Content uploaded by Paul M. Woods
Author content
All content in this area was uploaded by Paul M. Woods on Jun 26, 2015
Content may be subject to copyright.
arXiv:1505.04499v2 [astro-ph.SR] 31 May 2015
Mon. Not. R. Astron. Soc. 000, 000–000 (0000) Printed 4 June 2015 (MN L
A
T
EX style file v2.2)
Spitzer Infrared Spectrograph point source classification in
the Small Magellanic Cloud
Paul M. E. Ruffle,1,2⋆F. Kemper,2†O. C. Jones,1,3G. C. Sloan,4K. E. Kraemer,5
Paul M. Woods,6M. L. Boyer,7S. Srinivasan,2V. Antoniou,8E. Lagadec,9
M. Matsuura,10,11 I. McDonald,1J. M. Oliveira,12 B. A. Sargent,13 M. Sewi lo,14,15
R. Szczerba,16 J. Th. van Loon,12 K. Volk3and A. A. Zijlstra1
1Jodrell Bank Centre for Astrophysics, The University of Manchester, Alan Turing Building, Oxford Road, Manchester M13 9PL
2Academia Sinica, Institute of Astronomy and Astrophysics, Taipei 10617, Taiwan
3Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA
4Department of Astronomy, Cornell University, Ithaca, NY 14853, USA
5Institute for Scientific Research, Boston College, 140 Commonwealth Avenue, Chestnuthill, MA 02467, USA
6Astrophysics Research Centre, School of Mathematics & Physics, Queen’s University Belfast, University Road, Belfast, BT7 1NN
7Observational Cosmology Lab, Code 665, NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA
8Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA
9Laboratoire Lagrange, Universit´e de Nice - Sophia Antipolis, Observatoire de la Cˆote d’Azur, CNRS, 06304 Nice, France
10School of Physics and Astronomy, Cardiff University, The Parade, Cardiff CF24 3AA
11Department of Physics and Astronomy, University Col lege London, Gower Street, London WC1E 6BT
12School of Physical and Geographical Sciences, Lennard-Jones Laboratories, Keele University, Staffordshire ST5 5BG
13Laboratory for Multiwavelength Astrophysics, Rochester Institute of Technology, 54 Lomb Memorial Drive, Rochester, NY 14623, USA
14Space Science Institute, 4750 Walnut Street, Suite 205, Boulder, CO 80301, USA
15Johns Hopkins University, Department of Physics and Astronomy, 366 Bloomberg Center, 3400 N. Charles Street, Baltimore, MD 21218, USA
16N. Copernicus Astronomical Center, Rabianska 8, 87-100, Torun, Poland
4 June 2015
ABSTRACT
The Magellanic clouds are uniquely placed to study the stellar contribution to dust
emission. Individual stars can be resolved in these systems even in the mid-infrared,
and they are close enough to allow detection of infrared excess caused by dust. We have
searched the Spitzer Space Telescope data archive for all Infrared Spectrograph (IRS)
staring-mode observations of the Small Magellanic Cloud (SMC) and found that 209
Infrared Array Camera (IRAC) point sources within the footprint of the Surveying
the Agents of Galaxy Evolution in the Small Magellanic Cloud (SAGE-SMC) Spitzer
Legacy programme were targeted, within a total of 311 staring mode observations.
We classify these point sources using a decision tree method of object classification,
based on infrared spectral features, continuum and spectral energy distribution shape,
bolometric luminosity, cluster membership and variability information. We find 58
asymptotic giant branch (AGB) stars, 51 young stellar objects (YSOs), 4 post-AGB
objects, 22 Red Supergiants (RSGs), 27 stars (of which 23 are dusty OB stars), 24
planetary nebulae (PNe), 10 Wolf-Rayet (WR) stars, 3 H ii regions, 3 R Coronae Bore-
alis (R CrB) stars, 1 Blue Supergiant and 6 other objects, including 2 foreground AGB
stars. We use these classifications to evaluate the success of photometric classification
methods reported in the literature.
Key words: techniques: spectroscopic – surveys – galaxies: Small Magellanic Cloud
– stars: early-type, YSO, supergiants, AGB, post-AGB, planetary nebulae, late-type,
carbon, oxygen – ISM: dust, H ii regions – infrared: stars.
⋆Paul M. E. Ruffle passed away on 21 November 2013. His co- authors have finished the manuscript on his behalf, and would
c
0000 RAS
2Paul M. E. Ruffle et al.
1 INTRODUCTION
The Mega-Surveying the Agents of Galaxy Evolution (Mega-
SAGE) project has obtained infrared photometric and spec-
troscopic inventories of the Magellanic Clouds with the
Spitzer Space Telescope (hereafter Spitzer ), using Spitzer
and Herschel Legacy Programmes. The initial SAGE sur-
vey (Meixner et al. 2006) detected and catalogued ∼6.9 mil-
lion point sources in the Large Magellanic Cloud (LMC)1,
while the SAGE-SMC survey (Gordon et al. 2011) detected
and catalogued ∼2.2 million point sources in the Small
Magellanic Cloud (SMC)2. Both surveys used all bands of
the Infrared Array Camera (IRAC; 3.6, 4.5, 5.8, 8.0 µm;
Fazio et al. 2004) and the Multi-Band Imaging Photometer
for Spitzer (MIPS; 24, 70, 160 µm; Rieke et al. 2004) instru-
ments on board Spitzer (Werner et al. 2004). The resolution
of the IRAC observations is ∼2′′, while for the MIPS bands,
three different resolutions apply: 6′′, 18′′ , and 40′′ for the
24, 70 and 160 µm bands respectively (Gordon et al. 2011).
To follow up on these programmes, the SAGE-Spec project
(Kemper et al. 2010) obtained 196 staring-mode pointings
using Spitzer’s Infrared Spectrograph (IRS; Houck et al.
2004, 5.2–38 µm) of positions selected from the SAGE cat-
alogue. SAGE-Spec will relate SAGE photometry to the
spectral characteristics of different types of objects in both
Magellanic Clouds, and ultimately, allow us to classify pho-
tometric point sources in both the LMC and SMC. This
characterisation of the point sources observed in the SAGE-
Spec survey, and the IRS data archive, builds an inventory
of dusty sources and their interrelation in each of the Magel-
lanic Clouds. In a first step towards this goal, Woods et al.
(2011) classified the initial 196 LMC point sources, using a
decision tree method of object classification, based on in-
frared (IR) spectral features, continuum and spectral en-
ergy distribution (SED) shape, bolometric luminosity, clus-
ter membership and variability information. The initial clas-
sification of LMC objects is being extended to ∼1,000 point
sources, covering all archival IRS observations within the
SAGE footprint (Woods et al. in prep.).
To extend the LMC classifications to the SMC, we have
searched the Spitzer data archive for IRS staring-mode ob-
servations and found 311 spectra, yielding 209 unique and
genuine point sources with IRS data within the footprint of
the SAGE-SMC Spitzer Legacy programme. The data used
in the classification process are described in Section 2. In
Section 3 we discuss the classification method, and in Sec-
tion 4 we classify the 209 SMC point sources using the deci-
sion tree method. Finally, in Section 5 we compare spectral
versus colour classifications by means of colour-magnitude
diagrams (CMDs). We use our spectroscopic classifications
to test photometric classification methods, e.g. those by
Boyer et al. (2011); Sewi lo et al. (2013) and Matsuura et al.
(2013). The classification of each of these 209 sources is part
like to dedicate it to his memory. Paul was a very enthusiastic
scientist, and a wonderful friend with a great sense of humour.
We miss him tremendously.
†Email: ciska@asiaa.sinica.edu.tw
1http://irsa.ipac.caltech.edu/data/SPITZER/docs/
spitzermission/observingprograms/legacy/sage
2http://irsa.ipac.caltech.edu/data/SPITZER/docs/
spitzermission/observingprograms/legacy/sagesmc
of the data delivery of the SAGE-Spec Legacy project to the
Spitzer Science Center and the community.3These classifi-
cations will also be used to benchmark a colour-classification
scheme that will be applied to all point sources in the SAGE
and SAGE-SMC surveys (Marengo et al. in prep).
2 DATA PREPARATION
2.1 Spitzer IRS staring mode observations
The IRS on board Spitzer covers the wavelength range 5–
38 µm. For the low-resolution mode, the spectrum splits in
two bands, short-low (SL: 5.2–14.5 µm) and long-low (LL:
14.0–38.0 µm), with almost perpendicular slits. Each seg-
ment splits into a range covered at second order (SL2, LL2),
and one at first order (SL1, LL1). The resolution varies be-
tween 60 and 130. The high resolution mode covers the wave-
length range from 10–19.6 µm (short-high; SH) and from
18.7–37.2 µm (long-high; LH) with a spectral resolution of
R∼600.
We identified 311 Spitzer IRS low- and high-resolution
staring mode observations within the footprint of the SAGE-
SMC survey (Gordon et al. 2011), not necessarily associ-
ated with a point source. We numbered these SMC IRS 1–
311, by ordering them by observing program (Project ID;
PID) first, and then the Astronomical Observation Request
(AOR) number (Table 1). Where available (for SL and LL
observations only), the reduced spectra were downloaded
from the Cornell Atlas of Spitzer IRS Sources4(CASSIS;
Lebouteiller et al. 2011), in a full resolution grid, using the
optimal extraction method. At the moment, the CASSIS
database only contains SL and LL data, but it turns out
that there are no point sources in the SMC targeted with
only SH and LH, so we will use the SL and LL data only.
The data were downloaded in the Infrared Processing and
Analysis Center (IPAC) Table format5; the file names are
also given in Table 1. As an example, the upper left panel
of Fig. 1 shows the spectrum for point source SMC IRS 110,
with the SL2 and LL2 data from CASSIS in red and the SL1
and LL1 data in blue.
The CASSIS-reduced IRS data header provides the
original PI’s requested (REQ) position and the field of view
(FOV) position, i.e. where the telescope actually pointed
(usually, but not always coincident within 1′′ of the REQ
position). In many cases, however, neither of these two po-
sitions is the same as the position at which the spectrum is
actually extracted from the slit (EXT), using the optimal
extraction method. In order to check the spectrum position
for each source, slit images were generated by over-plotting
the IRS SL and LL slit positions (recorded in each Spitzer
AOR’s BCD FITS header) on a 360′′ ×360′′ image extracted
from the SAGE-SMC 8 µm image. The REQ and EXT posi-
tions were also over-plotted on the image. In cases where the
spectrum was extracted at the FOV position (and thus an
EXT position is lacking), the FOV position was overplotted
3http://irsa.ipac.caltech.edu/data/SPITZER/docs/
spitzermission/observingprograms/legacy/sagespec
4http://cassis.astro.cornell.edu
5http://irsa.ipac.caltech.edu/applications/DDGEN/Doc/
ipac tbl.html
c
0000 RAS, MNRAS 000, 000–000
Spitzer-IRS point source classification in the SMC 3
Table 1. IRS staring mode targets in the SMC. The first ten lines are presented to demonstrate the format of this table; the full table
is available on-line.
SMC IRS AOR PID PI spectrum position CASSIS file name
RA (h m s) Dec (◦ ′ ′′ )
1 3824640 18 Houck 00 46 40.32 −73 06 10.80 NULL
2 3824896 18 Houck 01 09 16.80 −73 12 03.60 NULL
3 4384000 63 Houck 01 24 07.68 −73 09 03.60 NULL
4 4384000 63 Houck 01 24 07.68 −73 09 03.60 NULL
5 4384000 63 Houck 01 24 07.68 −73 09 03.60 NULL
6 4384768 63 Houck 00 59 09.84 −72 10 51.60 NULL
7 4385024 63 Houck 00 58 52.25 −72 09 25.92 cassis tbl spcf 4385024 1.tbl
8 4385024 63 Houck 00 58 58.22 −72 09 50.76 cassis tbl spcf 4385024 2.tbl
9 4385024 63 Houck 00 58 58.80 −72 10 25.32 cassis tbl spcf 4385024 3.tbl
10 4385024 63 Houck 00 59 06.62 −72 10 25.68 cassis tbl spcf 4385024 4.tbl
Spectra for SSID: 110 AOR: 10663169 Target: MSX SMC 024
5 10 15 20 25 30 35
Wavelength (µm)
0
20
40
60
80
Fν (mJy)
90 180 270 36000 SSID: 110 PID: 3277 AOR: 10663169 Pointing: 1 Target: MSX SMC 024
0
90
180
270
360
00
00 42 52.2 −73 50 52 Requested Position
N
00 42 52.3 −73 50 51 Extracted Position
SED for SSID: 110 AOR: 10663169 Target: MSX SMC 024
0 10 20 30
Wavelength (µm)
0
20
40
60
80
100
120
140
λFλ (10−15 Wm−2)
Log SED for SSID: 110 AOR: 10663169 Target: MSX SMC 024
1 10
Wavelength (log µm)
0.1
1.0
10.0
100.0
λFλ (log 10−15 Wm−2)
Figure 1. From upper left to lower right: Example spectrum, slit image, SED and log SED plots for point source SMC IRS 110. The
spectrum plots were generated using CASSIS-reduced IRS data from the following low resolution modules: short-low 2nd order 5.2–
7.6 µm (red); short-low 1st order 7.6–14.0 µm (blue); long-low 2nd order 14.0–20.5 µm (red); long-low 1st order 20.5–37.0 µm(blue). Slit
images were created by over-plotting IRS short-low and long-low slit positions on a 360′′ ×360′′ image extracted from the SAGE-SMC
8µm image; the REQ and EXT or FOV positions were also over-plotted. SED and log SED plots (lower left and lower right panels,
respectively) were generated by combining the above IRS data with the following photometric data points: SAGE-SMC catalogue (green
diamonds): U,B,V,I(MCPS), J,H,KS(IRSF); 3.6, 4.5, 5.8, 8.0 µm (IRAC), 24 µm (MIPS); WISE catalogue (magenta triangles): J,
H,KS(2MASS), 3.4, 4.6, 12, 22 µm (WISE ).
c
0000 RAS, MNRAS 000, 000–000
4Paul M. E. Ruffle et al.
instead (see upper right panel of Fig. 1 for an example for
SMC IRS 110). The slit images are useful in determining
the origin of the emission seen in the IRS spectra. The coor-
dinates in Table 1 represent, if available, the EXT position.
The next preference is the FOV position, and if neither of
these is available, the REQ position is given.
2.2 Photometric matching
In order to find matching photometry for the IRS spectra,
we searched the SAGE-SMC Single Frame + Mosaic Pho-
tometry (SMP) Archive v1.5 (Gordon et al. 2011) available
on Gator6, using, in order of preference, the EXT, FOV
or REQ spectrum positions. We searched for IRAC point
source matches within 3′′ of the spectrum positions, which
corresponds with the pointing accuracy of the IRS mode
on Spitzer. In cases where the SAGE-SMC point source
catalogue did not provide a match, we also searched the
Spitzer Survey of the Small Magellanic Cloud (S3MC) cat-
alogue7(Bolatto et al. 2007) for IRAC matches within 3′′ .
We found three sources in S3MC without a SAGE-SMC cat-
alogue counterpart. Although the S3MC data are included
in the SAGE-SMC result, both teams used different point
source extraction pipelines and the source catalogues there-
fore do not provide a one-to-one match.
Of the original list of 311 IRS staring mode observations
within the SAGE-SMC footprint, we discarded all 44 spectra
for which we could not identify an IRAC point source within
3′′ in either the SAGE-SMC or the S3MC surveys. In cases
where multiple matches were present within 3′′, we man-
ually compared the magnitudes with the flux levels of the
spectra, and used the slit images to establish which source
was responsible for the spectrum. We also consolidated du-
plicate measurements of the spectrum of a given source, as
evidenced by their SAGE-SMC or S3MC identification, into
a single entry in our analysis; this is sufficient for spectral
identification purposes. This further reduced the number by
58 to a list of 209 unique Spitzer -IRAC point sources, with
either SAGE-SMC or S3MC identifications, for which IRS
staring mode observations are available. We compiled all
relevant information in a table available online. Table 2 de-
scribes the columns of the online table. Fig. 2 shows the
distribution of the 209 sources over the SMC.
Preferring the IRAC coordinates over the spectrum co-
ordinates, we then matched the IRAC point sources to a
number of other infrared and optical photometric surveys.
We obtained MIPS-[24], [70] and [160] matches, within a
search radius of 3′′, 9′′ , and 20′′, respectively, from the
SAGE-SMC survey (Gordon et al. 2011), corresponding to
half a resolution element in these bands. We also searched
the Wide-Field Infrared Survey Explorer (WISE ) All-Sky
Source Catalog for matches within 3′′. We also searched
for AKARI matches in the N3, N4, S7, S11, L15, and S22
bands within 3′′ using the catalogue provided by Ita et al.
6Gator is the general catalogue query engine provided by the
NASA/IPAC Infrared Science Archive, which is operated by the
Jet Propulsion Laboratory, California Institute of Technology, un-
der contract with the National Aeronautics and Space Adminis-
tration.
7At the time of writing only the S3MC Young Stellar Object Cat-
alog is available in the public domain (A. Bolatto, priv. comm.).
Table 3. Classification types used in the decision tree shown in
Fig. 3, and counts for a total of 209 SMC point sources. The
last section of the table shows a breakdown of other known types
(OTHER).
Code Ob ject type Count
YSO-1 Embedded Young Stellar Objects 14
YSO-2 Young Stellar Objects 5
YSO-3 Evolved Young Stellar Objects 22
YSO-4 HAeBe Young Stellar Objects 10
HII H ii regions 3
O-EAGB Early-type O-rich AGB stars 8
O-AGB Oxygen-rich AGB stars 11
RSG Red Supergiants 22
O-PAGB Oxygen-rich post-AGB stars 1
O-PN Oxygen-rich planetary nebulae 4
C-AGB Carbon-rich AGB stars 39
C-PAGB Carbon-rich post-AGB stars 3
C-PN Carbon-rich planetary nebulae 20
STAR Stellar photospheres 4
Dusty OB stars 23
RCRB R CrB stars 3
BSG Blue Supergiant 1
WR Wolf-Rayet stars 10
OTHER B[e] stars 2
Foreground stars 2
S stars 1
symbiotic stars 1
(2010). In the near-infrared, the Two Micron All Sky Sur-
vey (2MASS) Long Exposure (6X) survey was searched
for matches within 2′′ of the IRAC positions (SAGE-SMC
matches with 2MASS), and we also used this search ra-
dius with the InfraRed Survey Facility (IRSF) catalogue
(Kato et al. 2007). The Deep Near Infrared Survey (DENIS)
of the Southern Sky catalogue (3rd release; Epchtein et al.
1999) was also searched with a 2′′ search radius. In the op-
tical, many of our sources have matches in the Magellanic
Clouds Photometric Survey (MCPS; Zaritsky et al. 2002)
and the catalogue published by Massey (2002). In both cat-
alogues we looked for matches within 1.5′′ of the IRAC po-
sition. Some of the objects in our sample are actually too
bright for those two optical surveys, and a search of the TY-
CHO catalogue with a radius of 3′′ filled in some of these
gaps. All tabulated photometry is available from the online
database (see Table 2). We only provide the magnitudes for
the purpose of evaluating the shape of the SED. Further in-
formation, including the photometric uncertainties, can be
found in the respective source tables using the designations
provided, as well as Appendix A.
2.3 Bolometric magnitudes, variability and colour
classifications
For each source bolometric magnitudes were calculated via a
simple trapezoidal integration of the SED, to which a Wien
tail was fitted to the short-wavelength data, and a Rayleigh-
Jeans tail was fitted to the long-wavelength data. The fol-
lowing SED combinations were calculated:
(i) MCPS or Massey (2002) optical photometry; JHK S
photometry; and IRAC and MIPS-[24] photometry, all as
available (mbol phot in Table 2);
c
0000 RAS, MNRAS 000, 000–000
Spitzer-IRS point source classification in the SMC 5
Table 2. Numbering, names and description of the columns present in the classification table which is available on-line only.
Column Name Description
1 smc irs SMC IRS identification number of the target
2 name Name of point source targeted
3 sage spec class Source classification determined in this paper
4–5 ra spec, dec spec Position of the extracted spectrum
6 aor Spitzer Astronomical Observation Request
(AOR) number
7 pid Spitzer observing program identification number
8 pi Last name of the PI of the Spitzer PID
9 cassis file name Name of the file containing the CASSIS -reduced
Spitzer-IRS spectrum
10 irac des SAGE-SMC or S3MC IRAC point source designation matching
the extracted spectrum
11–12 ra ph, dec ph RA and Dec in degrees of the IRAC point source
13 dpos ph Distance in arcsec between the IRAC point source
and the position of the extracted spectrum
14–17 irac1, irac2, irac3, irac4 IRAC magnitudes in bands 1–4
18–20 tycho des, b tycho, v tycho TYCHO counterpart and its Band Vmagnitudes
21–25 m2002 des, u m2002, b m2002, v m2002, r m2002 Massey (2002) counterpart and its U,B,V,R
magnitudes
26–29 u mcps, b mcps, v mcps, i mcps Matching MCPS U,B,V,Imagnitudes
30–33 denis des, i denis, j denis, k denis DENIS counterpart and its I,J,KSmagnitudes
34–37 irsf des, j irsf, h irsf, k irsf IRSF counterpart and its J,H,KSmagnitudes
38–41 tmass des, j tmass, h tmass, k tmass 2MASS 6X counterpart and its J,H,KSmagnitudes
42–46 wise des, wise1, wise2, wise3, wise4 WISE counterpart and its magnitudes in the four
WISE bands
47–52 akari n3, akari n4, akari s7, akari s11, akari l15, akari l22 Matching AKARI magnitudes in bands N3, N4, S7,
S11, L15 and L22 from Ita et al. (2010)
53–54 mips24 des, mips24 SAGE-SMC MIPS-[24] designation and magnitude
55–56 mips70 des, mips70 SAGE-SMC MIPS-[70] designation and magnitude
57–58 mips160 des, mips160 SAGE-SMC MIPS-[160] designation and magnitude
59 mbol phot Mbol calculated by interpolation of JHK S, IRAC and
MIPS-[24] photometry, with a Wien and
Rayleigh-Jeans tail
60 mbol phwi as #59, but with WISE photometry added
61 mbol phsp as #59, but with the IRS spectrum added
62 mbol pwsp as #59, but with the IRS spectrum and WISE
photometry added
63–65 mbol mcd, lum mcd, teff mcd Mbol calculated using the SED fitting code from
McDonald et al. (2009, 2012); only good fits are included
66–67 id groen, per groen Source ID and variability period in days from
Groenewegen et al. (2009)
68–72 ogle3id, ogle3mean i, ogle3mean v, ogle3amp i, ogle3period Source ID and variability information from the
OGLE survey
73 boyer class Colour classification from Boyer et al. (2011)
74 matsuura class Colour classification from Matsuura et al. (2013)
75 sewilo class Colour classification from Sewi lo et al. (2013)
(ii) like (i) but combined with the WISE photometry
(mbol phwi);
(iii) like (i) but combined with the IRS spectrum
(mbol phsp);
(iv) like (i) but combined with both the WISE photome-
try and the IRS spectrum (mbol pwsp);
For sources where there is little reprocessing of the
optical emission, i.e. little infrared excess, bolometric
magnitudes were calculated using an SED fitting code
(McDonald et al. 2009, 2012). This code performs a χ2-
minimisation between the observed SED (corrected for inter-
stellar reddening) and a grid of bt-settl stellar atmosphere
models (Allard et al. 2011), which are scaled in flux to derive
a bolometric luminosity. This SED fitter only works effec-
tively where a Rayleigh-Jeans tail is a good description of
the 3 to 8 µm region, and provides a better fit to the opti-
cal and near-IR photometry than a Planck function. For the
most enshrouded stars, fitting the SED with ‘naked’ stellar
photosphere models leads to an underestimation of the tem-
perature and luminosity, due to circumstellar reddening, and
hence the integration method (above) for calculating Mbol
is preferred for very dusty sources. Experience shows that
good fits can be separated from bad fits in the NIR: if the
model and observations differ at I,J,Hor KSby more than
a magnitude in any band, the fit is considered bad. This
retains the cases where the difference between model and
observations in the MIR or FIR is large, but often in these
cases the excess emission is unrelated to the point sources.
c
0000 RAS, MNRAS 000, 000–000
6Paul M. E. Ruffle et al.
Figure 2. The SMC IRS targets distributed on the sky, overlaid upon a SAGE-SMC IRAC 8 µm map. The colour of the points represent
our classifications, according to the legend. All four YSO subcategories and the H ii regions are grouped together in red. The class “stars”
contains objects classified as STAR and the sub-category of dusty OB stars. The “O-rich evolved” category contains O-EAGB, O-AGB,
O-PAGB and O-PN objects, and likewise the “C-rich evolved” category groups together C-AGB, C-PAGB and C-PN objects. The Red
Supergiants are a group by themselves, and “other” contains all other categories (see Table 3).
In cases where it is related to the point source, making the
source very red, the values calculated by trapezoidal inte-
gration provide a better estimate of Mbol .Teff ,Mbol and L
for the good fits are included in the online table as teff mcd,
mbol mcd and lum mcd, respectively (see Table 2).
The sample was then matched to the Optical Gravita-
tional Lensing Experiment (OGLE-III) catalogue of long-
period variables in the SMC (Soszy´nski et al. 2011) and
Groenewegen et al. (2009) to obtain variability periods, and
the variability information is included in the on-line table
(see Table 2).
Finally, we included a number of colour classification
schemes for comparison. First, Boyer et al. (2011) have ex-
tended the classification scheme developed by Cioni et al.
(2006) to classify dusty mass-losing evolved stars into
subcategories, using IRAC, MIPS and NIR colours. We
checked our source list against their catalogue for matches.
Their classifications (O-AGB, C-AGB, x-AGB, aO-AGB,
RSG, RGB and FIR) are included in the on-line table as
boyer class (see Table 2). Definitions of these classes can be
found in Boyer et al. (2011). Furthermore, we also applied
the colour classification scheme proposed by Matsuura et al.
(2013) to the sources in our list. This classification scheme
is also designed to distinguish between various kind of very
red objects, to estimate the dust production rate. We ap-
plied the cuts described by Matsuura et al. (2013, Fig. 4, 5)
on our sample and list the classifications that follow from
these cuts (O-AGB, C-AGB, RSG) in our online table, as
matsuura class (see Table 2). The last colour classification
scheme we apply is the one proposed by Sewi lo et al. (2013)
for YSOs, who applied classification cuts in the five different
infrared CMDs, followed by visual inspection of images and
SED fitting to select YSO candidates from the SAGE-SMC
survey. We checked our source list against their catalogue
and identified ‘high-reliability’ and ‘probable’ YSO candi-
dates accordingly (sewilo class; Table 2).
3 THE CLASSIFICATION METHOD
To classify our sample of 209 SMC point sources for which
IRS staring mode data exist, we follow the method described
by Woods et al. (2011). Fig. 3 shows a restyled version of
the classification decision tree. We made enhancements to
the tree, which will be discussed in this section.
A literature search was performed for each object to
retrieve other information useful in the process of classifica-
tion, including (but not limited to) determination of stellar
type, luminosity, age of nascent cluster of stars (if the ob-
ject was found to be a member of a cluster), H αdetections,
etc. This information was used in addition to the spectro-
scopic data, the photometric matches and derived bolomet-
ric luminosity, and the variability data, described in Sec. 2,
to classify the sources. Any existing classification from the
literature was used as a starting point before our spectral
classification. Appendix A provides a brief summary of the
literature survey for each object.
As in Woods et al. (2011), we adopt the following
categories for our point source classification. Low- and
intermediate-mass (M < 8 M⊙) post-main-sequence stars
are classified by chemistry (O- or C-rich) and by evolution-
ary stage (asymptotic giant branch, post-asymptotic giant
c
0000 RAS, MNRAS 000, 000–000
Spitzer-IRS point source classification in the SMC 7
YES
NO
START
YES
NO
NO
YES
YES
NO
NO
YES
YES
YES
NO
YES
NO
YES
NO
NO
NO
YES
YES
NO
NO
YES
NO
NO
YES
YES
NO
YES
NO
YES YES
NO
YES NO
NO
YES
YES
NO
NO
YES
YES
NO
NO
NO
YES YES
Redshifted
features?
Featureless
continuum with
PAHs or atomic
emission lines?
Galaxy
Atomic
emission
lines?
Evolved
YSO-3
Spectrum
rises
>30 μm?
HII
region
Strong PAH features?
3.3 6.2 7.7 8.6
11.2 12.7 16.4 μm C-PN
O-PN
Solely
molecular
features?
Ice
absorption
at 15 μm?
C-rich features?
5.0 7.5 13.7 μm
absorption
C-AGB
Mbol <–7.1 and
SED peak ~1 μm?
O-AGB
RSG
Dust
features? SED peaks
longward
of 2 μm?
RCrB
Star
O-rich features?
10, 20 μm silicate
emis. or absorp.
C-rich features?
5.0 7.5 13.7 μm abs.
11.3 21 30 μm emis.
Other
21 μm feat. or strong
11.3 & 30 μm feats. or
double-peaked SED?
C-PAGB
C-PN
C-AGB
Atomic
emission lines?
Falling spectrum over
range 20–32 μm?Distinctly
double-peaked
SED?
O-PAGB
Atomic
emission
lines?
High
excitation
lines? [NeV]
[NeVI] [OIV]
HII
region
O-PN
Member of
young cluster?
RSG
Mbol <–7.1
and SED
peak ~1 μm?
O-AGB
Ice
absorption
at 15 μm?
Silicate
absorp-
tion?
Strong
10 μm
emission?
Other HAeBe
YSO-4 YSO-2 Embedded
YSO-1
Embedded
YSO-1
Figure 3. The logical steps of the classification decision tree, where Spitzer IRS spectra (λ= 5.2–38 µm), associated optical, NIR,
WISE, IRAC and MIPS photometry, luminosity, variability, age and other information are used to classify SMC infrared point sources.
See Table 3 for key to classification group codes. Figure restyled in appearance from the one published by Woods et al. (2011).
c
0000 RAS, MNRAS 000, 000–000
8Paul M. E. Ruffle et al.
branch and planetary nebula), hence our groupings O-AGB,
O-PAGB, OPN, C-AGB, C-PAGB, C-PN. We propose
an enhancement of the classification tree by Woods et al.
(2011) to include early-type O-rich AGB stars, namely O-
EAGB. These stars do not show any evidence for dust fea-
tures in their infrared spectra, but they do show long pe-
riod variability in OGLE and MACHO and some evidence
for continuum infrared excess. Although these stars are in
the early stages of AGB evolution, they are most likely more
evolved than genuine early-AGB (E-AGB) stars, which have
not yet started helium shell burning. E-AGB stars do not
normally show long period variability. We assume that O-
EAGB stars are thermally-pulsing AGB-type objects prior
to the onset of, or just beginning, significant dust forma-
tion. More massive and luminous red supergiants have a
class of their own, RSG. Young stellar objects can be clas-
sified phenomenologically into four groups, YSO-1, YSO-2,
YSO-3, YSO-4 (see Woods et al. 2011 for the definition of
these classes). Stars showing a stellar photosphere, but no
additional dust or gas features, or long-period variability,
are classified as STAR. One observational program focussed
on stars showing a far-infrared excess in their MIPS pho-
tometry (Sheets et al. 2013; Adams et al. 2013), due to the
illumination and heating of interstellar dust (the Pleiades ef-
fect); these objects are classified as a subcategory of STAR:
dusty OB stars. The tree also distinguishes galaxies (GAL;
even though none are actually found in the present work)
and H ii regions (HII), and we furthermore have a class for
R CrB stars (labeled RCRB in the classification table). Fi-
nally the classification OTHER exists for ob jects of known
type which do not belong in the categories above and do not
follow from the classification tree (e.g. B[e] stars). The na-
ture of these objects are usually identified by other means,
and as such reported in the astronomical literature.
3.1 Classification Process
Each source was classified independently by at least three of
the co-authors. In cases where the classification was unan-
imous, it was simply adopted as the final classification,
whereas in those cases where some discrepancies occurred,
we asked additional co-authors to classify the sources, to set-
tle the issue. In some cases, a discussion on the nature of the
source ensued. We aimed to reach consensus among the co-
authors on the nature of a source in cases where differences
in classification occurred.
The lead author of this work (Paul Ruffle) developed an
internal web browser-based classification tool, to facilitate
the classification process. The decision tree was built into
the tool as a series of questions, and the collected data were
available in tabulated form. For each source the classification
tool provided a slit image and plots of its spectrum, SED, log
SED and bolometric magnitudes (see Fig. 1 for examples).
Each author was free to use either the decision tree logic or
their own method of classification. There was also room for
the co-authors to add additional comments to the table.8
8Although we were able to use this tool to its conclusion and
generate the final classification table prior to Paul Ruffle’s unex-
pected death, the tool was intended to be available online indefin-
tely, for application on other data collections. Unfortunately, the
4 SAGE-SPEC POINT SOURCE
CLASSIFICATION
Table 3 lists the classification types, used in the decision tree
shown in Fig. 3, and counts for a total of 209 SMC point
sources, for which Spitzer-IRS staring mode observations are
available. The classifications are also shown overplotted on
the map of the SMC (Fig. 2).
Fig. 4 shows typical spectra of objects classified as one
of the four YSO type objects, as well as a typical IRS spec-
trum of an H ii region. The numbering 1–4 for the YSO
classes represent an evolutionary sequence, with YSO-1 be-
ing the most embedded and YSO-4 the most evolved type of
YSO, namely Herbig Ae Be (HAeBe) stars. This is evident
from the spectral appearance of the silicate feature, which
appears in absorption towards YSO-1 objects, then gradu-
ally in self-absorption (YSO-2), until it finally appears in
emission (YSO-3, YSO-4), for less embedded objects. The
spectra of the YSO-1 objects also show ice absorption fea-
tures, for instance the CO2ice feature at 15.2 µm, further
evidence of their early evolutionary phase. Polycyclic aro-
matic hydrocarbons (PAHs) are seen in the spectra of the
YSO-3, YSO-4 and H ii classes, indicative of a UV radiation
field. Atomic lines are also seen, particularly in YSO-3 and
Hii objects, and the latter category shows a rising contin-
uum indicative of cold dust in the vicinity of the ionizing
star.
Fig. 6 shows typical spectra in the group of oxygen-
rich evolved stars. The earliest type of O-rich AGB stars
are shown at the bottom of the plot (O-EAGB), and the
spectra shown here do not show any dust features, while the
signature of oxygen-rich molecular species may be present in
the spectra. A slight change of slope due to a small infrared
excess caused by thermal dust emission, may be visible in
the SED, however. Later type O-AGB stars, red supergiants
(RSG) and oxygen-rich post-AGB stars (O-PAGB), share
spectroscopic characteristics, such as the presence of silicate
emission features, although the detailed shape can be differ-
ent. To distinguish between the RSG and O-AGB category, a
bolometric luminosity cut of Mbol =−7.1 is used, while the
distinction between O-AGB and O-PAGB is based on the
presence of a detached shell where a double-peaked SED is
used as a criterion for the latter category. This is demon-
strated in the lower right panel of Fig. 5 showing the SED
of the sole O-PAGB object in the sample, LHA 115-S 38
(SMC IRS 257). O-rich PNe (O-PN) may still show a dis-
tinguishable oxygen-rich chemistry in their dust mineralogy
(although not in the case shown), but they are discriminated
from C-PN using the presence of PAH lines in the spectra
of the carbon-rich objects.
Fig. 7 gives an overview of the 5–38 µm spectral appear-
ance of carbon-rich evolved stars. Tracers of the carbon-rich
chemistry are the C2H2molecular absorption bands at 5.0,
7.5 and 13.7 µm, the SiC dust feature at 11.3 µm, the 21-
µm feature (which remains unidentified and really peaks at
20.1 µm), and the so-called 30-µm feature, which has re-
cently been suggested to be due to the same carbonaceous
internet service provider where this website was hosted recently
stopped providing this service and as of yet, the co-authors have
not been able to reconstruct and resurrect the website with the
classification tool.
c
0000 RAS, MNRAS 000, 000–000
Spitzer-IRS point source classification in the SMC 9
Figure 4. Example spectra in the YSO and H ii categories. From
bottom to top, we show examples of YSO-1 through 4, and a spec-
trum of an H ii region, in an evolutionary sequence from young to
more evolved. The spectra are labelled with their SMC IRS num-
ber. Discernible spectral features are indicated with tick marks
and labels at the top of the diagram. The heavily embedded YSO-
1 spectrum shows silicates and ices in absorption. The YSO-2
spectrum shows silicate in self-absorption. Silicate emission and
PAH features are visible in the spectrum of the YSO-3, YSO-4
and H ii objects, albeit i n different ratios. The H ii region has a
rising SED slope.
compound that carries the continuum (Otsuka et al. 2014),
while the MgS identification is under debate (Zhang et al.
2009; Lombaert et al. 2012). Again, the distinction between
the C-PAGB and C-AGB categories is based on whether
the SED is double-peaked, which is evidence for a detached
shell, indicating mass loss has stopped. Fig. 5 shows two
clear examples of this, namely SMC IRS 243 and SMC
IRS 268. SMC IRS 95 is confirmed to be a C-PAGB star
(Kraemer et al. 2006; van Loon et al. 2008), despite its ab-
sence of a double-peaked SED. C-PAGB and C-PN also show
the UV-excited PAH features, and in case of the C-PN, the
presence of atomic emission lines.
Fig. 8 shows typical spectra of a number of star-like cat-
egories, namely (from top to bottom) stellar photospheres,
with no discernible dust features in the spectrum; R CrB
stars, which only appear to show a dust continuum with
no spectral substructure; a Blue Supergiant in our sample,
which appears to have a strong far-infrared excess on top
of a stellar photosphere, and finally Wolf-Rayet stars, which
may form dust and show the corresponding infrared excess
10-5
10-4
10-3
10-2
10-1
1 10
10-5
10-4
10-3
10-2
10-1
1 10
λFλ (10-12 W m-2)
λ (µm)
SMC IRS 95 SMC IRS 243
SMC IRS 268 SMC IRS 257
Figure 5. SEDs of the four post-AGB stars in the sample. The
SMC IRS spectra 95, 243 and 268 correspond to C-PAGB ob-
jects, while SMC IRS 257 is the spectrum of a O-PAGB ob-
ject. The black lines represent the IRS spectra, while the red
diamonds correspond to the collected photometric measurements
for each source. The double-peaked structure used as a distin-
guishing feature is visible in the SEDs of 243, 268 and 257.
2MASS J00364631−7331351 (SMC IRS 95) does not show a dou-
ble peaked structure, but its C-PAGB nature is confirmed using
near-infrared spectroscopy and the PAH features in the IRS spec-
tra (See Appendix A).
in the spectrum. The Blue Supergiant and the Wolf-Rayet
star classifications are taken from the literature, and are not
based on the infrared spectroscopy. Similarly, Fig. 9 shows
the spectra of the object types group under OTHER, of
which the classifications are taken from the literature (see
Appendix A). The examples shown in Fig. 9 represent from
top to bottom a B[e] star, a foreground AGB star, an S star
and a symbiotic star. Their infrared spectra are not used to
achieve this classification, and the IRS data are plotted only
for illustration.
Figs. 10–12 show the 209 classified point sources on the
[8.0] versus J−[8.0] CMD and two different colour-colour di-
agrams (CCDs), overlayed on the SAGE-SMC point source
catalogue (Gordon et al. 2011) in gray scale.
The [8.0] versus J−[8.0] CMD (Fig. 10) shows a large
spread for the population of 209 ob jects. The stellar atmo-
spheres (STAR) and Wolf-Rayet stars have colours more or
less indistinguishable from the bulk of the SAGE-SMC cat-
alogue (with J−[8.0] ∼0–1 mag), and modest brightness at
8.0 µm, even though these stars are amongst the brightest
stars in the optical in the SMC. All other categories dis-
played are bright in the IRAC [8.0] band, and often show
considerable redness in their J−[8.0] colour. In this dia-
gram RSG, O-AGB/O-EAGB, C-AGB and YSOs (all classes
combined), are reasonably well separated from each other,
although there are some interlopers. It appears to be difficult
to separate PNe and YSOs on the one hand, and the most
extreme C-AGB stars and YSOs on the other hand. Distin-
c
0000 RAS, MNRAS 000, 000–000
10 Paul M. E. Ruffle et al.
Figure 6. Example spectra of different types of O-rich evolved
stars. The spectra are labelled with their SMC IRS number. At
the top the wavelengths of the silicate features and spectral lines
are labelled. At the bottom of the plot the spectrum of the E-
OAGB star closely resembles a stellar photosphere. The spectra of
the O-AGB, RSG and O-PAGB star all show the silicate emission
features. Distinguishing between these three types is not possible
from IRS spectroscopy alone, and additional information on SED
shape and bolometric luminosity is required. The top spectrum
is characeteristic of a O-PN.
guishing between the four YSO classes is also not possible
in this diagram.
Fig. 11 is a CCD composed of the four IRAC bands,
namely the [3.6]−[4.5] versus [5.8]−[8.0] colours. The advan-
tage of using a CCD is that it is distance independent. The
coverage of the sample of 209 objects fans out nicely over
colour-colour space. In this diagram, the stars and WR stars
no longer separate well from the rest of the sample, however
it now appears easier to separate YSOs from C-AGB stars on
the one hand and PNe on the other hand. However, C-PAGB
and O-PAGB stars probably overlap with the colour-colour
space taken up by YSOs, as they transition from the AGB
region to the PN region in the diagram. But as this phase
is short-lived, pollution of the colour-selected YSO sample
with post-AGB stars is limited. Subdivision within the YSO
category is still not possible, and also the O-(E)AGB objects
do not seem to occupy a unique part of colour-colour space.
Finally, Fig. 12 shows the J−KSversus [8.0]−[24] CCD.
In this CCD, the C-AGB objects are easily separated from
all other types of objects, as their J−KScolour quickly in-
creases, with increasing [8.0]−[24]. For O-(E)AGB stars and
Figure 7. Example spectra of C-rich evolved stars. From bottom
to top two C-AGB stars are shown, followed by a C-PAGB and a
C-PN. All four spectra are labelled with their SMC IRS number.
The tick marks at the top show the position of characteristic
spectral features. In the C-AGB spectra the molecular absorption
bands of C2H2are visible, as well as emission due to SiC and
possibly MgS (30µm, not in all sources). The C-PAGB object
no longer shows the molecular absoprtion bands, but the SiC
and the MgS (not always) are still visible. Other features include
PAH bands and the 21-µm feature. The infrared spectrum of C-
PN objects shows the spectral features due to PAHs and atomic
lines.
RSGs this increase is less steep, forming a separate branch in
the middle of the plot. O-PN and C-PN group together with
YSOs towards the top of the diagram, showing the highest
values of [8.0]−[24], against modest J−KSreddening.
5 COMPARISON WITH EXISTING COLOUR
CLASSIFICATIONS
These 209 spectral classifications will allow us to verify ex-
isting infrared photometric classification schemes that have
come out of recent studies of the Magellanic Clouds. We
compare our results with three distinct colour classifica-
tion schemes: a) the JHK Scolour classification scheme
for evolved stars by Cioni et al. (2006), expanded by
Boyer et al. (2011) to include mid-IR wavelengths; b) the
IRAC classification scheme for AGB stars and RSGs by
Matsuura et al. (2013), which is based on the previous
work by Matsuura et al. (2009) on the LMC; and c) the
Spitzer classification scheme to select YSO candidates by
c
0000 RAS, MNRAS 000, 000–000
Spitzer-IRS point source classification in the SMC 11
Figure 10. [8.0] vs J−[8.0] CMD. The gray scale in the background represents a density plot of the SAGE-SMC point source catalogue,
while the coloured symbols are the sources classified in this work. For clarity, the plot is duplicated with the coloured symbols spread
out over the left- and right-hand plot according to the legend.
Figure 11. [5.8]−[8.0] vs [3.6]−[4.5] all-IRAC CCD. Description as in Fig. 10.
Sewi lo et al. (2013), based on earlier work by Whitney et al.
(2008).
5.1 Boyer et al. (2011)
In order to classify all evolved stars in the SAGE-SMC
(Gordon et al. 2011) data, Boyer et al. (2011) devised a clas-
sification scheme, based on the 2MASS classification scheme
presented by Cioni et al. (2006). The basis for this classifica-
tion scheme is the Kversus J−KSCMD (Fig. 13), showing
the cuts for the RSG, O-AGB and C-AGB object classes, su-
perposed on the SAGE-SMC point sources (grey pixels). The
dotted line refers to the tip of the Red Giant Branch (RGB),
and Boyer et al. (2011) use this line to separate RGB stars
from the AGB and RSG categories. Photometrically classi-
fied objects from Boyer et al. (2011) are shown as coloured
pixels. At the red end of the diagram, some photometrically
classified C-AGB stars appear below the diagonal cut, or
c
0000 RAS, MNRAS 000, 000–000
12 Paul M. E. Ruffle et al.
Figure 12. [8.0]−[24] vs J−KSCCD. Description as in Fig. 10.
even below the dotted line corresponding to the tip of the
RGB. These objects are called extreme AGB stars, predom-
inantly carbon-rich, which are defined as J−[3.6] >3.1
mag (e.g., Blum et al. 2006). Often these objects are so red,
that they are not detected in 2MASS JHK S, and alterna-
tive selection criteria in the mid-infrared are required. The
reader is referred to Boyer et al. (2011) for a detailed de-
scription. The larger symbols in Fig. 13 represent the sample
of Spitzer-IRS spectroscopically classified ob jects described
in this work.
5.1.1 C-AGB stars
Objects showing the C-AGB spectral signature are
all classified as either C-AGB or x-AGB (most of
which are expected to be C-rich), according to the
Boyer et al. (2011) classification, although three of these
objects are not included in the catalogue published by
Boyer et al. (2011). These objects (OGLE SMC-LPV-
7488 (SMC IRS 44; SSTISAGEMA J004903.78−730520.1);
2MASS J01060330−7222322 (SMC IRS 109; SSTISAGEMA
J010603.27−722232.1); and 2MASS J00561639−7216413
(SMC IRS 129; SSTISAGEMA J005616.36−721641.3)) are
among a larger set of objects that had not been properly
matched between the IRAC Epochs in the mosaicked cata-
logue (Srinivasan et al. in prep.). These objects were thus
missing photometric measurements in the point source cata-
logue, in some bands, and were therefore not classified (cor-
rectly) by Boyer et al. (2011). The on-line table described
by Table 2 shows the correct photometry for these three
targets.
5.1.2 Red Supergiants
Most of the 21 spectroscopically classified RSGs indeed fall
within the RSG strip defined by Boyer et al. (2011). Only
three of the objects classified as an RSG by us were clas-
sified differently by Boyer et al. (2011) using only the pho-
tometry: HV 11417 (SMC IRS 115) is designated as a FIR
object by Boyer et al. (2011); Massey SMC 55188 (SMC IRS
232) as an O-AGB star; and IRAS F00483−7347 (SMC IRS
98) as an x-AGB star. HV 11417 is a variable star with
a period of 1092 days, and an amplitude in the I band of
1.9 mag (Soszy´nski et al. 2011). Due to this large ampli-
tude, the timing of the observations affects the IR colours
of the object significantly, which is why the photometry
measurements show a small positive slope between 8 and
24 µm ([8.0] −[24] >2.39 mag), causing it to be classified
as a FIR object by Boyer et al. (2011), while the slope of
the IRS spectrum is distinctly negative. If we disregard the
FIR colour cut, which is meant to separate AGB stars from
objects such as YSOs, PNe and background galaxies, this
object would have been classified as an O-AGB star, which
deviates from our determination of RSG, and from past clas-
sifications (Elias et al. 1980). The bolometric luminosity of
this object is close to the RSG/O-AGB boundary, and could
be affected by the variability of the object too. The large am-
plitude favours a classification as luminous AGB star over
a RSG. RSGs were separated from O-AGB stars using the
classical AGB bolometric magnitude limit (Fig. 3), but this
boundary is not absolute. AGB stars undergoing hot bottom
burning can be brighter than this limit, while less-evolved
RSGs can be fainter. This causes some disagreement in clas-
sifications of stars near the boundary.
Alternatively, the luminosity boundary between RSG
and O-AGB may not be properly represented by the cuts in
the KSvs J−KSdiagram from Boyer et al. (2011). Similary,
the misclassification of Massey SMC 55188 also suggests that
the bolometric cut we applied to distinguish between O-
AGB stars and RSGs does not correspond to the boundaries
between these two categories in the KSvs J−KSCMD.
IRAS F00483−7347 clearly shows an oxygen-rich chemistry
c
0000 RAS, MNRAS 000, 000–000
Spitzer-IRS point source classification in the SMC 13
Figure 13. The Boyer et al. (2011) photometric classification, based on the work by Cioni et al. (2006), applied to the SMC. Boyer et al.
(2011) begin with the J−KSclassification shown here, then use 3.6 and 8-µm photometry to recover and classify the dustiest sources.
The slanted lines in this KSvs J−KSCMD represent the boundaries of the C-AGB, O-AGB and RSG categories, and the horizontal
dotted line shows the boundary between the tip of the RGB and the earlier type AGB stars. In gray scale the SAGE-SMC point source
catalogue is shown in the form of a density plot. The lilac, yellow and dark pink dots represent the RSG, O-AGB and C-AGB objects,
as they are classified in the KSvs J−KSCMD. The light pink dots are the x-AGB stars, selected by Boyer et al. (2011), replacing any
KSvs J−KSclassification. The x-AGB objects fall mostly within the C-AGB boundaries. The coloured symbols represent the objects
spectrally classified in this work, following the legends and spread out over two panels for clarity.
in its spectrum, with the presence of the amorphous silicate
bands at 9.7 and 18 µm. The object is heavily embedded,
with a very red SED, and the 9.7-µm feature starts to show
signs of self-absorption. IRAS F00483−7347 demonstrates
that not all x-AGB objects are in fact carbon-rich AGB
stars.
5.1.3 O-AGB stars
All objects classified as O-EAGB in this work are similarly
classified as O-AGB by Boyer et al. (2011). The spectrally-
confirmed O-AGB objects, however, show a much wider
spread in colour-magnitude space than the defined strip, and
they encroach on the photometrically-defined RSG, C-AGB
and FIR categories. In particular, only two spectroscopi-
cally classified O-AGB stars, HV 12149 (SMC IRS 45) and
2MASS J00444463−7314076 (SMC IRS 259) are classified as
O-AGB based on photometry. One object (HV 1375; SMC
IRS 110) is classified as a C-AGB based on its photomet-
ric colours, and a further six ob jects (RAW 631, 2MASS
J00463159−7328464, 2MASS J00445256−7318258, BMB-B
75, IRAS F01066−7332 and HV 12956) are apparently suf-
ficiently red ([8.0] −[24] >2.39 mag) to be classified as
FIR, even though they are not background galaxies, YSOs
or PNe. These six objects are heavily embedded, and show
the effect of a high optical depth in the relative strength of
the 18 µm silicate band with respect to the 9.7-µm band.
They also show evidence for the presence of crystalline sili-
cates in their spectrum, another sign of high optical depth
(Kemper et al. 2001; Jones et al. 2012). Finally, HV 2232
(SMC IRS 230) and HV 11464 (SMC IRS 309) are photo-
metrically classified as Red Supergiants (Boyer et al. 2011)
(see the table described by Table 2).
5.1.4 Additional sources
The further nine interlopers in the C-AGB/x-AGB section
of the CMD defined by Boyer et al. (2011) are a mixed
bag of objects, all but one falling in the x-AGB category.
These objects reflect selection biases and include rare cat-
egories of previously known types such as R CrB stars
(2MASS J00461632−7411135, 2MASS J00571814−7242352
and OGLE SMC-SC10 107856), an S star (BFM 1) and a
B[e] star (Lin 250). These object types are not included
in the Boyer et al. (2011) classification scheme, and could
therefore never have agreed with the spectral classification.
True misclassifications are the objects thought to be C-rich
AGB stars, only to be revealed to be something else based
on their IRS spectroscopy. These include O-PAGB star
LHA 115-S 38, YSO-2 object 2MASS J01050732−7159427,
and RSG IRAS F00483−7347. NGC 346:KWBBe 200 is
a special case, as it was thought to be a B[e] supergiant
(Wisniewski et al. 2007a), but its classification has recently
been revised to a YSO (Whelan et al. 2013), in line with our
classification of YSO-3 (See Appendix A).
The FIR category defined by Boyer et al. (2011) was
introduced to exclude YSO s, compact H ii regions, PNe and
background galaxies from the AGB/RSG sample, by apply-
ing the [8] −[24] >2.39 mag cut, corresponding to a rising
continuum. Indeed, 14 out of 23 FIR objects are either C-PN
c
0000 RAS, MNRAS 000, 000–000
14 Paul M. E. Ruffle et al.
Figure 8. Example spectra of ob jects with stellar photospheres
or similar. Shown here from bottom to top are the spectra of a
WR star, a BSG, a R CrB star and a regular stellar photosphere,
labelled with their SMC IRS number. The WR star shows some
atomic lines in its spectrum (marked at the top of the diagram),
but the other spectra are rather featureless. The infrared excess
in the R CrB and BSG objects are caused by dust emission.
(three objects) or YSOs (11 objects), according to their IRS
spectra. The remaining nine FIR objects include the six O-
AGB stars discussed in Sec. 5.1.3, but also a symbiotic star
(SSTISAGEMA J005419.21−722909.7; SMC IRS 260) and a
C-PAGB (2MASS J00364631−7331351; SMC IRS 95), both
of which are classified based on their IRS spectra.
5.2 Matsuura et al. (2013)
The second classification scheme was proposed by
Matsuura et al. (2009) to identify mass-losing O-rich and
C-rich AGB stars, as well as Red Supergiants, in order to
estimate the dust production budget in the LMC, separated
out by C-rich and O-rich chemistries. The scheme is based
on the IRAC bands, using the [8.0] vs [3.6]−[8.0] CMD, and
separates out the O-rich AGB stars and RSGs on the one
hand, and C-rich AGB stars on the other hand (Fig. 14).
Matsuura et al. (2009) used only this diagram to classify
the IRAC point sources in the LMC, but for the SMC,
Matsuura et al. (2013) added an extra step to overcome pol-
lution between the categories for redder [3.6]−[8.0] sources.
Indeed, from Fig. 14 it is clear that the red part of the C-
AGB section is dominated by sources identified as YSOs (tan
diamonds), while a significant fraction of the O-AGB objects
Figure 9. Example IRS spectra of objects in the OTHER cat-
egory. From bottom to top we show the spectra of a symbiotic
star, an S star, a foreground (Galactic) O-AGB star and a B[e]
star, all labelled with their SMC IRS number. Relevant spectral
features due to silicates are labelled at the top of the figure.
(blue diamonds) also fall within the C-AGB section. In the
second step, MIPS and NIR data are included, and a com-
plex series of cuts is made in the KS–[24] vs KS–[8.0] CCD
(see Fig. 5 in Matsuura et al. 2013, ; the equations are not
provided). In cases where the classification using the KS–
[24] vs KS–[8.0] diagram (Fig. 15) deviates from the [8.0]
vs [3.6]−[8.0] classification, the KS–[24] vs KS–[8.0] classi-
fication takes preference. Most spectral classifications agree
with the photometric classification in the second step (see
Fig. 15). In case where sources were only classified in one of
the two steps, we used that classification. We have executed
this classification method for our 209 sources, and included
the results in the table described by Table 2.
5.2.1 Carbon-rich AGB stars
The two-step classification described by Matsuura et al.
(2013) correctly identifies most of the spectroscopically
classified C-AGB stars in our sample as such. In only
three cases was an object photometrically classified as
being oxygen-rich (RSG/O-AGB), while the spectroscopy
shows the carbon-rich nature of the source. These sources
are 2MASS J00515018−7250496 (SMC IRS 103); 2MASS
J00524017−7247276 (SMC IRS 127) and IRAS 00350−7436
(SMC IRS 238). IRAS 00350−7436 is among the brightest
mid-infrared objects in the SMC, and falls above the C-AGB
c
0000 RAS, MNRAS 000, 000–000
Spitzer-IRS point source classification in the SMC 15
Figure 14. First step of the Matsuura et al. (2013) photometric classification method, based on the work by Matsuura et al. (2009),
applied to the SMC. The solid lines in this [8.0] vs [3.6]-[8.0] CMD represent the boundaries of the C-AGB category, and provide a lower
boundary to the RSG/O-AGB category. The dashed line shows the position of the YSO limit used by Boyer et al. (2011). All colours
have the same meaning as in Fig. 13.
Figure 15. Second step of the Matsuura et al. (2013) photometric classification method. The solid lines in this KS–[24] vs KS–[8.0]
CCD represent the boundaries of the various evolved star categories. All colours have the same meaning as in Fig. 13.
cut in Fig. 4 of Matsuura et al. (2013), due to its exceptional
brightness.
Furthermore, the classification of several C-AGB stars
by Matsuura et al. (2013) disagrees with the spectral clas-
sification. These include: a C-rich post-AGB star (2MASS
J00364631−7331351; SMC IRS 95), four O-rich early-type
AGB stars (HV 1366, HV 11303, HV 838 and HV 12122;
SMC IRS 36, 38, 39 and 116, respectively), a WR star
(HD 5980; SMC IRS 281), a RSG (HV 11262; SMC IRS
111), a YSO-3 object (NGC 346:KWBBe 200; SMC IRS
19), the three R CrB stars (2MASS J00461632−7411135,
2MASS J00571814−7242352 and OGLE SMC-SC10 107856;
SMC IRS 94, 114 and 245, respectively), the two fore-
ground O-rich AGB stars NGC 362 SAW V16 and HV 206,
c
0000 RAS, MNRAS 000, 000–000
16 Paul M. E. Ruffle et al.
the S star BFM 1 and the symbiotic star SSTISAGEMA
J005419.21−722909.7. The last four objects are all in the
OTHER category, which contains subclasses the work by
Matsuura et al. (2013) does not seek to classify. The four
O-rich early-type objects have actually long been recognized
as such, in some cases their nature was already known in the
1980s, and they fall only slightly outside the boundaries in
Fig. 5 of Matsuura et al. (2013). The RSG HV 11262 has
very similar KS–[8.0] and KS–[24] colours to the four O-
EAGB objects. 2MASS J00364631−7331351 is not properly
filtered out by Fig. 5, as it falls just below the KS–[24] =
8 mag cutoff that is meant to exclude YSO, C-PN, and C-
PAGB objects from the C-AGB category. The R CrB stars
represent a rare class which, not surprisingly, overlaps in in-
frared colours with the C-AGB stars, as it is believed that
the dust in these stars is carbonaceous (Feast 1986). HD
5980 is the only WR star that is classified by Matsuura et al.
(2013). Although other WR stars do appear in Fig. 14, they
do not receive a classification as they are too faint in the [8.0]
band. HD 5980 is considerably brighter in the [8.0], and is
the only WR star with a MIPS-[24] detection in the SMC
(Bonanos et al. 2010). Closer inspection of the IRS spectrum
seems to suggest that a chance superposition with a compact
Hii region gives rise to this 24 µm detection, and is actually
not a detection of the WR star itself. Thus, the position of
HD 5980 in Fig. 15 and the classification following from it,
should be disregarded. Finally, NGC 346:KWBBe 200 is a
curious object, which, from its IRS spectrum appears to be a
YSO, but has characteristics in common with B[e] stars (See
Appendix A). It falls well outside the box defined for YSO,
C-PN and C-PAGB, perhaps due to its unusual nature.
5.2.2 Oxygen-rich AGB stars and RSGs
Since the main purpose of the classification scheme by
Matsuura et al. (2013) is to determine the dust production
by evolved stars distinguished by carbon-rich and oxygen-
rich chemistry, the subdivision in types of evolved stars
within the oxygen-rich class is less important. In fact, in the
first step, all types of oxygen-rich evolved stars (AGB stars
and RSGs) are lumped together as RSG/O-AGB (Fig. 14),
while in the second step a subdivision is made between
O-AGB (which also includes RSG) and O-AGB/O-PAGB
(Fig. 15). The latter category contains the more evolved
O-AGB stars. Thus, a total of three partially overlapping
categories exist. The RSG/O-AGB category contains ob-
jects that are classified as such in the first step, but re-
mained unclassified in the second step. This class contains
only three objects (2MASS J00515018−7250496, 2MASS
J00524017−7247276 and IRAS 00350−7436), which are all
C-AGB according to their IRS spectroscopy, and are already
discussed in Sec. 5.2.1.
The 24 objects classified as O-AGB in the second step
are indeed all O-AGB or RSG objects according to their IRS
spectroscopy. However, the category O-PAGB/O-AGB con-
tains a more diverse range of objects. Of the eight SMC IRS
objects classified by the method of Matsuura et al. (2013)
to be in this category only four are genuine O-rich evolved
stars: IRAS F00483−7347 (SMC IRS 98) is a RSG according
to our classification, LHA 115-S 38 (SMC IRS 257) is found
to be a O-PAGB ob ject, and 2MASS J00463159−7328464
(SMC IRS 121) and HV 12956 (SMC IRS 277) are O-AGB
stars. The other objects in this category include two carbon-
rich evolved stars, 2MASS J00444111−7321361 (SMC IRS
268), a C-PAGB object, and Lin 343 (SMC IRS 161), a C-
PN. Both of these objects should have fallen in the YSO/C-
PN/C-PAGB, but with KS–[24] just below 8 mag they both
just miss the cutoff. Since Fig. 15 does not show any O-
rich ob jects in the vicinity of the K–[24] = 8 mag cutoff
between C-rich and O-rich post-AGB objects, a case can be
made to lower the cutoff slightly. Finally, the remaining two
misclassified objects in the O-PAGB/O-AGB category by
Matsuura et al. (2013) are spectrally classified as B[e] stars
in the OTHER category. They are RMC 50 (SMC IRS 193)
and Lin 250 (SMC IRS 262).
When checking the reverse direction, we find that all
objects classified by us as O-AGB are indeed identified as
either O-AGB or as O-PAGB/O-AGB according to the clas-
sification scheme by Matsuura et al. (2013). However, this
classification scheme misclassifies all objects identified as O-
EAGB by us. According to the scheme by Matsuura et al.
(2013), they are either C-AGB (See Sect. 5.2.1; four out of
eight objects), or remain unclassified, as they have [3.6] −
[8.0] <0.7 mag, rendering them unclassifiable in Fig. 14, and
have KS−[8.0] <0.5 mag and KS−[24] <0.5 mag, which
causes them to fall outside all boundaries in Fig. 15. This
also happens to six of the 22 ob jects classified by us to be
RSGs; one is misclassified as a C-AGB (See Sect. 5.2.1), and
the other five do not receive a classification because their
colours are too blue (i.e. their mass-loss rates are too low).
Our sample of 209 targets with IRS spectra contains
only one oxygen-rich post-AGB star, which agrees with
the O-PAGB/O-AGB classification from Matsuura et al.
(2013).
5.3 Sewi lo et al. (2013)
The third classification scheme that we compare with is the
method developed by Whitney et al. (2008) and refined by
Sewi lo et al. (2013) to select YSO candidates. From the set
of CMDs used by Whitney et al. (2008), Sewi lo et al. (2013)
selected a combination of five different CMDs to select YSO
candidates in the SMC. Two of these diagrams are repro-
duced in this work, with the 209 IRS point sources overplot-
ted (Figs. 16 and 17). After the initial colour selection of
YSO candidates, Sewi lo et al. (2013) performed additional
tests, including a visual inspection of the imaging, a check
against the SIMBAD and other catalogues for known non-
YSO sources, and fitting against the YSO SED grid calcu-
lated by Robitaille et al. (2006). Sewi lo et al. (2013) arrive
at a list of approximately 1000 ‘high-reliability’ and ‘proba-
ble’ YSOs in the SMC.
Although Figs. 16 and 17 appear to show considerable
amount of pollution from non-YSO IRS staring mode tar-
gets, as well as many confirmed YSOs based on the IRS
data straying out of the defined boxes, it is the combina-
tion of the five CMDs, along with the additional non-colour
based checks that yields a highly reliable YSO candidate list.
Indeed, the Sewi lo et al. (2013) classification favours relia-
bility over completeness, and CMD areas with significant
pollution due to other sources (background galaxies, PNe)
have been excluded. The list of 209 IRS staring mode point
sources contains nine objects classified as probable YSOs by
Sewi lo et al. (2013), and 45 high-reliability YSO candidates.
c
0000 RAS, MNRAS 000, 000–000
Spitzer-IRS point source classification in the SMC 17
Figure 16. One of the CMD diagrams used in the Sewi lo et al. (2013) selection method for candidate YSOs, based on the work by
Whitney et al. (2008). This [8.0] vs [4.5]-[8.0] CMD diagram shows the cut used in solid black lines. In gray scale the SAGE-SMC point
source catalogue is shown in the form of a density plot. The light blue dots represent the YSO candidates finally selected by Sewi lo et al.
(2013). The coloured symbols show the location of the objects spectrally classified in this work, spread out over two panels for clarity.
(Sewi lo et al. 2013), based on (Whitney et al. 2008)
Figure 17. As Fig. 16, but now for [8.0] vs [8.0]-[24].
Of the high-reliability YSOs for which IRS observations are
available, we find that the vast majority are indeed YSOs
(covering all four classes). Only two objects turn out to
be something different: H i i region IRAS 00436−7321 (SMC
IRS 310) and C-PAGB object 2MASS J01054645−7147053
(SMC IRS 243). Among the probable YSO candidates, the
success rate is lower: four out of nine are not YSOs, upon
inspection of their IRS spectra. RMC 50 (SMC IRS 193) is
a B[e] star, while SMP SMC 11 (SMC IRS 32), LHA 115-N
43 (SMC IRS 155) and Lin 49 (SMC IRS 292) are actually
C-PN objects.
Furthermore, there are three YSO-3 type objects, as
spectrally classified, that are not identified as YSO candi-
dates by Sewi lo et al. (2013). These are NGC 346:KWBBe
c
0000 RAS, MNRAS 000, 000–000
18 Paul M. E. Ruffle et al.
200 (SMC IRS 19), 2MASS J00465185−7315248 (SMC IRS
269) and LHA 115-N 8 (SMC IRS 274).
6 SUMMARY
We have analysed all 311 Spitzer-IRS staring mode obser-
vations within the SAGE-SMC IRAC and MIPS coverage
of the SMC (Gordon et al. 2011). After removing IRS ob-
servations of extended emission, blank sky and duplicate
observations, we find that 209 unique IRAC point sources
were targeted. We applied the infrared spectroscopic clas-
sification method devised by Woods et al. (2011), with the
addition of one more category, namely O-EAGB stars (early-
type oxygen-rich AGB stars). We find that the Spitzer-IRS
staring mode sample of point sources in the SMC contains
51 YSOs, subdivided in 14 embedded YSOs (YSO-1), 5 less-
embedded YSOs (YSO-2), 22 evolved YSOs (YSO-3) and 10
HAeBe type objects (YSO-4). Furthermore, we find the sam-
ple contains 46 oxygen-rich evolved stars: 8 O-EAGB stars,
11 O-AGB stars, 22 RSGs, 1 O-PAGB object and 4 O-PN
objects. 62 objects turn out to be carbon-rich evolved stars,
namely 39 C-AGB stars, 3 C-PAGB objects and 20 C-PN
objects. The sample also includes 3 R CrB stars, 1 BSG
and 10 WR stars. 27 objects show stellar photospheres, 23
of which were selected based on their MIPS-[24] excess, and
labelled by us as dusty OB stars. It turns out that the 24-
µm emission is in general not related to the host OB star
(Adams et al. 2013; Sheets et al. 2013). Finally, the sample
includes a small number of other objects (OTHER), which
do not follow from the classification method. These include
2 B[e] stars, 2 foreground oxygen-rich AGB stars, an S star,
and a symbiotic star.
In this work, we have compared the resulting spectral
classifications with the outcome of photometric classifica-
tion schemes. It should be noted that the spectroscopic ob-
servations are obtained in 14 different observing programs,
with a diverse range of science goals. Thus, there is no ho-
mogeneous coverage of colour-magnitude space, and it will
be impossible to quantify the goodness of any given pho-
tometric classification method. Furthermore, the observing
programs tend to target rare types of sources in a dispro-
portionate amount, and some of these rare type of sources
(R CrB stars, WR stars, B[e] stars, etc.) are not included
in photometric classification schemes, precisely because they
are rare. Thus, these objects tend to pollute the classifica-
tion schemes discussed here, but statistically they are rather
insignificant.
We reviewed three different photometric classification
schemes for infrared sources in the SMC: the schemes by
Boyer et al. (2011) and Matsuura et al. (2013) for evolved
stars, and the classification scheme to select candidate YSOs
by Sewi lo et al. (2013). The latter scheme is not a pure
photometric classification, as it includes additional steps,
such as visual inspection of the direct environment of the
point source in imaging, checks against existing catalogues
and fitting against the grid of YSO SEDs calculated by
Robitaille et al. (2006). However, as discussed in Sec. 5.3,
the 54 overlapping sources from this work with the result-
ing YSO candidate list are mostly correctly classified. Only
a few sources are misclassified in either direction, i.e. three
spectroscopically confirmed YSOs were not on the candidate
list by Sewi lo et al. (2013), and six sources on the candidate
list were found to be something else upon inspection of their
IRS spectroscopy. All-in-all we conclude that the YSO can-
didate list produced by Sewi lo et al. (2013) is reliable, with
48/54 sources indeed being YSOs and the high-reliability
sources doing better than the probable sources. Only three
spectroscopically confirmed YSOs were missed due to un-
usual infrared colors.
The two photometric classification methods for evolved
stars can be directly compared to each other. The method
by Boyer et al. (2011) has its focus on identifying the en-
tire dusty evolved star population, while the Matsuura et al.
(2013) method is mainly driven by the motivation to de-
termine the carbon-rich and oxygen-rich dust production
rates. Thus, in the later method, correct identification of
lower mass-loss rate stars is not so important. Both meth-
ods can be used to estimate the integrated dust production
rate. Due to the low metallicity of the SMC, carbon-rich
evolved stars are more numerous (Blanco et al. 1978, 1980;
Lagadec et al. 2007), and the carbon stars have, on average,
the highest mass-loss rates. Thus, identifying carbon stars
correctly is important as they dominate the dust budget
(Matsuura et al. 2009; Boyer et al. 2012; Riebel et al. 2012).
Apart from rare object classes, Boyer et al. (2011) very ef-
ficiently separate C-AGB stars from the other classes, only
classifying one non-C-AGB star as C-AGB star, while classi-
fying all genuine C-AGB objects as C-AGB. Matsuura et al.
(2013) do slightly worse, with three genuine C-AGB objects
being classified as something else, and a number of objects,
including some rare types, incorrectly classified as C-AGB
star.
On the oxygen-rich side, we find that Matsuura et al.
(2013) perform better for the high-mass loss rate objects (O-
AGB stars), but that they perform rather poorly on the low-
mass rate objects (O-EAGB stars), while the performance
of the Boyer et al. (2011) classification method is reversed
because of overlap between high-mass AGB stars and RSGs
near the classical AGB limit.
Finally, we note that of the two steps involved in
the classification method by Matsuura et al. (2013), the
second step (Fig. 15) matches very well with the ac-
tual spectroscopic classification. It has been introduced by
Matsuura et al. (2013) as a correction on the first step from
Matsuura et al. (2009), but almost all photometric classi-
fications currently included in the online table described
by Table 2 correspond to the second step, and most of
the time match the spectroscopic classification, rendering
the first step practically unnecessary. Thus, in case of the
Matsuura et al. (2013) classification scheme, applying only
the second step, as demonstrated in Fig. 15, would suffice
for dusty sources.
ACKNOWLEDGEMENTS
The authors wish to thank Paul Ruffle’s partner Rose
Wheeler. Rose provided access to Paul’s notes and files,
which allowed us to finish this work. PMER thanks
Academia Sinica Institute of Astronomy and Astrophysics
(ASIAA) for their financial support and hospitality during
the preparation of this work. The authors thank David Whe-
lan for making available Spitzer spectra of point sources
c
0000 RAS, MNRAS 000, 000–000
Spitzer-IRS point source classification in the SMC 19
described in his 2013 paper. Astrophysics at JBCA is sup-
ported by STFC. F.K. acknowledges support from the for-
mer National Science Council and the Ministry of Science
and Technology in the form of grants NSC100-2112-M-
001-023-MY3 and MOST103-2112-M-001-033-. B.A.S. ac-
knowledges support from NASA ADP NNX11AB06G.
R.Sz. acknowledges support from the Polish NCN grant
2011/01/B/ST9/02031. The research presented here was
conducted within the scope of the HECOLS International
Associated Laboratory, supported in part by the Polish NCN
grant DEC-2013/08/M/ST9/00664 (E.L.; R.Sz). This work
is based (in part) on observations made with the Spitzer
Space Telescope, obtained from the NASA/IPAC Infrared
Science Archive, both of which are operated by the Jet
Propulsion Laboratory, California Institute of Technology
under a contract with the National Aeronautics and Space
Administration. This publication makes use of data prod-
ucts from the Wide-field Infrared Survey Explorer, which is
a joint project of the University of California, Los Angeles,
and the Jet Propulsion Laboratory/California Institute of
Technology, funded by the National Aeronautics and Space
Administration. Some of the data presented in this paper
were obtained from the Mikulski Archive for Space Tele-
scopes (MAST). STScI is operated by the Association of
Universities for Research in Astronomy, Inc., under NASA
contract NAS5-26555. Support for MAST for non-HST data
is provided by the NASA Office of Space Science via grant
NNX13AC07G and by other grants and contracts. This re-
search has also made use of the SAGE CASJobs database,
which is made possible by the Sloan Digital Sky Survey Col-
laboration; SAOImage DS9, developed by Smithsonian As-
trophysical Observatory; the VizieR catalogue access tool,
CDS, Strasbourg, France; the SIMBAD database, operated
at CDS, Strasbourg, France; and NASA’s Astrophysics Data
System Bibliographic Services.
REFERENCES
Abia C., Cunha K., Cristallo S., de Laverny P., Dom´ınguez
I., Recio-Blanco A., Smith V. V., Straniero O., 2011, ApJ,
737, L8
Adams J. J. et al., 2013, ApJ, 771, 112
Allard F., Homeier D., Freytag B., 2011, in ASP Conference
Series, Vol. 448, 16th Cambridge Workshop on Cool Stars,
Stellar Systems, and the Sun, Johns-Krull C., Browning
M. K., West A. A., eds., p. 91
Ardeberg A., Maurice E., 1977, A&AS, 30, 261
Arp B. H., 1959, AJ, 64, 254
Azzopardi M., Breysacher J., 1980, A&AS, 39, 19
Azzopardi M., Vigneau J., 1982, A&AS, 50, 291
Balona L. A., 1992, MNRAS, 256, 425
Barba R. H. et al., 1996, in Rev. Mex. A. A. Conference
Series, Niemela V., Morrell N., Pismis P., Torres-Peimbert
S., eds., Vol. 5, p. 85
Barba R. H., Niemela V. S., Baume G., Vazquez R. A.,
1995, ApJ, 446, L23
Barnbaum C., Kastner J. H., Morris M., Likkel L., 1991,
A&A, 251, 79
Basinski J. M., Bok B. J., Bok P. F., 1967, MNRAS, 137,
55
Battinelli P., 1991, A&A, 244, 69
Bayne G. et al., 2002, MNRAS, 331, 609
Bernard-Salas J., Peeters E., Sloan G. C., Gutenkunst S.,
Matsuura M., Tielens A. G. G. M., Zijlstra A. A., Houck
J. R., 2009, ApJ, 699, 1541
Bica E. L. D., Schmitt H. R., 1995, ApJS, 101, 41
Blanco B. M., Blanco V. M., McCarthy M. F., 1978, Na-
ture, 271, 638
Blanco V. M., Blanco B. M., McCarthy M. F., 1980, ApJ,
242, 938
Blanco V. M., Frogel J. A., McCarthy M. F., 1981, PASP,
93, 532
Blum R. D. et al., 2006, AJ, 132, 2034
Bolatto A. D. et al., 2007, ApJ, 655, 212
Bonanos A. Z. et al., 2010, AJ, 140, 416
Bouchet P., Lequeux J., Maurice E., Prevot L., Prevot-
Burnichon M. L., 1985, A&A, 149, 330
Boyer M. L. et al., 2012, ApJ, 748, 40
Boyer M. L. et al., 2011, AJ, 142, 103
Brett J. M., 1991, MNRAS, 249, 538
Breysacher J., Fran¸cois P., 2000, A&A, 361, 231
Breysacher J., Moffat A. F. J., Niemela V. S., 1982, ApJ,
257, 116
Breysacher J., Perrier C., 1980, A&A, 90, 207
Buscombe W., Kennedy P. M., 1962, J. R. As-
tron. Soc. Canada, 56, 113
Castilho B. V., Gregorio-Hetem J., Spite F., Spite M., Bar-
buy B., 1998, A&AS, 127, 139
Catchpole R. M., Feast M. W., 1981, MNRAS, 197, 385
Cellone S. A. et al., 1996, in Rev. Mex. A. A. Conference
Series, Niemela V., Morrell N., Pismis P., Torres-Peimbert
S., eds., Vol. 5, p. 123
Charmandaris V., Heydari-Malayeri M., Chatzopoulos E.,
2008, A&A, 487, 567
Chen Y.-P., Trager S. C., Peletier R. F., Lan¸con A.,
Vazdekis A., Prugniel P., Silva D. R., Gonneau A., 2014,
A&A, 565, A117
Chiosi E., Vallenari A., Held E. V., Rizzi L., Moretti A.,
2006, A&A, 452, 179
Cidale L., Zorec J., Tringaniello L., 2001, A&A, 368, 160
Cioni M.-R. L. et al., 2003, A&A, 406, 51
Cioni M.-R. L., Girardi L., Marigo P., Habing H. J., 2006,
A&A, 448, 77
Clark J. S., Bartlett E. S., Coe M. J., Dorda R., Haberl F.,
Lamb J. B., Negueruela I., Udalski A., 2013, A&A, 560,
A10
Cohen J. G., 1984, AJ, 89, 1779
Contursi A. et al., 2000, A&A, 362, 310
Crowther P. A., Hadfield L. J., 2006, A&A, 449, 711
Dachs J., 1970, A&A, 9, 95
de Boer K. S., Savage B. D., 1980, ApJ, 238, 86
De Smedt K., Van Winckel H., Kamath D., Karakas A. I.,
Siess L., Goriely S., Wood P., 2014, A&A, 563, L5
De Smedt K., Van Winckel H., Karakas A. I., Siess L.,
Goriely S., Wood P. R., 2012, A&A, 541, A67
Dopita M. A., Bell J. F., Chu Y.-H., Lozinskaya T. A.,
1994, ApJS, 93, 455
Dopita M. A., Lawrence C. J., Ford H. C., Webster B. L.,
1985, ApJ, 296, 390
Dufour R. J., Killen R. M., 1977, ApJ, 211, 68
Dufton P. L., McErlean N. D., Lennon D. J., Ryans R. S. I.,
2000, A&A, 353, 311
Eenens P. R. J., Morris P., 1996, in Rev. Mex. A. A. Con-
c
0000 RAS, MNRAS 000, 000–000
20 Paul M. E. Ruffle et al.
ference Series, Niemela V., Morrell N., Pismis P., Torres-
Peimbert S., eds., Vol. 5, p. 120
Elias J. H., Frogel J. A., Humphreys R. M., 1980, ApJ, 242,
L13
Elias J. H., Frogel J. A., Humphreys R. M., 1985, ApJS,
57, 91
Epchtein N. et al., 1999, A&A, 349, 236
Evans C. J., Crowther P. A., Fullerton A. W., Hillier D. J.,
2004a, ApJ, 610, 1021
Evans C. J., Howarth I. D., 2008, MNRAS, 386, 826
Evans C. J., Howarth I. D., Irwin M. J., Burnley A. W.,
Harries T. J., 2004b, MNRAS, 353, 601
Evans C. J., Lennon D. J., Smartt S. J., Trundle C., 2006,
A&A, 456, 623
Faccioli L., Alcock C., Cook K., Prochter G. E., Protopapas
P., Syphers D., 2007, AJ, 134, 1963
Fazio G. G. et al., 2004, ApJS, 154, 10
Feast M. W., 1979, MNRAS, 186, 831
Feast M. W., 1986, in Astrophysics and Space Science
Library, Vol. 128, IAU Colloq. 87: Hydrogen Deficient
Stars and Related Objects, Hunger K., Schoenberner D.,
Kameswara Rao N., eds., pp. 151–161
Feast M. W., Thackeray A. D., Wesselink A. J., 1960, MN-
RAS, 121, 337
Fitzpatrick E. L., Savage B. D., 1983, ApJ, 267, 93
Fitzpatrick E. L., Savage B. D., 1985, ApJ, 292, 122
Foellmi C. et al., 2008, Rev. Mex. A. A., 44, 3
Foellmi C., Moffat A. F. J., Guerrero M. A., 2003, MNRAS,
338, 360
Frogel J. A., Persson S. E., Cohen J. G., 1983, ApJS, 53,
713
Garc´ıa-Hern´andez D. A. et al., 2011, ApJ, 737, L30
Garc´ıa-Hern´andez D. A. et al., 2009, ApJ, 705, L31
Garmany C. D., Conti P. S., Massey P., 1987, AJ, 93, 1070
Georgiev L., Koenigsberger G., Hillier D. J., Morrell N.,
Barb´a R., Gamen R., 2011, AJ, 142, 191
Gonz´alez R. F., Koenigsberger G., 2014, A&A, 561, A105
Gordon K. D. et al., 2011, AJ, 142, 102
Gouliermis D. A., Bestenlehner J. M., Brandner W., Hen-
ning T., 2010, A&A, 515, A56
Gouliermis D. A., Dolphin A. E., Brandner W., Henning
T., 2006, ApJS, 166, 549
Groenewegen M. A. T., 2004, A&A, 425, 595
Groenewegen M. A. T., Blommaert J. A. D. L., 1998, A&A,
332, 25
Groenewegen M. A. T., Blommaert J. A. D. L., Cioni M.-
R., Okumura K., Habing H. J., Trams N. R., van Loon
J. T., 2000, Mem. Societa Astronomica Italiana, 71, 639
Groenewegen M. A. T., Sloan G. C., Soszy´nski I., Petersen
E. A., 2009, A&A, 506, 1277
Guerrero M. A., Chu Y.-H., 2008, ApJS, 177, 216
Gvaramadze V. V., Pflamm-Altenburg J., Kroupa P., 2011,
A&A, 525, A17
Henize K. G., 1956, ApJS, 2, 315
Henize K. G., Westerlund B. E., 1963, ApJ, 137, 747
Hennekemper E., Gouliermis D. A., Henning T., Brandner
W., Dolphin A. E., 2008, ApJ, 672, 914
Heydari-Malayeri M., Rauw G., Esslinger O., Beuzit J.-L.,
1997, A&A, 322, 554
Heydari-Malayeri M., Selier R., 2010, A&A, 517, A39
Hill V., 1997, A&A, 324, 435
Hill V., Barbuy B., Spite M., 1997, A&A, 323, 461
Hog E., Kuzmin A., Bastian U., Fabricius C., Kuimov K.,
Lindegren L., Makarov V. V., Roeser S., 1998, A&A, 335,
L65
Hoopes C. G., Sembach K. R., Howk J. C., Blair W. P.,
2001, ApJ, 558, L35
Hoopes C. G., Sembach K. R., Howk J. C., Savage B. D.,
Fullerton A. W., 2002, ApJ, 569, 233
Houck J. R. et al., 2004, ApJS, 154, 18
Hummel W., Szeifert T., G¨assler W., Muschielok B., Seifert
W., Appenzeller I., Rupprecht G., 1999, A&A, 352, L31
Humphreys R. M., 1979, ApJ, 231, 384
Hunter I. et al., 2009, A&A, 496, 841
Hunter I., Lennon D. J., Dufton P. L., Trundle C., Sim´on-
D´ıaz S., Smartt S. J., Ryans R. S. I., Evans C. J., 2008,
A&A, 479, 541
Hutchings J. B., 1966, MNRAS, 131, 299
Ita Y. et al., 2010, PASJ, 62, 273
Jacoby G. H., De Marco O., 2002, AJ, 123, 269
Jones O. C. et al., 2012, MNRAS, 427, 3209
Kamath D., Wood P. R., Soszy´nski I., Lebzelter T., 2010,
MNRAS, 408, 522
Kamath D., Wood P. R., Van Winckel H., 2014, MNRAS,
439, 22112270
Kato D. et al., 2007, PASJ, 59, 615
Keller S. C., Wood P. R., Bessell M. S., 1999, A&AS, 134,
489
Kemper F., Waters L. B. F. M., de Koter A., Tielens
A. G. G. M., 2001, A&A, 369, 132
Kemper F. et al., 2010, PASP, 122, 683
Koenigsberger G., 2004, Rev. Mex. A. A., 40, 107
Koenigsberger G., Auer L. H., Georgiev L., Guinan E.,
1998a, ApJ, 496, 934
Koenigsberger G., Fullerton A. W., Massa D., Auer L. H.,
2006, AJ, 132, 1527
Koenigsberger G., Georgiev L., Barb´a R., Tzvetanov Z.,
Walborn N. R., Niemela V. S., Morrell N., Schulte-
Ladbeck R., 2000, ApJ, 542, 428
Koenigsberger G., Georgiev L., Hillier D. J., Morrell N.,
Barb´a R., Gamen R., 2010, AJ, 139, 2600
Koenigsberger G. et al., 2001, AJ, 121, 267