ebook img

Extremely Red Submillimeter Galaxies: New z>~4-6 Candidates Discovered using ALMA and Jansky VLA PDF

1.6 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Extremely Red Submillimeter Galaxies: New z>~4-6 Candidates Discovered using ALMA and Jansky VLA

Draftversion January9,2017 PreprinttypesetusingLATEXstyleemulateapjv.01/23/15 EXTREMELY RED SUBMILLIMETER GALAXIES: NEW Z &4–6 CANDIDATES DISCOVERED USING ALMA AND JANSKY VLA Soh Ikarashi1, R.J. Ivison2,3, KarinaI. Caputi1, Koichiro Nakanishi4,5,6, ClaudiaD. P. Lagos7, M.L.N. Ashby8, Itziar Aretxaga9, JamesS. Dunlop2, Bunyo Hatsukade4, DavidH. Hughes9, Daisuke Iono4,5, Takuma Izumi10, Ryohei Kawabe4,5, Kotaro Kohno10,11, Kentaro Motohara10, Kouji Ohta12, Yoichi Tamura10, Hideki Umehata10,13, GrantW. Wilson14, Kiyoto Yabe15, MinS. Yun14 1KapteynAstronomicalInstitute,UniversityofGroningen,P.O.Box800,9700AVGroningen,TheNetherlands 7 2InstituteforAstronomy,UniversityofEdinburgh,RoyalObservatory,BlackfordHill,EdinburghEH93HJ,UK 3EuropeanSouthernObservatory,KarlSchwarzschildStr.2,D-85748Garching,Germany 1 4NationalAstronomicalObservatoryofJapan,Mitaka,Tokyo181-8588,Japan 0 5TheGraduateUniversityforAdvancedStudies(SOKENDAI),2-21-1Osawa,Mitaka,Tokyo181-8588,Japan 2 6JointALMAObservatory,AlonsodeCordova3107,Vitacura,Santiago7630355,Chile 7InternationalCentreforRadioAstronomyResearch,UniversityofWesternAustralia,7Fairway,Crawley6009,PerthWA,Australia n 8Harvard-SmithsonianCenterforAstrophysics,60GardenSt.,Cambridge,MA02138,USA a 9InstitutoNacionaldeAstrof´ısica,O´pticayElectro´nica(INAOE),Aptdo.Postal51y216,72000Puebla,Mexico 10InstituteofAstronomy,UniversityofTokyo,2-21-1Osawa,Mitaka,Tokyo181-0015,Japan J 11ResearchCenterfortheEarlyUniverse,SchoolofScience,UniversityofTokyo,7-3-1Hongo,Bunkyo,Tokyo113-0033,Japan 5 12DepartmentofAstronomy,KyotoUniversity,Kitashirakawa-Oiwake-Cho,Sakyo-ku,Kyoto606-8502,Japan 13TheOpenUniversityofJapan,2-11Wakaba,Mihama-ku,Chiba261-8586 14DepartmentofAstronomy,UniversityofMassachusetts,Amherst,MA01003,USA ] A 15KavliInstituteforthePhysicsandMathematicsoftheUniverse(WPI),TheUniversityofTokyo,5-1-5Kashiwanoha,Kashiwa,Chiba, 277-8583,Japan G Draft version January 9, 2017 . h ABSTRACT p - We present the detailed characterization of two extremely red submillimeter galaxies (SMGs), o ASXDF1100.053.1and 231.1, with the Atacama Large Millimeter/submillimeter Array (ALMA) and r the Jansky Very Large Array(VLA). These SMGs were selected originally using AzTEC at 1100µm, t s and are observed by Herschel to be faint at 100–500µm. Their (sub)millimeter colors are as red as a – or redder – than known z & 5 SMGs; indeed, ASXDF1100.053.1 is redder than HFLS3, which lies [ at z = 6.3. They are also faint and red in the near-/mid-infrared: 1µJy at IRAC 4.5µm and ∼ 1 <0.2µJy in the Ks filter. These SMGs are also faint in the radio waveband, where F6GHz =4.5µJy v for ASXDF1100.053.1 and F = 28µJy for ASXDF1100.231.1, suggestive of z = 6.5+1.4 and 1.4GHz −1.1 8 z =4.1+0.6 for ASXDF1100.053.1 and 231.1,respectively. ASXDF1100.231.1 has a flux excess in the 4 −0.7 3.6-µm filter, probably due to Hα emission at z = 4–5. Derived properties of ASXDF1100.053.1 for 4 z =5.5–7.5 and 231.1 for z =3.5–5.5 are as follows: their infrared luminosities are [6.5 7.4] 1012 1 − × 0 and[4.2−4.5]×1012L⊙;theirstellarmassesare[0.9−2]×1011and[0.4−3]×1010M⊙;theircircularized half-lightradiiinthe ALMAmapsare 1and.0.2kpc( 2–3kpcfor90%ofthe totalflux). Lastly, . 1 their surface infrared luminosity densit∼ies, ΣIR, are 1 ∼1012 and & 1.5 1013 L⊙kpc−2, similar to 0 ∼ × × values seen for local (U)LIRGs. These data suggest that ASXDF1100.053.1 and 231.1 are compact 7 SMGs at z &4 and can plausibly evolve into z &3 compact quiescent galaxies. 1 Keywords: submillimeter: galaxies — galaxies: evolution — galaxies: formation — galaxies: high- : v redshift i X r 1. INTRODUCTION LargeArray(VLA) inordertopinpointsourcepositions a (Ivison et al. 1998, 2000, 2002, 2005, 2007; Smail et al. Submillimeter(submm)galaxies(SMGs)withinfrared 1999;Borys et al.2004;Pope et al.2006;Aretxaga et al. (IR, rest-frame 8–1000µm) luminosities, LIR &1012L⊙, 2011;Biggs et al.2011;Yun et al.2012). Intensivestud- are routinely detected in deep continuum images at iesofradio-brightSMGswereabletoyieldspectroscopic λ =850–1300µm using ground-basedsingle-dishtele- obs redshifts for those out to z 3 (e.g. Chapman et al. scopes. Even out to z 7, there is no significant loss of ∼ ∼ 2003, 2005). However, at that time radio sensitivities sensitivity to these SMGs, given the strong negative K- could not detect SMGs beyond z 3, and as many as correctionintheRayleigh-Jeanstailoftheirdustspectral ∼ half of SMGs lacked reliable radio counterparts (see e.g. energy distributions (SEDs) (e.g. Blain et al. 2002). Ivison et al. 2007; Biggs et al. 2011, cf. Lindner et al. Despite 20 years of deep submm surveys since 2011). Later attempts to determine SMG positions and Smail et al. (1997), our knowledge of the upper half redshifts using near-andmid-IRimagingcouldnotfully of the redshift distribution of SMGs remains incom- overcome the bias towards lower redshifts, since the K plete. Early attempts to determine redshifts were corrections there are no more favorable than those in conducted towards SMGs with radio counterparts, be- the radio regime, such that high-redshift sources are cause low-resolution(sub)mm images obtained with sin- muchfainter(e.g.Wardlow et al.2011;Yun et al.2012). gle dishes require high-resolution radio continuum maps Millimeter spectroscopic surveys toward gravitationally- from radio interferometers such as the Jansky Very lensed, dusty, star-forming galaxies, taking advantage of 2 Ikarashi et al. theirapparentultrabrightness,revealedaredshiftdistri- 2014; Asboth et al. 2016; Ivison et al. 2016). Too faint bution stretching out to z 5.8 (e.g. Vieira et al. 2013; atoptical/near-IRwavelengthsto allowmeaningful esti- ∼ Weiß et al. 2013; Strandet et al. 2016). These surveys mationoftheirredshiftsusingclassicalphotometrictech- suggested a larger fraction of SMGs at z & 3 than pre- niques,wehaveinsteaddeterminedphotometricredshifts vious studies of unlensed SMGs, perhaps partly because usingdeep radio/submm/far-IRimages fromthe Janksy they were selected at 1.3mm rather than the traditional VLA,ALMA,SCUBA-2andHerschel,respectively,aim- 0.8–1.1mm, but also because the requirement for high ingtorevealwhetherthesegalaxiesareindeedlocatedat magnification favors galaxies with a long line of sight. very high redshifts — obvious candidate progenitors of We need to reveal the intrinsic redshift distributions of themassivepassivegalaxiesatz &3. Weadoptthrough- unlensed SMGs in large contiguous maps to determine out a cosmology with H = 70kms−1Mpc−1, Ω = 0.3 0 M their abundance in the early Universe and to study the and Ω = 0.7, and all magnitudes refer to the AB sys- Λ evolution of the most massive galaxies via abundance tem. matching with other populations, and with cosmologi- cal predictions (e.g. Hayward et al. 2013; Cowley et al. 2. THETARGETS:ASXDF1100.053.1AND231.1 2015). ASXDF1100.053.1 and 231.1 are the brightest Early (sub)mm interferometric imaging of intrinsi- andsecond-brightest1100-µm-selectedALMA-identified cally bright SMGs, conducted with the IRAM Plateau SMGs among the z & 3 candidates discovered in our de Bure interferometer (PdBI) and the Submillime- ALMA Cycle-1 program (2012.1.00326.S: PI. Ikarashi). ter Array (SMA) (e.g. Gear et al. 2000; Iono et al. The parent sample consists of 221 SMGs discoveredin a 2006; Younger et al. 2007; Dannerbauer et al. 2008; deep AzTEC/ASTE 1100-µmmap covering 950arcmin2 Younger et al. 2009), pinpointed the positions of SMGs, of the Subaru/XMM-Newton Deep Field (SXDF) (e.g. including radio-faint ones, and resulted in the discovery Furusawa et al. 2008), which includes the UKIDSS Ul- of SMGs at z & 4–5 (e.g. Capak et al. 2011). Subse- tra Deep Survey(UDS) field (e.g.Lawrence et al.2007). quently, surveyswith PdBI andthe Combined Arrayfor In our ALMA program, we targeted 30 SMGs from this ResearchinMillimeter-waveAstronomy(CARMA)indi- parent sample, selected on the basis of their faintness in catedthattheredshiftdistributionofintrinsicallybright 1.4-GHz VLA imaging (5σ . 35µJy, Arumugam et al. SMGs most likely stretches to z 6 (Smolˇci´c et al. 2016)andSPIRE250-µmimages(3σ .18.3mJy, ∼ confusion 2012). Oliver et al. 2012), aiming to reveal the tail of the SMG The capabilities of the Atacama Large Millime- redshift distribution. The faintness of these two SMGs ter/submm Array (ALMA) now enable astronomers to at optical/near-/mid-IR wavelengths suggests z & 4–5 rapidly pinpointthe positions of largesamples ofSMGs, (Ikarashi et al. 2015; Fig. 1). The submm (250, 350, with no strong biases (though see Zhang et al. 2016). 500 and 850µm)/mm (1100µm)/radio (1.4GHz) colors ALMA submm continuum imaging surveys towards ofASXDF1100.053.1and231.1are as redas – or redder LABOCA 870-µm-selected SMGs (e.g. Hodge et al. – than known z & 5 SMGs, which suggests that these 2013;Simpson et al.2014)andAzTEC1100-µm-selected new SMGs could lie at z &5 (Fig. 1). We thus focus on SMGs (Ikarashi et al.2015) haveuncovereda number of thesetwoSMGs forapilotstudy ofcandidateextremely radio-faint SMGs. Some of these radio-faint SMGs have high-redshift SMGs. been too faint at optical/near-IR wavelengths to permit estimation of their redshifts using standard techniques 3. DATAANDPHOTOMETRY (Simpson et al. 2014; Ikarashi et al. 2015). Some could Here we describe the observational data used in this lie at very high redshifts, i.e. z & 5; alternatively, they paper. Our images are shown in Figs 2 and 3, and mea- could be heavily dust-obscured SMGs at more moder- surements are listed in Table. 1. ate redshifts, z 3–5. The redshifts of these SMGs ≈ remains a puzzle, with important implications for our 3.1. ALMA 1100-µm continuum understanding of early galaxy evolution. ALMA mm-wave continuum imaging of z & 3 candi- We first describe the ALMA data taken in Cycle 1 dateSMGshaverevealedsurprisinglycompactsizes,sup- (S. Ikarashietal. in preparation: see also Ikarashi et al. portingtheideathatz &3SMGscouldevolveintocom- 2015). Theseobservationswerecarriedoutwithanarray pact quiescent galaxies at z 2 (Ikarashi et al. 2015). configurationsimilartoC32-3,with25working12-man- ∼ The latest intensive optical/near-/mid-IR extragalactic tennas covering uv distances up to 400kλ. On-source ∼ surveys have reported compact quiescent galaxies up to observation times (per target) were 3.6–4.5 minutes. z 4 (Straatman et al. 2015). In order to understand ThetwoSMGswerealsoobservedaspartofanALMA ∼ the formation phase of these massive, passive galaxies continuumimagingsurveyof333brightAzTECSMGsin at z & 3, surveys and studies of SMGs z & 4–5 are as Cycle 2 (2013.1.00781: PI. Hatsukade). These observa- important today as they ever were. tions were carriedout in arrayconfigurations C34-5 and In this paper, we present a detailed multi- C34-7, with 37–38 working 12-m antennas covering uv wavelength analysis of two ALMA-identified galaxies, distances up to 1500kλ. On-source observation times ∼ ASXDF1100.053.1andASXDF1100.231.1,detectedorig- per source were 0.6 minutes. inally in a deep ASTE/AzTEC survey at 1100µm We combined the ALMA data obtained in Cycles 1 (Ikarashi et al.2015). TheseSMGswereselectedforfur- and 2. Synthesized beams were then 0′′.46 0′′.35 (PA, ther scrutiny on the basis of their secure non-detections 69◦) and 0′′.57 0′′.48 (PA, 82◦) for ASXD×F1100.053.1 × inHerschel100–500-µmimages,whichgivethemostuse- and 231.1, respectively, with sensitivities of 70 and ful constraints on redness at submm wavelengths (see 63µJybeam−1 (1σ). ASXDF1100.053.1 and 231.1 were also Cox et al. 2011; Riechers et al. 2013; Dowell et al. detected with S /N = 27 and 29, respectively, with peak Extremely red SMGs 3 total flux densities, F = 3.51 0.15 and 2.28 around ASXDF1100.053.1 and 231.1, where the resid- 1100µm ± ± 0.08mJy. ual maps have been deblending based on the posi- Both ASXDF1100.053.1 and 231.1 appear to be sin- tions of known VLA-1.4 GHz and MIPS 24-µm sources. gle, unblended SMGs, with no signs of multiplic- The respective 250-, 350- and 500-µm flux densities of ity; their ALMA 1100-µm flux densities are consis- ASXDF1100.053.1 in the residual images are 1.3, 0.1 tent (within 1σ) with those measuredby AzTEC/ASTE and4.3mJybeam−1,andthoseofASXDF1100.−231.1are (S.Ikarashietal. in preparation). 0.6, 4.0 and 0.8 mJybeam−1. 3.2. Jansky Very Large Array radio continuum 3.4. SCUBA-2 850-µm continuum 3.2.1. Classic VLA 1.4-GHz continuum Both ASXDF1100.053.1 and 231.1 are detected in the deepSCUBA-2850-µmmapoftheSCUBA-2Cosmology The accurate SMG positions from our ALMA images Legacy Survey Data Release 1. In Geach et al. (2016) enable us to exploit existing deep VLA radio continuum they are referred to as UDS0186 and UDS0206, with maps. ASXDF1100.231.1 was detected at 3.3σ in an ex- 850-µm flux densities of 4.8 1.1 and 4.5 1.1 mJy, isting wide, deep VLA 1.4-GHz image of the SXDF field ± ± respectively. TherespectiveoffsetsbetweentheirALMA (Arumugam et al. 2016); ASXDF1100.053.1 was not de- 1100-µm and SCUBA-2 850-µm positions are 2.5 and tected. The r.m.s. noise of the map is 6–8µJybeam−1, 5.7 arcsec, consistent with the SCUBA-2 positional off- and the FWHM synthesized beam is 1′′.5. set distributions (Simpson et al. 2015b). ∼ 3.2.2. Jansky VLA 6-GHz continuum 3.5. Spitzer mid-IR continuum In order to measure the radio flux density of We use the deep Spitzer IRAC 3.6- and 4.5-µm ASXDF1100.053.1, we conducted new extremely deep maps from the Spitzer Extended Deep Survey (SEDS; JanskyVLA observations. The data wereobtained from Ashby et al.2013)andIRAC5.8-and8.0-µmandMIPS 2015 February to April with the Jansky VLA in its 24-µm data from the Spitzer UKIDSS Ultra Deep Sur- B configuration, using the new 3-bit samplers1, with vey (SpUDS; PI. J. Dunlop; see e.g. Caputi et al. 2011). the WIDAR correlator, covering an almost contiguous IRAC counterpartsofASXDF1100.053.1and231.1were 4-GHz band across 4–8GHz (several spectral windows foundat(RA,Dec)=(02h16m48.19s, 04◦58′59.6′′)and covering a total of 0.25GHz were discarded due to (02h17m59.62s, 04◦46′59.7′′), respec−tively, with offsets radio-frequency inter≈ference). The phase center was set from the ALMA−positions of 0′′.2 and 0′′.5. Photomet- to be the position of ASXDF1100.053.1. The FWHM ric measurementsperformed atthe IRAC positions with field of view (FoV) covers a circular area of radius fixedaperturesandaperturecorrectionsfor2.4′′-φ(IRAC 3.7arcmin in the final map. The total observation time 3.6 and 4.5µm); 2.8′′-φ (IRAC 5.8 and 8.0µm). The was 14hr, of which 10.1hr were spent on-source. We MIPS 24-µm upper limit (3σ) is based on photometry chose J0239 0234as the gain calibrator, using 3C48 as atrandompositions including anaperture correctionfor the bandpas−scalibratorandto setthe fluxdensity scale. 7′′-φ. For <2-σ detections in IRAC maps, we adopt 2σ We reduced the data with CASA and imaged using a upper limits. naturalweightingscheme. The resultingmapreachesan r.m.s.noiselevelof1.1µJybeam−1andhasasynthesized 3.6. Optical/near-IR continuum beamsizeof1′′.5 1′′.2(PA,16◦.2). Giventhecolorcor- We use optical/near-IRimages at B, V, Rc, i′ and z′- × rectionbetweenν =1.4and6GHz foraradiospectral obs bandsfromthe SubaruTelescope(Furusawa et al.2008) index, α= 0.8, the sensitivity of the new Jansky VLA and near-IR images at J, H and K -bands from the − s 6-GHz map is more than 2 deeper than the old VLA UKIRT IR Deep Sky Survey (UKIDSS; Lawrence et al. × 1.4-GHz map. In the new 6-GHz map we detect emis- 2007). We measured fluxes with fixed apertures at the sion at the position of ASXDF1100.053.1: 4.5 1.1µJy positionsoftheIRACcounterpartsandappliedaperture ± (4.0σ). The source characteristics are summarized in corrections: 2′′-φapertureforBthroughK . Errorswere s Table 2. derived from random aperture photometry. Again, For <2-σ detections, we adopt 2σ upper limits. 3.3. Herschel/SPIRE 250–500-µm continuum 4. RADIO/MM-WAVEPHOTOMETRICREDSHIFTS We use the Herschel/SPIRE 250, 350 and 500-µm maps in the UKIDSS UDS field, provided as part The existence of SMGs which are extremely faint of the HerMES (Oliver et al. 2012) 2nd data release at optical to mid-IR wavelengths has long been rec- (DR2). Armed with ALMA positions, accurate to < ognized (e.g. Hughes et al. 1998; Ivison et al. 2000; 0.1′′, it is clear that both of ASXDF1100.053.1 and Wang et al. 2009; Weiß et al. 2009; Walter et al. 2012) 231.1 were not detected in deep imaging by Herschel and radio/submm colors have been used to estimate PACS and SPIRE images (see Figs 2 and 3): the the redshifts of heavily dust-obscured SMGs (e.g. respective flux densities in the 250-, 350- and 500- Carilli & Yun 1999; Hughes et al. 2002; Ivison et al. µm maps are 4.2, 6.3 and 9.5mJybeam−1, and those 2005; Aretxaga et al. 2003, 2005, 2007), exploiting the at the position of ASXDF1100.231.1 are 0.7, 5.2 and tight correlation between radio and far-IR luminosities 3.1mJybeam−1. These values are below the 3-σ lim- seen for local galaxies (Condon 1992). its measured in residual SPIRE maps of 5′ 5′ areas × 4.1. Method 1 We acknowledge funding towards the 3-bit samplers used in We estimate the radio/submm photometric redshifts thisworkfromERCAdvancedGrant321302, COSMICISM. of ASXDF1100.053.1 and ASXDF1100.231.1 by fitting 4 Ikarashi et al. Table 1 PhotometricDataofASXDF1100.053.1and231.1. ASXDF1100.053.1 ASXDF1100.231.1 Wavelength Flux(µJy) Flux(µJy) Reference SuprimeCamB-band(0.45µm) <0.014 <0.016 1 SuprimeCamV-band(0.55µm) <0.023 <0.021 1 SuprimeCamRc-band(0.66µm) <0.027 <0.025 1 SuprimeCami′-band(0.77µm) <0.027 <0.025 1 SuprimeCamz′-band(0.92µm) 0.067±0.036 <0.064 1 WFCAMJ-band(1.2µm) <0.14 <0.14 2 WFCAMH-band(1.6µm) <0.23 <0.23 2 WFCAMKs (2.2µm) <0.19 <0.19 2 IRAC3.6µm 0.61±0.14 1.00±0.17 3 IRAC4.5µm 1.43±0.17 0.93±0.22 3 IRAC5.8µm 3.5±1.9 <3.5 4 IRAC8.0µm <4.7 7.4±2.6 4 MIPS24µm <66 <66 4 PACS110µm <2400 <2400 5 PACS160µm <5000 <5000 5 SPIRE250µm <9600 <8800 5 SPIRE350µm <7700 <9800 5 SPIRE500µm <10000 <10000 5 SCUBA2850µm 4800±1100 4500±1100 6 ALMA1100µm 3510±150 2280±80 7 JVLA6GHz 4.46±1.1 ··· 7 VLA1.4GHz <17.8 27.6±8.7 8 Notes. 2-and3-σ upperlimitsarepresentedfor(stellar)emissionat0.45–8.0µmanddust/synchrotron emissionat24µmthrough1.4GHz,respectively. References: (1)Furusawaetal.(2008);(2)Lawrenceetal.(2007);(3)Ashbyetal.(2013);(4)Caputietal.(2011), (5)Oliveretal.(2012);(6)Geachetal.(2016);(7)thiswork;(8)Arumugametal.(2016). Table 2 JVLAobservations. Observationdate 2015February16 March2,9,17and30 April2 Frequency 4–8GHz Phasecenter(J2000) RA=02h16m48s Dec.=−04◦58′59′′ Gaincalibrator J0239−0234 Fluxdensitycalibrator 3C48 Bandpass calibrator 3C48 Arrayconfiguration B Projectedbaselines 0.2–11km Primarybeam 7.3arcmin(FWHM)at6GHz Synthesizedbeamsize 1.5′′×1.1′′ (PA,16◦.2) Mapnoiselevel 1.1µJybeam−1 dust SED templates to ALMA 1100-µm,SCUBA-2 850- µm and (J)VLA 6- or 1.4-GHz flux densities. Whenmakingradio/(sub)mmphotometricredshiftes- timates,obtainingstrongconstraintsaroundthe peak of the dust SEDs is important, to exclude spurious SED models which return dubious redshift estimates due to the degeneracy between redshift and dust temperature (e.g. Blain et al. 2002). For most SMGs at z 2–3, the ≈ HerschelSPIREimagesat250,350and500µmcoverthe Figure 1. Comparison of submm/mm/radio colors for redshifted dust SED peak. ASXDF1100.053.1 and 231.1 ASXDF1100.053.1 and 231.1 with colors of known z & 5 are not detected in the Herschel SPIRE maps (Figs 2 SMGs from the literature. Submm/mm/radio colors for and 3) and we have therefore included 3σ-upper limits ASXDF1100.053.1 and 231.1 are markedby redcrosses or arrows fromthe SPIREdataat250,350and500µmassurvival based on fluxes in Table. 1. The highest-redshift SMG known, HFLS3atz=6.3(Riechersetal.2013),andknownz∼5SMGs, functions (Isobe et al. 1986), as was done for SCUBA AzTEC1andAzTEC3(Smolˇci´cetal.2015),aremarkedbyblack 450-µm upper limits in radio/submm photometric red- points. The blue, orange and green lines mark the color track as shift estimates in Aretxaga et al. (2007). Survival func- a function of redshift of the average SED of 99 ALMA-identified SMGs(Swinbanketal.2014),SEDtemplatesofArp220andM82 tion enables us to derive redshift probability densities in (Silvaetal.1998),respectively. the entire of a redshift range avoiding drastic changes due to upper limits in fluxes. Radio/submm photometric redshifts typically have larger uncertainties than optical/near-IR photometric Extremely red SMGs 5 Figure 2. Multi-wavelengthimagesofASXDF1100.053.1. Topright: RGBimage(R,GandBbeing1100,500and350µm,respectively) around ASXDF1100.053.1. Yellow circles mark AzTEC 1100-µm sources. First and second rows from top: ALMA, AzTEC, SPIRE, PACS and Spitzer images. The black circle marks the AzTEC position of ASXDF1100.053.1 and the beam size of the AzTEC/ASTE image (30′′). The red cross marks the ALMA position of ASXDF1100.053.1. The small cyan circle marks the position of a Spitzer and HerschelbrightsourcenearASXDF1100.053.1. Third and fourth rows from top: JanskyVLA,ALMA,IRAC,UKIRTandSubaruimages ofASXDF1100.053.1. ThebluecirclemarkstheALMApositionofASXDF1100.053.1. redshifts because of the lack of clear SED features, such probability density distributions in order to achieve a as continuum breaks. Since redshift estimates using ra- redshift probability density distribution weightedby the dio/(sub)mm colors depend on the adopted dust SED, likelihood of each SED temperature. weneedtouseSEDsrepresentativeofourtargetpopula- Themulti-variateGaussianprobabilitydistribution,Φ, tion,i.e.galaxieswithsimilarIRluminositiesandsimilar for k colors, is given by redshifts. Here, we adopt the SED template made from of Φ(c c )= i 0 99 ALMA-identified SMGs, derived from deep Her- − 1 schel and ALMA submm and VLA radio data pre- (2π)−k/2 A−1 1/2exp( (c c )′A−1(c c )) Surv, i 0 i 0 sented in Swinbank et al. (2014). The SEDs of ALMA- | | −2 − − ×Y ν identified SMGs were fitted with a library of 185 (1) SEDs from Chary & Elbaz (2001), Rieke et al. (2009), Draine et al.(2007),Ivison et al.(2010)andCarilli et al. where A is a covariance matrix. Here we assume that (2011), adopting optical/near-IR photometric redshifts any non-diagonal elements in the covariance matrix are fromSimpson et al.(2014). Thedusttemperatureofthe zero,therefore (c c )′A−1(c c ) canbe substituted i 0 i 0 best-fit SED of each ALMA-identified SMG is listed in by standard χ2. S−urv is a survi−valfunction (Isobe et al. the paper. We picked SEDs randomly from the parent 1986). The survival function is expressed using an com- SED library along with the dust temperature distribu- plementary error function as tion (19–52K) for the ALMA-identified SMGs derived ipnroSbwaibnibliatynkdeetnsailt.y(2d0i1st4r)ib2u,taionnd fcoarlcuealacthedchtohseenredSsEhDift. Surv = 1 ∞ e−t2/2dt. (2) We bootstrapped this process and combined the derived √2π Z(ci(λ)−cobs(λ))/σobs 2 The reformatted SED templates with dust temper- WeassumethatthefluxdensityerrorsfollowGaussian atures used in Swinbanketal. (2014) are distributed at distributions. The finalredshiftprobabilitydistribution, http://astro.dur.ac.uk/˜ams/HSOdeblend/templates/ P(z), of any galaxy is the sum of the individual proba- 6 Ikarashi et al. Figure 3. Multi-wavelengthimagesofASXDF1100.231.1. bilities from the SEDs, or explicitly SMGs,regardlessoftherarityorotherwiseofcoldSEDs. P(z)=a Φ(c c ), (3) 4.3. Benchmark tests of the redshift estimates i 0 − Xi,∀z It is informative to perform some benchmark tests, using SMGs with known spectroscopic redshifts to as- where a is the normalization constant, such that sess whether our method returns sensible values and 0zmaxP(z)=1 where zmax =10. The asymmetric error to evaluate systematics in our photometric redshift es- bRars (z−,z+) correspond to 68% confidence levels such timates. We have found seven bright or lensed SMGs TthhaetseRzcz−+alPcu(lza)tidozns=fol0l.o6w8 tahnedm(ezt+ho−dozl−o)gyisprmesinenimteidzedin. with10C00O-µmspeacntrdoscraodpiioc rpehdosthoimftsetrwyhiicnh thhaevelitSePraItRuEre, ∼ (Ivison et al. 2010; Ikarashi et al. 2011; Riechers et al. Hughes et al. (2002) and that on the survival function. 2013; Wardlow et al. 2013; Messias et al. 2014). Fig. 5 shows comparisons of their radio/(sub)mm photometric 4.2. Resulting redshift estimates redshifts and their spectroscopic redshifts. All except The radio/submm photometric redshift probability HFLS3 show good agreement. The under-estimation of distributions for ASXDF1100.053.1and231.1are shown the redshift when using the radio/(sub)mm method for with black curves in Fig. 4 along with fit-SEDs. The re- HFLS3 can be explained by its abnormally high dust sultant photometric redshifts for ASXDF1100.053.1 and temperature (56K). As the probability density distribu- 231.1 are 6.5+1.4 and 4.1+0.6, respectively. tion shows (see middle in Fig. 5), there is a small local −1.1 −0.7 Probability densities, Φ, for each SED with T =19, peak aroundthe spectroscopic redshift with a T similar d d 32 and 52K given by Equation 1, are shown at the bot- to that of HFLS3. tom of Fig. 4, with the aim of understanding the con- In addition to benchmark tests with a spectro- tributions of these SEDs to the combined photometric scopic sample, we also conducted another benchmark redshift probability distributions. The Φ density plots test using 46 radio-detected ALMA-identified SMGs indicate that the low probabilities of low-redshift solu- from ALESS with optical/near-IRphotometric redshifts tions in the combined redshift probability distribution, (Simpson et al. 2014; Swinbank et al. 2014). ALESS P(z), are due to two factors: (1) the rarityof coldSEDs sources were originally 880-µm-selected SMGs and are due to the dust temperature distribution, and (2) the expected to be drawn from the same population as cold SEDs give poor fits. The Φ plots also demonstrate ASXDF1100.053.1and231.1. Weestimateradio/submm thatsolutionsforcoldSEDsarelessplausibleforthetwo photometric redshifts using SPIRE 250–500-µm,ALMA Extremely red SMGs 7 Figure 4. Radio/(sub)mmphotometricredshiftofASXDF1100.053.1and231.1. Top: ObservedphotometricdataandmodelSEDs. Red opensquaresmarkphotometricdatausedinourradio/(sub)mmphotometricredshiftestimates: JVLA6-GHz,ALMA1100-µm,SCUBA2 850-µm, and upper limits inthe SPIRE bands. Black open squares markphotometric data not used inour redshiftestimation. The red lineisthe best-fit SED at the best-fit redshiftinphotometric redshiftestimation. The grey-shaded area marks arange of allfit-SEDs at all redshifts. The black line represents the averaged SED of ALMA-identified SMGs at the best-fit redshift presented in §4.4. Middle: Redshiftprobabilitydensitydistributionsofradio/(sub)mmphotometricredshift. Theblackhatched curvemarkstheredshiftprobability density distribution. The grey curve shows that without the SCUBA2850-µm data. The bluelinemarks that of a singleSED template, the average SED of ALMA-identified SMGs. The derived photometric redshift for each estimate is displayed in the panels. The orange hatched area marks a redshift range where the mid-IR color of ASXDF1100.231.1 is explained by the redshifted Hα emission line in the IRAC3.6-µmband,asdiscussedin§6. Bottom: Probabilitydensities(Φ)forindividualSEDsof19,32and52Kusingall(sub)mm/radio bandswhichgivethecontributionsofdifferentTd temperaturetothecombinedphotometricredshift. 880-µm and VLA 1.4-GHz flux densities (see these flux do not suffer strong systematics. densities in Table A1 of Swinbank et al. 2014). A comparison of their radio/submm-estimated photomet- 4.4. Cross-checking photometric redshifts ric redshifts and optical/near-IR estimates is shown in We first derived photometric redshifts using the Fig. 5. We derived ∆z = (zradio zopt )/(1+zopt ). photo− photo photo average SED of 99 ALMA-identified SMGs from Itsmedianand1σ dispersionare 0.01and0.27,respec- Swinbank et al. (2014). This redshift estimate is ex- − tively. Weshouldnotethatthereisnocontaminationbe- pected to give us the most reliable redshift for typical tween optical/near-IR photometric redshifts of . 4 and SMGsbutwillunder-estimatetheuncertaintyduetothe radio/(sub)mm photometric redshifts of &5. plausible diversity of SEDs. The derived redshift prob- These benchmark tests suggest that our ra- ability density distributions for ASXDF1100.053.1 and dio/(sub)mm photometric redshifts using multi-SEDs 231.1,basedontheaverageSED,showresultsconsistent 8 Ikarashi et al. Figure 5. Radio/(sub)mmphotometricredshiftestimation. Left: Comparisonofradio/(sub)mmphotometricredshiftwithspectroscopic redshift obtained via CO for six bright or lensed SMGs from the literature (Ivisonetal. 2010; Ikarashietal. 2011; Riechersetal. 2013; Wardlowetal. 2013; Messiasetal. 2014). Middle: Redshift probability density distribution for HFLS3, shown in order to explain what happens whenestimatingitsredshiftusingradio/(sub)mm photometry. Right: Comparisonofradio/(sub)mm photometricredshiftswith optical/near-IR photometricredshiftsfor46ALMA-identifiedSMGswithradiodetections (Simpsonetal.2014;Swinbanketal.2014). with radio/submm-basedphotometric redshifts (Fig. 4): cal (U)LIRGs, based on observed synchrotron flux den- The respective photometric redshifts based on the aver- sities. They expected a stronger B, > 600µG, based on age SED are 6.7+1.0 and 4.0+0.5. measurements of Zeeman splitting in OH masers. Given −1.0 −0.4 Redshiftprobabilitydensitiesbasedonourredshiftes- that ASXDF1100.053.1 and 231.1 have compact mm- timates without SCUBA-2 850-µm data are shown in wave sizes and surface IR luminosity densities similar to Fig. 4. The 850-µm detection allows a smaller uncer- those of local ULIRGs (see 5 and 6.2), these studies § § tainty and sharpens the redshift probability densities. also support a strong B for our sample. This implies that the model 850-µmflux densities of the Inthispaper,weinvestigatehowtheeffectoftheCMB SMGsbasedonphotometricredshiftsfromALMA1100- on radio emission contributes to radio/mm photometric µm and (J)VLA radio colors and the upper limits at redshift estimates where B = 100 and 300µG: 300µG SPIRE bands are consistent with the observed 850-µm is taken as the value for SMGs, and 100µG is used to flux densities. examine what happens if the magnetic field is weaker. Wedeterminedthepredictedsuppressionofnon-thermal emission by the CMB using the equations and assump- 4.5. Possible effects of the CMB on redshift estimation tions provided in Murphy (2009): we modeled the syn- IntheveryearlyUniverse,thecosmicmicrowaveback- chrotronemissionbysubtractingfree-freeanddustemis- ground(CMB) can have effects on observedsubmm and sion, where we model free-free emission from L based IR radio flux densities (see e.g. Zhang et al. 2016). Here on the equation (16) in the literature. we discuss possible contributions of the CMB to the ra- Fig. 6 shows the resulting redshift probability den- dio/submm photometric redshifts using toy models for sity distribution, including the CMB effects, for both CMB effects. ASXDF1100.053.1 and 231.1: (1) probability density On the basis of the predictions of CMB effects on without any CMB effects; (2) with the CMB effect at observed total submm flux densities (da Cunha et al. (sub)mm wavelengths; (3) with both of the CMB ef- 2013 deals with the effect on total flux densities, and fects (B = 300µG); (4) with both of the CMB effects Zhang et al. 2016 explores the spatially resolved ef- (B = 100µG). We see that taking the effects of the fects), we took into account two CMB effects on ob- CMB into account pushes the photometric redshifts to served submm flux densities: the effect on intrinsic far- higher values. However,as we do not havespectroscopic IR/submmdustSEDs,andonthedetectabilityofSMGs redshifts for the two SMGs, we cannot determine how against the CMB background. We evaluate these effects strong this effect really is. onthe observedflux densities at1100,850,500,350and 5. ALMAMM-WAVESOURCESIZES 250µm for the T of each SED, in the same manner as d da Cunha et al. (2013). The first millimetric size measurements of Observed radio flux densities of distant galaxies are ASXDF1100.053.1 and 231.1 were determined us- expected to get fainter as a function of redshift due to ing our ALMA Cycle-1 data, which covered up to a suppressionofsynchrotronemissionbyinverseCompton uv distance of 400kλ, with a synthesized beam size (IC) losses off the CMB (Murphy 2009). The suppres- of 0′′.70 (FW∼HM – Ikarashi et al. 2015). These ∼ sion off radio flux densities by the CMB depends on the visibility data assumed Gaussian profiles and suggested strength of the magnetic field (B) in a galaxy, about compactmillimetric sizes: 0.28+−00..0044 and0.12+−00..0088 arcsec which we know very little. (FWHM) for ASXDF1100.053.1 and 231.1, respectively. On the basis of observational studies of SMGs in the In this section, we re-assess their millimetric sizes, literature,Murphy(2009)suggestedthatSMGscanhave combining Cycle-1 and -2 data, which now cover up to a strong B, potentially &300µG. McBride et al. (2014) 1500kλ. reporteda minimum B strengthof &150–500µG for lo- In Fig. 7, ALMA maps for ASXDF11100.053.1 and Extremely red SMGs 9 Figure 6. Redshift probability densities of radio/(sub)mm photometric redshift estimates when including the effects of the CMB for ASXDF1100.053.1and231.1. ThegreycurveisaprobabilitydensitydistributionwithoutanyCMBeffects, i.e.itisthesameplotshown in Fig. 4. The red curve shows a probability density distribution with the CMB effect on observed (sub)mm flux densities. The green curveisaprobabilitydensitydistributionwiththeCMBeffectsonboth(sub)mmandradiofluxdensitiesforB=300µG.Thebluecurve is for B =100µG. The redshiftrange is extended to z =15 forASXDF1100.053.1 due to the largeprobability of z ≥10. For the direct comparisonwiththeresultshowninFig.4,theprobabilitydensitiesarescaledinz=0–10. 231.1 are shown. These maps were generated from the 100 independent sets of visibility data generated from combined ALMA 1100-µm data, cleaning down to the the actual ALMA data. In the simulations, we input 1-σ depth in a circle with a radius of 1 arcsec using the a Gaussian model with the same flux as one of the real clean task in CASA. The pixel scale is 0′′.05 pixel−1. sources,thenimagedthesedata,generatingcleancompo- nent maps. We then measure the enclosed flux densities 5.1. Mm-Wave size measurements in visibility data in the same manner as that for the real sources. We re- First,wemeasuremm-wavesizesofASXDF1100.053.1 peatedthisprocesswithsourcesizesbetween0′′.025and and 231.1 with the ALMA visibility data, in the same 0′′.800 in steps of 0′′.025 (FWHM) to reconstruct ob- manner as Ikarashi et al. (2015). We use uv-distance servedenclosed flux densities in eachbin. We adopted a versus amplitude plots (hereafter uv-amp plots) for the fluxdensityerrorinthe simulationwithanenclosedflux measurements (Fig. 8). Modelling sources with uv-amp density closest to a real measured flux density in each plots helps us to avoid underestimating their flux densi- radius bin as the error for the real measurements. We ties, since we can interpolate/extrapolate across incom- refer the readerto A where we describe the simulations § pletevisibilitycoverage. WeassumesymmetricGaussian in more detail. profiles, as is usually done in the literature. Circular- ForASXDF1100.053.1,basedontheobtainedenclosed ized effective radii estimated using uv plots are useful, fluxdensityplotandthetotalfluxdensity,wedetermine evenforsourceswithasymmetricprofiles(Ikarashi et al. Rc,e of 0.17+−00..0021 arcsec. Since the half width half maxi- 2015). Binsizesadoptedinuv-distanceare100kλoutto mum of a symmetric Gaussian is equivalent to Rc,e, the 500kλ and 500kλ between 500–1500kλ. The estimated sizeobtainedfromthecleancomponentmapisconsistent sizes of ASXDF1100.053.1 and 231.1 are then 0′′.33 and with that from the uv-amp plot. 0′′.15 (FWHM), respectively (Fig. 8). Correcting these For ASXDF1100.231.1, the flux density in the cen- ‘raw’ mm-wave sizes for systematic effects using Monte ter pixel is 1.19+0.27mJybeam−1. This corresponds to −0.24 Carlosimulations,themm-wavesizesarethen0′′.34+−00..0022 52+−76% of its total flux density. From the obtained en- and 0′′.18+0.04 for ASXDF1100.053.1 and 231.1, respec- closedfluxdensityplot,withlinearinterpolation,wefind −0.04 tively, consistent with our previous measurements. R = 0.025+0.015 arcsec, meaning the half light radius c,e −0.00 of ASXDF1100.231.1 is 0′′.04. R determined via 5.2. Mm-Wave size measurements in clean component ≤ c,e the clean component map is approximately 2 smaller maps × than that determined from the uv-ampplot, R =0.09 c,e Next, we derive R for ASXDF1100.053.1 and 231.1 arcsec. c,e usingALMA cleancomponentmaps, asshownin Fig.7. The enclosed flux density plot suggests that These maps were generated from the combined ALMA ASXDF1100.231.1 cannot be modeled with single 1100-µmdataby runningthe cleantaskinCASA.Our Gaussian profile: ASXDF1100.231.1 appears to com- motivation is to measure R directly, without any as- prise a compact, intense mm emitting region, located c,e sumed model, exploiting the high signal-to-noise ratios in its center region, and a fainter, extended region. of 30. The different R values determined using the uv-amp c,e F∼ig. 9 shows enclosed flux densities as a function of plot and the clean component map can thereby be radiusforASXDF1100.053.1and231.1,measuredinthe understood. clean component maps. Total flux densities of the two Monte Carlo simulations of source size measurements SMGs in the clean component maps are consistent with in ALMA clean component maps are described in Ap- the fluxes measured in the beam-convolved ALMA con- pendix A. According to these simulations, this method tinuum images listed in Table2, despite the potential of measuring source sizes is useful down to R = c,e absence of any < 1σ components. Flux density errors 0′′.025. The simulations show that the measured sizes are estimated based on Monte Carlo simulations using of ASXDF1100.053.1 and 231.1 are not expected to suf- 10 Ikarashi et al. Figure 7. ALMAimagesofASXDF1100.053.1and231.1. 1stcolumn: CleanedALMA1100-µmcontinuumimagestakeninALMACycles 1and2. Synthesizedbeamsofthecombineddataare0′′.46×0′′.35(PA,69◦)and0′′.57×0′′.48(PA,82◦)forASXDF1100.053.1and231.1, respectively. Contoursareshownat5,10,15,20and25σ. ThefluxdensityunitismJybeam−1. Ther.m.s.noiselevelisshownatbottom ineachpanel. 2nd column: Cleancomponent maps ofthecombined ALMAimagesareshowninthemiddlepanel. Thecleancomponent mapswereobtainedbycleaningdownto1σ. 3rdcolumn: SynthesizedbeamsforthecombinedALMAimages. 4thcolumn: Residualmaps aftersubtractingthecleancomponents convolvedwiththesynthesizedbeams. Thepixelscaleis0′′.05inallimages. Figure 8. ALMAuv-distanceversusamplitudeplotsofASXDF1100.053.1and231.1. Blacksolidpointsaretheobserveddata. Binning sizes inuv-distance are100kλout to 500kλand 500kλ between 500–1500kλ. The black lineisauv-amp model of the best-fit Gaussian component. Thebluelineandshadedareaarepossiblesolutionsforthecorrectedsourcesize,witherrorsbasedonMonteCarlosimulations. fer large systematic errors. adoptedaChabrierIMF(Chabrier2003)andstellarpop- ulation synthesis models by Bruzual & Charlot (2003). 6. ONTHENATUREOFASXDF1100.053.1AND231.1 Dustextinctionwasconsideredaccordingtotheprescrip- HerewedeterminethepropertiesofASXDF1100.053.1 tion by Calzetti et al. (2000). We adopted metallicity, and 231.1 from multi-wavelength data, adopting the ra- Z⊙ = 0.02. We performed the analysis using the code, dio/submmphotometric redshifts. We discuss the possi- Le Phare (Arnouts et al.1999;Ilbert et al. 2006). These bleroleofthetwoSMGs,whicharefaintintheHerschel SMGs are detected in only three or four filters in the bands, and at optical/near-/mid-IR and radio wave- available optical/mid-IR broad band images, similar to lengths, in the context of galaxy evolution. the extremely red mm source analyzed by Caputi et al. (2014). We derived optical/near-IR properties at fixed 6.1. Optical/near-IR SED fitting redshifts of z = 5.5, 6.5 and 7.5 for ASXDF1100.053.1, andz =3.5,4.5and5.5forASXDF1100.231.1,toreduce Inordertocharacterizethe optical/near-IRproperties parameter space. These values span a range of approx- of ASXDF1100.053.1 and 231.1, we conducted an SED- imately 1 around the best radio/submm photometric fitting analysis across optical–mid-IR wavelengths. We ±

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.