Spitzer SAGE survey of the Large Magellanic Cloud II: Evolved Stars and Infrared Color Magnitude Diagrams

Abstract

Color-magnitude diagrams (CMDs) are presented for the Spitzer SAGE (Surveying the Agents of a Galaxy's Evolution) survey of the Large Magellanic Cloud (LMC). IRAC and MIPS 24 um epoch one data are presented. These data represent the deepest, widest mid-infrared CMDs of their kind ever produced in the LMC. Combined with the 2MASS survey, the diagrams are used to delineate the evolved stellar populations in the Large Magellanic Cloud as well as Galactic foreground and extragalactic background populations. Some 32000 evolved stars brighter than the tip of the red giant branch are identified. Of these, approximately 17500 are classified as oxygen-rich, 7000 carbon-rich, and another 1200 as ``extreme'' asymptotic giant branch (AGB) stars. Brighter members of the latter group have been called ``obscured'' AGB stars in the literature owing to their dusty circumstellar envelopes. A large number (1200) of luminous oxygen--rich AGB stars/M supergiants are also identified. Finally, there is strong evidence from the 24 um MIPS channel that previously unexplored, lower luminosity oxygen-rich AGB stars contribute significantly to the mass loss budget of the LMC (1200 such sources are identified).

Full text

Page 1

arXiv:astro-ph/0608189v1 8 Aug 2006
Spitzer SAGE survey of the Large Magellanic Cloud II: Evolved
Stars and Infrared Color Magnitude Diagrams
R. D. Blum1, J. R. Mould2, K. A. Olsen1, J. A. Frogel3, M. Werner4, M. Meixner5, F.
Markwick–Kemper6, R. Indebetouw6, B. Whitney7, M. Meade8, B. Babler8, E. B.
Churchwell8, K. Gordon9, C. Engelbracht9, B–Q For9, K. Misselt9, U. Vijh5, C. Leitherer5,
K. Volk10, S. Points1, W. Reach11, J. L. Hora12, J–P. Bernard13, F. Boulanger14, S. Bracker8,
M. Cohen15, Y. Fukui16, J. Gallagher8, V. Gorjian11, J. Harris9, D. Kelly9, A. Kawamura16,
W. B. Latter17, S. Madden12, A. Mizuno16, N. Mizuno16, A. Nota5, M. S. Oey18, T.
Onishi16, R. Paladini13, N. Panagia5, P. Perez-Gonzalez9, H. Shibai16, S. Sato16, L. Smith19,
L. Staveley-Smith20, A.G.G.M. Tielens21, T. Ueta22, S. Van Dyk11, and D. Zaritsky9
ABSTRACT
Color–magnitude diagrams (CMDs) are presented for the Spitzer SAGE (Sur-
veying the Agents of a Galaxy’s Evolution) survey of the Large Magellanic Cloud
(LMC). IRAC and MIPS 24 µm epoch one data are presented. These data rep-
resent the deepest, widest mid–infrared CMDs of their kind ever produced in
the LMC. Combined with the 2MASS survey, the diagrams are used to delineate
the evolved stellar populations in the Large Magellanic Cloud as well as Galac-
tic foreground and extragalactic background populations. Some 32000 evolved
stars brighter than the tip of the red giant branch are identified. Of these, ap-
proximately 17500 are classified as oxygen–rich, 7000 carbon–rich, and another
1200 as “extreme” asymptotic giant branch (AGB) stars. Brighter members of
the latter group have been called “obscured” AGB stars in the literature ow-
ing to their dusty circumstellar envelopes. A large number (1200) of luminous
oxygen–rich AGB stars/M supergiants are also identified. Finally, there is strong
evidence from the 24 µm MIPS channel that previously unexplored, lower lumi-
nosity oxygen–rich AGB stars contribute significantly to the mass loss budget of
the LMC (1200 such sources are identified).
Subject headings: stars: AGB and post-AGB,stars: mass loss,stars: carbon,infrared:
stars,(galaxies:) Magellanic Clouds

Page 2

– 2 –
1.Introduction
The Spitzer Space Telescope (Spitzer), with its rapid wide field mapping capability and
sensitivity, allows us to survey the thermal emission produced by mass loss from individual
evolved stars in the nearest galaxies. Meixner et al. (2006) presented the Spitzer SAGE
(Surveying the Agents of a Galaxy’s Evolution) survey of the Large Magellanic Cloud (LMC)
along with a detailed description of its goals, expected extent in depth and coverage and
some preliminary results. The SAGE survey is designed to enable studies of the life–cycle of
1Cerro Tololo Interamerican Observatory, Casilla 603, La Serena, Chile
2NOAO, PO Box 26732, Tucson AZ 85726-6732
3AURA, Inc., 1200 New York Ave. NW, Suite 350, Washington D.C. 20005
4Jet Propulsion Lab, 4800 oak Grove Dr., MS 264–767, Pasadena, CA 91109
5Space Telescope Science Institute, 3700 San Martin Way, Baltimore, MD 21218
6Department of Astronomy, University of Virginia, PO Box 3818, Charlottesville, VA 22903
7Space Science Institute, 308 Morningside Ave., Madison, WI 53716
8Department of Astronomy, 475 North Charter St., University of Wisconsin, Madison, WI 53706
9Steward Observatory, University of Arizona, 933 North Cherry Ave., Tucson, AZ 85719
10Gemini Observatory, 670 North A’ohuku Place, Hilo, HI 96720
11Spitzer Science Center, California Institute of Technology, 220-6, Pasadena, CA, 91125
12Center for Astrophysics, 60 Garden St., MS 67 , Harvard University, Cambridge, MA 02138
13Centre d’´Etude Spatiale des Rayonnements, CNRS, 9 av. du Colonel Roche, BP 4346, 31028 Toulouse,
France
14Institut d’Astrophysique Spatiale, Universit´ e Paris-XI, 91405 Orsay Cedex, France
15Radio Astronomy Laboratory, 601 Campbell Hall, University of California at Berkeley, Berkeley, CA
94720
16Department of Astrophysics, Nagoya University, Chikusa-ku, Nagoya 464-8602, Japan
17Caltech, NASA Herschel Science Center, MS 100–22,Pasadena, CA 91125
18Department of Astronomy, University of Michigan, 830 Dennison Bldg., Ann Arbor, MI 48109
19Department of Physics and Astronomy, University College London, Gower Street, London WC1E 6BT
20Australia Telescope National Facility, CSIRO, P. O. Box 76, Epping NSW 1710, Australia
21Kapteyn Institute, P.O. Box 800, NL-9700 AV Groningen, Netherlands
22NASA Ames Research Center/SOFIA, Mail Stop 211-3, Moffett Field, CA 94035

Page 3

– 3 –
baryonic matter, as traced by dust emission, in the LMC. In this work, we focus on one corner
stone of this life–cycle, the dusty evolved star population, which is returning matter to the
interstellar medium. Coupled with detailed star formation history (SFH) studies at optical
wavelengths, SAGE provides the opportunity to link specific LMC populations to their dust
content as a function of key parameters such as metallicity and age. Such analyses can, in
turn, be used to validate predictive models of infrared emission in high redshift galaxies. In
the present paper, we focus on the necessary first step of identification and characterization
of the dusty evolved star populations.
SAGE provides a global view of the dust producing stellar populations in the LMC
that will contribute to our understanding of evolved stars. Essentially all significant, dusty,
mass-losing stars in the LMC will be detected by SAGE as it provides the deepest, widest
survey in the near– and mid-infrared wavelengths (3–24 microns) produced to date for this
nearby galaxy. These wavelengths are particularly well suited to the investigation of the late
stages of stellar evolution because once beyond the near–infrared (five microns), emission
from circumstellar dust can become the dominant source of emission over the radiation from
the photosphere. By eight microns, many LMC sources show appreciable emission from
dust, and the MIPS 24 micron band is extremely sensitive to cool envelopes with excess dust
emission even for sources with little evidence of strong dust excess at shorter wavelengths.
SAGE builds on and leverages a wealth of past observations of evolved stars in the LMC.
Optical (MCPS; see Zaritsky et al. 2004, and http://ngala.as.arizona.edu/dennis/mcsurvey.html)
and near–infrared surveys, for example, DENIS (Epchtein et al. 1994) and 2MASS (Skrutskie
et al. 2006), have provided a global view of the LMC at shorter wavelengths. See Cioni et al.
(2006) and Nikolaev & Weinberg (2000) for LMC data from these two surveys, respectively.
The Infrared Astronomy Satellite (IRAS, Neugebauer et al. 1984) revolutionized the study
of evolved stars by observing objects with luminous dusty envelopes at long wavelengths.
This survey has provided samples of stars in the LMC to be observed in greater detail up to
the present, including spectroscopic observations with Spitzer; see Zijlstra et al. (2006) and
Markwick–Kemper et al. (2005). However, IRAS was only sensitive enough to observe the
brightest sources in the LMC (Schwering 1989) from which several hundreds of mass–losing
evolved star candidates could be identified (Loup et al. 1997). The successful Midcourse
Space Experiment (MSX, Price et al. 2001) was about four times more sensitive and pro-
vided a full LMC survey (Egan et al. 2001). SAGE should be roughly 1000 times more
sensitive than MSX when complete (Meixner et al. 2006, the present data are roughly 400
times more sensitive). While near–infrared and optical surveys of the LMC have identified
lower mass evolved stars, only SAGE has the depth to detect such stars at wavelengths
necessary to explicitly explore their dust properties. The ISOCAM Magellanic Cloud Mini–
Survey (see Cioni et al. 2003) has provided a preliminary mid–infrared view of the LMC over

Page 4

– 4 –
a small field of view (0.3 square degrees) near the depth of SAGE.
The SAGE sensitivity and areal coverage of the entire LMC will allow a detailed quan-
titative derivation of the global mass loss budget from all stellar populations when combined
with existing and future mid–infrared spectroscopic observations of the asymptotic giant
branch (AGB) stars and supergiants; for example, see van Loon et al. (1999), van Loon et
al. (2005), Markwick–Kemper et al. (2005), and Zijlstra et al. (2006). Several studies have
derived detailed SFHs in the LMC from a variety of field locations and clusters (Holtzman
et al. 1999; Olsen 1999) using principally optical color–magnitude diagrams (CMDs) and
sophisticated stellar models and fitting techniques (e.g. Harris & Zaritsky 2001). Recently,
Cioni et al. (2006) made a global SFH calculation using the 2MASS and DENIS surveys.
Their new models still have difficulties producing the correct mix of observed carbon–rich
AGB stars (C–rich or C–stars) versus oxygen–rich AGB stars (O–rich or M stars), but their
extension to these wavelengths of the general technique is a great advance. With SAGE,
we will be able to associate the mass–loss and chemical properties of the AGB stars in the
LMC with their known evolutionary state from existing SFH studies. This coupling can
then be used to investigate, with precision, the evolutionary status of galaxies which lie at
much larger distances through the development and testing of detailed population synthesis
models (for example, see Mouhcine & Lan¸ con 2003).
In this paper, we continue the analysis begun by Meixner et al. (2006) by presenting
CMDs of the survey data for the first of two epochs. We lay the ground work for the detailed
analysis to come when the full SAGE survey is available by dissecting the observed CMDs
and indicating where various LMC and fore and background populations reside in these
diagrams. The present paper covers approximately 49 square degrees in the IRAC bands
and MIPS 24 µm band.
2. The SAGE Catalog and Epoch One Source List
The SAGE epoch one catalog is discussed by Meixner et al. (2006). The present source
list contains the first epoch data for the IRAC (3.6 µm, 4.5 µm, 5.8 µm, and 8.0 µm)
bands processed by the IRAC pipeline as of May 04, 2006. This amounts to a coverage
of approximately 49 square degrees (Figure 1). The current IRAC source list includes a
merged source list with the 2MASS point source catalog. Details of the SAGE IRAC pipeline
processing are given by Meixner et al. (2006); the pipeline is based on the GLIMPSE pipeline
(Benjamin et al. 2003). Briefly, the IRAC data were taken such that a typical point on the
sky was observed twice within a ∼ one degree square “tile.” The frames are analyzed with a
custom DAOPHOT (Stetson 1987) package developed by the GLIMPSE team. The catalog

Page 5

– 5 –
photometry is based upon combining the individual frame photometry (i.e., photometry is
not done on the mosaic tiles). Each tile itself overlaps at the edges with adjacent tiles
and the comparison of source photometry between tiles results in more detections in the
overlap regions. The IRAC bands and 2MASS catalog photometry are merged in the so–
called “cross–band” merge step of the pipeline which is based upon positional matching (see
Meixner et al. 2006, and references therein for details) of the final flux calibrated sources in
each IRAC and 2MASS band. Systematic offsets in the IRAC–2MASS merging are less than
0.3′′.
The combined IRAC–2MASS source list analyzed in this paper contains approximately
four million sources with a 3.6 µm data point (the deepest IRAC band in terms of SAGE
photometry). There are approximately 820000 sources with J and [3.6] magnitudes of which
nearly 650000 have errors less than 0.1 mag in each of J and [3.6]. An additional 10000
sources have individual J and [3.6] errors less than 0.2 mag and the remainder have errors
of less than 0.3 mag. The present dataset further has 250000 8 µm sources which also have
3.6 µm data (90% have [8.0] and [3.6] photometry with errors less than 0.1 mag). The
analogous numbers for J and 8 µm are 230000 and 91%. The IRAC uncertainties are those
reported by the custom DAOPHOT package used in the pipeline and so include photon
statistics and fitting uncertainties. The 2MASS errors are taken from the 2MASS point
source catalog. The initial frames delivered from the Spitzer Science Center (SSC) are flux
calibrated (SSC pipeline version s12.0). A network of 238 absolute calibration stars has been
developed near the area observed by SAGE (137 stars overlap the SAGE survey area) using
the identical technique by which IRAC primary standards were established (Cohen et al.
2003). Comparison between predicted and observed magnitudes for stars in this network
indicate that ensemble systematic uncertainties in any IRAC band do not exceed the five
percent level.
The present source list for the MIPS 24 µm band covers a similar area as for the IRAC
source list (Figure 2). The 24 µm source list (64800 sources) is for epoch one data processed
as of May, 19 2006. Details of the MIPS processing are given by Gordon et al. (2005) and
Meixner et al. (2006). The MIPS data were observed in sets of scans with each set covering
a four degree by 25 arcminute strip. The photometry was extracted via PSF fitting using
the StarFinder program (Diolaiti et al. 2000).
The IRAC and MIPS data presented here were merged after assembly of the individual
catalogs by the respective pipeline teams. The IRAC sources (those with valid 8 µm magni-
tudes) were matched within 1′′of the MIPS 24 µm sources, choosing the closest companion
if multiple possibilities existed. This resulted in 27741 matches with a an average difference
in position of 0.42′′(and systematic offsets in RA of −0.024′′and DEC of −0.004′′) and a

Page 6

– 6 –
distribution of offsets with one sigma standard deviation of approximately 0.25′′. The 24
µm pixel size is 2.6′′and the diffraction limited λ/D is approximately 6′′. Crowding induced
errors in the IRAC photometry and the matching of sources between IRAC and MIPS are
discussed in appendix A below.
About 42% of the MIPS sources have IRAC matches. What about the other 58%? This
number of non–matches suggests a large number of very red sources toward the LMC. As
the analysis below will suggest, some of these may be background galaxies not detected by
IRAC. Others could be embedded sources in the LMC, and many could be slightly extended
sources which have no point–like IRAC counterpart. Detailed analysis of the 24 µm and
other MIPS channels will be the subject of a subsequent SAGE paper.
3. Color–Magnitude Diagrams
Figures 3–6 present four Hess diagrams (Hess diagrams are a two dimensional histogram
of the CMD which give the number of stars per unit magnitude and color with the grayscale
representations graphically showing the relative source densities) and their corresponding
CMDs produced with IRAC, MIPS, and 2MASS photometry. The Hess diagrams are binned
into 200×200 pixels (i.e., bins of color and magnitude).
There is a wealth of information in these diagrams. Egan et al. (2001) presented near–
infrared+mid–infrared CMDs for the entire LMC from the MSX satellite and 2MASS (Skrut-
skie et al. 2006). The MSX sensitivity is approximately eight magnitudes brighter than
SAGE, so that the present survey will greatly extend the mid–infrared view of the LMC
begun by IRAS and MSX.
In the following, all near–infrared photometry has been taken from the 2MASS point
source catalog (Skrutskie et al. 2006). These data have been merged with the SAGE pho-
tometry using the SAGE IRAC pipeline (see above). The diagrams in this paper use IRAC
and MIPS photometry converted to magnitudes using the following zero points (Reach et
al. 2005; Engelbracht et al. 2006): [3.6] zero mag = 280.9 Jy, λ◦= 3.55 µm; [8.0] zero mag
= 64.13 Jy, λ◦= 7.872 µm, [24] zero mag = 7.15 Jy, λ◦= 23.68 µm.
3.1.IRAC and 2MASS Color–Magnitude Diagrams
Earlier “all–LMC” surveys by IRAS and MSX had sensitivities 1000 times less than
SAGE. While these surveys could detect the brightest sources in the LMC, it has been more
difficult to put them in the overall context of the entire stellar population. In the following

Page 7

– 7 –
sections we analize the SAGE CMDs with an emphasis on identifying the main evolved star
populations and enumerating their relative contributions to the total source counts. This
basic accounting will help guide future detailed studies using the SAGE data base as well as
provide a comprehensive means for selecting follow–up targets for spectroscopic study.
3.1.1.
J−[3.6] vs. [3.6]
The 3.6 µm channel is by far the most sensitive of the IRAC bands and as such provides
the deepest photometry to compare to 2MASS. Nikolaev & Weinberg (2000) presented the
2MASS CMD for the LMC. Comparison of their Figure 3 with the J−[3.6] vs. [3.6] diagram
shown in Figure 3 shows the same sequences with perhaps slightly more “contrast” in the
features owing to the longer baseline in wavelength. Starting with J−[3.6] vs. [3.6], the
first prominent finger (J−[3.6]∼<0.5) reaching to bright magnitudes ([3.6]∼6) corresponds to
region “B” of Nikolaev & Weinberg (2000), young A–G supergiants. There is clearly a slope
to this finger indicating predominantly an LMC population (foreground sequences appear
vertical due to the varying distance of the sources which smears out their magnitudes but
not their colors). The brightest objects in this sequence include Galactic K and G dwarfs.
To the blue and fainter of this feature lies the OB star locus in the LMC (region “A” of
Nikolaev & Weinberg 2000).
The next sequence to the red is the vertical one reaching to bright magnitudes. This
finger consists mainly of foreground dwarfs and giants (region “C” of Nikolaev & Weinberg
2000). In fact, Nikolaev & Weinberg (2000) find∼>70% of the stars in this region of the
diagram are Galactic foreground. The importance of this contamination declines at lower
brightnesses relative to the rest of the diagram as is clearly seen in the Hess diagram. Simple
models of Galactic structure (for example, Blum et al. 1995) predict only several late M
stars in the LMC foreground owing to their rarity. The remaining sequences in the J−[3.6]
vs. [3.6] diagram are predominantly LMC red giant branch stars (RGB), AGB stars and
late–type (mostly M) supergiants (SG).
Cioni et al. (2006) computed star formation history models of the LMC using the 2MASS
CMD and new stellar models (Marigo et al. 2003). The Marigo et al. (2003) models could,
for the first time, reach the observed red colors of the C–stars and thus explain the origin
of their location in the CMD. Based on their analysis of the 2MASS CMD, Cioni et al.
(2006) divided the region above the tip of the RGB (TRGB) into O–rich and C–rich zones
using relations for the 2MASS J − K color and K magnitude. In Figures 3–6 the blue
and red points correspond to Cioni et al.’s photometric division of O–rich and C–rich stars,
respectively, using 2MASS colors (their equations 5–7). The C–star and O–rich star loci are

Page 8

– 8 –
well defined and separated in our CMDs, lending support to the Cioni et al. (2006) criterion.
The green points in Figure 3 (and following figures) represent M SG and luminous,
O–rich M stars. The locus was obtained by using a cut in the J − K vs. K plane which
paralleled the Cioni et al. (2006) relation for O–rich AGB stars, but was displaced to the blue
enough to encompass the obvious sequence in Figure 3 (left panel) for the J−[3.6] vs. [3.6]
diagram. Cross reference with SIMBAD (here and elsewhere in this work, SIMBAD searches
are associated with a 5′′search radius) shows that several of the Elias, Frogel, & Humphreys
(1985) supergiants fall on this sequence as well as luminous M stars identified by Massey
(2002). Nikolaev & Weinberg (2000) confirm the presence of known M SG on the associated
J − K sequence. These luminous, late–type stars are young and thus trace regions in the
LMC of recent star formation (M SG have ages of about 10 Myr while the most luminous
AGB stars are a few hundreds of Myr). This is seen in Figures 1 and 2; contrast the clumpy
and non–uniform distribution of the young stars with the more smooth distribution of the
older C–stars and O–stars. The distribution of young M supergiants qualitatively matches
the structure seen in other tracers of star formation in the LMC; see, for example, the Hα
map of Gaustad et al. (2001) and the UV map of Smith et al. (1987). These and other
tracers are conveniently summarized in Figure 1 of Meixner et al. (2006).
The yellow points in Figure 3 represent a group of objects which show the effects of
dusty circumstellar envelopes. These “extreme” AGB stars are discussed in the next section
where their J−[8.0] color is used to define their selection. Likewise, a sequence which may
be dominated by background galaxies is given by the magenta points in Figure 3. These
sources are defined by their J−[8.0] color also (see below), and shown in the J−[3.6] vs. [3.6]
CMD to demonstrate how their excess color is dominated by longer wavelength emission.
3.1.2.
J−[8.0] vs. [8.0]
C–stars have J−[3.6] color which extend up to about 3.0 (J − K∼<2.1); at this point,
the sequence continues at nearly constant [3.6] brightness (see Figure 3) to very red colors
(whereas the K magnitudes break sharply to fainter brightnesses here). In the J−[8.0] vs.
[8.0] CMD shown in Figure 4, the analogous break occurs at J−[8.0] ≈ 3.5 with the [8.0]
data points trending to brighter magnitudes, but with different slope than before. We have
drawn this “break” in the apparent C–star relation arbitrarily at J−[3.6] = 3.1 and coded
the redder objects with [3.6] ≥ 10.5 as yellow points in Figures 3–6. We call these stars
“extreme” AGB stars. The brightest members of this sequence (IRAS sources) and the red
stars above it have been the subject of numerous earlier studies to search for and characterize
obscured AGB stars. See, for example, Loup et al. (1997); van Loon et al. (1999). While most

Page 9

– 9 –
of the objects in this sequence are C–stars, not all of them are (see, for example, Zijlstra et
al. 2006); spectroscopic confirmation is required to confidently place any particular source in
the C–star or M–star category. A number of investigators have also studied representatives
of the “extreme” star sequence in detail at shorter wavelengths. Hughes & Wood (1990)
investigated the AGB stars which span the “visible” C–star locus and bluest part of the
present “extreme” star locus (J−[8.0]∼<4).
The separation of the stellar sequences becomes more apparent in the J−[8.0] vs. [8.0]
diagram. The prominent middle vertical sequence (“C”) now breaks into two clear sequences
with foreground late K giants further to the red than earlier G type giants.
An obvious sequence of sources becomes evident in the J−[8.0] vs. [8.0] diagram. These
points are color coded in magenta. As pointed out by Meixner et al. (2006), comparison of
the [8.0] number counts to those for deep extragalactic fields (Fazio et al. 2004) indicates
these are predominantly background galaxies. The same sources are shown in the J−[3.6]
diagram where they merge with the bottom of the LMC RGB. This indicates a strong 8 µm
excess, presumably due to polycyclic aromatic hydrocarbons (PAHs); see, for example, Dale
et al. (2005). This population is the bright component of a much larger population which
is discussed in the next section. The number of sources of various populations in the CMDs
discussed in this section and below are summarized in Table 1.
3.2. The IRAC and MIPS Color–Magnitude Diagrams
3.2.1.[3.6]-[8.0] vs. [8.0]
The [3.6]−[8.0] vs. [8.0] Hess diagram and CMD are shown in Figure 5. The LMC and
foreground stars are compressed, generally, to the blue except for the extreme AGB stars. In
fact, some of these are so extreme that they are not detected at J in 2MASS. The brightest
([8.0]∼<7), reddest ([3.6]−[8.0]∼>1.5) objects (both yellow and cyan) are typically IRAS
sources and/or MSX (Egan et al. 2001) LMC sources, a number of which have already been
observed spectroscopically with the Spitzer IRS (Zijlstra et al. 2006; Markwick–Kemper et
al. 2005, see also the following section).
The faint cloud of objects centered at 3, 12.5 (cyan points) in Figure 5, is likely domi-
nated by background galaxies. These objects have no J−band counterparts and so appear
to form the fainter extension of the magenta sources discussed above (all objects in this dia-
gram colored cyan have no J band counterpart in the 2MASS point source catalog whether
they are bright, red AGB stars or faint red galaxy candidates). Young stellar objects (YSOs)
have similar colors as those for the background galaxies discussed here (see Whitney et al.

Page 10

– 10 –
2004; Meixner et al. 2006). Detection of YSOs will therefore require a careful analysis of
background counts, consideration of the spatial location (i.e. in or near young clusters) or
clustering of such objects, and most likely, follow up spectroscopy. The investigation of YSOs
in the LMC is a major focus of the SAGE team and will be discussed in detail in future
publications.
We have queried the SWIRE extragalactic survey database (Lonsdale et al. 2004) to
estimate the number of background sources expected. Table 2 shows the number of sources
per square degree with 10 < [8.0] < 13.5 mag and [3.6]−[8.0] > 1.5 mag for three of the
SWIRE fields and our SAGE data. A direct comparison between SAGE and SWIRE is
difficult for several reasons. The SAGE data are not complete; thus the SAGE counts
should be considered a lower limit. The SWIRE photometry comes from Sextractor (Bertin
& Arnouts 1996) while SAGE uses DAOPHOT. The criteria for detecting extended objects
is different for these two methods. It is possible that slightly extended, faint, objects are
extracted by DAOPHOT as point sources. In Table 2, we compare the SAGE counts to the
SWIRE counts for which the extracted objects are classified as point–like, indeterminate,
and slightly extended in [3.6] (as determined from the SWIRE database 3.6 µm extended
source flag with values −1, 0, and 1 respectively).
A large fraction of the faint red objects appear to be background galaxies. One of the
SWIRE fields is plotted in Figure 7. This CMD looks qualitatively the same as Figure 5
in the region fainter than [8.0] = 10 mag and for [3.6]−[8.0] > 1.5 mag; the distribution of
magnitudes and colors is similar including the rather sharp redward cut–off at [3.6]−[8.0] =
3.5 mag and the small number of sources brighter than [8.0] = 11 mag. The SAGE data have
more counts per square degree than all the SWIRE fields. This suggests some of the faint, red
objects belong to the LMC; however a definitive count of the LMC background will require
observations in fields near the LMC. Indeed, the surface density of the putative background
population towards the LMC appears uniform (magenta and cyan points, Figure 1) compared
to the LMC stellar populations.
3.2.2. [8.0]−[24] vs. [24]
Moving to longer wavelengths, the [8.0]−[24] vs. [24] diagram (Figure 6) further com-
presses the bluer stellar sequences such that only stars with very strong emission in [24]
stand out. In Figure 6, Spitzer IRS photometry is over–plotted on the SAGE data for a
sample of sources investigated by Markwick–Kemper et al. (2006). Many of these are highly
obscured objects. Markwick–Kemper et al. (2006) have classified the objects according to
the scheme of Kraemer et al. (2002). This scheme identifies objects with silicate features in

Page 11

– 11 –
the circumstellar envelope (“S”, typically O–rich stars in the LMC, but could also be young,
embedded stars) and carbonaceous features (“C”, typically C–stars). In addition, objects
can be classified as planetary nebulae (PNe or “P”).
The population of sources identified above with background galaxies (same sources from
the [3.6]−[8.0] vs. [8.0] CMD with no J−band counterparts, cyan points) is well differentiated
from stars in this diagram. We have argued above that most of these are probably background
galaxies (the brightest sources include known LMC objects; see below). It is clear that our
counts for [24] are not complete for all [8.0]−[24] as evidenced by the sloping cut–off at the
red limit of Figure 6 for [24] < 11 mag and [8.0]−[24] > 3 mag. Nevertheless, a comparison
to the SWIRE data shows that we should expect approximately 200–400 galaxies per square
degree (see Table 2) with 7 < [24] < 11 mag. This compares with approximately 250 per
square degree for the SAGE data.
As mentioned in the preceding section, some LMC YSOs are expected to be included in
the SAGE source counts at all brightness levels. This can already be verified for the brightest,
reddest objects. The IRS sources plotted in Figure 6 include four “S” objects which are
generally the brightest, reddest objects and lie in the part of the CMD where massive YSOs
are expected (Whitney et al. 2004; Meixner et al. 2006). Two of these sources are associated
with young objects and may indeed be embedded, massive YSOs. Source MSX LMC 1200,
the brightest “S” source in Figure 6, is identified as a likely compact H II region by van Loon
et al. (2001). Source MSX LMC 1786 is identified with a molecular cloud (Johansson et al.
1998) by Egan et al. (2001). The other two “S” sources (MSX LMC 906 and MSX LMC 1436)
have no further identification apart from 2MASS catalog id’s. Model colors and magnitudes
for massive YSOs (Whitney et al. 2004) also show that these objects can have the same
colors as those near the bright, red end of the “extreme” star sequences in Figures 5 and 6,
though such objects should be rare even compared to AGB stars. The context, i.e. whether
an object is found in or around a region of star formation, will be important in determining
the it’s precise nature (particularly those sources which are completely obscured at shorter
wavelengths).
The narrow vertical sequence at [8.0]−[24] = −0.2 is composed predominantly of stars
with J−[8.0] ≤ 1.0 (i.e. blue LMC stars and foreground objects). A sequence of points falls
near [24] = 10 but in between the main stellar and extragalactic loci. These sources have
blue colors at shorter wavelengths and so are not obvious on any of the previous diagrams as
they merge with the numerous LMC AGB stars at shorter wavelengths. But the sequence
stands out at 24 µm, trending to larger values of [8.0]−[24] excess (see Figure 6 and the tip
of the sequence marked by an “F”). This sequence is discussed further below.

Page 12

– 12 –
4.Discussion
The Hess diagrams and CMDs presented in Figures 3–6 can be used to assess the relative
importance to the mass loss budget in the LMC. Detailed model calculations will be made
once the full SAGE data set is in hand and there are appropriate template spectra to assess
the stellar envelope dust properties as a function of location in these diagrams. The SAGE
team and others have existing and planned Spitzer and ground based spectroscopic follow–up
programs underway. Some of the brightest mass–losing sources in the LMC have already
been analyzed (see, for example, Zijlstra et al. 2006). These sources are prolific, but rare
mass–losing objects. The full impact of the less luminous stars on the AGB has yet to be
quantified.
The effect of dust in shaping the infrared CMDs is clear at a glance from Figures 3–6 and
considering the 2MASS H −K vs. K CMD (Nikolaev & Weinberg 2000). In the latter, even
at Ks (2.1 µm), the tail of extreme AGB stars decreases in brightness for redder colors. This
suggests the primary effect at short wavelengths is circumstellar extinction. In the J−[3.6]
vs. [3.6] diagram of Figure 3, the extreme star branch has essentially constant brightness.
By 8 µm, the sequence is increasing in brightness clearly indicating that the circumstellar
envelopes are exhibiting successively more excess emission. The [8.0]−[24] vs. [24] CMD
shows that the majority of the reddest objects are the fainter objects we have identified
with background galaxies. However, at the brightest and reddest this sequence seems to
merge with the luminous stellar sources (i.e. the “extreme” AGB and SG stars). Indeed a
survey of the SIMBAD catalog shows a number of these objects to be PNe and other bright,
stellar, IRAS sources (one Wolf–Rayet star and several objects associated with HII regions,
as well as obscured AGB stars). Of the 236 sources with [8.0]−[24] > 3.0 mag and [24] <
7.0 mag, 10 are known PNe and 13 are classified as IRAS or IR sources (of which Loup et
al. 1997, discuss four as obscured AGB stars), The majority have no other counterparts in
the SIMBAD data base.
This general picture is confirmed by the Spitzer IRS sources plotted in Figure 6. The
stars identified as “C” lie among the “extreme” stars where many C–rich objects are ex-
pected23. The “S”, or silicate, sources lie among these, but also slightly to the red and
at brighter magnitudes where the M supergiants (green points) and luminous AGB stars
predominate. The IRS PNe lie at bright magnitudes and red colors as do several massive
YSOs as discussed in § 3.2.2. Clearly SAGE will provide a host of new mass losing objects
to follow up in detail.
23One object, LI–LM 603, had anomalous IRAC/MIPS colors. In this case, we substituted the IRS
spectrophotometry.

Page 13

– 13 –
While there are differences in detail, the location of the basic LMC AGB and supergiant
sequences correspond remarkably well to those predicted from Galactic models (moved to
the distance of the LMC, distance modulus=18.5) for the analogous objects (Wainscoat et
al. 1992; Cohen 1993, 1994, 1995). In particular, this includes the general location, extent,
and slope of the “extreme” AGB stars in Figures 3 and 4, as well as the position of the most
luminous supergiants and O–rich objects above the extreme stars. The predicted positions
of the Galactic objects such as PNe and HII regions are also in rough agreement with the
objects identified above in the [8.0]−[24] vs.[24] CMD (Figure 6).
Figure 3 shows the most dominant Spitzer population by number is the red giants. The
RGB is the most densely populated sequence with a tip magnitude of TRGB([3.6])= 11.85.
There are approximately 650000 stars on the RGB below the TRGB in the epoch one data
set. The peak number density (2400 stars in a 0.0475×0.0625 square magnitude pixel) occurs
at J−[3.6], [3.6] = 0.8, 15.6. The RGB number density drops rapidly to a value ∼ 25% of
its value just to the faint side of the tip. Approximately 74% of the stars in the J−[3.6] vs.
[3.6] CMD above the TRGB are O–rich, C–rich, and SG types. Roughly 12% of stars above
the TRGB are foreground giants and dwarfs with the remainder being mostly blue LMC
supergiants according to the discussion in §3.
The AGB stars and supergiants can be divided up among those with strong mass loss
(the so called “extreme” stars) and those whose colors suggest weaker mass loss or less dusty
envelopes. Of the latter, most of the objects can be classified as O–rich or C–rich based on
the J −K vs. K classification of Cioni et al. (2006). The red tail of C–stars (literally where
Figure 3 shows red points) stretches to approximately J−[3.6] = 3.1 where the density of
objects drops significantly toward redder colors (see Figure 3). This color limit is a useful
indicator of the extreme or obscured AGB stars as it corresponds roughly to the point in the
J − K vs. K CMD where the AGB stars become fainter at redder color (exhibiting strong
circumstellar extinction). The C/O star ratio (not including the “extreme” stars) is 0.39
(see Table 1). This value is somewhat lower than given by Cioni et al. (2006); however, the
data analized by Cioni et al. (2006) includes stars at over a larger area than considered here
and for which the number of C–stars is significant. A full analysis of the C/O star ratio with
position will follow in a subsequent paper.
According to the models of Marigo et al. (2003), the position of the C–stars in the red
tail (Marigo et al. 2003, discuss this in terms of J−K) is due to cooler temperatures resulting
from their molecular opacity differences compared to the O–rich stars. The opacity difference
is a direct result of the changing molecular equilibrium as dredge up changes the ratio of C/O
(and hence the important molecular constituents in the C–rich phase compared to the O–rich
phase). The colors are thus “photospheric” in nature, and not until the obvious break in the

Page 14

– 14 –
CMDs (e.g., J−[3.6] = 3.1) does dust significantly affect the colors. The majority of stars
in the red tail of Figure 3 are optically visible C–stars with objective prism identifications
given by Blanco & McCarthy (1990) and Kontizas et al. (2001). About 25% of the sources
identified as C–stars here have no previous identification in the SIMBAD database (but are
2MASS sources).
Comparison of the [3.6]−[8.0] vs. [8.0] and [8.0]−[24] vs. [24] CMDs suggests the relative
importance of mass loss can be seen among the different AGB stars. The green SGs lie to
the blue of the yellow extreme stars in the former diagram, but generally to the red in the
latter. The SGs thus appear to have cooler, perhaps more extended, dust envelopes. The red
“visible” C–stars generally have little or no 24 µm excess, whereas a subset of the blue O–rich
AGB stars exhibit increasing amounts of excess emission and presumably mass loss. This
sequence is approximately at constant [8.0] magnitude (≈ 10.5). This group of AGB stars
is well above the TRGB in the shorter wavelength CMDs and represents the brightest tip of
stars classified as O–rich just at the transition to the “visible” C–star locus in the J−[8.0]
vs. [8.0] CMD (Figure 4). The group of stars is clearly visible as a slight enhancement in
density in the Hess diagram (Figure 4) above and to the red of the thin finger of AGB stars
which rise above the TRGB (i.e. at J−[8.0], [8.0] ≈ 1.5, 10.5). Fainter O–rich AGB stars
are not detected at high signal–to–noise in this preliminary data set at 24 µm. Even so, this
is striking evidence for mass loss in lower luminosity stars than has generally been explored
in the LMC.
Van Loon et al. (1999) derived mass loss rates from a sample of 57 stars in the LMC cho-
sen by their infrared excess. Their analysis suggests lower luminosity AGB stars have mass
loss rates∼>10−7M⊙yr−1. The SAGE 24 µm data reach somewhat below the luminosities
probed by van Loon et al., and the epoch one data will be combined with epoch two photom-
etry, allowing SAGE to reach even fainter sources at 24 µm (0.4 mag). The fact that the low
luminosity mass losing sources we have identified above (Figure 6, “F”) span a range of 24
µm excess from no excess to ∼ 2 mag suggests they represent the faintest population of dusty
sources with significant mass loss (10−7∼>M⊙yr−1∼>10−8) in the LMC. Thus, the full SAGE
catalog should detect the all the important mass losing sources in the LMC. Adding up the
first epoch SAGE sources identified here as “extreme” AGB stars, SG/luminous AGB, and
the present lower luminosity O–rich AGB stars, we find approximately 10% of the evolved
stars above the TRGB are likely significant mass–loss sources.
In Figure 8, the average spectral energy distribution (SED) is plotted for these low
luminosity AGB sources. The sample of stars is arbitrarily divided into three bins of varying
[8.0]−[24] color ([8.0]−[24] ≤ 0.67, 0.67 ≤ [8.0]−[24] < 1.34, 1.34 ≤ [8.0]−[24] < 2.0) to
show the increasing excess emission along the sequence at constant 8 µm brightness. There

Page 15

– 15 –
are 632, 439, and 75 stars in these bins, respectively. To contribute to a given bin, a star
is required to have valid photometry for each 2MASS, IRAC, and MIPS (24 µm) band;
hence, there are more stars plotted in Figure 6 than represented in Figure 8 since the former
includes objects for which there is no required H, [4.5], or [5.8] photometry.
The sources in all bins have very similar SEDs up to 8 µm, but then exhibit increasing
24 µm flux indicative of successively more cool dust emission which we suggest is due to
increasing mass loss among otherwise very similar objects. Careful scrutiny of Figure 8
shows that the stars with the least amount of 24 µm excess are the brightest at shorter
wavelengths, while those with the most excess are fainter at the shorter wavelengths. The
cool dust envelopes are likely providing additional extinction at the shorter wavelengths.
While not shown in the figure, arbitrarily scaling the SEDs to match in the near infrared
shows that they are nearly identical until about 6.8 µm. At this point, a clear but small
excess begins to show at 8 µm. This fact, and the fact that no fainter population of obvious
mass losing stars is evident in the much deeper CMD shown in Figure 4 provides support
for our claim that SAGE will detect all the important mass losing sources in the LMC.
Could these objects be chance alignments of “normal” AGB stars and background
(dusty) galaxies at 24 µm? To estimate the possible number of chance alignments, con-
sider the SWIRE number counts given above. The IRAC–24 µm search radius was 1′′. The
total area covered by all 24 µm sources is then 28000 sources × the area per 24 µm source =
0.007 square degrees. Multiplying this by the number of SWIRE galaxies per square degree
(253) gives two chance alignments per 28000 24 µm sources).
5. Summary
Epoch one data for the SAGE survey of the entire LMC have been presented for the
Spitzer IRAC bands and the MIPS 24 µm band. For the first time, an infrared view is
presented which places the bright IRAS and MSX sources in context. The Spitzer (and
2MASS) CMDs are dominated by red giants (650000 red giants to the present epoch one
limit of the survey) and AGB stars.
For epoch one of the SAGE survey, we find approximately 18000 oxygen–rich AGB stars,
7000 carbon–rich stars, 1200 late–type supergiants (or luminous oxygen–rich stars), and 1200
“extreme” AGB stars (optically obscured or enshrouded AGB stars) which represent the high
mass loss rate evolved star population in LMC. Of the approximately 30000 evolved stars
above the tip of the red giant branch, some 10%, can be readily identified with dusty mass–
losing envelopes. The C/O star ratio is approximately 40% (46% if all the “extreme” stars