198 OApJ. . .238L.149H The AstrophysicalJournal,238:L149-L153,1980June15 © 1980.TheAmericanAstronomicalSociety.Allrightsreserved.PrintedinU.S.A. globular clusterstarswerechemicallyhomogeneous, dence thathighlyevolvedglobularclusterstarsexhibit but variousrecentinvestigationshaveprovidedevi- andcolorranges.Withtheexceptionofco ences amongfeaturesduetoCH,NH,CN,andCO. a widerangeoflinestrengthdifferenceswithinsmall magnitude diagrams(CMDs)havebeenqualitatively clusters withnarrowgiantbranchesintheircolor- ments totheobservablesurfacelayersorprimordial Centauri andM22,theevidencehasinvolveddiffer- inhomogeneities atthetimeofstarformation(the attributed eithertomixingofinternallyproducedele- Such spectraldifferencesamonghighlyevolvedstarsin in asinglecluster.Sincethemixingmechanismscur- reason forexcludingtheoperationofbothmechanisms extensive literatureonthissubjectisreviewedby rently knownappeartooperateforMy<+lmag McClure 1979andKraft1979).Thereisnoapriori photometry toMu^+2.5(Hesser,Hartwick,and primordial explanationswithinagivencluster. insight intotherelativeimportanceofmixingand the observationstofainterMy’smightprovidesome without comparable^-excess (i.e.,heavyelement) appear atM>+0.5mag.For47TuesubgiantsDDO variations. Carbonetal.(1977) andKraft(1979)have McClure 1976,1977;Hesser1978)andlow-dispersion (Sweigart andMengel1979;Kraft1979),extensionof SIT vidiconspectroscopytoMy~+3.5mag(Hesser tory, whichisoperatedbyAURA, Inc.,undercontractwiththe 1978) revealedthatCNstrengthvariationsarepresent U.S. NationalScienceFoundation. v 1 Until afewyearsagoitwaswidelybelievedthat VisitingAstronomer,CerroTololo Inter-AmericanObserva- Evidence hasbeenaccumulatingthatdifferencesdo © American Astronomical Society •Provided bythe NASAAstrophysics Data System Subject headings:clusters:globular—spectrophotometrystarsabundances parable tothoserequiredexplainobservationsofhighlyevolvedstarsinthesamecluster.Ifa primordial originofthevariationsistoberejected,ourobservationsrequireidentificationa mixing mechanismorotherprocesscapableofaffectingthemain-sequenceatmosphericabundances suggest thatthestar-to-starnitrogenabundancesdifferbyfactorsof^5,i.e.,amountscom- of CNOgroupelementsindependentlyheavierones. (NGC 104)revealdifferencesinCNX3883(0,0)bandstrengths.Spectrumsynthesiscalculations Low-resolution CTIOSITvidiconspectraofsevenupper-main-sequencestarsin47Tucanae CN VARIATIONSAMONGMAIN-SEQUENCE47TUCANAESTARS I. INTRODUCTION :PopulationII—subdwarfs Dominion AstrophysicalObservatory,HerzbergInstituteofAstrophysics Received 1979December28;accepted1980February26 Astronomy Program,UniversityofMaryland 1 James E.Hesser ABSTRACT R. A.Bell AND L149 -1 found evidenceforstar-to-starvariationsinNHand cluster M92,althoughBell,Dickens,andGustafsson NGC 6397starsbrighterthanMy^—0.7mag.Varia- argue thatCHweakeningoccursonlyinM92and fainter 47TuestarswithMv^4.3magwerealso CH strengthsamongsubgiantsinthemoremetal-poor noted intheoriginalTololovidiconsurvey(Hesser among metal-poorM22subgiants(HesserandHarris tions ofCNandCHstrengthshavealsobeenfound importance oftheconclusionthatspectralvariations warranted confirmatoryobservationsbeforeinterpre- sufficient toplacefaithintheresult,itwasfeltthat (1979) haveusedDDOphotometryandspectrato could occuronthemainsequenceofaglobularcluster to ^200-300countsperpixel)ofthespectrawas of HesserandHartwick(1977)weresecuredduring tation. Procedures werevirtuallyidenticaltothoseusedearlier 1978). Eventhoughthestatisticalaccuracy(equivalent 1979). vidicon pixelcorrespondsto^6Âandtheresolutionis tions of^2''5seeingandintermittentlightcloud. the darkhoursof1979August2-4(UT),undercondi- and aslitwidthof590jum(2''5)wereemployed.Each and theRCA4804SITvidicon(Atwoodetal.1979). ~3 pixels.Ouraimwastoutilize thelimiteddarktime extend thesurveytoasmany starsaspossiblefor differences overshort-wavelength intervalsinastronger as efficientlypossibleto seek evidenceofspectral and weakerCNstarfrom the earliersurveyandto Grating 09fromthe1.5mspectrograph(^200Âmm) (Hesser 1978)withtheCTIO4mRCspectrograph Differences amongCN(0,0)bandstrengthsinfour Additional observationsofstarsfromtheCMDstudy II. OBSERVATIONS 198 OApJ. . .238L.149H place thedataonanapproximaterelativefluxscale. 0.66, Mv~+5.4mag)alsofelldirectlyontheslit. marized inTable1.Reductionsusedthecurrentversion The narrowslitcoupledwithseeingeffects,guiding only onefluxstandardstarpernightwereusedto future selectionofthebestcandidatesforhigher- stars 1-9018or1-9004.Inaddition,duringoneofour unmistakeably strongerCN(0,0)featuresthando of thesoftwareoriginallydescribedbySchalleretal. errors, butcomparisonsofadjacentspectralfeatures errors, anddifferentialrefractioncanresultinslope resolution studies.Thus,asbefore,observationsof L150 Although theresultantspectrumwasnearedgeof where itcanbeseenthatstars3-2195and3-2153show frame lines^lO"distantonoppositesidesofeachstar. are unaffected.Datafortheobservedstarssum- rithms arenotentirelysatisfactory,itsuggestsmedium- the vidiconframe,wherestandardextractionalgo- (1978), andincludedskysubtractionfromstar-free 1979 exposures,star3-2274[F=18.56,(B—F)g 3-2153 isalsogiventofacilitatecomparison ofthedifferencesatCN(0,0)X3883bandhead,although attentioniscalledtostar1-9018 and precisionofthesespectraare suchthatonlystrong,broadfeaturesspanningseveralpixelsarestatistically reliable. Fig. 2isduetoverystrong,low-excitation linesofFeiandanMgline,iscommoninstars ofthisspectraltype.Theresolution as aCNweakstarwitha.B—V more likethatof3-2153.Thestrongfeatureat^3825Àinallstars andinthesynthesizedspectraof pgP Fig. 1.—CTIOSITvidiconspectra forthesevenfaint47TuestarsofTable1.Asuperposition of thetracingsstars1-9004and Our finalspectraforallstarsareshowninFigure1, © American Astronomical Society •Provided bythe NASAAstrophysics Data System MESSER ANDBELL 47 Tue. havior ofCNonthemainsequence47Tue.Inview for theanalysisofgiantsin47Tuecanbeusedtoreach by Dickens,Bell,andGustafsson(1979,hereafterDBG) among starslocatedbelowthekneeofturnoffin some semiquantitativeconclusionsregardingthebe- there isnolongeranydoubtthatCNdifferencesoccur to-strong CN.Withtheincreaseddatasamplewefeel group ofobservedspectra. B —VvaluesofthestarsinTable1,wecomputed of thenoiseindataandsimilarityV representative spectrawhichcouldbecomparedtothe using themethodsandprogramsofGustafssonetal. flux constantmodel5750/4.0/—0.8(T^ff/logg/[A/H]) of [m/H]=—0.8foundbyDBG,wecomputedthe DBG wefindthatthestarsinsamplehaveT{« 5750 Kandlogg~4.0.Adoptingthemetalabundance e{ Synthetic spectrumcalculationsfollowingthoseused Extrapolating thesubgiantbranchofFigure5 III. CALCULATIONS Vol. 238 198 OApJ. . .238L.149H -1 profile of15Âhalf-width,whichwasfoundempirically noise inFigure1.Theoverall agreementofthese for the3883ÂCNfeature. The effectonthe4215Â /Vvalues werescaledbyafactorthatvariedfrom0.70 puted fortheregion3750-4250Â,withNabun- calculated HandKlineprofiles.Inordertomatchthe spectra wereconvolvedwithatriangularinstrumental of 0.1Â. by 0.8dexrelativetothesolarvalues,whichforC,N, 4.5/0.0 and5500/4.5/0.0)weretakenfromtheunpub- that enhancementoftheNabundancebyafactor at 3750Âto1.04850beforeplottinginFigure2. and thespectrawerecomputedatwavelengthintervals solar abundancemodels.ADopplerbroadeningvelocity dance beingincreasedbyfactorsof0.5and1.0dex. quently showntohaverfandloggvaluessimilar logarithmic scaleofH=12.00.Thismodelissubse- slopes oftheobservedspectra(cf.§II),computed to givereasonableagreementbetweenobservedand those oftheobservedstars.Otherspectrawerecom- and Oare8.62,8.00,8.86,respectively,onthe length interval3000-7200Âforthemodel5750/4.0/ and Bell. lished gridofdwarfmodelsEriksson,Gustafsson, in themodelcalculations.Twoothermodels(6000/ for [^4/H]=—0.5andused—0.8elsewhere was availablefor[^4/H]=—0.8,weusedtheODF 3-2234. CN bandisprobablytooweak tobedistinguishedfrom Spectra werealsocomputedforthe6000and5500K 3-2153 17.40 (thermal andmicroturbulent)of2kmswasused, hour angleattheendofexposure(Exp.)isalwayswest. except 1-9004and1-9018themagnitudesarephotographic.The 3-2195. form “figurenumber—identificationnumber.”Forallstars 3-2110. 3-2188. 1.0 dexdoesproducereadily observableconsequences No. 3,1980 (1975). Sincenoopacitydistributionfunction(ODF) 1-9004. 1-9018. —0.8, withtheabundancesofallmetalsbeingreduced e a © American Astronomical Society •Provided bythe NASAAstrophysics Data System IdentificationnumbersfromHesserandHartwick1977inthe The spectrumofthemodel5750/4.0/—0.8shows A syntheticspectrumwascomputedforthewave- For comparisonwiththeobservations,synthetic Summary of47TucanaeTurnoffRegion (mag) (UT) 17.51 17.48 17.53 17.50 17.58 0.57 17.54 0.62 V B-VDate Vidicon Observations* 0.62 0.60 0.61 0.60 0.54 TABLE 1 1979 Aug4 1977 Oct13 1977 Oct13 1977 Oct13 1979 Aug2 1979 Aug2 1977 Oct13 1977 Oct13 1979 Aug4 1977 Oct13 1979 Aug3 1979 Aug3 1979 Aug3 1977 Oct13 1977 Oct13 CN VARIATIONSIN47TUCSTARS (min) Exp. 30 30 30 30 30 30 30 30 21 25 25 30 30 30 15 hm Angle Hour 0 25 0 17 3 26 0 47 2 52 2 19 3 59 0 55 439 5 12 1 18 1 03 1 49 1 40 1 13 perature andspectraldata.) Peterson andKing1975; 1976)intheSW precise enoughtoallowusfindanaccurateCabun- be minimizedonceweareabletoobtainbettertem- dance. UncertaintiesintheexactNandCrangeswill satisfactory for[w/H]=—0.8,butthedataarenot indicates thatthecomputedG-bandstrengthsare least afactorof10.(ComparisonFigs.1and2 dance inthestrongCNstarsmustbeenhancedbyat and Canterna1980),calculationsshowtheNabun- found inthegiantsbyDBG,viz.,[N/^4]=+0.7. may haveNabundanceenhancementsaslargethose shortward oftheKlineresemblesthatN-en- The models6000/4.5/0.0{thickline)and5500/4.5/0.0{thin instrumental profileof15Âhalf-width.Thespectrumthe Similarly, if[w/H]isaslow—1.2(Pilachowski, riched syntheticspectrumquitewell.Onthebasisof reasonable. Inparticular,thespectrumof3-2153just calculated spectrawiththeobservationsseemsquite are plottedusingadifferentzerofortheordinate. Can terna,andWallerstein1979;Pilachowski,Sneden, these comparisonswefeelthatsomedwarfsin47Tue model 5750/4.0/—0.8hasbeencomputedfortwovaluesofthe N abundance,[N/^4]=0.0{thinline)and+1.0{thickline). Fig. 2.—Examplesofsynthesizedspectra,convolvedwithan Our observationsweremade at f=14'(fidai44'; t IV. THEMEMBERSHIPQUESTION L151 198 OApJ. . .238L.149H -2 2 2 2 = 2 using themodelofBahcallandSoneira(1979).Com- of skyweexpect^0.01stars.Asimilarresultisfound matched bythespectrumofaPopulationIstar;and, function (HesserandHartwick1977)leadstoaprob- bining theseestimateswiththe47Tueluminosity if so,doanyotherpropertiesofthatPopulationIstar power-law halo,anexponentialdiskwithascaleheight disagree withtheobservedpropertiesofstrong a strongCN47Tuestar,suchas3-2153,canbe sample. ability of<10thatafieldstarcontaminatesour function) yieldsnumbersandcolordistributionsof ment withthemean(B—F)of0.56mag47Tue possible tomatchthespectrumof3-2153witha sample. TheB—Vcolorofthemodel5500/4.5/0.0 square degreeinthedirectionof47Tue.Thus increasing withage,andLuyten’s1968luminosity without significantlyaffectingthe syntheticspectra. for thesedwarfmodelshasnotyetbeentestedindetail, is 0.71mag.Althoughthereliabilityofcolorcalculations Bell (1979)andGustafsson(1978).Themodel of themodelsusingmethodsGustafssonand that thespectrumofmodel5500/4.5/0.0matches A Fbandsampledbyourobservationsover6arcmin B —F<0.675mag.theirmodelpredicts29starsper faint starsinaccordwithKron’s(1979)observations Bergh (1980)havefoundthattheirmodel(usinga by thestarcountswhenconservativeestimatesof loper starsareexpected,aconclusionstronglysupported occupied byfiveTable1stars,6arcmin,<0.2inter- by thestarsofTable1.Thus,inareaonsky stars visibleintheCMD(HesserandHartwick1977, et al.1968)andCMDstudies(HesserHartwick Tue sample.Useoflogg=4.2would giveagreementinF,caic Mho\ 3.86oxMy—3.93,i.e., slightlybrighterthanthe47 5750/4.0/—0.8 yieldsB—V=0.58mag,inagree- that of3-2153.Consequentlywecomputedthecolors models, where,bycomparisonwithFigure1,itisseen included inFigure2calculatedspectraofPopulationI realistic galactichalomodels.Pritchetandvanden functioneffectsareincluded. lie intheAF,A(Z3—F)main-sequencebandoccupied stars isknownfromstarcounts(Lindsay1967;King slightly hotterPopulationImodel,itstillappearsthat this lattercolorissubstantiallyredderthanthemean In therange17mag