Empirical calibration of the near-IR CaII triplet–I. The stellar library and index definition

24 Pages·2008·1.7 MB·English

Mon.Not.R.Astron.Soc.000,1–24() Printed1February2008 (MNLaTEXstylefilev1.4)
Empirical calibration of the near-IR Ca triplet – I. The
II
stellar library and index definition
⋆
A.J. Cenarro,1 N. Cardiel,1 J. Gorgas,1 R.F. Peletier,2 A. Vazdekis,3
and F. Prada.4
1Dept. de Astrof´ısica, Fac. de CienciasF´ısicas, Universidad Complutense de Madrid, 28040 Madrid, Spain.
2School of Physicsand Astronomy, University of Nottingham, University Park, Nottingham NG7 2RD, UK
1 3Dept. of Physics, Universityof Durham, South Road, Durham DH1 3LE, UK
0 4Observatoriode Calar Alto, Almer´ıa, Spain
0
2

p
e
S
ABSTRACT
0
1
A new stellar library at the near-IR spectral region developed for the empirical

1 calibration of the Caii triplet and stellar population synthesis modeling is presented.
v The library covers the range λλ 8348-9020 ˚A at 1.5 ˚A (FWHM) spectral resolution,
7 and consists of 706 stars spanning a wide range in atmospheric parameters. We have
5 defined a new set of near-IR indices, CaT∗, CaT and PaT, which mostly overcome
1 the limitations of previous definitions, the former being specially suited for the mea-
9 surement of the Caii triplet strength corrected for the contamination from Paschen
0
lines. We also present a comparative study of the new and the previous calcium in-
1
dices, as well as the corresponding transformations between the different systems. A
0
thoroughanalysisofthesourcesofindexerrorsandtheproceduretocalculatethemis
/
h given.Finally,indexanderrormeasurementsforthewholestellarlibraryareprovided
p together with the final spectra.
-
o Key words: stars: abundances – stars: fundamental parameters – globular clusters:
r
t general – galaxies: stellar content.
s
a
:
v
i
X 1 INTRODUCTION star formation histories of, at least, an important fraction
r of early-type galaxies is much more complex and hetero-
a Withthispaperwestartaseries dedicatedtotheempirical
geneous. The apparent age spread among elliptical galax-
calibration of the near-infrared Caii triplet. The ultimate
ies (Gonz´alez 1993; Faber et al. 1995; Jørgensen 1999), the
aimofthisworkforusistousethestrengthoftheCalines
distributionofelementabundances(Worthey1998; Peletier
in this spectral range to investigate the stellar content of
1999;Trageretal.2000a)andtheinterpretationofthescal-
early-type galaxies. However, we expect that researchers in
ingrelations(likethecolour–magnitudeorMg –σrelations;
different areas (e.g. starburst and active galaxies, globular 2
Bower,Lucey&Ellis1992; Bender,Burstein&Faber1993;
clusters, and stellar astrophysics) can make use of some of
Pedraz et al. 1999; Terlevich et al. 1999; Kuntschner 2000;
theresults and information that will be presented through-
Trager et al. 2000b), are some of the main issues in the
out these papers. These include a new stellar library, a set
present debate about the evolutionary status of early-type
of homogeneous atmospheric parameters, empirical fitting
galaxies. A major common obstacle to tackle with the pre-
functions describing the behaviour of the Ca triplet with
vious problems is how to disentangle age and metallicity
thestellarparameters,andstellarpopulationmodelpredic-
effectsintheintegratedspectrumofacompositestellarpop-
tions.
ulation.
Traditionally, elliptical galaxies have been thought to
be a uniform class of objects, with global properties chang- Themeasurementandinterpretationofline-strengthin-
ing smoothly with mass and hosting old and coeval stellar dices in the spectra of early-type galaxies has been a fun-
population. However, over the last decade, a growing body damental tool to address the above topics, the Lick/IDS
of evidence is indicating that the formation processes and (ImageDissectorScanner)system(Wortheyetal.1994 and
referencestherein)beingthemostwidely used(butseealso
Rose 1994). The Lick system makes use of strong spec-
⋆ E-mail:cen@astrax.fis.ucm.es tral features in the blue spectral range (from λ 4100 ˚A to
(cid:13)c RAS
2 A.J. Cenarro et al.
λ 6300 ˚A), although only the main absorption lines in the The outline of the series is as follows. In this first paper we
narrowerλλ4800–5400 ˚Arangehavebeenextensivelymea- introducethenewstellar library anddefineanewset of in-
suredwithhighprecisioninthespectraofearly-typegalax- dices to quantify the strength of the Ca triplet. Paper II is
ies. More recently, new spectral indices have been defined dedicatedtothedeterminationoftheinputatmosphericpa-
and calibrated in the bluer region (Jones & Worthey 1995; rametersforthestellar sample. Thesearesomeofthebasic
Worthey&Ottaviani1997; Vazdekis&Arimoto1999;Gor- ingredients of the empirical fitting functions and the spec-
gas et al. 1999), with enlarged capability to decouple age tral synthesis analysis that will be presented in Paper III
and metallicity effects. However, the spectral range of in- and IV, respectively. In this latter paper, we will compute
terest is still quite narrow. It is obvious that enlarging the thepredictions for single-burst stellar population models in
wavelengthcoveragewill allow ustoinvestigatetherelative two parallel ways: providing the Ca triplet strength using
contributionsofdifferentstellartypestothecompositespec- thefittingfunctionsofPaperIII,andsupplyingspectralen-
trum.Asan example,acomparison between themean ages ergy distributions in the λλ 8348–9020 ˚A range (following
derived using, as an age discriminant, features in different Vazdekis1999).
spectral regions, can provide important constraints about In Section 2 of this paper we review previous work on
thestarformationhistoryofthegalaxies(S´anchez–Bl´azquez the Caii triplet behaviour and its applications to different
et al. 2000). Although an important effort has been made astrophysical problems. Section 3 describes the new stellar
to extend the stellar population studies to the ultraviolet library aswellas anoverviewof theobservation andreduc-
region (e.g. Ponder et al. 1998), this is not the case for the tion procedures. A thorough discussion of the Caii index
near-infrared. In fact, it is remarkable that, although the definition is given in Section 4, where we define new im-
features in the near-infrared spectral range where already proved Caii indices which overcome the problems affecting
includedintheearlyanalysesoftheextragalacticoldstellar previous definitions. This section also includes a compara-
populations(e.g.Spinrad&Taylor1971),30yearslater,and tivestudyofthesensitivityoftheCaiiindicestoeffectslike
despite the advent of modern CCD detectors, the potential thesignal-to-noiseratioorthespectralresolution,andaset
ofthisspectralrangetoinvestigatethestellarpopulationof ofcalibrationsforconvertingbetweendifferentindexdefini-
early-typegalaxies is still almost unexploited. tions. In Section 5, an extensive analysis of index errors is
The Caii triplet is one of the most prominent features presented, discussing the different sources of random errors
in the near-IR spectrum of cool stars and its potential to andsystematiceffects.Asasupplementofthissection,Ap-
study the properties of stellar populations has been exten- pendix A provides accurate formulae for the computation
sively acknowledged in the literature (see Section 2.2 for a of random errors in the index measurements, together with
reviewofpreviousworksonthesubject).However,areliable an analytical estimate of errors from signal-to-noise ratios.
analysis of the Ca triplet measurements in integrated spec- Finally, in Section 6 we refer the reader to a web page in
trarestsonthecomparisonofthedatawiththepredictions which we providea database containing the spectra for the
of stellar population models. The accuracy of such predic- wholelibrary,an electronic tablelisting fullinformation for
tionsishighlydependentontheinputcalibrationofthecal- eachstar,andpublicfortranroutinestocomputethenew
ciumline-strengthsintermsofthemainatmosphericstellar indices and their associated errors.
parameters (namely effective temperature, surface gravity
and metallicity). The quality of this calibration has been
the major drawback of previous stellar population models
which haveincluded predictions for the strength of the cal- 2 PREVIOUS WORKS ON THE NEAR-IR
cium triplet (Vazdekis et al. 1996; Idiart, Th´evenin & de CAII TRIPLET
Freitas Pacheco 1997, hereafter ITD; Mayya 1997; Garc´ıa–
Inapioneerwork,Merrill(1934) (seealsoWilson&Merrill
Vargas,Moll´a&Bressan1998;Schiavon,Barbuy&Bruzual
1937) explored for the first time the near-infrared region of
2000; Moll´a & Garc´ıa–Vargas 2000). Previous calibrations
stellarspectra,identifyingthemostrelevantabsorptionfea-
havebeeneithertheoretical(based onmodelatmospheres), turesand noting that the lines of Caiiλλ 8498, 8542, 8662
with their associated uncertainties, or based on empirical
are prominent in types A–M and may have an interesting
stellarlibrarieswithapoorcoverageoftheatmosphericpa-
relationship to absolute magnitude. This tentative predic-
rameter space. Apart from this major problem, there are
tion was confirmed by the subsequent works of Keenan &
other factors which have compromised, with different im-
Hynek(1945)andParsons(1964).Inparticular,thislastau-
pacts, thereliability of thesepreviouspapers. They include
thorshowedhowthestrengthoftheCalines(blendedwith
stellar libraries with too few stars, problematic index def-
Paschen lines for early–type stars) increases with absolute
initions (e.g., they cannot be used for all spectral types),
magnitude(i.e.beinglargerforsupergiants)atfixedspectral
uncertainties in the input stellar parameters (which trans-
type. Later works by O’Connell (1973), Anderson (1974),
lateintounknownuncertaintiesin thederivedpredictions),
Cohen (1978), and others (see below) confirmed these find-
and illnesses in the fitting procedures (a proper statistical
ings, establishing the near-infrared Ca triplet as one of the
analysis of the computed coefficients is seldom followed).
most striking luminosity-sensitive feature in the spectra of
Throughoutthisworkwewillreviewandcommentindetail
F–M stars.
all theseproblems and how we havecoped with them.
The main motivation of the present series of papers
is, therefore, to provide a reliable calibration of the near-
2.1 Calibrations of the calcium line-strengths
infrared Caii triplet that makes it possible to accurately
calculatethebehaviourofthecalciumline-strengthsinstel- Aftertheseearlystudies,severalauthorshavetriedtomodel
lar populations with a wide range of ages and metallicities. the behaviour of the Ca triplet with the basic atmospheric
(cid:13)c RAS,MNRAS000,1–24
The near-IR Caii triplet: stellar library 3
parameters (T , logg and [Fe/H]). Two main lines of ap- sphere and, therefore, non–Local Thermodynamic Equilib-
eff
proach have been followed, either using empirical stellar li- rium (NLTE)models are required.For thisreason, thefirst
braries, or using thepredictions of stellar atmosphere mod- attemptstoanalyzethesensitivityoftheCastrengthtothe
els. atmospheric parameters were restricted to the line wings.
In this sense, Smith & Drake (1987, 1990) concluded that
thedependenceonmetallicityshouldbelargerthanthatre-
2.1.1 Empirical calibrations portedbytheempiricalstudies(likeJAJ),andthateffective
temperatureshadanonnegligibleeffect.Erdelyi–Mendes&
Among the works following the empirical method, we must
Barbuy(1991),alsousingLTEmodels,extendedtheanaly-
remark on that of Jones, Alloin & Jones (1984, hereafter
sis to cooler stars and included the significant contribution
JAJ), in which they found that the Ca triplet strength
ofmolecularbands.TheyagreedwithSmith&Drakeinthat
strongly correlates with gravity (in the sense that it in-
metallicity,ratherthangravity,wasthemainparameter,the
creasesforgiantandsupergiantstars).Theyalsonotedthat
[Fe/H] dependencebeing stronger for lower gravities.
theresidualsofthisrelation correlate weakly withmetallic-
Using NLTE models, Jørgensen, Carlsson & Johnson
ity.This stronggravity dependencewasindependentlycon-
(1992) carried out, for the first time, the computation of
firmed by Carter, Visvanathan & Pickles (1986) and Alloin
theline-cores, derivingfullequivalent-widths(theyshowed,
& Bica (1989), although theydid not agree in their conclu-
however, that the effects of departures from LTE are neg-
sions about the metallicity dependence, being stronger in
ligible since the equivalent widths are dominated by the
thelatter study.
linewings).Concerningthesensitivitytostellarparameters,
An important step to quantify the g and [Fe/H] de-
they found that the gravity dependenceof the Ca triplet is
pendence was given by the comprehensive study of D´ıaz,
influenced by T and [Fe/H] in a complicated way, They
eff
Terlevich & Terlevich (1989, hereafter DTT). Using a stel-
provided fitting functions for stars in the range 4000 K
lar library with an enlarged range in gravity and metal-
≤T ≤ 6600 K and −1.0≤[A/H]≤+0.2 dex.
eff
licity, these authors quantified the biparametric behaviour
Recently, Chmielewski (2000) has presented a compre-
with logg and [Fe/H], noting that, in the high-metallicity
hensiveanalysisofthetheoreticalmodellingoftheCalines.
range, the strength of the Ca triplet depends only on grav-
He stresses that previous works did not take into account
ity, whilst, for low-metallicity stars, [Fe/H] is the main pa-
thecontributionofhydrogenPaschenlinestotheintegrated
rameter. Zhou (1991, hereafter ZHO) carried out a similar
equivalent widths, showing that this contribution is signif-
analysis, using a higher spectral resolution (2 ˚A) and ex-
icant for dwarfs hotter than 5800 K and giants with T
eff
tendingthesample of stars byincluding cooler stars (upto
above ∼ 5500 K. Interestingly, this effect explains most of
M7). He found that the effect of temperature, found to be
the discrepancies between the previous empirical and theo-
negligible in the previous work, was important at low T .
eff reticalworksconcerningthetemperaturedependenceofthe
Healso noted that thegravity dependencewas stronger for
Ca lines. His main conclusions about the Ca line-strength
giants than for dwarfs.
sensitivitytothestellarparametersarethatthegravityde-
Another relevant study is that of Mallik (1994, 1997).
pendence only holds for logg < 3 dex, metallicity is a sig-
Throughtheanalysisofanamplestellarlibraryathighreso-
nificant parameter in all cases, and temperature effects can
lution(0.4˚A),thisauthorconfirmedthestrongdependence
only be neglected for dwarfs between 5000 and 6000 K.
on logg (and showed that it was stronger as metallicity in-
Unfortunately,thesetheoreticalstudies,beingveryim-
creases)andthemilderdependenceon[Fe/H](but,remark-
portanttohelptounderstandthebehaviouroftheCatriplet
ably,moreimportantforsupergiantstars,inagreementwith
and to check the empirical studies, are restricted to cool
DTT). Unfortunately, he does not provideany fitting func-
stars(typicallybetweenFandK)and,obviously,theircon-
tion,restrictingthepotentialapplicationofthesequalitative
clusions can not be extrapolated to later spectral types to
results. ITDhavederivedempirical functions which predict
construct stellar population synthesis models of, relatively
the strength of the Ca triplet in terms of the three atmo-
old, stellar systems.
spheric parameters. Using a library with a good coverage
in metallicity but lacking a representative sample of super-
giants, they conclude that the gravity dependenceis not as 2.2 Applications of the Ca triplet to the study of
strong as previously reported, and that metallicity is the stellar systems
main parameter.
Overtheyears,thenear–IRCatriplethasbeenusedexten-
It is clear that there are some apparent contradictions
sively to address a numberof topics in different areas.
among the conclusions of these papers. A critical analysis
of these will be presented in Paper III of this series, where
wewillshowthattheapparentdiscrepanciesaremainlydue 2.2.1 Stellar astrophysics
tothedifferentrangesofstellarparametersintheemployed
Inthefieldofstellarastrophysics,thetriplet,togetherwith
stellar libraries, the diversity of index definitions (see Sec-
otherspectralfeatures inthisspectral range,hasbeenused
tion 4.2), and differences in the fitting procedures (e.g. in
forthespectralclassification ofstars(Sharpless1956;Bouw
most cases, a proper statistical approach was not applied).
1981; Kirkpatrick,Henry & McCarthy 1991; Ginestet et al.
1994; Munari & Tomasella 1999). These features have also
been widely used as chromospheric activity indicators (e.g.
2.1.2 Theoretical calibrations
Linskyetal.1979;Dempseyetal.1993;Montesetal.1998).
The theoretical modelling of theCa triplet lines is not sim- Furthermore, its sensitivity to surface gravity has been ex-
ple since the line cores are formed in the stellar chromo- ploited to identify supergiants in the Galaxy (e.g. Garz´on
(cid:13)c RAS,MNRAS000,1–24
4 A.J. Cenarro et al.
et al. 1997) or in Local Group galaxies (Mantegazza 1992; Vargaset al. 1997; Gonz´alez Delgado et al. 1997; Gallagher
Humphreyset al. 1988; Massey 1998). & Smith 1999; Goudfrooij et al. 2001).
2.2.4 Early-type galaxies
2.2.2 Globular clusters
ThefirstpapersanalyzingtheCatripletinearly-typegalax-
Inspiteofthefirstempiricalpapersthatremarkedthatthe
iestriedtomakeuseofitsgravitysensitivitytoconstrainthe
Ca triplet was mainly gravity driven (see above), Arman-
dwarf/giant ratiointheintegratedlightofoldstellarpopu-
droff & Zinn (1988, hereafter A&Z), in a pioneering study,
lations. The main debate at that time was focussed on the
found that, in the integrated light of galactic globular clus-
possible dwarf enrichment in the nucleus of M 31, as com-
ters,theCaequivalentwidthwaswellcorrelatedwithmetal-
pared toits bulgeorto low luminosity ellipticals like M32.
licity,indicatingthatthisfeaturecouldbeafairmetallicity
WhileFaber&French(1980)and,tosomedegree,Carteret
indicator for old, and approximately coeval, stellar popula-
al.(1986)favoredadwarf-enrichedstellarpopulation,Cohen
tions, provided that themetallicity was lower than solar.
(1978) and Alloin & Bica (1989) found that metallicity ef-
Subsequentworksinthisfieldconcentratedinthestudy
fectsandcontaminationbyamolecularbandcouldaccount
of individual red giants of globular clusters. Armandroff &
fortheapparentenrichment,mainlybasedonmeasurements
Da Costa (1991, hereafter A&D) derived an empirical re-
of thestrength of the NaIfeature at λ 8190 ˚A.
lation between the cluster metallicity and the “reduced”
′ One interesting result from these first studies was that
equivalentwidthW =EW(CaT)+c(V −V ),wherecis
HB
itwasfoundthatthestrengthoftheCatripletdidnotvary
a constant term, EW(CaT) is thepseudo–equivalent width
much among early-type galaxies (Cohen 1979). Note that
(typically the sum of the two strongest lines of the triplet),
although Cohen (1978) found that the nucleus of M 31 ex-
andV istheV magnitudeofthehorizontalbranch.Note
HB
thatW′isamonotonicfunctionofmetallicitybecausegrav- hibitedstrongerCalinesthanM32,Faber&French(1980)
reportedtheoppositebehaviour.Later,Bica&Alloin(1987,
ityeffectsareremovedbythelasttermoftheequation.This
hereafter B&A)measured theCa strength in thespectraof
relation has been recalibrated by several authors (e.g. Da
62galacticnuclei(fromEtoSc),concludingthattheequiva-
Costa & Armandroff 1995; Geisler et al. 1995; Rutledge et
lentwidthswerenotrelatedtogalaxytypesorluminosities.
al. 1997ab, hereafter RHS) and extensively used to derive
ThiswasconfirmedbythestudiesofTerlevichetal(1990a),
metallicities of galactic globular clusters (Armandroff, Da
in which they found a small spread in the Ca strengths of
Costa & Zinn 1992; Da Costa, Armandroff & Norris 1992;
14normalgalaxynuclei,andHoudashelt(1995),whonoted
Buonanno et al. 1995; Suntzeff & Kraft 1996; Rosenberg
that the Ca equivalent widths of a sample of 34 early-type
et al. 1998), clusters and individual stars in the Magellanic
galaxies did not vary significantly among galaxies of differ-
Clouds(Olszewskietal.1991;Suntzeffetal.1992;DaCosta
ent color or absolute magnitude.
& Hatzidimitriou 1998; Cole, Smecker–Hane & Gallagher
The apparent uniformity of the Ca measurement in el-
2000), and Local Group dwarf spheroidals (Suntzeff et al.
lipticalgalaxiesissomewhat surprisingsince:i)Itcontrasts
1993; Smecker–Hane et al. 1999). A comprehensive analy-
withthemetallicitydependencefoundforthestellarsamples
sis of the different methods to measure cluster metallicities
andthereportedbehaviouringalacticglobularclusters(see
using thetheCa triplet can befound in RHS.
above),ii)Previousstellarpopulationmodelspredictahigh
sensitivity of the Ca triplet to the metallicity of metal-rich
2.2.3 Active galaxies and extragalactic Hii regions oldstellarpopulations(Garc´ıa–Vargas etal.1998;Schiavon
et al. 2000), and iii) It has been argued that, in contrast to
One of the fields in which the study of the Ca triplet has other α-elements, Ca abundance is not enhanced compared
had a wide application is that of active galaxies and star- toFeinbright elliptical galaxies (O’Connell1976; Vazdekis
burst regions. Terlevich, D´ıaz & Terlevich (1990a) studied et al. 1997; Worthey 1998; Moll´a & Garc´ıa–Vargas 2000),
the Ca triplet in a sample of normal and active galaxies, andvariations oftheFeline-strengthsamong ellipticals, al-
concluding that LINERS and Seyfert 2 objects exhibit Ca thoughdifficulttodetect,arenotnegligible(e.g.Gorgas,Ef-
strengths equal or larger than those found in normal ellip- stathiou&Arag´on–Salamanca1990;Gonz´alez1993;Davies,
ticals (in contrast with the blueabsorption lines, which are Sadler & Peletier 1993; Kuntschner 2000). Therefore, more
usually diluted in activegalaxies). Following theconclusion reliable predictions of the Ca triplet behaviour in the in-
ofDTTthatCaequivalentwidthsabove9˚Aareonlyfound tegrated stellar population of high metallicity systems and
inredsupergiantstars,theyinterpretthisresultasastrong newhigh-qualitymeasurementsarehardlyneededtoclarify
evidencefortheoccurrenceofnuclearstarburstsinthecen- thispoint.
tralregionsofactivegalaxies.Thissameapproachhasbeen This problem is closely related to the radial behaviour
followed by a number of authors (Forbes, Boisson & Ward of the Ca strengths within galaxies. Given the measured
1992; Garc´ıa–Vargas etal.1993;Gonz´alezDelgado&P´erez metallicitygradientsusuallyfoundinellipticalgalaxies (see
1996ab;Heckmanetal.1997;P´erezetal.2000)tostudythe Gonz´alez & Gorgas 1996, and references therein), even us-
evolutionary status of the nuclei and circumnuclear regions ingFeline-strengths,oneshouldexpecttofindunambiguous
of active galaxies through thepresence of red supergiants. negativegradientsinCa.However,Cohen(1979) andBoro-
The Ca triplet has also been detected in extragalactic son & Thompson (1991) found negligible gradients; Carter
Hiiregions(Terlevichetal.1990b,1996),beinginterpreted et al (1986) and Peletier et al. (1999) measured either null
as due to the recent star formation burst, and it has been or positive gradients, and Delisle & Hardy (1992, hereafter
used to study the star formation history of extragalactic D&H) found negative gradients in many galaxies of their
super-starclusters(Prada,Greve&McKeith1994; Garc´ıa– sample.
(cid:13)c RAS,MNRAS000,1–24
The near-IR Caii triplet: stellar library 5
Anotherimportantdifficultyariseswhentryingtocom- of DTT and 43 of the55 stars of thesample of ITD. More-
paretheabsolutevaluesoftheCatripletmeasuredinearly- over,withtheaimoffillinggapsintheparameterspace,we
typegalaxieswiththeactualvaluespredictedbystellarpop- also included stars from several other compilations, mainly
ulation synthesis models (e.g. Peletier et al. 1999; Moll´a & focusingonO,BandAtypes(37hotstarsfromthesample
Garc´ıa–Vargas 2000). The uncertainties of the theoretical of Andrillat, Jaschek & Jaschek 1995), late M types, metal
or empirical Ca behaviourimplemented in such models, to- poor,metalrichandchemicallypeculiarstars(40starsfrom
getherwithotherproblemssuchasobservationaluncertain- a list kindly provided by G. Worthey, private communica-
ties,corrections from velocitydispersion ortransformations tion, and 102 stars from Jones 1997). The presence of hot
between different index definitions, make very difficult to stars in the stellar library allows us to expand the predic-
extract any significant conclusion from such a comparison. tions of our models to young stellar populations. Also, and
as it will be shown in Paper IV, very cool stars are neces-
sary to reproduce the integrated spectra of old stellar sys-
tems at the near-infrared spectral region over a wide range
3 THE NEAR-IR STELLAR LIBRARY ofages.Finally,sincecalciumisanα-element,29additional
stars with high or low Ca/Fe ratios from the catalogue by
3.1 Previous libraries in the near-IR spectral Th´evenin(1998) wereobservedinordertoanalyzetheCaii
range
index dependenceon relative abundances (see Paper III).
PreviousstellarlibrariesprovidingspectraintheCaiitriplet The new stellar library covers the following ranges in
atmospheric parameters: T from 2750 to 38400 K, logg
spectralrangearelisted inTable1.Hereweonlyincludeli- eff
brarieswithaspectralresolutionbetterthan10˚A(FWHM). from 0to 5.12 dex,and [Fe/H] from −3.45 to+0.60 dex.It
is important to note that the stars of the stellar library do
For alist of lower resolution libraries we refer thereader to
nothomogeneouslycovertheparameterspace.Infact,most
Munari&Tomasella(1999).Wehavealsoexcludedfromthis
of thestars (∼ 83 per cent) have metallicities from −1.0 to
list some libraries which are too specific for thepurposes of
+0.5 dex, and the widest gravity range is also within this
this work, like those of Fluks et al. (1994) (only M stars),
metallicityinterval.Thefinalatmosphericparameterswhich
Allen & Strom (1995) (stars from two open clusters), RHS
have been adopted for each star and the method to derive
(huge collection of red giants from galactic globular clus-
them are discussed in detail in Paper II of this series. We
ters),orMontes&Mart´ın (1998)(highresolutionlibraryin
refer the reader to that paper for an HR diagram of the
a limited spectral range). It is clear from this table that no
whole sample.
previous library provides simultaneously a broad coverage
of stellar metallicities and effective temperatures. In fact,
the only libraries including a good fraction of low metallic-
3.3 Observations and data reduction
ity stars are those of DTT and ITD, but the spectral type
ranges spanned in these cases are rather limited, especially Thespectraofthestellarlibrarywereobtainedduringato-
in the latter case (as we will see in Paper IV, the relative tal 21 nights in six observing runs from 1996 to 1997 using
contributionofstarscolderthanK3tothenear-IRspectrum the JKT (Jacobus Kapteyn Telescope), INT (Isaac New-
ofanoldstellarpopulationisveryimportant).Ontheother ton Telescope) and WHT (William Herschel Telescope) at
hand,librariescomposedofstarsofallspectraltypesdonot theRoquedelosMuchachosObservatory(LaPalma,Spain)
attainabroadrangeinmetallicity(e.g.Munari&Tomasella andthe3.5-mtelescopeatCalarAltoObservatory(Almer´ıa,
1999).Thelowfrequencyofsupergiantstarsisanadditional Spain).It isimportant tonotethatmost ofthestars(∼93
problemofsomelibraries(JAJ;ITD).Itmustbenotedthat percent),includingallthefieldstarsandthebrightestones
someof theapparent discrepanciesamong differentauthors from open and globular clusters, were observed during the
who have modelled the behaviour of the CaT index with threerunsattheJKTusingthesameinstrumentalconfigu-
theatmosphericparametersaremainlyduetothisdifferent ration. This ensures a high homogeneity among thedata of
coverageoftheparameterspacebythecalibratingstars(see thesethreeruns(seetheanalysisinSection5.3).Theinstru-
Paper III). mentalconfigurationattheothertelescopeswasalsochosen
to be as similar as possible to that at the JKT, obtaining
spectral resolutions (FWHM) for the six runs in the range
from 1.07 ˚Ato2.13 ˚A.Afulldescription oftheseandother
3.2 The sample
instrumentaldetailsisgiveninTable2.Wemustnotethat,
We have observed a new stellar library of 706 stars at the as it will be explained in Section 5.2, spectra from runs 1,
near-infrared spectral region (λλ 8348-9020 ˚A). It includes 3 and 4 were broadened to the spectral resolution of run 2
421 of the 424 stars with known atmospheric parameters (1.5 ˚A). Hence, and with the exception of thefew stars ob-
of the Lick/IDS Library (Burstein et al. 1984; Faber et al. served in runs 5 and 6, the whole library has a common
1985; Burstein, Faber& Gonz´alez 1986; Gorgas et al. 1993; resolution. N stands for the number of stars observed in
obs
Worthey et al. 1994). This subsample spans a wide range each run.
in spectral types and luminosity classes. Most of them are Typical exposure times varied from a few seconds for
field stars from thesolar neighbourhood, butstars covering bright stars to 1800 s for the faintest cluster stars. These
a wide range in age (from open clusters) and with differ- ensured typicalvalues of SN(˚A) ∼ 100 ˚A−1 (signal-to-noise
entmetallicities(fromgalacticglobularclusters)arealsoin- ratio per angstrom) for field and open cluster stars, and
cluded.Inordertoobtainalargesampleofstarsincommon SN(˚A) ≥ 15 for the globular cluster stars. In order to test
with other previous works devoted to the Caii triplet, our the derived random errors of the final index measurements
stellar library was enlarged to include 105 of the 106 stars (see Section 5.1), we performed multiple observations of a
(cid:13)c RAS,MNRAS000,1–24
6 A.J. Cenarro et al.
Table 1.Medium-resolutionstellarlibrariesintheCaiitripletspectralregion.
Reference No.stars Resolution Spectral [Fe/H] Comments
(FWHM,˚A) types range
Jones etal.(1984) 62 3 B–M5 −0.60,+0.23
D´ıazetal.(1989) 106 3.5 F5–M1 −2.70,+0.55
Zhou(1991) 144 2 F5–M7 −2.28,+0.60 2starswith[Fe/H]<−0.61
Andrillatetal.(1995) 76 1.2 O5–G0 ?
SeroteRoosetal.(1996) 21 1.25 B3–M5 −0.15,+0.39 Onlygiantsandsupergiants
Carquillatetal.(1997) 36 2 F5–M4 ?
Idiartetal.(1997) 55 ∼2 A1–K3 −3.15,+0.35
Mallik(1997) 146 0.4 F7–M4 −3.0,+1.01 1starswith[Fe/H]<−1.6
Munari&Tomasella(1999) 131 0.43 O4–M8 −0.54,+0.30
Montesetal.(1999) 130 ∼0.7 F0–M8 −2.74,+0.31 3starswith[Fe/H]<−0.5
Thiswork 706 1.5 O6–M8 −3.45,+0.60
Table 2.Observational configurations
Run Dates Telescope Spectrograph Detector Dispersion ∆λ Slitwidth FWHM Nobs
(˚A/pix) (˚A) (arcsec) (˚A)
1 20–26Sep1996 JKT1.0m RBS TEK#4 0.85 8331–9200 1.5 1.18 264
2 15–19Jan1997 JKT1.0m RBS TEK#4 0.85 8331–9200 1.5 1.50 248
3 25–29Jun1997 JKT1.0m RBS TEK#4 0.85 8331–9200 1.5 1.07 362
4 13Aug1997 INT2.5m IDS TEK#3 0.85 8185–9055 1.5 1.28 33
5 4Aug1996 WHT4.2m ISIS TEK#2 0.79 8222–9031 1.2 2.13 12
6 17–18Nov1996 CAHA3.5m TWIN SITe#4d 0.81 8300–9912 2.1 2.11 15
subsampleofstars (15–20) within eachparticular runand altitude)duringtheday.Later,usingthealt-azimuthcoordi-
in common with other different runs. Finally, to perform a natesofeachstar,theclosestflat-fieldframeoverthewhole
reliable flux calibration, several (3 or 4) spectrophotomet- setwasfinallyusedasitsownflat-fieldcalibrationframe.It
ric standards were observed along each night at different isimportanttohighlighttheimportanceofanaccurateflat-
zenithal distances. field correction when the fringe pattern becomes relevant,
The reduction of the data was performed with as it is usually the case in this spectral range. Depending
REDumcE† (Cardiel1999),whichallowsaparalleltreatment on the properties of the CCD, the fringe effect introduced
of data and error frames (see more details in Section 5) high frequency structures with an amplitude of up to 7 per
and, therefore, produces an associated error spectrum for cent of the true flux, but they virtually disappeared by us-
each individual data spectrum. We carried out a standard ing ourprocedure. Flux differences between thenormalized
reduction procedure for spectroscopic data: bias and dark fringepatternsofdifferentflat-fieldframesallowsanestima-
subtraction, cosmic ray cleaning, flat-fielding, C-distortion tion of the uncertainties introduced in the fringe correction
correction, wavelength calibration, S-distortion correction, procedure. These are always below 0.6 per cent.
sky subtraction, atmospheric extinction correction, spec- Concerning the wavelength calibration, we only ob-
trum extraction, and relative flux calibration. We did not tained comparison arc frames for a previously selected sub-
attempt to obtain absolute fluxes since both the evolution- sample of stars covering all thespectral typesand luminos-
arysynthesiscodesandtheline-strengthindicesonlyrequire ityclassesineachrun.Theselectedspectrawerewavelength
relative fluxes. Cluster stars were also corrected from inter- calibrated withtheirown arcexposurestakingintoaccount
stellarreddeningusingthecolorexcessesfromGorgasetal. their radial velocities, whereas the calibration of any other
(1993) and Worthey et al. (1994) and the averaged extinc- star was performed by a comparison with the most similar,
tion curveof Savage & Mathis (1979). alreadycalibrated,referencespectrum.Thisworkingproce-
InordertooptimizetheobservingtimeduringtheJKT dure is based on the expected constancy of the functional
runs,wedecidednottoacquireflat-fieldandcomparisonarc formofthewavelengthcalibrationpolynomialwithinacon-
frames for each individual exposure of a library star. Con- sideredobservingrun(whichhasbeenwidelytestedbycom-
cerning the flat-field correction, after checking that small paring the derived polynomials for each run). In this sense,
variations of the CCD temperature do not affect the flat- the algorithm that we used is as follows: after applying a
field structure and considering that it exclusively depends test x-shift (in pixels) to any previous wavelength calibra-
ontheposition ofthetelescope,weobtainedacompleteset tion polynomial, we obtained a new polynomial which was
of flat-field exposures by pointing at a grid of positions on used to calibrate the spectrum. Next, the calibrated spec-
the dome (with a resolution of 30◦ in azimuth and 15◦ in trumwascorrectedfromitsownradialvelocityand,finally,
thespectrumwascross-correlated withareferencespectrun
ofsimilarspectraltypeandluminosityclass,inordertode-
† http://www.ucm.es/info/Astrof/reduceme/reduceme.html rivethewavelengthoffset betweenbothspectra.Byrepeat-
(cid:13)c RAS,MNRAS000,1–24
The near-IR Caii triplet: stellar library 7
ingthisprocedure,itispossibletoobtainthedependenceof spectral types (from F5 to M2 approximately) and for all
thewavelengthoffsetasafunctionofthetestx-shiftsand,as luminosity classes (see Fig. 1c).
aconsequence,wederivetherequiredx-shiftcorresponding ThehydrogenPaschenseries(λλ8359.0,8374.4,8392.4,
to a nullwavelength offset. Uncertainties in thewavelength 8413.3, 8438.0, 8467.3, 8502.5, 8545.4, 8598.4, 8665.0,
calibration are estimated in Section 5. 8750.5, 8862.8, 9014.9 ˚A, from P22 to P10 respectively) is
It is also important to account for the presence of apparent in stars hotter than G3 types. Depending on the
telluric absorptions, mainly at the red end of the spec- luminosity class, the strength of this series reaches a max-
tral range (the strongest lines are locate at λλ 8952, 8972, imum for F or A type stars (Andrillat et al. 1995). In the
8980, 8992 ˚A). Fortunately, since these strong H O lines earliest spectraltypes,wheretheCaiistrengthbecomesin-
2
do not affect the Caii triplet region and the observations significant, the relative depths of the Paschen lines show a
wereperformedunderdryconditions,wedidnotcorrectfor smooth sequence with wavelength (see Fig. 1a). However,
theircontamination and it is possible that our finalspectra due to the fact that, for low and intermediate spectral res-
present unremoved features at the red end. An illustrative olution, the Paschen lines P13, P15 and P16 overlap with
example of the effect of telluric lines in the Caii triplet re- the Caii triplet, these three Paschen lines stand out in the
gion is given in Stevenson (1994) and Chmielewski (2000). smoothsequenceforAandFtypes(Fig.1b).Thefactthat
theCaiilinesareblendedwiththePaschenlineshasalways
been an obstacle to measure the calcium triplet in warm
stars (see B&A and Chmielewsky 2000).
4 CA II TRIPLET INDICES DEFINITION StarscoolerthanearlyMtypesexhibitmolecularbands
thatchangetheslopeofthelocal continuum(Fig.1d).The
Line-strengthindiceshavebeenwidelydefinedtoobtainob-
strongest onesarethetriple-headedbandat λλ8432, 8442,
jective measurements of any relevant spectral feature. For
8452˚Aandthedouble-headedbandsatλλ8859.6,8868.5˚A
the Ca triplet, there exist previous index definitions from
and λλ 8937.4, 8949.8 ˚A of titanium oxide (TiO). Other
otherauthorswhichwereoptimizedtomeasurethelinesfor
molecularfeaturesarethebandsequenceofTiOatλλ8373,
a narrow range of spectral types (mainly G and K-types).
8386,8420,8457,8472,8506,8513,8558,8569˚Aandseveral
These previousindices, however,are not appropriate for all
vanadium oxide (VO) bands at λλ 8521, 8538, 8574, 8597,
the spectral types. Hot and cold stars show strong spec-
8605, 8624, 8649, 8668 ˚A (see Kirkpatrick et al. 1991 and
tral features which were not taken into account for those
references therein). Thestrength of thesefeatures increases
indexdefinitionsand,asaconsequence,mostofthecalcium
as the temperature decreases, being more prominent for gi-
measurements for these spectral types become unrealistic.
ants than for dwarfs. The spectra of late M type stars are
Moreover,someofthepreviousdefinitionshavenotbeenop-
dominated by strong molecular bands showing very weak
timizedforcompositestellarsystems.Someofthemrequire
Caii lines.
spectrawithhighsignal-to-noiseratio,becomingunpractical
forfaintobjects.Besides,someindicesshowastrongdepen-
denceonspectralresolution,orvelocitydispersionbroaden- 4.2 Previous index definitions
ing. In this sense, one should not forget that, in the study
Severalpreviousworks haveestablished different indexdef-
of stellar populations, to measure true equivalent widths is
not as important as to obtain robust index measurements, initions to measure the strength of the Caii triplet. In par-
ticular, the most commonly used indices are those defined
specially inspectrawithlowsignal-to-noiseratios orwitha
by JAJ, B&A, A&Z, DTT, A&D, ZHO, D&H, and RHS.
wide range of spectral resolutions.
Mostoftheseindicesweredefinedaccordingtotheclassical
In order to cope with the above problems, we have de-
cided to introduce some new Caii triplet index definitions. definition of line-strength indices, that is, by establishing a
central bandpass covering the spectral feature of interest,
Section 4.1 presents an overview of the strongest spectral
features in the range from 8350 to 9020 ˚A. In Section 4.2 and two other bandpasses (at the red and blue sides of the
we analyze the behaviour and limitations of previous Caii central region) which are used to trace a local continuum
reference level through a linear fit to the mean values in
tripletindexdefinitionsfordifferentspectraltypes.Thenew
both bands (for more details about thedefinition and com-
indicesare definedin Section 4.3, whereas Section 4.4 is re-
putation of classical indices see Appendix A1). In Table 3
served to study the dependence of the new and old index
we list the bandpasses limits of theprevious definitions.
definitionsentheS/Nratio, thecombinedeffectofspectral
Itmustbenotedthatsomeoftheseindicesslightlydif-
resolution and velocity dispersion broadening, and the flux
fer from the classical definition. On the one hand, the con-
calibration. Finally, in Section 4.5 we present comparisons
and calibrations between different Caii indexsystems. tinuum by ZHO is computed from the mean value of the 5
highest pixels in each continuum bandpass, whereas in the
index by JAJ, it is chosen relative to the maximum flux
4.1 The λλ 8350-9020 ˚A spectral region around two selected wavelengths. It is important to stress
thatthiskindofindexdefinitionsisuselessforlowsignal-to-
Prior to any index definition, we have explored how the noise spectra (S/N), leading to a spurious negative correla-
main absorption features in our wavelength range change tionbetweenindexvalueandS/Nratio.Also,sincetheclas-
along the spectral type sequence (Fig. 1). Apart from sev- sical index definition refers tothemeasurement of aunique
eral atomic lines of intermediate strength, as those of Fe i spectralfeature,previousdefinitionsaregivenforeachCaii
(λλ 8514.1, 8674.8, 8688.6, 8824.2 ˚A), Mg i (8806.8 ˚A) line. In this sense, the most general Caii triplet index is
and Ti i (8435.0 ˚A), the Caii triplet (λλ 8498.02, 8542.09, usuallycomputedasthesumofthethreesingle-lineindices,
8662.14 ˚A) is the strongest feature over a wide range of althoughsomeauthors(DTT,A&DandZHO)prefertouse
(cid:13)c RAS,MNRAS000,1–24
8 A.J. Cenarro et al.
Figure 1. Spectra of the stars HD 186568 (B8 III), HD 89025 (F0 III), HD 216228 (K0 III) and HD 114961 (M7 III) in the spectral
rangeofthestellarlibrary.Thestrongestfeaturesinthisregionaremarked:thePaschenSeries(fromP11toP20),theCaiitriplet(Ca1,
Ca2andCa3),severalmetalliclines(Fei,MgiandTii),molecularbands(TiOandVO)andtelluricabsorptions.
the sum of the two strongest lines of the triplet (λλ 8498.0 bandpassesbyDTTandZHO(whichareroughlythesame)
and 8542.1 ˚A) or even a weighted combination of the three arelocated tooclose tothecontinuumbreak,causing arise
lines(RHS)inordertooptimizethesensitivityoftheindex in the continuum level at the position of the Ca lines (see
to thesignal-to-noise ratio. Figs.2iand2l).Also,whentherelativedistancebetweenthe
Fig. 2 illustrates both the bandpass position and the twocontinuumbandpassesistoolarge,theindexcanbein-
predictedcontinuumofsomedefinitionsoverthreedifferent sensitivetolocalvariationsinthetruecontinuum(JAJand
spectraltypes.TheindicesbyB&AandRHShavenotbeen DTT),whereasindiceswithcontinuumsidebandswhichare
included since theirbluebandpasses are out of ourspectral too close (A&Z) are highly sensitive to velocity dispersion
range.AlsotheindexbyA&Dhasbeenexcludedduetoits broadening (see Section 4.4.2). As far as the width of the
similarity to theindex by A&Z. sidebandsisconcerned,bandpassesthataretoowide(D&H)
are not appropriate when the spectral region is full of ab-
In order to predict a reliable local continuum,the con-
sorption lines which decrease the continuum level. Indices
tinuumbandpassesshouldbelocatedinspectralregionsfree
with two narrow sidebands (JAJ) are very sensitive to the
from strong absorption features. In this sense, all the pre-
S/N ratio and to residuals from sky subtraction. Another
vious definitions are optimized for G and K spectral types
importantfactor isthewidthofthecentralbands,since, as
(see Figs. 2b, 2e, 2h, 2k and 2n). However, because therel-
wewillseeinSection4.4,itmainlydeterminesthesensitiv-
ative contribution of the earliest spectral types to the inte-
ityoftheindextotheS/Nratioandthevelocitydispersion
grated spectrum of stellar systems was expected to be neg-
broadening. Finally, and not the least important, it is ob-
ligible at this spectral range, most of the previous calcium
vious that any Ca index definition following the classical
indices were defined without taking into account the wave-
systemisunavoidablyaffectedbyPaschencontaminationin
length position of the Paschen series. The indices by A&Z,
the central bandpass for stars hotter than G1 and G3, for
DTT, and D&H clearly suffer from such a limitation deriv-
dwarfs and giants respectively.
ing a continuum level below the true one for hot stars (see
Figs. 2d, 2g, and 2m). On the other hand, because of the
presence of molecular bands in late type stars, some defi-
4.3 New indices definitions
nitions show unreliable continuum levels for these spectral
types. In particular, the red bandpass of the index by JAJ Althoughthepreviousindexdefinitionshavebeenveryuse-
falls in a strong TiO absorption (Fig. 2c), while the blue ful to increase the understanding of the behaviour of the
(cid:13)c RAS,MNRAS000,1–24
The near-IR Caii triplet: stellar library 9
Figure 2. Previous calcium index definitions over different spectral types. Indices (from top to bottom) correspond to the systems of
JAJ, A&Z, DTT, ZHO and D&H. The representative spectra (from left to right) are those of HD161817 (A2 VI), HD25329 (K1 Vsb)
andHD148783 (M6III). Greyandopen bands mark,respectively, continuum andcentral bandpasses, whereas thesolidline represents
thelocalpseudo-continuum computedbyanerrorweightedleast-squaresfittoallthepixelsinthecontinuum bands.
Caiitriplet bothin individualstarsand stellar populations providemeasurementsascloseaspossibletothetrueequiv-
(seeSection2),inthisworkwehavedecidedtodefineanew alentwidths.Otherindices,likethoseofRHSaremuchmore
setofimprovedindices,speciallydesignedtobemeasuredin appropriate for that purpose.
theintegrated spectra of galaxies. With thenewdefinitions
we havetried to alleviate difficulties such as thecontinuum The new indices have been defined according to a new
definition, the Paschen contamination, and the sensitivity type of line-strength index concept: the generic index. It is
to the S/N ratio and the velocity dispersion (spectral res- a natural generalization of theclassical definition which in-
olution). Note that the aim of these new indices is not to cludes the following requirements: it is characterized by an
arbitrary number of continuum and spectral-feature band-
(cid:13)c RAS,MNRAS000,1–24
10 A.J. Cenarro et al.
Table3.Bandpasseslimitsofpreviouscalciumindices.Codesare Table4.BandpasseslimitsforthegenericindicesCaTandPaT.
the following: JAJ (Jones, Alloin& Jones, 1984), B&A (Bica &
Alloin,1987),A&Z(Armandroff&Zinn,1988),DTT(D´ıaz,Ter- CaTcentral PaTcentral Continuum
levich&Terlevich,1989),A&D(Armandroff&DaCosta,1991), bandpasses (˚A) bandpasses (˚A) bandpasses (˚A)
ZHO(Zhou1991),D&H(Delisle&Hardy,1992),andRHS(Rut-
ledge et al., 1997a). Due to the subjective continuum definition Ca18484.0–8513.0 Pa18461.0–8474.0 8474.0–8484.0
byJAJ,bandpasses limitsforthisindexareasgiveninD&H. Ca28522.0–8562.0 Pa28577.0–8619.0 8563.0–8577.0
Ca38642.0–8682.0 Pa38730.0–8772.0 8619.0–8642.0
8700.0–8725.0
Index Central Continuum
bandpass (˚A) bandpasses (˚A) 8776.0–8792.0
Ca1(JAJ) 8483.0–8511.0 8633.0–8637.0, 8903.0–8907.0
Ca2(JAJ) 8517.0–8559.0 8633.0–8637.0, 8903.0–8907.0 (iii) The use of an error weighted least-squares fit in the
Ca3(JAJ) 8634.0–8683.0 8633.0–8637.0, 8903.0–8907.0 determinationofthelocalpseudo-continuumisspeciallyad-
vantageous when we measure the near-IR absorption fea-
Ca1(B&A) 8476.0–8520.0 8040.0–8160.0, 8786.0–8844.0
tures. In this spectral region, the presence of sky emission
Ca2(B&A) 8520.0–8564.0 8040.0–8160.0, 8786.0–8844.0
Ca3(B&A) 8640.0–8700.0 8040.0–8160.0, 8786.0–8844.0 linesandtelluricabsorptionsimpliesthatthesignal-to-noise
ratio,asafunctionofwavelength,isahighlyinhomogeneous
Ca1(A&Z) 8490.0–8506.0 8474.0–8489.0, 8521.0–8531.0
function.
Ca2(A&Z) 8532.0–8552.0 8521.0–8531.0, 8555.0–8595.0
Ca3(A&Z) 8653.0–8671.0 8626.0–8650.0, 8695.0–8725.0 Following the notation from Cohen (1978) for the cal-
cium triplet,thenewgenericindex will bereferred as CaT.
Ca1(DTT) 8483.0–8513.0 8447.5–8462.5, 8842.5–8857.5
It is defined by establishing five continuum bandpasses,
Ca2(DTT) 8527.0–8557.0 8447.5–8462.5, 8842.5–8857.5
Ca3(DTT) 8647.0–8677.0 8447.5–8462.5, 8842.5–8857.5 and three central bandpasses covering each calcium line
(Ca1, Ca2 and Ca3). The continuum regions were carefully
Ca1(ZHO) 8488.0–8508.0 8447.0–8462.0, 8631.0–8644.0
chosen to optimize the continuum level for all the spec-
Ca2(ZHO) 8532.0–8552.0 8447.0–8462.0, 8631.0–8644.0
tral types, even when the spectra were broadened up to
Ca3(ZHO) 8652.0–8672.0 8447.0–8462.0, 8631.0–8644.0 σ = 300 km s−1. In order to calibrate the CaT contam-
Ca2(A&D) 8532.0–8552.0 8474.0–8489.0, 8559.0–8595.0 ination by the Paschen series in early spectral types, we
Ca3(A&D) 8653.0–8671.0 8626.0–8647.0, 8695.0–8754.0 defined a new generic index, namely PaT, which measures
the strength of three Paschen lines free from the calcium
Ca1(D&H) 8483.0–8511.0 8559.0–8634.0, 8683.0–8758.0
Ca2(D&H) 8517.0–8559.0 8559.0–8634.0, 8683.0–8758.0 contamination. The continuum bandpasses are the same as
Ca3(D&H) 8634.0–8683.0 8559.0–8634.0, 8683.0–8758.0 in CaT whereas the spectral-feature bandpasses (Pa1, Pa2
andPa3) arecenteredonthelinesP17, P14andP12of the
Ca1(RHS) 8490.0–8506.0 8346.0–8489.0, 8563.0–8642.0
series. In both definitions, the multiplicative factors equal
Ca2(RHS) 8532.0–8552.0 8346.0–8489.0, 8563.0–8642.0
unityand theanalytical expressions are
Ca3(RHS) 8653.0–8671.0 8563.0–8642.0, 8697.0–8754.0
CaT=Ca1+Ca2+Ca3, and (1)
passes, the contribution of each spectral-feature bandpass PaT=Pa1+Pa2+Pa3. (2)
can be modified by defining a multiplicative factor for each
ThebandpasseslimitsfortheCaTandPaTindicesare
one, and the pseudo-continuum is derived by using an er-
listedinTable4,andthebandpassespositionsandpredicted
ror weighted least-squares fit to all the pixels of the con-
continuaareillustratedinFig.3fordifferentspectraltypes.
tinuum bandpasses. Although at first sight it could seem
It is important to note that, apart from rough correc-
thatthisapproachisexactlythesamethattheoneobtained
tionsappliedtotheintegratedspectraofyoungstarclusters
byaddingseveralclassical indices(weightedwiththecorre-
(Alloin &Bica 1989) andtheoretical considerations onsyn-
sponding multiplicative factors), the fact that all the spec-
theticspectra(Chmielewski2000), fewpreviousworkshave
tralfeaturessharethesamecontinuumhasanimportantef-
faced thePaschen contamination in individual stars.
fect in thecomputation of theindexerror as it is explained
Finally, we have defined a new calcium triplet index,
in Appendix A2. It is also convenient to highlight that the namelyCaT∗,whichexpressesthestrengthoftheCaiilines
requirementsincorporatedinthedefinitionofgenericindices
correctedfromthecontaminationbyPaschenlines.Thenew
translate intoimportant improvements:
index was defined by imposing the following requirements.
(i) Generic indices are specially suited for the measure- First, the values of CaT∗ should be very similar to those
ment of adjacent spectral features, where the use of a com- of CaT for late-type stars. Secondly, since the true calcium
∗
mon continuum level can be useful. They are also highly strengthinhotstarsisveryloworevennull,theCaT index
recommendedwhenthespectralregionofinterestisdensely should tend to values around zero for the earliest spectral
populated by spectral features, since many thin continuum types. Note that, since the CaT index in hot stars is actu-
bandpasses at adequate locations make it possible to avoid allymeasuringthestrengthofthePaschenlinesratherthan
thepresenceoftheotherspectralfeatures.Moreover,alarge thoseofCa,suchanindex(aswellasthoseusedinprevious
numberof continuum bands guarantees a low sensitivity to works)canlead towrongconclusions wheninterpretingthe
theS/N ratio and a robust continuum definition. integrated spectra of young stellar populations.
(ii) As we will prove later, the multiplicative factors are To estimate the hydrogen contribution to the calcium
useful to remove the contamination by other spectral fea- index, we have compared the CaT and PaT indices for hot
tures (eitherin absorption or in emission). stars,whichshowapuresmoothPaschenseriesintheirspec-
(cid:13)c RAS,MNRAS000,1–24