|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
METALLICITY, DISTANCE, AND DISTRIBUTION OF POPULOUS
CLUSTERS IN THIE LARGE MVAGEtLLANIC CLOUD
AARON J. GROCHOLSKI
Ai DISSERTATION PR i : 11 i TED) TO THE GRADUATE SCHOOL,
OF TH-IE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THIE REQUIIR~I, blt1TS FOR TH-IE DEGREE OF
DOCTOR OF PHILOSOPHY-
UNIVERSITY OF FLORIDA
A~aron J. Grocholski
""Space... it seems to go on and on forever. But then you get to
the end and a gorilla starts throwing barrels at you."
i i i J. Fry
P. :IENOWILEDC. Iill ITS
First and foremost, I would like to thank my adviser, Dr. Ata Sarajedini. Through
incredible patience with my questions (and my forgetting to write down the answers)
and giving me: the <.iil :i to travel to many meetings and'l 1.1 ...|.. he has made:
me the astronomer that I amn today. I feel that this dissertation is as much a reflection
of his abilities as it is of mine (so, hopefully it is good).
In doing the research that went in to this dissertation, I have had the ..1
to work with a number of astronomers. Dr. A~ndrew Cole served as a bit of a second
adviser, teaching me how to process stellar spectra and better understand why they look
the way they do. Drs. Doug Geisler and Verne Smith also contributed greatly to ( i: :
ter 3, providing the rawt~ data as well as many useful .. i .. on the numerous drafts
of that i .:i : For ::i:.:i i :- 4, Dlrs. Knut Olsen and Glenn Tiede offered considerable
input on the: target list, data ?.| i: li..:: and i:i :i : 1il ?ir : Conor Mancone, with his
hard work on the cluster ages, helped to make the results of that chapter considerably
more accurate than they would have been otherwise.
I would also like to thank my dissertation committee (Drs. : c. I~ r II ~F~red
H-amnann, Elizabeth Lada, and Vicki Sarajedini) for .::r;::: up with my many e-mails
and changes in scheduling my defense as i i as reading through / I i i .. I 1. quite
thoroughly) my dissertation.
Joanna Levine... or rather, Dr. Joanna Levine, was an iI i dissertation buddy.
Being able to share the fun and excitement of : i-:: submitting, and i; i=::.1i:::u (only
twno days apart) wMith her made the: whole process a little more bearable. She also
talked my adviser in to sending me with her to Chile on my first real 1 .: : I::. run,
where: I got to see the LMC and SMC in person for the: first time. On that trip, I also
learned from a flight attendant that Joanna and I were living in sin and that they have:
special customs forms for that.
The: future: Dr. I'. 1:i Barker has been a great office late, never complaining
about my .: 7 basic astronomy questions or random inr:" 1.:. ... 1...
I r: -::- -. etc, when I'm lost in thought or annoyed at my data, .. :::1.::r 1. or the -;::i-I-
'L' template. iir doubtful that we will ever be at the same institution .: :r. but
I know that if I ever have any questions about _-.i I or synthetic color-magnitude
diagrams, he is the first person I'll ask.
I would like to not thank Jeff Julian and M~faj. Dr. Steve Novotny, USAF. Without
them (7. 1:. me to lunch at Sonny's for all you can eat ribs so often, I probably
would have finished my dissertation a long time ago. But, the ribs were tasty...
And, finally, I want to thank my parents. In 28 years (and I i: for many
more), I have had nothing but :: -: t from them, whatever path I chose to take. Even
with all the: big words I have learned while writing up my dissertation, I still cannot
find the words to i what this has meant to me. i 1,:- -1 you.
My research ( i: .;: ri'i :l my salary, travel, and publication money) was supported
by: i CAI~iERE grant A i :: --48 to Ata Sarajedini. This dissertation is brought to
you by Thomnpson's Teeth. The only teeth strong enough to eat other teeth.
TABLE OF CONTENTS
ACKNOWLEDGMENTS . . iv
LIST OF TABLES . . viii
LIST OF FIGURES . . ix
ABSTRACT . ........................... . xiii
1 INTRODUCTION . 1
2 K-BAND RED CLUMP MAGNITUDE AS A DISTANCE INDICATOR 7
2.1 Introduction . . 7
2.2 The Data . . 9
2.2.1 Open Clusters .. . . 9
2.2.2 Globular Clusters . . 15
2.3 Results and Discussion . . 17
2.3.1 Cluster Data . . 17
2.3.2 Field Star Data . . . 17
2.3.3 Comparison With Theoretical Models . . 19
2.4 Application As a Distance Indicator . . 23
2.5 Conclusions . .............. ...... . 25
2.6 Beyond Grocholski & Sarajedini (2002) . . 26
3 ABUNDANCES AND VELOCITIES OF A SAMPLE OF LMC CLUSTERS 31
3.1 Introduction . . 31
3.2 D ata . . . . . . ... . 36
3.2.1 Target Selection . . 36
3.2.2 Acqui siti on . . 39
3.2.3 Processing . . 41
3.2.4 Radial Velocities ......... ... 42
3.2.5 Equivalent Widths and Abundances . . 45
3.3 Analysis . . 49
3.3.1 Cluster Membership. . . 50
3.3.2 Cluster Properties . . 56
126.96.36.199 Metallicities . . 57
188.8.131.52 Kinematics . . 62
3.4 Comparison with Previous Work . . 63
3.5 Summary . .................. . 71
3.6 Notes on Individual Clusters . . 73
3.6.1 NGC 1718 . . 73
3.6.2 NGC 1846. .. .... ......_ 75
3.6.3 NGC 1861. . . 75
4 DISTANCES AND DISTRIBUTION OF POPULOUS LMC CLUSTERS 149
4.1 Introduction . . 149
4.2 D ata . . . . . . ... . .152
4.2.1 Observations . . 152
4.2.2 Reduction . . 153
4.2.3 Photometry . . 155
4.3 Cluster Ages and Abundances . . 158
4.4 Apparent and Absolute K-band RC Magnitudes . . 161
4.5 Cluster Distances and the Distance to the LMC . . 167
4.5.1 Absolute Distance Moduli . . 167
4.5.2 LMC Cluster Distribution . . 167
4.5.3 The Distance to the LMC Center . . 173
4.5.4 Systematic Errors . . 175
4.6 Comparison to Previous Distances . . 176
4.7 Summary . . 177
5 SUMMARY . ......................... .. .180
REFERENCES . . 186
BIOGRAPHICAL SKETCH . . 193
LIST OF TABLES
2-1 Open and Globular Cluster Information .....
2-2 LMC Cluster Information ....
3-1 LMC Target Cluster Information .....
3-2 CaT Line and Continuum Bandpasses .....
3-3 Derived LMC Cluster Properties .....
3-4 Published LMC Cluster Metallicities ....
3-5 Metallicities of Young and Intermediate-Age Stellar Populations .
3-6 Positions and Measured Values for Cluster Members ....
3-7 Positions and Measured Values for Field Stars ....
4-1 Exposure Times at Each Dither Point ....
4-2 LMC Cluster Sample Information ....
4-3 LMC Cluster Ages and Metallicities .....
4-4 Calculated Red Clump Values and Cluster Distances .....
4-5 LMC Globular Cluster Information ....
4-6 LMC Center Distances .....
4-7 Effect of LMC Geometry ....
LIST OF FIGURES
2-1 Comparison of open cluster ages ....
2-2 Near-IR open cluster CMDs ....
2-3 Near-IR globular cluster CMDs ....
2-4 Comparison of MRC and M,RC ....
2-5 Effects of age on MRC ....
2-6 Effects of [Fe/H] on MRIC ....
2-7 Age effects on solar neighborhood RC stars .
2-8 Intrinsic red clump color .....
2-9 Near-IR CMDs for Hodge 4 and NGC 1651
3-1 Schematic diagram of the LMC ....
3-2 Sample of spectra from RGB stars in our targ~
3-3 The xy positions of our target stars in the Hod
3-4 Radial velocities for our spectroscopic targets
3-5 Hodge 11 target star metallicities ....
3-6 CW vs. V VHB for Hodge 11 .....
3-7 CMD for the entire Hodge 11 field ....
3-8 Positions on the sky and derived metallicities :
3-9 Cluster metallcity vs. position angle .....
3-10 Cluster metallicity vs radial distance ....
3-11 Cluster radial velocity vs. position angle ..
3-12 Metallicity comparison with OSSH ....
3-13 Metallicity distribution of LMC clusters ...
et clusters . . 43
ge 11 field . . 51
in Hodge 11 . . 52
for our target clusters . 59
3-14 Cluster CMD for NGC 1718 .....
3-15 IC 2146 cluster member selection ...
3-16 IC 2146 cluster and field CMD ....
3-17 NGC 1651 cluster member selection .
. . 101
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
cluster member selection .
cluster and field CMD ...
. . 105
. . 107
. . 109
. . 111
. . 113
. . 114
. . 115
. .. 127
. .. 129
-51 Reticulum cluster member selection ..
-52 Reticulum cluster and field CMD ...
-53 SL 396 cluster member selection ...
-54 SL 396 cluster and field CMD .....
-55 SL 41 cluster member selection ....
-56 SL 41 cluster and field CMD ....
-57 SL 4 cluster member selection .....
-58 SL 4 cluster and field CMD ....
-59 Hodge 4 cluster member selection ...
-60 Hodge 4 cluster and field CMD ....
-61 Hodge 3 cluster member selection ...
-62 Hodge 3 cluster and field CMD ....
-63 SL 61 cluster member selection ....
-64 SL 61 cluster and field CMD ....
-65 SL 663 cluster member selection ...
-66 SL 663 cluster and field CMD .....
-67 SL 869 cluster member selection ...
3-68 SL 869 cluster and field CMD . . 130
4-1 K'-band images for all target clusters . . 155
4-2 Optical photometry for NGC 1651 and NGC 2173 . .. ..161
4-3 Near-infrared CMDs . ...... .. .. 164
4-4 Schematic diagram . ...... ... 69
4-5 Cluster distances as a function of their position .... .. .. 170
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of ii :: i ..- ii:y
M\IETAILLICi i, DISTANCE, AND DISTRIBUTION OF POPULOUS
CLUSTERS IN TH(E ILARG~E M~AGELLANIC CLOUD
Alaron J. Grocholski
Chair: Ata :.:
Major Department: Astronomy
The ILarge i' il.: 1 1 : Cloud ii li' i: ), with its proximity to the li' i y '.;..1, and
location well awvay from the Galactic }.7 :::- offers us an excellent ;- ..: i:: to study
the < ? i. : of tidal interactions on the formation and evolution of :il i populations in
a satellite galaxy. In this work we i:: :::1 results from a program aimed at determining
the ages, kinematics, metallicities, distances and spatial distribution of populous
clusters in the LMC.
To further our understanding of the LMCI cluster system, wet have: acquired near-
infrared; 1e i )i photometry and spectroscopy and combined these data with optical
and NIR photomnetry from the: literature. WVith FORS2 on the: VLT, wne obtained NIR
spectra for more than 800 stars in and around 29 LMC clusters that span a large
range of ages (~ 1-13 Gyr) and metallicities (--0.3 2 T Ir rr -2.0o). Wie use:
these spectra to calculate r0 i1 :: velocities and metallicities, and identify more than
members in 28 clusters. U I:--.published id:..(. ::: ;:y, VLT FORS2 images, and
archival HST WJ\FPC2 images, we compiled deep optical photometry for 15 clusters.
These data extend below each cluster's main : i::: :: turnoff and, combined with
our abundances, l l. .. us to break the well known age. : .< : i .. : degeneracy and
determine ..... ..1 .>cal ages via main sequence -i.::::, (:"l .12). As the first step in our
i .:: calculations, we use JK~, photometry of 141 Galactic open clusters from the
-.1 rlSS to calibrate: the K-band luminosity of core helium burning red clumnp stars
: ) as a function of age and metallicity. Next, with ISP"I on the : i ii:0 4m telescope,
we imaged 17 ILMC clusters in the NIR (JK') down to K' ~ 18._5, or about 1.5
belowv the RiC, allowing us to measure the apparent K-band RiC magnitude of each
cluster. WZe combine the LMC cluster ages and abundances with our RC calibration to
predict ': for each cluster and thereby calculate accurate distances and i I.T .-e the
geometry of the cluster system.
velocities are in good argeemetnt wMith previous results and showv that
the ILMCL clusters rotate with diski-like kiinematics. Our abundances indicate that the
: 11. :j i: :::i.:: .. :: of the: more: metal-rich clusters is much tighter than i : .:: 7y
believed, with no tail toward solar ::: -: :i i: :i:: The peak of this distribution is similar
to that of the bar, which is in good agreement wiith dynamical models that :: that
both the restart of cluster formation and the formation of the bar occurred as a result of
the first close encounter with the SMCI (~ 4 Gyr ago). Cluster ages derived from MISF
range from ~1-3 Gyr for all clusters in our sample except ESO 121-St: (~ 9 Gyr),
the only known cluster in the LMC: with an age between ~3-13 Gyr. The intermediate
age range, as i i as the age of ESO 121-SCO3, is similar to previous results.
Finally, we find that the .i:r ,7 distribution of the LMIC: cluster system is in
good agreement with thick, inclined disk geometry found from LMC field stars. In
.:iii :~ using RRE L~yrae based distances, we find that the old globular clusters are
also consistent wiith this geometry. Given the: .1: i -like kinemnatics of the entire cluster
system, this implies that the :' i: 's 1: i formed at about the same time as the old
clusters, ~ 13 Gyr ago. Combining the LMC geometry with cluster ii: : r:::. we:
calculate a distance to the center of the ii li' ii C of (m2 :'j, ')0 18.40 + 0.04 f 0.08, which
is 0.1 mag shorter than the commonly accepted ILMCI distance.
The current scenario of hierarchical formation suggests that large galaxies like the
Milky Way (MW), and in particular their spheroid components (bulge and halo), are
built up through the accretion of dwarf satellite galaxies and protogalactic fragments
(e.g., C~te: et al. 2000; Zentner & Bullock 2003). It is clear from an observational
standpoint that the chemical enrichment and star formation histories of galaxies in
general, whether large or small, at high or low redshift, are dominated by interactions
and merger events (Abraham 1999; Schweizer 1999). Thus, understanding the effects
of gravitational interactions on the evolution of satellite galaxies is an important piece
in the puzzle of large galaxy formation.
The MW and its satellite galaxies are a prime example of interactions and mergers
in action, with signatures of the many stages of hierarchical formation visible in and
around the Galaxy. The most striking evidence of this is the cannibalization of the
Sagittarius dwarf galaxy (Sgr, Ibata et al. 1994). Lying ~ 25 kpc from us and almost
directly behind the Galactic center, Sgr is elongated along its orbit around the MW
as a result of tidal stripping. Also visible throughout the MW are many tidal streams
(e.g., Majewski et al. 2006), all of which are remnants of dwarf galaxies that were
long ago disrupted by and accreted into our Galaxy. Even halo globular clusters that
have large Galactocentric radii (X 8 kpc) likely formed in satellite galaxies before
being absorbed by the MW (Searle & Zinn 1978). While signatures of past accretions
abound in the Galaxy, the fact that these systems were disrupted well in the past
makes it difficult if not impossible to uncover the formation history of their parent
galaxies. Similarly, the location of Sgr almost directly behind the Galactic bulge
causes widespread contamination by foreground stars, rendering observations of this
galaxy difficult to interpret. In contrast, the Large Magellanic Cloud (LMC) is a
nearby galaxy that shows signatures of interactions, but remains mostly intact, and is
relatively free from foreground contamination. Since its proximity allows us to resolve
individual stars and determine their physical properties, the LMC offers us an excellent
environment in which to study the effects of tidal forces on the formation history of a
Using gas dynamical N-body simulations, Bekki et al. (2004) model the orbits of
the LMC and Small Magellanic Cloud (SMC) in their paths around the MW and pay
special attention to the effects of tidal interactions between the Clouds. Prior to 5 Gyr
ago, the LMC orbited the MW every ~ 2 Gyr with a highly eccentric orbit that ranged
in Galactocentric radius from ~ 50-150 kpc. Originally, the SMC orbited the MW on
a similar path to, but independent from, the LMC. Its smaller Galactocentric radius (~
50-100 kpc) resulted in a faster orbit, with the SMC completing one revolution around
the MW every 1.5 Gyr. Approximately 5 Gyr ago, the LMC and SMC passed within
25 kpc of each other, an encounter that caused a small decay in the LMC's orbit and
an increase in the radius of the SMC's orbit, which ultimately led to the Magellanic
Clouds becoming bound ~ 1 Gyr later. Since becoming bound, the LMC and SMC
have had a number of close encounters (g 10 kpc, see Fig. 1 in Bekki et al. 2004),
with their first close encounter (6.4 kpc, 3.6 Gyr ago) having a significant effect on the
star formation history of the LMC (see below).
Quite possibly, the most remarkable signature of interactions in the LMC-SMC-
MW system is seen at radio wavelengths; column density maps of H I show a complex
envelope of gas in and around the Magellanic Clouds. The LMC and SMC are
connected by a substantial bridge of material that was likely stripped from the SMC in
a previous encounter. In addition, both the short Leading Arm and longer Magellanic
Stream, which trails the Magellanic Clouds in their orbit around the MW and stretches
~ 100" across the sky, are the result of tidal stripping of material from the Magellanic
Clouds by the MW (Putman et al. 2003).
More subtle, but just as revealing, are the markers of interactions found in the
LMC's stellar populations. Studying carbon star kinematics, Alves & Nelson (2000)
showed that the disk of the LMC is flared, with the disk scale height increasing from
0.3 kpc at a R = 0.5 kpc to 1.6 kpc at R = 5.6 kpc. Using an expanded sample
of carbon stars, van der Marel et al. (2002) find similar results and also show that
the LMC's disk, with V/o a 2.9 + 0.9, is much thicker than the MW thin disk
(V/o a 9.8) and slightly thicker than the MW thick disk (V/o a 3.9). In addition,
both Olsen & Salyk (2002) and Nikolaev et al. (2004) use field stars as relative
distance indicators to show that the LMC disk may also be warped. These results are
in agreement with the N-body simulations by Weinberg (2000) which predict that the
Galaxy is a significant driver of the LMC's evolution and that tidal forces from the
MW will heat (thicken) and possibly warp the disk of the LMC.
However, the most impressive feature of the LMC that is likely linked to gravita-
tional forces is its cluster formation history. The LMC is known to have a population
of old, metal-poor globular clusters that formed ~ 13 Gyr ago and a more recent epoch
of intermediate metallicity cluster formation that began ~ 3 Gyr ago and has continued
to the present. In between these two epochs is the well known "age gap" in which
only one cluster, ESO 121-SCO3 (ESO 121; 9 Gyr), is known to reside (e.g., Da Costa
1991; Geisler et al. 1997; Da Costa 2002). Similar to the intermediate-age clus-
ters, stars in the LMC bar formed only within the past ~ 5 Gyr (Cole et al. 2005),
whereas star formation histories for the disk of the LMC show that field stars had a
constant, although low, star formation rate during the cluster age gap (Holtzman et al.
1999; Smecker-Hane et al. 2002). The aforementioned model of Bekki et al. (2004)
shows that, while the cause of the apparent end of cluster formation ~ 13 Gyr ago
is unknown, the first very close encounter between the LMC and SMC ~ 4 Gyr ago
would have caused the formation of the LMC bar as well as "dramatic gas cloud
collisions" that resulted in the restart of cluster formation in the LMC; continued
strong interactions have sustained the LMC's cluster formation. During the age gap,
weak interactions between LMC, SMC, and MW would have only been sufficient to
support star formation in the field. Thus, the star formation history of the LMC, and,
in particular, the star formation that has occurred in within the last ~ 4 Gyr, is a direct
result of tidal forces acting on the LMC.
As mentioned above, tracers of the LMC's field populations show that the spatial
distribution of the field stars has been affected by interactions; in contrast, the 3-
dimensional distribution of clusters has not been fully explored. Typically, the LMC
is treated as a planar galaxy that can be assumed to lie at a single distance from us.
However, the proximity of the LMC (~ 48 kpc) combined with the fact that it is
inclined appreciably with respect to the plane of the sky (e.g., Caldwell & Coulson
1986) leads to a significant distance gradient across the face of the LMC. Recently,
both van der Marel & Cioni (2001) and Olsen & Salyk (2002) have used field stars
to trace the geometry of the LMC, where they have assumed that variations in the
brightness of their observed fields were due to differences in distance (see Chapter 4
for more detail). Using this method, both authors find that the LMC field populations
lie in a disk that is inclined ~ 35" (0" is face on) with the northeast portion of the
LMC closer to the MW than the southwest. For stars lying 5 kpc from the center of
the LMC, this inclination can lead to difference in distance from the MW of ~ 8 kpc.
While relative distances have been used to show that the LMC field stars lie in a thick,
inclined disk, the distribution of the clusters has only been inferred from kinematics.
Schommer et al. (1992) calculated velocities for ~ 80 populous clusters and found that
the entire cluster system rotates with disk-like kinematics, while no clusters appear to
reside in a pressure supported halo. We note that, although both the cluster system and
the majority of field stars reside in the disk of the LMC, there does seem to exist a
tenuous halo of metal-poor RR Lyrae stars (Borissova et al. 2004).
Star clusters are an important tool for studying the structure and formation
history of a galaxy because they have one main advantage over field stars. Whereas
a sample of field stars may cover a wide range of ages, all stars in a given cluster
can be considered to be coeval, allowing the cluster's age to be readily determined
from deep color-magnitude diagrams. Thus, clusters place a much needed time-stamp
on a variety of events in the history of a galaxy. For example, since clusters contain
a record of their host galaxy's chemical abundances at the time of their formation,
they permit us to place tight constraints on the age-metallicity relation of the galaxy.
This is particularly important when we consider that tidal forces can result in a
galaxy experiencing infall or outflow of material, which may leave markers of these
interactions on the galaxy's age-metallicity relation. In addition, given knowledge of
the kinematics and distribution of the cluster system, it may be possible to determine
the timescale of formation of features, such as the disk of the galaxy. Finally, for
many standard candles, their absolute brightness varies as a function of both age and
metallicity. Due to the fact that the ages of field stars are difficult to determine, it is
usually only possible to calculate relative distances from field populations. Clusters, on
the other hand, with their available ages and abundances, enable us to properly apply a
standard candle and thereby calculate accurate absolute cluster distances. Therefore, we
can use a sample of clusters not only to explore their spatial distribution, but we can
combine this distribution with their absolute distances to determine an accurate distance
to their host galaxy. For the LMC, an accurate distance is of particular importance
due to its use as the zeropoint in the extragalactic distance scale (e.g., Freedman et al.
In this dissertation, we present the results from a program designed to better
understand the ages, kinematics, abundances, distances, and spatial distribution
of populous clusters in the LMC. First, in Chapter 2, we have developed the core
helium burning red clump (RC) stars as a standard candle. Specifically, we calibrated
the absolute K-band magnitude of the red clump as a function of cluster age and
metallicity for a sample of Galactic open clusters. Next, in Chapter 3, we obtained
moderate-resolution near-infrared spectra for a large number of stars in and around a
sample of LMC clusters. With these data we were able to identify cluster members
and subsequently determine cluster abundances and velocities. In Chapter 4, we have
acquired near-infrared (JK) images for a number of intermediate-age LMC clusters
and used the resulting photometry to measure the apparent K-band RC magnitude
for these clusters. We combine newly calculated ages from deep optical photometry
(see ~4.3) with the abundances from Chapter 3 and the RC calibration presented in
Chapter 2 to predict the absolute K-band magnitude of the RC for our LMC clusters.
Absolute cluster distances are then readily calculated and the spatial distribution of the
cluster system is explored, in addition to determining the absolute distance to the LMC.
Finally, in Chapter 5, we summarize our results.
K-BAND RED CLUMP MAGNITUDE AS A DISTANCE INDICATOR
During the past few years, the helium burning red clump (RC) has gained
considerable attention for its potential as a standard candle. The primary advantage
of the RC is the ease with which it can be recognized in the color-magnitude diagram
(CMD). However, there is currently a great deal of controversy in the literature
regarding the appropriate treatment of possible metallicity and age effects on the I-band
absolute magnitude of the RC ( 0C). There are two schools of thought on this issue.
The first assumes a constant value for 0R which is then used to facilitate a single-step
distance determination via knowledge of the apparent RC magnitude and the extinction
(e.g., Paczyliski & Stanek 1998; Stanek & Garnavich 1998) The second approach is
founded on the claim that both age and metal abundance have a significant influence on
the luminosity of RC stars (e.g., Cole 1998; Sarajedini 1999, hereafter, S99) and must
be accounted for in determining 0R and therefore the distance.
Both Paczyliski & Stanek (1998) and Stanek & Garnavich (1998) use Hipparcos
RC stars with parallax errors of less than 10% to calculate the I-band absolute magni-
tude of the solar neighborhood red clump. In their analysis, Paczyliski & Stanek (1998)
find that ~IC shows no variation with color over the range 0.8 < (V -I)o < 1.4 and,
from a Gaussian fit to the RC luminosity function, find ~IC = -0.28 + 0.09. Follow-
ing the same methodology and building upon the earlier work, Stanek & Garnavich
(1998) find a similar result with MRC = -0.23 + 0.03. With this calibration, a sin-
gle step calculation is then used to determine the distance to the Galactic center
(Paczyniski & Stanek 1998) and M31 (Stanek & Garnavich 1998). Both of these inves-
tigations found little or no variation in Myr of the RC stars with color; this was taken to
imply that MRC does not vary significantly with metallicity.
In contrast, theoretical models from Girardi & Salaris (2001) and the earlier
models of Seidel, Demarque, & Weinberg (1987, see also Cole 1998) show that MRC
is dependent on both age and metallicity, becoming fainter as both increase. These
models are in good agreement with the observations presented by S99. Using published
photometry for eight open clusters, S99's most important result is that while MRC iS
less sensitive to metal abundance than MR~C, both still retain a considerable dependence
on the age and metallicity of the stellar population. As a result, the single-step method
of applying the solar-neighborhood MRC to populations with a different age-metallicity
mix could be problematic.
Alves (2000) also uses the Hipparcos RC for his calibration; however, he relies
upon the K-band luminosity (MrK) of RC stars in the hope that, since the K-band is less
sensitive to extinction (and possibly metallicity as well) than the I-band, it might make
a better choice as a standard candle. Alves (2000) restricts his RC stars to those that
have metallicities from high resolution spectroscopic data. For this group of 238 RC
stars, he finds a peak value ofMRC = 1.61 +0.03 with no correlation between [Fe/H]
and Mr. However, he is not able to explore the effect of age on MRC due to the lack
of such information for the individual stars in his sample.
These previous works prompted us to combine the approaches of S99 and Alves
(2000) to investigate the influence of age and metal abundance on MRC for a number of
open clusters with well-known distances and metallicities. Our findings were originally
published in Grocholski & Sarajedini (2002, hereafter, GSO2), however, the results
relied on photometry from the Second Incremental Data Release of the Two Micron
All Sky Survey (2MASS). Since the original publication of our paper, the All Sky Data
Release of the 2MASS catalog has been made publicly available and, in this chapter,
we have updated the work presented GSO2, based on the newest 2MASS photometry.
Additionally, in @ 2.6, we have updated our initial application of the RC calibration to
the LMC (Sarajedini et al. 2002). We note that the improved photometry from the All
Sky Release of 2MASS has had very little effect on the calculations in either GSO2 or
Sarajedini et al. (2002) and has not changed any of the conclusions in these papers.
In @ 2.2 we discuss the observational data used in calibrating the K-band lumi-
nosity of the red clump. Section 2.3 compares our data with the results of theoretical
models and presents a discussion of the results. We test the utility of our results for
calculating the distance to a Galactic open cluster in @ 2.4 and our conclusions are
summarized in @ 2.5. Finally, in @ 2.6, we discuss the work of Sarajedini et al. (2002),
which sought to apply our RC calibration to determining the distance to a pair of LMC
2.2 The Data
2.2.1 Open Clusters
In the present study, the most important criterion that the observational data
must fulfill is that of internal consistency. For example, we must ensure that the
distance moduli, reddenings, ages, and metallicities of all of the clusters in our sample
have been determined using the same techniques. In addition, it is imperative that
the infrared photometry we rely upon be measured and calibrated in a consistent
manner. For the former, we use the database of open cluster properties as measured by
Twarog, Ashman, & Anthony-Twarog (1997), supplemented by cluster ages from the
WEBDA database, and for the latter, we utilize the All Sky Release of the Two Micron
All Sky Survey (2MASS) Point Source Catalog. We now discuss each of these in more
Twarog et al. (1997), have compiled a list of 76 open clusters for which they
provide reddenings, distance moduli, and metallicities. For the purposes of the
present paper, we limit ourselves to distance moduli derived via the technique of
main sequence fitting (MSF) so as to remain independent of methods that rely on the
luminosity of the RC. Their metallicities have all been measured on the same system
and the reddenings have been determined using an internally consistent method. The
vast majority of these values are consistent with those found in the literature, except for
the reddening of NGC 6819 for which the Twarog et al. (1997) value is much higher
than other published values. As a result, we have decided to adopt the S99 reddening
for NGC 6819 instead of the apparently discrepant value tabulated by Twarog et al.
(1997). In addition, because the determination of the reddening and distance modulus
is coupled, we also adopt the S99 distance modulus for NGC 6819.
The ages of the open clusters have been obtained from WEBDA, which is a
compilation of open cluster data from various sources. To check the reliability of the
WEBDA ages, we compare them with the cluster ages determined by S99 in Fig. 2-1.
S99 presents isochrone-fitting ages for eight open clusters that have been determined
in a consistent manner using the Bertelli et al. (1994) theoretical isochrones. The left
panel of Fig. 2-1 plots the ages from S99 versus the ages in WEBDA, where both
axes are in log space and the dashed line represents a zero age difference between
the systems. From this plot, it is evident that there is a systematic offset between the
two systems with the WEBDA ages being younger than those of S99. The average
difference, Alog (Age) = 0.191, is used to shift the ages given in WEBDA onto the
S99 system. The right panel of Fig. 2-1 shows the ages from S99 plotted against
the shifted WEBDA ages; it is clear from Fig. 2-1 that the shifted ages are in better
agreement with those of S99; as a result, we will apply this shift to nine of the clusters
in our study and use the S99 ages for the five clusters that are common to both studies.
For the open clusters in the Twarog et al. (1997) study that possess MSF dis-
tances, we extracted JHKs photometry from the All Sky Release of the 2MASS
Point Source Catalog. As noted above, these data have been obtained using similar
instruments and reduced with the same pipeline techniques. For each cluster, we have
10.0 *~ / *
8,8 9.0 9.2 9.4 9.6 9.8 10.0 8.8 9.0 9.2 9,4 9.6 9.8 10.0
Log age (WEBDA ages) Log age (Shifted WEBDA ages)
Figure 2-1: Comparison of open cluster ages. Ages from WEBDA (left panel) plotted
against those from S99. Due to the systematic difference between the sys-
tems, we shift the WEBDA ages older by Alog (Age) = 0.191 (right panel)
to place them on the same system as S99.
utilized the same criteria for the 2MASS data retrieval. The field size is originally
set to 30' in radius and then reduced to fields as small as 5' in radius in an attempt to
isolate the cluster stars. The sources are limited to a brightness of 6th magnitude or
fainter due to saturation effects at the bright end (Carpenter 2000). Lastly, we have
extracted only the highest quality photometry from the 2MASS catalog, which provides
a read flag (rd~fg) indicating how the photometry of each star was measured. We have
chosen to exclude any source that has a read flag of zero in any band since this implies
that the source was not detected in that band and the magnitude given is an upper limit.
We note that the vast majority of the stellar magnitudes used in this study are based on
point-spread-function fitting (i.e., rd~fig = 2); however, in order to include the brighter
red clumps of more nearby clusters, we have had to use the aperture photometry in a
minority of cases.
The 2MASS program uses a K-short (Ks) filter for their observations. We have
chosen to convert these magnitudes to the K-band adopting the Bessell & Brett (1988,
hereafter, BB) system, which is also used in the theoretical models of Girardi et al.
(2000) and Girardi & Salaris (2001). The transformation equations are derived by
Carpenter (2001) and are adopted as follows:
(J -KBB) = [(J- Ks) (-0.01 1 +0.005)]/(0.972 +0.006) (2-1)
KBB = [Ks (-0.044 +0.003)] (0.000 +0.005) (J- KBB). (2-2)
We note, however, that transformation to the Koornneef (1983) K-band (used
by Alves 2000; @ 2.3.2) would have a negligible effect on our results. To cor-
rect for the interstellar reddening, we adopt the extinction law determined by
Cardelli, Clayton, & Mathis (1989), which, using their value of Rv = 3.1, gives
AK = 0.11Av and AJ = 0.28Av. From this, it is a simple matter to calculate the
absolute K-band magnitude and dereddened J-K color of the open cluster stars.
We have determined the RC luminosity for our clusters by taking the median
value ofM~K for all stars within a standard sized box placed around the RC. We use
the median value of MK along with a constant box size in an attempt to eliminate any
selection effects that may occur in choosing the location of the RC and to limit the
effect of outliers on My~. Fig. 2-2 shows the CMDs for all 14 clusters, focused on the
RC and main sequence turnoff. The box used to select the RC stars for each cluster
is also shown. We note that, where available, we have used published optical CMDs
for our clusters to help isolate the approximate RC location. The uncertainty in Mf is
calculated by combining the standard error about the mean K-magnitude for all stars
inside the RC boxes along with the errors in E(B V) and (m M),, all added in
quadrature. Except where otherwise noted, we adopt 20% of the value as the error in
E(B V) and 10% of the value as the error in (m -M),.
).0 0.5 1.0
Figure 2--2: Near-IR open cluster CMI~s. Infrared CM13s for the 14 open clusters in
our sample are shown, with a box indicating the location of the cluster's
red 7:r:::.;. All stars within the box are: used in calculating the median K
::: ,l-, of the red 1
o o o
40 Co a oc oo
D o -
o M oo
2~O 3 co a .
0. 0.5 1. .
(J K (J-K)
Figr 2-2 Nea-I opncutrCo otne
I I I 1
47 Tuc li,' NGC 362 O
OO O vO OO
Figure 2-2: Near-IR open cluster CMDs Continued.
0.0 0.5 1.0
Near-IR globular cluster CMDs.
clusters in our sample.
Same as Fig.
2-2, but for the globular
2.2.2 Globular Clusters
It is difficult to ensure that the distances, ages, and metallicities of globular clus-
ters with RCs are on the same system as those of the open clusters. Fortunately, the
ages and metallicities of the two globulars in our sample 47 Tuc and NGC 362 are
sufficiently different from the bulk of the open clusters that small systematic discrep-
ancies in these quantities should not be a significant hindrance to the interpretation of
the results. In any case, we have decided to adopt literature values for the basic cluster
parameters and use the globular cluster RCs as a consistency check.
In the case of 47 Tuc, we adopt the metallicity quoted by Carretta & Gratton
(1997) of [Fe/H] = -0.70 + 0.07, which happens to be very close to the Zinn & West
(1984) value. For the distance modulus and reddening, we average the published
values listed in Table 2 of Zoccali et al. (2001) to obtain (m -M); = 13.45 +0.21 and
E(B V) = 0.044 & 0.008, where the errors represent half of the range of tabulated
values. Lastly, for the age of 47 Tuc, we adopt the oldest age for which the models
predict the presence of a RC at its metallicity 12 Gyr.
For NGC 362, we adopt a similar approach. The metal abundance of [Fe/H] =
-1.15 + 0.06 is taken from Carretta & Gratton (1997), which is approximately 0. 1
dex more metal-rich than the Zinn & West (1984) value. Our search of the literature
has revealed distance moduli that range from (m -M);- = 14.49 (Zinn 1985) to 14.95
(Burki & Meylan 1986, see also Bolte 1987) and reddenings in the range E(B V)
= 0.032 (VandenBerg 2000) to 0.08 (Alcaino 1976) leading to adopted values of
14.70 + 0.23 and 0.048 + 0.024 for the apparent distance modulus and reddening of
NGC 362, respectively. Once again, we adopt an age of 12 Gyr.
The RCs of these globulars have been isolated in the 2MASS point source catalog
in the same way as for the open clusters. The [M~K, (J- K)o] CMDs for 47 Tuc and
NGC 362 are shown in Figure 2-3 along with the box used to define their RCs. All
of the relevant observational parameters for the open and globular clusters are listed in
Table 2-1. Open and Globular Cluster Information
Name Log Age (m-M~)Va E(B-F)a [Fe/H]a O([Fe/H])a h1g o(Mr() (J-K)o o(J-K)o
NGC 752 9.24 8.35 0.04 -0.088 0.018 -1.538 0.118 0.603 0.011
NGC 1817 8.80 12.15 0.26 -0.268 0.023 -1.875 0.181 0.565 0.028
NGC 2099 8.73 11.55 0.27 0.089 0.073 -2.111 0.185 0.555 0.029
NGC 2204 9.28b 13.30 0.08 -0.338 0.120 -1.607 0.114 0.612 0.010
Be 39 9. 88b 13.50 0.11 -0.177 0.032 -1.595 0.122 0.677 0.013
NGC 2360 8.94 10.35 0.09 -0.150 0.026 -1.177 0.120 0.604 0.012
NGC 2420 9.24 12.10 0.05 -0.266 0.017 -1.690 0.115 0.617 0.009
NGC 2477 9.04 11.55 0.23 0.019 0.047 -1.436 0.163 0.597 0.024
NGC 2506 9.24 12.60 0.05 -0.376 0.029 -1.573 0.107 0.619 0.007
NGC 2527 8.84 9.30 0.09 -0.080 0.090 -1.700 0.125 0.554 0.014
NGC 2539 8.76 10.75 0.09 0.137 0.028 -1.564 0.125 0.540 0.016
M 67 9.60b 9.80 0.04 0.000 0.092 -1.687 0.106 0.668 0.011
NGC 6791 9.98b 13.40 0.15 0.150 0.041 -1.422 0.133 0.687 0.016
NGC 6819 9.42b 12.44b 0.16b 0.074 0.035 -1.658 0.136 0.648 0.017
47 Tue 10.08 13.45C 0.044C -0.70d 0.07d -1.340 0.211 0.538 0.016
NGC 362 10.08 14.70" 0.048C -1.15d 0.06d -0.831 0.240 0.440 0.029
aFrom Twarog et al. (1997) unless otherwise noted
bFrom Sarajedini (1999)
CSee 6 2.2.2
dFrom Carretta & Gratton (1997)
2.3 Results and Discussion
2.3.1 Cluster Data
As described in @ 2.1, we are interested in exploring the dependence of Me on
[Fe/H] and age. Plotted in Figure 2-4 are the Mye values for the 14 open clusters
(open cirtles) and the two globulars (filled cirtles) in our sample versus the logarithm
of the age (top right panel) and the metallicity (top left panel). The red clumps start
out very bright at young ages and decrease in brightness by almost 1 mag as the cluster
ages approach 109 yr after which they brighten by ~ 0.5 mag. Then, the RCs become
slightly fainter as the cluster ages increase up to 1010 yr. The top two panels of Figure
2-4 also include MRC (open squares) from S99. Keeping in mind that the numbers of
clusters is small, over the age and metallicity range common to both studies, the K-
band absolute magnitude of the RC exhibits less sensitivity to age and metal abundance
than does MRq
2.3.2 Field Star Data
Alves (2000) reports K-band absolute magnitudes of solar-neighborhood stars
in the Hipparcos catalog along with parallaxes and proper motions. Following the
I I I I I I I I I I I
MK (This Paper)
MK (This Paper)
o Alves (2000)
e This paper
Figure 2-4: Comparison ofM c and MfRc. Upper panels show the variation of the
RC absolute magnitude as a function of [Fe/H] (top left panel) and age
(top right panel). The open circles represent K-band absolute magnitudes
(Mc) for the 14 open clusters while the filled circles signify Mc values
for the two globular clusters in the present sample. The open squares des-
ignate Mf~c values from S99. In the bottom panel, MK for Hipparcos Solar
neighborhood red clump stars from Alves (2000) (open circles) are com-
pared with Mrc for clusters in the present paper (filled cirles). These two
data sets show remarkable agreement in their mean K-band magnitudes.
analysis of Alves (2000), we have limited ourselves to stars with 2.2 < (V K)o < 2.5
and -2.5 < 2W < -0.8 in Table 1 of Alves' paper in order to isolate a sample of
nearby RC stars. We have plotted 2& versus [Fe/H] for these stars in the bottom panel
of Figure 2-4 (open circles). For comparison, the open and globular cluster data from
the present work (filled circles) are also shown. Keeping in mind that there are far
fewer open clusters than field stars in Figure 2-4, we find good consistency in the
locations of the two samples. Alves (2000) finds (2W(RC))= -1.61 +0.03, while the
open clusters in our sample give (2W(RC))= -1.61 +0.06. This is remarkable given
the fact that the open cluster distances are based on the main sequence fitting results
of Twarog et al. (1997) and the field stars are on the Hipparcos distance scale. Both
show mean K-band magnitudes of ~-1.6 and very little if any dependence on metal
abundance over the same range.
2.3.3 Comparison With Theoretical Models
Leo Girardi has kindly provided us with theoretical models that represent the
median magnitude of the red clump as a function of age and metal abundance
(Girardi & Salaris 2001; Crowl et al. 2001). Figures 2-5 and 2-6 show these the-
oretical models in the K-band compared with our open and globular cluster data. The
nine panels of Figure 2-5 display the K-band luminosity of the red clump as a function
of metallicity for a range of ages. The five panels in Figure 2-6 illustrate the variation
of the K-band luminosity with age for a range of metal abundances.
In both of these figures, clusters that are similar in metallicity or age to the model
plotted in each panel are represented by filled circles while the remaining clusters are
denoted by open circles. Figures 2-5 and 2-6 suggest that at ages younger than ~2
Gyr, the red clump luminosity is greatly dependent on the age of the cluster and shows
little effect from the metallicity whereas clusters older than ~2 Gyr show the exact
opposite, having little age dependence while still showing the effects of metallicity.
Log Age = 8.6 Log Age = 8.8 Log Age = 9
Log Age = 9.2 Log Age = 9.4 Log Age = 9.6
Il 11 l
Log Age = 9.8 Log Age = 10 Log Age = 10.1
I III I
0.0 -1.0 -0.5
0.0 -1.0 -0.5
Effects of age on MW. Plotted is the observed variation of M with the
logarithm of the age as compared with the predictions of theoretical mod-
els (Girardi et al. 2000) for the indicated metallicities. The filled circles
represent clusters with ages that are within +0. 1 dex of the model age
in each panel. For the upper left and lower right panels, the filled circles
represent clusters with log(Age) < 8.7 and log(Age) > 10.05, respectively.
The remaining clusters in each panel are marked by open circles.
To facilitate a more detailed comparison between the models and the observations,
we utilize an interpolation routine based on low order polynomials to compare the
theoretical M~K values with the observed ones. As a test of the interpolation, we have
applied it to the observational data alone comparing the interpolated MPC values to
9.0 9.5 1 0.0 9.0 9.5 1 0.0
Log Age Log Age
Effects of [Fe/H] on Mf. Plotted is the observed variation of My with
metal abundance as compared with the predictions of theoretical models
(Girardi et al. 2000) for specific ages. The filled circles denote the clusters
with [Fe/H];n,, < [Fe/H] < [Fe/H];;;a, where [Fe/H];ni,, and [Fe/H];na are
halfway between the model shown and the next lower and next higher
metallicity models, respectively. For the models of [Fe/H] = -1.3 and
[Fe/H] = 0.2, clusters having [Fe/H] < -1.0 and [Fe/H] > 0.1, respec-
tively, are marked with filled circles. The remaining clusters in each panel
are shown by open circles.
the actual values at the age and abundance of each cluster. We find that the rms of
the residuals is negligible in MC with no systematic trends as a function of age or
abundance. We have also tested the interpolation routine on the theoretical models with
similar encouraging results. The accuracy of the interpolation allows us to compare the
My values predicted by the models for a given age and metallicity to the observed
Mc for each cluster. From this comparison, we find that the rms deviation of the
theoretical models from the observations is 0.16 mag, with no systematic variation
in the residuals as a function of age or metallicity. This deviation is slightly larger
Log Age = 8.6 Log Age = 8.8 Log Age = 9
o o a
0 a ~ ~
Log Age = 9.8 Log Age = 104 Log Age = 10.1
co a co ao
0 o a
I, I I I I II
-0.6 -0.4 -0.2 0.0 -0.6 -0.4 -0.2 0.0 -0.6 -0.4 -0.2 0.0
[Fe/H] [Fe/H] [Fe/H]
Figure 2-7: Age effects on solar neighborhood RC stars. We plot the K-band absolute
magnitude of solar-neighborhood RC stars with Hipparcos parallaxes vs.
their metallicities. The solid lines represent the predictions of theoretical
models constructed by Girardi et al. (2000). The models suggest that the
vertical spread in M~K can be explained by variations in the ages of the
than the mean error in the MC values of the 16 clusters, which we find to be 0.13
mag. Given that the mean deviation of the models from the observations is roughly
consistent with the errors inherent in the latter, it is reasonable to conclude that the
models are generally consistent with the observational data.
Figure 2-7 shows the Girardi models plotted along with the Alves (2000) field
red clump star data. The models reinforce the conclusion drawn by Alves (2000) that
2& is insensitive to metallicity for nearby stars in this abundance range. Furthermore,
given that a typical 2& error in Alves (2000) data is 0.11 mag, this figure suggests
that the vertical spread in the 2A values is mainly the result of age effects among the
field stars. Both the 108.8 and 109.2-9.4 year isochrones agree with the majority of the
data. However, given the expectation that stars in the Solar neighborhood are likely to
be near Solar-age, most of the Hipparcos stars in Figure 2-7 probably have Log ages
between 9.2 and 9.6 (1.6 to 4.0 Gyr).
It is interesting to note that the ages of the solar neighborhood stars as predicted
by the models show a lack of stars around 109 yr (Figure 2-7). In contrast, using
a model of the solar neighborhood RC that assumes a constant star formation rate,
Girardi & Salaris (2001) expect an age distribution for the RC stars that peaks at 1 Gyr
with approximately 60% of the stars having this age. The discrepancy in this result
with the apparent ages of the Hipparcos RC stars likely indicates a non constant star
formation rate in the solar neighborhood; this is not surprising if the formation of stars
is triggered by density waves traveling through the solar neighborhood, which is an
intrinsically episodic process.
2.4 Application As a Distance Indicator
An important aspect of this study is the application of the K-band RC absolute
magnitude as a distance indicator. To optimize this application in the present work,
we seek a range of age and abundance over which variations in MPC are minimized.
Inspecting Figure 2-4, we see that if the age of the stellar population is in the range
2gAgeg6 Gyr and the metal abundance is between -0.5g[Fe/H]g0.0, then the
intrinsic variation in MFC is minimized suggesting that uncertainties in our knowledge
of these properties are inconsequential in the determination of the distance. On the
basis of these considerations, we have selected the open cluster NGC 2158. This
cluster possesses 2MASS photometry, and it is included in the study of Twarog et al.
(1997), so we have a metallicity value ([Fe/H] = -0.24 & 0.06) that is on the same
system as the other clusters in this study. The age shift described in @ 2.2.1 is also
applied to NGC 2158 giving us an age of 1.6 +0.2 Gyr. We note in passing that NGC
2158 was not included as part of our M~K(RC) calibration because the distance given
in Twarog et al. (1997) was determined using the magnitude of the RC and not main
For the reddening toward NGC 2158 we can utilize the data in Table 2-1 to
parameterize the intrinsic color of the RC [(J -K)o] in terms of the metal abundance
and age. Figure 2-8 shows (J-K)o versus [Fe/H] (left panel) and age (right panel)
for the clusters in our sample. Using the interpolation discussed in @ 2.3.3, we
can determine the intrinsic color of NGC 2158 given its metallicity and age, for
which we find (J -K)o = 0.618 +0.003. We calculate the error in (J -K)o by
determining the uncertainty resulting from oage and GT[Fe/H] and adding these in
quadrature. Comparing the implied intrinsic color of the RC with the apparent color,
(J -K) = 0.837 +0.005, we find E(J -K) = 0.219 +0.006. Converting this to a
color excess in the optical regime, we find E(B V) = 0.42 +0.012, which is in good
agreement with published values (e.g., Christian et al. 1985; Twarog et al. 1997). The
preceding method represents an internally consistent formalism which can be utilized to
estimate the reddening of a cluster.
The interpolation on M using only the open cluster data predicts MZK =
-1.67 + 0.09 for NGC 2158. Along with E(B V) = 0.42 and the apparent RC
K-band magnitude, K(RC) = 11.53 + 0.02, we find (m M); = 14.35 + 0.09. Our
distance modulus for NGC 2158 agrees within the errors with the main sequence fitting
modulus of (m -M);- = 14.4 & 0.2 found by Christian et al. (1985), but is slightly
lower than that determined by Twarog et al. (1997) of (m -M);- = 14.5.
-1,0 -0.5 0.0 9.0 9.5 10.0
[Fe/H] Log Age
Figure 2-8: Intrinsic red clump color. The intrinsic color of the RC is plotted as a
function of [Fe/H] (left) and age (right), where the open circles represent
the open clusters while the filled circles are the globulars.
In this work, we have sought to establish the K-band absolute magnitude of the
helium burning red clump stars (Mf) as a distance indicator. To facilitate this, we
have utilized infrared photometry from the 2MASS catalog along with distances,
metallicities, and ages for 14 open clusters and two globular clusters. Our sample
encompasses an age range from 0.63 Gyr to 12 Gyr and metallicities from -1.15 to
0.15 dex. Based on an analysis of these data, we draw the following conclusions.
1. There is a statistically significant range ofMF values among the star clusters
in our sample. In particular, for the 14 open clusters, we calculate (MF)= -1.61 with
a standard deviation of 0.22 mag. In contrast, the mean error in these MW values is
2. Upon inspection of Figures 2-5 and 2-6, we find that for clusters younger than
~2 Gyr, MW is insensitive to metallicity but shows a dependence on age. In contrast,
for clusters older than ~2 Gyr, Mye is influenced primarily by the metallicity of the
population and shows little or no dependence on the age.
3. In general, MRC is less sensitive to age and metallicity than MRC over the
parameter range common to both this paper and Sarajedini (1999) from which the M,~RC
values are taken.
4. Over comparable metallicity and age ranges, our average MRC value of -1.61
mag is consistent with that of Alves (2000) which is based on solar-neighborhood
RC stars with Hipparcos parallaxes. We also suggest that the significant scatter in
the Alves (2000) M~K data is likely due to a range of ages between ~1.6 and ~4 Gyr
among these stars.
5. The theoretical RC models based on the formalism of Girardi et al. (2000)
agree reasonably well with our observational data, indicating that age plays an impor-
tant role in determining MRC for younger populations while metallicity mainly affects
6. Using the K-band absolute magnitude of the RC, we are able to compute
the distance to the open cluster NGC 2158. Adopting an age of 1.6 + 0.2 Gyr and
[Fe/H] = -0.24 & 0.06, our calibration yields a distance of (m M), = 14.35 + 0.09.
7. When determining distances for star clusters having -0.5 < [Fe/H] < 0.0 and
109.2 < age < 109.9, one can ignore the interpolation discussed in @ 2.3.3 and simply
use (M~K(RC)) = -1.61 +0.04.
2.6 Beyond Grocholski & Sarajedini (2002)
As discussed in the introduction, we ultimately want to apply our RC calibration
to determining the distance to the LMC. In Sarajedini et al. (2002) we tested the
feasibility of applying the RC calibration presented in @ 2.3.3 to the stellar populations
of the LMC. Over three nights in December 2001, using OSIRIS on the CTIO 4m,
we obtained JKs photometry of the LMC clusters Hodge 4 and NGC 1651 down to
a photometric depth of Ks ~19, or about 2 magnitudes below the RC. Images were
processed using a number of IRAF scripts, which performed the following tasks. The
nine dithered images of each cluster were dark subtracted and median combined,
without shifting, so as to remove any objects from the combined frame, leaving only an
exposure of the sky. We created a flat-field image by normalizing this sky frame. The
original images were then sky subtracted and flat-fielded using these calibration frames.
After processing, spatial offsets in the dither pattern were calculated and the images
were shifted and average combined resulting in a pair of images for each cluster with
signal-to-noise ratios equivalent to a single 198 s exposure in J and a single 270 s
exposure in K.
Standard star images were combined in the same fashion as the science frames
and they were analyzed with the QPHOT software package in IRAF, which was used to
measure instrumental magnitudes using a 15 pixel radius aperture. Standard stars were
observed over the course of three nights and we have combined these observations by
offsetting the measured instrumental magnitudes to the same zero point using stars
in common between the three nights. The combined data set of seven standard star
observations was fitted using a least-squares algorithm to calculate the extinction
coefficients and zero points. This procedure yielded the following equations:
j J = (2.03 + 0.03) +t (0.06 + 0.02)XJ, (2-3)
ks Ks = (2.23 + 0.03) +t (0.09 + 0.02)XK, (2-4)
where the lower case letters represent the instrumental magnitudes, X is the air mass,
and the capital letters are the apparent magnitudes.
The combined JKs cluster images were measured using the aperture photometry
routines in DAOPHOT (Stetson 1994) and calibrated using equations 2-3 and 2-4. Ks
magnitudes are converted to the K-band using equations 2-1 and 2-2 and Figure 2-9
shows the resulting [K,(J-K)] CMDs. Both of these clusters show a well populated
red giant branch (RGB) and an easily identifiable helium burning RC. The box in each
Table 2-2. LMC Cluster Information
Cluster [Fe/H] Age (Gyr) K(RC) Alg(RC) E(J-K) (m-M~)o
Hodge 4 -0.17 10.04 1.7 10.3 16.90 +0.02 -1.64 10.17 0.03 10.01 18.52 +0.17
NGC 1651 -0.07 +0.10 1.8 +0.3 17.03 +0.02 -1.54+0.12 0.06 +0.01 18.53 +0.12
panel indicates the stars that were used in calculating the median apparent K-band
magnitude of each cluster RC, which we find to be 16.90 + 0.02 mag and 17.03 + 0.02
mag for Hodge 4 and NGC 1651, respectively. As mentioned in the previous section,
knowledge of a cluster's age and metallicity are necessary to determining its distance
using M Work by Tiede, Martini, & Frogel (1997) showed that the slope of a
cluster's RGB in the near infrared is related to it's [Fe/H]. In Figure 2-9 we mark
the cluster RGB stars with filled circles and the solid line denotes the linear fit to
these stars. Utilizing these fits with the calibration of Tiede et al. (1997), we find
[Fe/H] = -0. 17 + 0.04 for Hodge 4 and [Fe/H] = -0.07 + 0. 10 for NGC 1651.
Additionally, we used published optical photometry of the MSTO for these clusters
(Sarajedini 1998, Hodge 4 and Mould et al. 1997, NGC 1651) to calculate ages
via MSF with theoretical isochrones (Girardi et al. 2000). The resulting ages are
1.7 + 0.3 Gyr for Hodge 4 and 1.8 + 0.3 Gyr for NGC 1651. Combining ages and
metallicities with our RC calibration, we find for Hodge 4, MC = -1.64 +0. 17 and
for NGC 1651, MC = -1.54 & 0. 12. Since they are sufficiently close to each other,
we averaged the reddening values derived from both the Burstein & Heiles (1982) and
Schlegel, Finkbeiner, & Davis (1998) dust maps, for each cluster. Using the reddening
relations quoted by Schlegel et al. (1998), A,- = 3.1E(B V), AK = 0. 11A,-, and
AJ = 0.28Ar-, we convert E(B V) to E(J -K). These values, along with other derived
cluster parameters, are given in Table 2-2.
Finally, combining our apparent and absolute K-band RC magnitudes with the
derived reddenings, we calculate the absolute cluster distances to be (m -M~o =
18.52 + 0. 17 and (m M1o = 18.53 + 0. 12 for Hodge 4 and NGC 1651, respectively.
These numbers are in good agreement with the LMC distance of 18.515 + 0.085 mag
published by Clementini et al. (2003). This value is derived by averaging published
LMC distances determined through a variety of distance indicators. Additionally,
these cluster distances are in good agreement with the LMC geometry determined by
van der Marel & Cioni (2001) and Olsen & Salyk (2002), which indicates that Hodge
4, in the Northeast portion of the LMC, should be closer to us than NGC 1651, found
in the Southwest.
o OO oOO
16 gs oo
OO" OO p p
0.0 0.5 1.0
Figure 2-9:) Near-IR ::- CMls f~or the : i:: clusters Hodge 41 and :: 1651. The
: : 1 are used to outline: the cluster R~s. All stars within these boxes
are used in .:1 ::1 :i::- the :i -::i- K--band RC magnitude. ( i::.i= :- RGB
stars are denoted by the filled circles with the solid i showing the
ABUNDANCES AND VELOCITIES OF A SAMPLE OF LMC CLUSTERS
In the current paradigm of galaxy formation, it is believed that the formation
history of spiral galaxy spheroids, such as the Milky Way (MW) halo and bulge, may
be dominated by the accretion/merger of smaller, satellite galaxies (e.g., Searle & Zinn
1978; Zentner & Bullock 2003). This type of galactic interaction is currently seen
in the Sagittarius dwarf galaxy (Sgr), which is in the midst of being cannibalized
by the MW. However, due to its location on the opposite side of the Galaxy from
us (Ibata et al. 1994), contamination by MW foreground stars makes it difficult to
study stellar populations in Sgr. In contrast, both the Large Magellanic Cloud (LMC)
and Small Magellanic Cloud (SMC), two satellite galaxies that may eventually be
consumed into the MW halo, suffer little from foreground contamination due to
their direction on the sky, which places them well out of the plane of the MW. In
addition, the relative proximity of these galaxies allows us to easily resolve stellar
populations in the Magellanic Clouds down below their oldest main sequence turnoffs
(MSTOs). Thus, the LMC and SMC offer us a golden opportunity to study the effects
of dynamical interactions on the formation and evolution of satellite galaxies; this
information plays an integral part in discovering the secrets of spiral galaxy formation.
One of the most direct ways to determine the chemical evolution history (CEH)
and star formation history (SFH) of a galaxy is through the study of its star clusters,
which preserve a record of their host galaxy's chemical abundances at the time of
their formation. The LMC star clusters continue to play a critical role in shaping
our understanding of the age-metallicity relation of irregular galaxies. The rich star
cluster system of the LMC is also a unique resource for many experiments in stellar
and galactic astronomy, largely due to the fact that the LMC harbors well populated
clusters that occupy regions of the age-metallicity plane that are devoid of MW
clusters. Thus, LMC clusters have been widely studied as a test of stellar evolution
models at intermediate metallicity and age (e.g., Brocato et al. 1994; Ferraro et al.
1995; Bertelli et al. 2003) and as empirical templates of simple stellar populations for
applications to population synthesis models of unresolved galaxies (e.g., Beasley et al.
2002; Leonardi & Rose 2003; Maraston 2005).
The LMC cluster system, however, is well known to show a puzzling age distri-
bution, with a handful of old (~ 13 Gyr), metal-poor globular clusters; a number of
intermediate age (1-3 Gyr), relatively metal-rich populous clusters; and, apparently,
only one cluster, ESO 121-SCO3 (~ 9 Gyr, hereafter ESO 121), that falls into the
LMC's so-called "age gap" (e.g., Da Costa 1991; Geisler et al. 1997; Da Costa 2002).
We note that the LMC bar seems to show a formation history very similar to that of
the clusters (Cole et al. 2005), while field SFHs derived from deep color-magnitude
diagrams (CMDs) suggest that stars in the LMC disk had a constant, albeit low, star
formation rate during the cluster age gap (Holtzman et al. 1999; Smecker-Hane et al.
2002). While the cause of the cessation of cluster formation (the beginning of the
age gap) is not known, dynamical simulations by Bekki et al. (2004) suggest that the
recent burst of cluster formation is linked to the first very close encounter between the
Clouds about 4 Gyr ago, which would have induced "dramatic gas cloud collisions,"
allowing the LMC to begin a new epoch of cluster and star formation; strong tidal
interactions between the Clouds have likely sustained the enhanced cluster formation.
Bekki et al. (2004) also find that the close encounter between the Clouds would have
been sufficient to cause the formation of the LMC bar around the time of the new
epoch of cluster formation, giving rise to the similar SFHs seen in the cluster system
and the bar. In addition to enhancing star formation, tidal forces can result in the infall
or outflow of material, thereby affecting the CEH of the LMC and, at the same time,
leaving behind a signature of the interaction. Thus, accurate knowledge of the ages
and metallicities of LMC clusters is necessary to fully understand the formation and
dynamical history of this galaxy.
While age and metallicity estimates from isochrone fitting to CMDs exist for
a large number of clusters, the degeneracy between age and metallicity makes these
estimates inherently uncertain in the absence of solid metallicity measurements based
on spectroscopic data. Integrated light has been used to measure [Fe/H] for many of
these clusters; however, these values are often problematic, since the cluster light can
be dominated by a few luminous stars, and the results are susceptible to small-number
statistical effects. In recent years, high spectral resolution studies of a few prominent
clusters have been undertaken, yielding, for the first time, detailed abundance estimates
of a wide variety of elements, including iron, for individual stars within these clusters
(Hill 2004; Johnson et al. 2006). This work is highly valuable, but because of the large
investment in telescope time necessary to obtain data of sufficiently high signal-to-noise
ratio (S/N), it has necessarily been limited to only a few stars in a few clusters; most
of these targets are very old, leaving the vast majority of young and intermediate age
Moderate-resolution studies are an excellent complement to high-resolution work
for a couple of reasons. First, the multi-object capability available for many moderate-
resolution spectrographs makes it possible to observe many potential cluster members
in a given field. This increases the probability of observing true cluster members and
facilitates their identification, even in sparse clusters. Second, less integration time is
needed to achieve the desired S/N at moderate resolution, allowing the observation of
many more targets in a given amount of time. Thus, with moderate-resolution spectra
we can observe a large number of targets in a short period of time and thereby create
an overview of a galaxy's global metallicity distribution, both spatial and temporal.
This approach is particularly important for the LMC, since its metallicity distribution is
very broad and the intrinsic shape is not very well known.
To date, the only large-scale spectroscopic metallicity determination for LMC clus-
ters based on individual cluster stars has been the landmark study by (Olszewski et al.
1991, hereafter, OSSH; Suntzeff et al. 1992). They obtained medium-resolution spectra
of red giant branch (RGB) stars in ~ 80 clusters at a wavelength of a 8600 A+, cen-
tered on the very prominent triplet of Ca II (CaT) lines. Their work was motivated by
the recognition that the CaT lines were proving to be a reliable metallicity indicator
for Galactic globular clusters (e.g., Armandroff & Zinn 1988; Armandroff & Da Costa
1991) Additionally, this spectral feature is easily measured in distant targets and at
medium resolution since the CaT lines are extremely strong and RGB stars are near
their brightest in the near-infrared. Using the CaT, OSSH calculated metallicities and
radial velocities for 72 of their target clusters. Analysis of the metallicity distribution
showed that the mean [Fe/H] values for all clusters in the inner (radius < 5 ) and outer
(radius > 5 ) LMC are almost identical (-0.29 + 0.2 and -0.42 + 0.2, respectively),
suggesting the presence of little, if any, radial metallicity gradient, in sharp contrast
to what is seen in the MW (e.g., Friel et al. 2002) and M33 (Tiede et al. 2004). Us-
ing radial velocities from the OSSH sample, Schommer et al. (1992) found that the
LMC cluster system rotates as a disk, with no indication that any of the clusters have
kinematics consistent with that of a pressure-supported halo.
However, the results of OSSH present some difficulties owing to the limitations
of technology at the time. The use of a single-slit spectrograph severely limited the
number of targets observed toward each cluster. Additionally, the distance of the LMC
paired with a 4m telescope required that they observe the brightest stars in the clusters.
Many of these stars are M giants, which have spectra contaminated by TiO (although
it may not be significant until spectral type MS or later), or carbon stars, neither of
which are suitable for using the CaT to determine [Fe/H]. Thus, the combination of a
single-slit spectrograph with a midsized telescope made it difficult for OSSH to build
up the number of target stars necessary to differentiate between cluster members and
field stars. Most of the resulting cluster values are based on only one or two stars; in
some cases, there are metallicity or radial velocity discrepancies between the few stars
measured, and it is unclear which of the values to rely on.
The interpretation of the OSSH results is further complicated by subsequent
advances both in knowledge of the globular cluster metallicity scale to which the
CaT strengths are referred (Rutledge et al. 1997a) and in the standard procedure used
to remove gravity and temperature dependencies from the CaT equivalent widths
(Rutledge et al. 1997b). It is not a simple matter to rederive abundances from the
equivalent widths of OSSH because of the lack of homogeneous photometry for many
of the clusters; mapping the OSSH abundances to a modern abundance scale (e.g., that
defined at the metal-poor end by Carretta & Gratton 1997 and near-solar metallicity by
Friel et al. 2002) is insufficient, because the transformation is nonlinear and random
metallicity errors tend to be greatly magnified (see, Cole et al. 2005).
In an effort to produce a modern and reliable catalog of LMC cluster metallicities,
we have obtained near-infrared spectra of an average of eight stars in each of 28 LMC
clusters. We have taken advantage of the multiplex capability and extraordinary image
quality and light-gathering power of the European Southern Observatory's (ESO)
8.2m Very Large Telescope (VLT) and of the great strides in the interpretation and
calibration of CaT spectroscopy made in the past 15 years to provide accurate cluster
abundances with mean random errors of 0.04 dex. Here we present our derived cluster
metallicities and radial velocities and compare these results to previously published
spectroscopic metallicities. The metallicity distribution of several hundred non-cluster
LMC field stars will be presented in a forthcoming paper (A. A. Cole et al., in prepara-
tion). The current chapter is laid out as follows: Section 3.2 discusses the observations
and data processing. In ~3.3 we present the derived cluster properties, and comparisons
to previous works are detailed in ~3.4. Finally, in ~3.5 we summarize our results. We
note that these data and results were originally presented in Grocholski et al. 2006.
3.2.1 Target Selection
We observed 28 prominent star clusters scattered across the face of the LMC,
in environments ranging from the dense central bar to the low-density regions near
the tidal radius (a 29th cluster was observed, but it appears to be too young to apply
the CaT method; see ~3.6.2). Our observations were aimed at clusters rich enough
and sufficiently large and diffuse to give us confidence in harvesting at least four
definite cluster members from which to derive the cluster metallicity. In order to obtain
leverage on the LMC age-metallicity relation, we included clusters from SWB class
IVB-VII, spanning the age range of clusters containing bright, well-populated RGBs
(Persson et al. 1983; Ferraro et al. 1995). Our sample was intentionally biased towards
those clusters with conflicting or uncertain previous abundance measurements, those
thought to lie near the edge of the age gap, and those whose radial velocities might
provide new insight into the dynamical history of the LMC-SMC system, based on
their location. Our targets and their positions, sizes, integrated V magnitudes, and SWB
types are listed in Table 3-1. A schematic of the LMC is presented in Fig. 4-4. Shown
are near-infrared isopleths from van der Marel (2001; solid ellipses), at semi-major
axis values of 1 1 5, 2 3 4 6 and 8 Prominent H I features (dashed lines;
Staveley-Smith et al. 2003) and the two largest centers of LMC star formation (30 Dor
and N11; open circles) are also plotted. Finally, the rotation center of intermediate-age
stars is denoted by the open square (van der Marel et al. 2002), and the H I rotation
center from Kim et al. (1998) is plotted as the open triangle. Our target clusters are
plotted with solid symbols, with the exception of NGC 1841, which lies farther south
than the area covered by this diagram.
~ BJ~ -P
Schematic diagram of the LMC showing the location of our target clusters
along with prominent features. Filled symbols represent the target clus-
ters, with symbol size directly related to V magnitude and shape denoting
SWB type, where the triangles, squares, and pentagons are type V, VI,
and VII, respectively. Note that NGC 1841 (declination ~ -84") is out-
side of the range of this plot and NGC 1861 (SWB type IVB) is marked
by a filled triangle. Near-infrared isopleths from van der Marel (2001)
are marked by solid lines while the dashed lines outline major H I fea-
tures (see Staveley-Smith et al. 2003. The H I rotation center (Kim et al.
1998) is marked with the open triangle, and the rotation center of the
intermediate-age stars (van der Marel et al. 2002) is shown by the open
square. Finally, the two largest H II regions are marked by open circles.
Pre-images of our target fields in the V and I bands were taken by ESO Paranal
staff in the fall of 2004, several months prior to our observing run. The pre-images
were processed within IRAF, and stars were identified and photometered using the
aperture photometry routines in DAOPHOT (Stetson 1987). Stars were cataloged using
the FIND routine in DAOPHOT and photometered with an aperture size of 3 pixels.
The V- and I-band data were matched to form colors. Red giant targets were chosen
based on the instrumental CMD, and each candidate was visually inspected to ensure
location within the cluster radius (judged by eye) and freedom from contamination by
very nearby bright neighbors. In each cluster we looked for maximum packing of the
S8"' long slits into the cluster area and for the best possible coverage of the magnitude
range from the horizontal branch/red clump (V M 19.2) to the tip of the RGB (V m
16.4). The positions of each target were defined on the astrometric system of the
FORS2 pre-images so that the slits could be centered as accurately as possible, and the
slit identifications were defined using the FORS Instrument Mask Simulator software
provided by ESO; the slit masks were cut on Paranal by the FORS2 team.
The spectroscopic observations were carried out with FORS2 in visitor mode at
the Antu (VLT-UT1) 8.2 m telescope at ESO's Paranal Observatory during the first half
of the nights of 21-24 December 2004; weather conditions were very clear and stable
during all four nights, with seeing typically 0."5-1"'0. We used the FORS2 spectrograph
in mask exchange unit (MXU) mode, with the 1028z+29 grism and OG590+32 order
blocking filter. The MXU slit mask configuration allows the placement of more slits
on the sky than the 19 movable slits provided in Multi-Object Spectrograph mode. We
used slits that were 1" wide and 8"' long (7"' in a few cases), and, as mentioned above,
targets were selected so as to maximize the number of likely cluster members observed;
typically 10 stars inside our estimated cluster radius were observed, with an additional
~ 20 stars outside of this radius that appeared to be LMC field red giants based on our
FORS2 uses a pair of 2k x 4k MIT Lincoln Laboratory CCDs, and the target
clusters were centered on the upper (master) CCD, which has a readout noise of 2.9
electrons, while the lower (secondary) CCD, with a readout noise of 3.15 electrons,
was used to observe field stars. The only exception to this was the Hodge 11-SL 869
field, where, with a rotation of the instrument, we were able to center Hodge 11 in
the master CCD and SL 869 in the secondary CCD. Both CCDs have an inverse gain
of 0.7e- ADU Pixels were binned 2 x2, yielding a plate scale of 0."25 pixel ,
and the resulting spectra cover 1750 A+, with a central wavelength of 8440 A+ and a
dispersion of ~0.85 A pixell (resolution of 2-3 A+). While the FORS2 field of view
is 6(8 across, it is limited to 4(8 of usable width in the dispersion direction in order to
keep important spectral features from falling off the ends of the CCD.
Each field was observed twice, with offsets of 2" between exposures, to ameliorate
the effects of cosmic rays, bad pixels, and sky fringing. The total exposure time in
each setup was either 2 x 300, 2 x 500 or 2 x 600 s. Both the readout time (26
s) and setup time per field (some 6-10 minutes) were very quick and allowed us to
obtain longer exposures than originally planned in many cases. For most of our targets
with short exposure times (300 s) we combined the spectra so as to improve the S/N.
However, with the longer exposures (500 and 600 s) we found that the S/N in a single
exposure was adequate, and cosmic rays and bad pixels were not a problem, so we
have used only one of the pair of exposures in our analysis. Column 8 of Table 3-1
gives the total exposure time that we have used in our analysis of each cluster.
Calibration exposures were taken in daytime under the FORS2 Instrument Team's
standard calibration plan. These comprise lamp flat-field exposures with two different
illumination configurations and He-Ne-Ar lamp exposures for each mask. Two lamp
settings are required for the flat fields because of parasitic light in the internal FORS2
In addition to the LMC clusters, we observed four Galactic star clusters (47
Tuc, M67, NGC 2298, and NGC 288), three of which are a subsample of the CaT
calibration clusters in Cole et al. (2004, hereafter, CO4). Since we used the same
instrument setup as CO4, we expected to use their CaT calibration, and these three
clusters were observed to serve as a check on the validity of that approach. Processing
of these three clusters shows that our results are identical to within the errors; thus,
we use the CaT calibration of CO4 rather than deriving our own CaT calibration
Image processing was performed with a variety of tasks in IRAF. The IRAF task
ccdproc was used to fit and subtract the overscan region, trim the images, fix bad pix-
els, and flat field each image with the appropriate dome flats. The flat-fielded images
were then corrected for distortions in order to facilitate extraction and dispersion cor-
rection of the spectra. The distortion correction is a two-step process, whereby first the
image of each slitlet is rectified to a constant range of y-pixel (spatial direction) values
on the CCD, and then the bright sky lines are traced along each slitlet and brought
perpendicular to the dispersion direction. The amount of the distortion is minimal near
the center of the field of view and increases toward the edges; in all cases it is fit with
a polynomial that is at most quadratic in y and linear in x. Although the distortion
corrections are small, they greatly reduce the residuals left after sky subtraction and
improve the precision and accuracy of the dispersion solution (see below).
Once distortion corrections were completed, the task pall (in the HYDRA
package) was used to define the sky background and extract the stellar spectra into one
dimension. The sky level was defined by performing a linear fit across the dispersion
direction to sky "windows" on each side of the star. This procedure presented few
difficulties, since the target stars were usually bright compared to the sky and the
seeing disks were small compared to the length of the slitlets. The only problems
arose when the star fell near the top or bottom of the slitlet; in these cases the sky
regions were chosen interactively, and we found for all of these spectra that the
resulting sky subtraction was indistinguishable from that of more centrally located
stars. While daily arc lamp exposures are available for dispersion-correcting the
spectra, telescope flexure during the night, along with small slit centering errors, makes
this a less desirable method for correcting the spectra. As such, more than 30 OH
night-sky emission lines (Osterbrock & Martel 1992) were used by the IRAF tasks
identify/, refspectra, and dispcor to calculate and apply the dispersion solution for each
spectrum, which was found to be ~ 0.85 A+ pixels with a characteristic rms scatter
of ~ 0.06 A+. For the short (300 s) exposure data, we processed both sets of images
for each pointing and combined the dispersion-corrected spectra using combine to
improve the S/Ns for these stars. In a few cases we found that averaging the stellar
spectra actually decreased the S/N; for these stars we chose to use the higher quality
of the two individual spectra in place of the averaged spectrum. All spectra were then
continuum-normalized by fitting a polynomial to the stellar continuum, excluding
strong absorption features (both telluric and stellar). For the final spectra, S/Ns are
typically 25-50 pixels with some stars as high as ~ 90 pixels and, in only a few
cases, as low as ~ 15 pixel Sample spectra showing the CaT region are presented in
3.2.4 Radial Velocities
Accurate radial velocities for our target stars are important for two reasons. First
and foremost, since a cluster's velocity dispersion is expected to be relatively small
compared to the surrounding field and its mean velocity quite possibly distinct from
the field, radial velocities are an excellent tool for determining cluster membership. In
[Fe/H] = -1.84
[Fe/H] = -1.31
[Fe/H] = -0.41
.Hodge 11 #9
.NGC 2019 #10
.IC 2146 #5
. ~Fe I F
EEW = 5.57
EEW = 6.83
EEW = 9.24
Fe I Fe I
Sample of spectra from RGB stars in our target clusters covering a range
in metallicities. The three CaT lines, along with some nearby Fe I lines,
are marked for reference; the change in CaT line strength with [Fe/H] is
readily visible. Calculated summed equivalent widths and metallicities for
each star are given.
addition, our equivalent-width measuring program uses radial velocities to derive the
expected CaT line centers.
Radial velocities for all target stars were determined through cross-correlation
with 30 template stars using the IRAF task fxcor (Tonry & Davis 1979), and we have
chosen to use template spectra from CO4. The template stars were observed as a part
of their CaT calibration program; thus, their sample offers a good match to the spectral
types of our target stars. In addition, their observations were made with a telescope and
instrument setup that is almost identical to ours. CO4 chose template stars for which
reliable published radial velocity measurements were available. Template velocities
came from the following sources: 11 stars from NGC 2298, NGC 1904, and NGC
4590 (Geisler et al. 1995); 8 stars from Berkeley 20 and Berkeley 39 (Friel et al.
2002); 2 stars from Melotte 66 (Friel & Janes 1993); 6 stars from M67 (Mathieu et al.
1986); and 3 stars from 47 Tuc (Mayor et al. 1983). In addition to calculating relative
radial velocities, fxcor uses information about the observatory location and the date
and time of the observations (once the ESO header has been appropriately reformatted)
to correct the derived velocities to the heliocentric reference frame. For a star's final
heliocentric radial velocity, we adopt the average value of each cross-correlation result.
We find good agreement among the template-derived velocities, with a typical standard
deviation of ~ 6 km sl for each star.
When the stellar image is significantly smaller than the slit width, systematic er-
rors due to imprecise alignment of the slit center and the stellar centroid can dominate
the error budget in the radial velocity measurements. With the grism and CCD used
here, an offset of 1 pixel across the 4 pixel wide slit would introduce an error in the
measured velocity of a 30 km s We follow the approach of Tolstoy et al. (2001)
in applying a correction to each measured radial velocity based on the individual slit
offsets; following CO4, we measure the slit offsets using acquisition (so-called through-
slit) images taken immediately prior to the spectroscopic measurement and estimate
a precision of m0.14 pixels on the measured offset value. This introduces an error
of + 4.2 km sl and, added in quadrature with the error resulting from the velocity
cross-correlations, gives an error of roughly 7.5 km s We adopt this as the error in
measuring the radial velocity of an individual star.
Table 3-2. CaT Line and Continuum Bandpasses
Feature Line Bandpass (A+) Blue Continuum (8+) Red Continuum (8+)
Ca II h8498 8490 8506 8474 8489 8521 8531
Ca II h8542 8532 8552 8521 8531 8555 8595
Ca II h8662 8653 8671 8626 8650 8695 8725
3.2.5 Equivalent Widths and Abundances
To measure the equivalent widths of the CaT lines, we have used a previously
written FORTRAN program (see CO4 for details). However, since this region of a
star's spectrum can be contaminated by weak metal lines and, in some cases, weak
molecular bands, measuring the true equivalent width of the CaT lines at all but
the highest spectral resolutions is impossible. Instead, we follow the method of
Armandroff & Zinn (1988) and define continuum bandpasses on either side of each
CaT feature. In this wavelength range, the continuum slope of a red giant star is
virtually flat; thus, the "pseudo-continuum" for each CaT line is easily defined by
a linear fit to the mean value in each pair of continuum windows. The "pseudo-
equivalent width" is then calculated by fitting the sum of a Gaussian and a Lorentzian,
required to have a common line center, to each CaT line with respect to the "pseudo-
continuum." For reference, the rest wavelengths of the line and continuum bandpasses,
as defined by Armandroff & Zinn (1988), are listed in Table 3-2. For many years it
has been known that even at the moderate spectral resolutions used here, a Gaussian
fit to the CaT lines is susceptible to loss of sensitivity at high metallicity because
the Gaussian fails to accurately measure the extremely broad wings of the lines (see
discussion in Rutledge et al. 1997b). We follow the procedure established in CO4
and add a Lorentzian profile to the Gaussian in order to recover sensitivity to the full
range of metallicities. Errors in the equivalent width measurements were estimated by
measuring the rms scatter of the data about the fits.
A number of previous authors have calibrated the relationship between the
strengths of the three CaT lines and stellar abundance using a variety of methods (see
Table 3 in Rutledge et al. 1997a). In all cases, a linear combination of the individual
line strengths was used to produce the summed equivalent width, CW, with weighting
and inclusion of lines (some authors dropped the weakest line, 8498 A+) varying based
on the quality of their data. Since the quality of our data is such that all three lines are
well measured, we adopt the same definition for CW as CO4,
CW EW8498 +tEW8542 +tEW8662. (3-1)
It is well known that TiO, which has a strong absorption band beginning near 8440
A+ (e.g., Cenarro et al. 2001), can affect the spectra of cool (~ MS or later), metal-rich
stars. This absorption feature, which depresses the "pseudo-continuum" around the
CaT lines and results in an underestimation of the measured equivalent widths, was
noted by OSSH in some of their LMC spectra. During processing, we checked each
spectrum for the appearance of this TiO absorption band and found no evidence that
TiO had affected any of our observations.
Both theoretical (Jarrgensen et al. 1992) and empirical (Cenarro et al. 2002)
studies have shown that effective temperature, surface gravity, and metallicity all
play significant roles in determining the CaT line strengths. However, it is well
established that for red giants of a given metallicity, there is a linear relationship
between a star's absolute magnitude and CW (Armandroff & Da Costa 1991), where
stars farther up the RGB have larger CW values. This is primarily due to the change
in surface gravity as a star moves along the RGB; stars near the bottom of the RGB
have smaller radii, thus larger surface gravities, which increases the H- opacity. Since
H- is the dominant opacity source in red giants, increasing the H- opacity depresses
the "pseudo-continuum", which in turn drives down the measured value for CW. To
remove the effects of luminosity on CW, similar to previous authors, we define a
reduced equivalent width, W', as
W' CW +t p(V VHB), (3-2)
where the introduction of the brightness of a cluster's horizontal branch (HB), VHB,
removes any dependence on cluster distance or reddening (see the thorough discussion
in Rutledge et al. 1997b). Due to the fact that a majority of our clusters are too
young and metal-rich to have a fully formed HB, we instead adopt the median value
of the core helium-burning red clump (RC) stars for these clusters (see ~3 for more
information). Values for p have been derived empirically by previous authors, with the
most robust determination being that of Rutledge et al. (1997b). Utilizing stars from 52
Galactic globular clusters, they found a metallicity-independent value of P = 0.64+0.02
A+ mag covering clusters in the range -2.1 g [Fe/H] g -0.6. Similarly, CO4 found
p = 0.66 + 0.03 for the globular clusters in their sample. However, when their open
clusters were included, the slope steepened to P = 0.73 + 0.04. This steepening of the
relationship between W' and V VHB with [Fe/H] is in qualitative agreement with the
theoretical results of Jorrgensen et al. (1992). Since our target clusters span an age and
metallicity range similar to the entire calibration cluster sample observed by CO4, for
p we have chosen to adopt their value of 0.73, which is based on both their open and
globular calibration clusters. To validate this approach, as mentioned in ~3.2.2, during
our science observations we observed a subsample of the calibration clusters used by
CO4 and found that, to within the errors, our measurements are identical to theirs, as is
expected, given that essentially the same instrument setup was used in both programs.
Before proceeding to the last step of the CaT calibration, we need to address
the issue of possible age effects on these calculations. As noted by previous authors
(e.g., Da Costa & Hatzidimitriou 1998; CO4; Koch et al. 2006), the age of a stellar
population affects the luminosity of core helium burning stars and may introduce
systematic errors in determining V VHB and, therefore, metallicities derived via
the CaT method. Experiments by CO4 and Koch et al. (2006) have shown that age
effects brought about by using an inappropriate VHB for any given RGB star will
typically cause errors in [Fe/H] on the order of + 0.05 dex, but these errors can,
in extreme cases, be as large as + 0.1 dex. One can avoid this type of uncertainty
by observing populous clusters, since this allows the correlation of a given RGB
star to a specific HB/RC, which is composed of stars of the appropriate age and,
therefore, has a well-defined mean magnitude. However, Da Costa & Hatzidimitriou
(1998) still had to address the issue of age effects for their sample of SMC clusters
due to the fact that many of their target clusters were considerably younger than the
Galactic globular clusters used in the CaT calibration of Da Costa & Armandroff
(1995); thus, they sought to correct for the difference in age between the target and
calibration clusters. Using adopted cluster ages, along with theoretical isochrones,
Da Costa & Hatzidimitriou (1998) estimated the change in VHB from the old to the
young populations, thereby creating age-corrected metallicities for their targets. Their
corrections were of the order of 0.05 dex, which is smaller than the precision of the
abundances. In contrast to Da Costa & Hatzidimitriou (1998), we have made no
attempt to calculate any age corrections for the following reason. We use the CaT
calibration of CO4, which is based on a sample of both globular and open clusters,
covering a wide range of ages and metallicities. With the inclusion of younger clusters,
the variation of VHB with age is built into the CaT calibration, specifically in Eq. 3-2,
and the steeper value for p than what has been found by authors only considering
globular clusters. Thus, age corrections are not required for our abundance data.
Finally, Rutledge et al. (1997a) showed that for MW globular clusters there is a
linear relationship between a cluster's reduced equivalent width and its metallicity on
the Carretta & Gratton (1997) abundance scale. CO4 extended this calibration to cover
a larger range of ages (2.5 Gyr g age g 13 Gyr) and metallicities (-2 g [Fe/H] g
-0.2) than previous authors, and, because their calibration is closer in parameter space
to our cluster sample, we adopt their relationship, where
[Fe/H] = (-2.966 + 0.032) +t (0.362 + 0.014)W'. (3-3)
We note that, while this calibration actually combines two metallicity scales
(Carretta & Gratton 1997 for the globular clusters and Friel et al. 2002 for the open
clusters), CO4 find no evidence of age effects on the calibration or any significant
deviation from a linear fit to suggest that these two populations are not ultimately on
the same [Fe/H] scale (see their Figure 4). Although some of our clusters are likely
younger than the 2.5 Gyr age limit established in the calibration of CO4, the CaT line
strengths for red giants of ~1 Gyr are not expected to deviate strongly from a simple
extrapolation of the fitting formula (based on the empirical fitting functions from
Cenarro et al. 2002 applied to isochrones published in Girardi et al. 2000), so we use
the above calibration for all of our clusters.
As mentioned in ~3.2.5, knowledge of the relative brightness of each target star
and the cluster HB is imperative to the accurate calculation of W' and thus [Fe/H] for
each star. To determine V yHB we utilized the preimages necessary for creating the
slit masks used by FORS2. Small-aperture photometry was performed on these V- and
I-band images so as to allow us to create cluster CMDs below the core helium-burning
RC stars. For the younger clusters in our sample, yHB was measured as the median
magnitude of cluster RC stars. Cluster stars were isolated from the field by selecting
stars within the inner half of the apparent cluster radius. We then placed a standard-
sized box (0.8 mag in V and 0.2 mag in V I) around each cluster RC and used
only the stars within this box in our calculation of VH. Regarding clusters with bona
fide HBs, i. e. old clusters, we compared our instrumental photometry to published
photometry and calculated a rough zero point for our data, allowing the conversion
of published PHB values onto our instrumental system. Literature sources for the five
old clusters are as follows: NGC 1841, Alcaino et al. (1996); NGC 2019, Olsen et al.
(1998); NGC 2257 and Hodge 11, Johnson et al. (1999); and Reticulum, Walker
(1992a). Errors in VHB are taken as the standard error of the median for clusters in
which we measured the RC directly; for the HB in the old clusters we adopt 0.1 mag.
We note that although we have not calibrated our photometry onto a standard system,
the V-I color term for the FORS2 filter system is expected to be small (<0.02 mag),
thus having little effect on the relative brightnesses of our target stars over the small
range of colors covered by the RGB.
3.3.1 Cluster Membership
We use a combination of three criteria to isolate cluster members from field
stars. This process is identical for all clusters, so we illustrate the process using
Hodge 11. First, the cluster centers and radii are chosen by eye, based primarily on
the photometric catalog. As an example, Fig. 3-3 shows xy-positions for all stars
photometered in the Hodge 11 field, with large filled symbols denoting our target stars
and the large open circle representing the adopted cluster radius; target stars marked
in blue (see figure legends for a discussion of the color coding used in Figs. 3-3
through 3-5 and Fig. 3-7) are considered non-members due to their distance from the
cluster center. We note that stars outside of the cluster radius were observed so as to
define parameters for the LMC field, which aids in isolating cluster members. Next,
radial velocity versus distance is plotted in Fig. 3-4. Stars moving at the velocity of
Hodge 11 are easily identified due to their smaller velocity dispersion and lower mean
velocity than that of the field stars. Our velocity cut, denoted by the horizontal lines,
has been chosen to represent the expected observed velocity dispersion in each cluster.
To determine this, we have adopted an intrinsic cluster velocity dispersion of 5 km
s-1 and added this in quadrature with our adopted radial velocity error, 7.5 km s-1
which results in an expected dispersion of ~ 9 km s-1. Thus, we have rounded this
up and adopted a width of + 10 km s-1 for our radial velocity cut. The cluster radius
150UU I I I I
.~ 'i~...(i. .~....'.~':" ':i~'Z.
I : '" :'::
'' .I ''r.: rrr .z ~i~ .
\, r~'' i j~T~;~r
'' *:' ;: .-rp: ~~C ~~L".
'1..;. i:i::. .... ~'..:' %~~:3 :~2e; iSl
"''"~'~' :~::Z4r: ~-~~
; ~fP~: ve.s-~n
... r. i;
'~' : (~C -
'~ 5~; .r--~~ .e
.: i. 5~
:~.'..... ': ~. ~
:r.i; :" 2':'~ rl:
. ~I .
. I I .
x position (pixels)
The xy positions of our target stars (large filled symbols) in the Hodge 11
field. The adopted cluster radius is marked by the large open circle, and
stars outside of this radius are considered non members. The color coding
of symbols in Figs. 3-3 through 3-5 and Fig. 3-7 is as follows: blue
points represent non members that are outside the cluster radius; teal and
green symbols represent non members that were cut because of discrepant
radial velocities and metallicities, respectively; and, finally, red symbols
denote cluster members.
350 Hodge 11
Radial velocities for our spectroscopic targets in Hodge 11, plotted as a
function of distance from the cluster center. The horizontal lines represent
our velocity cut and have a width of +10 km s The cluster radius is
shown by the vertical line, and the color coding of symbols is discussed
in Fig. 3-3. The error bars represent the random error in determining the
radial velocity for each star, where we have added in quadrature the slit
centering and cross-correlation errors.
(Fig. 3-4, vertical line) is marked for reference. Finally, Fig. 3-5 shows metallicity as
a function of distance for the stars in Hodge 11, with horizontal lines representing the
[Fe/H] cut that has been applied to these data. For the stars in six of our clusters we
have processed both sets of spectra and compared the two [Fe/H] measurements so as
I liii~iilll I!lalr'i.
50 100 150
Hodge 11 target star metallicities. Metallicities are plotted as a func-
tion of distance for all target stars in Hodge 11. The [Fe/H] cut of +0.20
dex is denoted by the horizontal lines. For this old, metal-poor cluster,
the field ([Fe/H] ~ -0.5) is easily distinguished from the cluster (red
symbols). We note that the color coding is the same as in Fig. 3-3. The
plotted error bars represent the random error in calculating [Fe/H], where
we have propagated the error in measuring the equivalent widths through
to directly determine the metallicity error for each star. Based on these data we find
G[~Fe/H] M 0.15 dex, which we adopt as the random error in [Fe/H] for each star. We
have rounded this up to + 0.20 dex for use as the metallicity cut shown in Fig. 3-5.
[Fe/H] = -1.84
1 0 -1 -2 -3 -4
Figure 3-6: CW vs. V yHB for Hodge 11; only stars considered to be cluster mem-
bers are plotted. The dashed line is an isoabundance line at the mean
metallicity of the cluster, [Fe/H] = -1.84, and has a slope P = 0.73.
Red symbols denote stars that have made all three cuts and are therefore considered
to be cluster members. Since we had no a priori membership information, up to this
point we have used a value for yHB that was derived from the entire field, rather than
just the cluster. Thus, we have recalculated W' (and [Fe/H]) using the appropriate
cluster yHB value. In Fig. 3-6 we present the traditional CW versus V yHB plot for
cluster members, with the dashed line representing the mean metallicity of Hodge
-~ ~ ~ ~ ~~ 1. 10-.5000. .
Figure 3-7 CMD for he entireHodge 11 ield, wit taret tr akda ecie
in~~~~~~~~ Fi.33 lse ebr leaogteRBadAB
11.Th CM i Fi. -7 hos al tar potmetre inth Hoge11 ied; lute
members~~~~~~~~~~~~~~~L= (rdsmos i nte G n smttcgan rnh(G) iue
3-1 thoug 3-8 peset te custr mmbe seecionposo h eann lses
inth fllwngfoma: ah lute s plt ve wofiurs wthth frt igrei
eac par(d ubrdfgue)soigtex-pstoso h age tr uprlf
panl) heicnti aia eoit n FeH esu itnc rmth lsercne
(upr ih adlwe efrspciel) ndfnalLWvrusV- H oralclse
members (lower right). The second figure for each cluster (even numbers) shows the
CMD for all stars in each pointing (cluster and field stars), with the spectroscopic
targets marked, using the color coding as discussed in Fig. 3-3. We note that these
figures are located at the end of this chapter.
In Table 3-6, also located at the end of this chapter, for all stars determined
to be members of the observed LMC clusters, we list the following information:
stellar identification number, right ascension and declination (as determined from the
preimages), heliocentric radial velocity and its associated error, V VHB, and CW,
along with the error in measuring CW. Although we do not discuss the field stars,
for completeness, in Table 3-7 we present our measured values for all field stars. In
this table, "primary" refers to all stars that fell on the same FORS2 chip as the target
cluster, but were found to be non-members of the cluster and "secondary" denotes the
stars that were observed on the non-cluster array. We note that Hodge 11 and SL 869
were observed in the same pointing, with Hodge 11 on the primary array and SL 869
on the secondary, hence the lack of secondary fields listed for either cluster.
3.3.2 Cluster Properties
Cluster properties derived from our data are presented in Table 3-3, with the
number of cluster stars given in column 2, the mean heliocentric radial velocities
and mean metallicities in columns 3 and 5, and their respective standard error of the
mean values in columns 4 and 6. For the clusters SL 4, SL 41, SL 396, Hodge 3, SL
663, and SL 869, we report the first spectroscopically derived metallicity and radial
velocity values based on individual stars within these clusters. In addition, NGC 1718
and NGC 2193 have no previously reported spectroscopic [Fe/H] values; however,
OSSH derived velocities for these two clusters. Of these eight clusters, NGC 1718
occupies a particularly interesting area of parameter space, as it is the most metal-poor
of our intermediate-age clusters, with a metallicity comparable to that of ESO 121 (see
Table 3-3. Derived LMC Cluster Properties
Cluster n Stars RV ofi [Fe H] ~Ielq
Name (lon s-1) (km s-1) (dex) (dex)
SL 4 5 227.1 3.6 -0.51 0.06
Reticulum 13 247.5 1.5 -1.57 0.03
NGC 1651 9 228.2 2.3 -0. 53 0.03
NGC 1652 7 275.7 1.3 -0.46 0.04
NGC 1841 16 210.3 0.9 -2.02 0.02
SL 41 6 229.3 1.3 -0.44 0.03
SL 61 8 221.9 2.0 -0.35 0.04
NGC 1718 3 278.4 2.2 -0. 80 0.03
NGC 1751 6 245.4 2.1 -0.44 0.05
NGC 1846 17 235.2 0.9 -0.49 0.03
SL 396 5 225.2 1.1 -0.39 0.05
NGC 1942 8 293.7 2.3 -0. 50 0.04
NGC 2019 5 280.6 2.3 -1.31 0.05
Hodge 4 7 310.8 1.9 -0.55 0.06
Hodge 3 7 277.4 0.8 -0.32 0.05
IC 2146 18 226.3 0.6 -0.41 0.02
SL 663 8 301.4 1.5 -0. 54 0.05
NGC 2121 12 232.5 1.2 -0. 50 0.03
NGC 2173 6 237.4 0.7 -0.42 0.03
NGC 2155 7 309.1 1.6 -0. 50 0.05
NGC 2162 5 322.6 3.5 -0.46 0.07
NGC 2203 9 245.5 1.4 -0.41 0.03
NGC 2193 5 291.2 2.0 -0.49 0.05
NGC 2213 6 242.7 1.2 -0. 52 0.04
Hodge 11 12 245.1 1.0 -1.84 0.04
SL 869 3 258.4 2.1 -0.40 0.04
NGC 2231 9 277.6 1.4 -0. 52 0.03
NGC 2257 16 301.6 0.8 -1.59 0.02
discussion in ~3.6.1). As mentioned previously, we have not derived values for NGC
1861, since it appears to be younger than 1 Gyr (see ~3.6.3).
Positions on the sky for each cluster are shown in Fig. 3-8, along with the
metallicity bin into which each cluster falls, represented by the color of the plotting
symbol. For two of the higher metallicity bins (orange and green symbols), the bin
size is roughly twice the standard error in [Fe/H], so it is possible that cluster errors
could "move" clusters between these and adjacent bins. The adopted center of the
LMC (oc = Sh27nz36s, 6 = -69 52'l2"'; van der Marel et al. 2002) is marked by
the filled square, and the dashed oval represents the 2" near-infrared isopleth from
van der Marel (2001), which roughly outlines the location of the LMC bar. Conversion
from right ascension and declination to Cartesian coordinates was performed using
a zenithal equidistant projection (e.g., van der Marel & Cioni 2001, their equations
1-4); for reference, lines of right ascension and declination are marked with dotted
lines. In Figs. 3-9 and 3-10 we further explore the metallicity-position relationship
for LMC clusters by plotting metallicity as a function of deprojected position angle
and radial distance (in kiloparsecs), respectively. We have corrected for projection
effects by adopting 34P7 as the inclination and 122P5 for the position angle of the
line of nodes of the LMC (van der Marel & Cioni 2001). In this rotated coordinate
system, a cluster with a position angle of zero lies along the line of nodes, and angles
increase counterclockwise; for reference, NGC 2019 has a position angle of ~ 8".
Radial distances were converted from angular separation to kiloparsecs by assuming
an LMC distance of (m -240 = 18.5 (~ 50 kpc); at this distance, l' is ~ 870 pc.
Combined, these three figures illustrate that, similar to what was found by OSSH (and
Geisler et al. 2003), there is no [Fe/H] gradient in terms of either position angle or
radial distance for the higher metallicity clusters in our sample. While we cannot make
strong comments on the metal-poor clusters due to our small sample size, it is well
known that a number of metal-poor clusters ([Fe/H] g -1.5) exist in the inner portions
of the LMC (e.g., OSSH), suggesting that neither the Population I nor the Population II
clusters exhibit a metallicity gradient.
In Fig. 3-10 we have over plotted both the MW open cluster metallicity gradient
from Friel et al. (2002, dashed line) and the M33 gradient from Tiede et al. (2004,
solid line). Neither of these disk abundance gradients resembles what we see among
the LMC clusters. The question of how to interpret this difference takes us to the
work of Zaritsky et al. (1994). They studied the H II region oxygen abundances in
39 disk galaxies. Their data suggest that disk abundance gradients are ubiquitous in
spiral galaxies. However, the presence of a classical bar in the galaxy one that
extends over a significant fraction of the disk length tends to weaken the gradient.
This observation seems to find support in the appearance of Fig. 3-10. In the case of
the LMC, the presence of a strong bar component may have diluted the metallicity
6" 20" 6" 00" 5" 40" 5" 20" 5" 00" 4" 40"
-0 I : I '...'. '... :' ...'....'...l...'...i... .'...I...'.. .' ....;I : '
so -N2162 N194~2 -- -2
5 N2 5 ... ....9 .--------4H dg 4-*
5 -NS..--N2193N2155 SL663 j -64*
~~~~~~~N1846/ N1718';""--. ._g
iv I'~~';"'~V"-...........19.N652 -...
..- SHd6g9e f: ,N2019 \\ N1751 .: -8
E 0 .. --- 5--- ... to ..e f,i 465.1.
g ~ N2213 N2\9 2'. -70"
---- .....'... 6 ......SL41. Sk.
.:: *....~. .. .
0 ...------."'~ '~""~'f6::--- 1 2T46"..'::'.
-5 .--N2203 SL61 .',
:- : *, -74.-6
*~[Fe/j~~~.t-l] +64 ... "
-10 9 0.406 ~[Fe/H]+ -0.}.0 .............-
0.50 >[Fes/lj3'] ~0.6D ',
p.,so [Fe/H3'' -1. 0 .. ......... :..
**il~ $ .. :, : : N1841 ... -800
-6 -4 -2 0 2 4 6
X (degrees from center)
Figure 3-8: Positions on the sky and derived metallicities for our target clusters.
Metallicity bins are given in the lower left corner of the plot. The adopted
LMC center is marked with the filled square, and the dashed line roughly
outlines the bar. See ~3.3.2 for a detailed discussion.
-0.5 e *
O 100 200 300
Deprojected Position Angle (degrees)
Figure 3-9: Cluster metallcity vs. position angle. We plot the metallicities of our
target clusters as a function of deprojected position angle, where we have
used the LMC geometry of van der Marel & Cioni (2001) to correct for
projection effects. This plot illustrates that there is no apparent relation
between position angle and metallicity in the LMC. The error bar shown
in the lower left corner of the plot illustrates the average random error in
[F e/H] .
5 10 15
Deprojected Distance from Center (kpc)
Cluster metallicity vs. radial distance. Cluster metallicities are plotted as
a function of deprojected distance (in kpc) from the center of the LMC.
We have assumed a distance of (m M~1o = 18.5. Overplotted are the
metallicity gradients observed in the MW open clusters (dashed line;
Friel et al. 2002) and M33 (solid line; Tiede et al. 2004), which help
to further illustrate that the LMC's cluster system lacks the metallicity
gradient typically seen in spiral galaxies. This flattened gradient is likely
caused by the presence of the central bar (Zaritsky et al. 1994). As in
Fig. 3-9, the average random error is illustrated by the error bar on the
gradient originally present in the star clusters, leading to a cluster population that
is well mixed. We note that this result is also consistent with the conclusion of
Pagel et al. (1978), who found little evidence for a gradient in oxygen abundance based
on a survey of H II regions within 4 kpc of the LMC center. The Pagel result, that
dlog(O/H)/dR = -0.03 + 0.02 dex kpc parallels our non-detection of a gradient in
To characterize the rotation of their clusters, Schommer et al. (1992) fit an
equation of the form
V(6) = if ,, [tan(6 8o) seci]" +ti }-o. + Em (3-4)
to their radial velocity data using a least-squares technique to derive the systemic
velocity (Vms), the amplitude of the rotation velocity (y ,), and the orientation of the
line of nodes (8o); they adopted an inclination of 27". Their best-fit parameters give
a rotation amplitude and dispersion consistent with the LMC clusters having disk-like
kinematics, with no indications of the existence of a pressure supported halo. We note
that, due to the non-circularity of the LMC, 8o in Eq. 3-4 is not the true orientation of
the line of nodes (the intersection of the plane of the sky and the plane of the LMC),
but rather it marks the line of maximum velocity gradient (van der Marel & Cioni
2001). More recently, van der Marel et al. (2002) used velocities of 1041 carbon
stars to study kinematics in the LMC. Similarly, they found that these stars exhibit a
disk-like rotation with V/o = 2.9 + 0.9, suggesting that these stars reside in a disk that
is slightly thicker than the MW thick disk (V/o a 3.9).
In Fig. 3-11 we have plotted galactocentric radial velocity versus position
angle on the sky for our sample, along with velocity data for all clusters listed in
Schommer et al. (1992). To be consistent with the approach of Schommer et al.
(1992), we have adopted the galactocentric velocity corrections calculated by
Feitzinger & Weiss (1979). Additionally, for this figure only, we have adopted
their LMC center (oc = Sh20nz40s, 6 = -69 14'10"; J2000.0) for use in calculating the
position angles of our clusters. We have used the standard astronomical convention in
which north has a position angle of zero and angles increase to the east; NGC 1942 has
a position angle of ~ 4" in this coordinate system. Data from Schommer et al. (1992)
are plotted as open circles, and our data are plotted as filled stars for the clusters with
previously unpublished velocities and filled circles for the remainder of our clusters;
overplotted on this figure (dashed line) is the rotation curve solution number 3 from
Schommer et al. (1992). For the clusters in common between these two data sets,
we find excellent agreement, with a mean offset of 0. 15 km s where our velocities
are faster than those of Schommer et al. (1992). Additionally, the derived velocities
for the six "new" clusters show that their motions are consistent with the findings of
Schommer et al. (1992) in that the LMC cluster system exhibits disk-like kinematics
that are very similar to the H I disk and has no obvious signature of a stellar halo.
3.4 Comparison with Previous Work
As mentioned in 93.1, OSSH and Suntzeff et al. (1992) have provided the only
previous large-scale, spectroscopic [Fe/H] calculations for clusters in the LMC. Similar
to our work, they utilized the CaT lines as a proxy for measuring Fe abundance
directly, but with two important differences: they used the absolute magnitude of their
stars, based on the spectral intensity at 8600 A+, as a surface gravity estimator instead
of V VHB, and their [Fe/H] calibration was based largely on the Zinn & West (1984)
metallicity system, with the addition of two open clusters that have metallicities derived
from various spectrophotometric indices (see their Table 7). This introduced two
systematic offsets that make it inappropriate to directly compare the OSSH values to
our work and other recently measured CaT abundances: first, the use ofM8S600 creates
a dependence on the relative distances of the calibrating clusters and the LMC, and the
globular cluster distance scale has been much revised in the post-Hipparcos era (Reid
100 200 300
Position Angle (degrees)
\ o oO
O h 0"0 \ \e ~,3)
Cluster radial velocity vs. position angle. Galactocentric radial velocities
as a function of position angle on the sky are plotted for the clusters in
our sample (filled symbols) as well as those from Schommer et al. (1992,
open circles). The six clusters in our sample with no previous velocity
determinations are plotted as filled stars and all others in our sample are
filled circles. Rotation curve solution number 3 from Schommer et al.
(1992) is overplotted as the dashed line, showing that both data sets are
consistent with circular rotation. We note that we have not plotted a rep-
resentative error bar since our plotting symbols are roughly the same size
as the average random velocity error.
1999). Second, it has been shown (e.g., Rutledge et al. 1997a) that the Zinn & West
scale is non-linear compared to the more recent Carretta & Gratton (1997) scale based
on high-resolution spectra of globular cluster red giants. To put the work of OSSH on
the Carretta & Gratton system, Cole et al. (2005) perform a non-linear least-squares fit
to calibration clusters in common with their work and that of OSSH. They find that
one can estimate the abundance of OSSH clusters on the metallicity system we have
used via the following conversion:
[Fe/H] a -0.212 +t 0.498 [Fe/H]OSSH 0. 128 [Fe/H/OSS.(35
This equation approximates the metallicity that OSSH would have derived from their
spectroscopic data and calibration procedure but with updated metallicities for their
calibration clusters; it does not attempt to account for any other differences in the
treatment of the data.
In columns 3 and 4 of Table 3-4 we list [Fe/H] for clusters in OSSH and
Suntzeff et al. (1992) in common with our target clusters, where column 3 gives
their published values and in column 4 we have converted their numbers onto our
metallicity system using Eq. 3-5. The number of stars used by OSSH in calculating
final cluster metallicities is given in parentheses in column 4, and our derived metallic-
ities are given in column 2 for reference. In Fig. 3-12 we plot the difference between
our metallicities and their converted [Fe/H] values as a function of our metallicities.
OSSH give their [Fe/H] errors for an individual star as 0.2 dex; therefore, deviations
between these data sets as large as + 0.2 are not unexpected, suggesting that these
results are in relative agreement, with no offset. We note, however, that even with the
use of Eq. 3-5, it is very difficult to directly compare the derived cluster abundances
because of the differences in target selection and calibration strategy.
While a direct comparison of [Fe/H] values is difficult, we can readily compare
the metallicity distributions of these two data sets. As such, in Fig. 3-13 we have
converted onto our system
d onto our systent using
Table 3-4. Publish
ed LMC Cluster Metallicities
[Fe H]a [Fe H]b [Fe H]
CaT CaT High-Res.
aFront OSSH, unless otherwise noted.
bFront OSSH, unless otherwise noted,
using Eq. 3-5.
CFront Suntzeff et al. (1992).
dFront Suntzeff et al. (1992), converted
eFroni Hill (2004)
fFront Johnson et al. (2006)
[Fe/H] (This work)
Metallicity comparison with OSSH. We compare derived metallicities for
clusters in common between our study and that of OSSH. We note that
[Fe/H] values from OSSH were converted onto the metallicity scale we
have used via Eq. 3-5. This comparison shows that, to within the errors,
there is relatively good agreement between our results and those of OSSH
(see ~3.4 for more details).
plotted the metallicity distribution of OSSH's raw data (top panel), converted [Fe/H]
values (middle panel), and our results (bottom panel). The dark shaded histogram
shows only the 20 clusters in common between the three panels, while the lighter
histogram plots all the clusters in each sample. From this figure it is clear that both
the raw and converted OSSH samples show an extended distribution of intermediate-
metallicity clusters, whereas our cluster sample exhibits a very tight distribution. For
the 20 clusters in common, we find a mean [Fe/H] = -0.47 with 0- = 0.06, while
the converted OSSH metallicities give [Fe/H] = -0.42 +0. 14. Our tight metallicity
distribution, with a lack of higher metallicity clusters ([Fe/H] X -0.30), is an important
feature of our data for the following reason. Chemical evolution models suggest that
metallicity is a rough estimator of age, in that younger stellar populations should be
more metal-rich than older populations, since there has been more time to process
material and enrich the interstellar medium. Thus, intermediate-age clusters should
be more metal-poor than younger stellar populations in the LMC. However, some
intermediate-age clusters in the sample of OSSH appeared to be more metal-rich than
much younger stellar populations in the LMC, which would indicate the presence
of a large spread of metallicities at any given age. In Table 3-5 we give the mean
metallicity and spread of our entire sample of intermediate-age clusters and all clusters
in OSSH with converted metallicities above -1.0 dex, along with published results for
a sample of younger stellar populations (e.g., B dwarfs, Rolleston et al. 2002; Cepheid
variables, Luck et al. 1998; young red giants, Smith et al. 2002) and intermediate age
RGB field stars in the LMC bar (Cole et al. 2005). This table shows that, as we would
expect from chemical enrichment models, the intermediate-age clusters are slightly
more metal-poor than the younger populations in the LMC. Thus, the much tighter
metallicity distribution seen in our clusters is in excellent agreement with the expected
chemical enrichment pattern in the LMC and alleviates the problem created by the high
metallicity tail of intermediate-age clusters in the OSSH results. In addition, Table
3-5 shows that our intermediate-age clusters have a mean metallicity and distribution
similar to that of the metal-rich component of the bar field studied by Cole et al.
(2005). The similarity between these two populations is in good agreement with the
models of Bekki et al. (2004), in which the formation of the LMC bar and the restart
of cluster formation (the end of the age gap) are both a result of the same very close
encounter with the SMC.
Finally, in Table 3-4 we have also included [Fe/H] values derived from high-
resolution spectra for NGC 1841 and NGC 2257 from Hill (2004) and NGC 2019
and Hodge 11 from Johnson et al. (2006). For the two clusters from Johnson et al.
(2006), we list [Fe/H] values that are the average of their metallicities determined from
Fe I and Fe II lines, and the number of stars observed in each cluster is given. Two
clusters, NGC 1841 and NGC 2019, show good agreement between our metallicities,
calculated from the CaT lines, and metallicities derived from fitting to high-resolution
spectra. In contrast, Hodge 11 and NGC 2257 show a roughly 0.3 dex offset between
these methods in the sense that our values are more metal-rich than the results from
high-resolution spectra. Similarly, a preliminary result for ESO 121, which is more
metal-rich than the aforementioned clusters, suggests an offset in the same direction,
where the CaT method gives a [Fe/H] value higher than what is measured with high-
resolution spectra (A. A. Cole, private communication). It has been suggested that
variations in [Ca/Fe] between calibrating clusters in the MW and target clusters in
the LMC may cause a breakdown in the utility of CaT lines as a metallicity indicator.
However, abundances based on high-resolution spectra show that [Ca/Fe] is typically
lower for LMC cluster giants than for MW giants of the same [Fe/H], which is in the
opposite direction of what is needed to explain the difference between CaT and high-
resolution results. We also note that, for low-metallicity stars, previous authors have
shown that metallicities derived from high-resolution spectra can vary considerably (0.3
dex is not uncommon), depending on which ionization stages, what temperature scale,
and what model atmospheres are being used (e.g., Johnson et al. 2006; Kraft & Ivans
.. Row [Fe/H] from OSSH
..Converted [Fe/H] from OSSH
: I I I------
[Fe/H] from this paper
Metallicity distribution of LMC clusters as determined by OSSH and
this paper. Published values from OSSH are given in the top panel, while
the middle panel shows their values converted onto our metallicity scale
using Eq. 3-5; in the bottom panel we have plotted our results. In all
three panels, the dark shaded region shows the distribution for the 20
clusters in common between OSSH and this paper, while the light shaded
region shows the entire cluster sample from each study. Our results in-
dicate that the LMC's intermediate-age cluster metallicity distribution is
actually much tighter than suggested by the results of OSSH.
Table 3-5. Metallicities of Young and Intermediate-Age Stellar Populations
Population Age Estimate [Fe/H] G[Fe/H] Reference
B dwarfs <20 -0.31 0.04 Rolleston et al. (2002)
Cepheid variables 10-60 -0.34 0.15 Luck et al. (1998)
Young RGB stars 200-1000 -0.45 0.10 Smith et al. (2002)
Intermediate-age clusters 1000 -3000 -0.48 0.09 This paper
Intermediate-age clusters 1000 -3000 -0.48 0.17 OSSH
Bar RGB stars, metal-rich 1000-5000 -0.37 0.15 Cole et al. (2005)
Bar RGB stars, metal-poor X5000 -1.08 0.47 Cole et al. (2005)
As discussed in 93.1, determining abundances for populous clusters within
the LMC is an important step in understanding the history of this satellite galaxy.
Accurate [Fe/H] values help to break the age-metallicity degeneracy that arises when
trying to fit theoretical isochrones to cluster CMDs, which allows the unequivocal
determination of cluster ages, thereby providing a clear picture of the LMC's cluster
age-metallicity relation. These clusters also serve to fill a region of the age-metallicity
plane that is void of MW clusters; this makes the LMC cluster system an important
testbed for a variety of stellar population models. Additionally, in a previous paper
Grocholski & Sarajedini 2002, we showed that knowledge of a cluster's age and
metallicity is essential to predicting the K-band luminosity of the RC for use as a
standard candle. In a future work we will use the metallicities derived herein to
determine distances to individual populous LMC clusters, which will allow us to
compare the cluster distribution to the LMC geometry calculated from field stars (e.g.,
van der Marel & Cioni 2001).
In this chapter we have presented the results of our spectroscopic study of the
near-infrared Ca II triplet lines in individual RGB stars in 28 populous LMC clusters.
Utilizing the multi-object spectrograph, FORS2, on the VLT, we have been able to
determine membership and calculate metallicities and radial velocities for, on average,
eight stars per cluster, with small random errors (1.6 km sl in velocity and 0.04 dex
in [Fe/H]). The number of cluster members observed, combined with the updated
CaT calibration of CO4 (they extended the calibration to younger and more metal rich
clusters than previous work), has allowed us to improve on the work of OSSH, which
is the only previous large scale spectroscopic study of individual cluster stars within the
LMC. The main results of our paper are as follows:
1. We report the first spectroscopically derived metallicities and radial velocities
for the following clusters: SL 4, SL 41, SL 396, SL 663, SL 869, and Hodge 3. In
addition, NGC 1718 and NGC 2193 have no previously reported spectroscopic [Fe/H]
2. NGC 1718 is the only cluster in our sample that falls into the range -1.3 <
[Fe/H] I -0.6. This metallicity region corresponds to the well known 3-13 Gyr "age
gap," within which there is only one cluster, ESO 121. However, unlike ESO 121, the
CMD of NGC 1718 suggests an age (~ 2 Gyr) much younger than the age gap; we
use archival Hubble Space Telescope (HST) Wide Field Planetary Camera (WFPC2)
photometry to investigate this point in the ~3.6.1. This age makes NGC 1718 one of
the most metal-poor intermediate-age clusters in the LMC.
3. The intermediate-age clusters in our sample show a very tight distribution,
with a mean metallicity of -0.48 dex (0 = 0.09) and no clusters with metallicities
approaching solar. While this is in contrast to previous cluster results, it suggests that
the formation history of the bar (mean [Fe/H] = -0.37, o = 0. 15; Cole et al. 2005)
is very similar to that of the clusters. This agrees well with the theoretical work of
Bekki et al. (2004), which indicates that a close encounter between the LMC and SMC
caused not only the restart of cluster formation in the LMC but the generation of the
central bar as well.
4. Similar to previous work, we find no evidence for the existence of a metallicity
gradient in the LMC cluster system. This is in stark contrast to the stellar populations
of both the MW and M33, which show that metallicity decreases as galactocentric
distance increases; the LMC's stellar bar is likely responsible for the well-mixed cluster
5. We find that our derived cluster velocities, including the six "new" clusters, are
in good agreement with the results of Schommer et al. (1992) in that the LMC cluster
system exhibits disk-like rotation with no clusters appearing to have halo kinematics.
6. Comparing our results for four clusters to [Fe/H] values recently derived
through high-resolution spectra, we find that two of the four clusters are in good
agreement, while the other two have [Fe/H] values derived via the CaT method that are
~ 0.3 dex more metal-rich than what is found from high-resolution spectra; a similar
effect is seen in preliminary results for an additional two LMC clusters. The source of
this difference is unclear, and it is not immediately explained by variations in [Ca/Fe]
between the CaT calibration clusters in the MW and the LMC target clusters. Further
high-resolution studies, especially covering the LMC's intermediate-age clusters, are
needed to fully address this issue.
3.6 Notes on Individual Clusters
3.6.1 NGC 1718
While only three of the stars observed in NGC 1718 appear to be cluster mem-
bers, these stars are, on average, 0.3 dex more metal-poor than all but one of the other
stars observed in this field. As mentioned in ~3.3.2, this causes NGC 1718 to occupy
an interesting position in the LMC's age-metallicity relation; its metallicity is com-
parable to that of ESO 121, which seems to be the only cluster residing in the LMC
having an age between ~ 3 and 13 Gyr (Da Costa 2002). The cluster CMD resulting
from our aperture photometry is not well populated around the MSTO, so we have
used archival HS77WFPC2 data (GO-5475) to create a cluster CMD reaching below the
MSTO. The images were reduced using the procedure outlined by Sarajedini (1998).
In summary, all detected stars on the Planetary Camera CCD were photometered in the
F450W and F555W filters using a small aperture. These were then corrected to a 0"'5
0.0 0.5 1.0 1.5
Cluster CMD for NGC 1718, based on aperture photometry of archival
HST/WFPC2 images. We overplot isochrones of 1.3, 2.0, and 2.5 Gyr
(top to bottom) from Girardi et al. (2002) that have a metallicity (~ -0.7
dex) similar to the value we have derived for this cluster (-0.8 dex).
Although this cluster has a metallicity similar to that of ESO 121, the
isochrones suggest an age of ~ 2.0 Gyr for this cluster, leaving ESO 121
as the only known LMC cluster with an age between approximately 3
and 13 Gyr.
radius, adjusted for the exposure time, and transformed to the standard system using
the equations from Holtzman et al. (1995). In Fig. 3-14 we present the CMD of NGC
1718 with isochrones from Girardi et al. (2002) overplotted; the isochrones have [Fe/H]
S-0.7, close to our measured cluster value of -0.8 dex, and ages ranging from 1.3 to
2.5 Gyr. This figure suggests that NGC 1718 has an age of roughly 2.0 Gyr, making
it an intermediate-age cluster and leaving ESO 121 as still the only cluster known to
occupy the LMC's cluster age gap. However, the existence of an intermediate-age
cluster at this low metallicity is intriguing, as it indicates that some pockets of unen-
riched material must have remained intact even though most of the gas that formed the
intermediate-age clusters was well mixed.
3.6.2 NGC 1846
Given the sloped appearance of the RC and the width of the RGB (see Fig. 3-28),
NGC 1846 is suffering from differential reddening, making it difficult to accurately
measure the true location of the cluster RC, as well as V VHB for target stars.
To address this problem, we make no adjustments to the instrumental magnitudes,
but we measure the median magnitude of the entire differentially reddened RC,
effectively measuring the RC at the mean reddening of the cluster. Since the amount of
extinction suffered by the RGB stars should be scattered about the mean reddening, this
approach smooths over the differential reddening, allowing us to accurately measure
the cluster metallicity. We note that this method increases the scatter in [Fe/H] for
cluster members; as such, we have relaxed the metallicity cut in our member selection
method to include all stars moving at the radial velocity of the cluster. For reference,
if V VHB for any given star is off by & 0.2 mag (we estimate that the differential
reddening is 0.4 mag in V), the effect on [Fe/H] for that star is roughly & 0.05.
3.6.3 NGC 1861
This cluster is listed as SWB type IVB, suggesting an age range of 0.4-0.8 Gyr
(Bica et al. 1996), which is roughly the age at which the RC first forms (~ 0.5 Gyr;
Girardi & Salaris 2001). Plotting a CMD of stars within the apparent cluster radius
reveals what appears to be a fairly young MSTO in addition to no obvious cluster RC
or RGB. Therefore, we assume that NGC 1861 is a young cluster and all observed
RGB stars are actually part of an older field population.
I l l
50 100 150
[Fe/H] = -0.41 *
1 0 -1 -2 -3
-t -i 4- -1 -
..I .. II .I
50 100 150
500 1000 1500
x position (pixels)
IC 2146 cluster member selection. In this figure we illustrate our clus-
ter member selection process for IC 2146, using a combination of a
star's distance from the cluster center (upper left), radial velocity (upper
right), and metallicity (lower left) to separate field stars (blue, teal, and
green points) from cluster members (red points; see text for a complete
discussion of the color code). The lower right panel plots the summed
equivalent width as a function of V VHB for all stars considered to be
cluster members; the dashed line is an isoabundance line at the mean
metallicity of the cluster.
IC 2146 cluster and field CMD. Shown is the instrumental CMD for the
entire IC 2146 field (cluster and surrounding field stars), with the target
stars marked as described in the text; cluster members (red points) lie
along the RGB, AGB or in the RC.
I I .' I .
1 I I I
NGC 1 651
20 40 60 80 100 120
[Fe/H] = -0.53
1 0 -1 -2 -3
500 1000 1500
x position (pixels)
20 40 60 80 100 120
Figure 3-17: NGC 1651 cluster member selection. Same as Fig. 3-15 except the plots
shown are for NGC 1651.
NGC 1651 **
.. .. ..
*I Y 9*
. .- ,
NGC 1651 cluster and field CMD. Same as Fig. 3-16 except the CMD
shown is for the entire NGC 1651 field.
50 100 150
[Fe/H] = -0.46
1 0 -1 -2 -3
~.. I ....I.
I0 10 5
500 1000 1500
x position (pixels)
Figure 3-19: NGC 1652 cluster member selection. Same as Fig. 3-15 except the plots
shown are for NGC 1652.
1 1 1 1 l i l l l i l l l i l l l i l
.... *, .
.s ..* *. *
... ''..*.-'.e..'I..'....."I'..... ''
NGC 1652 cluster and field CMD. Same as Fig.
shown is for the entire NGC 1652 field.
3-16 except the CMD
500 1000 1500
x position (pixels)
j I i
20 40 60 80 100 120 140
20 40 60 80 100 120 140
[Fe/H] = -0.80
1 0 -1 -2 -3
- -t -
Figure 3-21: NGC 1718 cluster member selection. Same as Fig. 3-15 except the plots
shown are for NGC 1718.
10 I I I
.*:: ;* *,*,~ g
to~, .- :.. *g
.c 14 -" : -. --.. -
..' ~ ~ ** n 5r q:'. .:. ** .
18 ~ ~ ~ ~ ~ ~ : ...I I I . .
-1.5t -10- .
v-i 2' (inst)
Fiue32:NG 78cute n il CMD.. Same as Fi. 31 xeph M
shw sfr the' enie G 11fed
- ... ~-.,
. r S~C~;.-*
:s..Y;i.~*~: ~; -"~ iL:.::
c "~' I.
;- .1~ i
20 40 60 80 100 120 140
[Fe/H] = -0.44
1 0 -1 -2 -3
500 1000 1500
x position (pixels)
20 40 60 80 100 120 140
Figure 3-23: NGC 1751 cluster member selection. Same as Fig. 3-15 except the plots
shown are for NGC 1751.
16 -~ ..
18 *,- '"
Figre -24 NG 171 custr ad feld
shon i fo th enireNGC
0.0 0.5 1.0
CMD. Same as Fig. 3-16 except the CMD