Level of helium enhancement among M3's horizontal-branch stars

1.2 MB·

by  A. A. R. Valcarce

Astronomy&Astrophysicsmanuscriptno.avalcarceM3_2016v2 (cid:13)cESO2016
May3,2016
Level of helium enhancement
among M3’s horizontal branch stars
A.A.R.Valcarce1,2,3,M.Catelan1,2,3,J.Alonso-García4,3,R.ContrerasRamos1,3,andS.Alves1,3
1 PontificiaUniversidadCatólicadeChile,InstitutodeAstrofísica,FacultaddeFísica,Av.VicuñaMackenna4860,782-0436Macul,
Santiago,Chile
2 PontificiaUniversidadCatólicadeChile,CentrodeAstroingeniería,Av.VicuñaMackenna4860,782-0436Macul,Santiago,Chile
3 MillenniumInstituteofAstrophysics,Santiago,Chile
6
4 UniversidaddeAntofagasta,UnidaddeAstronomía,FacultadCs.Básicas,Av.U.deAntofagasta02800,Antofagasta,Chile
1
0 ReceivedApril22,2015;acceptedJanuary25,2016
2

y
ABSTRACT
a
M
Context.Thecolorandluminositydistributionofhorizontalbranch(HB)starsinglobularclusters(GCs)aresensitiveprobesofthe
originalheliumabundances ofthoseclusters.Inthissense,recentlythedistributionsofHBstarsinGCcolor-magnitude diagrams

2 (CMDs)havebeenextensivelyusedasindicatorsofpossiblevariationsintheheliumcontentY amongthedifferentgenerationsof
starswithinindividual GCs. However, recent analyses based onvisual and near-ultraviolet(UV) CMDshaveprovided conflicting

] results.
R
Aims.Toclarifythesituation,weaddresstheoptimumrangesofapplicability(intermsoftheT rangecoveredbytheHBstars)for
eff
S visualandnear-UVCMDs,asfarasapplicationofthis“HBYtest”goes.
. Methods.WeconsideredbothStrömgrenandHubbleSpaceTelescope(HST)bandpasses.Inparticular,wefocusontheF336Wfilter
h
of the HST, but also discuss several bluer UV bandpasses, such as F160BW, F255W, and F300W. Using the Princeton-Goddard-
p
PUC(PGPUC)code,wecomputedalargesetofzero-ageHB(ZAHB)lociandHBevolutionarymodelsformassesrangingfrom
-
o MHB = 0.582to0.800M⊙,assuminganinitialheliumabundanceY = 0.246,0.256,and0.266,withaglobalmetallicityZ = 0.001.
r TheresultsofthesecalculationswerecomparedagainsttheobservationsofM3(NGC5272),withspecialattentionontheyvs.(b−y)
t andF336Wvs.(F336W−F555W)CMDs.
s
a Results.Ourresultsindicatethat,fromanevolutionaryperspective,thedistributionsofHBstarsintheyvs.(b−y)planecanbea
[ reliableindicatoroftheHecontentincoolblueHB(BHB)stars,particularlywhenadifferentialcomparisonbetweenblueandred
HBstarsiscarriedoutintherangeTeff .8300K.Conversely,wedemonstratethatCMDsusingtheF336Wfilterhaveamuchless
2 straightforwardinterpretationatthecoolendoftheBHBbecausethedistributionsofHBstarsintheF336Wvs.(F336W−F555W)
v plane,forinstance,areaffectedbyatripledegeneracyeffect.Inotherwords,thepositionofanHBstarinsuchaCMDisexactlythe
7 sameforagivenchemicalcompositionformultiplecombinationsoftheparametersY,M ,andagealongtheHBevolutionarytrack.
HB
4 OtherHSTUVfiltersdonotappeartobeasseverelyaffectedbythisdegeneracyeffect,towhichvisualbandpassesarealsoimmune.
7 Ontheotherhand,suchnear-UVCMDscanbeextremelyusefulforthehotteststarsalongthecoolBHBend.
6 Conclusions.BasedonareanalysisofthedistributionofHBstarsintheyvs.(b−y)plane,wefindthatthecoolestBHBstarsinM3
0 (i.e.,thosewithT <8300K)areverylikelyenhancedinheliumby∆Y ≈0.01,comparedwiththeredHBstarsinthesamecluster.
eff
1. Using near-UV HSTphotometry, on the other hand, we findevidence of a progressive increase inY withincreasing temperature,
0 reaching∆Y ≈0.02atTeff ≈10900K.
6 Keywords. globularclusters:general-–globularclusters:individual(M3=NGC5272)-–Hertzsprung–RussellandC–Mdiagrams-
1
–stars:abundances-–stars:evolution-–stars:horizontal-branch
:
v
i
X
1. Introduction NGC2808showsthreediscreteMScomponents,stronglysug-
r gestinglargeHevariationsfromonesuchcomponenttothenext
aGlobularstarclusters(GCs)arefarfrombeingthesimplestellar
(Piottoetal.2007),eventhoughtheclusterdoesnothavealarge
populationsthattheywereoncethoughtto be.Thepresenceof
spread in metallicity (Grattonetal. 2011). In like vein, a bluer
multiplepopulations,correspondingtodifferentepisodesofstar MScomponentinωCenalsoappearstorequireamuchhigher
formation within individual GCs, is now routinely revealed by Heabundancethanwouldhavebeenimpliedbytheirdifference
photometry and spectroscopy alike. One important open prob-
inmetallicity(Piottoetal.2005).Eventhoughitisbelievedthat
lemisthedegreetowhichthedifferentpopulationsmaydifferin
mostofthestellarpopulationsofGCsfollowahelium-to-metal
theirheliumcontent,andhowanyinternalspreadinthehelium enrichment law (Y ≈ Y + ∆Y/∆Z × Z, with ∆Y/∆Z ≈ 2.0;
p
abundanceY maydifferfromoneclustertothenext.
Renzini1994;Salarisetal.2004;Valcarceetal. 2013),thedis-
The discovery of multiple main sequences (MSs) in crete main sequences and the different chemical compositions
NGC 2808 (Norris 2004; D’Antonaetal. 2005; Piottoetal. observedamongstarsinthesameGCsuggestedtheexistenceof
2007) and ω Centauri (NGC 5139; Bedinetal. 2004; helium-richstarsineveryMilkyWayGC(Grattonetal.2012).
Bellinietal. 2009, 2010) has reopened the question about the Nevertheless,thefractionoftheseHe-richstarsinGCsandthe
initial helium abundanceof the GC stars (vandenBergh 1965,
1967; Sandage&Wildey 1967; Hartwick 1968). Remarkably,
Articlenumber,page1of13
A&Aproofs:manuscriptno.avalcarceM3_2016v2
amountofheliumenrichmentineachclusterarestillamatterof andnear-UVregimes,forsettingconstraintsontheinitialhelium
debate. abundancesofthedifferentstellarpopulationsinGCs.
Because the surface helium abundance can only be mea-
sured spectroscopically in stars with an effective temperature
T & 8000 K, only GCs presenting relatively blue horizontal
eff
branch (HB) extensions can harbor stars with sufficiently high 2. Estimating∆Y fromHBpopulations
temperatures to enable direct spectroscopic estimations of the
helium abundances. However, these stars have passed through Since the discovery of multiple populations in GCs, two main
the whole MS, red giant branch (RGB), and pre-HB phases, approacheshavebeenusedto estimatedifferencesinthe initial
whichhasmodifiedtheirsurfacechemicalcompositionthrough helium abundances in the HB population. The first approach
the effects of internal mixing (e.g., Sweigart&Mengel 1979; involves the computation of synthetic HBs, which can then
Vandenberg&Smith 1988), and they might also have been af- be compared to the distribution of stars observed along the
fectedbystellarrotation(e.g.,Mengel&Gross1976),massloss HB (e.g., Rood 1973; Catelanetal. 1998; D’Antonaetal.
(e.g.,Peterson1982;DeLaRezaetal.1996),and/orinteraction 2002; Caloi&D’Antona 2008; D’Antona&Caloi 2008;
with other bodies (planets or stars, Soker 1998; Siess&Livio Dalessandroetal. 2011, 2013; Salarisetal. 2013; Leietal.
1999),amongothereffects(seeCatelan2009). 2013a,b).Inthesecondapproach,referenceevolutionarylociare
EvenforthesehotHBstars,spectroscopicmeasurementsof directly overplotted on the empirical CMD data (Catelanetal.
helium unfortunately remain complex because these lines are 2009, hereafter C09; Brownetal. 2010; Valcarceetal. 2014,
weak and a high signal-to-noise ratio (S/N) is required to de- hereafterV14).
rive accurate results. Moreover, it is well established that GC
Bothapproachesarebasedonextensivecalculationsofevo-
HB stars with T & 11500 K show higher metal and lower
eff lutionarymodelsforGCstars,calculatedaccordingtothechem-
helium abundances than all the other stars of the same GC.
ical composition of the GC under study. More specifically, the
This is caused by the effects of metal levitation and He sed-
complete evolution of a star from the main sequence to the
imentation (Grundahletal. 1999; Behr 2003), which limit the
ZAHBiscalculated,mostoftenwithoutmassloss,with there-
range of temperatures for measuring helium to a window of
sultingmaximumZAHBmassdependingontheGCproperties
∆T ≈ 3500 K. Only recently, Villanovaetal. (2012) and
eff (e.g., for 13 Gyr, Z = 0.001, and Y = 0.245, the maximum
Marinoetal.(2013)haveobtainedhigh-precisionspectralmea-
ZAHBmassis 0.810M ).ThisreferenceZAHBmodelis then
surementsofheliumabundancesinthestarsofM4(NGC6121) ⊙
usedasastartingpointtocreatei)theZAHBlocusforallstars
and NGC 2808, respectively. We still lack a complete under-
with the same M and initial chemical composition (i.e., all
standing, however, of how these photospheric abundances are cHe
starsbornatthesametime),andii)theinitialmodelforHBstars
relatedtotheinitialabundancesafterthewholepreviousevolu-
atagiventotalmass(M )withthesameM anddifferenten-
tion, given in particular their dependenceon details of the first HB cHe
velope mass (i.e., due to mass loss during the RGB evolution;
dredge-upepisodeatthebaseoftheRGB.Thisphenomenonis
Serenelli&Weiss2005).Thenwecomputeforeach M value
onlypoorlydescribedbycanonicalevolutionarymodelswithout HB
theevolutionfromtheZAHBtotheheliumcoreexhaustion(also
extramixing(e.g.,Grattonetal.2000).Otherspectroscopiche-
called terminal-ageHB, or TAHB). The modelsare then trans-
liumabundancestudiesincludethoseofDupreeetal.(2011)and
formedtotheempiricalCMDplanes,usingsuitablebolometric
Pasquinietal. (2011) for RGB stars in ω Cen and NGC 2808,
correctionsandcolortransformations.
respectively. These authors used chromospheric helium lines
(whichhavemoretraditionallybeenusedasdiagnosticsofmass InthecaseofsyntheticHBpopulations,thecolorandabso-
lossinRGBstars)thatrequireacarefultreatmentofthestellar lutemagnitudedistributionsofstarsalongtheHBarethensim-
atmosphere(Dupree&Avrett2013). Thesedifficultieshavefa- ulated, using Monte Carlo methods. This requires an assump-
voredindirectmethodstobeproposedasprobesoftheHeabun- tiononthemass-lossdistributionontheRGB,asthisdefinesthe
danceinGCstars. resulting distribution in masses on the HB phase proper. Most
Forthispurpose,theeffectsofdifferentinitialheliumabun- commonly, mass-loss efficiency is treated as a free parameter
thatis constraineda posteriorito match the observations.Stars
dancesoncolor-magnitudediagrams(CMDs)havebeenknown
arethusfedintothe evolutionarytracksofdifferentmassesac-
since the late 1960s (Simoda&Iben 1968; Aizenmanetal.
1969; Iben 1974; Simoda&Iben 1970; see also Valcarceetal. cordingtotheexpectedevolutionaryspeedalongtheHBphase.
Photometric errors are then added to the photometric data for
2012, for a recent summary).These studies revealed that, even
though these effects are present in all the evolutionary se- the “synthetic stars” to more realistically compare them with
quencesalongtheCMD,theeffectofHeisamplifiedintheHB observationaldatasets. Dozensofsimulationsaretypicallyrun,
for which some or all of the free parameters are varied to find
phaseandis alsoeasierto detectbecauseofthehighluminosi-
the distribution that most closely describes the empirical data.
tiesofHBstars(Faulkner&Iben1966;Iben&Faulkner1968;
Iben&Rood1970;Sweigart1987).Specifically,anincreasein Someofthesecomparisonsaremadeusingcolorandmagnitude
Y increases the luminosity of all HB stars with T . 16000 histograms,butin othercases, onlya visualinspectionis used,
eff
K, but it also decreases the luminosity of these stars for T & whichcansometimesleadtomisinterpretations.
eff
16000 K. These effects are due to different efficiencies of the Whilethereferenceevolutionarylociusedinthesecondap-
H-burningshell and to the differencesin the helium-coremass proachcanalsobeobtainedfromsyntheticHBmodels,theyare
(M ;Valcarceetal.2012)withchangingY. moredirectlyretrievedfromtheHBevolutionarytracksproper.
cHe
However, depending on the set of filters used to transform We have already encountered two such reference loci, namely
fromthetheoreticaltotheobservationalplane,theseeffectscan the ZAHB and TAHB. Additional loci may easily be defined.
become more or less evident in the HB part of the CMD, and They represent the fraction of the entire HB lifetime that a
thusmoreorlesseasilydetectable.Inthispaper,weexplorethe given HB has already gone through.For instance, C09 defined
effects on the HB of using two different sets of filters, Ström- the middle-age HB (MAHB) locus that closely corresponds to
grenandHubbleSpaceTelescope(WFPC2),coveringthevisual starsthatarehalf-waythroughtheirHBevolution,andalsothe
Articlenumber,page2of13
A.A.R.Valcarceetal.:HeenhancementamongM3’sHBstars
90AHB,whichcorrespondstostarsthathavecompleted90%of consensusas yeton whetherthereis a He spread,andif so, by
theirlivesasHBstars.1 howmuch.One of ourmain goalsis to explainthe reasonsfor
IfagivenGCcontainsasinglepopulationofstars,onewould seemingly conflicting results based on similar samples of HB
then expect to find ≈ 50% of all HB stars between the ZAHB starsinthisonecluster.
and MAHB, ≈ 40% between the MAHB and 90AHB, ≈ 10% In the following, we assume that there is no difference
between the 90AHB and TAHB, etc. (see also C09). However, in either Z or C+N+O among stars in M3 (σ = 0.03;
[Fe/H]
if there are several populations of HB stars with similar tem- Snedenetal. 2004) so that we can compare the expected and
peraturesbutdifferentinitialheliumabundances,thesepercent- observedluminositydifferencesusingdifferentsetsoffilters.
ages should change as a result of the increase in luminosity of Figure 1 showsthe HB populationof M3 on a CMD based
HB stars with the increase of the initial helium abundance. In ontheyvs.y−bfiltersoftheStrömgrensystem.Similarly,Fig.2
particular, if there is a spread in the initial helium abundance showsthe F336W vs. F336W −F555W CMD, usingthe HST
∆Y = 0.01in a GC andmodelsfora single Y are comparedto Wide-FieldandPlanetaryCamera2(WFPC2)visible(F555W)
photometric data, the percentages of stars located between the and near-UV (F336W) filters, respectively. M3 observational
MAHB and 90AHB locishould be largerthan expectedon the data in the Strömgren filters are taken from Grundahletal.
basisofsingle-Ymodels.Thustheevolutionarylocimethod,like (1998,1999)asdescribedinC09withatotalof138non-variable
the synthetic HB populationsmethod,can be a powerfulprobe HB stars, while data in the HST filters were kindly provided
ofthepresenceofpopulationswithdifferentHeabundancesthat byE.Dalessandro(2014,priv.comm.)withatotalof155non-
co-existinaGC. variable HB stars. In these figures are also plotted theoretical
While HB stars with different Y values can coex- ZAHB loci for a chemical composition according to the M3
ist over a given temperature range, theoretical arguments metallicity (Z = 0.001 and [α/Fe] = 0.30;see C09) and three
(Sweigart&Gross1976)aswellasrecentspectroscopicobser- initial helium abundances Y = 0.246, 0.256, and 0.266 as red,
vations(Marinoetal. 2011; Villanovaetal. 2012; Marinoetal. blue, and green lines, respectively. HB evolutionary tracks for
2013) alike suggest a tendency for blue HB (BHB) stars to be Y = 0.246 are also plotted as red dot-dashed lines for some
preferentiallyHe-enhanced,comparedwithredHB(RHB)stars masses (M = 0.582, 0.596, 0.609, 0.632, 0.652, 0.665, and
HB
in the same cluster.2 As a consequence,if a spread in He is in- 0.800M ).Thesearemarkedwithcirclesatthreestagesoftheir
⊙
deed present within a GC with both blue and red HB compo- evolution–ZAHB,MAHB,and90AHB–withdifferentcolors
nents,onewouldexpecttheBHBtobenoticeablybrighterthan beingusedfordifferentM values.Thesefiguresareexplained
HB
theRHBinthesamecluster.Thesizeoftheeffect,overtheoth- inmoredetailbelow.
erwise ratherflat partofthe HB, is about0.03mag in Johnson
V orStrömgrenyforadifferenceinY of0.01.Wecallthiscom-
3.1.Previousstudies
parisonbetweenthebrightnessofRHBandBHBstarsusingthe
ZAHBandotherreferencelocitheHBY test. ThemorphologyoftheHBpopulationofM3([Fe/H]= −1.50)
It is worth mentioning that stars with high helium abun- wasstudiedinD13togetherwiththatoftwootherGCs,namely
dances could in principle also have suffered an increase in Z M13=NGC6205andM79=NGC1904,whichhaveverysim-
or C+N+O, as suggestedrecently by Jangetal. (2014) in their ilar [Fe/H] ratios (−1.53 and −1.60, respectively; Harris 1996,
proposed explanation of the Oosterhoff dichotomy observedin 2010edition).D13comparedtheresultsofsyntheticHBcalcu-
Milky Way GCs (Oosterhoff 1939; see also Catelan 2009, for lations to HST near-UV observations that were obtained with
a recent review and references). This can also favor helium- WFPC2. These simulations require four free parameters (the
normal and helium-rich stars sharing similar colors on the HB minimum Y value, the range of helium abundances, the mean
ofa givenGC,andthereforeservesasawarningagainstsolely value of the mass lost along the RGB, and the spread around
usingcolorinformationasanindicatorofHespreadsamongGC thismeanvalue)foran assumedageof12Gyr.Theamountof
stars.3 In particular, the percentages between ZAHB, MAHB, mass lost on the RGB is also expected to be a function of Y
90AHB,andTAHBlociareexpectedtochange,whencompared (e.g., Catelan&deFreitasPacheco 1993). Clearly, the number
totheexpectationsforasinglestellarpopulationbecauseofthe of possible parameter combinationsand the resulting synthetic
differenceinluminositythatisbroughtaboutbythepresenceof distributionsis verylarge,even thoughin D13 it is onlypossi-
aheliumspread. bletoexaminethebest-fittingsimulationforeachGC.ForM3,
thesesyntheticHBcalculationssuggest,accordingtoD13,that
HBstarswith0.2.(F336W−F555W).0.5haveadispersion
3. TheM3caserevisited
inY of0.02,butthedispersioninY advocatedbyD13forM13
HereweconsiderM3(NGC5272)becausethisclusterhasbeen isonly0.01overthesamecolorrange.Althoughsmall,thepres-
the subject of numerous recent studies of the HB population enceofsuchadifferenceisnoteworthy,giventhesimilarbright-
(C09; Dalessandroetal. 2013, hereafter D13), without a clear nessextensionsinF336WoftheHBpopulationsinbothclusters.
Thisdifferencemightplausiblybetracedbacktobothclusters’
1 C09point out thattheMAHB (originallydefined astheHB ridge- strikinglydifferentHBmorphologies:M3andM13constitutea
line in the CMD) and 50AHB (defined in complete analogy with the “classical”second-parametercase,andoneofthefirstforwhich
90AHB)locimaydifferslightly,butherewedefinetheMAHBasbe- chemical anomalies, as opposed to age, were suggested as the
ingsynonymouswith50AHB. explanation(Catelan&deFreitasPacheco1995).
2 BHB and RHB stars are terms that are often used to refer to stars
The population of HB stars was studied by C09 using the
bluer and redder, respectively, than the RR Lyrae instability strip, re-
visualfilters of the Strömgrensystem, andthe resultingCMDs
spectively.
are compared to different reference evolutionary loci (ZAHB,
3 Thereareother possibilitiesforHe-richandHe-poor starstoshare
MAHB,90AHB,andTAHB;seeSect.2).Ascanbeseenfrom
similarcolorsontheHB.Forexample,starswiththeexactsameZand
C+N+Obutdifferentmass-losshistoriesmayverywellhaveexactlythe Fig.2inC09,theevolutionofHBstarsinthesefilters(My, Mb,
sameZAHBtemperature,eventhoughtheirluminositiesmaybevastly and Mv) implies that, in the canonical scenario, brighter stars
different. (at a given temperature) are also more evolved stars, at least
Articlenumber,page3of13
A&Aproofs:manuscriptno.avalcarceM3_2016v2
15.55
15.6
15.65
PGPUC (Y=0.246)
14.5
PGPUC (Y=0.256)
PGPUC (Y=0.266)
15
15.5
16
16.5
0 0.2 0.4
(b-y)
Fig.1.M3HBstarscomparedtotheoreticalHBmodelswithZ=0.001intheStrömgrenyvs.(b−y)plane.Theupperpanelisazoom-inaround
theZAHBlocus.PGPUCZAHBlociareshownforheliumabundancesY =0.246(redline),0.256(blueline,shownintheupperpanelonly,for
clarity),and0.266(greenline).C09andBaSTIZAHBlociforY =0.246arealsoshownasmagentaandblacklines,respectively,withdifferent
distancemoduliinordertomatchthePGPUCZAHBlocuswithY =0.246.TheoreticalHBevolutionarytracksforY =0.246areshownasred
dot-dashedlinesfor M =0.582,0.596,0.609,0.632,0.652,0.665,and0.800M ,whichhavebeenmarkedwithcirclesattheZAHB,MAHB,
HB ⊙
and90AHB(colorsdependontheHBmass,forclarity).AnadditionalHBtrackforY = 0.256and M = 0.582M isalsoshowninthetop
HB ⊙
panel(bluedot-dashedlineswithpinkcircles).Verticaldot-dashedlinesindicatesomeT valuesattheZAHBlocusforY =0.246.Thedistance
eff
moduluswasselectedasdescribedinSect.3.3andFig.6.ThedatacomefromC09.
for T . 9500 K. However, since an increase in Y also in- vationaldatasetsusedbytheseauthors.Figure3isazoom-inof
eff
ducesan increasein the HB luminosityat similar temperatures Fig.2inthecolorrange0.2.(F336W−F555W).0.8.Inthese
(see,e.g.,Sweigart1987),anincreaseinY canbemimickedby figures we show HB evolutionarytracks (red dot-dashedlines)
differentagesalongtheHBevolutionarytracks(t ),andvice- with masses ranging from M = 0.800 down to 0.582M ,
HB HB ⊙
versa. However, as mentioned before, theoretical evolutionary with a progenitor mass of 0.800M , Y = 0.246, Z = 0.001,
⊙
modelspredictthat50%ofallHBstarsmustbefoundbetween and [α/Fe] = +0.3. To show the different evolutionary stages,
theZAHBandMAHBloci,or,similarly,betweentheZAHBlo- each HB track has been marked with coloredcircles at the po-
cuswithprimordialY valueandtheZAHBlocuswithaY value sitions in their evolution correspondingto the ZAHB, MAHB,
increasedby0.01,asthelatter’sZAHBcloselyoverlapsthefor- and 90AHB, the colors of the circles depending on the stellar
mer’sMAHB.Basedonthis, themainresultinC09isthatany mass. The distance modulus (m − M) = 14.909 is selected to
heliumspreadinM3mustaccordinglybe∆Y .0.01. reproduce the expected proportions of stars according to evo-
In summary,the proposedlevel of He enhancementamong lutionarystage along the HB phase (see Sect. 3.3).4 These fig-
M3 HB stars proposed by D13 is in conflict with the smaller ures also show the BaSTI (Pietrinfernietal. 2004) and C09
onesuggestedin C09 forstars overthe same rangeineffective ZAHBloci (in blackand magentacolors, respectively),shifted
temperature(i.e.,basicallyforthesamestars).Belowweaddress according to the indicated distance moduli in order to match
somepossiblereasonsfortheseconflictingresults. the PGPUC ZAHB locus for Y = 0.246 (red line). Note that
the different required shifts, hence distance moduli, are due to
differences in the adopted chemical compositions (in C09, the
3.2.Comparingmodelsinthevisualandnear-UV Y = 0.230 ZAHB locus is used), bolometric corrections (C09
andPGPUCmodelsuseClemetal.2004,whereasBaSTImod-
To look further into the reasons for the noted discrepancy be-
tween the results of C09 and D13, we show in Figs. 1 and 2 4 Eventhoughthedistancemodulusislowerthanreportedinthelitera-
HB evolutionary tracks and reference loci in CMDs based on ture(e.g., VandenBergetal.2013),thisdoesnotaffectthemainresults
thesamefiltersasusedinC09andD13togetherwiththeobser- ofourpaper.
Articlenumber,page4of13
A.A.R.Valcarceetal.:HeenhancementamongM3’sHBstars
15.9
16
15.6
15.8
16
16.2
16.4
-0.5 0 0.5 1
Fig.2.SameasFig.1,butfortheHSTF336Wvs.(F336W−F555W)plane.Theoreticalmodelsfor M = 0.665and0.632 M arenotshown
HB ⊙
hereforclarity(seeFig.3).TheobservationaldataarethesameasinD13,askindlyprovidedbyE.Dalessandro(2014,priv.comm.).
elsuseCastelli&Kurucz2004),and/orinputphysics(morede-
tails can be found in Sect. 2.7 of Valcarceetal. 2012). Blue
andgreenlinesrepresentthePGPUCZAHBlociforY = 0.256
and Y = 0.266, respectively. All these ZAHB loci are based
onthesamechemicalcompositionmentionedbefore,exceptfor
the C09 model, whose higher metallicity (Z = 0.002)explains
its further extension toward red colors, compared to the other
ZAHB loci shown. Finally, vertical dash-dotted lines schemat-
ically indicate some specific T values (based on the ZAHB
eff
loci), as were used to compare different sets of filters. For the
sake of clarity, evolutionarytracks for 0.665 and 0.632M are
⊙
notshowninFig.2,andneitheristheC09ZAHBlocus.
A quick inspection of these figures reveals that the Y-t
HB
degeneracy existing in the HB with the filters used in Fig. 1
(y vs. b − y plane) can induce a misinterpretation not larger
than ∆Y = 0.01 (compare the blue line in the upper panel
with the overplottedevolutionarytracks) because roughly50%
of the HB stellar population with Y = 0.246 must be brighter
thanthe ZAHB locuswith Y = 0.256in the temperaturerange
5700K.T .8300K,regardlessoftheexistenceofahelium
eff
spread.
However, the filters used in Fig. 2 (F336W vs. F336W −
F555Wplane)presentaY −t −M tripledegeneracyatthe
HB HB
HBlevel.Specifically,fromredtobluecolors,theZAHBlocus
Fig. 3. Zoom-in of Fig. 2 around the loop shown by the ZAHB loci. startswith0.800M⊙atacoloraround0.6,andwhenthemassis
HereareshownallthecomputedHBevolutionarytracks.Notethedif- decreasedto∼0.665M⊙,thecoloris0.4–,butforlowermasses,
ficultydisentanglingavariationinY frommassandevolutionarystate, theZAHBlocusbeginstodevelopaloop.Inotherwords,afur-
atcolorsaround(F336W−F555W)≈0.4. therdecreaseinmass,insteadofproducingabluerstructurewith
roughlythe same brightnessas in Fig. 1, gives rise to a redder
Articlenumber,page5of13
A&Aproofs:manuscriptno.avalcarceM3_2016v2
Fig.4.SyntheticCMDsattheHBlevelfrommodelswithZ=0.001inthefiltersofinterestoftheStrömgren(toppanels)andWFPC2(middleand
bottompanels)systems.TheZAHBlocusforY = 0.246isrepresentedasablackline.Auniformmassdeviaterangingfrom0.595to0.705M
⊙
with2100starsisusedtorepresentahelium-normalpopulation(Y = 0.246,lightgraycircles).Helium-richpopulations(Y = 0.256,0.266)are
createdwith1000starswithmassesrangingfrom0.595to0.645M (darkgraycircles).StarsidentifiedasRHBstarswithY =0.246areplotted
⊙
asredcircles.BHBstarsareplottedincyanforY =0.246,blueforY =0.256,andgreenforY =0.266.Inthesesimulations,photometricerrors
are included, following thephotometric data, at the level of 0.01 mag ineach bandpass. Panelsg, h, and i arezoom-ins of panels d, e, and f,
respectively.Eachpanelisaccompaniedbyhistogramsshowingthedistributionincolor(top)andmagnitude(right)ofthedifferentcomponents.
andbrighterCMDposition.Apeakinbrightnessisreachedfor in and around the temperature range where the loop occurs,
a ZAHB mass of 0.652M . When the mass is decreased even as indicated in Fig. 2 – that is, between about 5800K and
⊙
more, one finally returns to a color of 0.4, which occurs for 8300K. Unfortunately,many M3 HB stars live at this temper-
M =0.632M .Finally,theusualmonotonicbehaviorisagain ature range, including all of its RR Lyrae and the redder of its
HB ⊙
recoveredforevenlowermassesorhighertemperatures. BHB stars. Note that 72±7% of the M3 total HB population
hasT < 8300K.Inaddition,asaresultofphotometricerrors
eff
This non-monotonic behavior and the resulting triple de- (see the error bars plotted in Fig. 3), this effect potentially af-
generacy in HB parametersseriously hinders the interpretation fectsthe evolutionaryinterpretationof notonlythose few stars
of photometricobservationswith the F336Wfilter, particularly
Articlenumber,page6of13
A.A.R.Valcarceetal.:HeenhancementamongM3’sHBstars
that are located within the exactboundariesof these loops, but devoidofsucha loopfeature(Fig.5). F300Wisnoteworthyin
alsoofstarsintheirimmediateneighborhood,inthecolorrange thiscontext,astherelevantCMDsappeartoattempttodevelop
0.34.(F336W−F555W).0.45,whichcomprisesasignificant asimilarloop,butdonotquitesucceedatthetask.
fraction(28±5%)ofthenon-variableHBstarsinM3.Notethat Clearly, then, the loop feature affecting F336W is caused
this fraction increases to 55± 6% when RR Lyrae stars occu- solely by the corresponding bolometric corrections. Over the
pyingsimilarcolorsarealsoincluded(Catelanetal.2001,their corresponding temperature region, and as we have mentioned
Table1forr<50′′). previously, the degeneracy involves a parameter in addition to
The preceding analysis can be further explored using syn- Y and t , namely the mass M , as stars with different M
HB HB HB
theticHBmodels,asshowninFig.4,wheresyntheticCMDsat valuesbutthesame(F336W−F555W)canhavevastlydifferent
theHBlevelareshownintheStrömgrenandWFPC2 systems. absolute magnitudes in F336W (Figs. 2-3). Close examination
In each case, a uniform mass deviate was used with 200 HB of these figures, alongwith the synthetic HBs shownin Fig. 4,
stars per0.01M , leadingto T andlogL valuesfor eachstar suggeststhatthistripledegeneracyphenomenoncanlead toan
⊙ eff
which are then transformedinto the observationalplanesusing incorrect reading of the He abundance that may be as high as
theClemetal.(2004)andCastelli&Kurucz(2004)bolometric ∆Y = 0.02, in the specific case under study. Note, in particu-
correctionsforyvs.(b−y)andF336Wvs.(F336W−F555W), lar, that HB stars with Y = 0.246 starting their evolution close
respectively. The photometric uncertainties, in all bandpasses, totheloopregionwillbecomeasbrightastheZAHBlocusfor
are at the levelof 0.01magon averageand were also modeled Y = 0.266bythetimetheyreachtheMAHB(see evolutionary
after normaldeviates. Theleft panels(a, d, andg) show a stel- tracksforMHB =0.665and0.652M⊙inFig.3).
lar populationwith Y = 0.246with masses between 0.595and Based on this HB Y test of the presence of He enhance-
0.705M (lightgraycircles),andpanelsbande(candf)showa ment and internal spreads in GCs, our suggestion would be to
⊙
helium-enrichedstellarpopulationwithY = 0.256(Y = 0.266) avoid relying excessivelyon the F336Wfilter overthe temper-
withmassesbetween0.595to0.645M (darkgraycircles). ature range 5800 . T [K] . 8300. Other near- and far-UV
⊙ eff
To show how the high percentage of BHB and RHB stars filters can be safely used, as can F336W outside this tempera-
around the aforementioned loop feature mix in magnitude and turerange.Naturally,fortemperaturesinexcessoftheGrundahl
color for (F336W−F555W), RHB stars are cleanly selected in jumpatTeff ≃11500K,itisnecessarytoproperlytakeintoac-
theStrömgrensystemasthosewith0.28≤ (b−y),andBHBas counttheeffectsofheliumsedimentationandradiativelevitation
thosewith0.02≤(b−y)≤0.13andy≤15.7.Thesamestarsare ofheavierelements(Grundahletal.1999),asdone(albeitinan
then accordinglylabeled in the CMDs that use the WFPC2 fil- approximateway)forNGC2808byDalessandroetal.(2011).
ters.TheRHBpopulationswithY =0.246areshownasredcir-
cles, whileBHBpopulationswithY = 0.246,0.256,and0.266
3.3.Onthe∆Y levelamongM3’sBHBstars
areshownascyan,blue,andgreencircles,respectively.Asex-
pected,intheStrömgrenyvs.(b−y)CMD,thecolordistribution Followingtheseresults,weusedthemethoddescribedinV14to
separatesalmostcompletelyBHBstarswithY =0.256or0.266 obtainanewestimateofthedifferencein(initial)heliumcontent
fromRHBstarswithY = 0.246,exceptforasmallfraction(of betweenRHB(0.28<b−y<0.44)andBHB(0.060<b−y <
about 5% or less) corresponding to highly evolved BHB stars 0.135) stars in M3. While we used the same data in the visual
(tHB > 90%).Theupperpanelsshowthattheyvs.(b−y)plane StrömgrenbandsasinC09,herewerestrictthebluestcolorfor
canbeusedreliablyonlyfor(b−y) & 0.06(Teff . 8300K)be- BHB stars as described in the previous section, and study the
cause of the observationalerrors. Thus, unless the photometric variation in the relative proportions of stars distributed in the
errors are smaller than indicated and/or these errors are prop- CMD between the ZAHB, MAHB, 90AHB, and TAHB loci in
erly taken into account in the analysis, no safe constraints can greaterquantitativedetailfordifferentadopteddistancemoduli.
beplacedonthelevelofHeenhancementforsuchhotHBstars We allow for the possibility of a formal solution in which the
basedsolely(asinC09)ontheyvs.(b−y)CMD. distancemodulus,asinferredonthebasisof theBHB starsfor
For the F336W vs. (F336W − F555W) plane, we find, as anassumedY value,isdifferentfromtheonederivedusingthe
expected, that the color distribution around the loop cannot be RHBstarsforthesameY.Sincephysicallybothdistancemod-
used to separate BHB from RHB stars in the range 0.34 . uli shouldobviouslybe the same, any such differencesmay be
(F336W−F555W) . 0.45(8300 & Teff[K] & 5800).In turn, indicativeofadifferenceinintrinsicluminosity,henceintheHe
luminositydistributionscanbeusedtoidentifytheexistenceof abundance,betweenthetwoHBcomponents.
a helium-enrichedBHB population,butonly if RHB and BHB The resultsare shownin Fig. 6, where theleft panelsshow
populationshavesimilarnumbersofstarspopulatingtherelevant thevariationsinthenormalizedfractionofstarsatdifferentevo-
color range and if the statistics is high. Since it is well known lutionarystagesforRHBstars(upperpanel)andBHBstars(bot-
thathelium-enrichedpopulationsoftencompriseonlyafraction tompanel)asafunctionoftheadopteddistancemodulus.Since
of the helium-normal population in GCs, a reliable characteri- photometric errors can induce variations in these fractions, we
zation of the stars around the loop feature in terms of their He consideredatthesametimethenormalizedfractionofstarsthat
enhancementlevelsisverychallenging. are(1)fainterthantheMAHB(thinblacklines),(2)betweenthe
IstheloopfeatureonlyachallengeforF336W,orareother MAHB and the 90AHB (magenta lines), and (3) brighter than
UVfilterssubjecttothesameproblem? the 90AHB (cyan lines). The corresponding normalized frac-
To answer this question, we applied the same HB models tions are expected to be around 50%, 40%, and 10% for each
as shown in Fig. 2–3 to other UV and optical filters in the ofthesethreeranges,respectively.
HST set, using bolometric corrections from Castelli&Kurucz TherightpanelsinthesamefigureshowtheCMDsofRHB
(2004). The filters included in this analysis are the following: (reddots) and BHB (blue dots)stars using the distance moduli
in the far- and near-UV, F122W, F160BW, F170W, F185W, thatbestreproducetheexpectedevolutionaryproportionsshown
F218W, F255W, F300W, and F380W; in the visible, F439W, intheleftpanels(dottedanddashedverticallines,respectively).
F450W,F555W,F606W,F702W,F814W,andF814W.Interest- TheoreticalmodelsareshownforY = 0.246(HBtracksingray
ingly, CMDs producedusing all of these filters are completely lines,ZAHBlocusasaredline,MAHBlocusasabluedashed
Articlenumber,page7of13
A&Aproofs:manuscriptno.avalcarceM3_2016v2
Fig.5.ZAHBlociforY = 0.246andZ = 0.001fordifferentcombinationsofHST-WFPC2filters,asindicatedintheinsets.F336Wistheonly
filterproducingtheloopalongtheZAHBlocus(thinblacklineintheupperrightpanel).
line,90AHBlocusasadottedmagentaline,andTAHBlocusas errors,ofasmallpopulationwithlowerY,and/oroftheneedofa
alonggreendashedline)andforY = 0.256(ZAHBlocusonly, smallincreaseinthemetallicityofPGPUCmodels.Incidentally,
shownasablueline).InthebottomrightpaneltheZAHBlocus such a formal difference in distance moduli corresponds very
forY = 0.256usesthedistancemodulusobtainedfortheRHB closelytowhatwouldbeexpectediftheBHBcomponentwere
starsonlyasareferencelocus. moreHe-richthantheRHBcomponent,by∆Y ≈0.010±0.002
(wheretheindicatederrorbaristheformalvalueincludingonly
Figure6suggeststhattheRHBandBHBstellarpopulations
thePoissoniancomponent).AssumingthislevelofHeenhance-
doindeedappeartobebestcharacterizedbydifferencedistance
ment for the BHB component, the same distance modulus as
moduli. Specifically, the formal distance moduli for the RHB
obtainedfor the RHB providesan excellent match to the BHB
andBHBpopulationsare14.909+0.002and14.872±0.002mag,
−0.006 componentaswell,asshowninthebottomrightpanelofFig.6.
respectively;moreexactly,thedifferenceamountsto0.037mag
in y. Motivatedby the HB simulationsshown in the top panels Unfortunately, this method cannot be as straightforwardly
ofFig.4,weattributethefewRHBstarsthatarefainterthanthe used with the F336W filter because of the difficulties that are
ZAHB locus with Y = 0.246 as a consequenceof photometric broughtabout by the triple degeneracyeffect mentionedprevi-
Articlenumber,page8of13
A.A.R.Valcarceetal.:HeenhancementamongM3’sHBstars
1
0.8
0.6
0.4
0.2
1
0.8
0.6
0.4
0.2
14.85 14.9
Fig.6.DistancemodulusdeterminationfortheStrömgremfiltersaccordingtothefractionofstarsineachevolutionarystagealongtheHB,for
Z =0.001andY =0.246.Theupperandlowerpanelsontheleftshowthenormalizedfractionsforstarswith0.28<(b−y)<0.44(RHBstars)
and0.060 < (b−y) < 0.135(BHBstars),respectively,forfiltery.Colorlinesrepresentthenormalizedfractionofstarsthatare(1)fainterthan
theZAHBlocus(redlines),(2)betweentheZAHBandMAHBloci(bluelines),(3)betweentheMAHBand90AHBloci(magentalines),(4)
betweenthe90AHBandTAHBloci(cyanlines),(5)fainterthantheMAHBlocus(thinblacklines),and(6)fainterthanthe90AHBlocus(thick
blacklines).Verticaldottedlinesrepresentthedistancemodulususedintheupperrightpanel(andFig.1),andtheverticaldashedlineinthelower
rightpanelrepresentsthedistancemodulusthatshouldbeselectedifonlyBHBstarsareconsidered.InthebottompaneltheZAHBlocuswith
∆Y =0.01isshiftedaccordingtothedistancemodulusobtainedfromtheRHBpopulationforcomparison.
ously.ThisisnotonlyafeatureobservedintheF336Wfilter,but ZAHB locuswith Y = 0.256.This can plausibly be associated
alsoinfilterssamplingthesamewavelengthrange(seeSect.4). withanM3componentwithasmallheliumenrichment,thatis,
Inparticular,Fig.7showsM3HBstarsintheStrömgremuvs. with ∆Y ≈ 0.01, with ∆Y possibly increasing by an additional
(u−y)planewithstarscolor-codedasinFig.6.Clearly,because 0.005−0.01asthebluelimitofthisrangeisapproached.Over
RHBandBHBstarsoverlapat(u−y) ≈ 1.75,aninterpretation thiscolorrange17±4%ofthenon-variableHBstarsarelocated,
oftheseCMDsisfarfrombeingstraightforward. whichcorrespondsto11±2%ofthetotalHBpopulationinM3
Note, however,that overthe colorrange0.10 . (F336W− (consideringRRLyraestars;Catelanetal.2001).Similarly,over
F555W).0.30,Fig.2revealsalackofHBstarsfainterthanthe the color range −0.32 . (F336W−F555W) . 0.10, the same
Articlenumber,page9of13
A&Aproofs:manuscriptno.avalcarceM3_2016v2
Phase T range[K] ∆Y %non-variables %totalHB Maindiagnostic
eff
RHB+V <7500 <0.01 32±5% 58±6% Strömgren(visual)
coolBHB 7500↔ 8300 0.01 21±4% 13±3% Strömgren(visual)
BHB 8300↔ 9300 <0.02 25±4% 15±3% WFPC2(nearUV)
hotBHB 9300↔10900 0.02 16±3% 10±2% WFPC2(nearUV)
eBHB >10900 ? 6±2% 4±1%
Table1.AmountofHeenhancementofM3BHBstarscomparedtoRHBstars.Fractionsarecalculatedconsideringonlynon-variablestarsand
thatthetotalHBpopulationfollowsarelationB : V : R = 0.47 : 0.37 : 0.16,whereB,V,Rstandforthenumberofblue,variable(RRLyrae),
andredHBstars,respectively(Catelanetal.2001).
figureshowsalackofHBstarsfainterthantheZAHBlocuswith
Y = 0.266thatcanbeassociatedwithanM3componentwitha
largerheliumenrichment(∆Y ≈ 0.02),comprising16±3%of
thenon-variableHBstars(10±2%ofthetotalHBpopulation).
Based on these results, we cannot rule out at present that M3
HBstarsthatarehotterthantheGrundahljumpmayhavebeen
subject to a higher level of He enhancement than their cooler
BHBcounterparts–butthesecompriseonlyabout4±1%ofthe
totalHBpopulationinM3(C09).
In Table 1 we summarize the level of He enhancement in-
ferred from Figs. 1 and 2, separating the different M3 compo-
nents according to their temperatures and colors. The number
fractions were determined assuming that the non-variable HB
starscomprise63%ofthewholeHBpopulations(Catelanetal.
2001).
Note that, even though multiple populations show differ-
ent radial distribution in some GCs (e.g., Lardoetal. 2011;
Miloneetal. 2012; Bellinietal. 2013) that may be reflected
in different BHB-to-RHB number ratios (Catelanetal. 2001;
Iannicolaetal. 2009), we did not observe a large difference in
theseratioswhentheStrömgrenphotometry(withafieldofview
of3′.75×3′.75Grundahletal.1999)wascomparedwiththeHST
WFPC2 photometry(with afield ofviewof 2′.5×2′.5),asboth
samplebasicallycoverthesame(central)regionofthecluster.
This means that for the range of temperatures covered by Fig.7.M3HBstarscomparedtotheoreticalHBmodelswithZ=0.001
theHBY test, asappliedinC09,in V14,andinthispaper,the intheStrömgrenuvs.(u−y)plane.Lineshavethesamemeaningasin
visual bandpasses of the Strömgren and Johnson filter systems Fig.1.RedandbluepointsrepresenttheRHBandBHBstars,respec-
clearlyhaveamuchmorestraightforwardinterpretationattem- tively,thatareusedtoobtainthedistancemodulusinFig.6.
peraturesbelow∼ 8300K,moredirectlyconstrainingthelevel
ofHeenhancementamongM3(andM4;seeV14)starsalikeat
∆Y .0.01,whichconfirmstheC09resultsforthebulkoftheHB arepredictedbyourmodels.LoopsalongtheZAHBarefound
inmostofthecaseswherefilterscoveringthesamewavelength
starsinthecluster.Athighertemperatures,however,thesesame
regimeareusedtodefinecolorindices.
bandpasseslosesensitivity,anditbecomesnecessarytoresortto
InFig.8weshowtheZAHBlociforY = 0.246,Z = 0.001
filters coveringthe UV regime,some of which(F336Win par-
ticular)leadtoCMDsthataredifficulttointerpretatlowertem- as black lines in the theoretical plane (panel a) and in empir-
ical planes associated with these four different filter systems,
peratures.UsingdataobtainedwiththeWFPCF336W,wewere
namely:Johnson-CousinsU vs. U −V (panel b); Strömgren u
abletosuccessfullyidentifysmallpopulationsofhotterHBstars
vs.u−y (panelc);HST/WFPC2 F336Wvs. F336W−F555W
in M3 thatverylikely haveslightly enhancedhelium,reaching
(panel d); Sloan u vs. u − g (panel e). To compare the range
upto0.02in∆Y. Thissupportssimilarconclusionsreachedby
of effective temperatures restricting the loop in the F336W vs.
D13forthishotHBcomponent.
F336W−F555Wplanewithothersetsoffilters,weshowlines
representing T = 6125K (red dashed lines), 6600K (green
eff
4. OntheZAHBloopinUbandsofadditionalfilter dottedlines)and7800K(bluedashedline)inthese panels.As
canbeseen, theSloanu vs. u−g planealsoshowsa loopfea-
sets
turealongtheZAHBlocus,butinthiscaseitisevenmoreim-
Since thisloopfeatureisobservedin the HST/WFPC2 F336W pressive,extendingasitdoesupto∼ 9500K.FortheJohnson-
vs. (F336W− F555W) plane (Fig. 2) and in the Strömgrem u Cousinssystem,theU−V colorindexmonotonicallydecreases
vs. (u − y) plane (Fig. 7), we now address its possible pres- alongtheZAHBasT increases,whichmakesthesefiltersim-
eff
ence in two other filter systems equipped with bandpasses that mune to the aforementioned triple degeneracy effect, and thus
cover a similar wavelength regime as F336W or Strömgren u, affordsacleanerseparationbetweenBHBandRHBstarsthanis
namely Johnson-Cousins (Johnson 1955; Cousins 1976) and thecasewiththeotherthreefiltersystemsthatwestudied.
Sloan(Fukugitaetal.1996).However,weemphasizethatthese What is the origin of this loop phenomenon, and why is
arenottheonlyfiltercombinationsforwhichloop-likefeatures it seen more prominently in some filter systems than in oth-
Articlenumber,page10of13