DRAFTVERSIONFEBRUARY5,2008 PreprinttypesetusingLATEXstyleemulateapjv.6/22/04 OPTICALANDINFRAREDNON–DETECTIONOFTHEZ=10GALAXYBEHINDABELL1835 GRAHAMP.SMITH,1DAVIDJ.SAND,1 EIICHIEGAMI,2 DANIELSTERN,3ANDPETERR.EISENHARDT3 DraftversionFebruary5,2008 ABSTRACT Gravitationallensingbymassivegalaxyclustersisapowerfultoolforthediscoveryandstudyofhighred- shift galaxies, including those at z≥6 likely responsible for cosmic re–ionization. Pelló et al. recently used thistechniquetodiscoveracandidategravitationallymagnifiedgalaxyatz=10behindthemassiveclusterlens Abell1835(z=0.25). We presentnew Keck (LRIS) andSpitzer SpaceTelescope (IRAC) observationsof the z=10candidate(hereafter#1916)togetherwithare-analysisofarchivalopticalandnear-infraredimagingfrom 6 theHubbleSpaceTelescopeandVLTrespectively.Ouranalysisthereforeextendsfromtheatmosphericcut-off 0 at λ ≃0.35µm out to λ ≃5µm with Spitzer/IRAC. The z=10 galaxy is not detected in any of these data, obs obs 0 includinganindependentreductionofPellóetal.’sdiscoveryH-andK-bandimaging. Weconcludethatthere 2 is nostatistically reliable evidenceforthe existenceof#1916. We also assess the implicationsofourresults n forground-basednear-infraredsearchesforgravitationallymagnifiedgalaxiesatz∼>7. Thebroadconclusionis a thatsuchexperimentsremainfeasible,assumingthatspace-basedopticalandmid-infraredimagingareavail- J abletobreakthedegeneracywithlowredshiftinterlopers(e.g.z∼2- 3)whenfittingspectraltemplatestothe 9 photometricdata. Subjectheadings:cosmology:observations—earlyuniverse—galaxies: evolution—galaxies: formation— 1 infrared:galaxies v 1 8 1. INTRODUCTION ThefaintendoftheluminosityfunctionofLyman-αemitters 1 atz=5hasalsobeenconstrainedwiththehelpofgravitational 1 Observations of distant QSOs (Becker et al. 2001; Fan et lensing (Santos et al. 2004; Ellis et al. 2001). Extension of 0 al.2002)andthecosmicmicrowavebackground(Kogutetal. 6 2003)togethersuggestthattheuniversewasre-ionizedsome- thesetechniquestoz∼>7isthereforeanimportantelementof 0 where betweenz≃6 and z≃20. Searchingforthe sourcesof observationalstudiesofcosmicre-ionization. / re-ionizing photons is currently an intense observational ef- Pelló et al. (2004 – hereafter P04) reported a gravitation- h fort. Most searches naturally concentrate on luminous sys- allymagnified(µ∼25- 100)galaxyatz=10(hereafter#1916, p - tems, i.e. QSOs and luminous galaxies, at z∼6- 8 as these followingP04’snomenclature)behindtheforegroundgalaxy o shouldbeeasiertodetectthanlessluminousandmoredistant clusterA1835(z=0.25). Thisinterpretationisbasedonnon- r objects. HoweverQSOslikelyproducedinsufficientphotons detection in optical imaging from the ground (3σ limits in st toaccomplishre-ionizationalone(Fanetal.2001;Bargeret a 0.6′′ diameter aperture: V≥27.4, R≥27.5, I≥26.9) and v:a gala.l2ax0i0e3s)b,aanseddthoenssammaellmsaamybpeletsrufreoomfltuhmeiHnuobubsl(eLS∼>p0a.c3eLT⋆z=e3l.e8)- sapnadcethe(3sσhalpimeiotfitnheac0o.n2t′i′nuduiammaetteλroabsp≥er1tµurme:(uRs7in02g≥a271..25)′′, i scopeUltraDeepField(hereafterHST UDF;Bouwensetal. aperture: (J- H)≥0.6, (H- K)=- 0.5±0.4) which is reminis- X cent of the Lyman-break selection technique (Steidel et al. 2004;Bunkeretal.2004;Yanetal.2004).Thisraisestheim- r 1996).P04corroboratedtheputativeLyman-breakredshifted portant possibilities that re-ionization either occurred much a toλ ≃1.3µmwithanemissionlineatλ =1.3375µmwith earlier,orthatthebulkofthere-ionizingphotonswereemit- obs obs tedbysub-luminousgalaxies,i.e.L∼<0.1L⋆. integrated flux of (4.1±0.5)×10- 18ergcm- 2s- 1, which they interpret as Lyman-α. Lower redshift interpretations of the The UDF studies operate close to the detection threshold line ([OII] at z=2.59; [OIII] at z=1.68; Hα at z=1.04) were of the deepest optical/near-infrared imaging available. It is discardedbyP04largelyonthebasisofthelowprobabilityof thereforedifficult to envisage substantial progressin the de- tectionofmoreremoteand/orlessluminousgalaxiesviadeep solutionsatz∼<7whenfittingsyntheticspectralenergydistri- butions(SEDs)totheirphotometricdata. Themostlikelyof imagingof“blankfields”withthecurrentgenerationoftele- these lower-redshiftsolutions (z=2.59) was further excluded scopes. WiththeadventoftheJamesWebbSpaceTelescope on the basis of the dust extinction required to fit the photo- still some years ahead, the magnifying power of massive metric data (A ≥2), and the absence of doublet structure in galaxyclusterlensesisthereforeamuchneededboostforthe V theobservedemissionline. discoverypower of HST and large ground-basedtelescopes. Basedsolelyonthephotometry(opticalnon-detection,red Indeed,thegalaxyredshiftrecordhasbeenbrokenonseveral (J- H) and blue (H- K) colors) the z=10 interpretation of occasionswith the help of the gravitationalmagnificationof #1916 is plausible. However P04’s preference for this so- distantgalaxiesby foregroundgalaxyclusters(Mellier etal. lution over the lower redshift alternatives was controversial 1991; Franx et al. 1997; Hu et al. 2002; Kneib et al. 2004). fromtheoutset.Forexample,theemissionlinedoesnothave 1DepartmentofAstronomy,CaliforniaInstituteofTechnology,MailCode the characteristic P-Cygni profile of Lyman-α, and the dif- 105–24,Pasadena,CA91125.–Email:[email protected] ferent photometric apertures adopted in the optical (0.6′′ – 2StewardObservatory,UniversityofArizona,933NorthCherryAvenue, smallerthantheground-basedseeingdisc)andnear-infrared Tucson,AZ85721. (1.5′′ – 3× the seeing disk) may suppress the likelihood of 3JetPropulsionLaboratory,CaliforniaInstituteofTechnology,4800Oak lower-redshiftsolutionswhenfittingsyntheticSEDs. Bremer GroveDrive,MS169–327,Pasadena,CA91109. 2 Non–detectionofthez=10Galaxy etal.havealsosuggestedthat#1916mayeithernotexist,or TABLE1 be intrinsically variable, based on their non-detection in the SUMMARYOFPHOTOMETRY H-bandwith NIRI on Gemini-North,H(3σ)>26, in contrast toP04’s4σdetectionofH=25.00±0.25withISAAConVLT. Filter Telescope/ FWHM AperturePhotometrya The spectroscopic identification of #1916 is also in doubt. Instrument (′′) Pellóetal.b ThisPaper Weatherleyetal.(2004)re-analyzedP04’sspectroscopicdata V CFHT/12k 0.76 ≥27.5(0.6′′) ... R CFHT/12k 0.69 ≥27.6(0.6′′) ... andfailedto detectthe emissionline atλ =1.3375µm,cit- ing spuriouspositive flux arising from varoibasble hot pixelsin IR702 CHFSHT/TW/1F2PkC2 00..7187 ≥≥2276..20((00..26′′′′))c ≥27.0..(.0.5′′) theISAACarrayasthelikelysourceofthediscrepancy. J VLT/ISAAC 0.65 ≥25.6(1.5′′) ... A1835 has been used previously as a gravitational tele- H VLT/ISAAC 0.50 25.00±0.25(1.5′′) ≥25.0(1.5′′) scope, for example targeting sub-millimeter galaxies (here- K VLT/ISAAC 0.38 25.51±0.36(1.5′′) ≥25.0(1.5′′) after SMGs; e.g. Smail et al. 2002) and galaxies with 3.6µm Spitzer/IRAC 1.7 ... ≥24.3(5.1′′) extremely red optical/near-infrared colors (Smith et al. 4.5µm Spitzer/IRAC 1.7 ... ≥24.3(5.1′′) 2002). One of the galaxies detected in these surveys, SMMJ14011+0252, lies at z=2.56 and suffers an estimated a Eachnumberinparenthesesisthediameteroftheapertureusedforthe mexitnindcrteiocnenotfd1is.c8o∼<vAerVy∼<o6f.5ga(lIavxiysognroeutpasl.at2z0∼002)-.3Basesaoricniagteind rrees-spceaclteivdetophaopthoomtoemtriectrmiceaapsuerretumreenotfs.2′I′nd§ia4mPe0t4e’rs(∼op3titciamlensotnh-edeseteecintigondsisacr)e: V≥26.2,R≥26.3,I≥24.7. with gravitationally lensed SMGs (Kneib et al. 2004; Bo- bWeconvertallofP04’sopticaldetectionlimitsandtheir1σJ-banddetec- rys et al. 2004), and the strong clustering of SMGs (Blain tionto3σlimits. et al. 2004), #1916 is plausibly at a similar redshift to cP04donotstatewhethertheirR702 detectionlimitisintheVegaorAB SMMJ14011+0252,andmayalsobeobscuredbydust. Fur- system. Wehaveassumedtheformerandconverted ittoABinthistable. P04 also do not explain how they reduced the WFPC2 data, specifically thercircumstantialevidenceforalowerredshiftinterpretation whetherthefinalpixelscale wasdifferent fromthenative 0.0996′′/pixof of#1916comesfromRichardetal.(2003)whousedthesame theWFCdetectors. Inthistablewehaveassumedthatthefourpixelsover spectroscopicdataaspresentedbyP04todiscoverastrongly which this detection limit is measured (see P04 §2.1) subtend a solid an- gleof0.0996′′×0.0996′′. Giventheseuncertaintiesandtheabsenceofthis reddened star-forming galaxy at z=1.68. This redshift coin- detection limitfromP04’sTable1andFig.3,weignoreP04’slimitwhen cideswiththe[OIII]interpretationofPO4’sputativeemission attemptingtoreproducetheirphotometricredshiftresultsin§4. lineatλ =1.3375µm. obs In this paper, we address three questions: (i) is #1916 at z∼2- 3?; (ii) is #1916 intrinsically variable?; (iii) does #1916 exist? These tests exploit new optical and mid- inatethe z=10interpretation(e.g.Stern etal. 2000). Incon- infrared observations using the Keck-I 10-m telescope and trast,opticalnon-detectionwouldhaveseveralalternativein- theSpitzerSpaceTelescoperespectively,plusanindependent terpretations,including: agalaxyatz=10asperP04;adusty reduction of P04’s H- and K-band imaging data from VLT. galaxyat z∼2- 3; non-existenceof #1916. The mid-infrared Throughout this article we assume that the emission line at observations (§2.4) should therefore help to constrain the λ =1.3375µmis a false detection(Weatherleyetal. 2004). amountof energyre-radiatedby dustin the z∼2- 3 interpre- obs In §2 we present the new Keck and Spitzer data, explain in tation, and our re-reductionof P04’s VLT data also helps to detailthe re-reductionof the archivalVLT/ISAAC data, and clarify the possibility that #1916 may not exist, or may be summarizethe archivalHubbleSpaceTelescope (HST)data. variable(Bremeretal.2004). TheKeckandSpitzerdatade- Then in §3 we describe the analysis and key results, focus- scribedinthissectionwerecollectedandanalyzedinparallel ing on the three questions posed above; this section closes withthosepresentedbyBremeretal.(2004). withasummaryofthecurrentobservationalstatusof#1916. Finally, we discuss the implications of our results for future 2.1. KeckSpectroscopy ground-basednear-infraredsearchesforgalaxiesatz∼>7(§4) As part of a broad effort to secure spectroscopic redshifts andsummarizeourconclusionsin(§5). ofgravitationalarcsspanningseveralobservationalprograms We assume H0=65kms- 1Mpc- 1, ΩM=0.3 and ΩΛ=0.7 (Smithetal.2001,2002,2005;Sandetal.2002,2004,2005; throughout. Unless otherwise stated all error bars are at Edge et al. 2003; Sharon et al. 2005), we observed A1835 1σ significance; photometric detection limits are at 3σ sig- withtheLowResolutionImagerSpectrograph(LRIS;Okeet nificance; upper and lower limits on colors are based on al.1995)inmulti-slitmodeontheKeck-I10-mtelescope4on 3σ detection thresholds in the non-detection filter. Mag- UT2004March29.Asinglemaskwasobserved,includinga nitudes are stated in the AB system; conversion between slittargeting#1916.Thepurposeofthisslitwastosearchfor the AB and Vega systems for the specific filters used in line emission in the 0.35≤λ ≤0.95µm wavelength range. this paper are as follows:- ∆B=BAB- BVega=- 0.1, ∆V=0.1, For example, if #1916 doesoibnsdeed lie at z≃2.6 (§1), then ∆∆R3.=6µ0m.2=,2∆.8,F∆7024W.5µ=m0=.33,.2∆.I=0.5, ∆J=0.9, ∆H=1.4, ∆K=1.9, LyTmhaen-oαbsmeravyatbieondsetteoctatalebdle3a.t6λ-kobss,≃sp0l.i4t4iµnmto.two exposures, using the D560 dichroic with the 400/8500 grating and the 2. OBSERVATIONSANDDATAREDUCTION 400/3400grism. On the redside the spectraldispersionwas We describe new and archival observations of #1916 in 1.86Å/pixel with a pixel scale of 0.214′′/pixel and on the order of increasing wavelength, spanning the observed opti- bluesidethespectraldispersionwas1.09Å/pixelwithapixel cal, near-infraredand mid-infrared: spectroscopywith LRIS scale of 0.135′′/pixel. Overhead conditions were moderate on Keck-I (0.35≤λ ≤0.95µm); imaging with WFPC2 on- obs (FWHM≃1′′),andprobablynotphotometric,howeveraflux boardHST (λ =0.7µm);near-infraredimagingwithISAAC obs on ESO’s VLT (λ =1.6µm and 2.2µm); IRAC/Spitzer ob- obs 4 The W. M. Keck Observatory is operated as a scientific partnership servationsatλ =3.6µmand4.5µm. obs amongtheCalifornia Institute ofTechnology, theUniversity ofCalifornia, The detectionof any opticalflux from #1916would elim- andNASA. Smith,G.P.,etal. 3 H (1.6um) K (2.2um) 3.6um FIG. 1.—Infraredimagesofthez=10candidateat1.6,2.2,and3.6µmrespectively. Thetwoleftpanelsarebasedonourindependentre-reductionofP04’s ISAACdatadescribedin§2.3. Thewhitecirclesmarkthepositionof#1916fromP04–thereisnoobvioussignoffluxinanyofthesepanels. Formal3σ detectionlimitsare:H≥25,K≥25,F≤0.75µJy.NorthisupandEastleft.Eachpanelis23′′×16′′. calibration was obtained using the spectrophotometric stan- Deep near-infrared imaging of A1835 was obtained by dard star HZ44 (Oke et al. 1990). The data were de-biased, P04 with the ISAAC 1024×1024Hawaii Rockwellarrayon flat-fielded,sky-subtracted,extractedandcalibratedinastan- ESO’s 8-m VLT7 in February 2003. We reduced indepen- dardmannerwithinIRAF5. dently the H- and K-band data using standard IRAF tasks, No flux at observed optical wavelengths has yet been de- paying careful attention to rejection of cosmetic defects in tected at the position of #1916 (P04; Lehnert et al. 2004; theISAACarray,includingbadpixels,andtoconservingthe §2.2). We therefore did not expect to detect any contin- noisepropertiesofthedata.Wedetailkeyfeaturesofthedata uumemission,andconcentratedinsteadonsearchingforfaint reductionbelow. emissionlinesoflargeequivalentwidth. Visualinspectionof thereduced2Ddatarevealedneithercontinuumnorlineemis- (i) Flat-fielding and sky-subtraction were combined into a sion. Toestimatethesensitivitylimitweextracteda1Dtrace singlestepusingamedianoftheeightframestemporally fromthe center of the slit correspondingto the fullwidth of adjacentto eachscienceframe. We referto thisstep as theseeingdisk,andestimatedthe3σ detectionlimitper5-A “flat-fielding”,andtotherollingtemporalmedianframes spectral resolution element to be ∼4.5×10- 19ergs- 1cm- 2 at as“sky-flats”. λ =0.44µm–i.e.theobservedwavelengthatwhichLyman- obs (ii) Flat-fielding was performed twice. First on the dark- αwouldbefoundif#1916isatz=2.6. subtracted frameswhich were then registered and aver- 2.2. ArchivalHubbleSpaceTelescopeImaging aged to producea first-pass reducedframe. This frame was then used to mask out flux from identified sources A1835 has also been observed through the F702W filter from the individual dark-subtracted frames, and these withtheWFPC2cameraon-boardHST6.Wereferthereader maskedframeswerethenusedtoconstructthesky-flats toSmithetal.(2005)fordetailsofthesedataandtheirreduc- in a second-pass reduction. This approach minimizes tion.P04donotdetect#1916inthesedata(Table1),although the loss of flux from objects with angular extents com- theyneitherexplainhowtheyreducedthedatanorhowthede- parable with the size of the dither pattern. This is im- tectionlimitwascalculated.Here,weuseSmithetal.’s(2005) portantwhen searching for faint objects along lines-of- reducedframewhichhasapixel-scaleof0.0498′′/pixelafter sight throughthe crowded cores of rich galaxy clusters drizzling. To simplify the analysis, we re-bin the data back becausethelightfrombrightclustergalaxieseffectively to the original pixel-scale of 0.0996′′/pixel to minimize the form a spatially varying background against which the impactofpixel-to-pixelcorrelationsinthebackgroundnoise faint sourcesare detected. The goalof the second-pass whenestimatingthesensitivitylimitofthedata. flat-fieldistoconservethis“background”. Visual inspection reveals no obvious flux at the position of #1916 in these data. To quantify this non-detection, we (iii) Independent bad pixel masks were made by sigma- follow the same procedureas P04 – we measured the back- clipping both the darks and the sky-flats. The former groundnoiseinaperturesplacedrandomlyintoblankskyre- identifies22,020pixels(2.1%ofthetotalarray)asstatic, gionsofbothframesneartothepositionof#1916.Weensure i.e.“bad”,andthelatteridentifiesthesame22,020pix- consistencywithotherwavelengthsbymatchingthediameter els plus an additional 12,373 pixels (a further 1.2% of of these apertures to a diameter 3× that of the seeing disk, the total array)as bad. The latter mask was adoptedas i.e. 0.5′′. We obtaina 3σ sensitivity limit in thatapertureof thefiducialbadpixelmask. R =27.0(Table1). 702 (iv) Detectorbiasresidualswereremovedbysubtractingthe 2.3. ArchivalVLT/ISAACImaging medianalongrowsinindividualflat-fieldedframesafter masking identified sources, in a manner similar to that describedbyLabbéetal.(2003). 5 IRAFisdistributedbytheNationalOpticalObservatories,whichisop- erated by the Association of Universities for Research in Astronomy, Inc. (v) The individualframes were integer pixel aligned. This (AURA)undercooperativeagreementwiththeNationalScienceFoundation. 6BasedinpartonobservationswiththeNASA/ESAHubbleSpaceTele- has the important benefit of minimizing pixel-to-pixel scopeobtainedattheSpaceTelescopeScienceInstitute, whichisoperated bytheAssociationofUniversities forResearchinAstronomy,Imc.,under 7 Basedinpartonobservations collected withtheESOVLT-UT1Antu NASAcontractNAS5-26555. Telescope. 4 Non–detectionofthez=10Galaxy correlations in the noise properties of each frame and 3.1. Is#1916atz∼2- 3? thus the final stacked frame. Calculation of the back- This test concentrates on the optical data because the de- groundnoise is therefore simplified relative to a reduc- tection of any flux shortward of the putative Lyman limit of tionschemebasedonsub-integerpixelalignmentofthe a galaxy at z≃10 would immediately discount that interpre- individualframes. tation. The red observed optical/near-infrared spectral en- ergydistributiondescribedbyP04couldthenbenaturallyex- (vi) No frameswere rejected whenmakingthe final combi- plainedbyadustygalaxyatz∼2- 3, perhapsassociatedwith nation of the reduced, aligned frames. Two versionsof the SMGsthat lie within ∼30′′ (∼200- 300kpcin projection thefinalstackedframeweremade:astraightaverageand atz∼2- 3)of#1916(Ivisonetal.2000;Smailetal.2005). a weightedaverage– the weightofan individualframe Ournewnon-detectionof#1916withLRIS(§2.1),coupled was proportionalto (σ×rms)- 2, where σ is the FWHM with confirmationof P04’s non-detectionwith HST/WFPC2 of the seeing disk, and rms is the rootmean square per and Lehnert et al.’s non-detection in the V-band with pixelofthenoiseineachframe.Theweightedversionof VLT/FORSaremutuallyconsistentinthesensethatnooptical thefinalframehasslightlybetterimagequalitythanthe fluxhasbeendetectedatthispositiontodate. Howeverthese straight average. We therefore adopt it for the analysis non-detectionsareconsistentwithallofthefollowing: z=10, describedbelow. extreme dust obscuration at z∼2- 3, an intrinsically variable source,andnon-existence. Theresultofthistestistherefore The final reduced H- and K-band frames have seeing inconclusive. ofFWHM=(0.45±0.01)′′andFWHM=(0.34±0.02)′′respec- tively.Photometriccalibrationwasachievedwiththestandard 3.2. Is#1916IntrinsicallyVariable? star observationsthat were interspersedwith the science ob- TheobjectiveofthissectionistotestBremeretal.’s(2004) servationsaspartofP04’soriginalprogram.Weshowextracts proposalthat#1916isintrinsicallyvariable.IfP04’sphotom- from the reduced H- or K-bandframes in Fig. 1. Visual in- etry(H=25.00±0.25andK=25.51±0.51)isreproducibleus- spectionofboththefitsframesandFig.1revealsnoobvious ingourindependentreductionoftheirnear-infrareddata,then fluxatthelocationof#1916ineithertheH-orK-band. Fol- lowingP04weagainrandomlyinsert1.5′′diameterapertures the variable hypothesiswould be supported. If not, then the ideathat#1916doesnotexistwouldgaincredibility(§3.3). (roughly3timestheseeingdisc)intoblankskyregionsnear We attempt to reproduce P04’s analysis using SExtractor tothepositionof#1916,obtaining3σsensitivitylimitsinthis (Bertin & Arnouts 1996). SExtractor was configured to lo- apertureof:H=25.0andK=25.0. cate all sources with at least 7 pixels that are ≥0.75σ per pixel above the background – i.e. a signal-to-noise ratio of 2.4. Spitzer/IRACImaging ∼>2 perresolutionelement, basedon the H-bandseeingdisk A1835 was observed with the InfraRed Array Camera ofFWHM=0.45±0.01′′(§2.3)andthe0.15′′/pixscaleofthe (IRAC;Fazioetal.2004)on-boardSpitzer8onUT2004Jan- ISAACpixels.Wealsosmoothedthedatawithagaussianfil- uary 16 in the 3.6, 4.5, 5.8 and 8.0µm channels. Here we terthatmatchedtheFWHMoftheobservedpointsources,i.e. discussthetwoshortestwavelength,moresensitive,observa- agaussianofFWHM=3pixels.InthisconfigurationSExtrac- tions. Twelve and eighteen 200-second exposures were ac- torfailedtodetectasourceatthepositionof#1916.Wethere- cumulatedat3.6µmand4.5µmrespectively,usingthesmall- foreexperimentedwithdifferentsmoothingschemes,bothin- stepcyclingditherpattern.TheBasicCalibratedData(BCD) creasinganddecreasingthefullwidthofthegaussianfilter. A were combined using custom routines to produce the final “detection”wasonlypossiblewiththesmallestavailablefil- stackedframewithapixelscaleof0.6′′/pixel. Were-binned ter–FWHM=1.5pixels,i.e.halfthewidthoftheseeingdisk thedatabacktotheoriginalpixelscaleof1.2′′/pixeltoelim- – yielding H=25.3±0.6. Experimentation with block filters inatecorrelationsinthebackgroundnoise. producedsimilarresultsinthata“detection”wasnotpossible Visual inspection of the final frames again indicates that withanyofthestandardSExtractorblockfilters: 3×3,5×5, thereisnofluxatthepositionof#1916(Fig.1). Toquantify 7×7pixels. WealsoanalyzedtheK-banddatainexactlythe this non-detection we follow the same procedure as P04, as same mannerand failed to detectanythingat the position of describedin §2.2. We used 5.1′′ diameter apertures, i.e. 3× #1916withanygaussianorblockfilter. theseeingdiskoftheIRACobservationstoobtain3σ sensi- TheH-bandsegmentationmapproducedwhensmoothing tivity limits of: F(3.6µm)=0.75µJy and F(4.5µm)=0.75µJy with the FWHM=1.5 pixel gaussian reveals that the “detec- respectively. tion”isveryelongated,withawidthof1- 2pixelsandalength of∼5pixels.Theorientationofthesepixelsisconsistentwith the orientation of #1916 reported by P04. It is importantto 3. ANALYSISANDRESULTS stress thatthe motivationfor filteringdata with a kernelthat Theobjectiveofthissectionistoanswerthethreequestions matchestheresolutionelementofthedataistosuppressfalse posedin§1:(i)is#1916atz∼2- 3?;(ii)is#1916intrinsically detections. The collection of pixels identified by SExtractor variable?; (iii) does #1916 exist? Preliminary inspection of atthepositionof#1916wasonly“detectable”withasmooth- the data in §2 indicates that no flux is detected at the posi- ing kernel that has a linear scale half that of the resolution tionof#1916atanywavelengthtodate.Combiningthiswith elementofthedata.Itisthereforeinstructivetoconsiderhow Bremeretal.’smoresensitivenon-detectionofH(3σ)>26.0, manysuch∼2σblobsexistwithintheISAACdata.Inasingle itistemptingtoleaptothethirdquestionandreply“no”.We 1.5′′ diameter aperture (i.e. matching that used for the pho- adoptamoreconservativeapproach. tometry described above) placed randomly in these H-band data, there is a 5% chance of detecting a 2σ noise fluctua- 8ThisworkisbasedinpartonobservationsmadewiththeSpitzerSpace tion – i.e. a spurious detection. However the ISAAC array Telescope, which is operated bythe Jet Propulsion Laboratory, California InstituteofTechnologyunderacontractwithNASA. (1024×1024pixels, each pixel0.15′′×0.15′′) containsof or- Smith,G.P.,etal. 5 der104 independentphotometricaperturesof1.5′′ diameter. prior to producingthe final stacked frame, then the noise in TheH-bandframethereforecontains∼500noisefluctuations the stacked frame is correlated. Such pixel-to-pixelcorrela- of2σsignificance. tionsaregenerallyabsentfrominteger-pixelaligneddata,thus Sadly, the only reasonable conclusion to draw from this simplifying the noise properties of the final frame. Neither analysis is that #1916 is not detected in our independentre- sub-integer nor integer pixel alignment is intrinsically cor- duction of P04’s data. We therefore place 3σ limits on the rect. The relevantissue is correctmeasurementof the noise fluxatthispositionof:H≥25.0andK≥25.0(§2.3). Theonly ineachcase–thisiscriticaltoassessaccuratelythestatistical wavelengthatwhichtwodirectlycomparableobservationsare significance of sources detected close to the sensitivity limit availableisintheH-band.Combiningournon-detectionwith of the data. Specifically, if the pixel-to-pixelcorrelationsin thatofBremeretal.(2004),weconcludethatthereisnoev- sub-integer pixel aligned data are not included in the error idence for variability of #1916, and that (if it exists) its H- analysis, then the noiseis under-estimatedandthe statistical bandflux isfainterthanH=26at3σ significance(Bremeret significanceofthe detectionover-estimated(Casertanoetal. al.2004). 2000). Weintegerpixelalignedtheindividualframesin§2.3 in order to simplify the error analysis. We now estimate by 3.3. Does#1916exist? how much we would have under-estimated the noise if we Theresultsoftheprecedingtwosectionswerederivedfrom had sub-integerpixelaligned the individualframesand then non-detectionof#1916acrossthebroadestwavelengthrange ignoredthepixel-to-pixelcorrelationswhencalculatingnoise todate: 0.35≤λ ≤5µm. Wenowcombineallofthesenon- level. Thisisachievedbysimplysub-integeraligningthein- obs detections to address the question of whether #1916 exists. dividualframesandre-combiningthemusingaweightedav- The data force us to conclude that there is no statistically erage. Ignoringany resulting pixel-to-pixelcorrelations, we soundevidencethat#1916exists. Thebalanceofprobability obtaina3σthresholdofH=25.2,whichisslightlyfainterthan isthat#1916wasafalse detectionin P04’sdiscoveryobser- ourthresholdofH=25.0. vations. New observational data yielding statistically sound Insummary,itisplausiblethatthedifferencebetweenour detections are required before P04’s claim that #1916 is the near-infrarednon-detectionandP04’sdetectionof#1916us- most distant galaxy yet discovered may be resurrected. We ingthesamerawdatacanatleastinpartbeexplainedbythe considerthisunlikely,butwehopetobesurprisedbyPellóet efficiency of bad pixel identification and treatment of corre- al.’sforthcomingHST observations. latednoise. 4. DISCUSSION 4.2. ImplicationsforFutureWork 4.1. TheNear-infraredNon-detection We now discuss the implications of our results for future Wefirstdiscusspossiblereasonsforthedifferencebetween work, focusingon the feasibility of searchesfor gravitation- our non-detection of #1916 and P04’s ∼3- 4σ near-infrared ally magnified stellar systems at z∼>7 using ground-based detections. Wesingleouttwodatareductionsteps(§2.3)for near-infrared data. We begin by noting that, had their pho- discussion–theefficiencyofbadpixelrejectionandthecon- tometry been reliable, P04’s original z=10 interpretation of servationofnoiseproperties. #1916 was plausible, based solely on the photometric data. We therefore adopt P04’s optical and near-infrared photom- 4.1.1. Efficiencyofbadpixelrejection etry as being representative of what may be expected from Theefficiencywithwhichbadpixelsareidentifiedcanaf- similarfutureexperiments–i.e.opticalnon-detectionsinsev- fect estimates of the signal coming from faint sources – if eral filters based on a few hours of observationswith a 4-m some bad pixels are not identified, they could enhance the class telescope, a non-detection in a single red optical filter flux detected by SExtractor. In §2.3 we used two different withHST,∼3- 4σ detectionsintwonear-infraredfilters,and methodstoidentifybadpixels,findingasmallbutpotentially possiblyobservationswithSpitzer/IRAC.Specifically,wein- importantdifferencebetweenthetwomethods. Toassessthe vestigatethedegeneracybetweenz∼>7interpretationsofsuch impact of reduced efficiency of bad pixel identification we datasetswithlowerredshiftalternatives,andhowsuchdegen- madeamaskimagethatcontainedthe12,373badpixelsiden- eraciesmightbebroken. tifiedonlyinthebadpixelmaskgeneratedfromthesky-flats. We use Version 1.1 of HYPER-Z9 (Bolzonella et al. 2000) We then made one copy of the mask per science frame and to fit standard Bruzual & Charlot (1993) single stellar pop- integer-pixel shifted them to match the observed dither pat- ulation models (Burst, E, Sa, Sc, Im) to the photometric tern. Finally we took the weighted average of these aligned data. We assume a Calzetti et al. (2000) extinction law, al- maskframesandsummedthepixelcountsina1.5′′diameter low dust extinction in #1916 to lie in the range 0≤A ≤4, V aperture centered on #1916. From this we concluded that 7 adopt E(B- V)=0.03 for extinction within the Milky Way badpixelsthatareonlyidentifiedin our“sky-flat”badpixel (Schlegel et al. 1998), and use Madau’s (1995) prescription mapsfallwithinthefinalphotometricaperture. We estimate for absorption by the inter-galactic medium. In each case that if not identified and excluded from the analysis, these described below, if we fit the models to all available photo- pixelscouldincreasethefluxestimatesbyseveraltenthsofa metric information across the full redshift range (0≤z≤11), magnitude. thenweobtainχ2∼<1(allχ2 valuesarequotedperdegreeof freedom) for all redshifts beyond z≃7. This is because the 4.1.2. Conservationofnoiseproperties modelsaredefinedtohavezerofluxshort-wardoftheLyman Theapproachtore-sampling(ornot)theindividualframes limit (λrest=0.0912µm). At z∼>6, the models also have neg- during the data reduction process, especially the alignment ligible flux short-ward of Lyman-α(λ =0.1215µm)due to rest of individual frames affects the noise properties of the fi- theLymanforestsandGunnPetersonabsorption. Thesetwo nal stacked frame. If the individual frames are re-sampled, for example by sub-integerpixel aligningthem immediately 9Availablefromhttp://webast.ast.obs-mip.fr/hyperz 6 Non–detectionofthez=10Galaxy FIG. 2.— LEFT: Reducedχ2 asafunctionofphotometricredshiftandthefiltersusedinthemodelfits. Thereddashedcurvehasbeenoffset+0.2inthe Y-directionforclarity. ThispaneldemonstratesthatP04’sdiscoveryphotometry(VRIJHK)isdegeneratebetweenz≃2.6andz=10(bluesolidcurve). Adding insensitivespace-baseddetectionthresholdsfromHST andSpitzer/IRAC(reddashedandblackdot-dashedcurvesrespectively)liftsthisdegeneracy. RIGHT: Thebest-fitspectraltemplatestotheR702JHK-bandphotometryatz=2.6(redsolid)andz=10(bluedashed). TheIRACdetectionthresholdsarealsomarkedto illustratethepowerofthesedatatodiscriminatebetweenlow-andhigh-redshiftinterpretationsoftheshorterwavelengthdata. spectralfeaturesareprogressivelyredshiftedlong-wardofthe HST dataisnowclearlyevidentinthenewχ2distribution,as observedopticalfiltersathighredshift,renderingtheshorter showninFig.2. TheonlyacceptablefittotheR JHK-band 702 wavelength detection limits irrelevantto the fits. The good- dataisatz≃10. ness of fit is therefore systematically over-estimated at high We show the best fit SEDs at z=2.6 and z=10, based on redshift if all photometric information is included in the fit the R JHK-band data in the right panel of Fig. 2. This 702 acrossthe fullredshiftrange. We thereforefit the modelsto demonstratesthepotentialpowerofIRACphotometrytofur- thedatainaseriesofredshiftchunks–onlythephotometric ther discriminate between low and high-redshift interpreta- informationthatlieslongwardofred-shiftedLyman-αiscon- tions of candidate high-redshift galaxies. The SED of low sidered in each redshift chunk. The results described below redshiftsolutions(e.g.z≃2.6)isredatλ ∼3- 5µm,andthe obs are insensitive to whether the Lyman limit is substituted for SEDofhighredshiftsolutions(z≃10)isblueatsimilarwave- Lyman-α. lengths. We addthedetectionlimitsobtainedatλ =3.6µm obs First, we fit the models to P04’s VRIJHK-band ground- and 4.5µm to the R JHK-band data and re-fit the spectral 702 basedphotometry(i.e.theVRIJ-banddetectionlimitsandthe templates.TheresultisshownintheleftpanelofFig.2–the HK-banddetectionslistedincolumn4ofTable1). Notethat reducedχ2 of the z≃2.6 solution is now in excess of 2, and P04useddifferentsizedaperturesforthenear-infrared(1.5′′) theonlyredshiftswhichachieveanacceptablefittothe data and optical (0.6′′) photometryrespectively. We find two ac- are9.5∼<z∼<11.5. ceptablesolutions:z≃2.6andz≃10(Fig.2),neitherofwhich In summary, a single sensitive non-detection derived require any dust-obscurationwithin #1916, the latter having from HST imaging, combined with deep ground-based the lower formal χ2 value. We also scale P04’s ground- near-infrared imaging and ∼2,400 second integrations with basedopticalnon-detectionstothatappropriatefora consis- Spitzer/IRACat3.6and4.5µmcanbreakthedegeneracybe- tentphotometricapertureofthreetimestheseeingdiskatall tweenlowandhigh-redshiftinterpretationsofcandidatez∼>7 wavelengths(V≥26.2,R≥26.3,I≥24.7).Thismarginallyim- stellarsystems(seealsoEgamietal.2004;Eylesetal.2005). provesthe goodnessof fit of the z≃2.6 solution, and is oth- Whilstthisisnotanexhaustivestudy,itdemonstratesthatde- erwise indistinguishable from the original fit. We retain the spitethedemiseof#1916,searchesforgravitationallymagni- matchedphotometricaperturesfortheremainderofthisanal- fied galaxiesat extremelyhigh redshiftsusing ground-based ysis. near-infrareddata remain feasible when combinedwith sen- To improve the discrimination between low- and high- sitivespace-basedopticalandmid-infrareddata. redshift solutions, we add the sensitive HST/WFPC2 de- tection threshold of R ≥27.0 (Table 1) to the photomet- 702 5. CONCLUSIONS ric constraints. We fit the synthetic SEDs to the combined VRR IJHK dataset. Thegoodnessoffit of the z≃2.59so- We have analyzed new and archival observations of the 702 z=10 galaxy #1916 behind the foreground galaxy clus- lutionismarginallyworserelativetotheVRIJHK-bandanal- ysis, becausethe moststringentopticalnon-detectioncomes ter lens A1835 (z=0.25) spanning 0.35≤λobs≤5µm. Sta- from the R -band. However, in general, the χ2 as a func- tistically significant flux is not detected in any of these 702 data, including our independent re-analysis of P04’s dis- tion of redshift is indistinguishable between this fit and the VRIJHK-bandfits. Thisis probablybecausethe χ2 is dom- covery H- and K-band data. The 3σ detection thresh- inatedby the non-detectionsin six outof the eightobserved olds are: F(λobs=0.44µm)∼>4.5×10- 19ergs- 1cm- 2 per spec- filters. To test this we re-fit the models, limiting the data to tral resolution element, R702≥27.0, H≥25.0, K≥25.0, justtheR JHK-bands,i.e.themostsensitivenon-detection F(3.6µm)≤0.75µJyandF(4.5µm)≤0.75µJy,wherethepho- 702 (R ), the reddest non-detection (J) and the two detections tometriclimitsarecalculatedinapertureswithadiameter3× 702 (HK). The impact of the sensitive detection limit from the thatoftheseeingdisk. CombiningtheseresultswiththoseofBremeretal.(2004) Smith,G.P.,etal. 7 and Weatherley et al. (2004), we are thereforeforcedby the with10-mclasstelescopesremainapowerfultoolforthedis- data to concludethat there is no statistically soundevidence covery of intrinsically faint galaxies (L∼<0.1L⋆) at z>7 that fortheexistenceof#1916.Wealsoshowthatinefficientbad- maybe responsibleforcosmic re-ionization. Futuresurveys pixel rejection and issues relating to the calculation of the shouldcombinethesedatawithsensitivespace-basedoptical backgroundnoisecanbroadlyaccountforthedifferencesbe- andmid-infraredobservations. tweenP04’snear-infraredphotometryandourownusingthe samedata. Thebalanceofprobabilityisthereforethat#1916 ACKNOWLEDGMENTS was a false detectionin P04’sanalysis. Froma broaderper- We acknowledge the bold efforts of Roser Pelló and col- spective, the demise of #1916warns of the hazards of oper- laborators, and GPS acknowledges cordial discussions with ating close to the detection threshold of deep ground-based RoserinLausanneduringthesummerof2004.Specialthanks near-infrareddata. gotoJean-PaulKneibformakingtherawnear-infraredimag- The need for gravitational magnification to boost the ob- ing data available. GPS acknowledges the Caltech Opti- servedfluxoffaint(L∼<0.1L⋆)galaxiesatz∼6–8andpopula- cal Observatories TAC for enthusiastic support, and thanks tionsofgalaxiesatstillhigherredshifts(see§1) isundimin- RichardEllisandAvishayGal-Yamforhelpfulcommentson ished by our results on #1916. However an important issue draftsofthemanuscript. ThanksalsogotoMalcolmBremer, is whether it is feasible to find such galaxies using ground- Michael Cooper, Joe Jensen, Dan Stark, Chuck Steidel and based facilities. We explore this issue using HYPERZ to fit Dave Thompson for a variety of discussions and assistance. syntheticspectraltemplatestorepresentativedata. Ourmain DJS thanks Dawn Erb, Alice Shapley, Naveen Reddy and conclusionsarethatdeepnear-infraredimagingsimilartothat Tommaso Treu for assistance with the LRIS data reduction, presented by P04 in combinationwith a single sensitive op- and acknowledges financial support from NASA’s Graduate tical non-detection from HST imaging and moderate depth StudentResearchProgramunderNASAgrantNAGT-50449. Spitzer/IRACimagingat3.6and4.5µmcandiscriminatebe- DSandPRMEacknowledgesupportfromNASA.Werecog- tween low (e.g. z∼2- 3) and high (e.g. z∼>7) redshift solu- nizeandacknowledgetheculturalroleandreverencethatthe tionswithstrongstatisticalsignificance.Insummary,ground- summit of Mauna Kea has within the Hawaiian community. based near-infraredsurveysof massive galaxycluster lenses Wearefortunatetoconductobservationsfromthismountain. REFERENCES BargerA.J.,CowieL.L.,CapakP.,AlexanderD.M.,BauerF.E.,Fernandez KneibJ.-P.,EllisR.S.,SantosM.R.,RichardJ.,2004,ApJ,607,697 E.,BrandtW.N.,GarmireG.P.,HornschemeierA.E.,2003,AJ,126,632 KneibJ.-P.,vanderWerfP.P.,KraibergKnudsenK.,SmailI.,BlainA.W., Becker R.H., Fan X., White R.L., Strauss M.A., Narayanan V.K., Lupton FrayerD.,BarnardV.,IvisonR.,2004,MNRAS,349,1211 R.H.,GunnJ.E.,AnnisJ.,BahcallN.A.,BrinkmannJ.,etal.,2001,AJ, KogutA.,SpergelD.N.,BarnesC.,BennettC.L.,HalpernM.,HinshawG., 122,2850 Jarosik N., Limon M., Meyer S.S., Page L., Tucker G.S., Wollack E., BolzonellaM.,MirallesJ.-M.,PellóR.,2000,A&A,363,476 WrightE.L.,2003,ApJS,148,161 BlainA.W.,ChapmanS.C.,SmailI.,IvisonR.,2004,ApJ,611,725 Labbé I., Franx M., Rudnick G., Förster Schreiber N.M., Rix H.-W., BorysC.,ChapmanS.,DonahueM.,FahlmanG.,HalpernM.,KneibJ.-P., Moorwood A., van Dokkum, P.G., van der Werf P., Röttgering H., NewburyP.,ScottD.,SmithG.P.,2004,MNRAS,352,759 vanStarkenburgL.,etal.,2003,AJ,125,1107 Bouwens R.J., Thompson R.I., Illingworth G.D., Franx M., van Dokkum LehnertM.D.,FörsterSchreiberN.M.,BremerM.E.,2004,ApJ,submitted, P.G.,FanX.,DickinsonM.E.,EisensteinD.J.,RiekeM.J.,2004,ApJ,616, astro-ph/0412432 L79 BremerM.N.,JensenJ.B.,LehnertM.D.,Förster-Schreiber N.M.,Douglas MadauP.,1995,ApJ,441,18 L.,2004,ApJ,615,L1 MellierY.,FortB.,SoucailG.,MathezG.,CaillouxM.,1991,ApJ,380,334 BruzualA.G.,CharlotS.,1993,ApJ,405,358 OkeJ.B.,1990,AJ,99,1621 BunkerA.J.,StanwayE.R.,EllisR.S.,McMahonR.G.,2004,MNRAS,355, OkeJ.B.,CohenJ.G.,CarrM.,CromerJ.,DingizianA.,HarrisF.H.,1995, 374 PASP,107,375 Calzetti D., Armus L., Bolin R.C., Kinney A.L., Koornneef J., Storchi- PellóR.,SchaererD.,RichardJ.,leBorgneJ.-F.,KneibJ.-P.,2004,A&A, BergmannT.,2000,ApJ,553,682 416,L35 Casertano S., de Mello D., Dickinson M., Ferguson H.C., Fruchter A.S., RichardJ.,SchaererD.,PellóR.,LeBorgneJ.-F.,KneibJ.-P.,2003,A&A, Gonzalez-LopezliraR.A.,HeyerI.,HookR.N.,LevayZ.,LucasR.A.,et 412,57 al.,2000,AJ,1202747 SandD.J.,TreuT.,EllisR.S.,2002,ApJ,574,L129 Edge A.C., Smith G.P., Sand D.J., Treu T., Ebeling H., Allen S.W., van SandD.J.,TreuT.,SmithG.P.,EllisR.S.,2004,ApJ,604,88 DokkumP.G.,2003,ApJ,599,L69 SandD.J.,TreuT.,EllisR.S.,SmithG.P.,2005,ApJ,accepted EgamiE.,Kneib,J.-P.,RiekeG.H.,EllisR.S.,RichardJ.,RigbyJ.,Papovich SantosM.R.,EllisR.S.,KneibJ.-P.,RichardJ.,KuijkenK.,2004,ApJ,606, C.,StarkD.,SantosM.R.,HuangJ.-S.,etal.,2005,ApJ,618,L5 683 EllisR.S.,SantosM.R.,KneibJ.-P.,KuijkenK.,2001,ApJ,560,L119 SchlegelD.J.,FinkbeinerD.P.,DavisM.,1998,ApJ,500,525 Eyles L.,Bunker A.,Stanway E.,Lacy M., Ellis R.S.,Doherty M., 2005, SharonK.,OfekE.O.,SmithG.P.,Broadhurst T.,MaozD.,Kochanek C., MNRAS,submitted,astro-ph/0502385 OguriM.,SutoY.,InadaN.,FalcoE.E.,2005,ApJL,629,73 FanX.,NarayananV.K.,StraussM.A.,WhiteR.L.,BeckerR.H.,Pentericci SmailI.,IvisonR.J.,BlainA.W.,KneibJ.-P.,2002,MNRAS,331,495 L.,RixH.-W.,2002,AJ,123,1247 SmailI.,SmithG.P.,IvisonR.J.,2005,ApJ,inpress,astro-ph/0506176 FanX.,Narayanan V.K.,Lupton R.H.,Strauss M.A.,KnappG.R.,Becker SmithG.P.,KneibJ.-P.,EbelingH.,CsozkeO.,SmailI.,2001,ApJ,552,493 R.H.,WhiteR.L.,PentericciL.,LeggettS.K.,ZoltánH.,etal.,2001,AJ, Smith G.P., Smail I., Kneib J.-P., Davis C.J., Takamiya M., Ebeling H., 122,2833 CzoskeO.,2002,MNRAS,333,L16 FazioG.G.,HoraJ.L.,AllenL.E.,AshbyM.L.N.,BarmbyP.,DeutschL.K., HuangJ.-S.,KleinerS.,MarengoM.,MegeathS.T.,etal.,2004,ApJS, SmithG.P.,KneibJ.-P.,SmailI.,MazzottaP.,EbelingH.,CzoskeO.,2005, 154,10 MNRAS,359,417 Franx M., Illingworth G.D., Kelson D.D., van Dokkum P.G., Tran K.-V., Steidel C.C.,Giavalisco M.,Pettini M.,Dickinson M.E.,Adelberger K.L., 1997,ApJ,486,L75 1996,ApJ,462,L17 HuE.M.,Cowie L.L.,McMahonR.G.,CapakP.,IwamuroF.,KneibJ.-P., Stern D., Eisenhardt P., Spinrad H., Dawson S., van Breugel W.,Dey A., MaiharaT.,MotoharaK.,2002,ApJ,568,L75 deVriesW.,StanfordS.A.,2000,Nature,408,560 IvisonR.J.,SmailI.,BargerA.J.,KneibJ.-P.,BlainA.W.,OwenF.N.,Kerr WeatherleyS.J.,WarrenS.J.,BabbedgeT.S.R.,2004,A&A,428,L29 T.H.,CowieL.L.,2000,MNRAS,315,209 YanH.,WindhorstR.,2004,ApJ,600,L1