Mon.Not.R.Astron.Soc.000,1–9(2011) Printed6January2011 (MNLATEXstylefilev2.2) Galaxy Environments in the UKIDSS Ultra Deep Survey (UDS) R. W. Chuter1⋆, O. Almaini1, W. G. Hartley1, R. J. McLure2, J. S. Dunlop2, S. Foucaud3, C. J. Conselice1, C. Simpson4, M. Cirasuolo2 and E. J. Bradshaw1 1UniversityofNottingham,SchoolofPhysics&Astronomy,Nottingham,NG72RD 2InstituteforAstronomy,UniversityofEdinburgh,RoyalObservatory,EdinburghEH93HJ 3DepartmentofEarthSciences,NationalTaiwanNormalUniversity,No.88,Section4,TingzhouRoad,WenshanDistrict,Taipei11677,Taiwan 1 4AstrophysicsResearchInstitute,LiverpoolJohnMooresUniversity,TwelveQuaysHouse,EgertonWharf,BirkenheadCH411LD 1 0 2 n Accepted2010December20.Received2010December8;inoriginalform2010August4 a J 4 ABSTRACT Wepresentastudyofgalaxyenvironmentstoz∼2,basedonasampleofover33,000K-band ] selectedgalaxiesdetectedintheUKIDSSUltraDeepSurvey(UDS).Thecombinationofin- O frareddepthandareaintheUDSallowsustoextendpreviousstudiesofgalaxyenvironment C to z > 1 without the strong biases associated with optical galaxy selection. We study the . environmentsofgalaxiesdividedbyrestframe(U −B)colours,inadditionto‘passive’and h ‘star-forming’subsetsbasedontemplatefitting.Wefindthatgalaxycolourisstronglycorre- p latedwithgalaxyoverdensityonsmallscales(< 1Mpcdiameter),withred/passivegalaxies - o residinginsignificantlydenserenvironmentsthanblue/star-forminggalaxiestoz ∼ 1.5.On r smaller scales(< 0.5Mpc diameter)we also finda relationshipbetweengalaxyluminosity st and environment, with the most luminous blue galaxies at z ∼ 1 inhabiting environments a comparable to red, passive systems at the same redshift. Monte Carlo simulations demon- [ stratethattheseconclusionsarerobusttotheuncertaintiesintroducedbyphotometricredshift 1 errors. v Keywords: GalaxyEvolution,Environment,DeepSurveys–Infrared:Galaxies. 9 4 8 0 . 1 1 INTRODUCTION Hα (Baloghetal. 2004) found that it’s strength does not depend 0 on environment but that the fraction of galaxies with equivalent Ithaslongbeenknownthatthepropertiesofgalaxiesdependonthe 1 width, W0(Hα) >4A˚ is environmentally dependant, decreasing environmentinwhichtheyarelocated.Elliptical,non-star-forming 1 withincreasingdensity.Theyalsonotedthatemissionlinefraction : galaxies occupy more dense regions of space than star-forming, appearstodependonboththelocalenvironment(∼1Mpc)andon v disc-dominatedgalaxies,givingrisetotheso-calledmorphology- i thelargescalestructure(∼5Mpc). X densityrelation(Oemler1974;Dressler1980).Thephysicalorigin Studies at higher redshifts (z ∼ 1) have used surveys such ofthisrelationisstillsubjecttodebate,withdisagreementmainly r as DEEP2 (Davisetal. 2003) and VVDS (LeFe`vreetal. 2005). a centeringonwhethertherelationarisesduetointernalorexternal DEEP2 investigations (Cooperetal. 2006; Cooperetal. 2007) processes(naturevsnurture). usedtheprojectedthirdnearestneighbourstatistic,studyinggalaxy Most recent low redshift studies (Kauffmannetal. 2004; properties and thecolour-density relation respectively. They con- Baloghetal. 2004) utilise the Sloan Digital Sky Survey (SDSS) cluded that there is a strong dependence on rest frame (U −B) or the Two-degree-Field Galaxy Redshift Survey (2dFGRS) colour, with blue galaxies occupying lower density regions but to conduct statistical investigations of galaxy environments. showing a strong increase in mean local density with luminosity Kauffmannetal.(2004)constrainedthespecificstarformationrate atz∼1.Thistheyconcludeisconsistentwiththerapidquenching (SSFR)usingthe4000A˚ breakandfoundthattheSSFR(andnu- ofstarformationbyAGNorsupernovafeedback,asrampressure clear activity) depend most strongly on local density, from star- stripping,harassmentandtidalinteractions,whichoccurpreferen- forminggalaxiesatlowdensitiestopredominantlyinactivesystems tially in clusters, would be insufficient to explain these findings. athighdensities. Cooperetal.(2007)alsoobservedthatthefractionofgalaxieson Studies by vanDerWeletal. (2008) and Bamfordetal. theredsequenceincreaseswithlocaldensity,asinthelocalUni- (2008)foundthatstructure,colourandmorphologyaremainlyde- verse,butthisweakenswithredshiftanddisappearsbyz∼1.3. pendent on galaxy mass but that at fixed mass, colour and, to a TheVIMOSVLTDeepSurvey(VVDS)investigatedthered- lesserextent,morphologyaresensitivetoenvironment.Studiesof shiftandluminosityevolutionofthegalaxycolour-densityrelation uptoz∼1.5(Cucciatietal.2006).InagreementwithCooperetal. ⋆ E-mail:[email protected] (2007) they found that the local colour-density relation progres- (cid:13)c 2011RAS 2 R. W. Chuter et al. sively weakens and possibly reverses in the highest redshift bin 1.5 (1.2<z<1.5).Thismayimplythatquenchingofstarformationwas more efficient in high density regions. The VVDS team also ob- 1 servedthatthecolour-densityrelationsdepend onluminosityand found that at fixed luminosity there is a decrease in the number est of red objects as a function of redshift in high density regions. B)r 0.5 − Thisimpliedthatstarformationendsatearliercosmicepochsfor (U moreluminous/massivegalaxies,whichisconsistentwithdownsiz- 0 ing(Cowieetal.1996).Wenote,however,thattheVVDSsurvey isbasedonopticalI-bandselection,andassuchwillbestrongly −0.5 biased against red, passive galaxies at z > 1. Conclusions from 0 500 1000 −20 −22 −24 deep K-selected samples suggest that the galaxy colour bimodal- N MK (AB) ity is present to at least z∼1.5 (e.g. Cirasuoloetal. 2007) and Figure1.RestFrame(U−B)coloursforUDSgalaxiesbetween0.75< may be still be present at z∼2 (Cassataetal. 2008; Krieketal. z<1.75.ShownasahistogramandasafunctionofabsoluteK-bandmag- 2008; Williamsetal. 2009). Furthermore red galaxies have been nitude.Thedivisionbetweenredandblueisshown,usingthered-sequence seen to strongly cluster at z>1.5(Daddietal. 2003; Quadrietal. slopederivedbyHartleyetal.(2010). 2007;Hartleyetal.2008;Hartleyetal.2010)whichsuggeststhat acolour-densityrelationmayalsoexistatthesehigherredshifts. A number of physical processes may be responsible for the targetdepthofK=25(AB)overasingle4-pointingmosaicofthe observed environmental trends. Mergers or tidal interactions can Wide-fieldcamera(WFCAM,Casalietal.2007),givingtheUDS teargalacticdiscsapartandarelikelytoplayanimportantrolein forming the most massive galaxies observed today (Toomreetal. anareaof0.88x0.88degrees.The5σ,ABdepthswithin2′′aper- turesfortheJ,HandK-bandsare23.7,23.5and23.7respectively 1972; Faroukietal. 1981). Other processes such as gas stripping fortheDR3,makingitthedeepestnearinfraredsurveyoversucha canseverelyreducethestarformationratebyremovingthecoldgas largeareaatthetimeofrelease.Fordetailsofthestackingproce- from galaxies falling into massive dark-matter halos (Gunnetal. dure,mosaicing,catalogueextractionanddepthestimationwere- 1972;Dekel&Birnboim.2006).Feedbackisalsothoughttoplay ferthereadertoAlmainietal.(inprep.)andFoucaudetal.(2007). amajor role, either fromAGN or supernovae (e.g. Benson 2005; The field is also covered by deep optical data in the B, V, R, i′ Springeletal. 2005). These processes may heat or eject the gas withingalaxiesandthuseffectivelyterminateanyfurtherstarfor- and z′ -bands with depths of BAB=28.4, VAB=27.8, RAB=27.7, mation, which can rapidly lead to the build-up of the galaxy red i′AB=27.7 and zA′B=26.7 from the Subaru-XMM Deep Survey sequence. Finally infalling cold gas in low mass dark matter ha- (SXDS) (Furusawaetal. 2008, 3σ, 2′′ diameter). Data from the losmayfalldirectlyontothegalaxy, whereas inhigh masshalos SpitzerLegacyProgram(SpUDS,PI:Dunlop)reaching5σ depths the gas is thought to be heated by shocks and therefore remains of24.2and24.0(AB)at3.6µmand4.5µmrespectivelyandU-band supported(Whiteetal.1991;Birnboimetal.2003).Akeygoalof datafromCFHTMegacam(UAB=25.5;Foucaudetal.inprep)are alsoutilised,whichresultsinaco-incidentareaof0.63deg2 after observational extragalactic astronomy is to disentangle which of masking. theseprocessesareresponsibleforestablishingthebimodalgalaxy Thegalaxysampleusedhereisprimarilybasedonselectionin populationsobservedinthelocalUniverse. With the recent advent of deep, wide-field infrared imaging theK-bandimage,uponwhichweimposeacutatKAB=23tomin- imisephotometricerrorsandspurioussourcesandalsotoensurea in the UKIDSS UDS we can now extend studies of galaxy envi- highlevelofcompleteness(≃100percent)andreliablephotomet- ronmentstoz>1.Selectionintheinfraredavoidsthemajorbiases ricredshifts(seeCirasuoloetal.2010;Hartleyetal.2010).Bright againstdustyand/orevolvedstellarpopulations,allowingustoin- starsaretriviallyremovedfromthecombinedcataloguebyexclud- vestigatewhethercorrelationsobservedatlowredshiftalsooccur ingobjectsonastellarlocusdefinedby2-arcsecand3-arcsecaper- at high redshift and how these change over time. The large con- tiguousareaofthissurveyalsoallowsustoprobeawiderangeof tures,whichiseffectivetoK <18.1.Thefainterstarsareremoved environmentsusinglargesamplesofgalaxies. usinga(B−z′)−(z′−K)diagram(Daddietal.2004)andthe The paper is structured as follows: §2 outlines the data and criterion(z′ −K) < 0.3 (B −z′)−0.5. These cuts and care- fulmaskingofbrightsaturatedstarsandsurroundingcontaminated selectioncriteriausedinthiswork.§3discussesthemethod used regionsleave33,765galaxiesinoursample. toestimategalaxyenvironments.Theresultsarethenpresentedin §4and§5,with§6summarisingourconclusions.Throughoutthis paperweassumeaΛCDMcosmologywithΩm=0.3,ΩΛ=0.7and H0=71kms−1Mpc−1. 2.2 PhotometricRedshifts Themagnitudes from the DR3catalogue were used todetermine photometric redshifts and stellar ages by χ2 minimisation using a large set of templates. This was performed with a code based 2 DATAANDSAMPLESELECTION largely on the HYPERZ package (Bolzonellaetal. 2000) using bothaveragelocalgalaxySEDsandtemplatesfromtheK20survey. 2.1 TheUKIDSSUltraDeepSurvey Six SEDs of observed starbursts from Kinneyetal. (1996) were This work has been performed using the third data release alsoused toimprove thecharacterisation of young blue galaxies. (DR3) of the UKIRT (United Kingdom Infra-Red Telescope) In- This yeilded photometric redshifts with a δz/(1+z)=0.008 with a frared Deep Sky Survey, Ultra-Deep Survey (UKIDSS, UDS; standard deviation of σ=0.034 after the exclusion of outliers (for Lawrenceetal.2007;Almainietalinprep).TheUKIDSSproject more detail see Cirasuoloetal. 2007; Cirasuoloetal. 2010 and consistsof5sub-surveysofwhichtheUDSisthedeepest,witha references within).Thisalsoprovided rest framemagnitudes and (cid:13)c 2011RAS,MNRAS000,1–9 GalaxyEnvironmentsin theUKIDSS UltraDeep Survey(UDS) 3 colours,ofwhichtherestframe(U−B)andK-bandabsolutemag- regionsandthenumberofrandomswithintheaperturearecounted nitudeareutilisedhere.Sourceswithanunacceptablefit(χ2 >15) (NAper).Thenumber of galaxieswithintheaperture isthennor- r areremovedfromour sampleasthesearelikelytobeunreliable. malisedtogivethefinaldensitymeasurement,ρ/ρ : r Thisremoves4%ofthegalaxysample,themajorityoftheseareei- ρ NAper NTot therQSOs(36%),cross-talk(26%)ortheminormembersofpairs = g × r (3) ormergers(23%),withtheremainderconsistinglargelyofobjects ρr NrAper NgTot withverylowsurfacebrightness.Thefractionofotherwiseuseful whereNTotandNTotarethentotalnumberofrandompointsand r g objectsrejectedistherefore<0.6%. galaxiesrespectively,sothatρ/ρ =1correspondstoadensitycon- r sistentwiththatofarandomdistributionofgalaxies.Thismethod waschosentobethebasisofthisworkasitisconceptuallysim- 2.3 PassiveSample ple,andasconcludedbyCooperetal.(2005),thistechniquehasa Todefineapassivegalaxysubsetwithminimalcontaminationfrom distinctadvantageinfieldsmaskedbyalargenumberofholes.The dustystarformingobjectsweuseasubsetofgalaxiesoutlinedin nearest neighbour method would require the exclusion of a large Hartleyetal.(2010).Templateswereusedtofiteitheraninstanta- fractionofdataclosetoholesandfieldedges. neousburstparameterisedbyanage,oranexponentiallydecaying The nth nearest neighbour method was first employed by star-formation rate parameterised by an age and τ, the e-folding Dressler (1980), this calculates the distance to the nth nearest timeintheexponentiallydecliningstar-formationrate,suchthat, galaxy,D inMpcandisexpressedhereasasurfacedensity, n SFR=SFR0×e−age/τ (1) Σ = n (4) n πD2 n where SFR is the star-formation rate at the time of observation Thesurfacedensity,Σ isthenrenormalisedsuchthat, andSFR0 wastheinitialvalue.Wedefineaconservativepassive n sample as galaxies that are simultaneously old (age>1Gyr) and Σ δ = n (5) have ongoing star formation with SFR 6 0.1% of SFR0, and n Σ¯ astarformingsamplewithSFR>10%ofSFR0,with3947and whereΣ¯ isthemediandensityofgalaxieswithinthefield.Tore- 22,158galaxiesineachsamplerespectively. Todefinetheredsequenceweperformedaχ2minimisationto ducetheeffect of theedges, thedistancetothenearest edge was calculatedandifthiswaslessthanthedistancetothethirdnear- fitanequationoftheform(U−B)=a×M +b totheold,burst K estneighbour thentheobjectwasremovedfromthesample.This galaxies,definingtheredsampletobeallgalaxieswithin3σofthis methodwasonlyusedinthisworktotesttheprimaryfindingsof fit(seeHartleyetal.(2010)foramoredetaileddescription).Inthis theaperturemethod.Theresultsarepresentedintheappendix. work, to separate redand blue galaxies, weuse thered-sequence slopefromHartleyetal.(2010)butchoosethedivisionbetweenthe twopopulationstofittheminimumintheoverallcolourbimodality (asshowninthehistograminFigure1).Thisleadstoadividingline 4 RESULTS inthecolour-magnitudediagramasfollows: Belowweexploretherelationshipbetweengalaxycoloursanden- (U−B)=−7.09×10−3M +0.52 (2) vironmentasafunctionofredshift.Relativelybroadredshiftbins K areusedtominimisethecontaminationduetophotometricredshift Thisboundarywasfoundtoseparatetheredandbluepopulations errors.Thesesourcesofuncertaintyareexploredfurtherinsection effectivelytoz∼1.75.Athigherredshiftthebimodalityingalaxy 5. coloursislessclear,whichmayinpartbeduetophotometricer- rors.Afullexaminationofthisissueandtheevolutionofthered sequence willbe presented in Cirasuoloet al. (inprep). Previous 4.1 RedandBlueEnvironments studieshavefoundevidenceforanevolutioninthelocationofthe redsequencewithredshift(e.gBrammeretal.2009).Forsimplic- Figure2showshistogramsofgalaxyenvironmentswithintwoaper- ity we choose not to model the red sequence in such detail and tures of 1Mpc and 500kpc diameter. These are divided into four instead use the fixed colour selection boundary given above. We redshiftbinsbetweenz=0.25andz=2.25andintoredandblue note, however, that using an evolving boundary made no signifi- populations.AKS-testwasperformedonthesamplestoassessthe cantdifferencetoanyoftheconclusionspresentedinthiswork. difference between the red and blue populations, withall but the Table1showstheresultingnumber ofredandbluegalaxies final redshift bin showing highly significant differences between assignedtoeachphotometricredshiftbin,includingtheconserva- thetwopopulations.Inthefirstthreeredshiftbinswecanexclude tivesubsamplesofpassiveandactivelystar-forminggalaxies. thenullhypothesisthat theredandbluegalaxiesaredrawnfrom thesameunderlyingpopulationwithasignificanceinexcessof5σ significance,fallingto<2σand<1σsignificanceatz >1.75. Figure 3 is a plot of the mean density of red, blue, passive 3 ENVIRONMENTALMEASUREMENT (black)andstarforming(cyan)galaxiesinbinsofabsoluteK-band Weusedtwomethodstocalculategalaxyenvironment: Countsin magnitude.Errorbarsarederivedfromtheerroronthemeanden- an Aperture and nth Nearest Neighbour. In both methods all the sityof galaxies within agiven bin. Sources of error are explored galaxieswithinaphotometricredshiftbinarecollapseddownonto furtherinSection5.Thepassiveandstarforminggalaxiesarede- a2Dplaneandtheredshiftinformationwithinthebinisnotutilised finedinSection2.Asbeforetheyareplottedinfourredshiftbins anyfurther.Intheaperturemethodaperturesof1Mpc,500kpcand but with an additional 250kpc aperture. This plot illustrates that 250kpcdiameter(physical)areplacedoneachgalaxyandthenum- redand/or passive galaxiesresideinsignificantlydenser environ- berofgalaxieswithinthataperturearecounted(NAper).Asample mentsthanblueand/orstar-forminggalaxiesfromthepresentday g of∼100,000randomgalaxiesarethenputdownintheunmasked toz ∼1.5,andthisdifferenceisapparentatallluminosities.This (cid:13)c 2011RAS,MNRAS000,1–9 4 R. W. Chuter et al. 1.5 1.8 1.4 0.25<z<0.75 1Mpc 0.25<z<0.75 500kpc 2.5 0.25<z<0.75 250kpc 1.6 1.3 2 > 1.4 ρr1.2 ρ/ < 1.1 1.2 1.5 1 1 1 0.9 1.5 1.8 1.4 0.75<z<1.25 1Mpc 0.75<z<1.25 500kpc 2.5 0.75<z<1.25 250kpc 1.6 1.3 2 > 1.4 ρr1.2 ρ/ < 1.1 1.2 1.5 1 1 1 0.9 1.5 1.8 1.4 1.25<z<1.75 1Mpc 1.25<z<1.75 500kpc 2.5 1.25<z<1.75 250kpc 1.6 1.3 2 > 1.4 ρr1.2 ρ/ < 1.1 1.2 1.5 1 1 1 0.9 1.5 1.8 M (AB) K 1.4 1.75<z<2.25 1Mpc 1.75<z<2.25 500kpc 2.5 1.75<z<2.25 250kpc 1.6 1.3 2 > 1.4 ρr1.2 ρ/ < 1.1 1.2 1.5 1 1 1 0.9 −18 −20 −22 −24 −18 −20 −22 −24 −18 −20 −22 −24 M (AB) M (AB) M (AB) K K K Blue Red Star Forming Passive Figure3.Theaveragegalaxyoverdensity asafunctionofK-bandluminosity, displayedinfourredshiftbins(toptobottom)and usingprojectedapertures ofdiameter1Mpc,500kpcand250kpc(threecolumns).Galaxies aredisplayedinred,blue,passiveand star-formingsubsets,asdefinedinSection2.Notethechangeinscaleforeachcolumn.Theenvironmentsaredefinedsothatρ/ρr=1 correspondstoadensityconsistentwiththatofarandomdistributionofgalaxies. 0.25<z<0.75 0.75<z<1.25 1.25<z<1.75 1.75<z<2.25 Red 3,469 3,500 2,565 481 Blue 7,367 7,255 4,674 1,711 Passive 1,623 1,395 722 87 Star-forming 6,938 6,741 5,048 1,817 Table1.Tableshowingthenumberofgalaxiesineachsampleandineachredshiftrange,includingourconservativesubsetsofpassive andactivelystar-forminggalaxies. (cid:13)c 2011RAS,MNRAS000,1–9 GalaxyEnvironmentsin theUKIDSS UltraDeep Survey(UDS) 5 0.12 0.1 0.25<z<0.75 1Mpc 0.25<z<0.75 500kpc 1.3 0.25<z<0.75 1Mpc 1.6 0.25<z<0.75 500kpc 0.08 >10σ >10σ 1.2 1.5 f0.06 ρρ</>r 1.4 1.1 1.3 0.04 1.2 0.02 1 1.1 0.102 0.1 0.75<z<1.25 1Mpc 0.75<z<1.25 500kpc 1.3 0.75<z<1.25 1Mpc 1.6 0.75<z<1.25 500kpc 0.08 ∼5σ ∼8σ 1.2 1.5 f0.06 ρρ</>r 1.4 1.1 1.3 0.04 1.2 0.02 1 1.1 0.102 0.1 1.25<z<1.75 1Mpc 1.25<z<1.75 500kpc 1.3 1.25<z<1.75 1Mpc 1.6 1.25<z<1.75 500kpc 0.08 ∼5σ ∼4σ 1.2 1.5 f0.06 ρρ</>r 1.4 1.1 1.3 0.04 1.2 0.02 1 1.1 0.102 0.1 1.75<z<2.25 1Mpc 1.75<z<2.25 500kpc 1.3 1.75<z<2.25 1Mpc 1.6 1.75<z<2.25 500kpc 0.08 <2σ <1σ 1.2 1.5 f0.06 ρρ</>r 1.4 1.1 1.3 0.04 1.2 0.02 1 1.1 0 0 1 2 3 4 0 1 2 3 4 0.4 0.6 0.8 0.4 0.6 0.8 ρ/ρ ρ/ρ (U−B) (U−B) r r rest rest Figure2.Histogramsofthedensityofgalaxieswithin1Mpcand500kpc −22.5<KMag<−21.5 −24<KMag<−22.5 diameteraperturescomparedtoarandomsampleforred(dashedline)and blue(thickline)galaxies.TheσvaluesareobtainedbyperformingaKS Figure4.Theaveragegalaxydensityina1Mpcand500kpcdiameteraper- test,representing thesignificance inrejecting thenull-hypothesis thatthe turescomparedtoarandomsampleasafunctionof(U −B)restframe samplesaredrawnfromthesameunderlyingpopulation. colour.Weseparatehigh(black)andlow(green)luminositysubsets.Note thechangeiny-axisvaluesbetweenthetwodifferentaperturesizes. iscomparabletowhathasbeenfoundinthelocaluniversebyother studies(Kauffmannetal.2004;vanDerWeletal.2008). nosityforbluegalaxiesatz∼1.Ourresultsappeartoextendthese Figure 3 also shows that passive galaxies (shown in black) findingstohigher redshift,suggesting thattheepoch 1 < z < 2 follow a similar density profile to red galaxies but are on aver- representsakeyperiodoftransformationofmassivegalaxiesfrom ageinslightlydenser environments. Thissupportstheconclusion thebluecloudontothepassiveredsequence. thatpassivegalaxieswithintheredpopulationareresponsiblefor theenhancedenvironmentscomparedtobluestar-formingobjects. The environments of galaxies that were red but not in the strict 4.2 Colour-DensityRelation ‘passive’samplewerealsoinvestigatedandthesewerefoundtolie in-betweentheredandbluegalaxyenvironments,aswouldbeex- InFigure4wedisplayaveragedensityasafunctionof(U −B) pected(thesearenotshownforclarity).Theactivelystar-forming rest-framecolour, withfaint (-22.5< M <-21.5) andluminous K galaxies exhibit the same environmental dependence as the blue (-24<M <-22.5)objectsshowningreenandblackrespectively. K galaxies,followingthesameluminosity-densityprofile. Thisshowsacleartrendforgalaxiesbelowz∼2,withreddergalax- In addition to the clear separation of red and blue galaxies, iesoccupyingdenseraverageenvironments.Inthelowestredshift wealsofindageneral trendofincreasinggalaxydensitywithlu- binthischangeoccursveryabruptlyattheboundaryinthecolour minosity, particularly for blue galaxiesand on smaller scales. In- bimodality at (U −B) ∼0.7 (see Figure 1). There is also a rest specting the twointermediate redshift bins in Figure3, on scales general trend for more luminous galaxies to occupy denser envi- below500kpcwefindthatthemostluminousbluegalaxiesappear ronments. to inhabit environments approaching those of red/passive galax- In summary, we have found strong evidence that red galax- ies.Theseresultsareconsistent withthefindingsof Cooperetal. ies are on average found in denser local environments than blue (2007),whoobservedastrongincreaseinlocaldensitywithlumi- galaxies, extending the comparison to a higher redshift than any (cid:13)c 2011RAS,MNRAS000,1–9 6 R. W. Chuter et al. 1.8 1.8 1.6 4 1.7 1.25<z<1.75 500kpc 1.7 1.25<z<1.75 500kpc 0.25<z<0.75 1Mpc 2 0.25<z<0.75 500kpc 3.5 0.25<z<0.75 250kpc 1.6 Q1 1.6 Q2 1.4 1.8 3 ρρ</>r111...345 111...345 ρρ</>r1.21 111...2461 12..552 1 1.2 1.2 1.6 4 1.1 1.1 0.75<z<1.25 1Mpc 2 0.75<z<1.25 500kpc 3.5 0.75<z<1.25 250kpc 1.4 1.8 1 1 3 11..46 1.25<z<1.75 500kpc 11..46 1.25<z<1.75 500kpc ρρ</>r1.2 11..46 2.52 1.2 Q4 1.2 Q3 1.2 1 1.5 1 1 1 1 ρρ</>r 1.6 1.25<z<1.75 1Mpc 2 1.25<z<1.75 500kpc 3.54 1.25<z<1.75 250kpc 1.6 1.6 1.4 1.8 3 11..24 11..24 ρρ</>r1.2 11..46 2.52 1 1.61 1 1.2 1.5 Q1 1 Q2 1.4 1.25<z<1.75 500kpc 1 1.6 4 20 1.2 1.75<z<2.25 1Mpc 2 1.75<z<2.25 500kpc 3.5 1.75<z<2.25 250kpc y (1000 pixles) 1105 ρρ</>r11..461 ρρ/r11..241 1111....24681 12..5523 5 1 Q4 Q3 1.2 −22 −22.5 −23 −23.5 −24 −22 −22.5 −23 −23.5 −24 −22 −22.5 −23 −23.5 −24 M (AB) M (AB) M (AB) 1 K K K 5 10 15 20 −18 −20 −22 −24 x (1000 pixels) MK (AB) Figure6.AsFigure3,butusingonlythebrightestgalaxieswithM 6-22 K (AB)astracersofgalaxydensity.Notethechangeiny-axisvaluesbetween Figure5.ShowingthedensityofredandbluegalaxiesagainstK-bandlu- thethreedifferentaperturesizes minosityinfourquadrantsoftheUDSfieldshowninthebottomleft,using medianvaluesofxandytocalculatethepositionofthecentre.Theoverall resultforthefieldat1.25<z<1.75isshowninthebottomrightwiththe errorbarscalculatedfromtheerroronthemeanofthefourquadrants,to 5.1 CosmicVariance giveanestimateofthecosmicvariance. To attempt to quantify the effects of cosmic variance on our re- sultswedividetheUDSfieldintofourquadrants.Themedianval- ues of the x and y positions of the galaxies were used to divide previousstudy.Cooperetal.(2007)probeduptoz ∼1.3,finding the field, to ensure that the number of galaxies in each quadrant thatredandbluegalaxiesatthisredshiftoccupyindistinguishable wascomparable. Theresultsforthegalaxiesintheredshiftrange environments.Incontrastwefindthatthereisasignificantdiffer- 1.25<z<1.75 are shown in Figure 5. The average density versus ence between red and blue galaxies between 1.25 < z < 1.75 K-bandabsolutemagnitudeisshownforthefourquadrants, with (see Figures 2 and 3). This difference is likely to be due to our thebottomleftpanelshowingtheprojecteddistributioninthefield. methodofselectinggalaxiesfromdeepinfraredimaging.Veryred, ThelowerrightpanelshowstheoriginalresultfromFigure3(red passive galaxies willhave been missed from the R-band selected and blue only) but with the error bars calculated from the error candidates used inthe DEEP2study. Thisselection effect can be onthemeanofthedensitiesinthefourquadrantsineachmagni- quantified using our UDS sample. Considering only the most lu- tudebin. Itcan beseen that thedistinctionbetween redand blue minousgalaxies(M 6-22.5)intheredshiftrange1<z<1.3,ifwe galaxyenvironmentsremains,althoughthereareclearlysignificant K imposetheselectioncriteriaofDEEP2(R 624.1) wefindthat variations across the field. We conclude that while field variance AB 75%ofredgalaxieswouldbemissedcomparedtoonly20%ofthe isclearlypresentinthesedatatheprimaryfindingsremainrobust. bluesample.Atfainterluminositiesvirtuallynored/passivegalax- Similar conclusions were drawn from other redshift bins and on iesareselected. otherscales,whicharenotshownintheinterestsofbrevity. Asacautionaryremark,however,wenotethatstrictlyspeak- ingwehaveonlytestedtheinternalfieldvariance,sincedespitethe relativelywidefieldoftheUDS(50×50comovingMpcatz∼1) wemaynevertheless bepronetolarge-scalecosmicvariancedue 5 INVESTIGATINGSOURCESOFERROR tounusualsuperstructures(Somervilleetal.2004).Testingagainst sucheffectswillrequirefurtherwide-areainfraredsurveys,suchas Inthissectionweconductanumberofteststoinvestigatevarious theforthcomingVIDEOandUltra-VISTAsurveys. source of systematic uncertainty. Weattempt toquantify thecos- micvarianceinourdatabydividingthesurveyintofourquadrants. Avolumelimitedstudyofthebrightergalaxiesisalsoperformed, 5.2 Volume-LimitedSample toallow aclearer comparison betweendifferent redshift bins. Fi- nallyweinvestigatetheeffectsofphotometricredshifterrorsusing Ourmeasurementofenvironmentaldensityisbasedonafluxlim- MonteCarlosimulations. ited survey, so by definition we are using fainter galaxies to de- (cid:13)c 2011RAS,MNRAS000,1–9 GalaxyEnvironmentsin theUKIDSS UltraDeep Survey(UDS) 7 3 1.35 0.75<z<1.25 2.52 5<z<1.25 5<z< 1.75 5 <z< 2.25 11.2.35 7 2 7 0. 1. 1. pc 1.2 M 1 o > Phot1.5 ρρ/r1.15 z < 1.1 1 1.05 0.5 1 −19 −20 −21 −22 −23 −24 0 M (AB) 0 0.5 1 1.5 2 2.5 3 K z Spec Figure8.MonteCarlosimulationshowingtheeffectofthephoto-zerrors Figure 7. Plot of the reliable spectroscopic redshifts against photomet- onourmeasurementofgalaxyoverdensity,basedon100simulations.The ric redshifts, highlighting the three highest redshift bins, 0.75<z<1.25, thickblacklineshowstheaverageofthesimulationsandthedashedblack 1.25<z<1.75and1.75<z<2.25,withAGNandradiogalaxiesremoved. lineshowstheoriginalresultfor0.75<z<1.25fromFigure3. Theredsquaresandbluecirclesindicatetheredandbluegalaxiesrespec- tively,asdefinedinsection2.2. 1.35 1.25<z<1.75 fine environments at low redshift. To investigate the influence of 1.3 this systematic effect we reduced the sample to only the bright- estgalaxieswithM 6-22(AB)andrepeatedtheenvironmental K 1.25 analysisusingthisvolume-limitedsample.Thisallowsthegalax- ies to be used as tracers in the measure of environment over the pc 1.2 samerangeinluminosityatallredshifts,toallowafairercompar- 1M > isonbetweenepochs.TheresultsareshowninFigure6,wherethe ρρ/r1.15 averagedensityversusK-bandluminosityisshownforcomparison < withtheflux-limitedstudyinFigure3.Asexpected, thevolume- 1.1 limitedstudyismuchnoisier(particularlyatlowredshift)butthe resultsareconsistent withtheprimaryfindingsofSection4.Red 1.05 and/or passive galaxies are found tooccupy denser environments onaveragetoz∼1.5. 1 −19 −20 −21 −22 −23 −24 M (AB) 5.3 MonteCarloSimulations K Monte Carlosimulations wereconducted toinvestigatetheeffect Figure9.AsFigure8butforthe1.25<z<1.75bin. ofphotometricredshifterrorsonourresults,explicitlyallowingfor the effects of outliers and catastrophic errors. Such errors would generallydilutedifferencesingalaxyenvironments,butcouldpo- δz<0.25atz ∼1(Cirasuoloetal.2010).Theresultingfractions tentially introduce fake overdensities if a large fraction of low- were then randomly applied to the full sample, shifting galaxies redshift structure is incorrectly assigned to a higher redshift bin. intonewbinstosimulatetheeffectofphotometricredshifterrors We note, of course, that the effects of photo-z errors are already (‘normal’and‘catastrophic’). presentinthedata,sothesesimulationscanonlyprovideanindi- This was repeated 100 times per redshift bin. Results are cationofthemagnitudeoftheshiftsduetotheseeffects. displayed for the redshift bins 0.75 < z < 1.25 (Fig.8) and The distribution of spectroscopic versus photometric red- 1.25<z <1.75(Fig.9).Thesesimulationsdemonstratethatpho- shiftsusedinthisfieldareshowninFigure7,withspectroscopic tometric redshift errors tend to dilute the distinction between the redshifts derived from a variety of sources (Yamadaetal. 2005; redandbluepopulationsratherthanintroducefakedifferences.The Simpsonetal. 2006; Akiyama et al., in prep; Simpson et al., in resultingadditionalsourcesoferroringalaxydensityarecompara- prep; Smailet al.,in prep) asoutlined in Cirasuoloet al. (2010). bletotheerrorsonthemeandisplayedpreviously.Onlyinhighest Weusethedistributionineachredshiftbin(higherredshiftbinsin- redshiftbin(z>1.75)dothesimulatedpopulationsintroducesig- dicatedinFigure7)toestimatethefractionofgalaxiesincorrectly nificantextrascatter(notshown).Giventhesefindings,andthelack reassignedfromoneredshiftbintoanother.Forsimplicitytwocat- ofspectroscopicredshiftsatz∼2,weurgecautionininterpreting egories were adopted: ‘catastrophic’ errorswere defined as those our preliminary results in this highest redshift bin. At lower red- withδz > 0.25,whileothersareclassifiedas‘normal’.Intheab- shifts,however,theenvironmentaldifferencesappearrobust. senceofcatastrophicerrorsweexpect>99.9%ofgalaxiestoshow Notingthatwehaveonly97spectroscopicredshiftsinthered- (cid:13)c 2011RAS,MNRAS000,1–9 8 R. W. Chuter et al. onrestframe(U−B)colours,aswellasusing‘passive’and‘star- 1.35 forming’galaxiesfromtemplatefitting.Weattempttoquantifythe 0.75<z<1.25 effects of cosmic variance, photometric redshift errors and flux- 1.3 limitbiasesontheresultingenvironmentalmeasurements.Wefind thefollowingprincipleresults: 1.25 (i) We find a strong relationship between rest-frame (U −B) Mpc 1.2 colour andgalaxy environment toz ∼ 1.5, withredgalaxiesre- 1 ρρ/>r1.15 ssicdailnegsbinelosiwgn1ifiMcpacn.tlTyhdeesnesreersuenltvsiraornemroebnutsstthtoanthbelueeffegcatlsaxoifefiseolnd < variance,flux-limitbiasesandphotometricredshifterrors.Theen- 1.1 vironmentsappearindistinguishablebyz ∼2,butthecurrentlack ofspectroscopicredshiftsatthisepochdoesnotallowarobusttest 1.05 ofthistantalisingsignalatpresent. (ii) Selecting ‘passive’ and ‘star-forming’ galaxies using tem- 1 platefittingyieldsconsistentresults,withthepassivesubsetoccu- −19 −20 −21 −22 −23 −24 pyingslightlydenserenvironmentsthantheglobalredpopulation M (AB) K atallepochs.Weconcludethatthepassivesubsetoftheredgalax- iesareresponsiblefortheenhancedenvironmentscomparedtoblue Figure10.TheresultsofMonteCarlosimulations basedontheextreme galaxies. scenarioinwhichallincorrectlyassignedphotometricredshiftsleadtoblue (iii) Onsmallscales(< 0.5Mpc)wefindevidenceforaposi- galaxies being incorrectly classified as red, and vice versa. The original measuresofenvironmentforredgalaxiesandbluegalaxiesasafunctionof tivecorrelationbetweengalaxyK-bandluminosity(agoodproxy luminosityareshownasthickredandbluelines.Blackdashed(dot-dashed) forstellarmass) andlocaldensity.Thistrendisparticularlyclear linesshowtheresultsofMonteCarlosimulationsforthered(blue)popu- for star-forming and blue galaxies, with the most luminous blue lationsrespectively. Thecorrespondinggreenandmagentalinesillustrate galaxiesatz∼1−2showingaverageenvironmentscomparableto theinfluenceofdoublingandtriplingthenumberofincorrectly assigned passiveredgalaxies.Thesefindingsappearinverygoodagreement galaxies. withthefindingsofCooperetal.(2007)atz∼1.Theseresultsare consistentwiththeidentificationofhigh-zgalaxiesintransitionto theredsequenceinthedensestenvironments. shiftbin1.25< z <1.75(with64between1.25< z <1.75), photo Toimproveonthisstudy,wenotethattheUDSimagingpro- whereenvironmentaldifferencesappeartobeverysignificant,we gramme will continue at UKIRTuntil 2012 and ultimatelyprobe ranadditionalsimulationsusingonlyhalfofthespectroscopicsam- ∼ 1 magnitude deeper than the dataused inthiswork. Thiswill ple.TheresultswereverysimilartoFigure9,indicatingthatweare be sufficient to detect L∗ galaxies to z ∼ 5 in addition to tens notsufferingfromlownumberstatistics. ofthousandsofsub-L∗galaxiesatlowerredshift.Furthermore,an Asanadditionaltest,weconductfurtherMonteCarlosimula- ongoing ESO Large Programme (UDSz) will soon transform the tionstoinvestigatetheeffectsofcontaminationofthebluegalaxies UDSprojectwiththeadditionof∼3000galaxyspectra.Thiswill bytheredpopulationandviceversa.Thesesimulationsweremo- increasethenumberofUDSspectroscopicredshiftsatz >1.5by tivated by the worry that incorrect redshifts may inevitably lead anorder of magnitude, whichwe expect todramaticallyimprove to errors in the resulting restframe (U − B) colours. We adopt thereliabilityofphotometricredshiftsandallowustoextendour anextremeapproach bysimulating100%misclassification,using studyofgalaxyenvironmentsandlarge-scalestructuretothecru- the contaminating fractions outlined above but moving only blue cialepochwhenthegalaxyredsequenceisfirstestablished. galaxiesintotheredsampleandviceversa.Wefindthatthesesim- ulationshavetheeffectofdilutingtheenvironmentforbothredand bluesimulations,butonceagainareinsufficienttoinfluenceanyof theresultspreviouslypresented.Toinvestigatefurther,thecontam- ACKNOWLEDGEMENTS inatingfractions areartificiallyincreased todouble andtriplethe Weare grateful tothestaff at UKIRTfor operating thetelescope maximalcontaminationindicatedbyspectroscopicredshifts.These withsuchdedication.WealsothanktheteamsatCASUandWFAU only servetofurther reduce theaveragegalaxydensities. Results forprocessingandarchivingthedata.RWCwouldliketothankthe forthe0.75<z <1.25binareillustratedinFigure10,andsimi- STFCfortheirfinancialsupport.JSDacknowledgesthesupportof larresultsarefoundforallotherredshiftbins. the Royal Society via aWolfson Research Merit award, and also Weconcludethatphotometricredshifterrors(includingcatas- thesupportoftheEuropeanResearchCouncilviatheawardofan trophic errors) have only served to dilute the differences in envi- AdvancedGrant. ronment,butanunrealisticproportionofmisclassifiedgalaxiesare requiredtosignificantlyinfluenceourprimaryfindings. REFERENCES BaldryI.K.etal.,2004,ApJ,600,681 6 CONCLUSIONS BaloghM.etal.,2004,MNRAS,348,1355 Inthispaperwehavepresentedastudyoftheenvironmentofover BamfordS.P.etal.,2008,MNRAS,393,1324 33,000K-bandselectedgalaxiesbetween0.25<z<2.25.Wehave BellE.Fetal.,2004,ApJ,608,752 used a simple measurement of projected galaxy overdensities to Benson, A.J., 2005, Royal Society of London Philosophical studytheenvironmentsofgalaxiesdividedintoredandbluebased TransactionsSeriesA,363,695 (cid:13)c 2011RAS,MNRAS000,1–9 GalaxyEnvironmentsin theUKIDSS UltraDeep Survey(UDS) 9 BirnboimY.,DekelA.,2003,MNRAS,345,349 Blanton, M.R., Eisenstein D., Hogg D.W., Schlegel D.J., & 0.2 0.25<z<0.75 33rd NN 0.3 0.25<z<0.75 9th NN BrinkmannJ.,2005,ApJ,629,143 0.25 BolzonellaM.,MirallesJ.-M.,&Pello´R.,2000,A&A,363,476 δ+)>n0.15 0.2 BCraasmalimMer.Get.Bal..,,e2t0a0l6.,,2A0&09A,,A4p3J7,,770767,L173 <log (110 0.1 00.1.15 CassataP.,etal.,2008,A&A,483,L39 0.05 0.05 CirasuoloM.etal.,2007,MNRAS,380,585 0 Cirasuolo,M.etal.,2010,MNRAS,401,1166 0 M −0.05 K CooperM.C.etal.,2005,ApJ,634,833 0.2 0.75<z<1.25 18th NN 0.3 0.75<z<1.25 5th NN CooperM.C.etal.,2006,MNRAS,370,198 0.25 CooperM.C.etal.,2007,MNRAS,376,1445 δ+)>n0.15 0.2 C8o3w9ieL.L.,SongailaA.,HuE.M.&CohenJ.G.,1996,ApJ,112, <log (110 0.1 00.1.15 Cucciati.O.etal.,2006,A&A,458,39 0.05 0.05 DaddiE.etal.,2003,ApJ,588,50 0 Daddi E., Cimatti A., Renzini A., Fontana A., Mignoli M., 0 M −00..0355 K PozzettiL.,TozziP.&ZamoraniG.,2004,ApJ,617,746 0.2 1.25<z<1.75 11th NN 0.3 1.25<z<1.75 3rd NN DavisM.etal.,2003,Proc.SPIE,4834,161 0.25 DekelA.,&BirnboimY.,2006,MNRAS,368,2 δ+)>n0.15 0.2 DElrbeaszs.leDr.Ae.t,a1l9.,8200,0A7p,JA,2&3A6,,345618,33 <log (110 0.1 00.1.15 FaroukiR.,Shapiro,S.L.,1981,ApJ,243,32 0.05 0.05 0 Foucaud,S.etal.,2007,MNRAS,376,L20 FurusawaH.etal.,2008,ApJS,176,1 0 M −00..0355 M K K GunnJ.E.,GottJ.R.,1972,ApJ,176,1 0.2 1.75<z<2.25 5th NN 0.3 1.75<z<2.25 1st NN 0.25 HartleyW.G.,etal.,2008,MNRAS,391,1301 HartleyW.G.,etal.,2010,MNRAS,407,1212 δ+)>n0.15 0.2 HKoaugfgfmDa.Wnn.,Get.eatl.a,l2.,020030,4A,pMJ,N5R85A,SL,5353,713 <log (110 0.1 00.1.15 KriekM.,vanderWelA.,vanDokkumP.G.,FranxM.,&Illing- 0.05 0.05 0 worthG.D.,2008,ApJ,682,896 0 −0.05 Kinney A.L., Calzetti D., Bohlin R.C., McQuade K., Storchi- −18 −20 −22 −24 −18 −20 −22 −24 M (AB) M (AB) K K BergmannT.,&SchmittH.R.,1996,ApJ,467,38 LawrenceA.etal.,2007,MNRAS,379,1599 Figure11.Thenearestneighbourdensityofredandbluegalaxieswiththe LeFe`vreO.etal.,2005,,439,877 nearestneighbourchosensothatthemeandistanceprobedwasadiameter OemlerA.,1974,ApJ,194,1 of1Mpcand500kpcineachredshiftbin. PoggiantiB.M.etal.,2009,ApJ,693,112 QuadriR.etal.,2007,ApJ,654,138 SimpsonC.etal.,2006,MNRAS,372,741 ThispaperhasbeentypesetfromaTEX/LATEXfilepreparedbythe SpringelV.,DiMatteoT.,Hernquist,L.,2005,ApJ,620,L20 author. ToomreA.&,ToomreJ.,1972,ApJ,178,623 vanDerWelA.etal.,2008,ApJ,675,13 WhiteS.D.M.,FrenkC.S.,1991,ApJ,379,52 WilliamsR.J.,QuadriR.F.,FranxM.,vanDokkumP.,&Labbe´I., 2009,ApJ,691,1879 YamadaT.etal.,2005,ApJ,634,52 YorkD.G.etal.,2000,AJ,120,1579 APPENDIX:NEARESTNEIGHBOURMETHOD In this appendix we use the nth nearest neighbour method as an alternative environmental estimator, to provide a test of the re- sultspresentedearlierandtoallowcomparisonwithpreviouswork (e.g.Cooperetal.2007).Thenthnearestneighbourwascalculated forallthegalaxiesasafunctionofabsolutemagnitude,usingthe methodoutlinedinSection3.Thevalueofnischosentoprobethe samemeanscaleastheaperturesmethod.Theresultsareshownin Figure11),whichappeargloballyconsistentwiththosederivedby theaperturemethod. (cid:13)c 2011RAS,MNRAS000,1–9