MNRAS000,1–10(2015) Preprint13January2017 CompiledusingMNRASLATEXstylefilev3.0 Discovery of the most metal-poor damped Lyman-α system(cid:63) Ryan J. Cooke1,2,5†, Max Pettini3 and Charles C. Steidel4 1UCO/LickObservatory,UniversityofCalifornia,SantaCruz,CA95064,USA 2CentreforExtragalacticAstronomy,DepartmentofPhysics,DurhamUniversity,SouthRoad,DurhamDH13LE,UK 3InstituteofAstronomy,MadingleyRoad,Cambridge,CB30HA 4CaliforniaInstituteofTechnology,MS249-17,Pasadena,CA91125,USA 5RoyalSocietyUniversityResearchFellow 7 1 0 Accepted.Received;inoriginalform 2 n ABSTRACT a J We report the discovery and analysis of the most metal-poor damped Lyman-α (DLA) sys- tem currently known, based on observations made with the Keck HIRES spectrograph. The 1 metal paucity of this system has only permitted the determination of three element abun- 1 dances: [C/H] = −3.43±0.06, [O/H] = −3.05±0.05, and [Si/H] = −3.21±0.05, as well ] as an upper limit on the abundance of iron: [Fe/H] (cid:54) −2.81. This DLA is among the most O carbon-poor environment currently known with detectable metals. By comparing the abun- C dancepatternofthisDLAtodetailedmodelsofmetal-freenucleosynthesis,wefindthatthe chemistryofthegasisconsistentwiththeyieldsofa20.5M metal-freestarthatendedits . (cid:12) h lifeasacore-collapsesupernova;theabundanceswemeasureareinconsistentwiththeyields p of pair-instability supernovae. Such a tight constraint on the mass of the progenitor Popu- o- lation III star is afforded by the well-determined C/O ratio, which we show depends almost r monotonicallyontheprogenitormasswhenthekineticenergyofthesupernovaexplosionis st Eexp (cid:38)1.5×1051erg.WefindthattheDLApresentedherehasjustcrossedthecritical‘transi- a tiondiscriminant’threshold,renderingtheDLAgasnowsuitableforlowmassstarformation. [ Wealsodiscussthechemistryofthissysteminthecontextofrecentmodelsthatsuggestsome 1 ofthemostmetal-poorDLAsaretheprecursorsofthe‘firstgalaxies’,andaretheantecedents v oftheultra-faintdwarfgalaxies. 3 Keywords: quasars:absorptionlines–ISM:abundances–stars:PopulationIII–galaxies: 0 1 dwarf 3 0 . 1 0 1 INTRODUCTION is still a matter of debate. Numerical simulations that follow the 7 collapse of primordial material from cosmological initial condi- Almosteveryastrophysicalenvironmentiscontaminatedbythenu- 1 tions originally suggested that Pop III stars were predominantly cleosynthesisofstars.Todate,onlyasmallhandfulofpristineenvi- : very massive, with typical masses in excess of 100 M (Bromm, v ronmentshavebeenidentified,includingamostlyionizedcloudof (cid:12) Xi gasatredshiftz(cid:39)3whichmaybeassociatedwithcoldflows(Fu- C&oNppoir,m&anLa2r0s0o2n)1.9M9o9d;eNrankcaamlcuurlaat&ioUnsmseumgguerast2a00so1m;Aewbehla,tBlroywaenr, magalli,O’Meara,&Prochaska2011),andamostlyneutralcloud r typical mass of the first stars, and indicate that a small multiple a of gas at redshift z (cid:39) 7 attributed to either a neutral protogalaxy ofPopIIIstarsareformedinagivenminihalo(Clark,Glover,& ortheintergalacticmedium(Simcoeetal.2012;butseeBosman Klessen2008;Turk,Abel,&O’Shea2009;Stacy,Greif,&Bromm &Becker2015foranalternativeinterpretation).Ifpocketsofab- 2010;Stacy&Bromm2013;Hiranoetal.2014;Stacy,Bromm,& solutelypristinegasstillexistatredshiftz ∼ 3,theremayalsobe Lee2016).Althoughtheformofthemassfunctionhasnotyetbeen someenvironmentsthatwereenrichedexclusivelybythefirstgen- pinneddown,ageneralconclusionborneoutoftheabovecosmo- erationofstars(alsocalledPopulationIII,orPopIII,stars). logicalsimulationsisthattheinitialmassfunctionofthefirststars ThenatureofPopIIIstars,inparticulartheirmassspectrum, istopheavy,withmostofthetotalstellarmassconcentratedinstars withmassesM(cid:38)10M . (cid:12) (cid:63) BasedonobservationscollectedattheW.M.KeckObservatorywhichis Inprinciple,themassspectrumofthefirststarscanbeinferred operatedasascientificpartnershipamongtheCaliforniaInstituteofTech- observationally,byidentifyingthechemicalfingerprintoftheele- nology,theUniversityofCaliforniaandtheNationalAeronauticsandSpace mentsthatweremadeduringthelifeofaPopIIIstar.Findingthis Administration.TheObservatorywasmadepossiblebythegenerousfinan- cialsupportoftheW.M.KeckFoundation. chemicalsignatureisachallengingprospect,sincearegionmust † email:[email protected] beidentifiedthatissolelyenrichedbythefirststars.Dedicatedsur- (cid:13)c 2015TheAuthors 2 Cookeetal. veystofindputativesecondgenerationstarsintheMilkyWayhave 6530Å,withsmallgapsnear4600Åand5600Å,correspondingto identified several excellent candidates (see Frebel & Norris 2015 thegapsbetweenthedetectormosaic.Allframeswerebinned2×2 forareview).Thesesearcheshavealsouncoveredastrikingdiver- duringreadout. sity of chemical abundance patterns, including many metal-poor The data were processed with the makee data reduction starsenhancedwithlightelementsrelativetoheavyelements(col- pipeline1.Thispipelinefirstsubtractsthedetectorbiaslevel.The lectivelyknownascarbon-enhancedmetal-poor[CEMP]stars;for locations of the echelle orders are traced using an exposure of a anoverview,seeBeers&Christlieb2005),astarwithapparently quartz lamp through a pinhole decker (HIRES decker D5). The noiron(Kelleretal.2014),andanalmostpristinestarwithroughly pixel-to-pixel variations and the blaze function were removed by solar-scaledchemicalabundances(Caffauetal.2011).Theabun- dividing the science frames by an exposure of a quartz lamp dancepatternsofthesestarsaresuccessfullyreproducedbymodels throughthescienceslit(deckerC1).Aone-dimensionalspectrum ofPopIIIstellarnucleosynthesis(Heger&Woosley2010;Cooke was optimally extracted from each reduced frame, and the data &Madau2014;Ishigakietal.2014;Marassietal.2014;Tominaga, were finally calibrated to a vacuum and heliocentric wavelength Iwamoto,&Nomoto2014),however,thereisstillsomeflexibility scale.Theindividualspectrawereresampledandcombinedusing inthemodelsduetotheuncertainexplosionmechanismofcore- the uves popler code2. Deviant pixels and ghosts were identified collapsesupernovae(seee.g.Janka2012). byvisualinspection,andmaskedpriortocombination. Asimilarquesthasbeenundertakentoidentifythechemical TheabsorptionlineswereanalyzedusingtheAbsorptionLIne signatureofthefirststarsamongthemostmetal-poordampedLyα Software(alis)package3,whichusestheatomicdatatabulatedby (DLA) systems (Erni et al. 2006; Pettini et al. 2008; Penprase et Morton(2003).BoththequasarcontinuumemissionandtheDLA al. 2010; Cooke et al. 2011a,b; Cooke, Pettini, & Murphy 2012). absorptionlineswerefitsimultaneously.Thequasarcontinuumwas DLAsarecloudsofmostlyneutralgasthatareobservedinabsorp- modeledwithaloworderLegendrepolynomiallocallyaroundeach tion against a bright background source, typically a quasar (for a absorptionline,whileeachDLAabsorptioncomponentwasmod- generalreviewofDLAs,seeWolfe,Gawiser,&Prochaska2005). eledwithaVoigtprofile.TheLyαprofileoftheDLAispresented DLAsaretypicallyobservedatredshiftz(cid:39)3(∼2GyraftertheBig in Figure 1 (black histogram) overlaid with the best fitting Voigt Bang)wheremanyoftherestframeultravioletabsorptionlinesof profile (red curve), corresponding to a neutral hydrogen column interestareconvenientlyredshiftedintotheopticalspectralrange. densityoflogN(Hi)/cm−2=20.32±0.05. The absorption lines associated with the most metal-poor DLAs We detect metal absorption lines of Cii, Oi and Siii, repro- aretypicallyveryweak,owingtothelowmetalabundance;insuch duced in Figure 2 together with the best fitting model profiles. systems,onlythemostabundantelementscanbereliablymeasured The absorption profiles can be described by three components, withcurrenttelescopefacilities(Cookeetal.2013). located in two distinct velocity intervals. The primary absorption Themostmetal-poorDLAsarethoughttobetheantecedents componentislocatedatz = 3.077588±0.000002,andexhibits abs of the lowest mass galaxies (Salvadori & Ferrara 2012; Webster, a turbulent Doppler parameter of b = 9.4 ± 0.2 km s−1. The 1 Bland-Hawthorn,&Sutherland2015;Cooke,Pettini,&Jorgenson two additional ‘satellite’ components are located at a velocity of 2015), as well as cold gas streams being accreted onto galaxies ∆v (cid:39) −68.9kms−1 and∆v (cid:39) −93.1kms−1 relativetothepri- 2 3 (Yuan & Cen 2016). Such an affiliation marks these DLAs as a marycomponent,withDopplerparametersb =10.8±0.5kms−1 2 promisingenvironmenttomeasure–athighredshift–thechemical and b = 15.9±0.7 km s−1, respectively. In addition to the tur- 3 abundancepatternoftheearliestgenerationofstars. bulent broadening quoted above, we assume that the line profiles In this paper, we present the discovery of the most metal- ofallabsorptioncomponentsarethermallybroadenedbygasata poorDLAcurrentlyknown,anddiscussthechemistryofthisnear- kinetictemperatureofT =104K(seeCooke,Pettini,&Jorgen- kin pristineenvironment.InSection2,wedescribethedetailsofour son2015)4.SomeoftheDLAabsorptionlinesshowninFigure2 observations and absorption line profile fitting. We discuss the aremildlyblendedwithabsorptionunrelatedtotheDLA.Wesi- chemistry of this newly discovered DLA in Section 3, and sum- multaneouslyfittheDLAabsorptionandtheunrelatedabsorption, marisethemainconclusionsofourworkinSection4. anddisplaythemodelprofileoftheunrelatedblendsinFigure2, representedbythelightpurpleline. Thecolumndensitiesofeachcomponentofthedetectedions arelistedinTable1,andaremostlydrivenbyCiiλ1334,Oiλ1302, 2 OBSERVATIONSANDANALYSIS andSiiiλ1260.Wealsolista3σupperlimitontheFeiicolumn Among the DLAs discovered by the Sloan Digital Sky Survey density, which is based on the non-detection of Feiiλ1260, inte- (Noterdaemeetal.2009),wefirstidentifiedtheextremelymetal- grated over the velocity interval of Component 1. The errors on poor DLA towards the z = 3.22 quasar J0903+2628 (R.A. = thecolumndensitymeasurementsarecomputedusingthediagonal em 09h03m33s.55,decl.=+26◦28(cid:48)36(cid:48).(cid:48)3;m =19.0)basedonthestrong, r dampedLyαabsorptionfeatureatredshiftz = 3.076,combined abs withthenon-detectionofthestrongestassociatedabsorptionlines 1 makeeisavailablefordownloadfrom: ofC,OandSi(seePettinietal.2008,foradescriptionofthetech- http://www.astro.caltech.edu/∼tb/ipacstaff/tab/makee/index.html nique). 2 uvespoplerismaintainedbyMichaelT.Murphy,andisavailablefrom We observed J0903+2628 with the Keck High Resolution thefollowingurl: http://astronomy.swin.edu.au/∼mmurphy/UVESpopler Spectrometer (HIRES; Vogt et al. 1994) on 2016 March 1, 2, 30 3 alisisavailablefordownloadfrom: foratotalexposuretimeof9×3600s,inseeingconditions(full https://github.com/rcooke-ast/ALIS widthathalfmaximum,FWHM(cid:39)0.8(cid:48)(cid:48))thatwerewell-matchedto thechosenslitwidth(0.861(cid:48)(cid:48);deckerC1).Thenominalinstrument 4 WenotethatthechoiceofTkin=104Kdoesnotimpactthederivedcol- umndensities,sincethethermalDopplerparameterismuchnarrowerthan resolutionofoursetupisR (cid:39) 49,000,assumedtobeaGaussian theturbulentDopplerparameterand,inanycase,allofthefittedabsorption profilewithafullwidthathalfmaximumofvfwhm=6.1kms−1.We linesareonthelinearpartofthecurveofgrowthwherethederivedcolumn usedthebluecross-dispersertocoverthewavelengthrange3700– densityisindependentoftheDopplerparameter. MNRAS000,1–10(2015) Themostmetal-poorDLA 3 +1.0 x u l F d e z +0.5 i l a m r o N 0.0 1195 1200 1205 1210 1215 1220 1225 1230 1235 Rest Wavelength (˚A) Figure1. TheHiLyαabsorptionlineoftheDLAtowardsJ0903+2628atzabs=3.07759(blackhistogram)isshownwithaVoigtprofilemodeloverlaid(red curve)correspondingtoatotalHicolumndensityoflogN(Hi)/cm−2=20.32.SeveralunrelatedblendsnearthecoreoftheLyαprofilearealsoincludedinthe fit.Theshortdashedgreenlineindicatesthezerolevelofthedata,whilethelongdashedbluelinerepresentsthenormalisedlevelofthequasarcontinuum. 1.0 1.0 1.0 0.5 0.5 0.8 C ii λ1036 C ii λ1334 O i λ1302 0.0 0.0 x u Fl 01..60 1.0 d e z 0.5 0.5 i l a m 0.4 O i λ988 Si ii λ1190 Si ii λ1193 or 0.0 0.0 N 1.0 1.0 0.2 0.5 0.5 Si ii λ1260 Si ii λ1304 Si ii λ1526 0.0 0.0 0.0 0.0 100 50 00.2 +50 1000.4 50 0 0.+650 100 05.80 0 +501.0 − − − − − − Velocity Relative to z =3.077588 (km s 1) abs − Figure2. AselectionofmetalabsorptionlinesoftheDLAtowardsJ0903+2628atzabs = 3.07759(blackhistogram).Ineachpanel,thesolidredcurve representthebestfittingmodelprofile(includingfittedblends).Thelightpurplecurvesshowninsomeofthepanelsrepresentsthemodelprofileofjust thefittedblends(i.e.absorptionthatisunrelatedtotheDLA).ThebestfittingDLAmodelprofileconsistsofthreecomponentslocatedatrelativevelocities v=0.0,−68.9,and−93.1kms−1(labelledComponents1,2and3inthetext,respectively),andindicatedbytheredtickmarksaboveeachspectrum.The red,greenandbluetickmarksintheOiλ988panelindicatethemodelabsorptioncomponentsfortheOitriplet,withrestframewavelengthsλ=988.7734Å, 988.6549Å,and988.5778Å,respectively(notethatone/twoofthegreen/bluetickmarksareofftheplot). MNRAS000,1–10(2015) 4 Cookeetal. Table1.ColumndensitiesforeachcomponentoftheDLAatzabs=3.07759towardsJ0903+2628 X log(cid:15)(X)(cid:12)a Component1 Component2 Component3 Totalb [X/H] [X/O] Hi 12.0 ... ... ... 20.32±0.05 ... ... Cii 8.43 13.05±0.03 12.99±0.04 12.72±0.08 13.32±0.03 −3.43±0.06 −0.38±0.03 Civ 8.43 ... ... ... (cid:54)12.56c ... ... Oi 8.69 13.74±0.02 13.57±0.03 ... 13.96±0.02 −3.05±0.05 ... Siii 7.51 12.40±0.01 12.23±0.03 12.54±0.02 12.62±0.01 −3.21±0.05 −0.16±0.02 Siiv 7.51 ... ... ... (cid:54)12.10c ... ... Feii 7.47 (cid:54)12.76 ... ... ... (cid:54)−2.81d (cid:54)+0.23e alog(cid:15)(X)=12+logN(X)/N(H).SolarvaluesaretakenfromAsplundetal.(2009). bThetotalcolumndensityonlyincludesthecomponentswhicharemostlyneutral(i.e.Component1and2). cTheupperlimitsonN(Civ)andN(Siiv)areintegratedoverthevelocityinterval−100(cid:54)v/kms−1(cid:54)+20relativetoComponent1. dTheupperlimiton[Fe/H]isbasedonthe[Fe/O]limitofComponent1,combinedwiththetotal[O/H]abundanceoftheDLA. eThe[Fe/O]abundanceisbasedonlyonthecolumndensitiesofComponent1. terms of the covariance matrix. We do not detect any absorption −0.38±0.03 and [Si/O] = −0.16±0.02. These abundances are originatingfromthehigherionizationabsorptionlinesofCivand somewhatlowerthanthetypicalvaluesseeninotherverymetal- Siiv.InTable1wequote3σupperlimitsonthecolumndensityof poorDLAs(i.e.DLAswith[Fe/H](cid:54)−2.0),whichhave[(cid:104)C/O(cid:105)]= theseions.Otherhighionizationabsorptionlinesofinterest(such −0.28±0.12 and [(cid:104)Si/O(cid:105)] = −0.08±0.10 (Cooke et al. 2011b), asCiiiλ977andSiiiiλ1206)areblendedwiththeLyαforest. wherethequoteduncertaintyrepresentsthe1σspreadinthevery WenotethatComponent3isonlydetectedinSiiiλ1260and metal-poorDLApopulation.Wealsofindthatthe[Si/O]abundance ismarginallydetectedinbothCiilines;thiscomponentisnotob- isidenticalbetweenComponents1and2,whereasthe[C/O]abun- servedinOi.Thus,thiscomponentmaybeanunfortunateblendof danceofthesecomponentsdiffersby0.11dex(a1.8σdifference). unrelatedfeatures,notphysicallyassociatedwiththeDLA..How- ThisdifferenceisstillwellwithinthescatterofallDLAmeasure- ever, if this component is associated with the DLA, it probably ments(seeSection3.3). arises in a region of ionized gas, since no absorption is detected We first compare the abundance pattern of this DLA to nu- inOi.Forthesereasons,weconsideronlythecolumndensityfrom cleosynthesismodelsofverymassivemetal-freestars(M (cid:39)140− absorption components 1 and 2 when assessing the relative ele- 260M )thatendedtheirlifeaspair-instabilitysupernovae(Heger (cid:12) mentcompositionofthisDLA.Thisensuresthatweonlyconsider & Woosley 2002). In these models, stars with a mass in this themostlyneutralgas,wheretheobservedionsareallthedomi- range produce element yields of −0.66 (cid:46) [C/O] (cid:46) −0.60 and nantionizationstage,thusobviatingtheneedtoperformionization +0.20 (cid:46) [Si/O] (cid:46) +0.80,whichareinconsistentwiththevalues corrections. wemeasure.Itthusseemsunlikelythatapair-instabilitysupernova enrichedthisDLA. We also compare the abundance pattern of this DLA to nu- cleosynthesis models of massive metal-free stars that ended their 3 RESULTS livesastype-IIcore-collapsesupernovae(Heger&Woosley2010). 3.1 Metal-poorchemistry Thissetofmodelsconsistsof120simulatedstarscoveringamass rangeM =10−100M ,withamassresolutionof∆M (cid:38)0.1M . (cid:12) (cid:12) The DLA reported here is the most metal-poor system currently Sincetheexplosionmechanismofatype-IIsupernovaisstillun- known, and has an oxygen abundance ([O/H] = −3.05) a factor certain,thesemodelsadoptaparameterised‘mixingandfallback’ of2lowerthanthenextmostmetal-poorDLA(reportedrecently scheme.Themixingbetweenstellarlayersduringtheexplosionis byCookeetal.2016;[O/H]= −2.85).Similarly,thissystemex- achieved by ‘smoothing’ the star over the mass (i.e. radial) coor- hibitsthelowestCandSiabundanceofanyknownDLA.Otherex- dinate4timeswithaboxcarfilterofaspecifiedwidth6;agridof tremelymetal-poorquasarabsorptionlinesystemsareknownwith 14 different widths are considered by Heger & Woosley (2010), lowerHicolumndensity(knownasLymanlimitsystems,LLSs), wherethemixingwidthisdefinedasafractionoftheHecoresize. including two systems with no detectable metals ([Z/H] (cid:46) −4.0; Then,theexplosionissimulatedasamovingpistonthatdeposits Fumagalli, O’Meara, & Prochaska 2011) and two systems with momentumataspecifiedmasscoordinate.Inwhatfollows,wecon- comparable (lower) metal abundances to the DLA reported here siderthestandardcaserecommendedbyHeger&Woosley(2010), (Crighton,O’Meara,&Murphy2016;Lehneretal.2016).Given whichplacesthepistonnearthebaseoftheoxygenburningshell, that the abundance measurements of LLSs require an ionization wheretheentropyperbaryonS (cid:39)4k ,wherek istheBoltzmann correction,DLAstendtoaffordsomewhathigherprecisionabun- constant.TheexplosionisparameteriBsedbytheBkineticenergyof dances.This,inturn,allowsforamoreinformativetestofnucle- thematerialthatescapesthebindingenergyofthestar,E (here- exp osynthesismodels. afterreferredtoastheexplosionenergy).Thesemodelsconsidera The relative element abundances5 of this DLA are [C/O] = gridof10valuesoftheexplosionenergyforeachstar,intherange 5 Inwhatfollows,weassumethatthemetalsandhydrogenofthisDLAare andabundancepattern(seeTable2ofCooke,Pettini,&Jorgenson2015; uniformlymixed.Thisassumptionreceivessomesupportfromobservations seealso,Prochaska2003). ofthreeverymetal-poorDLAs(Q0913+072,J1419+0829,J1558−0031) 6 Heger&Woosley(2010)adoptthisparameterisationasitprovidesagood whichexhibittwoabsorptioncomponentsthatshareasimilarmetallicity fittothehardX-rayandopticallightcurvesofSN1987A. MNRAS000,1–10(2015) Themostmetal-poorDLA 5 10 ) 8 g r e 1 6 5 0 1 ( 4 p x e E 2 0.2 g n xi Mi 0.1 0.0 19 20 21 22 2 4 6 8 10 0.0 0.1 0.2 Mass (M ) E (1051 erg) Mixing exp (cid:12) Figure3. Theobserved[C/O]and[Si/O]abundancesofthereportedDLAhavebeencombinedwiththeHeger&Woosley(2010)metal-freenucleosynthesis calculationstoestimatetheprogenitormass,explosionenergy(Eexp)andstellarmixingparameterofthemetal-freestarthatmighthaveenrichedtheDLA.The diagonalpanelsshowthemarginalisedprobabilityofeachmodelparametershownonthex-axis.Themiddleleft,bottomleft,andbottommiddlepanelsshow thetwodimensionalprojections,whichdemonstratethatthemodelparametersarenotdegeneratewithoneanother.Darkandlightshadesenclosethe68and 95percentconfidencecontoursrespectively.Thenon-contiguousregionsinthetwodimensionalprojectionsareduetosmallchangesinthenucleosynthesis ofsimilarmassstars;thesechangesarecausedbyslightdifferencesinthelocationofthevariousburningshells.Thecurrentobservationsplaceastrongbound onthemassoftheprogenitorstar,independentlyoftheexplosionenergyandstellarmixingparameter(seealso,Figure4);notethatthemixingparameteris unconstrainedbythecurrentdata. (0.3−10)×1051 erg.Thus,thismodelsuitecomprisesatotalof Itmayseemsurprisingthattheprogenitormassofthemodel 16800combinationsofthethreeparameters:Stellarmass,explo- Population III star is so well-determined relative to the other pa- sionenergy,andmixingwidth.Therangesofthemodelparame- rametersofthemodel.Thisstrongboundislargelyduetothewell- tersareobservationallymotivated(seethediscussionbyHeger& determined [C/O] value. The most abundant isotopes of C and O Woosley2010),andencompassallrealisticmodelvalues. (i.e. 12C and 16O) are primarily formed by helium burning, with Tofindtherangeofparametersthatareanacceptablesolution some16Oresultingfromneonburning(seee.g.Woosley&Weaver totheabundancepatternoftheDLAreportedhere,wehavelinearly 1995).Thesemodelssuggestthatthe[C/O]abundanceislargelyin- interpolatedthisthree-dimensionalspace,andconductedaMarkov sensitivetothedetailsoftheexplosionwhenE (cid:38)1.5×1051erg, exp chainMonteCarloanalysistosearchforthemostlikelysetofpa- andisalmostuniquelydependentonthemassoftheprogenitorstar rameters,usingtheemceesoftware(Foreman-Mackeyetal.2013). (see top panel of Figure 4). When the explosion energy is lower InFigure3,weplottheone-andtwo-dimensionalprojectionsof thanthisvalue,ahigherfractionof16O(relativeto12C)fallsback thesamplestoidentifythecovariancebetweenmodelparameters. ontothecompactremnant,therebycausingtheC/Oratiotoincrease (darkandlightshadesinthe2Dprojectionsrepresentthe68and (seee.g.Figure4ofHeger&Woosley2010).Thisismatchedbya 95percentconfidencecontours,respectively).Wefindthatthepro- decreaseintheSi/Oratio,sincemoreSifallsbackrelativetoO. genitormassiswell-determined,centeredonavalueM=20.5M , Themonotonicityoftherelationshipbetweenthe[C/O]abun- (cid:12) whilethesupernovaexplosionenergytendstowardstheupperend danceandtheprogenitormass(aswellastheinvarianceof[C/O] oftherangeconsidered,withafavouredvalueE ∼6–8×1051erg. withE )breaksdownabove∼ 35 M ,wherehigherC/Ovalues exp exp (cid:12) Themixingparameterisunconstrained.7 parameters,weareunabletofullyconstrainthebest-fittingmodelparame- ters.Themixingparametercannotbedeterminedwiththecurrentlyavail- 7 Sinceweonlyhavetwomeasuredrelativeabundancestofitthreemodel abledata. MNRAS000,1–10(2015) 6 Cookeetal. +0.2 derivedfromthemeasured[Si/O]ratio(seemiddlepanelofFig- ure4).Wenotethatatighterboundontheexplosionenergycould 0.0 beaffordedbythe[Fe/O]ratio(bottompanelofFigure4).Unfor- tunately,thecurrentlymeasuredupperlimitonthisratio([Fe/O](cid:54) ] 0.2 /CO −0.4 +th0e.2[F3e),/Ois]uanbaubnldeantoceprisovaildsoehainghinlyfosremnastiitviveebtooutnhdecohnoEseenxpm.Sixinincge [ − parameter, a measurement of the [Ni/Fe] abundance is needed to 0.6 breakthedegeneracy(seeCookeetal.2013).Thesemeasurements − mightbecomepossiblewiththenextgenerationof30+mtelescope 0.8 facilities(seeSection3.2). − Finally, we have only two relative element abundances with +0.2 whichtoconstrainthenucleosynthesismodels,andweareunable atthisstagetoruleoutthepossibilitythatthissystemmaybecon- 0.0 taminatedbyotherformsofnucleosynthesis,includingPopulation O] 0.2 II core-collapse supernovae. At a redshift of zabs = 3.07759, the / − Universe has aged by (cid:39) 1.5Gyr after the epoch of the first stars Si 0.4 (z (cid:39) 10−15;e.g.Maioetal.2010),andthereisenoughtimefor [ − asecondgenerationofstarstohavealreadyoperatedinthisDLA. 0.6 ItisneverthelessencouragingthatthechemistryofthisoneDLA − isingoodagreementwiththechemistryofabsorptionlinesystems 0.8 − at z (cid:39) 6 (Becker et al. 2012), which are captured only a few abs hundred Myr after the putative epoch of the first stars. It is also +0.2 worthnotingthatthelowmetallicityofthisDLAisconsistentwith 0.0 the metallicity regime expected for gas that is enriched solely by metal-freePopulationIIIstars(Smith&Sigurdsson2007;Bromm O] 0.2 &Yoshida2011;Wiseetal.2012;Cooke&Madau2014;Webster, − / Bland-Hawthorn,&Sutherland2015;Ritteretal.2016). Fe 0.4 [ − 0.6 − 3.2 Theantecedentsofmetal-poorstarsandtheultra-faint 0.8 dwarfgalaxies − 10 20 30 40 50 60 70 80 90 100 According to the Stellar Abundances for Galactic Archaeology Progenitor Star Mass (M ) (SAGA)database8(Sudaetal.2008),theonlystarcurrentlyknown (cid:12) withalower[C/H]abundancethantheDLAreportedhereisthe Figure 4. The calculated relative abundances of [C/O], [Si/O], Leo star (Caffau et al. 2011). The remaining stars in the SAGA and [Fe/O] for metal-free stars (Heger & Woosley 2010) are shown databaseareeitherredgiantbranchstars9ordisplayahigherabun- as a function of the progenitor star mass. Each curve is colour- danceof[C/H].ThestarwiththeclosestabundancestotheDLA coded by the kinetic energy released by the supernova explosion we report here is SDSS J0259+0057 (Aoki et al. 2013), which (blue→red = 1.8, 2.4, 3, 5, 10 × 1051 erg).Theblackcurveinthe hasacarbonabundance[C/H]= −3.33±0.22,andisnotcarbon- toppanelshowstherelationshipbetweenC/Oandprogenitorstarmassfor enhanced([C/Fe]=−0.02±0.22). starswithametallicityZ =10−4Z(cid:12)(Woosley&Weaver1995).Thegrey IftheDLAweconsiderherehasachemicalabundancepattern shadedregionsinthetopandmiddlepanelsshowtheabundancemeasure- thatissimilartotheMilkyWayhalostarSDSSJ0259+0057,wees- mentsoftheDLAreportedhere(darkandlightshadesrepresentthe68and timatethattheFeabundanceoftheDLAwouldbe[Fe/H](cid:39)−3.4. 95percentconfidencebounds,respectively). The Fe abundance of the DLA can also be estimated by sub- tracting the typical Si/Fe abundance of very metal-poor DLAs ([Si/Fe] = +0.32±0.09; Cooke et al. 2011b) from the observed arerecoveredbythemodels.Wenotethatmodelstarswithamass [Si/H] abundance, which would give [Fe/H] (cid:39) −3.5. We note, above∼35M(cid:12)areruledoutbythemeasured[Si/O]abundanceof however,thatthe[Fe/H]abundanceofthisDLAmaybeconsider- theDLAreportedhere(seemiddlepanelofFigure4). ablylowerthantheseestimates;thereexistsomeMilkyWayhalo Givenenoughdata,wespeculatethatthestrongmonotonicde- starswithslightlyhigher[C/H]thatexhibitastrongoverabundance pendenceoftheC/Oratioonprogenitormassmightallowarobust of [C/Fe], collectively known as CEMP stars (see e.g. Beers & measureofthePopIIIinitialmassfunction.Themainuncertainty governing the calculated [C/O] ratio is the 12C(α,γ)16O reaction ratethatisusedasinputintothestellarmodels;thevalueofthisrate 8 TheSAGAdatabaseismaintainedbyTakumaSuda,YutakaKatsuta,and adopted by Heger & Woosley (2010) is consistent with the latest ShimakoYamada,andisavailableatthefollowingaddress: http://sagadatabase.jp/wiki/doku.php empiricaldeterminationreportedbyAnetal.(2016).However,we cautionthatothereffectssuchasrotation(Hirschi2007;Ekstro¨m 9 Asastarascendstheredgiantbranch(RGB),carbonthathasbeenpro- cessedthroughtheCNcycleismixedtothestellarsurface,reducingthe et al. 2008; Joggerst, Woosley, & Heger 2009), mass loss via ro- surfaceabundanceofC(e.g.Grattonetal.2000).Asaresult,the[C/H] tation/winds(Meynet&Maeder2002),andthethree-dimensional abundancethatanupperRGBstarwasbornwithcanonlyberecoveredby modellingofmetal-freestars(Joggerst,Almgren,&Woosley2010) applyingacorrectionbasedonstellarevolutionmodels(Placcoetal.2014). mayalterthedependenceofC/Oontheprogenitormass. Aftercorrectingforthiseffect,Placcoetal.(2014)find∼10RGBstarsthat The somewhat weaker bound on Eexp shown in Figure 3 is exhibitacorrected[C/H]valuethatislowerthantheDLAwereporthere. MNRAS000,1–10(2015) Themostmetal-poorDLA 7 Christlieb2005).Measuringthedetailedchemicalabundancepat- Table2.CompilationofDLAC/OandO/Hmeasurements ternofthisDLA(includingtheabundancesofN,Mg,Al,andFe) willbecomefeasiblewiththenextgenerationof30+mtelescope QSO zabs [C/O] [O/H] refa facilities. Assuming that the abundances of these elements are in the same proportions as a typical very metal-poor DLA (see Ta- J0035−0918 2.34010 0.08±0.16 −2.44±0.07 1,2 HS0105+1619 2.53651 0.110±0.045 −1.776±0.021 3 ble13ofCookeetal.2011b),thisgoalcouldbeachievedwitha J0140−0839 3.69660 −0.30±0.08 −2.75±0.15 4 spectrumofS/N (cid:39) 150.10 J0311−1722 3.73400 −0.42±0.11 −2.29±0.10 5 Asstatedabove,thereisonlyonestarcurrentlyknownthatis J0903+2628 3.07759 −0.38±0.03 −3.05±0.05 6 morecarbon-poorthantheDLAreportedhere.Wenowspeculate Q0913+072 2.61829 −0.36±0.012 −2.416±0.011 3,7 how many DLAs might need to be observed before a DLA with J0953−0504 4.20287 −0.50±0.03 −2.55±0.10 2 an abundance as low as the Leo star may be found. The SAGA J1001+0343 3.07841 −0.41±0.03 −2.65±0.05 5 database lists 7 stars with a carbon abundance [C/H] (cid:54) −3.0, in- J1016+4040 2.81633 −0.21±0.05 −2.46±0.11 7 cludingtheLeostarwhichhasanabundance[C/H](cid:54)−3.8.There Q1111+1332 2.27094 −0.18±0.11 −1.92±0.08 1 are currently four DLAs with a carbon abundance [C/H] (cid:54) −3.0 Q1202+3235 4.97700 −0.33±0.11 −2.02±0.13 8 (seeSection3.3andTable2);thus,doublingthenumberofDLAs J1337+3153 3.16768 −0.19±0.11 −2.67±0.18 9 inthisextremelymetal-poorregimemaybesufficienttouncovera J1358+6522 3.06730 −0.27±0.06 −2.335±0.02 3,10 J1558+4053 2.55332 −0.06±0.07 −2.45±0.06 7 DLAofsimilarmetallicitytotheLeostar. Q2206−199 2.07624 −0.38±0.04 −2.07±0.05 7 GiventhemetalpaucityoftheDLAreportedhere,itiscon- J2155+1358 4.21244 −0.29±0.08 −1.80±0.11 11 ceivable that just a single (Population III) star is responsible for themetalenrichment.Indeed,hydrodynamicmodelsthattracethe a1:Cooke,Pettini,&Jorgenson(2015);2:Duttaetal.(2014);3:Cookeet internal feedback11 within idealized dwarf galaxies, do entertain al.(2014);4:Ellisonetal.(2010);5:Cookeetal.(2011b);6:Thiswork;7: this idea (Webster, Bland-Hawthorn, & Sutherland 2015; Bland- Pettinietal.(2008);8:Morrisonetal.(2016);9:Srianandetal.(2010);10: Hawthorn,Sutherland,&Webster2015).Inthisscenario,theDLA Cooke,Pettini,&Murphy(2012);11:Dessauges-Zavadskyetal.(2003). reported here would be considered an immediate progenitor of a “firstgalaxy”thatispoisedtoformitsfirstgenerationoflowmass stars (see also, Section 3.3). This scenario receives some support triangle).Forconvenience,inTable2weprovidealloftheDLA from abundance measurements of stars in the lowest luminosity measurementsusedinthiswork,scaledtoouradoptedsolarabun- and most metal-poor galaxies in the Local Group – the so-called dance scale (Asplund et al. 2009, see also the second column of ultra-faint dwarf galaxies (UFDs; Kirby et al. 2008; Norris et al. Table1). 2010;Simonetal.2011;Laietal.2011;Vargasetal.2013;Brown AlloftheobservationsdisplayedinFigure5aregenerallyin etal.2014);thepresumed[Fe/H]abundanceoftheDLAreported good mutual agreement, regardless of the different environments here([Fe/H](cid:39)−3.5)iscomparabletothemostFe-poorstarsinthe probedandthedifferenttechniquesused.Thereare,however,afew UFDs.12 DLAsandHiiregionsthatexhibitsomewhatenhancedC/Ovalues relativetothelocusdefinedbyMilkyWaystars.Similarenhance- ments have also been seen in the [C/α] ratios measured in some 3.3 MetallicityevolutionoftheC/Oratio LLSs (Lehner et al. 2016; see also, Lehner et al. 2013). The ori- Wenowextendthisdiscussiontothemetallicityevolutionofthe ginofsuchenhancementsisunclearatpresent,butcouldberelated C/Oratio.InFigure5,wepresentthelatestcomplementofC/Oand to the production of C by intermediate mass stars (Sharma et al. O/Hmeasurementsavailableintheliterature.Theseincludeabun- 2016). dancemeasurementsofstars(Bensby&Feltzing2006;Fabbianet Asdiscussedpreviouslyintheliterature(Henry,Edmunds,& al.2009;Nissenetal.2014)intheMilkyWaythinandthickdisc Ko¨ppen 2000; Carigi 2000; Akerman et al. 2004; Cescutti et al. (solid and open black circles respectively) and in the Milky Way 2009; Romano et al. 2010), the increase of [C/O] when [O/H] (cid:38) halo (black squares). To this we add C/O measurements of local −1.0 is thought to be due to the metallicity dependent winds of Hiiregions(Dufour,Shields,&Talbot1982;Garnettetal.1995; massivestarscombinedwiththedelayedreleaseofCfromlowand Kurtetal.1999;Estebanetal.2002,2009,2014;Garc´ıa-Rojas& intermediatemassstars.Theupturnof[C/O]when[O/H](cid:46) −1.5, Esteban 2007; Lo´pez-Sa´nchez et al. 2007; Berg et al. 2016), us- on the other hand, is still open to debate. Akerman et al. (2004, ing either optical recombination lines or ultraviolet collisionally seealso,Carigi&Peimbert2011)proposethattheincreasedC/O excited lines (blue diamonds and squares respectively).13 Finally, values seen at low O/H is the result of an increased carbon yield weoverplotmeasurementsofC/OandO/Hinthelowestmetallic- frommetal-freePopulationIIIstars.Thisscenariohaspreviously ityDLAs(redtriangles)andinthenewDLAreportedhere(green been used to explain the lowest metallicity population of CEMP stars(Umeda&Nomoto2003;Ryanetal.2005;Cooke&Madau 2014). 10 ThenextdetectableelementsincludeSandNi,whichwouldrequirea In addition to the C/O upturn at low O/H, we also draw at- S/N (cid:39) 600and2000,respectively,todetectanindividualabsorptionline tention to the scatter in the DLA [C/O] measurements. With the at5σconfidence.StackingtheavailableNiiilines(seeCookeetal.2013) improvedstatisticsofthedatacollectedinTable2,wenowseethat couldreducethisrequirementtoaS/N (cid:39) 1000. theDLAvaluesof[C/O]atagiven[O/H]differbymorethantheir 11 Includingbothradiativefeedbackfrommassivestarsandthesubsequent errors,suggestinganintrinsicdispersionintheseelementsrelative supernovafeedback. abundancesthatwasnotimmediatelyapparentinourearlierstud- 12 Wenotethatadirectcomparisonbetweenthe[Si/H]abundanceofthis DLAandthestarsinUFDsisnotpossible,since[Si/H]hasnotbeenmea- iesofcarbonandoxygeninverymetal-poorDLAs(Pettinietal. suredforthemostFe-poorstarsinUFDs(Vargasetal.2013). 2008;Cookeetal.2011b).Thetotalrangein[C/O]exhibitedby 13 Weapplyacorrectionof+0.24dextothe[O/H]abundancemeasured DLAswith[O/H](cid:46)−2.0spansafactorof∼4,from[C/O](cid:39)−0.5 usingcollisionallyexcitedlines(Estebanetal.2014;Steideletal.2016). to(cid:39)+0.1.Thereareseveralpossibilitiesthatcouldexplainsucha MNRAS000,1–10(2015) 8 Cookeetal. +0.5 0.0 ] O / C [ 0.5 − 1.0 − 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 +0.5 − − − − − − − [O/H] Figure5. Thechemicalevolutionof[C/O]isshownforvariousastrophysicalenvironments.Theblacksymbolsrepresentstarsthatarekinematicallyassociated withthethindisc(solidcircles),thickdisc(opencircles),andhalostars(squares).ThebluesymbolsareforHiiregions,measuredfromopticalrecombination lines(diamonds)orultravioletcollisionallyexcitedlines(squares).Theredtrianglesshowliteraturemeasurementsof[C/O]and[O/H]ofmetal-poorDLAs (seeTable2).Thegreentriangleisthenewmeasurementreportedherein.Thegreybandrepresentsthe‘transitiondiscriminant’proposedbyFrebel,Johnson, &Bromm(2007),wherethewidthofthebandindicatestheuncertaintyofthiszone.Gascloudstotheleftofthiszonearenotexpectedtoformalowmass (i.e.long-lived)generationofstars.Seethetextforthereferencestoallplotteddatasets. scatter.AsshowninthetoppanelofFigure4(seealso,Section3.1), fromapredominantlyhighmassgenerationof(PopulationIII)stars the[C/O]ratioissensitivetothemassofthestarsthatenrichedthe toalowmassgenerationof(PopulationII)stars.Thiscriterionis DLAs. If the first stars formed in isolation or in small multiples, basedonthefinestructurecoolinglinesof[Cii]and[Oi].Oncea asincurrentlyfavouredscenarios(Clark,Glover,&Klessen2008; criticalabundanceofCandOisreachedinacloudofcoldneutral Turk,Abel,&O’Shea2009;Stacy,Greif,&Bromm2010;Stacy gas(i.e.totherightofthegreybandshowninFigure5),thegasis &Bromm2013;Hiranoetal.2014;Stacy,Bromm,&Lee2016), abletofragmenttolowmassscalesandproducealong-livedgen- the intrinsic C/O scatter could be explained if these DLAs were erationofstars.Inprinciple,DLAsarenotrestrictedtotherightof enrichedbyasingleorasmallmultipleofPopIIIstars.Thepre- thisregion;aDLAthatoverlapsthisregionwouldbeanidealenvi- dicted mass range of the stars responsible for the enrichment of ronmenttoempiricallystudythetransitionfromPopulationIIIto thesemetal-poorDLAsis10 (cid:46) M/M (cid:46) 25,withabiastowards PopulationIIstarformation.TheDLAreportedhereliesveryclose (cid:12) highermassstars.Analternativepossibility,suggestedrecentlyby tothisregion,14 andappearstohavejustcrossedthethresholdfor Sharmaetal.(2016),isthattheC/Oscatterisduetoenrichment lowmassstarformation. byacombinationofmassiveandintermediatemass(PopulationII) Finally, given that: (1) The gas in DLAs is mostly neutral stars;thehighC/Ovaluesaretheresultofpreferentialenrichment (i.e.conducivetostarformation);(2)thisparticularDLAhasjust byintermediatemassstarswhilethelowC/Ovaluesareprimarily crossed into a chemical regime where a generation of long-lived duetomassivestars.TheobservedC/Oscatteristheresultofpoor starsisexpectedtoform;and(3)thepresumed[Fe/H]abundance mixingbetweenthesetwochannelsofcarbonproduction. oftheDLAreportedhereisconsistentwiththemostFe-poorstars seeninUFDs,wesuggestthatthisDLAcouldberepresentativeof We point out that massive Population II stars alone are less oneoftheantecedentsoftheUFDgalaxypopulation,asdiscussed likelytoproducetheobservedC/OscatterthanPopulationIIIstars, previouslyinSection3.2. eventhoughtherelationshipbetweenC/Oandprogenitorstarmass isnearlyidenticalforPopulationIIIstarsandstarswithprogenitor metallicity Z = 10−4 Z (see the black curve in the top panel of (cid:12) Figure4).ThisisbecausethestellarinitialmassfunctionofPop- 4 SUMMARYANDCONCLUSIONS ulationIIstarsismorefullysampled,unlikethesparselysampled Wereportthediscoveryofthemostmetal-poordampedLyman-α initial mass function of Population III stars, as discussed above. Likewise,wesuggestthatthescatterofC/Omeasurementsinthe system currently known, located at a redshift zabs = 3.07759, based on observations of the quasar J0903+2628 taken with the mostmetal-poorDLAsmightprovideameasureofthestellarini- KeckHIRESspectrograph.ThisDLAwasidentifiedaspartofour tialmassfunctionandmultiplicityofthefirststars.Measuringthe intrinsicscatterofC/Oinalmostpristineenvironmentsshouldbe ongoingprogrammetosearchforthenucleosyntheticimprintsof consideredakeygoaloffutureobservations. In Figure 5, we also overlay the ‘transition discriminant’ 14 TheDLAreportedhereexhibitsavalueofthetransitiondiscriminant, (Frebel,Johnson,&Bromm2007,seealsoBromm&Loeb2003), Dtrans=−3.19±0.04,whichliesjustabovethecriticalvaluecalculatedby shownbythesolidgreyshadedregion,whichmarksthetransition Frebel,Johnson,&Bromm(2007),Dtrans,crit=−3.5±0.2. MNRAS000,1–10(2015) Themostmetal-poorDLA 9 thefirststarsinneutralgasathighredshift.Wedrawthefollowing bepossibletomeasurethe[Ni/Fe]ratiotocomparableprecision, mainconclusions: thereby allowing a strong bound on the supernova explosion en- ergy(Cookeetal.2013).Finally,weestimatethatbymerelydou- (i)TheextremelylowmetallicityofthisDLAhasallowedonly blingthenumberofDLAswithacarbonabundance[C/H](cid:54)−3.0 the most abundant chemical elements to be detected. Based on (from4to8DLAs),itisstatisticallypossibletofindaDLAthatis Voigtprofilefitting,wededucethefollowingabundances:[C/H]= asmetal-poorastheLeostar.Furthermore,thefutureprospectof −3.43±0.06,[O/H]=−3.05±0.05,and[Si/H]=−3.21±0.05,and findingextremelymetal-poor(andpossibly,pristine)DLAsathigh anupperlimitontheironabundanceof[Fe/H](cid:54)−2.81.ThisDLA redshiftisverypromising;thesteepslopeofthequasarluminosity isafactorof∼2moreoxygen-poorthanthenextmostmetal-poor functioncombinedwiththecollectingareaofthe30+mtelescopes, DLAknown(Cookeetal.2016;[O/H]=−2.85). willensureaccessto∼100timesmorequasarsthanthosethatare (ii)ThisDLAexhibitsanabundanceofcarbonthatislowerthan withinreachofcurrentfacilities(seethediscussionbyCookeetal. anymetal-poorstarcurrentlyknown,otherthantheLeostar(Caf- 2016). fauetal.2011).Thelowmetallicityisconsistentwiththeviewthat thisDLAmayhavebeenenrichedsolelybytheproductsofafirst generationofstars. ACKNOWLEDGEMENTS (iii)Wehavecomparedtherelativechemicalabundancesofthis We are grateful to the staff astronomers at Keck Observatory for DLA to nucleosynthesis calculations of massive metal-free stars their assistance with the observations. We thank the anonymous that ended their lives as core-collapse supernovae. We show that theC/Oyieldofmassivemetal-freestarsisastrong(almostmono- referee for their prompt review, and for offering several helpful suggestions that improved the presentation of this paper. During tonic)functionoftheprogenitormass,andisvirtuallyindependent this work, R. J. C. was supported by a Royal Society University oftheotherparametersinthenucleosynthesismodellingwhenthe explosionenergyisE (cid:38)1.5×1051erg.Usingthesemodels,we Research Fellowship, and by NASA through Hubble Fellowship exp grantHST-HF-51338.001-A,awardedbytheSpaceTelescopeSci- estimatethattheDLAwasenrichedbyastarofmassM(cid:39)20.5M . (cid:12) enceInstitute,whichisoperatedbytheAssociationofUniversities Wealsoruleoutthepossibilitythattheenrichmentwasduetoastar forResearchinAstronomy,Inc.,forNASA,undercontractNAS5- thatendeditslifeasapair-instabilitysupernova. 26555. CCShasbeen supported bygrant AST-1313472from the (iv) The carbon and oxygen abundances of this DLA yield a U.S.NSF.ThisresearchwasalsosupportedbyaNASAKeckPI measure of the transition discriminant, D = −3.19 ± 0.04, trans DataAward,administeredbytheNASAExoplanetScienceInsti- whichisjustoverthecriticalthresholdforlowmassstarformation tute.DatapresentedhereinwereobtainedattheW.M.KeckOb- D =−3.5±0.2(Frebel,Johnson,&Bromm2007).Giventhe trans,crit servatory from telescope time partially allocated to the National largecolumndensityofneutralgashostedbytheDLA,whichmay Aeronautics and Space Administration through the agency’s sci- beconducivetostarformation,weproposethatthisenvironment entificpartnershipwiththeCaliforniaInstituteofTechnologyand mightrepresentanimmediateprecursortotheformationofa‘first theUniversityofCalifornia.Ourworkmadeuseofthematplotlib galaxy’(i.e.adarkmatterhalothathasnotyetformedageneration (Hunter2007), emcee(Foreman-Mackeyet al.2013), andcorner oflong-lived,lowmassstars). (Foreman-Mackeyetal.2016)pythonpackages,whichwegrate- (v)WecompileallavailableliteraturedeterminationsoftheC/O fullyacknowledge.TheObservatorywasmadepossiblebythegen- abundanceinverymetal-poorDLAs.Thescatterofthesemeasure- erousfinancialsupportoftheW.M.KeckFoundation.Wethankthe mentsislargerthanthemeasurementuncertainties,indicatingthat HawaiianpeoplefortheopportunitytoobservefromMaunaKea; thereisanintrinsicdispersioninthepopulation.Giventhesensitiv- withouttheirhospitality,thisworkwouldnothavebeenpossible. ityoftheC/Oratiotoprogenitormass,weproposethatthescatter can be explained if a single or small multiple of metal-free stars isresponsiblefortheenrichmentofextremelymetal-poorDLAs. WealsoproposethatthedistributionofC/Omeasurementsofex- REFERENCES tremelymetal-poorDLAscanbeusedtodetermineboththestellar AbelT.,BryanG.L.,NormanM.L.,2002,Sci,295,93 initialmassfunctionandthemultiplicityofthefirststars. AkermanC.J.,CarigiL.,NissenP.E.,PettiniM.,AsplundM.,2004,A&A, (vi)Recentnumericalmodelshavealsosuggestedthatthemost 414,931 metal-poor DLAs may have been enriched by just a single mas- AnZ.-D.,MaY.-G.,FanG.-T.,LiY.-J.,ChenZ.-P.,SunY.-Y.,2016,ApJ, 817,L5 sivestar,andcouldhavesimilarpropertiestotheprogenitorsofthe AokiW.,etal.,2013,AJ,145,13 Milky Way ultra-faint dwarf galaxies (Salvadori & Ferrara 2012; AsplundM.,GrevesseN.,SauvalA.J.,ScottP.,2009,ARA&A,47,481 Webster, Bland-Hawthorn, & Sutherland 2015). We note that the BeckerG.D.,SargentW.L.W.,RauchM.,CarswellR.F.,2012,ApJ,744, presumed[Fe/H]oftheDLAreportedhereisconsistentwiththe 91 mostFe-poorstarsfoundintheUFDgalaxies.Akeygoaloffuture BeersT.C.,ChristliebN.,2005,ARA&A,43,531 workwillbetopindowntheironabundanceoftheDLAreported BensbyT.,FeltzingS.,2006,MNRAS,367,1181 hereandofotherextremelymetal-poorDLAs. BergD.A.,SkillmanE.D.,HenryR.B.C.,ErbD.K.,CarigiL.,2016, Our work highlights the importance of identifying and mea- ApJ,827,126 Bland-HawthornJ.,SutherlandR.,WebsterD.,2015,ApJ,807,154 suringthedetailedchemicalabundancepatternsofthemostmetal- BosmanS.E.I.,BeckerG.D.,2015,MNRAS,452,1105 poorDLAs.Thehighprecisionandaccuracythatcanbeachieved BrommV.,CoppiP.S.,LarsonR.B.,1999,ApJ,527,L5 for the relative chemical abundances ((cid:46) 0.05 dex), allow for a BrommV.,LoebA.,2003,Natur,425,812 strongandinformativetestofnucleosynthesismodelsofmetal-free BrommV.,YoshidaN.,2011,ARA&A,49,373 stars.Inparticular,theC/Oratiooffersasensitiveconversiontothe BrownT.M.,etal.,2014,ApJ,796,91 progenitormassofthemetal-freestarsthatmayhaveenrichedthe CaffauE.,etal.,2011,Nature,477,67 mostmetal-poorDLAs.Withfuturetelescopefacilities,itmayalso CarigiL.,2000,RMxAA,36,171 MNRAS000,1–10(2015) 10 Cookeetal. CarigiL.,PeimbertM.,2011,RMxAA,47,139 Lo´pez-Sa´nchezA´.R.,EstebanC.,Garc´ıa-RojasJ.,PeimbertM.,Rodr´ıguez CescuttiG.,MatteucciF.,McWilliamA.,ChiappiniC.,2009,A&A,505, M.,2007,ApJ,656,168 605 MaioU.,CiardiB.,DolagK.,TornatoreL.,KhochfarS.,2010,MNRAS, ClarkP.C.,GloverS.C.O.,KlessenR.S.,2008,ApJ,672,757 407,1003 CookeR.J.,MadauP.,2014,ApJ,791,116 MarassiS.,ChiakiG.,SchneiderR.,LimongiM.,OmukaiK.,NozawaT., CookeR.,PettiniM.,SteidelC.C.,RudieG.C.,JorgensonR.A.,2011a, ChieffiA.,YoshidaN.,2014,ApJ,794,100 MNRAS,412,1047 MeynetG.,MaederA.,2002,A&A,390,561 CookeR.,PettiniM.,SteidelC.C.,RudieG.C.,NissenP.E.,2011b,MN- MorrisonS.,KulkarniV.P.,SomD.,DeMarcyB.,QuiretS.,Pe´rouxC., RAS,417,1534 2016,ApJ,830,158 CookeR.,PettiniM.,MurphyM.T.,2012,MNRAS,425,347 Morton,D.C.2003,ApJS,149,205 CookeR.,PettiniM.,JorgensonR.A.,MurphyM.T.,RudieG.C.,Steidel NakamuraF.,UmemuraM.,2001,ApJ,548,19 C.C.,2013,MNRAS,431,1625 NissenP.E.,ChenY.Q.,CarigiL.,SchusterW.J.,ZhaoG.,2014,A&A, CookeR.J.,PettiniM.,JorgensonR.A.,MurphyM.T.,SteidelC.C.,2014, 568,A25 ApJ,781,31 NorrisJ.E.,WyseR.F.G.,GilmoreG.,YongD.,FrebelA.,Wilkinson CookeR.J.,PettiniM.,JorgensonR.A.,2015,ApJ,800,12 M.I.,BelokurovV.,ZuckerD.B.,2010,ApJ,723,1632 CookeR.J.,PettiniM.,NollettK.M.,JorgensonR.,2016,ApJ,830,148 NoterdaemeP.,PetitjeanP.,LedouxC.,SrianandR.,2009,A&A,505,1087 CrightonN.H.M.,O’MearaJ.M.,MurphyM.T.,2016,MNRAS,457,L44 PenpraseB.E.,ProchaskaJ.X.,SargentW.L.W.,Toro-MartinezI.,Beeler Dessauges-ZavadskyM.,Pe´rouxC.,KimT.-S.,D’OdoricoS.,McMahon D.J.,2010,ApJ,721,1 R.G.,2003,MNRAS,345,447 PettiniM.,ZychB.J.,SteidelC.C.,ChaffeeF.H.,2008,MNRAS,385, DufourR.J.,ShieldsG.A.,TalbotR.J.,Jr.,1982,ApJ,252,461 2011 DuttaR.,SrianandR.,RahmaniH.,PetitjeanP.,NoterdaemeP.,LedouxC., PlaccoV.M.,FrebelA.,BeersT.C.,StancliffeR.J.,2014,ApJ,797,21 2014,MNRAS,440,307 Prochaska,J.X.2003,ApJ,582,49 Ekstro¨mS.,MeynetG.,ChiappiniC.,HirschiR.,MaederA.,2008,A&A, RitterJ.S.,Safranek-ShraderC.,Milosavljevic´M.,BrommV.,2016,MN- 489,685 RASaccepted EllisonS.L.,ProchaskaJ.X.,HennawiJ.,LopezS.,UsherC.,WolfeA.M., RomanoD.,KarakasA.I.,TosiM.,MatteucciF.,2010,A&A,522,A32 RussellD.M.,BennC.R.,2010,MNRAS,406,1435 RyanS.G.,AokiW.,NorrisJ.E.,BeersT.C.,2005,ApJ,635,349 Erni,P.,Richter,P.,Ledoux,C.,&Petitjean,P.2006,A&A,451,19 SalvadoriS.,FerraraA.,2012,MNRAS,421,L29 Esteban C., Peimbert M., Torres-Peimbert S., Rodr´ıguez M., 2002, ApJ, SharmaM.,TheunsT.,FrenkC.,CookeR.,2016,arXiv,arXiv:1611.03868 581,241 SimcoeR.A.,SullivanP.W.,CookseyK.L.,KaoM.M.,MatejekM.S., EstebanC.,BresolinF.,PeimbertM.,Garc´ıa-RojasJ.,PeimbertA.,Mesa- BurgasserA.J.,2012,Nature,492,79 DelgadoA.,2009,ApJ,700,654 SimonJ.D.,etal.,2011,ApJ,733,46 EstebanC.,Garc´ıa-RojasJ.,CarigiL.,PeimbertM.,BresolinF.,Lo´pez- SmithB.D.,SigurdssonS.,2007,ApJ,661,L5 Sa´nchezA.R.,Mesa-DelgadoA.,2014,MNRAS,443,624 SrianandR.,GuptaN.,PetitjeanP.,NoterdaemeP.,LedouxC.,2010,MN- FabbianD.,NissenP.E.,AsplundM.,PettiniM.,AkermanC.,2009,A&A, RAS,405,1888 500,1143 StacyA.,BrommV.,2013,MNRAS,433,1094 Foreman-MackeyD.,HoggD.W.,LangD.,GoodmanJ.,2013,PASP,125, StacyA.,BrommV.,LeeA.T.,2016,MNRAS,462,1307 306 StacyA.,GreifT.H.,BrommV.,2010,MNRAS,403,45 Foreman-MackeyD.,2016,JOSS,1, SteidelC.C.,StromA.L.,PettiniM.,RudieG.C.,ReddyN.A.,Trainor FrebelA.,JohnsonJ.L.,BrommV.,2007,MNRAS,380,L40 R.F.,2016,ApJ,826,159 FrebelA.,NorrisJ.E.,2015,ARA&A,53,631 SudaT.,etal.,2008,PASJ,60,1159 FumagalliM.,O’MearaJ.M.,ProchaskaJ.X.,2011,Sci,334,1245 TominagaN.,IwamotoN.,NomotoK.,2014,ApJ,785,98 Garc´ıa-RojasJ.,EstebanC.,2007,ApJ,670,457 TurkM.J.,AbelT.,O’SheaB.,2009,Sci,325,601 GarnettD.R.,SkillmanE.D.,DufourR.J.,PeimbertM.,Torres-Peimbert UmedaH.,NomotoK.,2003,Natur,422,871 S.,TerlevichR.,TerlevichE.,ShieldsG.A.,1995,ApJ,443,64 VargasL.C.,GehaM.,KirbyE.N.,SimonJ.D.,2013,ApJ,767,134 GrattonR.G.,SnedenC.,CarrettaE.,BragagliaA.,2000,A&A,354,169 VogtS.S.,etal.,1994,SPIE,2198,362 HegerA.,WoosleyS.E.,2002,ApJ,567,532 WebsterD.,Bland-HawthornJ.,SutherlandR.S.,2015,ApJ,804,110 HegerA.,WoosleyS.E.,2010,ApJ,724,341 WiseJ.H.,TurkM.J.,NormanM.L.,AbelT.,2012,ApJ,745,50 HenryR.B.C.,EdmundsM.G.,Ko¨ppenJ.,2000,ApJ,541,660 WolfeA.M.,GawiserE.,ProchaskaJ.X.,2005,ARA&A,43,861 HiranoS.,HosokawaT.,YoshidaN.,UmedaH.,OmukaiK.,ChiakiG., WoosleyS.E.,WeaverT.A.,1995,ApJS,101,181 YorkeH.W.,2014,ApJ,781,60 YuanS.,CenR.,2016,MNRAS,457,487 HirschiR.,2007,A&A,461,571 HunterJ.D.,2007,CSE,9,90 IshigakiM.N.,TominagaN.,KobayashiC.,NomotoK.,2014,ApJ,792, L32 JankaH.-T.,2012,ARNPS,62,407 JoggerstC.C.,AlmgrenA.,WoosleyS.E.,2010,ApJ,723,353 JoggerstC.C.,WoosleyS.E.,HegerA.,2009,ApJ,693,1780 KellerS.C.,etal.,2014,Natur,506,463 KirbyE.N.,SimonJ.D.,GehaM.,GuhathakurtaP.,FrebelA.,2008,ApJ, 685,L43 KurtC.M.,DufourR.J.,GarnettD.R.,SkillmanE.D.,MathisJ.S.,Peim- bertM.,Torres-PeimbertS.,RuizM.-T.,1999,ApJ,518,246 LaiD.K.,LeeY.S.,BolteM.,LucatelloS.,BeersT.C.,JohnsonJ.A., SivaraniT.,RockosiC.M.,2011,ApJ,738,51 LehnerN.,etal.,2013,ApJ,770,138 Lehner N., O’Meara J. M., Howk J. C., Prochaska J. X., Fumagalli M., 2016,preprint,arXiv:1608.02588 LimongiM.,ChieffiA.,2012,ApJS,199,38 MNRAS000,1–10(2015)