Table Of Content

DRAFTVERSIONFEBRUARY8,2016 PreprinttypesetusingLATEXstyleemulateapjv.5/2/11 THEGASMASSOFSTAR-FORMINGGALAXIESATZ ≈1.3 NISSIMKANEKAR1,SHIVSETHI2,K.S.DWARAKANATH2 DraftversionFebruary8,2016 ABSTRACT WereportaGiantMetrewaveRadioTelescope(GMRT)searchforHI 21cmemissionfromalargesample 6 of star-forminggalaxies at z ≈ 1.18−1.34, lying in sub-fields of the DEEP2 Redshift Survey. The search 1 was carriedoutby co-adding(“stacking”)the HI 21cm emission spectra of 857galaxies, after shifting each 0 galaxy’sHI 21cmspectrumtoits restframe. We obtainthe 3σ upperlimitSH < 2.5µJyon theaverageHI 2 21cmfluxdensityofthe857galaxies,atavelocityresolutionof≈315kms−1.I Thisyieldsthe3σconstraint b MHI <2.1×1010×[∆V/315km/s]1/2M⊙ontheaverageHImassofthe857stackedgalaxies,thefirstdirect e constraint on the atomic gas mass of galaxies at z > 1. The implied limit on the average atomic gas mass F fraction(relativetostars)isMGAS/M∗ < 0.5,comparabletothecoldmoleculargasmassfractioninsimilar star-forminggalaxiesat these redshifts. We find thatthe cosmologicalmassdensity of neutralatomicgasin 5 massivestar-forminggalaxiesatz ≈ 1.3isΩGAS < 3.7×10−4,significantlylowerthanΩGAS estimatesin both galaxies in the local Universe and damped Lyman-α absorbers at z ≥ 2.2. Massive blue star-forming ] A galaxiesthusdonotappeartodominatetheneutralatomicgascontentoftheUniverseatz ≈1.3. G Subjectheadings:galaxies:high-redshift—galaxies:ISM—galaxies:starformation—radiolines:galaxies . h p 1. INTRODUCTION of the molecular component, studies of CO in emission at - Overthelastdecade,opticalandinfraredstudiesoftheso- z ≈ 1.5−3 have indeed found evidence for massive reser- ro called“deepfields”(e.g.Giavaliscoetal.2004;Scovilleetal. vstoeilrlsaromf masosl)eicnulsatarrg-faosrm(MinHg2g&ala1x0ie1s0M(e⊙.g,.cDoamdpdaireatballe.t2o01th0e; t 2007; Newmanetal. 2013) have revolutionized our under- s Tacconietal.2010,2013). Unfortunately,COisonlyatracer standing of galaxy formation and evolution. Detailed stud- a of the bulk of the molecular gas, and the conversion factor [ ies of the cosmic star formationhistory have shown that the comovingstarformationrate(SFR)densityrisestowardsthe fromtheCOlineluminositytomoleculargasmassisknown 2 end of epoch of reionization and peaks in the redshift range todependongalaxytype,varyingbyafactorof≈5between v z ≈1−3,beforedecliningbyanorderofmagnitudefromz ≈ starburst and spiral galaxies, and by larger factors in low- 9 metallicitygalaxies(e.g.Genzeletal.2012;Carilli&Walter 1 to today (e.g. LeFloc’hetal. 2005; Hopkins&Beacom 0 2013). 2006;Bouwensetal.2009).Indeed,therangez ≈1−3isof- 9 Overall,thesituationtodayisthatlittleisknownaboutthe tenreferredtoasthe“epochofgalaxyassembly”,asroughly 7 neutralgas,especiallytheatomiccomponent,inhigh-z star- half of today’s stellar mass density was formed during this 0 forminggalaxies. Unfortunately,evenverydeepintegrations period (e.g. Reddyetal. 2008; Marchesinietal. 2009). It is 1. alsoknownthatthenatureofthegalaxiesundergoingstarfor- with today’s best radio telescopes will only be able to de- 0 mationevolvessignificantlyfromz ≈ 2totoday,viacosmic tectHI 21cmemissionfromeventhemostmassivegalaxies 6 downsizing(Cowieetal.1996): theSFRdensityatz ∼ 2is (MHI & few×1010M⊙)outtorelativelylowredshifts,z . 1 0.5(e.g.Verheijenetal.2007;Fernańdezetal.2013);detect- dominatedbymassivegalaxieswithhighSFRs,whilethatat v: z ≈ 0mostlyarisesinlower-masssystemswithlowerSFRs ingHI21cmemissionfromindividualgalaxiesatz &1will needthelargecollectingareaoftheSquareKilometreArray. i (e.g.LeFloc’hetal. 2005;Seymouretal. 2008). A tightre- X However,informationcanbeobtainedabouttheaveragegas lation(the“mainsequence”)hasbeenfoundbetweenthespe- properties of high-z galaxies by co-adding (“stacking”) the r cific SFR (sSFR) and the stellar mass of most high-z star- a forminggalaxies(Noeskeetal.2007),whileasmallfraction, HI21cmemissionsignalsofalargenumberofgalaxieswith known redshifts that lie within the primary beam of the ra- the starbursts, undergofarmoreefficientstar formationthan diotelescopebutthataretoofainttobedetectedindividually main-sequencegalaxiesofthesamestellarmass,withsSFRs largerbyanorderofmagnitude(e.g.Rodighieroetal.2011). (Zwaan2000;Chengaluretal.2001).SuchHI21cmstacking Theaboveremarkablerecentprogressinstudiesofstarsin canbeusedtoinfertheaverageHI massofthegalaxiesand thecosmologicalmassdensityinneutralgasinthesesystems high-z galaxieshasnotbeenmirroredinstudiesoftheother (e.g.Lahetal.2007;Changetal.2010;Delhaizeetal.2013; majorconstituentofgalaxies,theneutralgas,thefuelforthe Rheeetal. 2013). One can also determine average proper- star formation. Inthe localUniverse, thisgasisbest probed ties of the galaxy sample, including the dependence of the with emission studies in the HI 21cm line. Unfortunately, gasmass on variousattributes (e.g. SFR, stellar mass, envi- the weakness of the HI 21cm transition has meant that the ronment,andredshift),theaverageTully-Fisherrelation,etc highestredshiftatwhichit hasso far beendetectedin emis- (e.g.Fabelloetal.2011,2012;Meyeretal.2016). sionisz ≈ 0.25(e.g.Catinella&Cortese2015). Inthecase Stacking of the HI 21cm spectra of individual galaxies 1Swarnajayanti Fellow; NationalCentreforRadioAstrophysics,Tata with known redshifts, obtained using radio interferometry, Institute ofFundamental Research, Ganeshkhind, Pune-411007, India; has so far only been carried out at low redshifts, z . 0.4 [email protected] (e.g.Lahetal. 2007, 2009; Rheeetal. 2013). Thishasbeen 2DepartmentofAstronomyandAstrophysics, RamanResearchInsti- extended to slightly higher redshifts, z ≈ 0.8, via cross- tute,C.VRamanAvenue,Bengaluru,India 2 Kanekaretal. correlationofsingle-dishHI 21cmintensitymapswithdeep opticalimages(Changetal.2010;Masuietal.2013). Inthis Letter,wereportGiantMetrewaveRadioTelescope(GMRT) resultsfromthefirstHI21cmstackingexperimentatz > 1, obtaining tight constraints on the average gas mass of star- forminggalaxiesatz ≈1.3. 2. THEDEEP2SURVEYFIELDS The stacking approach critically requires a large sample of galaxies with accurately known redshifts in the target field. This is a significant restriction when selecting fields forsearchesforHI 21cmemissionatz ≈ 1−1.5,asthisis the “redshiftdesert” wherethe brightopticallines(e.g. Hα, OII, Hβ) that are typically used to identify galaxiesare redshifted to wavelengths& 7500A˚, into regionsfullof night- sky lines. Most spectroscopic galaxy samples in deep fields arehencedominatedbygalaxiesatz .0.7,oratz &2. The DEEP2GalaxyRedshiftSurveyisuniqueinthisregardasits redshiftcoverage extendsout to z ≈ 1.4 (Davisetal. 2003; Newmanetal.2013), duetotheuseofthehighspectralres- olution (R = 6000) mode of the DEIMOS spectrographon theKeck-IITelescope,allowingtheclearidentificationofthe OIIλ3727 doubletout to z ≈ 1.4. The DEEP2 Survey cov- FIG.1.—ThespectralRMSnoisevaluesforthe868targetgalaxies,after ers four fields with a total areal coverage of 2.8 deg.2, and correctingfortheirlocationwithintheGMRTprimarybeam. Theabscissa isanarbitrarysourcenumber; eachfieldislabeledbytheGMRTpointing hasyieldedaccurateredshiftsfor38,000massivegalaxiesin center(DEEP2sub-fields21,22,31/32,32/33)andthenumberofgalaxies theredshiftrange0.75 . z . 1.4(Newmanetal. 2013). A includedintheGMRTspectralcoverage. few thousand galaxies lie at z ≈ 1.2 − 1.4, for which the HI 21cm line is redshifted to ≈ 590−650 MHz, i.e. into totalon-sourcetimes on each field rangedfrom 500minutes theGMRT610MHzband.ThispicksouttheDEEP2Survey (field32)to804minutes(field22). fields as the ideal targets for a GMRT search for HI 21cm The data were analysed in “classic” AIPS, using standard emission at z ≈ 1.3. Three of the DEEP2 fields are of procedures. For each field, after initial data editing and cal- size 120′ × 30′, each consisting of three sub-fields of size ibration of the antenna bandpasses and gains, the visibility 36′×30′,whilethefourth,theExtendedGrothStrip(EGS), datawereaveragedtoproduceadatasetofspectralresolution is of size 120′ × 15′. The size of the sub-fields of the first ≈ 1 MHz. Accurate antenna-basedgains were then derived threeDEEP2fieldsarewell-matchedtothesizeoftheGMRT via a self-calibration procedure, first iterating between 3-D primarybeamat610MHz,whichhasafullwidthathalfmax- imagingandphase-onlyself-calibrationforafewrounds,fol- imum(FWHM)of≈ 43′. Hence, althoughthe EGShasex- lowedbyamplitude-and-phaseself-calibration,3-Dimaging, cellent ancillary multi-wavelength information, we chose to anddataediting.Theprocedurewasrepeateduntilthecontin- focuson the DEEP2sub-fields21, 22, 31/32and32/33(see uumimagedidnotimproveonadditionalself-calibration;the Newmanetal.2013),allwithfullspectroscopiccoverage,for continuumroot-mean-square(RMS)noisenearthecenterof thepresentGMRTpilotsurvey. theprimarybeamis≈23−35µJyinthreeoftheGMRTpoint- ings,but≈100µJyforsub-field22,probablyduetodynamic 3. OBSERVATIONSANDDATAANALYSIS rangeissuesduetobrightsourcesatorbelowthehalf-power We used the GMRT 610 MHz receivers to ob- pointoftheprimarybeam. serve the four DEEP2 sub-fields in 2011 Novem- For each field, the task UVSUB was then used to sub- ber (sub-fields 21 and 22, with pointing cen- tractthe finalimagefromthe calibratedspectral-linevisibil- ters of RA=16h48m00.0s, Dec=+34d56’00.0” and ities, after which any residual continuumwas subtractedvia RA=16h51m00.0s, Dec=+34d56’00.0”, respectively) a linear fit to each visibility spectrumusing the task UVLIN. and 2012 October and November (pointing midway be- The residualvisibilities were then shifted to the heliocentric tween sub-fields 31 and 32, and 32 and 33, with pointing frame, using the task CVEL. Polyhedral imaging was then centers of RA=23h28m00.0s, Dec=+00d09’00.0” and used, with natural weighting, to produce wide-field spectral RA=23h32m00.0s, Dec=+00d09’00.0”, respectively). A cubes, extending beyond the FWHM of the primary beam. bandwidthof33.33MHz,sub-dividedinto512channels,was Theangularresolutionofthefinalspectralcubesrangesfrom used for all observations, with 30 antennas, and the GMRT ≈ 9.7′′×7.8′′ to ≈ 7.2′′ ×5.5′′, correspondingto a spatial SoftwareBackendasthecorrelator. Thefrequencycoverage sizeof&60kpc×50kpcatz =1.33. was 601 − 634.33 MHz in 2011 and 621 − 654.33 MHz in 2012, implying a redshift coverage of ≈ 1.24 − 1.36 4. STACKINGTHEHI21CMSPECTRA and 1.18 − 1.28, respectively, with a velocity resolution of WestackedtheHI21cmspectraofallDEEP2galaxiesly- ≈ 32 km s−1. The different frequency settings were used ingwithintheFWHM oftheGMRTprimarybeamandwith to testthe possibilitythatradiofrequencyinterference(RFI) accuratelyknownredshifts(redshiftquality≥3intheDEEP2 might restrict the sensitivity in different parts of the GMRT 610 MHz band. Observations of 3C286 and 3C48 were 3 We assume a standard Λ-cold-dark-matter (ΛCDM) cosmology, with usedtocalibratethefluxdensityscale,andof0022+002and H0 = 67.8 km s−1 Mpc−1, Ωm = 0.304 and ΩΛ = 0.696 1609+266tocalibratetheantennagainsandbandpasses.The (PlanckCollaboration2015). Gasmassofstar-forminggalaxiesatz ≈1.3 3 FIG.2.—ThefinalGMRTHI21cmspectra,afterstackingtheHI21cmspectraof857DEEP2galaxies,at[A]theoriginalvelocityresolutionof≈32kms−1 and[B]avelocityresolutionof≈315kms−1,aftersmoothingandresampling.Noevidenceforastatisticallysignificantemissionfeatureisapparentineither spectrum. catalog;Newmanetal.2013)thatshifttheHI21cmlineinto in the stacked spectrum after smoothing to, and resampling theobservedbandforeachtargetfield. Weinitiallyexcluded at,arangeofvelocityresolutionsfrom≈100−500kms−1; 20channelsfromeachedgeoftheobservingband,wherethe nostatisticallysignificantfeaturewasdetectedatanyresolu- sensitivityissignificantlylowerthaninthebandcenter.Next, tion. Fig. 2[B] showsthespectrumsmoothedto, andresam- inordertohavesufficientspectralbaselinearoundanyputa- pledat,aresolutionof≈315kms−1;thishasanRMSnoise tivestackedspectralfeature,weonlyconsideredgalaxieswith of≈2.1µJyandshowsnoevidenceforHI21cmemission. redshifts≥ 1500kms−1 fromthe edgesofthe retainedfre- The3σupperlimitonthestrengthofHI21cmemissionin quency range. This yielded 206− 247 galaxies in the four thefinalspectrumataresolutionof≈315kms−1canbeused GMRTpointings,withatotalof868galaxies.HI21cmspec- to derive a limit on the average HI mass of the 857 stacked trawereobtainedbytakingacutthroughthedifferentspectral galaxies. However, our estimate of the RMS noise on this cubes at the location of each galaxy. A second-order base- spectrumisbasedononly10points. We henceuseda boot- line was fit to each spectrum and subtracted out to remove strapapproachtoobtainamorereliableestimateoftheRMS anyresidualcurvature,andeachspectrumwasthenscaledto noiseonthestackedspectrum,bystackingthe857spectraaf- accountforthe locationofthe galaxywithinthe GMRT pri- ter randomizing the galaxy redshifts; this was done for 100 marybeam.ThedistributionofspectralRMSnoisevaluesper independentrealizations and yielded an average RMS noise ≈32kms−1channelforthedifferentspectra,aftertheabove of≈2.5µJyper≈ 315kms−1 channel. Thisestimateofthe primarybeamcorrection,isshownin Fig.1; theRMS noise RMSnoisewillbeusedinthelateranalysis. atthecenterofeachfieldis≈240−260µJyper≈32kms−1 channel. 5. DISCUSSIONANDSUMMARY Each spectrum was tested for gaussianity, via both the The DEEP2 galaxy sample is selected at R-band, with a Kolmogorov-Smirnov rank-1 (K-S) and Anderson-Darling limitingR-magnitudeof24.1(Newmanetal. 2013). Forthe (A-D)tests;11spectrafoundtohavep-values<0.003inthe A-Dtestwereexcludedfromthestacking(notethatretaining medianredshiftofzmed =1.265ofour857galaxies,thiscor- these spectra does not significantly affect our results). The respondstoselectionatarest-framewavelengthof≈2950A˚. HI 21cmspectrumforeachretainedgalaxywasnextshifted As pointed out by Weineretal. (2009), the high-z DEEP2 to its rest frame and then interpolated onto a single velocity sample is hence highlybiased towardsstar-forminggalaxies axis,beforethespectraofeachfieldwereoptimallyco-added inthebluecloudofthegalaxycolor-magnitudediagram(e.g. with weightsdeterminedfromtheir RMSnoise values(after Willmeretal. 2006). Our target galaxies are thus expected theprimarybeamcorrection;seeFig.1)toproducethefinal topredominantlybebluestar-forminggalaxies. Cooperetal. stackedspectrumforeachfield. Thesefourspectrawerethen (2006) find that< 2%of DEEP2galaxiesatall redshiftslie co-added,againwithappropriateweightsbasedontheirRMS in clusters, with ≈ 67% in the field and ≈ 30% in small noisevalues,andasecond-orderbaselinesubtractedoutfrom groups; the fraction of cluster galaxies is even smaller for the result, to produce the final stacked HI 21cm spectrum. blue, star-forming galaxies. Ram-pressure stripping of the ThisisshowninFig.2[A]andhasanRMSnoiseof≈8.7µJy HIinourtargetgalaxies(asobservedinnearbyclusters,e.g. per≈32kms−1. Giovanelli&Haynes1983;Verheijenetal.2007)isthusun- The signal-to-noise ratio in a search for line emission is likelytobeaseriousissue. maximized on smoothing the spectrum to a resolution equal For the assumed ΛCDM cosmology and the median red- tothelineFWHM.WehencesearchedforHI21cmemission shiftzmed =1.265ofourstackedgalaxies,ournon-detection ofHI21cmemissioninthestackedspectrumofFig.2[B]im- 4 Kanekaretal. line frequency for z = 1.265) is 43.2′, implying a comov- ingareaof50.3Mpc×50.3Mpc. Ourstackinganalysisfirst excluded20 edge channels, and then galaxies with redshifts ≤ 1500 km s−1 from the retained channels; this implies an effectivebandwidthof≈23MHzcenteredat627.1MHz,i.e. a comovingline-of-sightdistance of 177.8 Mpc. The effec- tive comoving volume of a single GMRT pointing is hence V ≈ 3.54 × 105 Mpc3. The upper limit to the cosmological mass density of HI in star-forming galaxies is then ΩHI,SF = MHI × N/V, where N = 851/4 is the average number of DEEP2 galaxies in a single GMRT pointing. This yields ΩHI,SF < 2.8 × 10−4. To obtain the cosmological mass density of neutral gas in star-forming galaxies, ΩGAS,SF, we include a factor of 1.32 for helium to obtain ΩGAS,SF <3.7×10−4. Fig. 3 plots ΩGAS versus redshift from a variety of mea- surements at different redshifts. The figure shows that our 3σ upperlimittoΩGAS,SF issignificantlylower(byafactor & 1.5)thanthemeasuredΩGAS atbothz ≈ 0andz ≥ 2.2. It thus appears that blue star-forming galaxies do not domi- natethecosmologicalgasmassdensityinneutralatomicgas atz ≈1.3. FIG.3.— The cosmological mass density in neutral gas ΩGAS plotted WenotethatitispossiblethatthemostHI-richgalaxiesat as a function of redshift. The filled triangles are from HI 21cm emis- a given redshift do not dominate the star formation activity. sionsurveys(Zwaanetal.2005a;Martinetal.2010),theopensquaresfrom HI21cmemissionstackingatlowredshifts(Lahetal.2007;Delhaizeetal. Forexample,the“HImonsters”ofLeeetal.(2014),withHI 2013;Rheeetal.2013),andthefilledcirclesfromabsorption-selectedgalax- masses > 3×1010 M⊙ at z ≈ 0.04−0.08, have SFRs ≈ aoiebpsse,onMrbsgetIarIsraasbhtsoozwrbs≥erosu2ar.t2zne(≈Nwo0tGe.r7Md−aReT1m.e2ree(sRtualatlo,.ef2ot0ra1l2b.;l2u0Ce0r6isgt)ahartn-ofdnordematmianlp.ge2dg0aL1l5yax)m.ieasTn-haαet t0h.a5t−Ω5GMAS⊙a/ytrz. O≈ur1r.e3siusltdsothmuisnadtoednobtyrumleasosuivteth,elopwo-sssuibrfilaictye z ≈ 1.265; itisclear thatthisissignificantly lowerthantheΩGAS esti- brightnessgalaxieswithlowSFRs. matesatbothz=0andz≥2.2[seealsoNeelemanetal.(2016)]. In summary, we have used the GMRT to carry out a polnietshteheav3eσraugpepeHrIlimmaitssMoHfIt<he28.15×7 1st0a1c0kMed⊙g×al[∆axVie/s,31w5h]1e/r2e sHeIar2c1hcfmoreHmIis2si1ocnmsigemnailsssifornomat8z57≈sta1r.-2f6o5rmbiyngstgaaclkaixnigesthine we have assumed a Gaussian line shape with FWHM = four DEEP2 survey fields. Our non-detection of HI 21cm 315kms−1. NotethatthelargespatialextentoftheGMRT emissioninthestackedHI 21cmspectrumyieldsthe3σ up- synthesized beam (> 60 kpc ×50 kpc at z ≈ 1.3) implies per limit MHI < 2.1 × 1010 × [∆V/315km/s]M⊙ on the thatitisveryunlikelythatweareresolvingouttheHI21cm average HI mass of galaxies in our sample. Comparing this to the averagestellar mass of DEEP2 galaxiesyields the 3σ emission. Thisisthefirstdirectconstraintontheatomicgas massofstar-forminggalaxiesatz >1. upper limit MGAS/M∗ < 0.5 on the atomic gas mass frac- tionrelativetostars. Thisissimilartothemeasuredaverage Stellar mass estimates are available for galaxies in two cold molecular gas mass fraction in star-forminggalaxies at of our target fields (22 and 32) from K-band imaging (Bundyetal. 2006), with an average stellar mass of M∗ ≈ z ≈1.2,suggestingthattheneutralgasinthesegalaxiesmay 5th.1e×sam10e1a0vMer⊙a.geAsstseulmlarinmgathssatthoeunrogtivheesrtaanrgaettomgailcaxgiaessmhaavses bΩeGmASo,sStFly<in3th.7e×mo1l0e−cu4laornpthhaesceo.sFminoallolyg,icwaelmobatsasindethnesiltiymoift neutralgas in star-forminggalaxiesat z ≈ 1.3. This is sig- fraction(relativetostars,andafterincludingafactorof1.32 to account for the contribution of helium) of MGAS/M∗ < znifi≥ca2n.t2ly, ilnowdiecrattihnagnthesattimmaatsessivoefbΩluGeAsStaart-fboormthinzg=gal0axaineds 0.5, at 3σ significance. Note that Tacconietal. (2013) used do not dominate the gas content of the Universe during the COstudiesofstar-forminggalaxiesatz ≈ 1.2intheEGSto epochofgalaxyassembly. infer an average cold molecular gasfraction relative to stars of MMol/M∗ ≈ 0.33, comparable to our upper limit on the atomic gas fraction. Our results thus indicate that neutral gas in massive star-forming galaxies at z ≈ 1.3 is likely to be mostly in the molecular phase (e.g. Tacconietal. 2013; We thankKate Rubin, Ben Weiner, Xavier Prochaska and Lagosetal.2014). ClaudiaLagosfordiscussions, andKevinBundyforprovid- Since the DEEP2 galaxies were selected uniformly in R- ingthestellarmassestimates. WealsothanktheGMRTstaff band as a magnitude-limited sample (Newmanetal. 2013), who have made these observations possible. The GMRT is our non-detectionof HI 21cm emission can be used to con- runbytheNationalCentreforRadioAstrophysicsoftheTata strainthecosmologicaldensityofneutralgasinstar-forming Institute of Fundamental Research. NK acknowledges sup- galaxies at z ≈ 1.265. The FWHM of the GMRT primary port from the Department of Science and Technology via a beamat627.1MHz(correspondingtotheredshiftedHI21cm SwarnajayantiFellowship(DST/SJF/PSA-01/2012-13). REFERENCES Bouwens,R.J.,Illingworth,G.D.,Franx,M.,etal.2009,ApJ,705,936 Carilli,C.L.&Walter,F.2013,ARA&A,51,106 Bundy,K.,Ellis,R.S.,Conselice,C.J.,etal.2006,ApJ,651,120 Catinella,B.&Cortese,L.2015,MNRAS,446,3526 Gasmassofstar-forminggalaxiesatz ≈1.3 5 Chang,T.-C.,Pen,U.-L.,Bandura,K.,&Peterson,J.B.2010,Nature,466, Martin,A.M.,Papastergis,E.,Giovanelli,R.,etal.2010,ApJ,723,1359 463 Masui,K.W.,Switzer,E.R.,Banavar,N.,etal.2013,ApJ,763,L20 Chengalur,J.N.,Braun,R.,&Wieringa,M.2001,A&A,372,768 Meyer,S.A.,Meyer,M.,Obreschkow,D.,&Staveley-Smith,L.2016, Cooper,M.C.,Newman,J.A.,Croton,D.J.,etal.2006,MNRAS,370,198 MNRAS,455,3136 Cowie,L.L.,Songaila,A.,Hu,E.M.,&Cohen,J.G.1996,AJ,112,839 Neeleman,M.,Prochaska,J.X.,Ribaudo,J.,etal.2016,MNRAS,inpress, Crighton,N.H.M.,Murphy,M.T.,Prochaska,J.X.,etal.2015,MNRAS, (arXiv:1601.01691) 452,217 Newman,J.A.,Cooper,M.C.,Davis,M.,etal.2013,ApJS,208,5 Daddi,E.,Bournaud,F.,Walter,F.,etal.2010,ApJ,713,686 Noeske,K.G.,Weiner,B.J.,Faber,S.M.,etal.2007,ApJ,660,L43 Davis,M.,Faber,S.M.,Newman,J.A.,etal.2003,Proc.SPIE4834,161 Noterdaeme,P.,Laursen,P.,Petitjean,P.,etal.2012,A&A,540,63 Delhaize,J.,Meyer,M.J.,Staveley-Smith,L.,&Boyle,B.J.2013, PlanckCollaboration2015,A&A,submitted(arXiv:astro-ph/1502.01589) MNRAS,433,1398 Rao,S.M.,Turnshek,D.A.,&Nestor,D.B.2006,ApJ,636,610 Fabello,S.,Catinella,B.,Giovanelli,R.,etal.2011,MNRAS,411,993 Reddy,N.A.,Steidel,C.C.,Pettini,M.,etal.2008,ApJS,175,48 Fabello,S.,Kauffmann,G.,Catinella,B.,etal.2012,MNRAS,427,2841 Rhee,J.,Zwaan,M.A.,Briggs,F.H.,etal.2013,MNRAS,435,2693 Fernańdez,X.,vanGorkom,J.H.,Hess,K.M.,etal.2013,ApJL,770,L29 Rodighiero,G.,Daddi,E.,Baronchelli,I.,etal.2011,ApJ,739,L40 Genzel,R.,Tacconi,L.J.,Combes,F.,etal.2012,ApJ,746,69 Scoville,N.,Aussel,H.,Brusa,M.,etal.2007,ApJS,172,1 Giavalisco,M.,FergusonH.C.,KoekemoerA.M.,etal.2004,ApJ,600, Seymour,N.,Dwelly,T.,Moss,D.,etal.2008,MNRAS,386,1695 L93 Tacconi,L.J.,Genzel,R.,Combes,F.,etal.2010,Nature,463,786 Giovanelli,R.&Haynes,M.P.1983,AJ,88,881 Tacconi,L.J.,Neri,R.,Genzel,R.,etal.2013,ApJ,768,74 Hopkins,A.M.&Beacom,J.F.2006,ApJ,651,142 Verheijen,M.,vanGorkom,J.H.,Szomoru,A.,etal.2007,ApJ,668,L9 Lagos,C.D.P.,Baugh,C.M.,Zwaan,M.A.,etal.2014,MNRAS,440,920 Weiner,B.J.,Coil,A.L.,Prochaska,J.X.,etal.2009,ApJ,692,187 Lah,P.,Chengalur,J.N.,Briggs,F.H.,etal.2007,MNRAS,376,1357 Willmer,C.N.A.,Faber,S.M.,Koo,D.C.,etal.2006,ApJ,647,853 Lah,P.,Pracy,M.B.,Chengalur,J.N.,etal.2009,MNRAS,399,1447 Zwaan,M.A.2000,PhDthesis,Ph.D.Thesis,Groningen:Rijksuniversiteit, LeFloc’h,E.,Papovich,C.,Dole,H.,etal.2005,ApJ,632,169 2000 Lee,C.,Chung,A.,Yun,M.S.,etal.2014,MNRAS,441,1363 Zwaan,M.A.,Meyer,M.J.,Staveley-Smith,L.,&Webster,R.L.2005a, Marchesini,D.,vanDokkum,P.G.,Fo¨rsterSchreiber,N.M.,etal.2009, MNRAS,359,L30 ApJ,701,1765