Calcium and light-elements abundance variations from high resolution spectroscopy in globular clusters

0.18 MB·English

by  Eugenio Carretta

PreprinttypesetusingLATEXstyleemulateapjv.11/10/09
CALCIUM AND LIGHT-ELEMENTS ABUNDANCE VARIATIONS FROM HIGH RESOLUTION
SPECTROSCOPYIN GLOBULAR CLUSTERS 1
Eugenio Carretta2, Angela Bragaglia 2, Raffaele Gratton3, Sara Lucatello3,4, Michele Bellazzini2,
Valentina D’Orazi3
ABSTRACT
We use abundances of Ca, O, Na, Al from high resolution UVES spectra of 200 red giants in 17
0
globular clusters (GCs) to investigate the correlation found by Lee et al. (2009) between chemical
1
enrichment from SN II and star-to-star variations in light elements in GC stars. We find that (i) the
0 [Ca/H] variations between first and second generationstars are tiny in most GCs (∼0.02−0.03 dex,
2
comparable with typical observational errors). In addition, (ii) using a large sample of red giants

n in M 4 with abundances from UVES spectra from Marino et al. (2008), we find that Ca and Fe
a abundances in the two populations of Na-poor and Na-rich stars are identical. These facts suggest
J that the separation seen in color-magnitude diagrams using the U band or hk index (as observed

9 in NGC 1851 by Han et al. 2009) are not due to Ca variations. Small differences in [Ca/H] as
2 associated to hk variations might be due to a small systematic effect in abundance analysis, because
most O-poor/Na-rich (He-rich) stars have slightly larger [Fe/H] (by 0.027 dex on average, due to

] decreased H in the ratio) than first generation stars and are then located at redder positions in the
A
V,hk plane. While a few GCs (M 54, ω Cen, M 22,maybe even NGC 1851)do actually show various
G degree of metallicity spread, our findings eliminate the need of a close link between the enrichment
. by core-collapse SNe with the mechanism responsible for the Na-O anticorrelation.
h
p Subject headings: stars: abundances — stars: evolution — globular clusters: general
-
o
r 1. INTRODUCTION The primordial P component of first generation stars,
t
s Starsofdifferentstellargenerations5arecurrentlyrou- present in almost all Galactic GCs surveyed so far,
a has the high [O/Fe] and low [Na/Fe] ratios typical
tinely found in all Galactic globular clusters (GCs). A
[ of core-collapse supernovae (SNe II) nucleosynthesis,
common misunderstanding is to identify multiple se-
characteristic of halo field stars with similar metal-
1 quences in the color-magnitude diagrams (CMDs) as
licity. On the other hand, Carretta et al. (2009a)
v the only evidence of multiple stellar populations in GCs,
2 while they are only the tip of the iceberg, observable showed that two other components of second gener-
0 whendifferencesinchemicalcompositionand/orageare ation stars may be found: one dominant component
0 very large. However, all GCs display large spreads in has intermediate composition (I; observed in each GC),
0 abundancesoflightelements. Thesearedue tothe pres- and another has extremely (E) modified composition,
. showing large O-depletion and Na (and Al) enhance-
2 ence of two generations of stars, separated by a small
0 difference in age, not enough to appear as a smearing ments, present only in the most massive clusters6.
0 of the cluster turn-off, but clearly visible as a different This general pattern can be qualitatively produced by
1 chemical signature. This difference can be uncovered by proton-capture reactions in H-burning at high temper-
: spectroscopy (see Gratton et al. 2004 for a recent re- ature (Denisenkov & Denisenkova 1989; Langer et al.
v view). Since its discovery by the Lick-Texas group (see 1993). The fact that the same Na-O and Mg-Al
Xi the early review by Kraft 1994), the Na-O anticorrela- anticorrelations are found in unevolved stars in sev-
tion in GCs offers a powerful instrument to resolve age eralGCs(Gratton et al. 2001;Ramirez & Cohen 2002;
r
a differences as small as a few 107 yrs (should very mas- Carretta et al. 2004)tellsusthatthiscompositionisin-
sive stars be involved)by means of a signal as large as 1 herited from the ashes of a previous generation of stars.
full dex in abundance. Whenever the Na-O anticorrela- While the involved processes have been largely iden-
tion is observed in a GC, multiple stellar generations are tified, it is still not clear where these reactions oc-
present in the cluster. curred, the favorite sites being either main-sequence
core H-burning of fast rotating massive stars (FRMS,
1Based on data collected at the European Southern Observa- Decressin et al. 2007) or the hot bottom of the con-
tory, Chile, programmes 072.D-507,073.D-0211,072.D-0742,077.D- vective envelope in intermediate-mass asymptotic giant
0182 branch (AGB) stars (Ventura et al. 2001). Whichever
2INAF, Osservatorio Astronomico di Bologna, via Ranzani
the sites are, they can not synthesize iron or other el-
1, 40127, Bologna, Italy. [email protected], an-
[email protected],[email protected] ements heavier than Al. Indeed, apart from a few no-
3INAF, Osservatorio Astronomico di Padova, vi- table exceptions (ω Cen, see reference in Gratton et al.
colo dell’Osservatorio 5, 35122 Padova, Italy. raf- 2004; M 22, Marino et al. 2009; Da Costa et al. 2009;
[email protected],[email protected],
M 54, Carretta et al. 2010a; Terzan 5, Ferraro et al.
[email protected]
4Excellence Cluster Universe, Technische Universit¨at
Mu¨nchen,Boltzmannstr. 2,D-85748,Garching,Germany 6 The operational distinction between I and E components
5 We will use the words stellar “generation” and “population” (Carrettaetal. 2009a) is not relevant here and we only consider
interchangeably, forthereasonsexplainedlater. thetotal fraction(I+E)ofsecondgenerationstars.
2 Carretta et al.
2009; and possibly NGC 1851, Yong & Grundahl 2008) 3. RESULTS
the level in [Fe/H] or α and Fe-group elements is very
Sincewe haveonlyabout10starspercluster observed
constant in GCs (Carretta et al. 2009b).
with UVES, we combined the 17 GCs, separating first
However,arecentworkbyLee et al. (2009,fromnow
generation(P)starsfromsecondgeneration(I+E)stars.
L09) seems to challenge this consolidated scenario, be-
For direct comparisonwith L09, we used [Ca/H]instead
causeintheirCa-uvby surveythey foundaspreadinthe
of[Ca/Fe],afternormalizingthe individualvaluestothe
photometric hk index (including Ca ii H and K lines)
meanofeachcluster. InFigure1weshowthehistogram
for several GCs. They interpret this spread as due to
in [Ca/H] for P and I+E stars, with average values, er-
Ca abundance variations and claimthat this α−element
rors, and rms. About two thirds of stars belong to the
was produced by SN II in a past phase of the lifetime
second generation, and the average normalized [Ca/H]
of GCs. In turn, these should be assumed to be initially
values do not differ from those of first generation (at
muchmoremassivethanatpresent,elsethehighlyener-
about 1σ). A further check comes from the cumula-
getic SNe winds wouldhavebeen lostfromthe potential
tivedistributions,showninFigure2(upper panel),from
well. Moreover, they find that the Ca-strong group in
which we excluded the 12 stars of NGC 2808, the GC
GCs with evidence of multiple stellar populations from
in our sample showing the most extreme variations in
spectroscopyisassociatedtothepopulationofstarswith
He (Piotto et al. 2007; Bragaglia et al. 2010, and be-
lowerOand higherNa values, andvice versaconcerning
low in this Letter). The two distributions cannot be
the Ca-weak stars.
distinguished. Note that if we include also NGC 2808,
The need ofmassiveproto-clustersis commonto most
the probability of the Kolmogorov-Smirnov drops to
scenariosofchemicalevolutionofGCs,includingonepro-
0.07, still compatible with no significant difference, but
posed by us (Carretta et al. 2010b). However a corre-
possibly indicating some actual variations in this par-
lation between Ca and light elements would imply that
ticular cluster. However, when we show in the three
the second generation stars were formed from material
lowerpanels of Figure 2 the cumulative distributions for
enriched not only by FRMS or AGB stars, but also by
[O/Fe],[Na/Fe],and[Al/Fe],allelementsinvolvedinthe
SN II events, which is a substantial modification of the
(anti)correlationsinGCs,weclearlyseeastrikingdiffer-
current scenario.
ence (compare e.g. to supplementary Fig.14 of L09).
InthepresentLetterweexploitthelargewealthofdata
Our sample of 200 stars comes from 17 different GCs,
fromourongoingproject“Na-OanticorrelationandHB”
but there is already available in literature a sample of
(Carretta et al. 2006)andfromotherextensivespectro-
similarsizeforasingleGC,M4,forwhichMarino et al.
scopic surveys to compare [Ca/H] ratios obtained from
(2008) analyzed high resolution UVES spectra of 105
high resolution spectroscopy with the abundance distri-
RGB stars. This cluster is also in the L09 sample; from
bution of O, Na, Al in a sample of several hundred red
their supplementary Fig.14 (panel i) we see that almost
giantbranch(RGB)starsin17GCs. Thisallowstoshed
all stars defined Ca-strong are Na-rich (in our interpre-
light on the rˆole of Ca in the chemical evolution of GCs;
tation,secondgenerationI,E)andviceversafortheCa-
from our extensive spectroscopic database, we seek to
weak. However, when we divide the stars in the Marino
confirmwhether the spreadin the photometric hk index
et al. sample in P and I+E (using their same separation
is driven by a real spread in Ca abundance.
at [Na/Fe]=0.2) and compare the histograms and the
means of the two populations (see Figure 3, left panel),
2. ATMOSPHERICPARAMETERSANDABUNDANCE
ANALYSIS they look very similar7. This is supported by the cu-
mulative distributions (Figure 3, right panel): they are
The full description of the analysis is given in
indistinguishable, according to the KS test. A further
Carretta et al. (2009b); here we present only the abun-
checkwasdoneconsideringonlystarsinthetwoextreme
dances of Ca, which are based on a large number of
quartiles in Na abundance: the averageCa values in the
lines(typically10-12inthemostmetal-poorclusterslike
more Na-poor and in the more Na-rich stars perfectly
M15andalmost20inmetal-richGCs like47Tuc). The
agreewith the ones in Figure 3 for the I+E and P stars,
[Ca/Fe] ratios with the number of lines and the line-to-
respectively.
line r.m.s. scatter for each star are listed in Table 1
The stars called Ca-weak (Ca-strong) on the basis of
(completely availableonline only). InTable 2we report
theirhk indexdonotseemtohavelower (higher) [Ca/H]
the number of stars in each cluster, the average [Ca/Fe]
valuesfrom direct analysis ofhigh resolution spectra; this
and [Ca/H] values and the 1 σ scatter of the mean. The
demonstrates that thedifferences in the hk index mustbe
rms scatters in [Ca/Fe] are smaller than in [Ca/H] be-
produced by something other than Ca variations.
cause the sensitivity of Ca and Fe lines to atmospheric
It is important to stress that the observational effect
parameters is quite similar. Five of the clusters listed in
at the basis of the L09 claim corresponds to 1σ spreads
Table 1 are included in the sample by L09, who claim
in [Ca/H] of about 0.03 dex in the most extreme cases
significant spread in Ca for four of them.
(NGC 1851,NGC 2808, excluding the special cases of ω
The values listed in Table 2 indicate that the mean
Cenand M22 where the presence ofa significantspread
level of Ca is highly homogeneous in each GCs. The
inironhas alreadybeenestablishedwithhighresolution
averager.m.s. star-to-starscatterfor[Ca/Fe]is0.03dex
spectroscopy)andofabout0.02dexfortypicalcases(like
from 17 GCs, to be compared with average scatter of
M 4, M 5, and NGC 6752, see Table 2)8. In our view,
0.18 dex, 0.22 dex, 0.07 dex and 0.25 dex for O, Na,
these numbers call for two basic considerations:
Mg, and Al respectively. Note that the mean scatter in
Mg, produced by core collapse SNe but also involved in
7 As noticed by the referee, the same result holds using the 38
proton-capture reactions, is more than twice that of Ca,
starsinNGC6752analyzedbyYongetal. (2005).
which does not participate in any H-burning. 8 We used their supplementary Table 3 for the FWHM of the
Ca and light elements abundances 3
1. Such tiny spreads are similar in amplitude to sev- in He and light elements only, with no fresh production
eral uncertainties that are known to affect the of Fe and Ca. The sizable effect on the Kolmogorov-
abundance analysis (like e,g, variations in the He Smirnovtestofthe inclusionofNGC 2808inthe sample
content); they areusuallyneglectedsincethe over- (Sect. 3) lends further support to this scenario, as this
alluncertaintyonsinglemeasuresistypicallylarger cluster seems to display especially large variations in He
than this. We do not exclude that there might be abundance (Piotto et al. 2007; Bragagliaet al. 2010).
dispersions at these very low levels, but presently Hence, while it is quite possible that in NGC 1851
theycannotbeseparatedfromtheintrinsicerrors. some small intrinsic scatter in iron does actually exist
(Yong & Grundahl 2008),ourresultsdonotsupportthe
2. evenifweregardtheseverysmallvariationsasreal, ideathatcorecollapseSNecontributedtotheenrichment
somebasicalgebraindicatesthat,assumingamass ofsecondgenerationstarsinmostGCs,andindicatethat
of2×105 M⊙ andametallicityofZ =10−3forthe the production of the proton-capture elements occurred
gascloudthatgaveorigintothecluster,asingleSN in sites not able to synthesize Ca and Fe-peak elements.
IIeventwouldbesufficienttoproducetherequired
amount of Ca (and Fe). 5. DISCUSSIONANDCONCLUSIONS
4. THECASEFORNGC1851 Lee and coworkers devised a scenario for chemical en-
richment in clusters with multiple populations. It in-
If the tiny variations detected in most GCs are not
cludes two-phases. First, core collapse SNe from first
measuring intrinsic scatter in Ca, what might be the
generation stars inject in the intra-cluster medium ma-
cause for the variations in the hk index? L09 discussed terial enriched with Fe and α−elements; at least part of
severalissues to show that the only way to affect the hk
this materialisnotlostby the GCs owingto their larger
index isachangeinCa. Specifically, they madeastrong
mass at early epochs. Second generation stars form
case for NGC 1851, where the few “Ca-strong” stars
from this enriched material, incorporating the ejecta of
(with abundances from high resolution spectroscopy by
intermediate-mass AGB stars with longer evolutionary
Yong & Grundahl (2008) are segregated to the red-
times. The second generation stars are then produced
dest positions in the V,hk CMD. In a recent paper,
from matter enriched both in heavy and light elements.
Han et al. (2009) presented a split of the RGB in this
However,in our opinion, this scenario has to face sev-
cluster in the U,U −I plane, where the redder sequence
eral problems:
is populated by Ca-strong stars.
We simply note the striking similarity between the • Since there is clear evidence of multiple popula-
CMD by Han et al. with the U,U −B CMD for M 4 tions from spectroscopy in all GCs, except per-
by Marino et al. (2008). In both cases, the RGB is haps a few small ones with only a handful of
clearly split into two distinct sequences. However, the stars analyzed (e.g. Pal 12, Cohen 2004; Ter 7,
same dichotomy of RGB sequences in M 4 is explained Sbordone et al. 2007;bothbelongingtotheSagit-
by Marino et al. (2008) asthe Nenhancementaffecting tarius dwarf galaxy), any proposed scenario must
the U − B colors (through the CN and NH molecular accountforthe formationofthe generalityofGCs.
bands within the U filter), so that Na-rich (and likely
N rich) stars are separated from Na-Poor (and N-poor) • WhileasinglecorecollapseSNmightbeenoughto
stars in the CMDs. Since we demonstrated in Figure 3 produce the claimed scatter in [Ca/H], thousands
that in M 4 the [Ca/H]ratio is identical for the Na-poor (or even more) FRMS or massive AGB stars are
andNa-richstars,thesameconclusionmightbevalidfor required to produce the observed pattern for ele-
NGC 1851. ments involved in proton-capture processes. The
A further piece of the puzzle comes from impact of these two mechanisms can not be the
Carretta et al. (2009a), who confirmed results by same.
Yong et al. (2008) in NGC 6752: Na-poor stars define
a narrow N-poor sequence all confined to the bluest • Core collapse SNe should produce a (variable)
part of the RGB in NGC 6752, whereas the Na-rich enhancement in the mean abundances for other
(N-enhanced) stars are spread out to the red part of the α−elements. No significant variation has been ob-
RGB, reaching large N abundances. Now, regardless served for Mg and Si (apart from those explained
of the nature of the polluters, the ejected matter is by the Mg-Al cycle; see Fig. 3 in Carretta et al.
always enriched in the main outcome of H-burning, 2009b), or Ti (Gratton et al. 2004).
He. Bragaglia et al. (2010, in preparation) used a large
sample (about 1400 RGB stars in 19 GCs) to show that • InGCslikeM4,asplitinhkof45σ (hk)(measure-
p
stars with higher He content should have on average ment errors for the populations, L09, supplemen-
slightly larger [Fe/H] values the difference between the tary Fig. 10) between two populations translates
E and P stars being 0.027 dex in [Fe/H], even if there is in virtually no difference in the [Ca/H] ratios ob-
no difference in the overallmetal content Z. tained from a homogeneous large sample of stars
These findings suggest that the correlation observed with high resolution spectra.
by L09 between the (small) dispersion in [Ca/H] and
the (large) spread in hk may be related to variations • If FRMS are the polluters, the whole model would
fail, since these stars release the polluted matter
RGBoftheGCwehaveincommon,convertedthembacktoscatter in very short time, while still in MS. Hence, the
inhk assuminga Gaussian distribution, and further converted to
timescale is even shorter than that for production
scatter in [Ca/H] and [Fe/H] using the sensitivity of Ca on the
variationinhk (suppl. Sect. 4)andtherelationsinSect. 3. of core collapse SNe.
4 Carretta et al.
• The model by L09, adapted to explain the dou- systematic second-order effects due to the fact that sec-
ble RGB seen in NGC 1851, results in a number ond generation stars are He-enhanced, and appear more
ratio of first-to second generation stars equal to metal-richandhotterwhenanalyzedwithhighresolution
3 (75%/25%). From spectroscopy, on the other spectroscopy.
hand,itwasfoundthatthebulkofstarsinGCsbe- At the moment, we cannot offer a viable alternative
long to the second generation (Cohen et al. 2002; explanation for the dispersion in the hk index observed
Carretta et al. 2009b), with a number ratio typi- by Lee and collaborators, although this issue certainly
cally of about 0.5 (0.33/0.66). We stress that this requires further studies.
feature is not a prerogative of mono-metallic GCs,
the same value for the primordial fraction is also
Partial funding come from PRIN MIUR 2007, PRIN
found in ω Cen and M 54 (Carretta et al. 2010a),
INAF 2007, and the DFG cluster of excellence ”Origin
the two most massive clusters in the Galaxy, both
and Structure of the Universe”.
showing actual (large) dispersion in [Fe/H]: this
finding demonstrates that the relative fractions of REFERENCES
first and secondgenerationstars arenot relatedto
the enrichment of heavy elements. Bragaglia,A.,etal.2010,inpreparation
Carretta,E.,GrattonR.G.,Bragaglia,A.,Bonifacio,P.&
• Ther−processelementeuropium,whosemainpro- Pasquini,L.2004, A&A,416,925
duction site is expected to be core-collapse SNe, is Carretta,E.,Bragaglia,A.,GrattonR.G.,Leone,F.,
fairlyconstantwithin allthe studiedglobularclus- Recio-Blanco,A.,Lucatello,S.2006, A&A,450,523
Carrettaetal.2009a, A&A,505,117
ters (Gratton et al. 2004).
Carretta,E.,Bragaglia,A.,Gratton, R.G.,Lucatello, S.2009b,
• Finally, one of the strongest objection to the L09 A&A,505,139
Carrettaetal.2010a, ApJLetters,submitted
modelcomesfromthe runofAl. Withitsyieldex- Carretta,E.,Bragaglia,A.,Gratton, R.G.,Recio-Blanco,A.,
tremely dependent on the neutron excess (metal- Lucatello, S.,D’Orazi,V.,Cassisi,S.2010b,A&A,submitted
licity), any variations in the content of Fe (or Cohen,J.G.2004,AJ,127,1545
Cohen,J.G.,Briley,M.M.,Stetson, P.B.2002, AJ,123,2525
heavy element produced in SN II, like Ca) should
DaCosta,G.S.,Held,E.V.,Saviane, I.,Gullieuszik,M.2009,
be accompanied by clear changes in the primor-
ApJ,705,1481
dial Al abundance. As shown by Carretta et al. Decressin,T.,Meynet, G.,CharbonnelC.Prantzos,N.,Ekstrom,
(2009b) this is not always seen, in particular in S.2007, A&A,464,1029
clusters like M 4 where a bimodal distribution is Denisenkov,P.A.,&Denisenkova, S.N.1989, A.Tsir.,1538,11
Ferraro,F.R.etal.2009, Nature,462,483
claimed in the hk distribution. The Al distribu-
Gratton,R.G.etal.2001,A&A,369,87
tion is far from being bimodal in M 4, on the con- Gratton,R.G.,Sneden, C.,Carretta,E.2004, ARA&A,42,385
trary [Al/Fe] is quite constant and does not vary Han,S.-I.,Lee,Y.-W.,Joo,S.-J.,Sohn,S.-T.,Yoon,S.-J.,Kim,
between first and second generation stars. This H.-S.,Lee,J.-W.2009,ApJ,707,L190
is clearly seen also in other GCs like NGC 6171, Kraft,R.P.1994,PASP,106,553
Langer,G.E.,Hoffman,R.,&Sneden, C.1993,PASP,105,301
NGC 288, and NGC 6838 from our own data
Lee,J.W.,Kim,S.-L.,Lee,C.-U.,Youn,J.-H.2009,Nature,462,
(Carretta et al. 2009b)aswellinthelargersample 480
by Marino et al. (2008). This constancy, due to Marino,A.F.,Villanova,S.,Piotto,G.,Milone,A.P.,Momany,
the secondary (metallicity-dependent) production Y.,Bedin,L.R.,Medling,A.M.2008,A&A,490,625
of Al in SNe strongly precludes that any signifi- Marino,A.F.,Milone,A.,Piotto,G.,Villanova,S.,Bedin,L.,
Bellini,A.,Renzini,A.2009,A&A,505,1099
cant difference in heavy metal-content is provided
Piotto,G.,etal.2007,ApJ,661,L53
by SNe to stars in different stellar generations in Ramirez,S.,Cohen, J.G.2002,AJ,123,3277
GCs. Sbordone,L.,Bonifacio,P.,Buonanno, R.,Marconi,G.,Monaco,
L.,Zaggia,S.2007, A&A,465,815
We conclude that (i) the spreads in [Ca/H] are very Ventura,P.D’Antona,F.,Mazzitelli,I.,&Gratton, R.2001,
small; (ii) different distributions in O, Na, Al are clearly ApJ,550,L65
Yong,D.,Grundahl,F.2008,ApJ,672,L29
seenbetweenfirstandsecondgenerationstarsinallGCs
Yong,D.,Grundahl,F.,Johnson,J.A.,Asplund,M.2008, ApJ,
analyzed so far, on the contrary no statistically signifi- 684,1159
cant difference is found regarding Ca abundances from Yong,D.,Grundahl,F.,Nissen,P.E.,Jensen,H.R.,Lambert,
high resolutionspectroscopy; (iii) the small difference in D.L.2005,A&A,438,875
Ca observed in some cases might be well explained with
Ca and light elements abundances 5
Table 1
[Ca/Fe]and[Ca/H]ratiosforindividualstars. Completetableavailableonlineonly.
GCID star nr. [Ca/Fe] σ [Ca/H] flagPIE
NGG0104 5270 15 0.299 0.150 −0.473 2
NGG0104 12272 14 0.310 0.107 −0.437 2
NGG0104 13795 11 0.345 0.105 −0.487 1
NGG0104 14583 14 0.309 0.110 −0.375 2
NGG0104 17657 11 0.298 0.127 −0.546 3
NGG0104 18623 19 0.322 0.213 −0.451 2
a nristhenumberofCalinesineachstar.
bflagis1,2,and3fortheprimordialP,intermediateIandextremeEcomponent,respectively;flag0meansthateitherNaorOwasmissing
andthestarcouldbenoassignedtoacomponent.
6 Carretta et al.
Table 2
Mean[Ca/Fe],[Ca/H]ratiosandinternalerrors.
GCID nr [Ca/Fe] σ err1 err2 [Ca/H] σ
NGC0104 11 +0.31 0.01 0.03 0.03 −0.45 0.06
NGC0288 10 +0.41 0.02 0.02 0.03 −0.90 0.05
NGC1904 10 +0.28 0.01 0.02 0.02 −1.30 0.04
NGC2808 12 +0.34 0.02 0.03 0.03 −0.81 0.07
NGC3201 13 +0.30 0.03 0.02 0.02 −1.21 0.07
NGC4590 13 +0.26 0.03 0.03 0.03 −2.00 0.04
NGC5904 14 +0.38 0.02 0.02 0.03 −0.96 0.05
NGC6121 14 +0.42 0.03 0.02 0.03 −0.75 0.07
NGC6171 5 +0.41 0.02 0.03 0.03 −0.63 0.08
NGC6218 11 +0.42 0.03 0.03 0.03 −0.91 0.04
NGC6254 14 +0.34 0.04 0.02 0.02 −1.23 0.07
NGC6397 13 +0.28 0.03 0.03 0.03 −1.71 0.05
NGC6752 14 +0.40 0.03 0.03 0.03 −1.16 0.04
NGC6809 14 +0.36 0.04 0.02 0.02 −1.58 0.07
NGC6838 12 +0.31 0.06 0.03 0.03 −0.53 0.07
NGC7078 13 +0.25 0.05 0.03 0.03 −2.07 0.10
NGC7099 10 +0.31 0.04 0.03 0.03 −2.03 0.05
a
nristhenumberofstarsineachcluster.
b
err1for[Ca/Fe]accountforinternalerrorsinTeff,vt andEW,whereaserr2includesallsourcesofuncertainties(seeCarrettaetal. 2009b).
Ca and light elements abundances 7
Figure 1. Distributionof[Ca/H]ratiosfromhighresolutionUVESspectrafor202RGBstarsin17globularclusters. Thecross-hatched
red area is the histogram for stars of the primordial P component, while the blue hatched area indicates stars of the second generation
(I+Estellarcomponent, seetext.)
8 Carretta et al.
Figure 2. Cumulative distributions for stars of the P component (solid red lines) and for second generation stars (I+E components,
dottedbluelines). Fromtoptobottomthedistributionofabundancesof[Ca/H],[O/Fe],[Na/Fe]and[Al/Fe]areplotted. Inthetoppanel
NGC2808isexcluded(seetext); ifincludedtheKSprobabilitywouldbecome0.07.
Ca and light elements abundances 9
Figure 3. Distributionof[Ca/H]ratiosinM4(fromUVESspectrabyMarinoetal. 2008;leftpanel),andcumulativedistributionsfor
starsmoreNa-rich(blue)andmoreNa-poor(red)than[Na/Fe]=+0.2dex,thatweidentifywiththeI+EandPcomponents, respectively.