the Universe and its role in galaxy evolution.Advances in submillimetre imaging technology are just nowallowing us to take up the search once more, taking advantage ofhigher resolution possible with large terrestrial telescopes, and im-proved sensitivity and mapping capability in submillimetre detec-tor arrays. The SCUBA–2 camera is the state-of-the-art in submil-limetre wide-field instrumentation (Holland et al. 2006). The cam-era, now mounted on the 15 m James Clerk Maxwell Telescope(JCMT), consists of 5000 pixels in both 450 µ m and 850 µ m detec-tor arrays with an ′ × ′ field-of-view (16 × that of its predeces-sor, SCUBA). The increase in pixel number is the reward of devel-opments in submillimetre detector technology; SCUBA–2 utilizessuperconducting transition edge sensors (TES) to detect submil-limetre photons, with multiplexed superconducting quantum inter-ference device (SQUID) amplifiers handling read-out, analogousto an optical CCD. SCUBA–2 offers the capability to efficiently c (cid:13) , 000–000 µ m number counts and the CIB (cid:0) D ec li n a t i on ( J2000 ) (cid:1) D ec li n a t i on ( J2000 ) Figure 1. (left) SCUBA–2 450 µ m signal-to-noise ratio map of the COSMOS/CANDELS field. The map has been scaled to emphasize the visibility of 60sources detected at > σ significance (circled). The grey contours show the variation in the noise level, and are at σ = 2 , , , mJy beam − (the solidangle bounded by the σ = 5 mJy beam − contour is Ω ≈ arcmin ). (right) Herschel -SPIRE 500 µ m image of the same region, from the HerMESsurvey (Oliver et al. 2012). This map has been slightly smoothed with a Gaussian kernel to improve presentation. We show the limiting σ = 5 mJy beam − contour used for 450 µ m detection in the SCUBA–2 map and the position of the same galaxies in the left panel. This illustrates the ability of SCUBA–2 topush below the Herschel confusion limit at similar wavelengths, resolving confused emission into individual galaxies. map large (degree-scale) areas, and has the sensitivity to simulta-neously achieve deep (confusion limited) maps at both 450 µ m and850 µ m. At 450 µ m, the resolution attainable with the JCMT is afactor ∼ × finer than the 500 µ m resolution of Herschel , and theconfusion limit is ∼ × fainter.Here we present results from early science observations ofone of the seven components of the JCMT Legacy Survey : theSCUBA–2 Cosmology Legacy Survey (S2CLS) . The goal of theS2CLS is to fully exploit SCUBA–2’s mapping capabilities for thepurpose of exploring the high redshift Universe. The S2CLS willcover several well-studied extragalactic ‘legacy’ fields, includingthe United Kingdom Infrared Deep Sky Survey Ultra Deep Sur-vey field (UDS), the Cosmological Evolution field (COSMOS), theExtended Groth Strip, and the Great Observatories Origins DeepSurvey (North) fields. We present the deepest blank-field map at450 µ m yet produced (in the COSMOS field), and measure the fluxdistribution and abundance of the extragalactic sources revealedwithin it. In § § µ m number counts and eval-uate the contribution to the CIB at 450 µ m. We briefly discuss andsummarize our findings in § § Observations were conducted in Band 1 weather conditions( τ
225 GHz < . ) over 22 nights between 23 rd January and 20 th May 2012 totalling 50 hours of on-sky integration. The mappingcentre of the SCUBA–2 COSMOS/CANDELS field is α = h1 m s , δ = ◦ ′ ′′ , chosen to be in the footprint of the Hubble Space Telescope
CANDELS (Grogin et al. 2011; Koeke-moer et al. 2011) . A standard 3 arcmin diameter ‘daisy’ mappingpattern was used, which keeps the pointing centre on one of thefour SCUBA–2 sub-arrays at all times during exposure. Individual 30 min scans are reduced using the dynamic iterativemap-maker of the
SMURF package (Jenness et al. 2011; Chapin etal. 2012 in prep). Raw data are first flat-fielded using ramps brack-eting every science observation, scaling the data to units of pW.The signal recorded by each bolometer is then assumed to be alinear combination of: (a) a common mode signal dominated by at-mospheric water and ambient thermal emission; (b) the astronom-ical signal (attenuated by atmospheric extinction); and finally (c)a noise term, taken to be the combination of any additional sig-nal not accounted for by (a) and (b). The dynamic iterative mapmaker attempts to solve for these model components, refining themodel until convergence is met, an acceptable tolerance has beenreached, or a fixed number of iterations has been exhausted (in thiscase, 20). This culminates in time-streams for each bolometer thatshould contain only the astronomical signal, corrected for extinc-tion, plus noise. The signal from each bolometer’s time stream isthen re-gridded onto a map, according to the scan pattern, with thecontribution to a given pixel weighted according to its time-domainvariance (which is also used to estimate the χ tolerance in the fitderived by the map maker). Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey,http://candels.ucolick.org/c (cid:13) , 000–000
J. E. Geach et al.
SNR N u m be r o f p i x e l s Detection limit−5 0 5 10
Figure 2.
Histogram of values in the SCUBA–2 450 µ m signal-to-noise ra-tio map (Fig. 1), indicating the characteristic positive tail due to the pres-ence of real astronomical sources. The solid line is a Gaussian centred atzero with a width of σ = 1 , and the darker shaded histogram shows thehistogram of pixel values in a map constructed by inverting a random 50%of the input scans; we use this for simulations of completeness, describedin § σ = 3 . , which yields a rea-sonably complete and reliable catalogue (see § The sky opacity at JCMT has been obtained by fitting ex-tinction models to hundreds of standard calibrators observed sincethe commissioning of SCUBA–2 (Dempsey et al. 2012). The op-tical depth in the 450 µ m band was found to scale with the Cal-tech Submillimetre Observatory (CSO) 225 GHz optical depth as: τ = 26 . τ − . . Note that this scaling is slightly differ-ent from the original SCUBA relations (see Archibald et al. 2002;Dempsey et al. 2012).Filtering of the time-series is performed in the frequency do-main, with band-pass filters equivalent to angular scales of ′′ <θ < ′′ (i.e. frequencies of f = v/θ , where v is the scan speed).The reduction also includes the usual filtering steps of spike re-moval ( > σ deviations in a moving boxcar) and DC step correc-tions. Throughout the iterative map making process, bad bolome-ters (those significantly deviating from the model) are flagged anddo not contribute to the final map. Maps from independent scansare co-added in an optimal stack using the variance of the datacontributing to each pixel to weight spatially aligned pixels. Fi-nally, since we are interested in (generally faint) extragalactic pointsources, we apply a beam matched filter to improve point sourcedetectability, resulting in a map that is convolved with an estimateof the 450 µ m beam. The average exposure time over the nominal3 arcminute daisy mapping region (in practice there is usable databeyond this) is approximately ksec per ′′ × ′′ pixel. The flux calibration of SCUBA–2 data has been examined byanalysing all flux calibration observations since Summer 2011 un-til the date of observation. The derived beam-matched flux conver-sion factor (FCF) has been found to be reasonably stable over thisperiod, and the average FCFs agree (within error) to those derived from the subset of standard calibrators observed on the nights of theobservations presented here. Therefore we have adopted the canon-ical calibration of
FCF = 540 ± Jy beam − pW − here. Acorrection of ∼
10% is included in order to compensate for flux lostdue to filtering in the blank-field map. This is estimated by insertinga bright Gaussian point source into the time stream of each obser-vation to measure the response of the model source to filtering.
We present the 450 µ m signal-to-noise ratio map of the COS-MOS/CANDELS field in Fig. 1. For comparison, we also show a Herschel
SPIRE 500 µ m map of the same region to illustrate thegain in resolution that JCMT/SCUBA–2 offers at similar wave-lengths . The 450 µ m map has a radially varying sensitivity, whichis nearly uniform over the central 3 ′ (the nominal mapping area)and smoothly increases in the radial direction as the effective ex-posure time decreases for pixels at the edge of the scan pattern,which have fewer bolometers contributing to the accumulated ex-posure. The total area of the map considered for source extractionis 140 arcmin , where the rms noise is below 5 mJy beam − . Ahistogram of pixel values in the σ mJy beam − region isshown in Fig. 2.To identify extragalactic point sources, we search for pix-els in the (beam convolved) signal-to-noise ratio map with values > Σ thresh . If a peak is found, we record the peak-pixel sky co-ordinate, flux density and noise, mask-out a circular region equiv-alent to ≃ × the size of the 8 ′′ beam at 450 µ m, reduce Σ thresh by a small amount and then repeat the search. The floor value, be-low which we no longer trust the reality of ‘detections’ is chosen tobe the signal-to-noise level at which the contamination rate due tofalse detections (expected from pure Gaussian noise) exceeds 5%,corresponding to a significance of σ ≈ × − beam − . We project that the confusion limitis at ∼ − Completeness is estimated by injecting a noise model with ar-tificial point sources. To create maps with no astronomical sourcesbut approximately the same noise properties of the real map, wegenerate jackknife realisations of the map where, in each fake map,a random half of individual scans have their signal inverted beforeco-addition (e.g. Weiss et al. 2009). Fig. 2 shows the equivalent his-togram of signal-to-noise ratio values in the jackknife map, whichdemonstrates the clean removal of astronomical sources, and thesimilarity with pure Gaussian noise. The recovery rate of sourcesas a function of input flux and local noise gives the completenessfunction: fake sources in batches of 10 are inserted into thejackknife map, where each source selected from a uniform flux dis-tribution < ( S / mJy) < . The 2-dimensional completenessfunction is shown in Fig. 3.In addition to the completeness correction, this technique al-lows us to estimate the noise-dependent flux boosting that occursfor sources with true fluxes close to the noise limit of the map, and The
Herschel map was made from the Level 2.5 processed data productsdownloaded from the public
Herschel
Science Archive. The data were co-added with sky coverage used as an estimator for image noise level, and re-binned into the SCUBA–2 image reference frame, using nearest-neighboursampling. The 1 σ noise level of this SPIRE map (including confusion) is6.2 mJy beam − c (cid:13) , 000–000 µ m number counts and the CIB so we can construct an equivalent ‘surface’ in the noise–(measured)flux plane that can be use to de-boost the fluxes measured for pointsources in the real map (Table 1). The typical de-boosting correc-tion is B < S = 0 . mJy. Sources werethen extracted in exactly the same manner as the real data and com-pleteness and flux boosting corrected as described above and thencompared to the input distribution. This procedure was repeated100 times and the average recovered source counts compared tothe input model. The recovered differential and cumulative numbercounts were found to be consistent with the input number countrealisations, indicating that our source detection and completenesscorrections are not significantly biased. µ m emitters In Table 1 and Fig. 4 we present the number counts at 450 µ m, cor-rected for flux boosting and incompleteness. The differential countsare well-described by a Schechter function: dNdS = (cid:18) N ′ S ′ (cid:19) (cid:16) SS ′ (cid:17) − α exp (cid:18) N ′ S ′ (cid:19) , (1)with S ′ = 10 mJy (fixed at a well-measured part of the flux distri-bution), N ′ = (490 ± deg − and α = 3 . ± . .At flux densities above 20 mJy, the number counts from thissurvey are complemented by the equivalent measurements from Herschel surveys, which survey wider areas at 500 µ m to shallowerdepths, and so find the rarer, bright sources (nearby galaxies, ex-tremely luminous distant sources and gravitationally lensed galax-ies) that are not present in our map (Clements et al. 2010; Negrelloet al. 2010). We focus on two Herschel surveys; HerMES, whichhas obtained confusion limited maps reaching a detection limit of S ≈ mJy (Oliver et al. 2012) and the Herschel -ATLAS sur-vey, which has mapped several hundreds of square degrees at ashallower depth (Eales et al. 2010). As Fig. 4 shows, our 450 µ mcounts are in excellent agreement at ≈
20 mJy where the
Herschel and SCUBA–2 CLS survey flux distributions meet. Below
Her-schel ’s confusion limit the 500 µ m galaxy number counts have beeninferred statistically, by both stacking (B´ethermin et al. 2012b) andpixel fluctuation analyses (Glenn et al. 2010), again indicating con-sistency with the directly measured 450 µ m number counts in ap-proximately the same flux regime.Recently, Chen et al. (2012) presented SCUBA–2 450 µ m ob-servations of SMGs in the field of the lensing cluster A 370. Thebenefit of observing a lensing cluster is – provided a lens model isknown – the ability to probe further down the luminosity functionthan would otherwise be possible for the same flux limit, with faintbackground sources boosted by the cluster potential. We comparethe ‘delensed’ counts of 450 µ m emitters derived from 12 galaxiesin the field of A 370 in Fig. 4, indicating broad agreement with ourblank-field counts within the errors in the same flux range. Afterdelensing, Chen et al. (2012) are able to probe slightly fainter than I npu t f l u x ( m Jy ) N o i s e ( m J y ) C o m p l e t e n ess Figure 3.
Completeness of the 450 µ m catalogue as a function of local noiseand input flux based on input-and-recovery simulations using jackknife re-alisations of the map noise. Modelling the completeness as a 2-dimensionalfunction is required due to the radially varying sensitivity in the map (Fig.1). The same simulations allow us to estimate the difference between true(input) flux and recovered (i.e. observed) flux densities across the same pa-rameter space, and we use this information to correct the number countsaccordingly. our catalogue, and the number counts at the 4.5 mJy level are alsoconsistent with an extrapolation of our best fit Schechter functionto the same limit. µ m background light What fraction of the CIB at 450 µ m have we resolved into galax-ies? The integrated flux density of point sources detected at 450 µ m(corrected for completeness) is I ν (450 µ m) = (7 . ± . × − MJy sr − . The absolute intensity of the CIB at 450 µ m mea-sured by COBE –FIRAS is I ν (450 µ m) = 0 . ± . MJy sr − ,thus we have directly resolved ± % of the CIB at 450 µ m(the uncertainty is dominated by the COBE –FIRAS measurement;Fixsen et al. 1998). For comparison, the deepest
Herschel surveyshave directly resolved 5–6% of the CIB at 500 µ m (Oliver et al.2010; B´ethermin et al. 2012a). We show the integrated brightnessof the 450 µ m emitters, relative to the absolute intensity of the CIBin Fig. 4.To measure the contribution to the CIB at 450 µ m by galaxiesnot formally detected in the SCUBA–2 map, but which are knownto be infrared–bright galaxies, we stack the map at the position of1600 galaxies selected from a catalogue generated from the Spitzer -COSMOS MIPS 24 µ m image of the same region (Sanders et al.2007). First, we remove point sources from the 450 µ m map, usinga point spread function (PSF) constructed by averaging the two di-mensional profiles of sources detected at > σ . This PSF was thennormalised to the flux of each individual source in our catalogue,and subtracted from the map. This yields a residual map where theonly flux (in addition to that of noise) is contributed by sources notin our catalogue. The 450 µ m is then stacked at the position of the24 µ m sources, averaging the flux with a weight equivalent to theinverse of the variance of the map at each position.The average 450 µ m flux density of 24 µ m sources with mean24 µ m flux h S i = 0 . mJy is h S i = 2 . ± . mJy.The resulting contribution to the 450 µ m background is . ± c (cid:13) , 000–000 J. E. Geach et al. - . . Glenn et al. (2010) 500 m m P ( D ) Bethermin et al. (2012a) 500 m m stackLacey et al. (2008) semi−analyticBethermin et al. (2012b) parametricSchechter fitSCUBA−2 450 m m this workOliver et al. (2010) 500 m mClements et al. (2010) 500 m mChen et al. (2012) 450 m m delensed S (mJy) d N d S ( m Jy - deg r ee - ) COBE−FIRAS 450 m m background1 10 . . . . I n / C I B ( % ) SCUBA−2 450 m m resolvedSCUBA−2 450 m m (24 m m stack)Bethermin et al. (2012a) 500 m m (24 m m stack)Integrated Schechter fit S (mJy) I n ( > S ) ( M Jy s r - ) Figure 4. (left) Differential number counts of galaxies detected at 450 µ m (error bars are derived from Poisson statistics). We compare the 450 µ m counts tothose measured recently by Chen et al. (2012) and by Herschel at 500 µ m (B´ethermin et al. 2012a; Glenn et al. 2010; Oliver et al. 2010; Clements et al. 2010),and to the predictions of a numerical model of galaxy formation (Baugh et al. 2005; Lacey et al. 2008) and a parametric model of the evolution of the infraredluminosity density of the Universe (B´ethermin et al. 2012b). We fit the measured 450 µ m counts with a Schechter function; the reason it fails to reproduce theshape of the 500 µ m number counts (and those predicted by the models) at bright fluxes is because in this flux regime the counts have significant contributionsfrom (a) bright local star-forming galaxies and (b) distant galaxies boosted to high observed flux by gravitational lensing; the SCUBA–2 map is too smallto adequately sample these populations. Note – no 450 µ m/500 µ m colour correction has been made to the 500 µ m data. (right) Integrated surface brightnessof 450 µ m emitters relative to the CIB measured by COBE –FIRAS at 450 µ m (Fixsen et al. 1998). The directly measured number counts are well-fitted by aSchechter function, the extrapolation of which agrees with the CIB derived from a stack of 24 µ m-emitting galaxies not individually detected in the SCUBA–2map. Thus, we directly resolve 16 ±
7% of the CIB measured by
COBE –FIRAS into galaxies, with an additional ≈
40% contributed by 24 µ m-emitting galaxiesnot formally detected at 450 µ m. We project that 100% of the CIB at 450 µ m is recovered at a flux density of S > . mJy. . MJy sr − (the uncertainty is σ/ √ N , with σ the standard de-viation in the stack and N the sample size). A simple simulationwas performed to test whether the stacking methodology describedabove produces unbiased estimates of the submillimetre flux. Theresidual flux map was inverted (by multiplying by − ) and sim-ulated sources were inserted using the derived PSF as a model,with the input fluxes of the fake sources set to S = 10 S upto a maximum of S = 5 mJy. The positions were set to the24 µ m catalogue positions, rotated 90 degrees about the map cen-tre, thus preserving clustering information. The stacking procedurewas then repeated as for the real catalogue. The mean input fluxwas S = 1 . mJy per source, and the recovered stacked flux was S = 1 . ± . mJy. The recovered flux is slightly low comparedto the input flux at the 1.5 σ level, however this does not affect ourconclusions, given the uncertainties in the 450 µ m flux calibrationand the absolute measured value of the CIB at 450 µ m.Excluding those detected as bright point sources, the 24 µ m-selected galaxies contribute (2 . ± . × − MJy sr − , or ± % of the CIB at 450 µ m. Therefore, in addition to the directlydetected sources, in total we can account for ± % of theCIB at 450 µ m using the SCUBA–2 map. Note that our stackedvalue is in good agreement with the background derived from astack of 24 µ m-emitters in Herschel µ m images (B´ethermin etal. 2012b), and is also consistent with the intensity expected froman extrapolation of a Schechter function fit to the directly measurednumber counts (Fig. 4). Table 1.
Number counts of 450 µ m-selected galaxies. N indicates the rawnumber of galaxies in each bin ( δS = 5 mJy), and the completeness ( C )and de-boosting ( B ) corrections represent the mean corrections for galaxiesin each bin (note that each galaxy is de-boosted individually, with the cor-rection increasing for lower flux densities). Uncertainties in the differentialcounts are the 1 σ confidence range assuming Poisson statistics (Gehrels1986). S N dN/dS N ( >S ′ ) a hCi b hBi c (mJy) (mJy − deg − ) (deg − ) . . +62 . − . . +339 . − . . . . . +32 . − . . +172 . − . . . . . +16 . − . . +94 . − . . . . . +14 . − . . +71 . − . . . a S ′ corresponds to the lower edge of the bin, i.e. ( S − . mJy b average completeness correction applied c average flux de-boosting correction applied We compare our results to the phenomenological model ofB´ethermin et al. (2012b), who use a ‘backwards evolution’ param-eterisation of the the infrared luminosity density (as traced by dustystar-forming galaxies; see also Lagache et al. 2004). The B´etherminet al. (2012b) model assumes that the star formation modes ofgalaxies can be either described as ‘main sequence’ (i.e. SFR scales c (cid:13) , 000–000 µ m number counts and the CIB with stellar mass) or ‘starburst’, with spectral energy distributionsdefined by the latest stellar synthesis template libraries. The evo-lution of the luminosity functions of these two populations inte-grated over cosmic history provides good fits to the observed num-ber counts of galaxies at 24, 70, 100, 160, 250, 350, 500, 850,1100 µ m and 1.4 GHz (as well as integrated observables such as theevolution of the volume averaged star formation rate and cosmicinfrared background). Here we confirm that the number counts of450 µ m emitters predicted by the model is also in good agreementwith the measured 450 µ m number counts in the flux range probedby our SCUBA–2 survey.We also compare the measured counts to the GALFORM semi-analytic model of galaxy formation (Cole et al. 2000; Baugh 2006;Lacey et al. 2008; Almeida et al. 2011). This prescription predictsthe formation and evolution of galaxies within the Λ CDM modelof structure formation (Springel et al. 2005), and includes the keyphysics of the galaxy formation (and evolution) process: radiativecooling of gas within the dark matter halos, quiescent (by whichwe mean non-burst driven) star formation in the resultant discs,mergers, chemical enrichment of the stellar populations and inter-galactic medium and feedback from supernovae and active galacticnuclei. As Fig. 4 shows, the numerical model slightly over-predictsthe abundance of 450 µ m emitters in the flux range probed. Never-theless, the reasonable agreement between the shape of the countspredicted by GALFORM and the data is encouraging for models ofgalaxy formation that aim to reproduce the full range of emissionprocesses of galaxies at long wavelengths.The 8 arcsec resolution of the 450 µ m SCUBA–2 map allowsus to accurately identify the optical/near-infrared counterparts ofthe SMGs, and we have identified the most likely counterpart tothe majority of 450 µ m sources in our sample (I. G. Roseboom et al.2012 in prep). The wealth of legacy data available in the COSMOSfield then provides the means to estimate the redshift distributionof the population. We have used 13 bands of optical/near-infraredphotometry, including CFHT ugri , Subaru SuprimeCam z’ , VISTA YJHK , HST
F125W and F160W and
Spitzer
IRAC [3.6] and [4.5]to evaluate the photometric redshifts of all the identified galaxies(the typical 1 σ uncertainty based on the confidence level of thetemplate fit is δz = 0 . ). The redshift distribution is shown inFig. 5, indicating that the majority of our sample lie at z < , witha mean redshift of h z i = 1 . (a full analysis of the source iden-tification and redshift distribution is to be presented in Roseboomet al. 2012 in prep). This is a clear indication that the 450 µ m se-lection is probing a lower redshift population than previous 850 µ mselected samples, which have typical redshifts of h z i = 2 . (e.g.Chapman et al. 2005; Wardlow et al. 2010). The shape of the red-shift distributions predicted both by the phenomenological modeland numerical model described above (for galaxies at the same fluxlimit) are also in good agreement with the measured distribution;both models predict little contribution from galaxies at z > (al-though a high redshift ‘tail’ is present in both models).Assuming the directly detected sources representing 16 ± µ m are star-forming galaxies at h z i = 1 . , thentheir total (rest-frame 8–1000 µ m) luminosities are in the ultralu-minous class, L IR ≈ . × L ⊙ (Chary & Elbaz 2001). If thegalaxies contributing to the 24 µ m stack described in § µ m is emitted by galaxies with L IR > . × L ⊙ . This is broadly consistent with the picturethat at z ≈ , the star formation rate budget of the Universe is dom-inated by galaxies in the LIRG class, with star formation rates of Average 1 s uncertaintyin photometric redshift Bethermin et al. (2012b)Lacey et al. (2008)
Mean 850 m m redshiftWardlow et al. (2010) Redshift d N d z ( d e g - ) Figure 5.
Redshift distribution of 450 µ m-selected SMGs in our sample (notcorrected for completeness), derived from 13-band photometric redshift es-timates ( § h z i =1.3, and the vastmajority of the 450 µ m-selected SMGs lie at z < . For comparison, theaverage redshift of SMGs selected at 850 µ m is z ≈ . (Wardlow et al.2010), indicating the efficacy at which the 450 µ m selection samples a pop-ulation of SMGs at lower redshift and therefore an important complement toany census of the dusty Universe. We compare the shape of the redshift dis-tribution to the models of B´ethermin et al. (2012b) and Lacey et al. (2008)shown in Fig. 4 for galaxies with S > mJy (we have area-normalisedboth model distributions, since the observed redshift distribution containsno completeness correction). The average redshift and shape of both modeldistributions is in good agreement with observations, suggesting that – atthis flux limit – there is little contribution from galaxies at z > . order 10 M ⊙ yr − (Dole et al. 2006; Rodighiero et al. 2010; Mag-nelli et al. 2011).An extrapolation of the Schechter function fit to the directlymeasured number counts (which agrees well with the backgroundat S ≈ mJy, derived from the stack of 24 µ m sources), im-plies that 100% of the CIB at 450 µ m should be recovered at .
40% of the CIB at 450 µ m islikely to be emitted by galaxies with L IR < . × L ⊙ , im-plying galaxies star formation rates of a few tens of Solar massesper year. However, we cannot as yet rule out what fraction of theremaining CIB light might be emitted by faint 450 µ m emitters athigher redshifts; note that a galaxy with S ≈ mJy at z > has a typical luminosity of L IR > . × L ⊙ (Chary & Elbaz2001), again indicating the importance of LIRG-class galaxies inthe cosmic infrared budget. Characterizing the high redshift tail ofthe 450 µ m population is an important next step. The SCUBA–2 camera on the 15 m JCMT repesents the state-of-the-art in panoramic submillimetre imaging, and has recently begun c (cid:13) , 000–000 J. E. Geach et al. scientific observations in earnest. In this paper we have presentedresults from the first deep, blank-field cosmological map at 450 µ m( σ = 1 . mJy); part of the SCUBA–2 Cosmology Legacy Sur-vey, the largest of the seven JCMT Legacy Surveys. Using a 450 µ mmap of the well-studied extragalactic COSMOS/CANDELS field,we have(i) made the first unbiased, blank-field determination of thenumber counts of galaxies at 450 µ m, at a flux density limit of S > mJy. This probes below the confusion limit of Her-schel , complementing the number counts measured at fluxes above20 mJy over wider areas in major
Herschel submillimetre surveys;(ii) measured the contribution of these galaxies to the cosmic in-frared background at 450 µ m: we resolve 16% of the CIB into indi-vidual galaxies. The ability of SCUBA–2 to ‘pin-point’ the galaxiesresponsible for the emission of the CIB is a critical step in under-standing the properties of the galaxies that are forming the majorityof stars in the Universe at this epoch;(iii) an additional ≈
40% of the CIB can be recovered in theSCUBA–2 map by stacking
Spitzer
MIPS-detected 24 µ m emitters.Using this analysis we estimate that the majority ( ≈ µ m is emitted by galaxies with S > mJy;(iv) a preliminary analysis of the redshift distribution of the450 µ m emitters (based on high-quality photometric redshifts avail-able for this field) imply that the typical redshift of galaxies with S > mJy is h z i = 1 . , with the majority lying at z < .The typical luminosity of galaxies in our sample are estimated tobe in the ultraluminous class, with L IR > L ⊙ . If the galax-ies contributing to the statistical stack of 24 µ m emitters describedabove are at a similar redshift, then we project that the majorityof the CIB at 450 µ m is emitted by ‘LIRG’ class galaxies with L IR > . × L ⊙ .These are the first results of the S2CLS. The final goal of the surveywill be to map a quarter of a square degree to σ = 1 . mJy, anda wider, ten square degree area to σ = 1 . mJy, yielding > SMGs with which to (a) determine the submillimetre luminosityfunction and its evolution over cosmic time; (b) search for therarest, most luminous SMGs at high- z ; (c) resolve the 450 µ m back-ground; (d) accurately measure the clustering properties of SMGsand determine their typical environments (including rare protoclus-ters) and (e) build-up the large samples required to properly re-late SMGs to other star-forming (ultraviolet/near-infrared selected)populations at z ∼ and thus gain further insight into SMGs’ rolein the overall history of galaxy evolution. ACKNOWLEDGEMENTS
J.E.G. is supported by a Banting Fellowship, administered byNSERC. J.S.D. acknowledges the support of the European Re-search Council via the award of an Advanced Grant, and the sup-port of the Royal Society via a Wolfson Research Merit award.M.M. and I.G.R. acknowledge the support of STFC. The authorsthank M. B´ethermin for providing data on the 500 µ m stacking andparametric model, and J. Dempsey for advice on the SCUBA–2 cal-ibration. It is also a pleasure to thank the JCMT telescope operatorsJ. Hoge, J. Wouterloot and W. Montgomerie, without whom theseobservations would not be possible. The James Clerk MaxwellTelescope is operated by the Joint Astronomy Centre on behalf ofthe Science and Technology Facilities Council of the United King-dom, the Netherlands Organisation for Scientific Research, and theNational Research Council of Canada. Additional funds for the contruction of SCUBA-2 were provided by the Canada Foundationfor Innovation. Herschel is an ESA space observatory with scienceinstruments provided by European-led Principal Investigator con-sortia and with important participation from NASA. This work isbased in part on observations made with the
Spitzer Space Tele-scope , which is operated by the Jet Propulsion Laboratory, Califor-nia Institute of Technology under a contract with NASA.
REFERENCES
Almeida, C., et al., 2011, MNRAS, 417, 2057Archibald, E. N., et al., 2002, MNRAS, 336, 1Aretxaga, I., et al., 2007, MNRAS, 379, 1571Austermann, J. E., et al., 2010, MNRAS, 401, 160Barger, A. J., et al., 1998, Nature, 394, 248Baugh, C. M., et al., 2005, MNRAS, 356, 1191B´ethermin, M., et al., 2012a, A&A, 542, 58B´ethermin, M., et al., 2012b, ApJ, 757, L23Blain, A. W., Smail, I., Ivison, R. J., & Kneib, J.-P. 1999, MN-RAS, 302, 632Chapman, S. C., Blain, A. W., Smail, I., & Ivison, R. J. 2005, ApJ,622, 772Chary, R., Elbaz, D., 2001, ApJ, 556, 562Chen, C.-C., Cowie, L. L., Barger, A. J., Casey, C. M., Lee,N., Sanders, D. B., Wang, W.-H., Williams, J. P. 2012,arXiv:1209.4377Clements, D. L., et al., 2010, A&A, 518, 8Cole, S., Lacey, C. G., Baugh, C. M., Frenk, C. S., 2000, MNRAS,319, 168Coppin, K. E. K., et al., 2006, MNRAS, 372, 1621Coppin, K. E. K., et al., 2008, MNRAS, 389, 45Dav´e, R., Finlator, K., Oppenheimer, B. D., Fardal, M., Katz, N.,Keres, D., Weinberg, D. H., 2010, MNRAS, 404, 1355Dempsey, J. T., et al., 2012, SPIE Conference series 8452, Mil-limetre, Submillimetre and Far-infrared Detectors and Instru-mentation for Astronomy VI, arXiv1208.4622Devlin, M. J., et al., 1999, Nature, 458, 737Dole, H., et al., 2006, A&A, 451, 417Dunlop, J. S., et al., 2004, MNRAS, 350, 769Dunlop, J. S., 2011, Galaxy evolution: Infrared to millimeterwavelength perspective. Proceedings of a conference held atGuilin, China 25-29 October 2010. Edited by W. Wang, J. Lu,Z. Luo, Z. Yang, H. Hua, and Z. Chen. San Francisco: Astro-nomical Society of the Pacific, 209Eales, S., et al., 2010 PASP, 122, 499Fixsen, D. J., Dwek, E., Mather, J. C., Bennett, C. L., & Shafer,R. A., 1998, ApJ, 508, 123Gehrels, N., 1986, ApJ, 303, 336Glenn, J., et al., 2010, MNRAS, 409, 109Griffin, M. J., et al., 2010, A&A, 518, L3Grogin, N. A., et al., 2011, ApJS, 197, 35Hainline, L. J., et al., 2011, ApJ, 740, 96Hauser, M. G., & Dwek, E., 2001, ARA&A, 39 249Hickox, R. C., et al., 2012, MNRAS, 421, 284Holland, W., et al., 2006, Millimeter and submillimetre Detectorsand Instrumentation for Astronomy III. Edited by Zmuidzinas,J., Holland, W. S., Withington, S., Duncan, W. D., Proceedingsof the SPIE. 6275Hughes, D. H., et al., 1998, Nature, 394, 241Ivison, R. J., Smail, Ian, Barger, A. J., Kneib, J.-P., Blain, A. W.,Owen, F. N., Kerr, T. H., Cowie, L. L., 2000, MNRAS, 315, 209Ivison, R. J., et al., 2002, MNRAS, 337, 1 c (cid:13) , 000–000 µ m number counts and the CIB Ivison, R. J., et al., 2005, MNRAS, 364, 1025Ivison, R. J., et al., 2010, MNRAS, 404, 198Jenness, T., et al., 2011, Astronomical Society of the Pacific Con-ference Series, Astronomical Data Analysis Software and Sys-tems XX, 442, 281Knudsen, K. K., van der Werf, P. P. & Kneib, J.-P., 2008, MNRAS,384, 1611Koekemoer, A. M., et al., 2011, ApJS, 197, 36Lacey, C. G., Baugh, C. M., Frenk, C. S., Silva, L., Granato, G.L., Bressan, A., 2008, MNRAS, 385, 1155Magnelli, B., Elbaz, D., Chary, R. R., Dickinson, M., Le Borgne,D., Frayer, D. T., Willmer, C. N. A., 2011, A&A, 528, 35Michałowski, M., Hjorth, J., Watson, D., 2011, A&A, 514, 67Negrello, M., et al., 2010, Science, 330, 800Nguyen, H. T., et al., 2010, A&A, 518, L5Oliver, S. J., et al., 2010, A&A, 518, L21Oliver, S. J., et al., 2012, The Herschel Multi-tiered ExtragalacticSurvey: HerMES. MNRAS, 424, 1614Pascale, E., et al., 2008, ApJ, 681, 400Pilbratt, G. L., et al., 2010, A&A, 518, 1Puget, J.-L., et al., 1996, A&A, 308, L5Rodighiero, G., et al., 2010, A&A, 515, 8Sanders, D. B., et al., 2007, ApJS, 172, 86Scoville, N., et al.,2007, ApJS, 172, 1Smail, I., Ivison, R. J., & Blain, A. W., 1997, ApJ, 490, L5Smail, I., Ivison, R. J., Blain, A. W., Kneib, J.-P., 2002, MNRAS,331, 495Springel, V., et al., 2005, Nature, 435, 629Swinbank, A. M., et al., 2006, MNRAS, 371, 465Wardlow, J. L., et al.,2011, MNRAS, 415, 1479Weiss, A., et al., 2009, ApJ, 707, 1201 c (cid:13)000