arXiv:1111.6068v1 [astro-ph.GA] 25 Nov 2011
Accepted November 21, 2011 to Astrophysical Journal
Preprint typeset using LATEX style emulateapj v. 11/10/09
GLOBULAR CLUSTER ABUNDANCES FROM HIGH-RESOLUTION, INTEGRATED-LIGHT
SPECTROSCOPY. IV. THE LARGE MAGELLANIC CLOUD: α, FE-PEAK, LIGHT, AND HEAVY ELEMENTS1
Janet E. Colucci
Department of Astronomy and Astrophysics, 1156 High Street, UCO/Lick Observatory,
University of California, Santa Cruz, CA 95064; email@example.com
Rebecca A. Bernstein
Department of Astronomy and Astrophysics, 1156 High Street, UCO/Lick Observatory,
University of California, Santa Cruz, CA 95064; firstname.lastname@example.org
Scott A. Cameron
Science Department, 3000 College Heights Blvd., Cerro Coso Community College, Ridgecrest, CA 93555; email@example.com
The Observatories of the Carnegie Institute of Washington, 813 Santa Barbara Street, Pasadena, CA 91101-1292; firstname.lastname@example.org
Accepted November 21, 2011 to Astrophysical Journal
We present detailed chemical abundances in 8 clusters in the Large Magellanic Cloud (LMC). We
measure abundances of 22 elements for clusters spanning a range in age of 0.05 to 12 Gyr, providing
a comprehensive picture of the chemical enrichment and star formation history of the LMC. The
abundances were obtained from individual absorption lines using a new method for analysis of high
resolution (R ∼25,000) integrated light spectra of star clusters. This method was developed and
presented in Papers I, II, and III of this series. In this paper, we develop an additional integrated light
χ2-minimization spectral synthesis technique to facilitate measurement of weak (∼15 m˚ A) spectral
lines and abundances in low signal-to-noise ratio data (S/N∼30). Additionally, we supplement the
integrated light abundance measurements with detailed abundances that we measure for individual
stars in the youngest clusters (Age <2 Gyr) in our sample. In both the integrated light and stellar
abundances we find evolution of [α/Fe] with [Fe/H] and age. Fe-peak abundance ratios are similar
to those in the Milky Way, with the exception of [Cu/Fe] and [Mn/Fe], which are sub-solar at high
metallicities. The heavy elements Ba, La, Nd, Sm, and Eu are significantly enhanced in the youngest
clusters. Also, the heavy to light s-process ratio is elevated relative to the Milky Way ([Ba/Y] > +0.5)
and increases with decreasing age, indicating a strong contribution of low-metallicity AGB star ejecta
to the inter-stellar medium throughout the later history of the LMC. We also find a correlation of
integrated light Na and Al abundances with cluster mass, in the sense that more massive, older
clusters, are enriched in the light elements Na and Al with respect to Fe, which implies that these
clusters harbor star-to-star abundance variations as is common in the Milky Way.
intermediate age and young clusters have Na and Al abundances that are lower and more consistent
with LMC field stars. Our results can be used to constrain both future chemical evolution models for
the LMC and theories of GC formation.
Subject headings: galaxies: individual (LMC) — galaxies: star clusters: general — galaxies: abun-
dances — globular clusters: individual(NGC 2005, NGC 2019, NGC 1916, NGC
1978, NGC 1718, NGC 1866, NGC 1711, NGC 2100) — stars: abundances
We are conducting an ongoing program to obtain de-
tailed abundances from high resolution (R ∼25,000),
integrated light (IL) spectra of extragalactic star clus-
ters.In a series of papers, we have developed and
presented a method to obtain abundances of ∼20 ele-
ments from individual absorption lines in these spectra.
In McWilliam & Bernstein (2008) and Cameron (2009),
hereafter Paper I and Paper II, we developed this method
on a training set of 7 globular clusters (GCs) in the Milky
1This paper includes data gathered with the 6.5 meter Mag-
ellan Telescopes located at Las Campanas Observatory, Chile.
Way (MW), which have ages >10 Gyr. In Colucci et al.
(2011) and Colucci & Bernstein (2011, submitted), here-
after Paper III and CB11, we extended this method to
a training set of 8 massive, high surface brightness star
clusters in the Large Magellanic Cloud (LMC), for which
we obtained measurements of [Fe/H] and age. This sam-
ple of LMC clusters is critical for development of the IL
spectra method on clusters with ages between 0.05 and
10 Gyrs, because the MW has few, if any, massive high
surface brightness clusters in this age regime. In this
paper, we present detailed chemical abundances of ∼20
additional elements in the LMC training set clusters.
The IL method was designed to take advantage of
2 COLUCCI ET AL.
the high luminosities and surface brightness of unre-
solved GCs, as compared to individual stars, in order to
study chemical evolution and cluster properties in distant
galaxies (∼4 Mpc). Therefore, the purpose of developing
this technique is to use GCs to constrain galaxy evolu-
tion in the same way that stars have been used to con-
strain formation of the MW. It has been demonstrated
that star clusters trace the properties of major star for-
mation events in their host galaxies, and therefore they
are crucial tools for learning about more distant galaxies
where individual stars cannot be studied in detail (e.g.
Brodie & Strader 2006).
To this end, it is important to verify that our IL
spectral analysis provides accurate results for clusters
of all ages.The study of young clusters, like those
in our LMC training set, is particularly interesting
for several reasons.First, there are young massive
star clusters currently forming in nearby star-bursting
galaxies like M82, which can teach us about cluster
formation in conditions of high star formation rate
and intensity.These conditions are likely similar to
those present at high redshift, and are important for
understanding the formation of massive clusters in
general (e.g. Leitherer 2001).
interesting in light of the recent results that MW GCs
have small age and abundance spreads, and thus had
multiple, fast star formation episodes at early times (e.g.
Piotto et al. 2007; Mackey et al. 2008; Milone et al.
2009;Goudfrooij et al.
2008;Carretta et al. 2010;
de Mink et al. 2009; Conroy & Spergel 2011; Conroy
2011). By studying IL abundance trends with age and
mass in these young clusters we can try and constrain
cluster formation timescales and internal chemical
enrichment models. Additionally, in starbursting envi-
ronments young clusters allow us to study the intrinsic
cluster luminosity function and cluster disruption, which
can be used to constrain the contribution of cluster stars
to field star populations.
Our training set of clusters in the LMC not only pro-
vides a necessary test of the IL method on young clus-
ters, but also comprehensive information on the chemi-
cal evolution of the LMC itself over a wide range in age
and [Fe/H]. While the star formation history (SFH) of
the LMC field and cluster populations has been well-
studied photometrically and with low resolution spec-
tra (e.g. Harris & Zaritsky 2009; Olszewski et al. 1996;
Cioni et al. 2006; Carrera et al. 2008; Olszewski et al.
1991), detailed abundance information for the LMC is
much more limited than for the MW. This is simply due
to the long integration times required to obtain high S/N,
high resolution spectra of large samples of stars at the
distance of the LMC (D∼50 kpc).
From prior work with high resolution photometry, we
know that the LMC cluster system had its initial forma-
tion epic >10 Gyr ago, and at least one additional burst
2-4 Gyr ago (e.g. Bica et al. 1986; Olszewski et al. 1991;
Harris & Zaritsky 2009). There is only one cluster that
formed in the “age-gap” between 3-10 Gyr. The disk
field population seems to have had a nearly constant for-
mation rate over most of the history of the LMC, with
evidence of localized enhancement in star formation rate
1-4 Gyrs ago (e.g. Geha et al. 1998). The LMC also has
a high surface brightness bar, which shows an underly-
Young clusters are also
Decressin et al.
D’Ercole et al.
ing old population, and some evidence for a burst in star
formation ∼ 6 Gyr ago (Cole et al. 2000).
With detailed chemical abundances, we can further
constrain the SFH of the LMC. There are several use-
ful galaxy formation diagnostics that can be used when
large samples of elements are available. One common fo-
cus of detailed abundance studies is to obtain α-element
abundances (e.g. O, Mg, Ca, Si, Ti), which are pro-
duced mostly in Type II supernovae (SNe II), or mas-
sive stars, on short timescales. Large over-abundances
(with respect to Fe-peak elements) of α-elements are
therefore signatures of early, rapid or bursty star for-
mation (e.g Tinsley 1979; Woosley & Weaver 1995). An-
other useful diagnostic is the ratio of [Ba/Y]. The high
[Ba/Y] ratios observed in nearby dwarf galaxies (e.g.
Shetrone et al. 2001, 2003; Smecker-Hane & McWilliam
2002; McWilliam & Smecker-Hane 2005; Geisler et al.
2007; Tolstoy et al. 2009) reflect the metallicity depen-
dence of the s-process in asymptotic giant branch (AGB)
stars (e.g. Busso et al. 1999).
dence means that high [Ba/Y] ratios identify star forma-
tion that occurred in long-lived low-metallicity environ-
ments, or low star formation rates. High [Ba/Y] ratios
have previously been observed in individual stars in the
LMC (Pomp´ eia et al. 2008; Mucciarelli et al. 2008), and
we determine new [Ba/Y] ratios for the 8 clusters ana-
lyzed in this work.
In this paper, our goals are both to demonstrate
the accuracy of our abundance analysis when applied
to young clusters, and also to obtain new abundance
results for LMC clusters in order to study the chemical
evolution history of the LMC. In §2 we present the
observations and data reduction. In §3 we review the IL
abundance analysis, results from Paper III, and present
a new, supplementary IL spectral synthesis technique.
In §4, we present abundance results from the sample of
individual stars in CB11, which are used to verify the
precision of the IL method at young ages. IL chemical
abundance results are presented in §5. In §6 we discuss
our results compared to abundance analyses from the
literature. In §7, §8 and §9 we discuss the implications
of our results for massive cluster formation and the
formation and evolution of the LMC. Throughout this
work we will frequently refer to previous measurements
of abundances in individual LMC stars by Pomp´ eia et al.
(2008), Johnson et al. (2006), Mucciarelli et al. (2008),
Mucciarelli et al.(2010),
Luck & Lambert(1992),
Hill et al. (2000) and Hill et al. (1995), which will be
abbreviated as P08, J06, M08, M10, M11, LL92, RB89,
H00, and H95, respectively.
This metallicity depen-
Mucciarelli et al.
Russell & Bessell
2. OBSERVATIONS AND DATA REDUCTION
Integrated light spectra for all of the LMC clusters were
obtained by uniformly scanning the core regions of each
cluster. All spectra were obtained with the 2.5 m du Pont
telescope at Las Campanas in 2000 December and 2001
January, as described in Paper I, Paper II and Paper III,
with the exception of NGC 1718 which was observed with
the MIKE double echelle spectrograph (Bernstein et al.
2003) on the 6.5 m Magellan telescope in 2006 November.
The reduction of data from the du Pont telescope is de-
scribed fully in Paper I. The reduction of the data from
the Magellan telescope is described fully in Paper III.
LMC INTEGRATED LIGHT ABUNDANCES3
Figure 1. Example of χ2minimization technique for the Ca I 6439˚ A line for NGC 2019 (left panels) and NGC 1916 (right panels). The
top panels show the observed data as cyan filled circles, and solid black lines show the IL synthesized spectra corresponding to −0.3 <
[Ca/Fe] < +0.7. The thick black line in each figure shows the IL spectrum corresponding to the minimum in the reduced χ2, which can be
seen in the bottom panels as a function of [Ca/Fe].
The LMC training set includes 8 clusters that are well
enough sampled to have well-constrained [Fe/H] and ages
from our analysis in Paper III ( NGC 1916, NGC 2005,
NGC 1916, NGC 1718, NGC 1978, NGC 1866, NGC
1711, and NGC 2100). These clusters were chosen to
span the available range in age and [Fe/H] of high sur-
face brightness clusters found in the LMC. In Paper III
we divided the LMC clusters into three groups accord-
ing to age. “Old” clusters are those with ages >5 Gyr,
“intermediate age” clusters are those with ages 1−4 Gyr,
and “young” clusters are those with ages <1 Gyr. The
clusters in each group are listed in Table 1.
In CB11 we presented an analysis of the [Fe/H] of 10
cool giant stars in the intermediate and young clusters
of our sample. The stellar sample includes 2 stars in
NGC 1978, 3 stars in NGC 1866, 3 stars in NGC 1711,
and 2 stars in NGC 2100. All of the stellar spectra were
taken with the MIKE spectrograph on Magellan. The
description of the target selection, data reduction, and
stellar parameter analysis is presented in CB11.
3. INTEGRATED LIGHT ABUNDANCE ANALYSIS
The IL abundance analysis technique is described in
detail in Paper I, II and III and CB11, and will not
be repeated here. In § 3.1, we briefly summarize the
ages and [Fe/H] derived for the LMC training set clus-
ters in Paper III and CB11. In § 3.2-§ 3.3, we present
the line lists we employ and discuss a new extension of
the ILABUNDs code for determining abundances of in-
dividual lines with a χ2-minimization spectral synthesis
3.1. CMDs and Ages
In Paper III and CB11 we presented a detailed analy-
sis of the [Fe/H] and age solutions for each cluster in the
LMC training set. The [Fe/H] and age for each cluster
are used to create final, representative color magnitude
diagrams (CMDs) from which we get the stellar param-
eters needed for spectral synthesis and abundance deter-
mination (see Papers I, II and III). Here we use these
final representative CMDs with the age and [Fe/H] de-
rived in Paper III and CB11 for each cluster to calculate
abundances for the additional elements.
For the final [Fe/H] and ages we use the results from
Paper III for the old clusters, and the results from CB11
for the intermediate age and young clusters. These re-
sults are summarized in columns 3 and 4 in Table 1.
We constrained the age of old clusters to a range of ∼5
Gyrs, intermediate age clusters to ∼1-2 Gyr, and young
clusters to ∼0.4 Gyr. The acceptable age range for each
cluster leads to a range in derived [Fe/H], corresponding
to the different stellar populations at each age. For old
clusters, the difference in [Fe/H] is generally small (<0.05
dex) because there are only subtle changes in the stellar
populations at older ages. Therefore, for the old clusters
we use one best-fitting CMD to report abundances calcu-
lated for other elements. For intermediate age and young
clusters, which have more rapidly evolving stellar pop-
ulations, the difference in derived abundances between
CMDs of different ages can occasionally be larger than
the line-to-line statistical scatter (σlines) of the individ-
ual Fe lines (>0.1 dex). For intermediate age and young
4COLUCCI ET AL.
clusters, we adopt the mean abundances calculated from
two CMDs spanning the appropriate age range and re-
port an additional uncertainty on the abundances due to
the assumed age as σage.
3.2. EWs and Line Lists
(McWilliam et al. 1995b) to measure absorption line
equivalent widths (EWs) for individual lines in both the
IL and stellar spectra. We interactively fit low order
polynomials to continuum regions for each spectral or-
der. The line profiles are fit with single, double, or triple
Gaussians as needed.
Line lists and log gf values were taken from Paper I,
Paper II, Colucci et al. (2009), and Mel´ endez & Barbuy
(2009), and supplemented with the Kurucz atomic and
molecular line database (Kurucz 1997). The same line
list was used for analysis of both the cluster IL spectra
and the individual stellar spectra. The lines and EWs
measured in the cluster IL spectra are listed in Table 2,
and the EWs measured in the stellar spectra are listed
in Table 3.
In addition to the IL EW analysis used in our previous
work, we have implemented an IL line synthesis compo-
nent to ILABUNDs, which we describe in detail in § 3.3.
Lines with abundances determined using the synthesis
routine of ILABUNDs are labeled “SYN” in column 13
of Table 2, while abundances determined using our orig-
inal EW analysis are labeled “EW.” We have calculated
abundances with hyperfine splitting (hfs) of energy lev-
els taken into account for the elements Sc, V, Co, Cu,
Mn, Ba, Sr, La and Eu. These calculations are noted as
“HFS” in Table 2 and Table 3.
We useprogram GETJOB
3.3. ILABUNDs Line Synthesis
The development of the line synthesis routine for IL-
ABUNDs was motivated by the desire to make additional
measurements of poorly constrained elements by analyz-
ing weak features for which EWs are not otherwise recov-
erable. As for the analysis of all of the elements discussed
in this paper, the line synthesis routine is performed only
with the best-fitting CMD already identified from the
[Fe/H] analysis, as discussed above. We note that the
goal of our synthesis analysis is still to calculate abun-
dances for a given species using individual lines, not to
fit the entire spectrum at one time. Therefore, we limit
the synthesis region to ±10˚ A from the line of interest,
and supplement our standard line lists with the Kurucz
Like the EW calculation routine of ILABUNDs, the
synthesis component calculates spectra for each stellar
type in the CMDs using the 2010 version of MOOG
(Sneden 1973). The spectra for the representative stars
are then flux weighted according to the scheme described
for the EWs in Paper I. The combined synthesized IL
spectrum is then renormalized and convolved with the 1-
dimensional velocity dispersion (σv) of the cluster. The
measurement of σv for the LMC clusters follows the
method described in Colucci et al. (2009), and is pre-
sented in Colucci et al. (2012, in preparation).
A χ2-minimization scheme is used to determine the
best-fitting abundance for a given line. First, the con-
tinuum level is identified in a 30-40˚ A window as the
Figure 2. Example of the χ2-minimization technique for the Nd
II 5319˚ A line for NGC 1866. In the top panel, the observed data
is shown as filled circles, and the solid black lines show the IL syn-
thesized spectra corresponding to +0.0 < [Nd/Fe] < +1.6. Bottom
panel is the same as in Figure 1, only for Nd. This example shows
a region with lower S/N and weaker features than that of Figure
1, and that the abundance is well determined with this technique.
mean value with 3σ lower rejection. Next, a 1˚ A region
used for the χ2fitting must be identified by eye and ad-
justed on a cluster by cluster basis because of blending
effects that vary due to the different σvand metallicities
of the clusters. An example of how this region can vary
between clusters because of different σv is shown in the
top panels of Figure 1. The left panel shows the Ca I
6439˚ A line for NGC 2019 and the right panel shows the
same for NGC 1916. The larger σvof NGC 1916 makes
a larger region necessary for calculating the χ2. Note,
the Ca I 6439˚ A line is a strong, high S/N line for which
EWs can be accurately measured using GETJOB, and is
shown here only as a demonstration of the χ2technique.
For a given line, synthesized spectra are created for a
range of abundances and the reduced χ2is calculated for
each. The minimum in the χ2can then be identified,
as shown in the bottom panels of Figure 1. We deter-
mine the approximate uncertainty in abundance using
the range of values that are within ∼40% of the mini-
mum χ2, shown by the horizontal solid lines in the bot-
tom panels of Figure 1. EWs for synthesized lines are
calculated for the best fitting spectra and are listed in
Table 2. We find excellent agreement between the EWs
LMC INTEGRATED LIGHT ABUNDANCES5
measured using GETJOB and the EWs measured with
the χ2-minimization for the examples in Figure 1. For
both NGC 2019 and NGC 1916 the EWs measured using
the two techniques agree to within 1%.
As an example of a case in which the line synthesis is
particularly useful, we show the χ2-minimization for a
weak (15% flux decrement) Nd II line at 5319˚ A in Fig-
ure 2. The Nd II line falls into the category of being in a
low S/N region of the spectra. The line synthesis is useful
anywhere line depths are reduced. This can occur in low
S/N data, for clusters with high σv, and for weak tran-
sitions. While the uncertainty on the derived abundance
of these synthesized lines is higher even when using the
χ2-minimization technique, we are able to meaningfully
constrain abundances of lines with EWs as small as 20
m˚ A, or in data with S/N∼30. This additional technique
makes it possible for us to use the IL abundance analysis
method on lower S/N data of distant, unresolved clusters
in future work.
4. ACCURACY OF LMC IL ANALYSIS: COMPARISON TO
OUR SAMPLE OF INDIVIDUAL STARS IN YOUNG
To demonstrate that the IL method produces accu-
rate abundances for clusters with ages of 0.05 to 2 Gyr,
we first compare the abundances measured with our IL
abundance analysis method to the abundances measured
in individual stars. These comparisons for NGC 1978,
NGC 1866, NGC 1711, and NGC 2100 are shown in Fig-
ure 3. Wherever possible we have used the same anal-
ysis techniques, i.e. identical line lists, line parameters,
stellar atmospheres, and stellar spectral synthesis codes.
This eliminates many potential systematic offsets that
can be present when comparing our cluster IL results to
abundance analyses performed by other authors (see § 6
for more details).
In Figure 3 we show the comparison between the re-
sults obtained from individual stars in NGC 1978, NGC
1866, NGC 1711, and NGC 2100 to the results obtained
from the IL spectra in those same clusters. To evalu-
ate systematic offsets between the two methods, we have
performed linear least squares fits, with the slopes con-
strained to unity, to both the neutral and ionized species
in each cluster. To show the scatter around the fits we
also show the residuals in the bottom panels of the plots.
The derived offset and scatter for the neutral and ionized
lines for each cluster are listed in Table 4. We find that
the systematic offset in neutral species is <0.1 dex in all
cases except for NGC 1711, for which we only have three
elements available for comparison. The scatter about the
fits for neutral species is 0.40, 0.17, 0.18, and 0.29 dex
for NGC 1978, 1866, 1711, and 2100, respectively. For
ionized species the average systematic offset is <0.16 dex
in all cases, and the scatter is 0.33, 0.22, and 0.26 dex
for NGC 1978, 1866, and 2100, respectively. In both
the neutral and ionized cases the scatter about the fits
is comparable to the σTot of the cluster measurements,
and it is likely that with higher S/N IL spectra the scat-
ter would decrease. For example, the cluster for which
we have the highest S/N data in the IL spectra, NGC
1866, has the smallest scatter in both neutral and ion-
For each cluster in Figure 3 we can look for element
species that are outliers in the fits. It is interesting to
note if the outliers for all of the clusters tend to be of
the same species, which would mean that the IL analysis
method has a large systematic error for these particular
species. For NGC 1978 the outliers are Na I, Ti I, Mn
I and Nd II. For NGC 1866 the only outlier is Sm II.
For NGC 1711 we only have four elements available for
comparison, with no obvious outliers. For NGC 2100 the
outliers are Al I and Ba II. This comparison shows that
the outliers in each cluster are different, so it does not
appear that the IL method has a large systematic offset
for any particular element, with the caveat that we can
only make this detailed comparison for the clusters in
our sample that have ages < 2 Gyr, due to the ages of
our current sample of individual stars.
Finally, it is particularly interesting to look at the com-
parison for the youngest cluster in our sample, NGC
2100. For this cluster we were only able to measure an
upper limit in [Fe/H] (Paper III; CB11), which we found
to be ∼0.4 dex more metal-rich than the mean [Fe/H]
we found from the two stars c12 and b22. From inspec-
tion of Figure 3, we find that even though the [Fe/H] is
offset by +0.4 dex, the abundance ratios are extremely
consistent with the abundance ratios that we derive for
the individual stars. This is not unexpected, because
the relative abundances of elements with respect to Fe
change very slowly, even when the absolute changes are
more significant. This is also consistent with what we
found in Colucci et al. (2009) for old clusters in M31.
Therefore, the IL analysis method can provide reliable
measurements of the abundance ratios even when we can
only obtain a limit on the [Fe/H].
In conclusion, we find good agreement between the
abundance ratios measured from the cluster IL spectra
and the abundance ratios measured from the individual
stellar spectra. The largest discrepancies are found for
the cluster IL spectra with the lowest S/N. In this sam-
ple, there are no elements with abundances that are al-
ways predictably inconsistent between the IL and stellar
analyses. We conclude that in general the cluster IL anal-
ysis method results in abundances that are accurate to
?0.1 dex over the age range 0.05−2 Gyr. The accuracy
of the measured abundance for any species for any one
cluster naturally depends on the accuracy of the mea-
sured equivalent widths, which in turn depend on the
S/N of the data, the velocity dispersion of the cluster,
and the strength of the constraint on the CMD solution.
For the results in this paper, the abundances measured
for any one species in any one cluster have statistical un-
certainties of 0.05−0.3 dex (see columns for σlines, σage,
and σTotin Tables 5 through 12).
5. RESULTS & DISCUSSION : CHEMICAL ABUNDANCES
In the following sections we present abundance results
in our LMC clusters for ∼20 individual elements, includ-
ing α−elements, light elements, Fe-peak elements, and
neutron capture elements. We discuss the behavior of
similar groups of elements with respect to previous abun-
dance work in LMC clusters and field stars, as well as in
comparison to abundance work in MW clusters and field
stars. Many of the LMC abundances measured here are
the first such measurements for certain clusters, and will
be discussed further in §7, §8 and §10.
For the eight LMC clusters, we show the final IL abun-
dances, number of available spectral lines and the uncer-
6 COLUCCI ET AL.
−1.5 −1.0 −0.5 0.00.5 1.01.5
−1.5 −1.0−0.5 0.0 0.51.01.5
−1.5−1.0−0.5 0.0 0.51.01.5
Figure 3. Comparison of abundance ratios from IL and stellar analysis for NGC 1978, NGC 1866, NGC 1711, and NGC 2100. Black
circles and cyan squares show abundances for neutral and ionized species, respectively. The thick dashed line in the top panels of each plot
shows a linear least squares fit to the neutral species, with the slope of the fit constrained to unity. The thin dotted line shows the same fit
to the ionized species. The bottom panels in each plot show the residuals of the neutral and ionized species around their respective trend
lines. The thick dashed lines in the bottom plots mark residuals of ±0.1 dex, to guide the eye. In general the two methods agree to the
±0.1 dex level.
LMC INTEGRATED LIGHT ABUNDANCES7
tainty in the mean of Nlines (σlines=σ/?(Nlines− 1))
for each species in Tables 5 through 12. For the young
and intermediate age clusters we also report the uncer-
tainty in the [X/Fe] ratios that results from the uncer-
tainty in the age solution (σage). For the old clusters
the uncertainty associated with the age of the cluster is
small—usually <<0.05 dex for [X/Fe]— and so is not
included for the individual species. Note that for the
young and intermediate age clusters, age uncertainties
are the dominant systematic uncertainty in the abun-
dances, and the only systematic uncertainty that does
not have an analogy in any calculation of abundances for
individual stars. As in any stellar abundance analysis,
systematic uncertainties due to the details of the calcu-
lations (e.g., assumptions adopted in MOOG, the Kurucz
model atmospheres, α-enhancement effects on H− opac-
ity, gf-values) are unavoidable, difficult to quantify, and
best dealt with by making uniform comparison between
results (see §6). See previous papers in this series for
The results for the 10 stars in the LMC clusters NGC
1978, NGC 1866, NGC 1711 and NGC 2100 are reported
in Table 13. We also report the number of lines of each
species that are measured in each star, and the error in
the mean abundance.
For both the cluster IL and the individual stars, abun-
dances relative to Fe are given using the solar abun-
dance distribution of Asplund et al. (2005), with a so-
lar logǫ(Fe)=7.50. The only exception is that we use
the solar logǫ(O)=8.93 of Grevesse (1989). In the recent
literature, we note that there has been a convergence
of the solar oxygen abundance to a lower value, namely
logǫ(O)=8.66 (Asplund et al. 2005). However, this value
has not yet propagated into the Kurucz model atmo-
sphere grids and so we are forced to adopt 8.93 dex in
our calculations. Moreover, the results in the literature
to which we compare our work have also used a higher
value for all calculations and normalizations. For these
reasons, we have elected to use the higher value for inter-
nal consistency in our analysis and with the literature.
The abundance ratios for neutral species are reported
with respect to [Fe/H]I, and ionized species with respect
In Figures 4 through 16, we show the LMC IL cluster
abundances from our analysis (red squares) as a function
of [Fe/H] and age. The abundances measured in the indi-
vidual stars are shown as black crosses. LMC abundances
are shown with the MW IL GC training set abundances
from Paper II (gray squares). For reference, we show a
compilation of abundances measured for individual stars
in the LMC as small red points and for the MW as small
gray points. The small red and gray points that include
error bars correspond to stellar abundances from clus-
ters in the LMC and MW, respectively. References for
individual star abundances are located in the captions of
Figures 4 through 16. For the old sample of LMC clus-
ters the error bars on the IL abundances in Figures 4
through 16 correspond to the standard error in the mean
of the lines for each species. For the intermediate age and
young clusters the error bar corresponds to σTot, which
includes the σlines and σage added in quadrature. For
elements where only one clean line was measured, σlines
was estimated using the χ2-minimization technique de-
scribed in § 3.3.
5.1. Alpha Elements: Ca, Ti and Si
In this section, we present the abundances for α-
elements Ca I, Ti I, Ti II and Si I. These are the first-ever
α-element abundances for NGC 1916, NGC 1718, and
NGC 1711. Abundances of O I and Mg I are discussed
with the other light elements in §5.2. The abundances
for all clusters are plotted in Figure 4 as a function of
both [Fe/H] (left panels) and age (right panels).
The [Ca/Fe] for all clusters tends to decrease with de-
creasing age and increasing [Fe/H], as one would expect.
The old clusters (NGC 1916, NGC 2019, and NGC 2005)
have [Ca/Fe] in the range +0.2 to +0.4.
mediate age clusters (NGC 1718 and NGC 1978) have
[Ca/Fe] of −0.14 and +0.03, respectively, from IL. Indi-
vidual stars in NGC 1978 have [Ca/Fe]= +0.10 ± 0.13.
From both IL and individual star spectra, the youngest
clusters (age < 1Gyr) have a spread in [Ca/Fe] between
−0.19 and +0.18.
Our measurements, obtained by both IL and individual
stellar spectra, agree with previous results (see detailed
comparison in §6).
available for the LMC is still limited, our results add
to increasing evidence that there is generally a larger
range in [α/Fe] in LMC stars than in MW stars of simi-
lar metallicity (e.g. Venn et al. 2004; Pritzl et al. 2005;
Pomp´ eia et al. 2008; Smith et al. 2002).
could indicate incomplete mixing of the LMC ISM, or
possibly gas inflow or outflow.
In the data for this sample, there are more individual
lines available for Ca I than for any other α-element.
Therefore the Ca I measurements given here are the most
statistically significant of our IL α-element results. While
Ca is not a “pure” α-element — it is also produced in
small quantities in Type Ia supernovae (SNe Ia) — it is
the most accurate and consistent α-element for cluster
comparisons. We find that the most evolution in [Ca/Fe]
occurs between 10 and 2 Gyrs ago, with a decreasing
mean value over that range, and remains more or less
constant since 2 Gyr ago. This implies that the LMC
was dominated by SNe II enrichment early on, followed
by significant SNe Ia enrichment before the second burst
of star formation (∼ 8 Gyr later) formed the intermediate
age clusters (Harris & Zaritsky 2009). This is consistent
with the evolutionary timescale for both SNe-types.
Our [Si/Fe] measurements generally show the same be-
havior as our [Ca/Fe] measurements: a spread in [Si/Fe]
for older clusters and decreasing [Si/Fe] with decreas-
ing age. Fewer Si I transitions make Si I more difficult
to measure than Ca I. Nonetheless previous measure-
ments of [Si/Fe] in field and cluster stars in the LMC
are consistent with our results, including some indica-
tion that mean values for [Si/Fe] are ∼0.1-0.2 dex higher
than those of [Ca/Fe], as can be seen in Figure 4.
In Figure 5, we show the results for Ti I and Ti II, sep-
arated by ionization state. [Ti I/Fe] and [Ti II/Fe] gener-
ally show both a larger line-to-line scatter and scatter be-
tween clusters than [Ca/Fe] or [Si/Fe], which is likely due
to the S/N of the data. Despite the higher scatter, [Ti
I/Fe] and [Ti II/Fe] are generally consistent with [Ca/Fe]
and [Si/Fe] in that Ti decreases with decreasing age. Our
[Ti I/Fe] measurements are higher on average than pre-
vious measurements obtained for field and cluster stars,
but the higher uncertainties in our measurements make
While the abundance information
8 COLUCCI ET AL.
medium and large red squares show the old, intermediate, and young cluster LMC IL abundances, respectively. Large grey squares show the
MW abundances from Paper II. Small black crosses show the individual stars in the young LMC clusters. Small grey and red points show
MW and LMC stellar abundances from the literature. Data for MW stars are from Bensby et al. (2005), Venn et al. (2004), Pritzl et al.
(2005) and references therein. For the LMC stars, triangles show abundances from J06, M08, M10, M11 and circles show abundances from
P08. When possible the abundance ratios of other authors have been adjusted to be consistent with the solar abundance distribution of
Asplund et al. (2005) that was used in our analysis. Ages for MW clusters and old LMC clusters are set to 10 Gyrs, and ages of LMC
clusters of M08 are set to 2 Gyrs.
Left panels show Ca I and Si I ratios as a function of [Fe/H], and right panels show the same as a function of age. Small,
it difficult to determine if there is a systematic offset.
Our [Ti II/Fe] measurements overlap more with the field
and cluster star measurements.
Although there is considerable scatter between clus-
ters, we find that the mean [Ca/Fe], [Si/Fe], and [Ti/Fe]
for the old LMC clusters are consistent with measure-
ments for MW clusters (e.g. Pritzl et al. 2005; Cameron
2009). This is further evidence that, like the MW, the
LMC ISM was dominated by enrichment of SNe II when
the old clusters formed ∼ 10 Gyr ago. To clarify the over-
all trends, we have tabulated the mean [α/Fe] for the old
LMC clusters in Table 14, and the mean [α/Fe] for the
MW IL sample. This is particularly useful for compar-
isons with results in the literature from other abundance
analysis techniques. The mean [α/Fe] for the old LMC
clusters is only ∼0.04 dex lower than for the MW train-
ing set GCs, and consistent within the statistical uncer-
tainty. In Figure 6 we show the mean [α/Fe] as a function
of [Fe/H] for all of the LMC clusters, both old and young,
as well as field stars in the MW and LMC.
5.2. Light Elements: O, Na, Mg, Al
It is well known that MW GCs exhibit star-to-star
abundance variations for light elements involved in
high temperature proton-capture nucleosynthesis (see
Gratton et al. 2004, for a recent summary.). Recently,
we found indirect evidence for abundance variations in
M31 GCs (Colucci et al. 2009), and Mucciarelli et al.
(2009) confirmed that abundance variations for O, Na,
Mg, and Al are also present in three old, metal-poor
GCs in the LMC. This indicates that star-to-star
abundance variations are likely to be ubiquitous and
an integral part of massive cluster formation, not
just limited to the MW. When star-to-star abundance
variations are present, a fraction of stars in the cluster
can exhibit any of the following to varying degrees:
depleted O due to the ON-cycle, enriched Na due
to the NeNa-cycle, and depleted Mg and/or enriched
Al due to the MgAl-cycle.
have tried to connect these star-to-star abundance
variations with the observations of multiple populations
of stars in MW globular clusters (e.g. D’Ercole et al.
2008;Carretta et al.2010;
de Mink et al. 2009; Conroy & Spergel 2011).
As mentioned above, we have already shown that in-
direct evidence for abundance variations can be mea-
sured in cluster IL spectra. Specifically, the IL analysis
shows large scatter in [Mg/Fe] when compared to other
α-elements, as well as a lower mean [Mg/Fe] and signifi-
cantly elevated [Al/Fe] and [Na/Fe]. This was seen both
in our M31 GCs, mentioned above, and in Paper I and
Paper II for the MW IL sample. The same indications
of star-to-star abundance variations are now also clear in
the IL of our old LMC clusters.
For all of our LMC clusters, the Na and Al abundances
are shown in Figure 7, and the Mg and O abundances are
shown in Figure 8. The old LMC clusters in our sample
clearly have elevated [Na/Fe] (∼ +0.5 dex) compared to
Recently, several authors
Decressin et al. 2007;
LMC INTEGRATED LIGHT ABUNDANCES9
Figure 5. The same as Figure 4 for Ti I and Ti II.
Figure 6. Mean [α/Fe] calculated from Ca I, Ti I, Ti II, and Si
I. Symbols and data for single stars are the same as in Figure 4.
field stars and younger clusters. The intermediate age
clusters (NGC 1978 and NGC 1718) have significantly
lower [Na/Fe] with a wide range of values (∼−0.6 to
+0.1). In the younger clusters, [Na/Fe] is only available
in the IL of NGC 1866, where [Na/Fe]=+0.23± 0.2.
Because Na I is difficult to measure, it is worth not-
ing the features from which our results are derived. The
most accessible Na I features in cluster IL spectra are
the 5682/5688˚ A and 6154/6160˚ A doublets, which are
relatively weak (∼30 m˚ A ), and so are measured with
the line synthesis component of ILABUNDS. Stellar Na
I abundances are measured with the SYNTH routine in
MOOG. We have not corrected our cluster IL or stel-
lar abundances for non-LTE effects, but note that for
cool giants (Teff<5000 K) the corrections tabulated by
Gratton et al. (1999) are at most +15 % at high metal-
The range in [Na/Fe] found by Mucciarelli et al. (2009)
is consistent with our result, as can be seen in Figure 7.
Our measurements of [Na/Fe] from 10 individual stars in
the young clusters generally agree with the IL measure-
ments, as can be seen in Figure 7.
Also in Figure 7 are our [Al/Fe] results. Similar to
[Na/Fe], the old clusters clearly also have high [Al/Fe],
with a mean above +0.5 dex, as also true of some old
MW clusters. The intermediate age clusters have [Al/Fe]
abundances similar to [Na/Fe], with values significantly
lower than for the old clusters. For young clusters, the
IL and individual stars both show lower abundances than
the intermediate age clusters. This results in a trend of
decreasing [Al/Fe] with decreasing cluster age, as can
be seen in Figure 7. Like Na I, the Al I measurements
come from line synthesis in both cluster IL and individ-
ual stars. All of our cluster IL abundances are obtained
from the Al I 6696/6698˚ A doublet, while the stellar Al
I abundances are obtained both from the 6696/6698˚ A
doublet and the Al I 7835.31˚ A feature.
Figure 8 shows the results for [Mg/Fe].
generally consistent with other α-element abundances.
Mean values for [Mg/Fe] for the old clusters are in the
range +0.0 to +0.32. We do not see evidence for signifi-
cantly depleted [Mg/Fe] with respect to [α/Fe] in the old
Over the full age range of the sample, [Mg/Fe] gen-
erally decreases with decreasing cluster age. The only
exception to this trend is NGC 1718, for which we ob-
tain a significantly lower value of [Mg/Fe]= −0.9±0.30.
10COLUCCI ET AL.
Figure 7. Same as Figure 4 for the light elements Al and Na. Symbols and data for single stars are the same as in Figure 4. Additional
LMC individual star data (also triangles) are from Mucciarelli et al. (2009). Additional MW GC individual star data are from references
compiled in Carretta (2006).
Figure 8. Same as Figure 4 for the light elements O and Mg. Symbols and data for single stars are the same as in Figure 4. Additional
LMC individual star data (also triangles) are from Mucciarelli et al. (2009). Note that the cluster IL [O/Fe] values are measured from the
7771˚ A triplet, and that no non-LTE correction can currently be applied.
LMC INTEGRATED LIGHT ABUNDANCES11
tra of NGC 1718. Cyan circles show the data and thick black lines
show syntheses of [Mg/Fe] = −1.5,−1.0, and −0.5. All synthe-
sized spectra were created using a CMD with an age of 3 Gyr and
a metallicity of [Fe/H]= −0.73, corresponding to our oldest accept-
able solution for NGC 1718. Both the 4703˚ A and 5528˚ A features
are best fit by the [Mg/Fe]=−1.0 synthesis.
IL line syntheses for two Mg I features in the IL spec-
The low Mg I in this cluster is interesting, although dif-
ficult to interpret. We note that this cluster has one of
lowest masses in our sample. This may suggest that NGC
1718 formed in a poorly mixed environment in which high
mass SNe II (M> 35M⊙) had not contributed metals, as
the highest mass SNe are thought to produce Mg and
O most efficiently (Woosley & Weaver 1995). Our mea-
surement appears robust because we obtain very similar
abundances using spectral synthesis for two Mg I features
which are separated in wavelength by ∼1000˚ A. This is
shown in Figure 9, where we demonstrate that the best
fitting syntheses have [Mg/Fe]∼ −1.0.
For the young clusters in our sample, we are able
to measure Mg I in the cluster IL spectra of NGC
1866, and obtain [Mg/Fe]=−0.27 ± 0.2.
[Mg/Fe]=−0.03 ± 0.12 from the individual stars in this
cluster, which is consistent with the IL result, given the
large uncertainties. From the individual stars in NGC
1711 and NGC 2100 we obtain [Mg/Fe]=+0.08 ± 0.08
and −0.16 ± 0.04, respectively. Previous measurements
in individual stars (e.g. in NGC 1978) are consistent with
our results and are discussed further in §6.
O I is always difficult to measure in cluster stars,
but is particularly difficult in the IL because of the ve-
locity dispersion of the clusters (i.e. line broadening).
Our IL estimates for O I are obtained by line synthe-
sis of the 7771˚ A triplet; the 6300˚ A forbidden line is
consistently too weak.For the old clusters, even the
7771˚ A triplet is quite weak and we only obtain limits
on O of [O/Fe]<+1.0 for NGC 2019, NGC 1916, and
NGC 2005. For NGC 1866 (∼100 Myrs) we measure
[O/Fe]= +0.2 ± 0.3. It is important to note that the
7771˚ A triplet can have significant non-LTE effects in
cool stars, which must be kept in mind when comparing
to O abundances from the 6300˚ A forbidden line. Un-
fortunately, there is no obvious way to correct the 7771
˚ A IL abundances for non-LTE effects, so we report the
abundances with this caveat.
To these IL results, we can add several measurements
from the stars in young clusters. The O I abundances
for the individual stars were measured from the 6300
˚ A forbidden line. The stars in both the intermediate
age and young clusters have a constant value of [O/Fe]∼
+0.25, similar to previous results for other LMC stars
In summary, we find evidence for star-to-star abun-
dance variations in the old clusters in our sample in
the form of highly enriched [Na/Fe] and [Al/Fe]. The
[Mg/Fe] and [O/Fe] abundances are not indicative of
star-to-star abundance variations on their own, but are
still consistent with this picture. We find that [Na/Fe],
[Al/Fe] and [Mg/Fe] generally decrease with decreasing
cluster age. In §7 we discuss the implications of our mea-
surements further, including the dependence of the IL
abundance with cluster mass.
5.3. Fe-peak Elements
Fe peak elements are well-studied in individual stars
in the MW and LMC. In general, the abundances of Ni,
Cr, Sc, V, and Co tend to scale with Fe, so that the
[X/Fe] ratios for these elements are approximately so-
lar for [Fe/H]> −2.0, or occasionally sub-solar in the
LMC (e.g. P08). The abundance of Mn, however, has a
plateau value of [Mn/Fe]∼ −0.4 for [Fe/H]< −1.0, and
then increases to solar ratios between [Fe/H]=−1.0 and
We measure Ni I, Cr I, Sc II and V I in most of our
LMC sample.We find that the abundance ratios for
these elements are consistent with solar ratios across all
of the ages and metallicities of the clusters, as shown in
Figure 10. The scatter between cluster IL measurements
is largest for Sc II, similar to what was found for IL mea-
surements in the MW and in clusters in M31 (Cameron
2009; Colucci et al. 2009).
The Mn I abundances that we obtain for the LMC clus-
ters are fairly constant, with a mean value of [Mn/Fe]∼
−0.1 dex, as shown in Figure 11. The older, lower metal-
licity clusters do not reach the MW low of [Mn/Fe]∼
−0.4. However, they do overlap with our MW IL mea-
surements, so it is difficult to tell if the LMC ratios differ
The [Mn/Fe] of the intermediate age and young clus-
ters overlap with the higher metallicity MW field stars
of Venn et al. (2004). For the youngest clusters we have
only measured [Mn/Fe] in the individual stars, and find
that it is offset to lower ratios than MW field stars at
12COLUCCI ET AL.
similar metallicities (see black crosses in Figure 11). We
note that McWilliam et al. (2003) and Sbordone et al.
(2007) found a similar pattern of low [Mn/Fe] at high
metallicity in the Sagittarius dwarf galaxy. To explain
the observations in Sagittarius, McWilliam et al. (2003)
and Cescutti et al. (2008) proposed that both metallic-
ity dependent Mn-production in SNe II and SNe Ia as
well as significant gas loss through galactic winds would
be required. The gas-loss required in this case could also
explain the reduced effective star formation rate between
bursts that has been found for the LMC.
In Figure 11 we also show the [Co/Fe] abundances.
From the cluster IL we can measure Co I in one cluster,
NGC 2019. The rest of the Co I abundances that we
measure are obtained from the individual stars in the
intermediate age and young clusters. We find [Co/Fe] to
be approximately solar in our sample. The abundances
of P08, M08, and J06 are consistent with our results.
The last element shown in Figure 11 is Cu.
[Cu/Fe] ratios are significantly sub-solar (∼ −0.9 dex) at
all [Fe/H]. We are able to measure Cu I in the cluster IL
spectra of NGC 1866. All of the other measurements we
obtain for Cu I are from the individual stars in the inter-
mediate age and young clusters. Previous measurements
of [Cu/Fe] by J06 and P08 are also sub-solar at both
high and low metallicity, which is unlike the [Cu/Fe]∼ 0
ratios found in the MW for [Fe/H]> −1. Cu produc-
tion is thought to occur in SNe II, and is metallicity
dependent (e.g. Woosley & Weaver 1995; Bisterzo et al.
2005). As the abundance here is not seen to rise at any
[Fe/H], it appears that SNe II with higher [Fe/H] are not
impacting the overall metallicity of the LMC gas over
time. This could be due to inflows of metal-poor gas,
outflows of metal-rich gas, or low star formation rates
over time with few high mass stars. Low [Cu/Fe] was
also measured for stars in the Sagittarius dwarf galaxy
by McWilliam & Smecker-Hane (2005), where it was ar-
gued that this implied significant gas-loss after early gen-
erations of star formation.
In summary, most of the Fe-peak elements generally
follow the abundance trends that have been previously
measured in the LMC. We find depleted [Cu/Fe] and
[Mn/Fe] at high metallicity, which may reflect the lower
star formation efficiency of the LMC and metallicity-
dependent SNe yields.
5.4. Neutron Capture Elements
The abundance ratios of neutron capture elements
in different environments are particularly useful for
constraining chemical evolution models, especially the
contribution of AGB stars to the interstellar medium
(Sneden et al. 2008). These elements have been observed
to be critically sensitive to the star formation history of
a galaxy and, like α-elements, show different patterns in
dwarf galaxies than in the MW.
Many neutron capture elements have weak features in
individual stellar spectra, and so are particularly difficult
to measure in cluster IL. With EWs, we have been able to
measure Ba II in most of the LMC clusters but can only
measure Y II, Eu II, La II, Nd II, Sm II, and Sr II in 1−3
clusters each. Therefore, the line synthesis component of
ILABUNDs was used to measure or put upper limits on
these elements for most of the sample.
There is some scatter in abundance ratios for the light
Figure 10. Abundances for Fe-peak elements Ni, Cr, Sc and V.
Symbols are the same as in Figure 4.
LMC INTEGRATED LIGHT ABUNDANCES 13
Figure 11. Same as Figure 10 for Co, Mn and Cu. Symbols are
the same as in Figure 4.
s-process elements [Y/Fe] and [Sr/Fe] as shown in Fig-
ure 12 and 13. Sr II was measured in two old clusters; to
our knowledge, these are the first Sr II measurements in
LMC clusters to date. [Sr/Fe] is sub-solar in both cases,
and lower than both the IL MW measurement and the
mean [Sr/Fe] from LMC F-type supergiants (RB89).
From cluster IL, Y II measurements or upper limits
were made for 6 of the 8 clusters. From the individ-
ual stars, we are able to measure both Y I and Y II in
the intermediate age and young clusters. We find some
scatter in [Y/Fe] between clusters, and potentially some
star-to-star scatter within some of the younger clusters,
although our stellar samples in each cluster are small. In
general, at high metallicity and young ages the results
are consistent with solar ratios. We find similar [Y/Fe]
in the LMC clusters as we did for the MW clusters in
In Figure 14, we show our measurements of [Ba/Fe].
[Ba/Fe] is always super-solarin our sample, and the mean
value is higher for the young and intermediate age clus-
ters than it is for the old clusters. The [Ba/Fe] obtained
from individual stars in younger clusters are also high:
[Ba/Fe]∼ +0.9 dex. At low metallicity and older ages,
the [Ba/Fe] we obtain is consistent with previous stellar
measurements in both the MW and the LMC. At high
metallicity and younger ages, our measurements are sim-
ilar to the high [Ba/Fe] measured previously in the LMC
(P08; M08; M11). High values of [Ba/Fe] suggest that
intermediate mass AGB stars had a significant effect on
the chemical evolution of the LMC, which we discuss fur-
ther in § 8.
We measure La II from cluster IL for NGC 2019, NGC
1978, and NGC 1866, as shown in Figure 15. The rest of
our La II measurements are obtained from the individual
stars in the younger clusters. We find [La/Fe] to be ap-
proximately constant for all of the ages and metallicities
in our sample, with a mean value of [La/Fe]∼ +0.5. The
measurements of [La/Fe] of P08 and M08 are consistent
with our results.
As shown in Figure 15, we find more scatter in [Nd/Fe]
than we do for [La/Fe], but the pattern of high ratios for
all ages and metallicities is generally similar. From the
individual stars in the youngest clusters, NGC 1711 and
NGC 2100, there is some indication that the [Nd/Fe]
value decreased at late times in the LMC. High values of
[Nd/Fe] were also found in LMC clusters by J06; M08;
M10, and in LMC field stars by H95. Nd II is obtained
by line synthesis of the IL of four clusters, and by EW
analysis for the individual stars.
Figure 16 shows some of the first Sm II measurements
in the LMC, particularly for old stars. Although the
sample is small, like [La/Fe] and [Nd/Fe], [Sm/Fe] is rel-
atively constant over the full age and metallicity range.
There is little [Sm/Fe] abundance information available
for comparison, but LL92 and RB89 also found Sm to be
overabundant in LMC supergiants by ∼ +0.4 dex when
compared to MW supergiants. Here we find evidence
that this overabundance is also present at low metallici-
ties and old ages.
In Figure 16 we show our [Eu/Fe] abundances. We
find [Eu/Fe] to be consistently enhanced (∼0.5 dex) over
solar ratios for all of the clusters. Eu II abundances are
obtained from line synthesis in both the cluster IL and
the individual stars. These measurements are similar to
what has been found for stars in other LMC clusters
(J06; M08; M10). LL92 also find enhanced [Eu/Fe] with
respect to solar for very young LMC field stars, which
is in contrast to the declining or solar [Eu/Fe] found in
MW stars of similar ages and metallicity.
In § 8 we discuss further implications of the LMC clus-
ter neutron-capture abundances. We have demonstrated
with the LMC cluster sample that many neutron-capture
elements can be analyzed in high resolution IL spectra,
and that this analysis method holds promise for strong
constraints on these elements for distant, unresolved clus-
6. ACCURACY OF LMC ANALYSIS: COMPARISON TO
STELLAR RESULTS FROM THE LITERATURE
In §4 we demonstrated the accuracy of the IL analy-
sis method with our own sample of individual stars in
the youngest LMC clusters. In this section we further
demonstrate the accuracy with abundances from indi-
vidual stars in the literature.
14COLUCCI ET AL.
Figure 12. Same as Figure 4 for the neutron-capture element Y. For consistency with the Y II IL measurements, only the Y II stellar
measurements are shown. Symbols are the same as in Figure 4.
Figure 13. Abundances for the neutron-capture element Sr. The
small red cross shows the mean [Sr/Fe] of LMC F-type supergiants
from RB89. Other symbols are the same as in Figure 4.
NGC 2019 and NGC 2005.
study of the detailed chemical composition in old LMC
cluster stars was presented in J06. This work included
three RGB stars in NGC 2019 and three in NGC 2005.
As discussed in Paper III, we find the [Fe/H] for NGC
2019 is lower by ∼0.3 dex than in J06, and higher for
NGC 2005 by ∼0.25 dex.
As for the rest of the elements, we illustrate a compar-
ison of the [X/Fe] values for the two analyses of NGC
2019 in Figure 17. This comparison is similar to that
described in §4; perfect agreement corresponds to the
solid black 1:1 line, and we identify a small systematic
offset with a linear least squares fit. The uncertainties in
the measurements are shown by the vertical error bars,
which correspond to the error in the mean of the abun-
dance of each element (σ/√Nlines− 1). We note that
the abundances from J06 come from a small sample of
3 stars, which have uncertainties comparable to the IL
uncertainties due to the difficulty in obtaining high S/N
stellar spectra in the LMC. The points in Figure 17 there-
fore have horizontal error bars corresponding to the error
in the mean abundance for each species from J06. We
have removed the light elements Na, Al, and Mg from
the fit because we do not expect them to track the mea-
surements of J06 if star-to-star abundance variations are
present (see § 5.2).
From the dashed line in the top panel of Figure 17, we
see that for NGC 2019, the abundance ratios are very
consistent, and the systematic offsets in neutral and ion-
ized species are small (< 0.1 dex). Therefore, the overall
The first comprehensive
abundance distribution pattern derived for NGC 2019 is
very similar for the two analyses. The outliers in the
comparison are [Si/Fe] and [Y/Fe]. This is probably be-
cause the dispersion in Si between the 2 stars of J06 is
large, and that Y II is very difficult to measure.
Due to lower S/N of our IL spectra of NGC 2005, we
were only able to measure abundances for 10 elements.
As for NGC 2019, we remove the light elements Mg and
Na from the comparison for NGC 2005. When we com-
pare the remaining elements to those of J06, we find
a positive offset for the neutral and ionized species of
∼ +0.3 dex, just as we found for Fe in Paper III. When
this offset is accounted for, the abundance ratios agree
very well, as can be seen in the residual plot of Figure 18.
The exception is [Mn/Fe], which may be systematically
high for all three old clusters (see § 5.3).
While we have already discussed a
detailed comparison of stellar and IL analyses of our
own analyses of NGC 1978 in §4, it is also interesting
to compare our results to the recent work of M08 and
Ferraro et al. (2006), who presented the abundances of
∼20 different elements for a sample of 11 stars.
We compare the [X/Fe] of our IL analysis to the stellar
sample of M08 in Figure 19. We find a small, ∼ −0.1 dex
offset in neutral species, and a large positive offset in ion-
ized species, although the dispersion is large ∼0.4. The
scatter is consistent with what we find from the compar-
ison to our own stellar analysis in §4, and is likely driven
by the unavoidably poorer data quality of the IL spectra
of NGC 1978.
Like NGC 1978, for NGC 1866 we have
already demonstrated good agreement between IL analy-
sis and stellar analysis for our own sample. Here, we also
compare our IL analysis to the stellar sample of M11, as
shown in Figure 20. We find a small ∼ +0.1 dex offset
in the neutral species, and a larger offset of ∼ +0.4 dex
in the ionized species. As for the other clusters in our
sample, when these offsets are removed the abundance
pattern is very similar in both analyses. The remaining
outlier is [Cu/Fe], which is one of the more uncertain
measurements because it is obtained from a single line.
Using medium resolution spectra,
Jasniewicz & Thevenin (1994) estimated Ca, Ti, Cr and
Fe for NGC 2100. Their estimates for [Ca/Fe], [Ti/Fe]
and [Cr/Fe] are consistent with our measurements ob-
tained with high resolution spectra.
In conclusion, as expected, we find minor systematic
LMC INTEGRATED LIGHT ABUNDANCES15 Download full-text
Figure 14. Same as Figure 4 for the neutron-capture element Ba. Symbols are the same as in Figure 4.
Figure 15. Same as Figure 4 for the neutron-capture elements La and Nd. Symbols are the same as in Figure 4.