0807.2998/ms.tex
1: %\documentclass[12pt,preprint]{aastex}
2: %\documentclass[12pt, manuscript]{aastex}
3: \documentclass{emulateapj}
4: 
5: % mspre - based on ms07.tex   25 Jun 08 
6: 
7: \newcommand{\iso}{{\em ISO}}
8: \newcommand{\iras}{{\em IRAS}}
9: \newcommand{\mum}{\ifmmode{\rm \mu m}\else{$\mu$m}\fi}
10: 
11: %\submitted{Submitted to the Astrophysical Journal ... }
12: %\journalinfo{ }
13: 
14: \begin{document}
15: 
16: \title{The Magellanic zoo:  Mid-infrared {\it Spitzer} 
17: spectroscopy of evolved stars and circumstellar dust in 
18: the Magellanic Clouds}
19: 
20: \author{
21: G.~C.~Sloan\altaffilmark{1},
22: K.~E.~Kraemer\altaffilmark{2},
23: P.~R.~Wood\altaffilmark{3},
24: A.~A.~Zijlstra\altaffilmark{4},
25: J.~Bernard-Salas\altaffilmark{1},
26: D.~Devost\altaffilmark{1,5},
27: J.~R.~Houck\altaffilmark{1}
28: }
29: \altaffiltext{1}{Cornell University, Astronomy Department,
30:   Ithaca, NY 14853-6801, sloan@isc.astro.cornell.edu, 
31:   jbs@isc.astro.cornell.edu, jrh13@cornell.edu}
32: \altaffiltext{2}{Air Force Research Laboratory, Space Vehicles
33:    Directorate, 29 Randolph Road, Hanscom AFB, MA 01731}
34: \altaffiltext{3}{Research School of Astronomy and Astrophysics,
35:   Australian National University, Cotter Road, Weston Creek ACT 2611,
36:   Australia, wood@mso.anu.edu.au}
37: \altaffiltext{4}{Univ. of Manchester, School of Physics \& Astronomy,
38:    P.~O.~Box 88, Manchester M60 1QD, albert.zijlstra@manchester.ac.uk}
39: \altaffiltext{5}{Canada France Hawaii Telescope, 65-1238 Mamalahoa 
40:   Hwy, Kamuela, HI, 96743, devost@cfht.hawaii.edu} 
41: 
42: \begin{abstract}
43: 
44: We observed a sample of evolved stars in the Large and Small 
45: Magellanic Clouds (LMC and SMC) with the Infrared Spectrograph 
46: on the {\it Spitzer Space Telescope}.  Comparing samples from
47: the SMC, LMC, and the Galaxy reveals that the dust-production
48: rate depends on metallicity for oxygen-rich stars, but carbon
49: stars with similar pulsation properties produce similar
50: quantities of dust, regardless of their initial metallicity.
51: Other properties of the oxygen-rich stars also depend on
52: metallicity.  As the metallicity decreases, the fraction of 
53: naked (i.e.\ dust-free) stars increases, and among the naked
54: stars, the strength of the 8~\mum\ absorption band from SiO 
55: decreases.  Our sample includes several massive stars 
56: in the LMC with long pulsation periods which produce 
57: significant amounts of dust, probably because they are young 
58: and relatively metal rich.  Little alumina dust is seen in 
59: circumstellar shells in the SMC and LMC, unlike in Galactic 
60: samples.  Three oxygen-rich sources also show emission from 
61: magnesium-rich crystalline silicates.  Many also show an 
62: emission feature at 14~\mum.  The one S star in our sample 
63: shows a newly detected emission feature centered at 
64: 13.5~\mum.  At lower metallicity, carbon stars with similar 
65: amounts of amorphous carbon in their shells have stronger 
66: absorption from molecular acetylene (C$_2$H$_2$) and weaker 
67: emission from SiC and MgS dust, as discovered in previous 
68: studies.  
69: 
70: \end{abstract}
71: 
72: \keywords{circumstellar matter --- infrared:  stars }
73: 
74: 
75: \section{Introduction} % Sec. 1
76: 
77: In the Galaxy, stars on the asymptotic giant branch (AGB)
78: may be responsible for nearly 90\% of the dust injected into
79: the interstellar medium (ISM), with supergiants accounting
80: for most of the remainder \citep{geh89}.  How much these
81: relative contributions change in the early Universe is not
82: well understood.  This question has grown even more 
83: important, now that the Infrared Spectrograph 
84: \citep[IRS][]{hou04} on the {\it Spitzer Space Telescope} 
85: \citep{wer04a} has detected dust in galaxies at redshifts 
86: of 2.6--2.7 \citep{tep07,lut07}, corresponding to look-back 
87: times of $\sim$11 Gyr in a Universe 13.7 Gyr old.  Supernovae 
88: have been suggested as a possible source of dust in the early
89: Universe, \citep[e.g.][]{cmc67,ds80}, but the amount of 
90: dust they contribute remains controversial 
91: \citep[e.g.][]{sug06,mei07}.  The most massive stars that can
92: evolve to the AGB will begin producing dust only $\sim$100 Myr 
93: after the formation of the first stars in a galaxy.
94: This initial population will be metal poor.
95: 
96: Nearby galaxies within the Local Group give us the 
97: opportunity to study how the dust produced by individual
98: supergiants and AGB stars depends on metallicity.
99: The Large and Small Magellanic Clouds (LMC and SMC) lie at 
100: distances of 50 and 61 kpc, respectively 
101: \citep[e.g.][]{alv04,kw06}.  Measurements of the metallicity 
102: ([Fe/H]) of the LMC vary between $-0.3$ and $-0.4$, while the 
103: range in the SMC is between $-0.6$ and $-0.8$ \citep[][and 
104: references therein]{kw06}.
105: 
106: The {\it Infrared Space Observatory} ({\it ISO}) obtained
107: the first mid-infrared spectra of evolved stars in the LMC.
108: \cite{tra99} published spectra of 23 targets using the 
109: PHOT-S spectrometer, which covered the 2--5 and 
110: 6--12~\mum\ wavelength ranges at a resolution $R\sim$90 
111: ($R=\lambda/\Delta\lambda$), and the CAM-CVF 
112: spectrometer, which observed the 7--14~\mum\ wavelength range 
113: at a resolution of $\sim$40.  The formation of CO molecules
114: in stellar outflows will exhaust either C or O from the material
115: available to form dust and divides the dust chemistry around
116: these objects into oxygen-rich or carbon-rich.  The \iso\
117: sample includes spectra from both chemistries, and 
118: \cite{tra99} showed that they could separate these two
119: groups photometrically with color-color diagrams.  
120: 
121: The spectrometers on \iso\ allowed observations of only the 
122: brightest targets.  The improved sensitivity of the IRS on 
123: {\it Spitzer} has made possible more detailed studies of 
124: mass-loss and dust injection in both the LMC and SMC.  The 
125: five major programs have observed a combined total of 176 
126: sources in the LMC and 63 in the SMC, most of them evolved 
127: stars.  \cite{slo08a} reviewed these programs and the 
128: different criteria used to select targets.  
129: 
130: Program 3426 (P.I.\ Kastner) focused on the brightest 
131: sources in the LMC, using the classes defined from 
132: observations of the {\it Mid-Course Space Experiment} ({\it 
133: MSX}) by \cite{msx01} \citep{kas06,buc06}.  Program 3591 
134: (P.I.\ Kemper) targeted objects covering the stages of 
135: evolution of oxygen-rich stars from the AGB to planetary 
136: nebulae \citep{lei08,mk08}.  Program 3277 (P.I.\ Egan) 
137: selected targets in the SMC based on their infrared colors 
138: \citep{kra05,kra06,slo06b}.  Program 3505 (P.I.\ Wood) 
139: spanned a range of mass-loss rates in the LMC and the SMC.  
140: It included a young stellar object \citep[YSO;][]{vl05b}, 
141: but nearly all of the rest are carbon stars 
142: \citep{zij06,mat06,gro07,lag07}.  
143: The other three programs also observed more carbon stars than 
144: initially planned, with sources classified photometrically as 
145: OH/IR stars turning out to be carbon stars instead.  
146: 
147: This paper focuses on Program 200, (P.I.\ Sloan; hereafter
148: referred to as the MC\_DUST program).  The next section 
149: describes the sample and the ground-based support
150: observations.  In \S 3, we describe the IRS observations,
151: the data reduction, and the classification of the spectra.  
152: The MC\_DUST sample contains a rich variety of object types, 
153: each of which requires a different method of analysis.  We 
154: examine each of these types in turn in the following 
155: sections:  \S 4---stars with no associated dust (naked stars), 
156: \S 5---dusty oxygen-rich stars, \S 6---carbon stars, 
157: \S 7---red sources with emission from polycyclic aromatic 
158: hydrocarbons (PAHs).  Several targets show characteristics 
159: which place them in multiple categories, and as a consequence, 
160: they will be analyzed in more than one of the following 
161: sections.  In each of these sections, we examine individual 
162: sources and unusual spectral features, but we save a more 
163: general comparison of the various populations for \S 8.  
164: Finally, we draw conclusions in \S 9.
165: 
166: 
167: \section{The Sample} % Sec. 2
168: 
169: \subsection{Sample Selection} % Sec. 2.1
170: 
171: \begin{deluxetable}{lrrrr} % Table 1
172: \tablenum{1}
173: \tablecolumns{5}
174: \tablewidth{0pt}
175: \tablecaption{The MC\_DUST program\label{Tbl2}}
176: \tablehead{
177:   \colhead{Object} & \multicolumn{2}{c}{LMC} & \multicolumn{2}{c}{SMC} \\
178:   \colhead{Class} & \colhead{Planned} & \colhead{Observed} & 
179:                     \colhead{Planned} & \colhead{Observed} 
180: }
181: \startdata
182: AGB/M         &       7  &       7  &       4  &       4 \\
183: AGB/MS        &       3  &       3  &       4  &       4 \\
184: AGB/S         &       3  &       2  &       1  &       1 \\
185: AGB/C         &       6  &       4  &       4  &       1 \\
186: OH/IR star    &       3  &       3  & \nodata  & \nodata \\
187: C/IR star     &       3  &       2  & \nodata  & \nodata \\
188: RSG           &       3  &       4  & \nodata  &       1 \\
189: YSO           & \nodata  &       1  & \nodata  & \nodata \\
190: unknown       & \nodata  &       1  & \nodata  & \nodata \\
191: total         &      28  &      27  &      13  &      11 \\
192: \enddata
193: \end{deluxetable}
194: 
195: \begin{deluxetable*}{lllrrrlrr} % Table 2
196: %\begin{deluxetable}{lllrrrlrr} 
197: \tablenum{2}
198: \tablecolumns{9}
199: \tablewidth{0pt}
200: \small
201: \tablecaption{The MC\_Dust sample\label{Tbl3}}
202: \tablehead{
203:   \colhead{ } & \colhead{RA} & \colhead{Dec.} &
204:   \multicolumn{3}{c}{2MASS} & 
205:   \colhead{Source} & \colhead{Period} & \colhead{ } \\
206:   \colhead{Target} & \multicolumn{2}{c}{J2000.0} & \colhead{J} & 
207:   \colhead{H} & \colhead{K$_s$} & \colhead{Type} & \colhead{(days)} & 
208:   \colhead{Ref.\tablenotemark{a}}
209: }
210: \startdata
211: 
212: HV 11223          & 00 32 01.61 & $-$73 22 34.7 & 11.166 & 10.276 &  9.971 & AGB/MS  &    407: & 1\\
213: HV 1366           & 00 42 49.87 & $-$72 55 11.4 & 12.232 & 11.431 & 11.154 & AGB/M   &    305  & 5\\
214: BFM 1             & 00 47 19.24 & $-$72 40 04.5 & 12.661 & 11.647 & 11.027 & AGB/S   &    398  & 5\\
215: CV 78             & 00 49 03.94 & $-$73 05 19.9 & 11.649 & 10.613 & 10.129 & AGB/C   &    423  & 6\\
216: HV 11303          & 00 52 08.84 & $-$71 36 24.0 & 10.836 &  9.983 &  9.540 & AGB/MS  &    534  & 1\\
217: HV 11329          & 00 53 39.40 & $-$72 52 39.2 & 10.911 &  9.987 &  9.647 & AGB/M   &    377  & 5\\
218: HV 838            & 00 55 38.22 & $-$73 11 41.1 & 10.608 &  9.860 &  9.440 & AGB/M   &    622  & 5\\
219: HV 11366          & 00 56 54.77 & $-$72 14 08.6 & 11.066 & 10.154 &  9.870 & AGB/MS  &    366  & 1\\
220: HV 12149          & 00 58 50.17 & $-$72 18 35.6 &  9.963 &  9.022 &  8.609 & AGB/M   &    745  & 5\\
221: HV 1963           & 01 04 26.63 & $-$72 34 40.3 & 10.638 &  9.677 &  9.409 & AGB/MS  &    249  & 6\tablenotemark{b}\\
222: Massey SMC 59803  & 01 04 38.21 & $-$72 01 27.0 &  9.093 &  8.301 &  8.097 & RSG     & \nodata & \nodata\\
223: \\
224: IRAS 04509$-$6922 & 04 50 40.47 & $-$69 17 31.9 &  9.874 &  8.670 &  7.928 & (AGB/M) &   1290  & 3\\
225: IRAS 04516$-$6902 & 04 51 29.00 & $-$68 57 50.1 &  9.927 &  8.619 &  7.914 & (AGB/M) &   1170  & 5\\
226: IRAS 04530$-$6916 & 04 52 45.67 & $-$69 11 49.5 & 13.942 & 11.859 &  9.959 & (RSG)   &   1260: & 3\\
227: IRAS 04545$-$7000 & 04 54 10.06 & $-$69 55 58.3 & 16.510 & 12.815 & 10.402 & OH/IR   &   1270  & 3\\
228: HV 888            & 05 04 14.14 & $-$67 16 14.4 &  8.010 &  7.188 &  6.781 & RSG     &    850  & 1\\
229: HV 2310           & 05 06 27.68 & $-$68 12 03.7 & 10.455 &  9.635 &  9.111 & AGB/M   &    600  & 5\\
230: IRAS 05128$-$6455 & 05 13 04.57 & $-$64 51 40.3 & 14.553 & 12.832 & 11.281 & (C/IR)  &    708  & 4\\
231: MACHO 79.5505.26  & 05 14 29.83 & $-$68 54 33.8 & 12.302 & 11.363 & 10.923 & AGB/C   &     97  & 5\tablenotemark{b}\\
232: HV 5810           & 05 24 07.02 & $-$69 23 36.9 & 10.813 &  9.947 &  9.642 & AGB/M   &    374  & 5\\
233: WBP 17            & 05 26 19.88 & $-$69 41 37.3 & 12.342 & 11.315 & 10.796 & AGB/C   &    304  & 5\\
234: WBP 29            & 05 26 40.95 & $-$69 23 11.4 & 12.270 & 11.382 & 10.915 & AGB/C   &    247  & 5\tablenotemark{b}\\
235: WBP 42            & 05 27 04.73 & $-$69 38 16.3 & 13.800 & 12.192 & 11.009 & AGB/C   &    404  & 5\\
236: WOH G 339         & 05 27 10.24 & $-$69 36 26.7 & 10.399 &  9.544 &  9.159 & AGB/M   &    538  & 5\\
237: WBP 77            & 05 28 06.75 & $-$69 32 27.9 & 12.313 & 11.482 & 11.197 & AGB/S   &    212  & 5\\
238: WBP 104           & 05 28 26.73 & $-$69 14 43.5 & 12.364 & 11.422 & 11.047 & AGB/S   &    245  & 2\\
239: 2MASS J052832     & 05 28 32.10 & $-$69 29 08.5 & 11.429 & 10.571 & 10.326 & unknown &    110: & 5\\
240: HV 2572           & 05 28 36.71 & $-$69 20 04.1 &  9.795 &  8.836 &  8.447 & AGB/M   &    591: & 6\\
241: HV 2575           & 05 29 59.98 & $-$67 45 01.4 & 11.482 & 10.556 & 10.066 & AGB/MS  &    390  & 5\\
242: HV 2578           & 05 29 03.15 & $-$69 48 06.9 &  9.947 &  9.104 &  8.737 & AGB/M   &    653  & 5\tablenotemark{b}\\
243: IRAS 05300$-$6651 & 05 30 03.89 & $-$66 49 24.2 & 16.895 & 14.800 & 12.247 & C/IR    & \nodata & \nodata\\
244: HV 996            & 05 32 35.62 & $-$67 55 09.0 &  8.991 &  8.156 &  7.643 & RSG     &    760  & 1\\
245: IRAS 05329$-$6708 & 05 32 51.33 & $-$67 06 52.0 & 17.285 & 13.638 & 11.233 & OH/IR   &   1260  & 3\\
246: HV 12620          & 05 32 59.92 & $-$70 41 23.6 & 11.206 & 10.357 & 10.052 & AGB/MS  &    326  & 5\\
247: IRAS 05348$-$7024 & 05 34 15.99 & $-$70 22 52.5 & 17.659 & 15.488 & 12.851 & C/IR    & \nodata & \nodata\\
248: IRAS 05402$-$6956 & 05 39 44.89 & $-$69 55 18.1 & 14.344 & 11.740 &  9.789 & OH/IR   &   1390  & 3\\
249: HV 12667          & 05 49 13.36 & $-$70 42 40.7 & 10.823 & 10.019 &  9.445 & AGB/M   &    645  & 5\\
250: HV 12070          & 05 52 27.86 & $-$69 14 10.0 & 10.605 &  9.754 &  9.204 & AGB/MS  &    599  & 5\\
251: %\\
252: %(NGC 371 LE 31)   & 01 01 45.70 & $-$72 07 57.0 & 13.036 & 11.819 & 11.082 & AGB/C  &   (290)  & \\
253: %(NGC 371 LE 8)    & 01 05 57.30 & $-$72 00 48.0 & 12.732 & 11.672 & 11.097 & AGB/C  &   (330)  & \\
254: %(WBP 219)         & 05 26 40.40 & $-$69 23 11.0 & 12.563 & 11.452 & 10.676 & AGB/C  &   (294)  & \\
255: \enddata
256: \tablenotetext{a}{References for periods:  (1) \cite{wbf83}; (2) \cite{wbp85};
257:   (3) \cite{woo92}; (4) \cite{whi03}; (5) from MACHO data \citep{macho}; 
258:   (6) from OGLE data \citep{ogle97,ogle05}.}
259: \tablenotetext{b}{The period has changed significantly from previously 
260:   published values; see \S 2.3.}
261: \end{deluxetable*}
262: %\end{deluxetable}
263: 
264: The MC\_DUST program was a small (17.2 hour) program 
265: conducted as part of the IRS Guaranteed Time Observations 
266: (GTO).  It sampled sources in the LMC and SMC covering a 
267: range of classifications determined in both the optical and 
268: the infrared.  The targets were divided into seven classes:  
269: red supergiants (RSGs), four categories of optically 
270: identified AGB sources (M giants, MS stars, S stars, and C 
271: stars), and two categories of infrared sources: OH/IR stars 
272: and infrared carbon-rich sources (C/IR stars).  Table 1 
273: breaks down the observing program by object class, as 
274: planned and as actually executed.  The plan included 28 
275: sources in the LMC (20 oxygen-rich and eight carbon-rich), 
276: and 13 in the SMC (nine oxygen-rich and four carbon-rich).  
277: As with other programs in the Magellanic Clouds, the crowded 
278: regions resulted in several peak-up failures.  In our case, 
279: we lost three objects in the LMC and three in the SMC, but 
280: gained three serendipitious observations, two in the LMC and 
281: one in the SMC.
282: 
283: % Table 1 goes here.
284: 
285: The other spectroscopic samples in the Magellanic Clouds 
286: were selected primarily from near- and mid-infrared 
287: photometry, but most of the sources in the MC\_DUST sample 
288: were selected based on optical spectral classifications, 
289: supplemented with longer wavelength data.  While many of the 
290: object types in other programs turned out to be different 
291: than planned, our sources, for the most part, had the 
292: photospheric chemistries we had expected.  Another advantage 
293: was that we were able to observe several sources below the 
294: photometric limits of previous infrared surveys, allowing us 
295: to probe fainter sources on the AGB.
296: 
297: % Table 2 goes here.
298: 
299: Table 2 lists the individual targets in the MC\_DUST sample.
300: The targets with Harvard Variable numbers and the carbon star 
301: CV 78\footnote{CV = Cordoba Variable; \cite{wbf83} cite 
302: \cite{des59}.} were selected from the study by \cite{wbf83}.  
303: These targets account for most of the sources in the RSG, 
304: AGB/M and AGB/MS classes.  The AGB/S and AGB/C stars with 
305: WBP numbers were selected from the observations of 
306: \cite{wbp85}.  \cite{wbf83} identified MACHO 79.5505.26 as 
307: HV 5680, but HV 5680 is actually 16\arcsec\ to the northeast 
308: and a magnitude fainter at K.  The target we have named 
309: BFM 1 was the first S star identified in the SMC \citep{bfm81} 
310: and, as discussed below, is the only true S star in our 
311: sample.\footnote{Table 2 corrects the erroneous coordinates
312: for BFM 1 given by \cite{bfm81}.}  We also 
313: selected six oxygen-rich infrared sources first studied by 
314: \cite{woo92}.  They identified three as OH/IR stars.  The 
315: other three include a suspected supergiant (IRAS 04530) and 
316: two luminous and embedded evolved stars (IRAS 04509 and 
317: IRAS 04516).  Finally, we selected three sources believed 
318: to be infrared carbon stars (C/IR) from a sample studied by 
319: \cite{vl99}.  The types in parentheses in Table 2 have
320: changed, as discussed in \S 3.3 and 3.4 below.
321: 
322: Peak-up failures affected our sample of carbon stars most
323: significantly, reducing our LMC sample from eight to six
324: carbon stars and our SMC sample from four to one.  
325: Fortunately, other programs produced a surplus of carbon 
326: stars, which has resulted in a series of papers on carbon 
327: stars in the SMC \citep{slo06b,lag07} and the LMC 
328: \citep{zij06,lei08}.  Three of the peak-up failures resulted 
329: in serendipitious observations.  Instead of observing NGC 371 
330: LE 28 in the SMC, we observed Massey SMC 59803, a supergiant 
331: with a spectral class of K2--3 I \citep{lev06}.  In the LMC, 
332: we observed the giant WOH G 339 instead of WBP 51, and 
333: instead of WBP 116, we observed a source identified only in 
334: the 2MASS catalog (2MASS J05283210-6929084, or J052832 for 
335: short).  Little is known about this last source; it may even 
336: be a foreground object and not in the SMC.
337: 
338: \subsection{Ground-based Spectroscopy} % Sec. 2.2
339: 
340: \begin{figure*} % Fig. 1
341: \includegraphics[width=6.0in]{f01.eps}
342: \caption{Optical spectra of our sample of evolved stars in
343: the Magellanic Clouds.  The panels for HV 5810, BFM 1, and
344: WBP 17 are larger than the rest to illustrate typical 
345: molecular bands in M giants, S stars, and carbon stars, 
346: respectively.  Spectral classes in parentheses are 
347: inconsistent with the observed spectra.  We observed only 
348: two of the nine IRAS sources due to the heavy optical 
349: obscuration.}
350: \end{figure*}
351: 
352: To verify the overall chemistry of our targets, we obtained
353: optical spectra using the Double Beam Spectrograph on the
354: 2.3 m telescope of the Australian National University at 
355: Siding Spring Observatory (SSO) of all of our targets, except
356: for seven of the nine IRAS sources, due to their heavy optical 
357: obscuration.  These spectra covered the 0.45--1.05~\mum\ 
358: wavelength range with a resolution of 10\AA.  They were 
359: reduced using standard IRAF procedures with HR 718 as a flux 
360: standard and the weak-lined giant HD 26169 to remove 
361: telluric features.  Figure 1 presents the optical spectra.
362: We have expanded the panels for HV 5810, BFM 1, and WBP 17 to
363: use them as prototypes for M giants, S stars, and carbon
364: stars, respectively, and make room for labels for the visible
365: molecular bands.
366: 
367: % Figure 1 goes here.
368: 
369: Cyanogen (CN) bands dominate the spectra of the carbon stars
370: and make them easy to distinguish, with its characteristic
371: corrugations.  The bands which appear in this wavelength 
372: range arise from the A $^2\Pi$ $-$ X $^2\Sigma^+$ electronic 
373: transition \citep{hh79}.  The band systems are labelled with 
374: the change in vibrational state, and in each system the 
375: vibrational states involved increase to the red.  The 
376: spectrum of CV 78 includes H$\alpha$ emission at 0.656~\mum.
377: 
378: TiO band systems dominate the spectra of the M giants.  The
379: identifications of the band systems presented in Figure 1 are 
380: from \cite{hh79} and the models of M dwarfs by \cite{val98}.  
381: The spectra in the sample show a wide range of band strengths.
382: The spectra which are most obscured by dust show only the
383: reddest bands.  Some of the spectra show H$\alpha$ emission.
384: 
385: Figure 1 shows that the spectrum of one of our planned C/IR 
386: stars, IRAS 05128, is very similar to IRAS 04516 and clearly 
387: oxygen-rich.  \cite{tra99} classified it as a carbon star 
388: based on its \iso/PHOT-S spectrum, but subsequently, 
389: \cite{mat05} found strong H$_2$O absorption and possible weak 
390: SiO absorption in its 2.9--4.1~\mum\ spectrum.  Our optical 
391: spectrum confirms their conclusion that IRAS 05128 is 
392: oxygen-rich.  \cite{vl08} classify IRAS 05128 
393: as an M9 giant. 
394: 
395: \cite{bfm81} identified BFM 1 as an S star based on a strong
396: LaO band at 0.79~\mum.  Our spectrum clearly shows both LaO
397: and ZrO bands, but these are superimposed on a CN spectrum
398: (compare BFM 1 to CV 78 just below it).  The strong 
399: absorption band at 0.93~\mum\ is due to ZrO \citep[][and
400: references therein]{sw69} and arises from the b' $^3\Pi$ $-$
401: a $^3\Delta$ electronic transition.  It is close to the head 
402: of the CN A$-$X $\Delta$v=1 band system, but displaced 
403: $\sim$0.2~\mum\ to the red, making it easy to identify.
404: The head of the LaO A $^2\Pi_{3/2}$ $-$ X $^2\Sigma^+$
405: band system at 0.79~\mum\ \citep{wc78} coincides
406: with the head of the CN A$-$X $\Delta$v=2 band system, but 
407: the LaO stands out in two ways.  First, the LaO band makes
408: the head of the CN band too deep, and second, the A 
409: $^2\Pi_{1/2}$ $-$ X $^2\Sigma^+$ band system appears at 
410: 0.74~\mum.  Similarly, the ZrO B $^1\Pi$ $-$ X 
411: $^1\Sigma$ band system strengthens the apparent head of
412: the CN A$-$X $\Delta$v=3 band system at 0.69~\mum, but 
413: the ZrO $\gamma$ band system from 0.61 to 0.66~\mum\
414: makes the identification unambiguous.  The $\gamma$
415: system arises from the A $^3\Phi$ $-$ X $^3\Delta$ 
416: transition \citep{pd79}.
417: 
418: BFM 1 is the sole true S star in our sample.  WBP 77 shows
419: a hint of LaO absorption at 0.79~\mum\ but is otherwise
420: more consistent with an MS spectrum.  Both it and WBP 104 
421: show the 0.65 and 0.93~\mum\ ZrO bands seen in many MS 
422: spectra.  The spectrum of HV 12070, which is classified as an 
423: MS star, looks more like the spectrum of an M giant.  We have
424: noted the uncertainties in the classifications by putting
425: them in parentheses in Figure 1, but we have not changed them
426: for two reasons.  First, these sources are variable stars
427: and the spectra may have different characteristics at 
428: different phases.  Second, we do not distinguish between M
429: and MS stars in our analysis below; we consider them as one
430: group of oxygen-rich sources in our analysis.  However, the
431: fact that BFM 1 is our sole strong S star is significant.
432: It is the only source where C/O$\approx$1, which may
433: explain its unusual dust properties (see \S 5.6).
434: 
435: \subsection{Optical Light Curves} % Sec. 2.3
436: 
437: Many of the stars in our sample have light curves in the 
438: databases from the MACHO Project \citep[Massive Compact Halo 
439: Objects;][]{macho} and/or the Optical Gravitational Lensing 
440: Experiment \citep[OGLE;][]{ogle97,ogle05}.  Table 2 includes 
441: the periods we have determined from these data.  Most of these 
442: stars pulsate regularly, or semi-regularly, with periods 
443: covering a wide range, from $\sim$100 days to $\sim$1400.  
444: The red and blue lightcurves sometimes give slightly 
445: different periods.  We have found that the blue light curves 
446: generally produce more self-consistent periods.
447: 
448: Some of the periods have changed significantly from 
449: previously published results, primarily because some of the
450: older periods were based on data which did not sample the
451: light curves adequately.  Several periods measured by 
452: \cite{wbf83} have changed.  HV 1963 had a period of 330: 
453: days, but we measure a period of 249 days with the OGLE data.  
454: The MACHO data for MACHO 79.5505.26 give a period of 97 days,
455: compared to 151.  This source has a longer cycle of 1200 
456: days, so the shorter period must be an overtone mode.  The 
457: period of HV 2578 has changed from 470 to 247 days, based on 
458: the MACHO data.  HV 2572 previously had a period of 201 days.  
459: The OGLE data are poorly characterized, but a period of 591 
460: days fits the data better than any other.  \cite{wbp85}
461: measured the period of WBP 29 to be 360 days, but we find that
462: the MACHO data are more consistent with a 247-day period.
463: Several other periods have changed by amounts less than 10\%.
464: 
465: \subsection{Near-infrared Photometry} % Sec. 2.4
466: 
467: \begin{deluxetable*}{lrrrrr} % Table 3
468: %\begin{deluxetable}{lrrrrr}
469: \tablenum{3}
470: \tablecolumns{6}
471: \tablewidth{0pt}
472: \small
473: \tablecaption{Near-infrared SSO photometry\label{Tbl4}}
474: \tablehead{
475:   \colhead{ } & \multicolumn{4}{c}{Magnitude} & \colhead{Julian} \\
476:   \colhead{Target} & \colhead{J} & \colhead{H} & \colhead{K} &
477:   \colhead{L} & \colhead{Date}
478: }
479: \startdata
480: HV 11223          & 11.197$\pm$0.017 & 10.266$\pm$0.023 & 10.077$\pm$0.013 &  9.673$\pm$0.081 & 2453526\\
481: HV 1366           & 12.258$\pm$0.019 & 11.534$\pm$0.012 & 11.305$\pm$0.008 & 10.451$\pm$0.196 & 2453334\\
482: BFM 1             & 12.377$\pm$0.034 & 11.489$\pm$0.018 & 11.082$\pm$0.019 & 10.103$\pm$0.164 & 2453524\\
483: BFM 1             & 12.088$\pm$0.013 & 11.268$\pm$0.014 & 10.830$\pm$0.014 & 10.186$\pm$0.157 & 2453942\\
484: CV 78             & 11.099$\pm$0.036 & 10.168$\pm$0.033 &  9.880$\pm$0.012 &  9.223$\pm$0.058 & 2453527\\
485: HV 11303          & 10.770$\pm$0.022 &  9.846$\pm$0.010 &  9.626$\pm$0.012 &  9.298$\pm$0.067 & 2453530\\
486: HV 11329          & 11.068$\pm$0.013 & 10.007$\pm$0.013 &  9.800$\pm$0.012 &  9.454$\pm$0.074 & 2453527\\
487: HV 11329          & 10.991$\pm$0.012 & 10.000$\pm$0.007 &  9.781$\pm$0.018 &  9.752$\pm$0.113 & 2453915\\
488: HV 838            &  9.824$\pm$0.006 &  9.008$\pm$0.003 &  8.824$\pm$0.004 &  8.607$\pm$0.045 & 2453227\\
489: HV 11366          & 11.135$\pm$0.016 & 10.115$\pm$0.015 &  9.941$\pm$0.012 &  9.673$\pm$0.085 & 2453530\\
490: HV 12149          & 10.701$\pm$0.015 &  9.758$\pm$0.009 &  9.353$\pm$0.014 &  8.643$\pm$0.045 & 2453530\\
491: HV 1963           & 10.606$\pm$0.006 &  9.630$\pm$0.005 &  9.388$\pm$0.003 &  9.233$\pm$0.011 & 2453227\\
492: HV 1963           & 10.711$\pm$0.011 &  9.735$\pm$0.007 &  9.509$\pm$0.017 &  9.368$\pm$0.079 & 2453915\\
493: Massey SMC 59803  &  9.056$\pm$0.015 &  8.254$\pm$0.011 &  8.141$\pm$0.012 &  7.952$\pm$0.018 & 2453530\\
494: %MSX SMC 018       & 12.955$\pm$0.029 & 11.375$\pm$0.010 & 10.318$\pm$0.018 &  8.943$\pm$0.057 & 2453918\\
495: \\
496: IRAS 04509$-$6922 &  9.959$\pm$0.006 &  8.724$\pm$0.011 &  8.098$\pm$0.015 &  7.071$\pm$0.015 & 2453517\\
497: IRAS 04516$-$6902 & 10.229$\pm$0.008 &  8.943$\pm$0.010 &  8.334$\pm$0.015 &  7.280$\pm$0.017 & 2453517\\
498: IRAS 04530$-$6916 & 13.869$\pm$0.020 &  \nodata         &  9.768$\pm$0.033 &  7.678$\pm$0.025 & 2453135\\
499: IRAS 04545$-$7000 & 15.900$\pm$0.320 &  \nodata         & 10.119$\pm$0.006 &  8.083$\pm$0.053 & 2453135\\
500: IRAS 04545$-$7000 & 14.472$\pm$0.026 & 11.422$\pm$0.007 &  9.397$\pm$0.005 &  7.221$\pm$0.016 & 2453918\\
501: HV 888            &  7.885$\pm$0.008 &  7.014$\pm$0.009 &  6.788$\pm$0.026 &  6.349$\pm$0.017 & 2453532\\
502: HV 2310           & 10.756$\pm$0.004 &  \nodata         &  9.533$\pm$0.005 &  8.969$\pm$0.080 & 2453135\\
503: IRAS 05128$-$6455 & 13.316$\pm$0.010 & 11.451$\pm$0.010 & 10.052$\pm$0.014 &  8.186$\pm$0.032 & 2453531\\
504: MACHO 79.5505.26  & 12.419$\pm$0.026 &  \nodata         & 10.994$\pm$0.009 &  \nodata         & 2453135\\
505: HV 5810           & 11.289$\pm$0.009 & 10.307$\pm$0.012 & 10.000$\pm$0.016 &  9.534$\pm$0.135 & 2453530\\
506: WBP 17            & 12.454$\pm$0.007 & 11.322$\pm$0.011 & 10.792$\pm$0.015 & 10.322$\pm$0.186 & 2453516\\
507: WBP 29            & 12.555$\pm$0.010 & 11.411$\pm$0.011 & 10.806$\pm$0.020 &  \nodata         & 2453516\\
508: WBP 42            & 12.797$\pm$0.026 & 11.283$\pm$0.012 & 10.311$\pm$0.015 &  9.168$\pm$0.062 & 2453517\\
509: WBP 77            & 12.562$\pm$0.031 &  \nodata         & 11.359$\pm$0.004 &  \nodata         & 2453135\\
510: WBP 104           & 12.387$\pm$0.015 & 11.398$\pm$0.013 & 11.115$\pm$0.015 & 10.641$\pm$0.189 & 2453532\\
511: HV 2572           & 10.079$\pm$0.008 &  \nodata         &  8.854$\pm$0.006 &  8.418$\pm$0.039 & 2453135\\
512: HV 2575           & 11.446$\pm$0.007 & 10.433$\pm$0.010 & 10.091$\pm$0.015 &  9.501$\pm$0.089 & 2453530\\
513: HV 2575           & 11.391$\pm$0.006 & 10.438$\pm$0.004 & 10.117$\pm$0.004 &  9.649$\pm$0.072 & 2453919\\
514: HV 2578           & 10.085$\pm$0.012 &  9.146$\pm$0.013 &  8.790$\pm$0.015 &  8.199$\pm$0.039 & 2453527\\
515: IRAS 05300$-$6651 &  \nodata         & 14.972$\pm$0.107 & 12.532$\pm$0.044 &  9.437$\pm$0.086 & 2453526\\
516: HV 996            &  8.799$\pm$0.007 &  7.910$\pm$0.016 &  7.468$\pm$0.014 &  6.665$\pm$0.014 & 2453526\\
517: IRAS 05329$-$6708 & 15.096$\pm$0.170 &  \nodata         &  9.794$\pm$0.008 &  7.445$\pm$0.034 & 2453135\\
518: IRAS 05329$-$6708 & 15.618$\pm$0.074 & 12.557$\pm$0.015 & 10.504$\pm$0.014 &  8.159$\pm$0.023 & 2453915\\
519: HV 12620          & 11.197$\pm$0.007 & 10.302$\pm$0.011 & 10.104$\pm$0.019 &  9.472$\pm$0.085 & 2453517\\
520: IRAS 05348$-$7024 & \nodata          &  \nodata         & 13.579$\pm$0.200 &  9.226$\pm$0.080 & 2453135\\
521: IRAS 05402$-$6956 & 13.632$\pm$0.013 & 11.054$\pm$0.011 &  9.289$\pm$0.015 &  7.170$\pm$0.017 & 2453517\\
522: HV 12667          & 10.793$\pm$0.006 &  9.853$\pm$0.020 &  9.354$\pm$0.020 &  8.527$\pm$0.037 & 2453521\\
523: HV 12070          & 10.421$\pm$0.003 &  \nodata         &  9.157$\pm$0.004 &  8.596$\pm$0.041 & 2453135\\
524: \enddata
525: \end{deluxetable*}
526: %\end{deluxetable}
527: 
528: We obtained ground-based near-infrared photometry for our 
529: sources as close in time to the IRS observations as possible.
530: We used CASPIR \citep{caspir} on the 2.3-m telescope at SSO
531: to obtain images at J (effective wavelength 1.24~\mum), H 
532: (1.68~\mum), K (2.22~\mum), and narrow-band L (3.59~\mum).
533: \cite{slo06b} describe the calibration methods.  Table 3
534: presents the photometry.  It includes two entries
535: for four sources because we observed them twice with the 
536: IRS and twice from the ground.  The photometric data for
537: WBP 17 are averaged from measurements made before and after
538: the IRS spectroscopy.  
539: 
540: % Table 3 goes here.
541: 
542: 
543: \section{IRS Spectroscopy} % Sec. 3
544: 
545: \subsection{IRS Observations} % Sec. 3.1
546: 
547: \begin{deluxetable*}{lrrrrlrll} % Table 4
548: %\begin{deluxetable}{lrrrrlrll}
549: \tablenum{4}
550: \tablecolumns{9}
551: \tablewidth{0pt}
552: \small
553: \tablecaption{Low-resolution observations\label{Tbl5}}
554: \tablehead{
555:   \colhead{ } & \colhead{ } & \multicolumn{2}{c}{Time (s)} & 
556:   \colhead{IRS} & \colhead{Observed} & \colhead{Julian} & \colhead{Infrared} &
557:   \colhead{ } \\
558:   \colhead{Target} & \colhead{AOR key} & \colhead{SL} & \colhead{LL} & 
559:   \colhead{Camp.} & \colhead{(UT)} & \colhead{Date} & \colhead{Sp. Class} &
560:   \colhead{M$_{bol}$\tablenotemark{a}}
561: }
562: \startdata
563: 
564: HV 11223             &  6019584 &  112 &  480 & 21 & 2005 Jun 04 & 2453526 & 1.N       & $-$6.13\\
565: HV 1366              &  6017024 &  240 &  480 & 14 & 2004 Oct 25 & 2453304 & 1.N       & $-$4.97\\
566: BFM 1                &  6022912 &  240 &  480 & 21 & 2005 Jun 02 & 2453524 & 2.ST:     & $-$5.03\\
567: BFM 1 (follow-up)    & 17398784 & 1920 & 2880 & 33 & 2006 Jul 26 & 2453942 & 2.NO      & $-$5.23\\
568: CV 78                &  6019072 &  112 &  480 & 21 & 2005 Jun 05 & 2453527 & 2.CE      & $-$6.22\\
569: HV 11303             &  6017536 &  112 &  480 & 21 & 2005 Jun 08 & 2453530 & 1.N:      & $-$6.05\\
570: HV 11329             &  6017280 &  112 &  480 & 21 & 2005 Jun 05 & 2453527 & 1.NO      & $-$6.25\\
571: HV 11329 (follow-up) & 17399296 & 1440 & 2880 & 32 & 2006 Jun 28 & 2453915 & 1.NO      & $-$6.27\\
572: HV 838               &  6017792 &   56 &  480 & 11 & 2004 Aug 11 & 2453228 & 1.N:      & $-$7.37\tablenotemark{b}\\
573: HV 11366             &  6018304 &  112 &  480 & 21 & 2005 Jun 08 & 2453530 & 1.N:O::   & $-$6.22\\
574: HV 12149             &  6019328 &   56 &  480 & 21 & 2005 Jun 08 & 2453530 & 2.SE8     & $-$6.69\tablenotemark{c}\\
575: HV 1963              &  6018048 &  112 &  480 & 10 & 2004 Jul 18 & 2453205 & 1.N:O:    & $-$6.66\\
576: HV 1963 (follow-up)  & 17399040 &  960 & 1920 & 32 & 2006 Jun 28 & 2453915 & 1.N:O     & $-$6.56\\
577: Massey SMC 59803     &  6019840 &  240 &  720 & 21 & 2005 Jun 08 & 2453530 & 2.SE4u    & $-$8.11\\
578: \\
579: IRAS 04509$-$6922    &  6022400 &   56 &  120 & 21 & 2005 May 26 & 2453517 & 2.SE8f    & $-$7.66\\
580: IRAS 04516$-$6902    &  6020096 &   56 &  120 & 21 & 2005 May 26 & 2453517 & 2.SE6f    & $-$7.43\\
581: IRAS 04530$-$6916    &  6023936 &   56 &   56 &  6 & 2004 Apr 18 & 2453114 & 5.U/SA    & $-$7.78\\
582: IRAS 04545$-$7000    &  6020352 &   56 &  120 &  6 & 2004 Apr 18 & 2453114 & 3.SAxf    & $-$6.42\\
583: HV 888               &  6015488 &   56 &  120 & 22 & 2005 Jul 10 & 2453562 & 2.SE8tf   & $-$9.01\\
584: HV 2310              &  6014976 &  112 &  480 &  6 & 2004 Apr 18 & 2453114 & 2.SE5xf   & $-$6.13\\
585: IRAS 05128$-$6455    &  6024192 &   56 &  240 & 23 & 2005 Aug 09 & 2453592 & 2.SE5f    & $-$5.99\\
586: MACHO 79.5505.26     &  6016768 &  240 &  480 &  6 & 2004 Apr 16 & 2453112 & 1.N       & $-$4.50\\
587: HV 5810              &  6014464 &  112 &  480 & 21 & 2005 Jun 08 & 2453530 & 2.U:      & $-$5.64\\
588: WBP 17               &  6023425 &  240 &  720 & 20 & 2005 Aug 14 & 2453597 & 1.NC      & $-$4.63\\
589: WBP 29               &  6021632 &  240 &  720 & 20 & 2005 Apr 22 & 2453483 & 2.CE:     & $-$4.54\\
590: WBP 42               &  6023680 &  112 &  480 & 21 & 2005 May 26 & 2453517 & 2.CE      & $-$4.97\\
591: WOH G 339            &  6022144 &  112 &  480 & 21 & 2005 May 26 & 2453517 & 2.SE5     & $-$6.52\tablenotemark{d}\\
592: WBP 77               &  6021120 &  240 &  480 &  6 & 2004 Apr 16 & 2453112 & 2.SE1:    & $-$4.26\\
593: WBP 104              &  6021376 &  240 &  720 & 22 & 2005 Jul 10 & 2453562 & 1.NO::    & $-$4.53\\
594: 2MASS J052832        &  6020864 &  240 &  720 & 22 & 2005 Jul 10 & 2453562 & 1.NO      & $-$5.45\tablenotemark{d}\\
595: HV 2572              &  6014208 &   56 &  480 &  6 & 2004 Apr 18 & 2453114 & 2.SE7 tf: & $-$6.78\tablenotemark{e}\\
596: HV 2575              &  6016256 &  112 &  480 & 21 & 2005 Jun 08 & 2453530 & 1.N:      & $-$5.51\\
597: HV 2575 (follow-up)  & 17399552 & 1920 & 2880 & 32 & 2006 Jul 02 & 2453919 & 2.NO      & $-$5.51\\
598: HV 2578              &  6014720 &   56 &  480 & 21 & 2005 Jun 05 & 2453527 & 2.SE7tf   & $-$6.88\\
599: IRAS 05300$-$6651    &  6024704 &   56 &  240 & 21 & 2005 Jun 04 & 2453525 & 3.CR      & $-$5.17\\
600: HV 996               &  6015744 &   56 &  120 & 21 & 2005 Jun 04 & 2453526 & 2.SE7     & $-$8.30\\
601: IRAS 05329$-$6708    &  6023168 &   56 &  120 &  6 & 2004 Apr 18 & 2453114 & 3.SAxf    & $-$7.06\\
602: HV 12620             &  6016000 &  240 &  720 & 21 & 2005 May 26 & 2453517 & 1.N:      & $-$5.66\\
603: IRAS 05348$-$7024    &  6024448 &   56 &  480 &  6 & 2004 Apr 16 & 2453112 & 3.CR      & $-$5.62\\
604: IRAS 05402$-$6956    &  6020608 &   56 &  120 & 21 & 2005 May 26 & 2453517 & 3.SBxf    & $-$7.20\\
605: HV 12667             &  6015232 &   56 &  480 & 21 & 2005 May 30 & 2453521 & 2.SE6 tf: & $-$6.26\\
606: HV 12070             &  6016512 &   56 &  120 &  6 & 2004 Apr 16 & 2453112 & 2.SE6     & $-$6.48\\
607: %\\
608: %(NGC 371 LE 31)      &  6018560 &  240 &  720 & 21 & 2005 Jun 02 & 2453524 & \nodata   & \nodata \\
609: %(NGC 371 LE 8)       &  6018816 &  240 &  480 & 21 & 2005 Jun 02 & 2453524 & \nodata   & \nodata \\
610: %(WBP 219)            &  6021888 &  240 &  720 & 21 & 2005 May 26 & 2453517 & \nodata   & \nodata \\
611: \enddata
612: \tablenotetext{a}{Based on {\it Spitzer} spectra and SSO photometry (in Table 3), except where noted.}
613: \tablenotetext{b}{Using 2MASS and S$^3$MC photometry gives M$_{bol}$ = $-$6.73.}
614: \tablenotetext{c}{Using 2MASS and S$^3$MC photometry gives M$_{bol}$ = $-$7.38.}
615: \tablenotetext{d}{Using 2MASS and SAGE photometry to replace absent CASPIR photometry.}
616: \tablenotetext{e}{Using 2MASS and SAGE photometry gives M$_{bol}$ = $-$7.16.}
617: \end{deluxetable*}
618: %\end{deluxetable}
619: 
620: \begin{deluxetable*}{lrrrlr} % Table 5
621: %\begin{deluxetable}{lrrrlr}
622: \tablenum{5}
623: \tablecolumns{6}
624: \tablewidth{0pt}
625: \small
626: \tablecaption{Short-High observations\label{Tbl6}}
627: \tablehead{
628:   \colhead{ } & \colhead{ } & \colhead{SH Int.} & \colhead{IRS} & 
629:   \colhead{Observed} & \colhead{Julian} \\
630:   \colhead{Target} & \colhead{AOR key} & \colhead{Time (s)} & \colhead{Camp.} & 
631:   \colhead{(UT)} & \colhead{Date}
632: }
633: \startdata
634: IRAS 04545$-$7000    & 17399808 & 120 & 32 & 2006 Jul 01 & 2453918.4 \\
635: IRAS 05329$-$6708    & 17400320 &  60 & 32 & 2006 Jun 28 & 2453914.9 \\
636: %MSX SMC 018          & 17400832 & 120 & 32 & 2006 Jul 01 & 2453918.4 \\ % DROP
637: \enddata
638: \end{deluxetable*}
639: %\end{deluxetable}
640: 
641: We observed all of the sources in our sample with the 
642: low-resolution modules on the IRS.  These observations were 
643: carried out as part of the IRS GTO allocation, which was 
644: spread over the first 2.5 years of the {\it Spitzer}
645: mission.  As described below, we also followed 
646: up some targets in Cycle 3.  We observed four stars with
647: possible low-contrast excesses with longer integration 
648: times (PID 30332), and we followed up two sources in the 
649: MC\_DUST program with Short-High (SH) observations to verify 
650: the existence of new dust features at 14~\mum\ (PID 30345).
651: Table 4 summarizes the low-resolution observations, and 
652: Table 5 summarizes the SH observations.
653: 
654: % Tables 4 and 5 go here.
655: 
656: \subsection{Standard Reduction} % Sec. 3.2
657: 
658: Our data reduction began with the S15 version of the data 
659: pipeline released by the {\it Spitzer} Science Center (SSC).
660: Before extracting spectra from the flatfielded images, we
661: subtracted background images and corrected bad pixels.
662: For the Short-Low (SL) data, we used images with the source
663: in the other aperture as the background (aperture
664: differences).  Thus, the Short-Low order 1 (SL1) images 
665: served as background data for the SL order 2 (SL2) images, 
666: and vice versa.  For Long-Low, we obtained nod differences,
667: subtracting images with the source in one nod position from
668: images with the source in the same aperture, but in the
669: other nod position.  The next step was to correct bad
670: pixels by interpolating from their neighbors, using the
671: {\rm imclean} IDL package\footnote{Available from the SSC
672: as {\rm irsclean}.}.  Bad pixels are those flagged in the
673: bit-mask images provided with the data release and those
674: flagged as rogues in the rogue masks provided by the SSC.
675: 
676: We extracted spectra from the corrected images using the
677: {\rm profile}, {\rm ridge}, and {\rm extract} routines
678: available in the SSC's {\it Spitzer} IRS Custom Extractor
679: (SPICE).  The spectra were calibrated with spectral
680: corrections generated from observations of the K0 giant 
681: HR 6348 for SL and for LL, HR 6348 along with the late
682: K giants HD 166780 and HD 173511 \citep[see][for further
683: details]{slo08c}.  To eliminate pointing-induced
684: discontinuities between the orders, we normalized the
685: various segments to match at the wavelengths where they
686: overlapped.  We normalized segments upwards to the
687: presumably best-centered segment.  Finally, we trimmed
688: the segments to eliminate untrustworthy data at the ends.
689: 
690: The SH observations required a different reduction method.
691: We used S14 images, substracted dedicated sky images, and
692: performed full-slit extractions.  We calibrated using
693: observations of $\xi$ Dra (K2 III).
694: 
695: \subsection{Bolometric Magnitudes} % Sec. 3.3
696: 
697: Table 4 also includes the bolometric magnitudes of our
698: sources at the time of observation, which result from an 
699: integration of the IRS spectra and the near-infrared SSO 
700: photometry.  We have assumed a Wien tail to the blue and a 
701: Rayleigh-Jeans tail to the red, and a distance modulus of 
702: 18.5 to the LMC and 18.9 to the SMC.  
703: 
704: Where possible, we have also computed bolometric magnitudes 
705: by replacing the SSO data with the 2MASS photometry in Table 2 
706: and 3.6 and 4.5~\mum\ {\it Spitzer} photometry from
707: the SAGE survey of the LMC \citep{mei06} and the S$^3$MC
708: survey of the SMC \citep{bol07}.\footnote{SAGE and S$^3$MC
709: stand, respectively, for Surveying the Agents of Galactic
710: Evolution and the {\it Spitzer} Survey of the SMC.}  In 
711: those cases where the data give a significantly different 
712: result, Table 4 includes a note of explanation.
713: 
714: The observational distinction between red supergiants and 
715: oxygen-rich AGB stars is non-trivial.  Sources in the 
716: Magellanic Clouds have known distances, making luminosity
717: a logical discriminant (pulsation amplitude is another).  
718: \cite{wbf83} estimate that the classic upper limit for AGB 
719: sources lies at about M$_{bol}$=$-7.1$, but as \cite{woo92} 
720: discussed, some AGB sources can occasionally stray above that 
721: limit.  The bolometric luminosities in Table 4 suffer from 
722: the additional difficulty of variability.  We have observed 
723: the stars at only one phase, which blurs the $-7.1$-magnitude 
724: limit somewhat.
725: 
726: For the classifications in Tables 1 and 2, we assume all RSGs 
727: exceed the classic limit and show no large-amplitude 
728: pulsation, OH maser emission, or Li absorption at 0.67~\mum.  
729: The Li absorption indicates that the source has experienced 
730: hot-bottom burning, placing it at the high-mass end of the 
731: AGB \citep{smi95}.  
732: 
733: Four of our targets meet all of these constraints.  HV 888 
734: and HV 996 are firmly identified supergiants in the LMC 
735: \citep[e.g.][]{wbf83}, as is Massey SMC 59803 in the SMC 
736: \citep{lev06}.  The fourth source is IRAS 04530, but this
737: source is probably a young stellar object (YSO), as 
738: explained in \S 7.3.
739: 
740: Two more sources meet some, but not all, of the criteria for 
741: a supergiant.  IRAS 04509 and IRAS 04516 have bolometric 
742: luminosities over the AGB limit by our measurements (with 
743: both SSO and 2MASS photometry), but \cite{woo92} suggested 
744: that they were AGB sources, having measured their bolometric 
745: luminosities to be $-7.30$ and $-6.77$, and
746: their amplitudes at K to be 1.5 and 1.3 magnitudes, 
747: respectively.  They noted that very red AGB sources might 
748: exceed the $-7.1$-magnitude limit slightly.  However, they 
749: found no evidence of OH maser activity in IRAS 04509, and 
750: \cite{mar04} found none in IRAS 04516.  Our bolometric 
751: luminosities are well above the classic AGB limit, and
752: while both sources have no detected maser activity, their
753: K-band amplitudes are inconsistent with supergiant status.
754: These two sources are as luminous as AGB sources can be, and
755: we suspect that they are massive AGB sources.
756: 
757: \cite{woo92} and \cite{whi03} detected masers in all three
758: targets identified as OH/IR sources in Table 1.  IRAS 05329 
759: and IRAS 05402 have bolometric luminosities in some 
760: references over the AGB limit, but the presence of maser 
761: emission clarifies their status.  IRAS 04545 is under the AGB 
762: limit in most references.
763: 
764: Two sources show evidence of hot-bottom burning, HV 2572
765: \citep{smi95,wf00} and HV 12070 \citep{whi03}, making them
766: massive AGB stars.  \cite{wbf83} suggested that HV 2572
767: might be a foreground object, but its radial velocity is
768: more consistent with membership in the LMC (P.\ Wood,
769: private communication).
770: 
771: %                    woo92                   whi03
772: % target        <K>  M$_{bol}$  my M$_{bol}$ M$_{bol}$
773: % IRAS 04545  10.64   -6.15     -6.31 -6.42  -6.56   P=1216d  DelK=1.57  (I had P=1270)
774: % IRAS 05329   9.87   -6.86     -7.13 -7.06  -6.95   P=1262d  DelK=1.50
775: % IRAS 05402  10.12   -6.50     -7.14 -7.20  -6.77   P=1393d  DelK=1.80
776: 
777: % IRAS 04509   8.35   -7.30     -7.75 -7.66          woo92 - DelK=1.5
778: % IRAS 04516   8.86   -6.77     -7.65 -7.43          woo92 - DelK=1.3
779: % IRAS 04530   9.79   -7.62     -6.44 -7.78          woo92 - DelK=0.2
780: 
781: \subsection{Infrared Spectral Classification} % Sec. 3.4
782: 
783: The first step in our analysis is to classify each source 
784: into one of five broad categories based on the general 
785: properties of its infrared spectrum.  Each spectrum is then 
786: analyzed in more detail, using a method appropriate to its
787: classification.
788: 
789: We applied the classification method developed at 
790: Hanscom Field and introduced by \cite{kra02}.  This system 
791: places a spectrum into one of five numbered groups based on 
792: its overall color:  1 for blue spectra dominated by a stellar
793: continuum, 2 for dusty stars, 3 for spectra dominated by warm 
794: dust, and 4 and 5 for spectra dominated by progressively 
795: cooler dust.\footnote{These last two groups are distinguished
796: using data from 36 to 45~\mum, which are unavailable to the
797: IRS.  We simply classify the one red spectrum as group 5.}  
798: It then describes the dominant features in the spectra with 
799: one- or two-letter abbreviations:  N for naked stars, SE for 
800: silicate emission, SB for self-absorbed silicate emission, SA 
801: for silicate absorption, CE for carbon-rich dust emission, CR 
802: for optically thick and cooler carbon dust shells, and U for 
803: spectra showing the unidentified infrared (UIR) emission 
804: features arising from polycyclic aromatic hydrocarbons (PAHs) 
805: and related material.  We introduce a new classification, 
806: 2.ST, in \S 5.6.
807: 
808: Table 4 gives the infrared spectral classifications, which
809: include the basic classification plus additional subdivisions
810: and notes as described in the following sections.  The 
811: overall classification determines the method(s) applied to 
812: analyze the spectra and in which sections they are
813: discussed.  In \S 4, we discuss the naked stars (1.N and 
814: 1.NO) and some of the nearly naked spectra showing other dust 
815: features.  In \S 5 we examine oxygen-rich dust spectra (2.SE, 
816: 3.SE, 3.SB, and 3.SA), while \S 6 examines carbon-rich 
817: spectra (1.NC, 2.CE and 3.CR).  Finally, \S 7 looks at the 
818: spectra with red continua and/or contributions from PAHS.
819: 
820: Distinguishing the nature of the dust (oxygen-rich vs.
821: carbon-rich) is straightforward if there is plenty of dust,
822: but for sources with low-contrast excesses, not only can it
823: be difficult to determine what kind of dust is present, but
824: it can be a challenge just to decide if there is an excess
825: at all.  To assist, we have measured the dust emission 
826: contrast (DEC), which \cite{sp95} defined to be the 
827: ratio of dust emission to stellar emission between 7.67 and 
828: 14.03~\mum.  They applied this method to the database of
829: Low-Resolution Spectra (LRS) from {\it IRAS}.  \cite{sp98}
830: noted that all sources with DEC $>$ 0.08 had clear dust
831: excesses, while all sources with DEC$<$ 0.04 looked naked.  
832: The sources between these two limits were more problematic  
833: and could be described as ``fig-leaf'' sources, since they 
834: are neither naked nor fully clothed by a dust shell.
835: 
836: We follow their example, but with some changes.  As done 
837: previously, we assume the optically thin case, which implies 
838: that the contributions to the spectrum from the star and dust 
839: are additive.  The better coverage to shorter wavelengths of 
840: the IRS compared to the LRS allows us to shift the wavelength 
841: range for fitting the stellar contribution to 6.8--7.4~\mum, 
842: between an absorption band from water vapor at 
843: $\sim$6.6~\mum\ and the fundamental SiO band, which begins at 
844: 7.4~\mum.  As discussed below (\S 4), the SiO band is much 
845: weaker in the Magellanic Clouds than in the Galaxy, so 
846: instead of using a model spectrum with an SiO band, we simply 
847: fit and subtract a 3600 K Planck function to measure the DEC.
848: This also makes the method general enough to apply to carbon 
849: stars.
850: 
851: While \cite{sp98} set the minimum DEC for dusty stars to be 
852: 0.08, we find that in the MC\_DUST sample, sources with a DEC
853: as high as 0.10 do not have a sufficient excess to identify
854: unambiguously.  We classify sources with DEC$\le$0.04 as 1.N,
855: while the fig-leaf sources with 0.04$<$DEC$\le$0.10 are 1.N:.
856: Sources with higher dust emission contrasts are classified in 
857: group 2, and in all cases but one, can be placed into the 
858: appropriate sub-category (2.SE, 2.CE, etc.).
859: 
860: Three sources (BFM 1, HV 5810, and WBP 77) have DECs 
861: high enough to qualify as dusty, but we can see through the
862: dust and investigate their photospheres.
863: Consequently, we will examine their photospheric properties
864: in the section below on naked stars (\S 4) and their dust 
865: content in the appropriate section as well.  
866: % Similarly, the naked star WBP 17 is carbon-rich, 
867: % so it is analyzed in both \S 4 and \S 6.  NOW JUST \S 6
868: 
869: Two of our sources are identified as C/IR stars in Table 2,
870: IRAS 05300 and IRAS 05348.  \cite{vl06} list
871: IRAS 05348 as ``unknown type'', but our IRS data clearly show 
872: a carbon star.  As noted in \S 2.2, IRAS 05128 has previously 
873: been identified as carbon-rich based on its {\it ISO} 
874: spectrum, but both our optical and infrared spectra reveal an 
875: oxygen-rich source.
876: 
877: One source, MACHO 79.5505.26, has inconsistent optical and
878: infrared spectra.  The optical spectrum is clearly 
879: carbon-rich, but the infrared spectrum shows no dust or
880: absorption bands indicating its chemistry, which is unusual
881: for carbon stars.  For this reason, we analyze it both in
882: \S 4 and 6.  Similarly, the SMC supergiant Massey
883: SMC 59803 shows both silicate and PAH emission (analyzed 
884: separately in \S 5 and 7).
885: 
886: 
887: \section{Naked Stars} % Sec. 4
888: 
889: \begin{figure} % Fig. 2
890: \includegraphics[width=3.5in]{f02.eps}
891: \caption{IRS spectra of naked oxygen-rich stars in the SMC.  
892: The spectra have been normalized and shifted for clarity.  
893: The 8~\mum\ flux, determined by applying the bandpass of the
894: IRAC 8~\mum\ filter to the IRS data, is given after each 
895: source name in parentheses.}
896: \end{figure}
897: 
898: \begin{figure} % Fig. 3
899: \includegraphics[width=3.5in]{f03.eps}
900: \caption{IRS spectra of naked oxygen-rich stars in the LMC.  
901: As in Fig.\ 2, the spectra have been normalized and shifted.}
902: \end{figure}
903: 
904: Figures 2 and 3 present the spectra of the naked
905: sources in the SMC and LMC, respectively.  These include seven 
906: of the 11 SMC sources and five of the 27 LMC sources.  
907: 
908: % Figures 2-3 go here.
909: 
910: Many of these spectra are faint, which exposes us to
911: background issues that would have little effect on brighter
912: sources.  The Magellanic Clouds are crowded in the infrared,
913: with extended emission and a multitude of stray sources.
914: These have a greater impact on LL, with its 10\arcsec\ slit
915: and sensitivity to extended emission from cool dust.  In some 
916: of the spectra, the source is faint enough and blue enough 
917: that we failed to detect it in LL.  For this reason, our 
918: analysis will concentrate on the SL data.  Even though some 
919: of the LL data are excellent, we did not detect any 
920: significant spectral features beyond about 15~\mum.
921: 
922: \subsection{Method} % Sec. 4.1.
923: 
924: \begin{figure} % Fig. 4
925: \includegraphics[width=3.5in]{f04.eps}
926: \caption{IRS spectra of naked stars in the SMC, plotted in 
927: Rayleigh-Jeans units ($\lambda^2 F_{\nu}$), so that the 
928: Rayleigh-Jeans tail of a blackbody would be a horizontal 
929: line.  For each spectrum, the dotted lines give the fitted 
930: continuum and the best fit absorption from an average 
931: Galactic M giant.  Table 6 gives the equivalent width of
932: the fitted SiO band.}
933: \end{figure}
934: 
935: \begin{figure} % Fig. 5
936: \includegraphics[width=3.5in]{f05.eps}
937: \caption{IRS spectra of naked stars in the LMC, plotted in 
938: Rayleigh-Jeans units with the fitted fitted continuum and 
939: absorption spectra, as in Fig.\ 4.}
940: \end{figure}
941: 
942: \begin{figure} % Fig. 6
943: \includegraphics[width=3.5in]{f06.eps}
944: \caption{IRS spectra of stars in the SMC and LMC with
945: indications of low-contrast dust emission (in Rayleigh-Jeans
946: units).  The dotted lines are the fitted continua and 
947: absorption spectra, as in Fig.\ 4.  Spectra with full
948: wavelength coverage for these sources appear in later
949: figures:  BFM 1 in Fig.\ 8, HV 5810 in Fig. 22, and WBP 77
950: in Fig.\ 9.}
951: \end{figure}
952: 
953: \begin{deluxetable*}{lrrrl} % Table 6
954: %\begin{deluxetable}{lrrrl}
955: \tablenum{6}
956: \tablecolumns{5}
957: \tablewidth{0pt}
958: \tablecaption{Naked and nearly naked stars\label{Tbl7}}
959: \tablehead{
960:   \colhead{ } & \colhead{Dust Em. } & \colhead{Fitted} &
961:   \colhead{W$_{\lambda}$ SiO} & \colhead{Infrared} \\
962:   \colhead{Target} & \colhead{Contrast} & \colhead{T (K)} &
963:   \colhead{(\mum)} & \colhead{Sp. Class}
964: }
965: \startdata
966: HV 11223             & $-$0.02 $\pm$ 0.08 &  2290 $\pm$  510    & $<$ 0.18        & 1.N \\
967: HV 1366              &    0.02 $\pm$ 0.06 &  2270 $\pm$  560    & $<$ 0.20        & 1.N \\
968: BFM 1                &    0.24 $\pm$ 0.04 &  1460 $\pm$  110    & $<$ 0.16        & 2.ST: \\
969: BFM 1 (follow-up)    &    0.16 $\pm$ 0.02 &  1160 $\pm$   10    & 0.15 $\pm$ 0.05 & 2.N:O \\
970: HV 11303             &    0.07 $\pm$ 0.03 &  2130 $\pm$  240    & $<$ 0.03        & 1.N: \\
971: HV 11329             &    0.04 $\pm$ 0.05 &  1600 $\pm$   80    & 0.28 $\pm$ 0.13 & 1.NO \\
972: HV 11329 (follow-up) &    0.00 $\pm$ 0.02 &  2380 $\pm$  100    & 0.18 $\pm$ 0.03 & 1.NO \\
973: HV 838               &    0.10 $\pm$ 0.03 &  2320 $\pm$  320    & $<$ 0.02        & 1.N: \\
974: HV 11366             &    0.06 $\pm$ 0.09 &  2230 $\pm$  300    & 0.18 $\pm$ 0.19 & 1.N:O:: \\
975: HV 1963              &    0.07 $\pm$ 0.05 &  2160 $\pm$  120    & 0.21 $\pm$ 0.13 & 1.N:O: \\
976: HV 1963 (follow-up)  &    0.05 $\pm$ 0.01 &  1990 $\pm$   50    & 0.13 $\pm$ 0.03 & 1.N:O \\
977: \\
978: MACHO 79.5505.26     & $-$0.06 $\pm$ 0.04 &  4120 $\pm$ 2170    & $<$ 0.11        & 1.N \\
979: HV 5810              &    0.19 $\pm$ 0.08 &  1210 $\pm$  110    & 0.38 $\pm$ 0.17 & 2.U \\
980: %WBP 17               & $-$0.00 $\pm$ 0.04 & 10000 $\pm$ 2950    & \nodata         & 1.NC \\
981: WBP 77               &    0.19 $\pm$ 0.07 &  1130 $\pm$  120    & 0.33 $\pm$ 0.18 & 2.SE1: \\
982: WBP 104              & $-$0.11 $\pm$ 0.06 &  4080 $\pm$  890    & 0.20 $\pm$ 0.19 & 1.NO:: \\
983: 2MASS J052832        & $-$0.21 $\pm$ 0.05 & 10000 $\pm$ \nodata & 0.34 $\pm$ 0.15 & 1.NO \\
984: HV 2575              &    0.07 $\pm$ 0.07 &  2170 $\pm$  870    & 0.09 $\pm$ 0.17 & 1.N: \\
985: HV 2575 (follow-up)  &    0.07 $\pm$ 0.02 &  2260 $\pm$  180    & 0.24 $\pm$ 0.08 & 1.N:O \\
986: HV 12620             &    0.06 $\pm$ 0.03 &  1900 $\pm$   10    & $<$ 0.04        & 1.N: \\
987: \enddata
988: \end{deluxetable*}
989: %\end{deluxetable}
990: 
991: The Hanscom system includes two subclasses of naked stars, 
992: oxygen-rich (1.NO), and carbon-rich (1.NC).  The 1.NO spectra 
993: can be distinguished by the presence of the SiO fundamental 
994: band, with maximum absorption at 8.0~\mum.  For 1.NC spectra, 
995: the primary discriminant is the acetylene (C$_2$H$_2$) 
996: absorption band at 7.5~\mum, and in some cases, a narrower 
997: acetylene band at 13.7~\mum.  Those sources which cannot be
998: identified as 1.NO or 1.NC are classified simply as 1.N.
999: We examine the one 1.NC spectrum, of WBP 17, in \S 6.
1000: 
1001: The SiO fundamental has a bandhead at 7.4~\mum.  While the
1002: wavelength of maximum absorption is nearby, at 8.0~\mum, the
1003: band extends out as far as $\sim$11.0~\mum\ in some cases.  
1004: The band is broad enough, and the spectra are noisy enough, 
1005: that our usual technique of fitting a line segment to the
1006: continuum on either side of the band to integrate the
1007: equivalent width is inadequate.  First, the continuum is not 
1008: linear over such a broad wavelength range, and second, our
1009: estimated equivalent width has proven too sensitive to the
1010: vagueries of the continuum fit.  Instead, we have constructed
1011: a template transmission spectrum from Cycle 4 IRS GTO 
1012: observations of 10 naked M giants \citep[PID 40112;][]{slo08d}
1013: and fitted it to the spectrum to estimate the equivalent
1014: width.\footnote{The sample currently includes 13 reduced 
1015: spectra, but we did not use three with strong water vapor 
1016: absorption at 6.6~\mum\ when generating the SiO template.}   
1017: We estimate the continuum with a Planck function which 
1018: reproduces the color measured over the wavelength intervals 
1019: 6.10--6.45 and 11.5--13.0~\mum.  We then scale the 
1020: transmission spectrum until we find the minimum $\chi^2$ 
1021: difference, and measure the equivalent width of our fitted 
1022: spectrum from 7.4 to 11.5~\mum.  We estimate the uncertainty 
1023: in the fitted temperature by varying the long-wavelength 
1024: fitting interval, and we estimate the uncertainty in 
1025: equivalent width by finding the SiO strengths which 
1026: correspond to the minimum $\chi^2$ multiplied by the factor 
1027: $(1 + N^{-1/2})$, where $N$ is the number of pixels in the 
1028: measured wavelength range.  
1029: 
1030: Figures 4 and 5 illustrate the best fits to the naked stars, 
1031: and Figure 6 shows the three sources with both possible 
1032: low-contrast dust and photospheric absorption bands.  The 
1033: results of these fits appear in Table 6.
1034: 
1035: % Figures 4-6 and Table 6 go here.
1036: 
1037: \subsection{Results} % Sec. 4.2
1038: 
1039: \begin{figure} % Fig. 7
1040: \includegraphics[width=3.5in]{f07.eps}
1041: \caption{The follow-up spectra with longer integration times 
1042: of four of the MC\_DUST targets.  The originally observed
1043: spectrum appears in light gray, while the new spectrum is in 
1044: black.  The dotted lines are the continuum fitted to the new 
1045: spectrum and the best-fit absorption spectrum, as in Fig.\ 3.}
1046: \end{figure}
1047: 
1048: For each naked spectrum, Table 6 lists the dust emission 
1049: contrast, the temperature of the fitted Planck function, and 
1050: the equivalent width of the best-fitting SiO band.  One 
1051: should not overinterpret the temperatures.
1052: Spectra from sources which were not properly centered in the 
1053: slit will be subtly distorted, since the slit throughput is 
1054: a complex function of both wavelength and position in the 
1055: slit \citep{ks06}.  As Table 6 shows, the uncertainty in the
1056: temperature can be substantial in the noisier spectra.
1057: However, the temperatures can be useful if treated carefully.
1058: Table 6 reveals a correlation between higher dust emission 
1059: contrasts and lower temperatures, suggesting that sources 
1060: with lower blackbody temperatures do show low-contrast excess 
1061: emission from dust.
1062: 
1063: The classifications of naked stars in Tables 3 and 6 require
1064: some clarification.  If a colon follows the ``N", it indicates
1065: uncertainty about whether or not the source is naked (i.e.\ it
1066: is a ``fig-leaf'' source).  If the colon follows the ``O'',
1067: then the SiO band is detected with a S/N between $\sim$1 and 
1068: $\sim$2.
1069: 
1070: Figure 6 presents the three sources with a DEC too high to
1071: be naked and too low to identify the nature of the dust
1072: excess.  BFM 1, the S star in the SMC, has a DEC of 
1073: $\sim$0.20, indicating the presence of some dust.  The full
1074: spectrum of this source appears in \S 5, along with a 
1075: discussion of the interesting feature at 13--14~\mum.  
1076: HV 5810 shows an apparent SiO absorption feature, but it 
1077: also shows apparent emission features at 11.3, 12.0, and 
1078: 12.7~\mum, the positions of the out-of-plane C--H bending 
1079: modes in PAHs \citep[e.g.][]{atb89}.  The apparent SiO 
1080: absorption is more likely just the trough between the 
1081: 7--9~\mum\ and 11--13~\mum\ PAH complexes (despite a S/N
1082: ratio of $\sim$2).  This interpretation is strengthened by 
1083: the spectral structure at 8.5~\mum, which resembles the 
1084: emission feature from the C--H in-plane bending mode in PAHs.
1085: We discuss this source further in \S 7.2 below.
1086: 
1087: %could also mention lack of CO in HV 5810
1088: 
1089: Figure 7 presents the follow-up observations with longer
1090: integrations of four of our targets.  We have detected SiO 
1091: bands in all four of the follow-up spectra at a S/N of 
1092: $\sim$3 or better, which is an improvement over the original 
1093: spectra, where the limited S/N allowed only two detections
1094: with a S/N ratio of just $\sim$2.  The follow-up spectrum 
1095: of BFM 1 has changed compared to the original.  The apparent 
1096: 13--14~\mum\ emission feature is absent, while the SiO band
1097: is now clearly present, though weak.  The S/N ratio of the
1098: spectrum of HV 2575 is high enough to reveal weak absorption 
1099: between 6.5 and 7.4~\mum\ from water vapor \citep{cam02}.
1100: 
1101: % Figure 7 goes here.
1102: 
1103: We will return to the naked stars in \S 8.1 and 8.3 below.
1104: 
1105: 
1106: \section{Oxygen-rich Dust Sources} % Sec. 5
1107: 
1108: \begin{figure} % Fig. 8
1109: \includegraphics[width=3.5in]{f08.eps}
1110: \caption{Spectra of the three SMC sources showing 
1111: silicate or possibly related dust emission features.}
1112: \end{figure}
1113: 
1114: \begin{figure*} % Fig. 9
1115: \includegraphics[width=7.0in]{f09.eps}
1116: \caption{Spectra of the 12 LMC sources with silicate or
1117: related emission.}
1118: \end{figure*}
1119: 
1120: \begin{figure} % Fig. 10
1121: \includegraphics[width=3.5in]{f10.eps}
1122: \caption{Spectra of the three LMC sources showing silicate
1123: absorption features or self-absorbed silicate features at
1124: 10~\mum.  These spectra are analyzed in Fig.\ 12 and 13.}
1125: \end{figure}
1126: 
1127: The MC\_DUST sample contains 18 spectra dominated by silicate 
1128: dust and related oxygen-rich dust species, as Figures 8 through 
1129: 10 show.  The SMC sample has three oxygen-rich dust emission 
1130: sources, and no sources with enough dust opacity to drive the 
1131: 10~\mum\ feature into absorption.  The LMC sample includes 12 
1132: emission sources, one red source showing a self-absorbed 
1133: 10~\mum\ silicate feature, and two even redder sources with 
1134: silicate absorption at 10~\mum.  
1135: 
1136: % Figures 8-10 go here.
1137: 
1138: \subsection{Classification} % Sec. 5.1
1139: 
1140: \begin{figure} % Fig. 11 
1141: \includegraphics[width=3.5in]{f11.eps}
1142: \caption{The silicate dust sequence.  MC\_DUST sources are
1143: plotted with black filled circles.  The smallar gray circles
1144: show the distribution of Galactic AGB sources 
1145: \citep{sp95,sp98}, and the dotted curve is a power law:  
1146:   $F_{10}/F_{12} = 1.32 (F_{10}/F_{11})^{1.77}$.}
1147: \end{figure}
1148: 
1149: \begin{deluxetable*}{lrrrrrl} % Table 7
1150: %\begin{deluxetable}{lrrrrrl}
1151: \tablenum{7}
1152: \tablecolumns{7}
1153: \tablewidth{0pt}
1154: \tablecaption{Silicate emission analysis\label{Tbl11}}
1155: \tablehead{
1156:   \colhead{ } & \colhead{ } & \colhead{Dust Em.} & \colhead{ } & 
1157:   \colhead{ } & \colhead{Corrected} & \colhead{Infrared} \\
1158:   \colhead{Target} & \colhead{[7]$-$[15]} & \colhead{Contrast} & 
1159:   \colhead{$F_{10}/F_{11}$} & \colhead{$F_{10}/F_{12}$} & 
1160:   \colhead{$F_{10}/F_{11}$} & \colhead{Sp. Class}
1161: }
1162: \startdata
1163: BFM 1             & 0.41$\pm$0.07 & 0.24$\pm$0.04 & \nodata       & \nodata       & \nodata       & 2.ST\\
1164: BFM 1 (follow-up) & 0.62$\pm$0.06 & 0.16$\pm$0.02 & \nodata       & \nodata       & \nodata       & 2.NO\\
1165: HV 12149          & 0.88$\pm$0.02 & 0.80$\pm$0.04 & 1.39$\pm$0.05 & 2.29$\pm$0.13 & 1.68$\pm$0.11 & SE8\\
1166: Massey SMC 59803  & 0.58$\pm$0.01 & 0.23$\pm$0.01 & 0.77$\pm$0.03 & 1.08$\pm$0.04 & 1.19$\pm$0.06 & SE4\\
1167: \\
1168: IRAS 04509        & 2.12$\pm$0.01 & 3.96$\pm$0.05 & 1.17$\pm$0.02 & 2.06$\pm$0.07 & 1.59$\pm$0.06 & SE8\\
1169: IRAS 04516        & 2.00$\pm$0.01 & 3.22$\pm$0.02 & 1.02$\pm$0.01 & 1.47$\pm$0.04 & 1.38$\pm$0.04 & SE6\\
1170: IRAS 04545        & 2.22$\pm$0.01 & 1.51$\pm$0.01 & \nodata       & \nodata       & \nodata       & 3.SA\\
1171: HV 888            & 1.60$\pm$0.01 & 1.90$\pm$0.01 & 1.27$\pm$0.02 & 2.00$\pm$0.05 & 1.58$\pm$0.05 & SE8\\
1172: HV 2310           & 0.88$\pm$0.03 & 0.63$\pm$0.02 & 1.00$\pm$0.04 & 1.26$\pm$0.02 & 1.30$\pm$0.05 & SE5\\
1173: IRAS 05128        & 1.93$\pm$0.00 & 2.42$\pm$0.02 & 0.95$\pm$0.01 & 1.25$\pm$0.02 & 1.29$\pm$0.03 & SE5\\
1174: HV 5810           & 1.02$\pm$0.02 & 0.19$\pm$0.08 & \nodata       & \nodata       & \nodata       & 2.U\\
1175: WOH G 339         & 0.92$\pm$0.02 & 0.46$\pm$0.02 & 1.11$\pm$0.04 & 1.14$\pm$0.06 & 1.27$\pm$0.08 & SE5\\
1176: WBP 77            & 1.01$\pm$0.10 & 0.19$\pm$0.07 & 0.72$\pm$0.14 & 0.43$\pm$0.10 & 0.88$\pm$0.26 & SE1:\\
1177: HV 2572           & 1.06$\pm$0.01 & 0.80$\pm$0.02 & 1.23$\pm$0.02 & 1.73$\pm$0.03 & 1.49$\pm$0.03 & SE7\\
1178: HV 2578           & 1.05$\pm$0.01 & 0.85$\pm$0.03 & 1.24$\pm$0.03 & 1.85$\pm$0.06 & 1.53$\pm$0.06 & SE7\\
1179: HV 996            & 1.63$\pm$0.01 & 1.88$\pm$0.01 & 1.19$\pm$0.01 & 1.65$\pm$0.04 & 1.46$\pm$0.04 & SE7\\
1180: IRAS 05329        & 1.94$\pm$0.00 & 1.14$\pm$0.03 & \nodata       & \nodata       & \nodata       & 3.SB\\
1181: IRAS 05402        & 2.24$\pm$0.00 & 2.29$\pm$0.04 & \nodata       & \nodata       & \nodata       & 3.SA\\
1182: HV 12667          & 0.92$\pm$0.01 & 0.57$\pm$0.03 & 1.17$\pm$0.02 & 1.54$\pm$0.06 & 1.42$\pm$0.06 & SE6\\
1183: HV 12070          & 0.60$\pm$0.02 & 0.29$\pm$0.02 & 1.16$\pm$0.01 & 1.50$\pm$0.02 & 1.41$\pm$0.03 & SE6\\
1184: \enddata
1185: \end{deluxetable*}
1186: %\end{deluxetable}
1187: 
1188: The classification method introduced by \cite{sp95,sp98} 
1189: quickly distinguishes the composition of the emitting 
1190: oxygen-rich dust.  We apply that method here, fitting an 
1191: assumed photospheric spectrum to the blue end of the observed
1192: spectrum and subtracting it to isolate the dust contribution.  
1193: Because the photospheric spectra in the Magellanic Clouds 
1194: have weaker SiO absorption than in the Galactic sample (see 
1195: \S 4), we depart from the previous method and simply assume a 
1196: 3600 K Planck function for the star.  We fit it to the 
1197: spectrum in the interval 6.8--7.4~\mum\ and subtract it.  The 
1198: classification is based on the ratios of the flux from the 
1199: dust spectrum in small wavelength intervals centered at 10, 
1200: 11, and 12~\mum.  Figure 11 shows that the flux ratios from 
1201: oxygen-rich dust trace the {\it silicate dust sequence}.  We 
1202: classify the spectra based on their position up this 
1203: sequence, starting with SE1 in the lower left and running to 
1204: SE8 in the upper right.  The silicate emission index = 
1205: $10 (F_{11}/F_{12}) - 7.5$, where $F_{11}/F_{12}$ is the 
1206: corrected ratio found by locating the point on the power law 
1207: $F_{10}/F_{12} = 1.32 (F_{10}/F_{11})^{1.77}$ closest to the 
1208: actual value for a given spectrum.  Table 7 presents the 
1209: results.
1210: 
1211: We also measured the magnitude  of the stars in two narrow
1212: filters (6.8--7.4 and 14.4--15.0~\mum) to produce a [7]$-$[15]
1213: color.  Table 7 includes these colors; we will examine these 
1214: colors more closely when comparing other samples in a future 
1215: paper.
1216: 
1217: % Figure 11 goes here.
1218: 
1219: % Table 7 goes here.
1220: 
1221: \subsection{Crystalline silicate structure beyond 20~\mum} % Sec. 5.2
1222: 
1223: \begin{figure} % Fig. 12 
1224: \includegraphics[width=3.5in]{f12.eps}
1225: \caption{The crystalline silicate emission features apparent
1226: in three LL spectra.  The data plotted in gray are the 
1227: original spectra, while the data plotted in black are the
1228: spectra which have been defringed by boxcar smoothing, as
1229: explained in the text.  The thin black line is the spline
1230: fitted at wavelengths marked by diamonds.  Fig.\ 13 presents
1231: the resulting crystalline silicate spectra.}
1232: \end{figure}
1233: 
1234: \begin{figure} % Fig. 13 
1235: \includegraphics[width=3.5in]{f13.eps}
1236: \caption{The spline-subtracted crystalline silicate emission
1237: features from the spectra in Fig.\ 12.  In the top three 
1238: panels, the smoothed data are in black and the unsmoothed 
1239: data are in light gray.  The vertical lines mark reference 
1240: wavelengths at 23.25, 27.9, and 33.7~\mum.  The bottom panel 
1241: presents a model spectrum (black) composed of 50\% laboratory
1242: forsterite \citep[Mg$_2$SiO$_4$, dark gray;][]{fab01b} and 50\% 
1243: laboratory enstatite \citep[MgSiO$_3$, light gray;][]{jag98},
1244: both at 80 K and assuming a CDE2 shape distribution (see 
1245: text).  The vertical dotted lines in the bottom panel 
1246: illustrate how the observed peaks should shift from 80 K
1247: to 300 K, based on the laboratory work by \cite{sut06}.}
1248: \end{figure}
1249: 
1250: Figure 10 shows the spectra of the three most embedded 
1251: sources in the MC\_DUST sample, all in the LMC.  Two of these 
1252: sources show clear evidence of crystalline silicate emission 
1253: features at 23, 28, and 34~\mum.  The third, IRAS 05329, 
1254: shows strong fringing in LL1 that obscures the crystalline 
1255: feature at 23~\mum\ and partly hides the one at 28~\mum.  
1256: These crystalline features are much broader than the 
1257: instrumental resolution, so a boxcar smoothing algorithm
1258: applied to LL1 will mask the fringes without significantly
1259: degrading the emission features.  For IRAS 04545 and 
1260: IRAS 05402, we applied a 3-pixel boxcar to all wavelengths 
1261: and then applied an additional 5-pixel boxcar to the LL1 data
1262: to the blue of 22.88~\mum.  For IRAS 05329, with its stronger 
1263: fringing pattern, we applied a 5-pixel boxcar at all 
1264: wavelengths and a second 5-pixel boxcar to the blue of 
1265: 30.5~\mum.  We stitched and trimmed these defringed spectra 
1266: in the same manner as described in \S 3.1.  Figure 12 
1267: presents the results.  (Correcting the fringes changed the 
1268: segment-to-segment normalization.)
1269: 
1270: % Figure 12 goes here.
1271: 
1272: We fitted a spline to the data at 20.0, 21.3, 24.5, 25.5, 
1273: 31.6, and 36.5~\mum\ to produce the spectra in Figure 13.
1274: The bottom panel in Figure 13 plots a model constructed
1275: from laboratory spectra of forsterite \citep{fab01b} and 
1276: enstatite \citep{jag98}, the magnesium-rich endmembers of the 
1277: olivine (Mg$_2$SiO$_4$) and pyroxene (MgSiO$_3$) series, 
1278: respectively.  We have produced optical efficiencies from the
1279: published indices of refraction for a modified continuous 
1280: distribution of ellipsoids which uses quadratic weighting to 
1281: emphasize spherical grains (CDE2), following the methods
1282: described by \cite{fab01b}.  We found that a grain temperature
1283: of 80 K reproduced the average relative strengths of the 23 
1284: and 34~\mum\ features, which both arise from forsterite.
1285: In order to reproduce the strong emission feature at 28~\mum,
1286: our sample had to include equal parts enstatite and 
1287: forsterite.  Adding enstatite also improves the match to the 
1288: wavelength of the observed feature at 28~\mum.
1289: 
1290: % Figure 13 goes here.
1291: 
1292: Spectra from the SWS on \iso\ of Galactic sources of
1293: crystalline silicates serve as a useful comparison.
1294: \cite{mol02} found important differences between spectra
1295: produced by objects with known disks and objects where the
1296: emitting dust was in stellar outflows, particularly at 28 and 
1297: 34~\mum.  Their Figures 3, 4, and 5 present mean spectra from
1298: disk and outflow objects for the 23, 28, and 34~\mum\ 
1299: features, respectively.  For comparison, we concentrate on 
1300: IRAS 04545 and IRAS 05402 due to the higher quality of their 
1301: spectra compared to IRAS 05239.  In both spectra, the 
1302: 23~\mum\ feature shows a sharp peak at 23.25~\mum.  This 
1303: differs from both the Galactic disk sources, which show a 
1304: peak at 23.6~\mum, and the Galactic outflow sources, which 
1305: have a broader peak from 22.8 to 23.8~\mum.  The 28 and 
1306: 34~\mum\ features in the LMC sources both resemble the 
1307: Galactic disk sources.  In the LMC sources, the 28~\mum\ 
1308: feature peaks at 27.9~\mum\ and is as strong as the 23 and
1309: 34~\mum\ features.  In the Galactic samples, the 28~\mum\
1310: feature is largely absent in the outflow sources, but 
1311: strong in the disk sources.\footnote{The spectra produced 
1312: by \cite{mol02} show additional structure at $\sim$29.5 and 
1313: 30.5~\mum, but this spectral region is notoriously difficult 
1314: in the SWS data, due to problems in spectral bands 3D and 3E, 
1315: making this structure questionable \citep[][discuss this in 
1316: more detail]{slo03}.}  In the LMC sources, the 34~\mum\ 
1317: feature peaks at 33.7~\mum, as does the Galactic disk sample, 
1318: while the Galactic outflow sample shows an additional peak at 
1319: 32.9~\mum\ which appears as only a shoulder in the disk
1320: spectra.  Thus, while the 23~\mum\ feature differs somewhat
1321: from the Galactic sample, both the 28 and 34~\mum\ features
1322: closely resemble spectra from Galactic sources of crystalline
1323: silicates, especially those found in circumstellar disks.
1324: 
1325: Several effects can shift the wavelength of the observed
1326: features.  \cite{sut06} recently investigated how the emission 
1327: features from forsterite shift with temperature.  Using their 
1328: data, we could expect a shift to the red of 0.21~\mum\ in the 
1329: 23.25~\mum\ feature, 0.25~\mum\ at 27.9~\mum, and 0.22~\mum\ 
1330: at 33.7~\mum\ if we were to heat the grains in our observed 
1331: spectra from 80 K to 300 K, which is the temperature of the 
1332: measured laboratory samples.  The bottom panel of Figure 13 
1333: includes vertical dotted lines illustrating these shifts.
1334: 
1335: Another way to shift the features is to replace some of the
1336: Mg in the minerals with Fe.  \cite{chi02} and \cite{koi03}
1337: studied this effect in pyroxenes and olivines, respectively.
1338: In pyroxenes, replacing 20\% of the Mg with Fe shifts the 
1339: 28~\mum\ feature by 0.4~\mum, while a similar replacement in 
1340: olivines shifts the 23~\mum\ feature 0.3~\mum\ and the 
1341: 34~\mum\ feature 0.2~\mum.  
1342: 
1343: The shape of the grains also influences the wavelengths of
1344: the emission features, as \cite{min03} have thoroughly
1345: investigated.  Generally, a shift from spherical to
1346: non-spherical shape distributions moves the features to
1347: longer wavelengths.  Most examinations of this effect have
1348: assumed that no particular axis is favorably extended in a
1349: sample of grains.  \cite{slo06a} examined the effect of
1350: preferential non-sphericity along different axes and found
1351: that some shapes could shift the features to the blue, 
1352: instead of the red.
1353: 
1354: The solid vertical lines in Figure 13 mark the average
1355: peaks of the features in IRAS 04545 and IRAS 05402.  In
1356: the bottom panel, we should compare the peaks of the
1357: model spectrum to the {\it dotted} vertical lines, since
1358: these account for the difference in temperature between
1359: the laboratory and LMC data.  The observed position of the 
1360: 23~\mum\ forsterite feature shows no redshift compared to 
1361: the laboratory data, indicating Mg-rich grains with a CDE2 
1362: shape distribution are a good fit. The 28~\mum\ feature, 
1363: which arises from a mixture of enstatite and forsterite, is 
1364: $\sim$0.3~\mum\ to the red of the laboratory data, indicating 
1365: Fe inclusions of no more than $\sim$15\% or a slightly 
1366: different shape distribution.  Less can be said about the
1367: relative positions of the 34~\mum\ feature in the observed 
1368: and laboratory data because of the clear difference in 
1369: profile.  Overall, our observations of crystalline silicates 
1370: in the LMC are fully consistent with Mg-dominated grains, 
1371: much like those observed in the Galaxy.
1372: 
1373: \subsection{Crystalline silicate structure at 10~\mum} % Sec. 5.3
1374: 
1375: \begin{figure} % Fig. 14
1376: \includegraphics[width=3.5in]{f14.eps}
1377: \caption{Structured silicate emission at 10~\mum\ in HV 2310
1378: (top) and HV 12667 (bottom).  The IRS data (circles and error
1379: bars) have been smoothed with a 3-pixel boxcar.  The fitted
1380: model dust spectrum (solid black line) is a combination of 
1381: amorphous silicates \citep[dashed gray curve;][]{ohm92} and 
1382: forsterite \citep[Mg$_2$SiO$_4$; dashed black curve;][]{fab01b}. 
1383: For HV 2310, 7\% of the grains are crystalline, with a shape
1384: distribution including spherical, prolate (along the $x$ 
1385: axis), and oblate (extended along the $x$ and $z$ axes).  For 
1386: HV 12667, 2.5\% of the dust is crystalline, and the shape 
1387: distribution includes spherical grains and prolate grains 
1388: ($x$ axis).}
1389: \end{figure}
1390: 
1391: \cite{slo06a} published the spectrum of HV 2310, one of the
1392: first objects observed in the MC\_DUST program.  This 600-day
1393: Mira shows an unusual silicate emission spectrum with 
1394: structure in the 10--11~\mum\ region best fit with a mixture
1395: of 7\% crystalline forsterite and a blend of amorphous grains
1396: dominated by astronomical silicates \citep[as described 
1397: by][]{ohm92}.  They based their analysis on S11 pipeline
1398: output from the SSC, while the data here are S15.  While the 
1399: general shape is still indicative of crystalline silicates,
1400: the spectral details have changed enough to require a new
1401: analysis.
1402: 
1403: The MC\_DUST sample has revealed one more object showing
1404: crystalline silicate structure at 10~\mum, HV 12667.  This 
1405: source is a 645-day Mira, so it has similar pulsation 
1406: properties to HV 2310, and it also has a similar infrared
1407: brightness, 45 mJy at 8~\mum, compared to 36 mJy for HV 2310.
1408: 
1409: Figure 14 plots the continuum-subtracted spectra of HV 2310
1410: and HV 12667.  In both cases, we have fitted and subtracted
1411: a 3600 K Planck function at 5.3--5.8~\mum.  We tested the
1412: dependency of our results on the assumed stellar continuum
1413: by also using a model M6 giant with SiO absorption at 8~\mum,
1414: and we found no significant impact.  Figure 14 also includes 
1415: a rough fit to the dust spectrum, using a combination of 
1416: amorphous silicates \citep{ohm92} and forsterite 
1417: \citep{jag98}.  Following \cite{slo06a}, we have produced a 
1418: modified shape distribution in an effort to reproduce the 
1419: detailed structure of the 10~\mum\ silicate feature.  In the 
1420: case of HV 12667, we have used a distribution of spherical 
1421: grains, and prolate grains extended along the $x$ axis.  To 
1422: fit the observed spectrum, this population of crystalline 
1423: grains needs to account for 2.5\% of the dust; the rest is 
1424: amorphous silicates.  For HV 2310, the shape distribution 
1425: includes prolate grains ($x$ axis) and oblate grains ($x$ and 
1426: $z$ axes).  The crystalline component of the dust population 
1427: is 7\%, unchanged from before.  \cite{slo06a} also included 
1428: amorphous pyroxene and amorphous alumina to fit the blue and 
1429: red wings of the silicate profile, respectively.  We do not 
1430: include these components because we wish to concentrate on 
1431: the structure at the peak of the profile.
1432: 
1433: % Figure 14 goes here.
1434: 
1435: It is highly unlikely that these shape distributions fit the
1436: silicate profiles uniquely.  In fact, the changes in the 
1437: pipeline calibration have changed the spectrum of HV 2310
1438: enough that we have modified the shape distribution used by
1439: \cite{slo06a}.  The prolate grains in their shape 
1440: distribution were extended along the $z$ axis, vs.\ the $x$ 
1441: axis here.  This difference illustrates that the data are not
1442: of sufficient fidelity to pin these details down.  However,
1443: the fits to HV 2310 and HV 12667 demonstrate that the
1444: silicate profiles in both are consistent with forsterite and 
1445: that is is not necessary to invoke enstatite or minerals with 
1446: Fe components to fit the features.  
1447: 
1448: \subsection{The 13~\mum\ emission feature} % Sec. 5.4
1449: 
1450: The 13~\mum\ emission feature, first detected by 
1451: \cite{lmp86}, often appears in the spectra from Galactic
1452: oxygen-rich dust shells \citep{slo96}.  In our Magellanic
1453: samples, it is largely absent.  There may be a feature in
1454: the spectra of HV 2572 and HV 2578, but if so, it is weak,
1455: noisy, and difficult to measure.  The spectra of IRAS 04509, 
1456: IRAS 04516, and IRAS 05128 show features in this spectral 
1457: region, but they are centered closer to 14~\mum, as discussed 
1458: in the next section.
1459: 
1460: Our sample of oxygen-rich dust spectra is small, it may 
1461: contain some biases (e.g.\ for Mira variables, which are 
1462: less likely to show a 13~\mum\ feature than semiregular 
1463: variables), and some of the spectra are noisy, which can hide 
1464: low-contrast emission features.  Consequently, our conclusion
1465: that the prevalence and strength of the 13~\mum\ emission 
1466: feature is reduced in the Magellanic Clouds is somewhat
1467: preliminary.  A more thorough examination of oxygen-rich 
1468: circumstellar dust observed in other Magellanic programs is 
1469: in preparation.
1470: 
1471: \subsection{The 14~\mum\ emission feature} % Sec. 5.5
1472: 
1473: \begin{figure} % Fig. 15
1474: \includegraphics[width=3.5in]{f15.eps}
1475: \caption{The 14~\mum\ feature in IRAS 04545 (top) and 
1476: IRAS 05239 (bottom), in low-resolution IRS data (light
1477: gray) and SH data.  The SH spectra are plotted before and 
1478: after smoothing with a 7-pixel boxcar.
1479: IRAS 04545 increased 30\% in brightness,
1480: producing a high-quality SH spectrum which confirms the
1481: presence of the 14~\mum\ feature.  IRAS 05329, on the other
1482: hand, dimmed by 40\%, making confirmation more difficult.}
1483: \end{figure}
1484: 
1485: \cite{slo06a} previously noted that the spectrum of HV 2310 
1486: includes an emission feature at 14~\mum.  They were somewhat 
1487: cautious about the feature, since it appears at the boundary 
1488: between SL and LL, raising the possibility that it is an 
1489: artifact, even if no known effect in the IRS could produce 
1490: such a feature.  \cite{ec06} also detected a 14~\mum\ 
1491: emission feature in the IRS spectra of the Trojan asteroids 
1492: 624 Hektor and 921 Agamemnon, but they were also cautious and
1493: did not suggest a possible carrier.  Figures 9--10 reveal 
1494: that several sources in the MC\_DUST sample besides HV 2310 
1495: also have 14~\mum\ features.  
1496: 
1497: To test if the feature might be an artifact in the 
1498: low-resolution IRS spectra, we observed two 14~\mum\ sources,
1499: IRAS 04545 and IRAS 05329, with the SH module, as shown in 
1500: Figure 15.   Both sources are variable stars, and IRAS 04545
1501: increased in brightness by 30\% between the two observations,
1502: resulting in an excellent SH spectrum which clearly confirms
1503: the detection in SL and LL.  IRAS 05329 declined in 
1504: brightness by nearly 40\%, and as a consequence, the SH 
1505: spectrum is noisier.  While the SH data are consistent with
1506: the low-resolution data, they do not show an obvious 14~\mum\
1507: feature.  
1508: %It is interesting that a 13~\mum\ feature appears
1509: %in the SH data.  If this feature were real, its sudden
1510: %appearance between the two epochs would imply that the
1511: %feature is more variable than the dominant silicate emission
1512: %and absorption.
1513: 
1514: % Figure 15 goes here.
1515: 
1516: %At this time, the 14~\mum\ feature remains unidentified.
1517: %We have examined various pyroxene samples, since some 
1518: %produce features in the vicinity of 14~\mum, but none have
1519: %proven suitable.
1520: 
1521: \subsection{BFM 1} % Sec. 5.6
1522: 
1523: \begin{figure} % Fig. 16
1524: \includegraphics[width=3.5in]{f16.eps}
1525: \caption{The spectrum of BFM 1 from 9.5 to 16.5~\mum, as 
1526: originally observed (in black), and when followed up with 
1527: a longer integration time (in gray) 418 days later.  The 
1528: thin black curve is a spline fitted through the original
1529: spectrum to isolate the possible dust spectrum shown in the
1530: bottom panel.  In the bottom panel, the spectrum has
1531: been smoothed, and the fitted Gaussian 
1532: (dashed gray curve) is centered at 13.52~\mum\ and has a 
1533: FWHM of 0.53~\mum.} 
1534: \end{figure}
1535: 
1536: As shown in Figures 6, 7, and 8, BFM 1 has some unusual 
1537: spectral characteristics.  First, while the spectrum shows no
1538: substantial structure out to 12.5~\mum, it has a significant 
1539: dust emission contrast (0.24 when first observed, and 0.16 
1540: when followed up), indicating the presence of a featureless
1541: excess component.  Second, the original spectrum shows a
1542: previously unknown emission feature from 13 to 14~\mum.
1543: Third, the follow-up spectrum, taken 418 days later with a
1544: longer integration time, confirms the featureless dust 
1545: excess, but shows nothing where the 13--14~\mum\ feature
1546: had previously been.
1547: 
1548: BFM 1 is the only true S star in our sample, and it is a
1549: steady Mira variable with a 398-day period and an amplitude
1550: of about 4 magnitudes in the optical.  In S stars, the 
1551: C/O ratio is near unity, leaving little C or O for the 
1552: production of dust once CO has formed, which could lead to
1553: unusual dust chemistry.  Our spectra of BFM 1 certainly
1554: show unusual characteristics.  The featureless excess is
1555: not consistent with amorphous alumina, as that would leave
1556: a recognizable low-contrast feature in the 11--12~\mum\
1557: region.  Perhaps it is amorphous carbon, or some other
1558: dust component.
1559: 
1560: The possible feature at 13--14~\mum\ is also enigmatic.
1561: The follow-up spectrum did not confirm its presence, so
1562: either it disappeared in the interval between the 
1563: observations, or else it was an artifact to begin with.
1564: The two spectra were obtained 418 days apart, compared to
1565: the 398-day period of the star.  Thus the spectra were
1566: at nearly the same phase, ruling out some phase-dependent 
1567: phenomenon.  Nonetheless, we suspect the feature is real.  
1568: First, it appears in the spectra from both nod positions.  
1569: Second, it covers over 25 pixels in SL and is not the 
1570: product of any known artifact.  Third, similar features 
1571: appear in two of the 90 spectra in the sample of Galactic 
1572: S stars observed by the IRS in Cycle 3 \citep[Program 
1573: ID 30737, P.I.\ S.\ Hony]{slo08b}.
1574: 
1575: Figure 16 focuses on the 13--14~\mum\ feature in BFM 1.
1576: Fitting a spline to remove the continuum isolates a feature 
1577: which can be fit with a Gaussian centered at 13.52~\mum\ 
1578: with a full width at half maximum (FWHM) of 0.53~\mum.
1579: The unique spectrum of BFM 1 deserves a unique 
1580: classification.  We extend the Hanscom system with a 
1581: new classification, 2.ST, where the ``T'' indicates a
1582: feature centered at 13.5~\mum.
1583: 
1584: % Figure 16 goes here.
1585: 
1586: 
1587: \section{Carbon-rich Sources} % Sec. 6  NC-CE-CR 
1588: 
1589: \begin{figure} % Fig. 17
1590: \includegraphics[width=3.5in]{f17.eps}
1591: \caption{The six definite carbon stars in the MC\_DUST 
1592: sample.  CV 78, the top spectrum, is in the SMC; the rest 
1593: are in the LMC.  The spectra of WBP 17 and WBP 29 are
1594: truncated at longer wavelengths due to background issues.
1595: The spectrum of a seventh source, MACHO 79.5505.26, appears 
1596: in Fig.\ 2 and 4 and has an uncertain chemistry.}
1597: \end{figure}
1598: 
1599: The MC\_DUST sample includes one definite carbon star in the 
1600: SMC and five in the LMC.  Figure 17 presents the spectra of 
1601: these sources.  In addition, MACHO 79.5505.26 has a 
1602: carbon-rich optical spectrum but an ambiguous infrared 
1603: spectrum with no dust emission and only CO absorption at
1604: 5~\mum.  Its spectrum appears in Figures 2 and 4,   In the
1605: next section, we show that it has no carbon-rich absorption
1606: features.  We exclude it from further analysis as a carbon 
1607: star because the chemistry of the infrared component
1608: is uncertain.  
1609: 
1610: We have truncated the spectra of two sources in Figure 17
1611: to avoid some problems introduced by complex backgrounds.
1612: The LL1 data for WBP 17 are negative due to strong 
1613: background gradients.  A nearby bright carbon star 
1614: contaminates the LL data for WBP 29.
1615: 
1616: % Figure 17 goes here.
1617: 
1618: \subsection{The Manchester Method} % Sec. 6.1
1619: 
1620: \begin{deluxetable*}{lcccccccl} % Table 8
1621: %\begin{deluxetable}{lcccccccl}
1622: \tablenum{8}
1623: \tablecolumns{9}
1624: \tablewidth{0pt}
1625: \small
1626: %\rotate
1627: \tablecaption{Carbon star analysis\label{Tbl12}}
1628: \tablehead{
1629:   \colhead{ } & \colhead{[6.4]$-$[9.3]} & \colhead{[16.5]$-$[21.5]} & 
1630:   \colhead{7.5~\mum\ EW} & \colhead{13.7~\mum\ EW} &
1631:   \colhead{$\lambda_{SiC}$} & \colhead{ } & \colhead{ } & 
1632:   \colhead{Infrared} \\
1633:   \colhead{Target} & \colhead{(mag)} & \colhead{(mag)} & \colhead{(\mum)} &
1634:   \colhead{(\mum)} & \colhead{(\mum)} & \colhead{SiC/Cont.} & 
1635:   \colhead{MgS/Cont.} & \colhead{Sp. Class}
1636: }
1637: \startdata
1638: CV 78            &    0.204$\pm$0.048 &    0.178$\pm$0.061 & 0.030$\pm$0.030 & 0.024$\pm$0.034 & 11.12$\pm$0.15 & 0.197$\pm$0.019 & \nodata         & 2.CE \\
1639: MACHO 79.5505.26 &    0.208$\pm$0.044 & $-$0.307$\pm$0.350 & \nodata         & \nodata         & \nodata        & \nodata         & \nodata         & 1.N  \\
1640: WBP 17           & $-$0.113$\pm$0.047 & $-$1.027$\pm$0.770 & 0.304$\pm$0.014 & 0.157$\pm$0.032 & 11.43$\pm$0.39 & 0.092$\pm$0.031 & \nodata         & 1.NC \\
1641: WBP 29           &    0.084$\pm$0.030 &    \nodata         & 0.395$\pm$0.011 & \nodata         & 12.01$\pm$0.69 & 0.035$\pm$0.037 & \nodata         & 2.CE \\
1642: WBP 42           &    0.358$\pm$0.017 &    0.140$\pm$0.094 & 0.310$\pm$0.019 & \nodata         & 10.95$\pm$0.31 & 0.092$\pm$0.025 & \nodata         & 2.CE: \\
1643: IRAS 05300       &    1.061$\pm$0.006 &    0.280$\pm$0.011 & 0.152$\pm$0.003 & 0.047$\pm$0.010 & 11.32$\pm$0.08 & 0.098$\pm$0.005 & 0.255$\pm$0.014 & 3.CR \\
1644: IRAS 05348       &    1.050$\pm$0.006 &    0.295$\pm$0.010 & 0.013$\pm$0.004 & 0.047$\pm$0.001 & 11.28$\pm$0.05 & 0.137$\pm$0.004 & 0.238$\pm$0.012 & 3.CR \\
1645: \enddata
1646: \end{deluxetable*}
1647: %\end{deluxetable}
1648: 
1649: To analyze the carbon-rich spectra, we applied the Manchester 
1650: method introduced by \cite{slo06b} and also applied by 
1651: \cite{zij06} and \cite{lag07}.  These papers examined  a 
1652: total of 28 carbon stars in the LMC and 33 in the SMC, while 
1653: our sample adds five carbon stars in the LMC and one in the 
1654: SMC.  The Manchester method measures the spectral continuum 
1655: at four narrow wavelength intervals relatively free of 
1656: molecular band absorption or dust emission centered at 6.3, 
1657: 9.4, 16.5, and 21.5~\mum.  The first two wavelengths define 
1658: the [6.4]$-$[9.3] color, which measures the contribution from 
1659: warm dust relative to the stellar photosphere.  The 
1660: [16.5]$-$[21.5] color measures the temperature of any cool 
1661: dust contribution to the spectrum.  It also provides a 
1662: means of estimating the contribution under the MgS dust 
1663: emission feature, which peaks in the 26--30~\mum\ range.  
1664: 
1665: Gaseous acetylene (C$_2$H$_2$) produces two absorption
1666: bands visible in the IRS spectra:  a broad band centered at
1667: 7.5~\mum\ and typically $\sim$1.0~\mum\ wide, and a narrow
1668: band at 13.7~\mum\ \citep{mat06}.  This latter band is 
1669: actually the Q branch of the fundamental $\nu_5$ mode and 
1670: it is often centered in a broader band which includes the P 
1671: and R branches, but the broad component can be difficult to 
1672: recognize due to the presence of the SiC dust emission 
1673: feature at 11.3~\mum.
1674: 
1675: The Manchester method fits line segments to the continuum to 
1676: measure the equivalent width of the acetylene bands and the
1677: integrated flux of the SiC emission feature.  We express the 
1678: flux of the SiC band as a ratio to the continuum underneath.  
1679: To measure the strength of the MgS the Manchester method uses 
1680: the [16.5]$-$[21.5] color to extrapolate a cool blackbody 
1681: continuum beneath the feature, which peaks in the 
1682: 26--30~\mum\ range.  Table 8 presents the results.  
1683: 
1684: % Table 8 goes here
1685: 
1686: \subsection{Results} % Sec. 6.2
1687: 
1688: \begin{figure} % Fig. 18
1689: \includegraphics[width=3.5in]{f18.eps}
1690: \caption{The equivalent widths of the acetylene absorption 
1691: bands at 7.5~\mum\ (top) and 13.7~\mum\ (bottom), plotted vs.
1692: [6.4]$-$[9.3] color, which measures the strength of the 
1693: amorphous carbon component.  Galactic data are plotted with
1694: light gray circles, LMC data in dark gray diamonds, and SMC
1695: data in black squares.  The MC\_DUST sample are plotted
1696: with large symbols, while small symbols are from previously
1697: published samples.}
1698: \end{figure}
1699: 
1700: \begin{figure} % Fig. 19
1701: \includegraphics[width=3.5in]{f19.eps}
1702: \caption{The integrated fluxes of the SiC (top) and MgS 
1703: (bottom) dust features, divided by the underlying continuum 
1704: and plotted as a function of [6.4]$-$[9.3] color.  Symbols 
1705: are as defined in Fig.\ 18.}
1706: \end{figure}
1707: 
1708: Figures 18 and 19 plot the results for our sample (large
1709: symbols), along with the LMC sample of \cite{zij06}, the SMC 
1710: samples of \cite{slo06b} and \cite{lag07}, and the Galactic 
1711: sample observed by the SWS on \iso.  The Galactic sample is 
1712: defined by the carbon-rich classifications of \cite{kra02} 
1713: and uses the data in the SWS Atlas \citep{slo03}.  The small 
1714: MC\_DUST sample follows the trends set by the larger samples 
1715: of Magellanic carbon stars previously published.  As 
1716: metallicity decreases, stars with similar [6.3]$-$[9.4] 
1717: colors show increasing absorption from acetylene and reduced 
1718: emission from SiC and MgS dust.
1719: 
1720: % Figures 18 and 19 go here.
1721: 
1722: The strength of the 7.5~\mum\ acetylene band increases as the 
1723: stars go from a [6.4]$-$[9.3] color of 0.0 to $\sim$0.3 and 
1724: then decreases as the amorphous carbon component becomes more
1725: dominant, although the scatter at a given color is large.  
1726: The data appear to follow two separate tracks, but the
1727: significance is not clear with these sample sizes.  The
1728: strength of the 13.7~\mum\ band generally decreases steadily 
1729: from 0.10~\mum\ to $\sim$0.05~\mum\ at a [6.4]$-$[9.3] color 
1730: of 0.5 and remains roughly constant beyond.  In both cases, 
1731: the LMC and SMC sources usually have deeper bands compared to 
1732: the Galactic sources of the same color, but no clear 
1733: difference between the LMC and SMC is discernable.
1734: 
1735: The strength of the SiC feature follows two distinct tracks 
1736: with [6.4]$-$[9.3] color.  The lower track starts with no SiC
1737: feature at a color of 0.0 and slowly increases to a 
1738: $\sim$10\% strength at a color of 1.0.  The upper track climbs 
1739: to a $\sim$30\% strength at a color of $\sim$0.45, then drops 
1740: to meet the lower track at redder colors.  \cite{slo06b} 
1741: noted that five of their carbon stars in the SMC had enhanced 
1742: SiC strength compared to the rest:  MSX SMC 054, 091, 105, 
1743: 159, and 163.  In Figure 19, these five clearly fall on the 
1744: upper track, which includes nearly all of the Galactic 
1745: sources.  \cite{slo06b} were unable to find any other
1746: property that distinguished them from the rest of the SMC
1747: sample, except that they were more likely to show MgS 
1748: emission.  Additional SMC sources on the upper track
1749: include GM 780 from the sample of \cite{lag07} and CV 78
1750: from the MC\_DUST sample.  \cite{lag07} noted that GM 780
1751: was peculiar, with no acetylene absorption and a relatively
1752: high luminosity.  They argued that its C/O ratio was lower
1753: than most of the Magellanic sources and more like Galactic
1754: carbon stars, which might affect the fraction of SiC
1755: in its outflows.
1756: 
1757: The MgS strength in the SMC and LMC samples are similar
1758: to each other.  Few sources bluer than [6.4]$-$[9.3] $\sim$ 
1759: 0.7 having any measurable MgS dust, while most Galactic 
1760: sources do.  At redder colors, the three samples behave 
1761: similarly, with MgS strengths typically $\sim$40\% of the
1762: continuum, but with considerable scatter, especially in the
1763: LMC sample.
1764: 
1765: \subsection{Distance indicators} % Sec. 6.3
1766: 
1767: \begin{figure} % Fig. 20
1768: \includegraphics[width=3.5in]{f20.eps}
1769: \caption{A near-infrared color-magnitude diagram for the
1770: carbon stars in the SMC samples of \cite{slo06b} 
1771: and \cite{lag07}.  The dashed line shows
1772: a line fit to the SMC data while the dotted lines show 
1773: the standard deviation of the scatter about that line.}
1774: %NEED TO ADD LMC DATA AND EXPLAIN WHY THEY DON'T FIT.
1775: \end{figure}
1776: 
1777: \begin{figure} % Fig. 21
1778: \includegraphics[width=3.5in]{f21.eps}
1779: \caption{A mid-infrared color-magnitude diagram for the
1780: Magellanic and Galactic carbon stars examined by
1781: \cite{slo06b}, \cite{zij06}, and \cite{lag07}.  The
1782: solid line shows the running mean of the data as a function
1783: of [6.4]$-$[9.3] color.  The dashed line is a quadratic fit 
1784: to the data, with the dotted lines above and below showing 
1785: the (smoothed) standard deviation.}
1786: \end{figure}
1787: 
1788: Figure 20 plots the absolute magnitude at K vs.\ the J$-$K
1789: 2MASS colors for the carbon stars in the SMC samples of
1790: \cite{slo06b} and \cite{lag07}.  We have
1791: assumed a distance modulus of 18.93 to the SMC \citep{kw06}.
1792: The data follow a linear relation with a standard deviation
1793: of 0.51 magnitudes:
1794: 
1795: \begin{equation}
1796:   M_K = -9.18 + 0.395 (J-K).
1797: \end{equation}
1798: 
1799: % Figure 20 goes here.
1800: 
1801: Applying the synthetic photometry to the spectra to generate
1802: magnitudes at 6.4 and 9.3~\mum\ reveals a second infrared 
1803: color-magnitude relation.  Figure 21 plots the absolute 
1804: magnitude at 9.3~\mum\ as a function of [6.4]$-$[9.3] color 
1805: for the samples of carbon stars in the Galaxy and both 
1806: Magellanic Clouds of \cite{slo06b}, \cite{zij06}, and 
1807: \cite{lag07}.  The magnitudes at 9.3~\mum\ are based on a 
1808: flux density at zero magnitudes of 45.7 Jy and distance 
1809: moduli to the LMC and SMC of 18.54 and 18.93, respectively 
1810: \citep{kw06}.  The Galactic sample were observed with the 
1811: SWS on \iso, as described by \cite{slo06b}.  The distances 
1812: were estimated by \cite{bc05}.  Here, the relation is 
1813: quadratic:
1814: 
1815: \begin{equation}
1816:   M_{9.3~\mu m} = \Sigma a_i ([6.4]-[9.3])^i,
1817: \end{equation}
1818: 
1819: \noindent where $a_0 = -8.81$, $a_1 = -6.56$, and 
1820: $a_2 = 2.74$.  The standard deviation of the data about this 
1821: relation is 0.45 magnitudes.
1822: 
1823: % Figure 21 goes here.
1824: 
1825: 
1826: \section{PAH Emission Sources} % Sec. 7
1827: 
1828: \begin{figure} % Fig. 22
1829: \includegraphics[width=3.5in]{f22.eps}
1830: \caption{Spectra of the two red sources in the MC\_DUST 
1831: sample.  Both show PAH emission features.  A third 
1832: spectrum with PAH features, of Massey SMC 59803, appears in 
1833: Fig.\ 8.}
1834: \end{figure}
1835: 
1836: \begin{figure} % Fig. 23
1837: \includegraphics[width=3.5in]{f23.eps}
1838: \caption{The 8 and 24~\mum\ fields of the three PAH sources
1839: shown in Fig.\ 8 and 22.  The images of Massey SMC 59803 and 
1840: IRAS 04530 are plotted on a logarithmic scale, while HV 5810 
1841: is plotted on a linear scale.  Massey SMC 59803 is an 
1842: isolated supergiant, but the other two sources are associated
1843: with extended emission.}
1844: \end{figure}
1845: 
1846: Figure 22 presents the spectra of the two red sources in the
1847: MC\_DUST sample, IRAS 04530 and HV 5810.  Both of these 
1848: spectra have PAH emission features.  A third source, the
1849: supergiant Massey SMC 59803, also shows PAH emission 
1850: superimposed on an oxygen-rich dust spectrum.  Its spectrum
1851: appears in Figure 8.
1852: 
1853: % Figure 22 goes here.
1854: 
1855: Figure 23 includes 8 and 24~\mum\ images of the three PAH
1856: emission sources.  The data for Massey SMC 59803 were 
1857: obtained as part of the S$^3$MC survey by \cite{bol07}.  
1858: Data for the other two sources are from the SAGE survey 
1859: \citep{mei07}.  
1860: 
1861: % Figure 23 goes here.
1862: 
1863: \subsection{Massey SMC 59803} % Sec. 7.1
1864: 
1865: \begin{figure} % Fig. 24
1866: \includegraphics[width=3.5in]{f24.eps}
1867: \caption{The spectrum of Massey SMC 59803 with a fitted
1868: spline ({\it top}), and the residual PAH spectrum ({\it 
1869: bottom}).  The dashed vertical lines in the bottom panel
1870: are at wavelengths of 6.23, 7.70, 8.60, 11.26, an 
1871: 12.70~\mum\, the centers of the PAH features in a typical
1872: class A PAH spectrum \citep[in the classification system
1873: introduced by][]{pee02}.}
1874: \end{figure}
1875: 
1876: An examination of the 8 and 24~\mum\ images of Massey SMC
1877: 59803 reveals no associated extended emission.  The IRS data 
1878: are also consistent with a point source at all wavelengths.  
1879: To separate the contribution of PAHs from the spectrum, 
1880: we fit a spline through the ``continuum'' from the star and 
1881: amorphous silicate grains which produce the 10~\mum\ feature, 
1882: as illustrated in Figure 24 (top).  \cite{kel08} applied this 
1883: method to a sample of spectra from Herbig Ae/Be stars with 
1884: both silicate and PAH emission and successfully disentangled 
1885: the separate dust components in most cases.  
1886: 
1887: % Figure 24 goes here.
1888: 
1889: Figure 24 (bottom) shows the resulting PAH spectrum.  PAHs in 
1890: the spectra of supergiants are unusual, but not unheard of.  
1891: \cite{syl98} observed PAH emission at 8.6 and 11.3~\mum\ in 
1892: three supergiants in the h and $\chi$ Persei association.
1893: The residual PAH spectrum from Massey SMC 59803 shows 
1894: emission features centered at 6.21, 6.70, 8.13, 11.29, 12.86, 
1895: and 13.56~\mum.  The 6.7~\mum\ feature probably does not 
1896: arise from PAHs and is presently unidentified.  The remaining 
1897: features are from PAHs.  The strongest features are the C--C 
1898: complex centered at 8~\mum\ and the out-of-plane C--H solo 
1899: bending mode at 11.3~\mum.  
1900: 
1901: \cite{pee02} defined three classes of PAH spectra based 
1902: primarily on the position of the 8~\mum\ PAH complex.  Most 
1903: spectra fell into class A (peak at 7.65~\mum) or class B
1904: (7.85~\mum), but two class C spectra showed a feature 
1905: peaking at $\sim$8.2~\mum.  The detailed position and shape 
1906: of the 8~\mum\ complex depends critically on how the spline 
1907: fits the continuum on the blue side of the silicate feature,
1908: so we should not overinterpret this feature.  Nonetheless,
1909: there does appear to be an excess between $\sim$7.5 and 
1910: 9~\mum, and it is more consistent with class C than with
1911: the other classes.  The 11.3~\mum\ feature is also shifted
1912: from its nominal position in class A and B spectra of 
1913: 11.26~\mum\ \citep{slo07} to 11.29~\mum.
1914: 
1915: \cite{slo07} found that class C PAH spectra are associated 
1916: with cooler radiation fields.  Massey SMC 59803 has an 
1917: estimated effective temperature of 4100 K \citep{lev06}, and 
1918: its PAH spectrum, while noisy, is consistent with this 
1919: temperature.
1920: 
1921: \subsection{HV 5810} % Sec. 7.2
1922: 
1923: As discussed in \S 4.2 and illustrated in Figure 6, the 
1924: spectrum of HV 5810 shows emission from several C$-$H bending 
1925: modes in PAHs.  These include the in-plane bending mode at 
1926: 8.55 and the out-of-plane bending modes at 11.27, 12.78, and 
1927: 13.5~\mum\ in SL \citep[e.g.][]{atb89,hon01}.  The emission 
1928: at 12.78~\mum\ probably contains a contribution from [Ne II] 
1929: emission as well.  The LL portion of the spectrum in 
1930: Figure 24 reveals an emission complex in LL with peaks at 
1931: 16.5 and 17.35~\mum, which also arises from PAHs 
1932: \citep{mou00,wer04b}.
1933: 
1934: Examination of the spectral images reveals that the emission 
1935: features at 11.3 and 12.8~\mum\ extend the length of the SL 
1936: slit.  The emission at 8.6~\mum\ is also extended.  In the LL 
1937: images, the 16--18~\mum\ PAH complex is extended over about 
1938: $\sim$20$\arcsec$, while several forbidden emission lines are 
1939: extended over much larger areas:  [\ion{Ne}{3}] at 15.6~\mum, 
1940: [\ion{S}{3}] at 18.7 and 33.5~\mum, and [\ion{Si}{2}] at 
1941: 34.8~\mum.  The forbidden lines are commonly seen in many 
1942: images of other targets, but they usually subtract out when 
1943: the background is removed.  At longer wavelengths, the 
1944: spatial structure of HV 5810 grows more complex and extended 
1945: as well.
1946: 
1947: Figure 23 shows that HV 5810 is embedded in a complex
1948: region of extended emission.  We conclude that we have 
1949: observed a naked star, HV 5810, superimposed on a complicated
1950: background.  %It is possible that the 16--18~\mum\ PAH complex 
1951: %might be more localized to the immediate vicinity of HV 5810.
1952: 
1953: \subsection{IRAS 04530$-$6916} % Sec. 7.3
1954: 
1955: \begin{figure} % Fig. 25
1956: \includegraphics[width=3.5in]{f25.eps}
1957: \caption{The SL spectrum of IRAS 04530, scaled to match the
1958: \iras\ photometry.  In the top panel, the spline points are
1959: marked with diamonds, and the fitted spline with a smooth
1960: thin curve.  The spline-subtracted spectrum appears in the
1961: bottom panel, showing both silicate absorption and PAH
1962: emission.  The vertical dotted lines are at wavelengths of
1963: 6.23, 7.70, 8.60, 11.26, and 12.70~\mum.}
1964: \end{figure}
1965: 
1966: The observations of IRAS 04530 were partially mispointed in
1967: both SL and LL, requiring some special attention to complete
1968: the processing.  The source was mostly out of the slit in
1969: one of the SL nod positions, and in LL, we only observed
1970: a portion of the source, which is extended.  To stitch the
1971: spectral segments together, we normalized them to the 12 and
1972: 25~\mum\ fluxes from the \iras\ Faint Source Catalog (2.01
1973: and 4.97 Jy, respectively).  Because of the spectral tilts
1974: in mispointed spectra, this leaves a discontinuity at the
1975: SL/LL interface.  The overall continuum properties of the
1976: resulting spectrum should be considered provisional.
1977: Despite the limitations of the spectrum, the presence of
1978: various spectral features and their wavelengths are reliable.
1979: 
1980: The spectrum shows a combination of silicate absorption
1981: and emission from PAHs.  Although the IRS does not cover the 
1982: 35--45 \mum\ wavelength range that the Hanscom system used to 
1983: distinguish the reddest sources, the \iras\ data show that 
1984: the spectrum continues to rise between 25 and 60 \mum, making
1985: this spectrum class 5.U/SA \citep[this class was
1986: introduced by][]{ekp04}.
1987: 
1988: The combination of both PAH emission and silicate absorption 
1989: complicates the analysis of the PAH strengths, but it does
1990: not have a significant effect on their positions.  To examine
1991: the PAH features more closely, we fit a spline through the
1992: most likely continuum, as Figure 25 illustrates.  The
1993: resulting PAH spectrum has features peaking at 6.22, 7.65,
1994: 8.62, 11.27, and 12.78~\mum.  With the exception of the last
1995: feature, which may be affected by [\ion{Ne}{2}] emission, the
1996: wavelengths for these features correspond to a class A PAH
1997: spectrum in the scheme of \cite{pee02}.  This class of
1998: PAH spectrum is commonly seen in star formation regions.
1999: 
2000: % Figure 25 goes here.
2001: 
2002: IRAS 04530 has usually been treated as an evolved star. 
2003: \cite{woo92} measured its bolometric magnitude to be $-$7.62 
2004: and concluded that it was probably a supergiant due to its 
2005: high luminosity, low pulsation amplitude, and lack of maser 
2006: activity.  \cite{vl01} seconded this opinion, and on the 
2007: basis of these works it was included in this study.  
2008: \cite{vl05a} recently suggested that it may instead be a YSO. 
2009: Their optical spectrum shows strong hydrogen emission lines 
2010: as well as the [\ion{S}{3}] doublet and [\ion{Ca}{2}] triplet. 
2011: \cite{hod04} and \cite{ekp04} found that the U/SA 
2012: classification in the \iso\ PHOT-S and CAM-CVF databases 
2013: occurred almost exclusively in YSOs, star forming regions, 
2014: and starburst galaxies (the single exception being C* 1662).
2015: 
2016: The SAGE images at 8~\mum\ and 24~\mum\ in Figure 23
2017: reveal that IRAS 04530 lies at the edge of an extensive 
2018: region of 8~\mum\ emission probably associated with the 
2019: DEM L15 H II region.  The IRAC and MIPS photometry of 
2020: IRAS 04530 \citep[from the 2006 December release of the 
2021: SAGE point source catalog][]{mei06} and the 2MASS photometry 
2022: help clarify its evolutionary status.  IRAS 04530
2023: does not lie near the AGB population in the color-color
2024: and color-magnitude diagrams published from Galactic and 
2025: Magellanic Spitzer programs \citep{whi03a,whi03b,blu06,bol07}.
2026: 
2027: In summary, evidence from the new {\em Spitzer} observations, 
2028: both spectroscopic and photometric, support the suggestion of 
2029: \cite{vl05a} that IRAS 04530 is a YSO, not an evolved star.
2030: 
2031: 
2032: \section{Dependencies on Metallicity} % Sec. 8
2033: 
2034: The previous sections concentrated on the analysis 
2035: applied to the different classes of spectra and the results 
2036: for individual sources.  This section concentrates on the 
2037: samples as a whole, comparing the SMC and LMC samples to the 
2038: Galactic samples published elsewhere.  The dominant difference 
2039: between these samples is metallicity.
2040: 
2041: \subsection{Fraction of naked stars} % Sec. 8.1.
2042: 
2043: \begin{deluxetable}{llll} % Table 9
2044: \tablenum{9}
2045: \tablecolumns{4}
2046: \tablewidth{0pt}
2047: \tablecaption{Fraction of naked oxygen-rich stars\label{Tbl8}}
2048: \tablehead{ 
2049:   \colhead{ } & \multicolumn{3}{c}{Period (days)} \\
2050:   \colhead{Sample} & \colhead{$\le$250} & \colhead{250--700} & \colhead{$>$ 700}
2051: }
2052: \startdata
2053: Galaxy & 2 of 106 & 5 of 269 & 0 of 11 \\
2054: LMC    & 2 of 3   & 2 of 9   & 0 of 8 \\
2055: SMC    & 1 of 1   & 6 of 7   & 0 of 1 \\
2056: \enddata
2057: \end{deluxetable}
2058: 
2059: Our sources were selected using criteria similar to those for
2060: the Galactic samples of AGB stars examined by \cite{sp95} and 
2061: supergiants studied by \cite{sp98}.  The Galactic samples were 
2062: chosen by searching for optically identified variables of 
2063: classes associated with either the AGB or supergiants in the 
2064: LRS Atlas \citep{lrs86}.  \cite{sp95} found that 43\% of 
2065: irregular (Lb) variables, 19\% of semi-regular (SRb) 
2066: variables, and only 1\% of Mira variables (including SRa 
2067: variables) were naked, while \cite{sp98} found that 13 of 65, 
2068: or 20\%, of the supergiants were naked.  
2069: 
2070: The MC\_DUST sample includes five supergiants, all of which 
2071: are associated with circumstellar dust, but the situation is 
2072: different for the AGB.  In the LMC, three of the 19 oxygen-rich
2073: sources, or 16\%, are naked (we are excluding the naked
2074: star J052832, about which little is known), while in the SMC, 
2075: seven of nine oxygen-rich AGB sources, or 78\%, are naked.
2076: 
2077: We know the pulsation periods of enough of the various 
2078: samples to compare how the fraction of naked stars varies 
2079: with period in the Galaxy, LMC, and SMC.  Because the number 
2080: of naked stars is small, we divide the periods into only 
2081: three bins separated at 250 d, the approximate boundary 
2082: between most of the short-period semi-regulars and Miras, and 
2083: 700 d, the approximate period limit for Miras in the Galactic 
2084: sample.  For the Galactic sample, we use the catalog 
2085: published by \cite{sp98}.  Of the 611 oxygen-rich AGB stars 
2086: and supergiants in that catalog, 386 have periods, and only 
2087: seven of those are naked.  Two of the naked stars have 
2088: periods less than 250 days, and the other five fall in the 
2089: 250--700-day bin, as Table 9 shows.  If we limited the sample 
2090: to just Miras, the results are similar, with only two naked 
2091: Miras out of 229.
2092: 
2093: % Table 9 goes here.
2094: 
2095: Table 9 reveals that in the two lower-period bins, the
2096: percentage of naked stars increases as the metallicity of
2097: the sample decreases.  None of the stars
2098: with periods greater than 700 days are naked.  In the SMC,
2099: the only variable with a period less than 700 days and a
2100: dust excess is BFM 1, but even its dust shell is thin, and
2101: the dust spectrum is unusual.
2102: 
2103: \subsection{Dust Emission Contrast} % Sec. 8.2
2104: 
2105: \begin{figure} % Fig. 26 
2106: \includegraphics[width=3.5in]{f26.eps}
2107: \caption{Dust emission contrast among oxygen-rich souces as a
2108: function of period, for all variables with known periods in 
2109: the Galactic LRS sample \citep{sp95,sp98} and the MC\_DUST 
2110: sample.  Dust emission contrast is defined in \S 3.4.  In the 
2111: Galactic sample, Miras and the closely 
2112: related SRa variables are plotted as filled circles, while 
2113: the SRb and SRc variables are plotted as open circles.  LMC 
2114: sources appear as dark gray diamonds, and SMC sources as 
2115: black squares.  The four symbols outlined in black mark 
2116: suspected massive stars.  The two points with periods 
2117: $\sim$800 days are supergiants.  The two in the upper right 
2118: are massive AGB sources (\S 3.3).  The horizontal line at 
2119: DEC = 0.10 separates naked and dusty sources in the 
2120: Magellanic samples.}
2121: \end{figure}
2122: 
2123: \begin{deluxetable}{llll} % Table 10
2124: \tablenum{10}
2125: \tablecolumns{4}
2126: \tablewidth{0pt}
2127: \tablecaption{Dust emission contrasts\label{Tbl9}}
2128: \tablehead{ 
2129:   \colhead{ } & \multicolumn{3}{c}{Period (days)} \\
2130:   \colhead{Sample} & \colhead{$\le$250} & \colhead{250--700} & \colhead{$>$ 700}
2131: }
2132: \startdata
2133: Galaxy & 0.38$\pm$0.27    & 0.61$\pm$0.41 & 1.24$\pm$0.61 \\
2134: LMC    & 0.06$\pm$0.11    & 0.44$\pm$0.30 & 2.29$\pm$0.92 \\
2135: SMC    & 0.06$\pm$\nodata & 0.07$\pm$0.07 & 0.80$\pm$\nodata \\
2136: \enddata
2137: \end{deluxetable}
2138: 
2139: As part of the classification process in \S 5.1, we also 
2140: measured the dust emission contrast for each oxygen-rich 
2141: source, as defined for the naked stars in our sample.  
2142: Figure 26 plots the dust emission contrast as a function of 
2143: period in the Magellanic and Galactic samples for the naked 
2144: and oxygen-rich sources.  For periods between 250 and 700 
2145: days, the range dominated by Mira variables, the Galactic 
2146: sample shows a general trend of increasing dust emission 
2147: contrast with increasing period, but with a substantial 
2148: spread.  For periods less than 700 days, the Magellanic 
2149: sources show significantly less dust emission than their 
2150: Galactic counterparts.  This result is consistent with the 
2151: low SiO molecular band strength detected in naked oxygen-rich 
2152: stars in the Magellanic Clouds, since SiO is the basic 
2153: building block of silicate dust.
2154: 
2155: % Figure 26 goes here.
2156: 
2157: Magellanic sources with periods greater than 700 days 
2158: typically show large contributions from dust in their 
2159: spectra.  In particular, five sources lie on or appear to 
2160: continue the trend of DEC vs.\ period established by the 
2161: Galactic sample.  Two of these sources, HV 888 and HV 996,
2162: are definite supergiants (at DEC$\sim$1.9 and P$\sim$800 d).
2163: Two more, IRAS 04509 and IRAS 04512, are probably very 
2164: luminous AGB sources as discussed in \S 3.3.
2165: We believe these four sources are more massive than typical
2166: AGB sources and therefore younger.  Since they formed more
2167: recently, they are probably more metal rich than typical
2168: Magellanic sources, and they would be expected to mimic the
2169: more metal-rich Galactic sample.
2170: 
2171: Figure 26 includes three more LMC sources with very long
2172: pulsation periods to the lower right.  From top to bottom,
2173: these are IRAS 05402 (DEC$\sim$2.3), IRAS 04545 
2174: (DEC$\sim$1.5), and IRAS 05329 (DEC$\sim$1.1).  All three
2175: of these sources show silicate absorption at 10~\mum, so
2176: our DEC measurement, which is a simple sum, underestimates 
2177: the total silicate contribution.  Correcting for this 
2178: simplification would undoubtedly push these points upward in 
2179: Figure 26.  However, all three spectra show clear evidence for 
2180: crystalline silicate emission at longer wavelengths, which
2181: as described in \S 5.2, resemble the spectra from dusty
2182: Galactic objects with disks.  This semblance is far from
2183: direct evidence of a disk, but it might indicate that 
2184: presence of a reservoir of dust in addition to the dust 
2185: forming in the outflows from these stars at the current time.
2186: 
2187: Table 10 presents the mean dust emission contrasts for the 
2188: same period ranges used for Table 9.  For the stars with 
2189: periods $\le$700 days, the amount of dust drops from the 
2190: Galaxy to the LMC and drops further in the SMC.  For these 
2191: stars, the amount of dust clearly depends on metallicity.  
2192: As described above, the samples with periods $>$ 700 days 
2193: probably include young massive stars which blur the 
2194: dependence on metallicity seen at shorter periods.
2195: 
2196: % Table 10 goes here.
2197: 
2198: \subsection{SiO band strengths} % Sec. 8.3.
2199: 
2200: \begin{deluxetable}{lr} % Table 11
2201: \tablenum{11}
2202: \tablecolumns{2}
2203: \tablewidth{0pt}
2204: \tablecaption{Comparative SiO band strengths\label{Tbl10}}
2205: \tablehead{
2206:   \colhead{Sample} & \colhead{W (\mum)} 
2207: }
2208: \startdata
2209: SWS M giants & 0.35$\pm$0.11 \\
2210: IRS M giants & 0.28$\pm$0.06 \\
2211: LMC          & 0.26$\pm$0.11 \\
2212: SMC          & 0.16$\pm$0.08 \\
2213: \enddata
2214: \end{deluxetable}
2215: 
2216: \cite{mat05} obtained 3--4~\mum\ spectra of six oxygen-rich 
2217: evolved stars in the LMC, and they detected possible weak SiO 
2218: absorption in only one source, IRAS 05218, which had been 
2219: thought to be a carbon star \citep{tra99}.  They explained 
2220: the absence of this band by noting that silicon is not 
2221: produced in AGB stars, so its abundance in the photosphere 
2222: reflects the initial abundance of the star, which will be low 
2223: in a metal-poor system like the LMC.  But at least four of 
2224: their six targets have circumstellar dust shells, which make 
2225: it possible that the dust partially masks the SiO bands.
2226: 
2227: In the MC\_DUST sample, the SiO fundamental band is clearly
2228: present in the spectra of several of the less obscured 
2229: sources.  Table 11 compares the mean equivalent widths of the
2230: total LMC and SMC samples to two Galactic samples, one 
2231: based on SWS observations and a second based on IRS data.
2232: For the SMC and LMC samples, we use only those equivalent 
2233: widths in Table 6 with a S/N of $\sim$1 or better.  For 
2234: sources with only an upper limit, we use it in place of the 
2235: measured width.  For HV 2575 (first observation), we take the 
2236: upper limit to be 0.26~\mum.  The follow-up spectra are 
2237: treated as separate data.  MACHO 79.5505.26 is not included
2238: since it may be carbon rich.
2239: 
2240: % Table 11 goes here.
2241: 
2242: Using {\it ISO}/SWS data, \cite{her02} determined that the 
2243: strength of the SiO fundamental band increases from an
2244: equivalent width of $\sim$0.05~\mum\ in K0 giants to 
2245: $\sim$0.18~\mum\ at K5.  No M giant had an equivalent width 
2246: less than $\sim$0.15, while the maximum value increased past 
2247: 0.25~\mum\ at M5.  They measured the SiO equivalent width
2248: between 7.6 and 9.0~\mum, but the bandhead is at 7.4~\mum,
2249: and the band continues to 11.0~\mum\ in some cases.  We have
2250: remeasured the SiO fundamental band in their sample of 23
2251: M giants using a similar method to that applied to the IRS
2252: data, except that we integrate the equivalent width directly
2253: instead of fitting a template absorption band.  We shifted 
2254: the long-wavelength range for fitting the Planck function 
2255: from 11.5--13.0~\mum\ to 10.8--11.4~\mum\ to avoid hysteresis 
2256: effects in the SWS data, which tend to drive the flux up 
2257: slightly toward 12~\mum.  For similar reasons, we shifted the 
2258: long-wavelength boundary for integrating the equivalent width 
2259: from 11.5 to 11.1~\mum.  These shifts slightly reduce the 
2260: equivalent widths, but the Galactic sample still has stronger 
2261: SiO absorption than the Magellanic samples.  The mean SiO 
2262: equivalent width shows no strong dependency on optical 
2263: spectral type.
2264: 
2265: The second comparison sample of M giants comes from the
2266: ongoing IRS GTO Cycle 4 program used to generate the SiO 
2267: template described in \S 4.1.  At the time of this writing,
2268: spectra of 13 of the 20 planned M giants have been obtained 
2269: and reduced.  We extracted the SiO equivalent width using 
2270: wavelengths identical to those used for the Magellanic 
2271: samples, but as with the SWS data, we simply integrated the 
2272: feature, instead of fitting a template.  As with the SWS
2273: sample, the equivalent width does not depend strongly on 
2274: spectral type.  The difference between the two Galactic M 
2275: giant samples may result from different selection criteria,
2276: but it is within the standard deviations.
2277: 
2278: Table 11 shows a general trend toward decreasing SiO
2279: band strength with decreasing metallicity, although there
2280: is some overlap in the samples.  The LMC sample overlaps
2281: the IRS M giants.  The wide spread in the LMC sample may 
2282: indicate a range of metallicities, but the sample is not 
2283: large.  The SMC clearly shows weaker SiO bands than the 
2284: Galactic giants.  These results confirm the 
2285: metallicity-dependent trend first recognized by 
2286: \cite{mat05} in the 3~\mum\ spectral region.
2287: 
2288: \subsection{Oxygen-rich dust composition} % Sec. 8.4
2289: 
2290: \begin{figure} % Fig. 27 
2291: \includegraphics[width=3.5in]{f27.eps}
2292: \caption{The distribution of the MC\_DUST sample and
2293: Galactic comparison samples with silicate emission index.  In 
2294: the top panel, LMC sources are shaded with diagonal stripes, 
2295: while the two SMC sources are cross-hatched (at SE4 and SE8).
2296: Alumina-rich dust produces the SE1--3 spectra and is largely
2297: absent from the Magellanic samples.  The dust distribution in
2298: the Magellanic sample closely resembles the Galactic 
2299: supergiants, even though only about three of the 14 
2300: Magellanic sources plotted in the upper panel are actually 
2301: supergiants.}
2302: \end{figure}
2303: 
2304: \cite{es01} demonstrated with radiative transfer models that
2305: the amorphous alumina grains produce the broad, low-contrast
2306: features peaking $\sim$11--12~\mum, which correspond to lower
2307: SE indices (SE1--3), while amorphous silicate grains produce
2308: the classic 10~\mum\ silicate feature (and associated 
2309: 18~\mum\ feature) which corresponds to higher SE indices 
2310: (SE6--8).  The middle portion of the sequence (SE4--6) 
2311: generally shows structured emission features, with a peak at 
2312: 10~\mum\ and a shoulder at 11~\mum.  This structure could 
2313: arise from multiple causes, including mixtures of amorphous 
2314: alumina and silicates, self-absorbed silicates, or possibly, 
2315: crystalline silicate grains.  Given that alumina and 
2316: silicates both contribute to the spectra, it might be more
2317: appropriate to refer to the sequence as the oxygen-rich dust 
2318: sequence.
2319: %\cite{kra02} introduced a modified classification 
2320: %scheme, grouping the broad, low-contrast alumina features 
2321: %(SE1--3) as SEa, the structured silicate spectra (SE4--6) as 
2322: %SEb, and the classic silicate features (SE6--8) as SEc.
2323: 
2324: Figure 27 compares the distribution of silicate emission 
2325: sources in the Magellanic Clouds to the Galactic samples 
2326: examined by \cite{sp95,sp98}.  The Magellanic sample differs 
2327: markedly from the Galactic AGB sample by having only two
2328: sources from SE1 to SE4.  In fact, it looks more like the 
2329: Galactic supergiant sample than anything else, which is 
2330: somewhat surpising given that the sample displayed in 
2331: Figure 27 includes only three supergiants.  We had 
2332: intentionally observed ten stars classified as S or MS, 
2333: which we had expected to resemble the Galactic S and MS 
2334: stars, but most of these proved to be naked instead.  Of 
2335: the five S and MS stars in the LMC, three (WBP 104, 
2336: HV 2575, and HV 12620) are naked, while HV 12070 is an 
2337: SE6 and WBP 77 accounts for the lone SE1 source.  Of the 
2338: five in the SMC, only BFM 1 shows any sign of dust, but as 
2339: discussed in \S 5.6., it is very unusual dust.
2340: 
2341: The lack of emission from amorphous alumina dust in the 
2342: Magellanic Clouds is consistent with how the abundance of Al 
2343: depends on metallicity compared to other dust-producing 
2344: elements like Mg and Si.  The latter elements are both 
2345: $\alpha$-capture products, and while their abundance is
2346: lower in metal-poor stars, Al is even less abundant
2347: \citep[see][and references therein]{whe89}. 
2348: 
2349: The apparent absence of a 13~\mum\ feature in our sample
2350: provides further evidence of the relative lack of Al in the 
2351: Magellanic Clouds.  The carrier of the 13~\mum\ feature has 
2352: been the subject of some debate, with the focus on crystalline 
2353: alumina \citep[corundum, Al$_2$O$_3$][]{gla94,slo96} or spinel 
2354: \citep[MgAl$_2$O$_4$][]{po99,fab01a}.  \cite{slo03} and
2355: \cite{dep06} have provided strong evidence against spinel 
2356: as the carrier, but here, it is sufficient to note that in
2357: both proposed carriers, the Al--O bond is responsible for
2358: the 13~\mum\ feature.  Consequently, the lack of Al 
2359: in the Magellanic Clouds fully explains its absence.
2360: 
2361: The presence of the 14~\mum\ feature in our Magellanic
2362: sample provides some clues about its carrier.
2363: \cite{pit06} recently noted that the 14~\mum\ feature in 
2364: HV 2310 could be reproduced by AlN dust or possibly by 
2365: non-spherical corundum grains.  While the non-spherical 
2366: grain models discussed by \cite{dep06} do shift the emission 
2367: peak from 13~\mum\ to longer wavelengths, the features become 
2368: quite broad compared to the observed 14~\mum\ features.  In 
2369: addition, the low abundance of Al in the Magellanic Clouds
2370: make any Al-bearing dust species an unlikely carrier.
2371: The presence of the 14~\mum\ feature appears to be limited to 
2372: strong silicate emission sources, which suggests that its 
2373: carrier is probably related to silicates in some way.
2374: 
2375: % Figure 27 goes here.
2376: 
2377: % Our sample selected against faint oxygen-rich sources,
2378: % since these were less likely to be identified as Harvard
2379: % Variables and less likely to have been included in the
2380: % near-infrared follow-up surveys which formed the basis
2381: % for our sample.  
2382: 
2383: \subsection{Carbon mass-loss rates} % Sec. 8.5
2384: 
2385: \begin{figure} % Fig. 28
2386: \includegraphics[width=3.5in]{f28.eps}
2387: \caption{Dust mass-loss rate as a function of [6.4]$-$[9.3]
2388: color, including data from the samples of Magellanic carbon 
2389: stars presented by \cite{slo06b,zij06} and \cite{lag07} and 
2390: analyzed with radiative transfer models by \cite{gro07}.  We 
2391: have converted the total mass-loss rates to dust mass-loss 
2392: rates by dividing by their assumed gas-to-dust ratio of 200.  
2393: The solid line is a linear fit to the data.}
2394: \end{figure}
2395: 
2396: \begin{figure} % Fig. 29
2397: \includegraphics[width=3.5in]{f29.eps}
2398: \caption{[6.4]$-$[9.3] color as a function of period.  
2399: The data include the samples from Fig. 22 (small diamonds
2400: and squares), the MC\_DUST sample (large diamonds and
2401: squares), and the Galactic sample (small circles).
2402: The Galaxy, SMC, and LMC show no significant differences
2403: despite their different metallicities.}
2404: \end{figure}
2405: 
2406: Grains composed primarily of carbon atoms, usually described
2407: as amorphous carbon, dominate the the dust produced by carbon 
2408: stars \citep[e.g.][]{mr87,se95}, but amorphous carbon has no 
2409: easily discernable spectral features, making a direct 
2410: measurement of its contribution difficult.  The Manchester 
2411: method substitutes the [6.4]$-$[9.3] color, which will grow
2412: increasingly red as the contribution of amorphous carbon
2413: relative to the star increases.  \cite{gro07} applied 
2414: radiative transfer models to all of the Magellanic carbon 
2415: stars in the previously cited papers, and they found a good 
2416: correlation between total mass-loss rate and [6.4]$-$[9.3]
2417: color (see their Fig. 7).  What the radiative transfer models 
2418: actually probe, however, is the {\it dust} mass-loss rate.  
2419: The total mass-loss rate is simply inferred from the 
2420: gas-to-dust ratio, which \cite{gro07} assumed to be 200 in
2421: both the LMC and SMC.  We convert their mass-loss rates
2422: to dust mass-loss rates (by dividing by 200) and repeat their
2423: comparison to [6.4]$-$[9.3] color in Figure 28.  Both the LMC 
2424: and SMC follow the same relation.  Fitting a line to the data 
2425: gives log (mass-loss rate) $= -9.8 + 1.6 $ ([6.4]$-$[9.3]).  
2426: The scatter of the data about this line has a standard 
2427: deviation of 16\%.
2428: 
2429: % Figure 28 goes here.
2430: 
2431: This and previous studies have compared objects in the 
2432: Galaxy, SMC, and LMC at similar [6.4]$-$[9.3] color as though
2433: objects with similar amounts of amorphous carbon dust are 
2434: truly alike.  They might not be if the dust formation 
2435: efficiency depended on metallicity.  To examine this 
2436: possibility, we examined the pulsation periods of the sources 
2437: in the various samples, since those should be similar for 
2438: physically similar stars.  Figure 29 plots [6.4]$-$[9.3]
2439: color as a function of pulsation period, and it reveals no
2440: significant differences in the distribution of colors at a 
2441: given period between the samples.  In other words, carbon 
2442: stars with similar pulsation properties are embedded in 
2443: similar quantities of dust, regardless of their initial 
2444: metallicity.
2445: 
2446: % Figure 29 goes here.
2447: 
2448: 
2449: \section{Discussion and Conclusions} % Sec. 9
2450: 
2451: The similar dust production rates among the samples in
2452: Figure 29 have important consequences.  The production of 
2453: amorphous carbon is limited by the amount of excess carbon 
2454: compared to oxygen, but carbon is produced in the 
2455: triple-$\alpha$ process in AGB stars.  The production of 
2456: silicates and other oxygen-rich grains, on the other hand, 
2457: is limited by the abundances of heavy elements such as Si 
2458: and Al, and these are not produced in significant quantities 
2459: on the AGB.  The ready explanation for the results 
2460: presented in Figure 29 is that the carbon produced and 
2461: dredged up by AGB stars dominates the carbon with which they 
2462: formed.  It should follow, then, that even Population III 
2463: stars, if they are in the mass range to become carbon stars, 
2464: will produce approximately the same amount of carbon-rich 
2465: dust as their Population I counterparts.
2466: 
2467: \cite{bbm78,bbm80} discovered that in the SMC, more AGB
2468: stars became carbon rich than in the LMC \citep[see 
2469: also][]{ch03}.  \cite{rv81} explained this phenomenon as a 
2470: metallicity effect.  Stars with solar abundances will become 
2471: carbon stars if their initial mass $\ga$1.7~M$_{\sun}$, but
2472: at 1/5 solar metallicity, the mass limit has dropped to 
2473: 1.2~M$_{\sun}$, primarily because of the lack of oxygen in
2474: the atmosphere of the metal-poor star to balance the dredged
2475: up carbon.  Assuming a Salpeter initial mass function 
2476: \citep{sal55} and integrating from these carbon limits to
2477: 8~M${_\sun}$, we find that in the SMC, 70\% more of the
2478: AGB should be carbon-rich compared to a sample with solar
2479: metallicity.  Since we see no diminishment of carbon-rich 
2480: dust production at decreasing metallicity, we conclude that 
2481: the more metal-poor an initial population of stars, the more 
2482: carbon-rich dust it will contribute to the ISM, simply 
2483: because it produces more carbon stars.  
2484: 
2485: Many other properties of carbon stars do depend on 
2486: metallicity.  Our small sample reinforces results from
2487: previous studies of {\it Spitzer} spectroscopy in the 
2488: Magellanic Clouds, which show that as the metallicity
2489: decreases, emission from SiC and MgS dust decrease, 
2490: while absorption from acetylene gas increases.  The
2491: increased acetylene absorption probably arises from 
2492: higher C/O ratios in Magellanic carbon stars 
2493: \citep{vl99,mat02,mat05}, which in turn arise from a 
2494: combination of increased dredge-up efficiency \citep{woo81} 
2495: and less initial oxygen in the atmosphere \citep{vw93}.
2496: The additional acetylene we observe may add even more to
2497: the amount of carbon injected into the ISM by metal-poor
2498: populations.
2499: 
2500: The behavior of the oxygen-rich dust stands in stark contrast
2501: to the carbon-rich dust.  While the quantity of carbon-rich
2502: dust shows no discernable dependence on metallicity, 
2503: oxygen-rich dust shows many clear trends.  As metallicity
2504: decreases, (1) the fraction of naked stars increases, (2)
2505: SiO absorption decreases, (3) the silicate dust emission also 
2506: decreases, and (4) alumina-rich dust emission disappears 
2507: almost entirely.  The lower metallicity of the Magellanic
2508: Clouds provides a ready explanation for all of these effects.
2509: Si is not produced on the AGB, so its abundance scales with 
2510: overall metallicity.  As a result, we observe less SiO gas,
2511: which is the fundamental building block of silicate dust, 
2512: and less silicate dust as well.  The abundance of Al scales
2513: even more strongly with metallicity (\S 5.4.), so the
2514: lack of alumina dust should not be a surprise.  Not only 
2515: do we see few spectra in the SE1--3 range, but we see little
2516: or no emission from the 13~\mum\ feature, which arises from
2517: an Al--O bond.
2518: 
2519: Standard condensation sequences propose the formation of
2520: alumina-rich grains before silicates 
2521: \citep[e.g.][]{tie90,ste90}.  The lack of alumina dust in 
2522: our sample suggests the formation of alumina is not a
2523: prerequisite to the formation of silicates.  In other words,
2524: silicates can form without pre-existing alumina grains to 
2525: serve as seeds.
2526: 
2527: The oxygen-rich sample contains a wealth of interesting dust 
2528: features.  Five sources show crystalline silicate features 
2529: at 10~\mum\ or longer wavelengths, and these are all 
2530: consistent with Mg-rich pyroxene and olivine.  There is no 
2531: evidence for Fe in the silicates, much like the Galactic 
2532: sample.  Several sources have an unidentified 14~\mum\ 
2533: emission feature, and one source, the only true S star in 
2534: our sample, shows a previously unknown 13.5~\mum\ emission 
2535: feature.  These two features are newly discovered by the
2536: IRS, and their identification awaits further investigation.
2537: 
2538: The Magellanic Clouds present us with a scenario of steady or 
2539: increasing carbon-rich dust production as systems grow more 
2540: metal-poor, compared to declining rates of oxygen-rich dust 
2541: production.  This result measures what goes into the ISM, but 
2542: it contradicts observations of what is currently present in 
2543: the ISM in metal-poor systems.  Most lines of sight in the 
2544: SMC have unusual extinction curves without the commonly 
2545: observed 2175~\AA\ bump seen in the Galaxy, in common with 
2546: starburst galaxies and quasars \citep[e.g.][and references 
2547: therein]{gc98,pit00}.  The 2175~\AA\ bump is commonly 
2548: associated with graphitic carbonaceous material \citep[as 
2549: first proposed by][]{sd65}.  \cite{ds98} showed that certain 
2550: PAHs and PAH clusters can reproduce the 2175~\AA\ bump and 
2551: that these particles can result from interstellar processing 
2552: of hydrogenated amorphous carbon.  The recent study of 
2553: unusual PAH-like spectra by \cite{slo07} demonstrates that 
2554: hydrocarbons can exist with a range of aromatic-to-aliphatic 
2555: ratios.  Many of the sources of unusual PAH-like emission are 
2556: carbon-rich post-AGB objects.  All of these observations 
2557: support an evolutionary relationship from amorphous carbon 
2558: produced by carbon stars on the AGB and to interstellar PAHs 
2559: to the more graphitic material responsible for the 2175~\AA\ 
2560: bump.  The results in this paper demonstrate that carbon-rich 
2561: dust production does not decrease in more metal-poor samples, 
2562: and yet, metal-poor systems like the SMC are missing the 
2563: 2175~\AA\ bump that should result.
2564: 
2565: Either the missing carbon-rich dust did not form, or it has
2566: been destroyed.  \cite{gc98} suggested that harsher radiation 
2567: fields in metal-poor systems can destroy the carrier of the 
2568: 2175~\AA\ bump.  Stellar modelling shows that decreasing the 
2569: metallicity of a star results in less UV line blanketing and 
2570: an emitted spectrum with more UV emission \citep{giv02}, which 
2571: readily explains the harsher interstellar radiation observed 
2572: in metal-poor \ion{H}{2} regions \citep{mh02}.  \cite{gal08} 
2573: recently reached the opposite conclusion.  They used galactic 
2574: evolution models to investigate the absence of PAH emission 
2575: in blue compact dwarf galaxies and other metal-poor galaxies 
2576: \citep{eng05,wu06}, and concluded that these systems produce 
2577: significantly less carbon-rich dust than the Galaxy.  Our 
2578: sample of Magellanic carbon stars provides no evidence for 
2579: less carbon-rich dust formation in more primordial systems, 
2580: which leads to the conclusion that the lack of carbon-bearing
2581: dust in these systems results not from reduced production,
2582: but from enhanced destruction.
2583: 
2584: \acknowledgements
2585: 
2586: The authors thank the anonymous referee for carefully reading
2587: the manuscript and helping us to improve it.
2588: These observations were made with the {\it Spitzer Space
2589: Telescope}, which is operated by JPL, California Institute of
2590: Technology under NASA contract 1407 and supported by NASA
2591: through JPL (contract number 1257184).  This research has
2592: made use of the SIMBAD and VIZIER databases, operated at the
2593: Centre de Donn\'{e}es astronomiques de Strasbourg, and the
2594: Infrared Science Archive at the Infrared Processing and
2595: Analysis Center, which is operated by JPL.
2596: 
2597: \begin{thebibliography}{}
2598: 
2599: \bibitem[Alcock et al.(1996)]{macho} Alcock, C., et al.\ 1996, \apj,
2600:   461, 84
2601: \bibitem[Allamandola et al.(1989)]{atb89} Allamandola, L.~J., Tielens,
2602:   A.~G.~G.~M., \& Barker, J.~R.\ 1989, \apjs, 71, 733
2603: %\bibitem[Alves et al.(2002)]{alv02} Alves, D.~R., Rejkuba, M., Minniti, D.,
2604: %  \& Cook, K.~H., 2002, \apj, 573, L51
2605: \bibitem[Alves(2004)]{alv04} Alves, D.~R.\ 2004, New Astron.\ Rev.\ 2004, 
2606:   48, 659
2607: \bibitem[Bergeat \& Chevallier(2005)]{bc05} Bergeat, J., \& Chevallier, L.\
2608:   2005, \aap, 429, 235
2609: \bibitem[Bernard-Salas et al.(2006)]{jbs06} Bernard-Salas, J., Peeters, E.,
2610:   Sloan, G.~C., Cami, J., Guiles, S., \& Houck, J.~R.\ 2007, \apj, 652, L29
2611: \bibitem[Blanco et al.(1978)]{bbm78} Blanco, B.~M., Blanco, V.~M., \& McCarthy,
2612:   M.~F.\ 1978, Nature, 271, 638
2613: \bibitem[Blanco et al.(1980)]{bbm80} Blanco, V.~M., Blanco, B.~M., \& McCarthy,
2614:   M.~F.\ 1980, \apj, 242, 938
2615: \bibitem[Blanco et al.(1981)]{bfm81} Blanco, V.~M., Frogel, J.~A., \&
2616:   McCarthy, M.~F.\ 1981, \pasp, 93, 532
2617: \bibitem[Blum et al.(2006)]{blu06} Blum, R.~D., et al.\ 2006, AJ, 132, 2034
2618: \bibitem[Bolatto et al.(2007)]{bol07} Bolatto, A.~D., et al.\ 2007, \apj, 655,
2619:   212
2620: \bibitem[Buchanan et al.(2006)]{buc06} Buchanan, C.~L., Kastner, J.~H.,
2621:   Forrest, W.~J., Hrivnak, B.~J., Sahai, R., Egan, M., Frank, A., \& Barnbaum,
2622:   C.\ 2006, \aj, 132, 1890
2623: \bibitem[Cami(2002)]{cam02} Cami, J.\ 2002, Ph.D. Thesis, Univ. of Amsterdam
2624: \bibitem[Cernuschi et al.(1967)]{cmc67} Cernuschi, F., Marsiano, F., \&
2625:   Codina, S.\ 1967, Ann. d'Ap., 30, 1039
2626: \bibitem[Chihara et al.(2002)]{chi02} Chihara, H., Koike, C.,
2627:   Tsuchiyama, A., Tachibana, S., \& Sakamoto, D.\ 2002, \aap, 391, 267
2628: \bibitem[Cioni \& Habing(2003)]{ch03} Cioni, M.-R.~L.\ \& Habing, H.~J.\
2629:   2003, \aap, 402, 133
2630: \bibitem[DePew et al.(2006)]{dep06} DePew, K., Speck, A., \& Dijkstra, C.\
2631:   2006, \apj, 640, 971
2632: \bibitem[Dessy(1959)]{des59} Dessy, J.~L.\ 1959, Bol.\ Inst.\ Mat.\ Astr.\ 
2633:   Fis.\ Cordoba, Vol.\ 1, No.\ 2
2634: \bibitem[Duley \& Seahra(1998)]{ds98} Duley, W.~W.\ \& Seahra, S.\ 1998, \apj,
2635:   507, 874
2636: \bibitem[Dwek \& Scalo(1980)]{ds80} Dwek, E.\ \& Scalo, J.~M.\ 1980, \apj, 239, 
2637:   193
2638: \bibitem[Egan et al.(2001)]{msx01} Egan, M.~P., van Dyk, S.~D., \&
2639:   Price, S.~D.\ 2001, \aj, 122, 1844
2640: \bibitem[Egan et al.(2003)]{msx03} Egan, M. P., et al.\ 2003,
2641:   ``The Midcourse Space Experiment Point Source Catalog, Version 2.3,
2642:   Explanatory Guide,'' Air Force Research Laboratory Technical Report
2643:   AFRL-VS-TR-2003-1589
2644: \bibitem[Egan \& Sloan(2001)]{es01} Egan, M.~P. \& Sloan, G.~C.\ 2001, \apj,
2645:   558, 165
2646: \bibitem[Emery et al.(2006)]{ec06}  Emery, J.~P.,
2647:   Cruikshank, D.~P., van Cleve, J.\ 2006, Icarus, 182, 496
2648: \bibitem[Engelbracht et al.(2005)]{eng05} Engelbracht, C.~W., Gordon, K.~D., 
2649:   Rieke, G.~H., Werner, M.~W., Dale, D.~A., \& Latter, W.~B.\ 2005, \apj, 
2650:   628, L29
2651: \bibitem[Engelke et al.(2004)]{ekp04} Engelke, C.~W., Kraemer, K.~E., \& 
2652:   Price, S.~D.\ 2004, \apjs, 150, 343
2653: \bibitem[Fabian et al.(2001a)]{fab01a} Fabian, D., Posch, Th., Mutschke, H., 
2654:   Kerschbaum, F., \& Dorschner, J.\ 2001a, \aap, 373, 1125
2655: \bibitem[Fabian et al.(2001b)]{fab01b} Fabian, D., Henning, T., J\"{a}ger, C., 
2656:   Mutschke, H., Dorschner, J., \& Wehrhan, O.  2001, \aap, 378, 228
2657: \bibitem[Galliano et al.(2008)]{gal08} Galliano, F., Madden, S.~C., Tielens, 
2658:   A.~G.~G.~M., Peeters, E., \& Jones, A.~P.\ 2008, \apj, 679, 310
2659: \bibitem[Gehrz(1989)]{geh89} Gehrz, R.\ 1989, in Interstellar Dust, Proc. IAU
2660:   Symp. 135, ed. L.~J.\ Allamandola \& A.~G.~G.~M.\ Tielens (Dordrecht:  
2661:   Kluwer Academic Publishers), 445
2662: \bibitem[Gillett, et al.(1968)]{gil68} Gillett, F.~C., Low, F.~J.,
2663:   \& Stein, W.~A.\ 1968, \apj, 154, 677
2664: \bibitem[Giveon et al.(2002)]{giv02} Giveon, U., Sternberg, A., Lutz, D., 
2665:   Feuchtgruber, H., \& Pauldrach, A.~W.~A.\ 2002, \apj, 566, 880
2666: \bibitem[Glaccum(1994)]{gla94} Glaccum, W.\ 1994, in Proc.\ of the Airborne 
2667:   Astronomy Symp.\ on the Galactic Ecosystem:  From Gas to Stars to Dust, ed. 
2668:   M.~R.\ Haas, J.~A.\ Davidson, \& E.~F.\ Erickson (San Francisco:  ASP), 395
2669: \bibitem[Gordon \& Clayton(1998)]{gc98} Gordon, K.\ \& Clayton, G.~C.\ 
2670:   1998, \apj, 500, 816
2671: \bibitem[Groenewegen et al.(2007)]{gro07} Groenwegen, M.~A.~T., et al.\ 2007, 
2672:   \mnras, 376, 313
2673: \bibitem[Heras et al.(2002)]{her02} Heras, A.~M., Shipman, R.~F., Price, S.~D.,
2674:   de Graauw, Th., Walker, H.~J., Jourdain de Muizon, M., Kessler, M.~F., 
2675:   Prusti, T., Decin, L.; Vandenbussche, B., Waters, L.~B.~F.~M.\ 2002, \aap, 
2676:   394, 539
2677: \bibitem[Hodge et al.(2004)]{hod04} Hodge, T.~M., Kraemer, K.~E., Price, S.~D.,
2678:   \& Walker, H.~J.\ 2004, \apjs, 151, 299
2679: \bibitem[Hony et al.(2001)]{hon01} Hony, S., Van Kerckhoven, C., Peeters, E., 
2680:   Tielens, A.~G.~G.~M., Hudgins, D.~M., \& Allamandola, L.~J.\ 2001, \aap, 370, 
2681:   1030
2682: \bibitem[Houck et al.(2004)]{hou04} Houck, J.~R., et al.\ 2004, \apjs, 154, 18
2683: \bibitem[Huber \& Herzberg(1979)]{hh79} Huber, K.~P. \& Herzberg, G.\ 1979,
2684:   Molecular Spectra and Molecular Structure, Vol. IV.  Constants of Diatomic
2685:   Molecules (New York:  Van Nostrand Reinhold Co.)
2686: \bibitem[IRAS Science Team(1986)]{lrs86} IRAS Science Team 1986, \aaps, 65, 607
2687: \bibitem[J\"{a}ger et al.(1998)]{jag98} J\"{a}ger, C., Molster, F.~J.,
2688:   Dorschner, J., Henning, Th., Mutschke, H., Waters, L.~B.~F.~M.\ 1998,
2689:   \aap, 339, 904
2690: \bibitem[Kastner et al.(2006)]{kas06} Kastner, J.~H., Buchanan, C.~L.,
2691:   Sargent, B., \& Forrest, W.~J.\ 2006, \apj, 638, L29
2692: \bibitem[Keremedjiev \& Sloan(2006)]{ks06} Keremedjiev, M.~S. \& Sloan, G.~C.,
2693:   2006, IRS Technical Report 06001
2694: \bibitem[Keller et al.(2008)]{kel08} Keller, L.~D., Sloan, G.~C., Forrest,
2695:   W.~J., Ayala, S., D'Alessio, P., Shah, S., Calvet, N., Hartmann, L.,
2696:   Najita, J., Sargent, B., Li, A., Watson, D.~M., \& Chen, C.~H.\ 2008,
2697:   \apj, submitted
2698: \bibitem[Keller \& Wood(2006)]{kw06} Keller, S.~C.\ \& Wood, P.~R.\ 2006,
2699:   \apj, 642, 834
2700: \bibitem[Koike et al.(2003)]{koi03} Koike, C., Chihara, H., Tsuchiyama, A.,
2701:   Suto, H., Sogawa, H., \& Okuda, H.\ 2003, \aap, 399, 110
2702: \bibitem[Kraemer et al.(2002)]{kra02} Kraemer, K.~E., Sloan, G.~C., Price,
2703:   S.~D., \& Walker, H.~J.\ 2002, \apjs, 140, 389
2704: \bibitem[Kraemer et al.(2005)]{kra05} Kraemer, K.~E., Sloan, G.~C., Wood,
2705:   P.~R., Price, S.~D., \& Egan, M.~P.\ 2005, \apj, 631, L147
2706: \bibitem[Kraemer et al.(2006)]{kra06} Kraemer, K.~E., Sloan, G.~C.,
2707:   Bernard-Salas, J., Price, S.~D., Egan, M.~P., \& Wood, P.~R.\ 2006, \apj,
2708:   652, L25
2709: \bibitem[Lagadec et al.(2007)]{lag07} Lagadec, E., Zijlstra, A.~A., Sloan,
2710:   G.~C., Matsuura, M., Wood, P.~R., van Loon, J.~Th., Harris, G.~J., Blommaert,
2711:   J.~A.~D.~L., Hony, S., Groenewegen, M.~A.~T., Feast, M.~S., Whitelock, P.~A.,
2712:   Menzies, J.~W., \& Cioni, M.-R.\ 2007, \mnras, 376, 1270
2713: \bibitem[Lambert \& Clegg(1980)]{lc80} Lambert, D.~L.\ \& Clegg, R.~E.~S.\
2714:   1980, \mnras, 191, 367
2715: \bibitem[Leisenring et al.(2008)]{lei08} Leisenring, J.~M., Markwick-Kemper, 
2716:   F., \& Sloan, G.C.\ 2008, \apj, in press
2717: \bibitem[Levesque et al.(2006)]{lev06} Levesque, E.~M., Massey, P., Olsen, 
2718:   K.~A.~G., Plez, B., Meynet, G., \& Maeder, A.\ 2006, \apj, 645, 1102
2719: \bibitem[Little-Marenin \& Price(1986)]{lmp86} Little-Marenin, I.~R., \& 
2720:   Price, S.~D.\ 1986, in Summer School on Interstellar Processes:  Abstracts of 
2721:   Contributed Papers, ed.\ D.~J.\ Hollenbach \& H.~A.\ Thronson, NASA Tech.\ 
2722:   Memorandum 88342 (NASA Ames Research Center:  Moffett Field, CA), 137
2723: \bibitem[Lutz et al.(2007)]{lut07} Lutz, D., Sturm, E., Tacconi, L.~J., 
2724:   Valiante, E., Schweitzer, M., Netzer, H., Maiolino, R., Andreani, P., 
2725:   Shemmer, O., \& Veilleux, S.\ 2007, \apj, 661, L25
2726: \bibitem[Maeder(1992)]{mae92} Maeder, A.\ 1992,\aap, 264, 105
2727: \bibitem[Markwick-Kemper et al.(2008)]{mk08} Markwick-Kemper, et al.\ 2008, in
2728:   preparation
2729: \bibitem[Marshall et al.(2004)]{mar04} Marshall, J.~R., van Loon, J.~Th., 
2730:   Matsuura, M., Wood, P.~R., Zijlstra, A.~A., \& Whitelock, P.~A.\ 2004, 
2731:   \mnras, 355, 1348
2732: \bibitem[Martin \& Rogers(1987)]{mr87} Martin, P.~G. \& Rogers, C.\ 1987, \apj,
2733:   322, 374
2734: \bibitem[Mart\'{i}n-Hern\'{a}ndez et al.(2002)]{mh02} Mart\'{i}n-Hern\'{a}ndez, 
2735:   N.~L., Vermeij, R., Tielens, A.~G.~G.~M., van der Hulst, J.~M., \& Peeters, 
2736:   E.\ 2002, \aap, 389, 286
2737: \bibitem[Matsuura et al.(2002)]{mat02} Matsuura, M., Zijlstra, A.~A., van
2738:   Loon, J.~Th., Yamamura, I., Markwick, A.~J., Woods, P.~M., \& Waters,
2739:   L.~B.~F.~M. 2002, \apj, 580, L133
2740: \bibitem[Matsuura et al.(2005)]{mat05} Matsuura, M., Zijlstra, A.~A., van Loon,
2741:   J.~Th., Yamamura, I., Markwick, A.~J., Whitelock, P.~A., Woods, P.~M., 
2742:   Marshall, J.~R., Feast, M.~W., \& Waters, L.~B.~F.~M.\ 2005, \aap, 434, 691
2743: \bibitem[Matsuura et al.(2006)]{mat06} Matsuura, M., et al.\ 2006, \mnras, 371,
2744:   415
2745: \bibitem[McGregor et al.(1994)]{caspir} McGregor, P.~J., Hart, J.,
2746:   Hoadley, D., \& Bloxham, G.\ 1994, in Infrared Astronomy with Arrays,
2747:   ed. I.\ McLean (Dordrecht:  Kluwer), 299
2748: \bibitem[Meikle et al.(2007)]{mei07} Meikle, W.~P.~S., Mattila, S., Pastorello,
2749:   A., Gerardy, C.~L., Kotak, R., Sollerman, J., Van Dyk, S.~D., Farrah, D.,
2750:   Filippenko, A.~V., H\"{o}flich, P., Lundqvist, P., Pozzo, M., \& Wheeler,
2751:   J.~C.\ 2007, \apj, 665, 608
2752: \bibitem[Meixner et al.(2006)]{mei06} Meixner, M., et al.\ 2006, \aj, 132, 2268
2753: \bibitem[Min et al.(2003)]{min03} Min, M., Hovenier, J.~W., \& de Koter, A.\
2754:   2003, \aap, 404, 35
2755: \bibitem[Molster et al.(2002)]{mol02} Molster, F.~J., Waters, L.~B.~F.~M., 
2756:   Tielens, A.~G.~G.~M.\ 2002, \aap, 382, 222
2757: \bibitem[Moutou et al.(2000)]{mou00} Moutou, C., Verstraete, L, L\'{e}ger, A.,
2758:   Sellgren, K., \& Schmidt, W.\ 2000, \aap, 354, L17
2759: \bibitem[Ossenkopf et al.(1992)]{ohm92} Ossenkopf, V., Henning, T., \&
2760:   Mathis, J.~S.\ 1992, \aap, 261, 567
2761: %\bibitem[Paczy\'{n}ski(1971)]{pac71} Paczy\'{n}ski, B. 1971, Acta Astr., 21, 4
2762: %\bibitem[Peeters et al.(2004)]{pee04} Peeters, E., Mattioda, A.~L., Hudgins, 
2763: %  D.~M., \& Allamandola, L.~J.\ 2004, \apj, 617, L65
2764: \bibitem[Peeters et al.(2002)]{pee02} Peeters, E., Hony, S., Van Kerckhoven, 
2765:   C., Tielens, A.~G.~G.~M., Allamandola, L.~J., Hudgins, D.~M., \& 
2766:   Buauschlicher, C.~W.\ 2002, \aap, 390, 1089
2767: \bibitem[Phillips \& Davis(1979)]{pd79} Phillips, J.~G.\ \& Davis, S.~P.\ 
2768:   1979, \apj, 229, 867
2769: \bibitem[Pitman et al.(2006)]{pit06} Pitman, K.~M., Speck, A.~K., \&
2770:   Hofmeister, A.~M.\ 2006, \mnras, 371, 1744
2771: \bibitem[Pitman et al.(2000)]{pit00} Pitman, K.~M., Clayton, G.~C., \&
2772:   Gordon, K.~D.\ 2000, \pasp, 112, 537\\
2773: \bibitem[Posch et al.(1999)]{po99} Posch, T., Kerschbaum, F., Mutschke, H., 
2774:   Fabian, D., Dorschner, J., \& Hron, J. 1999, \aap, 352, 609
2775: \bibitem[Renzini \& Voli(1981)]{rv81} Renzini, A.\ \& Voli, M.\ 1981, \aap,
2776:   94, 175
2777: \bibitem[Salpeter(1955)]{sal55} Salpeter, E.~E.\ 1955, \apj, 121, 161
2778: \bibitem[Skrutskie et al.(2006)]{skr06} Skrutskie, M.~F., et al.\ 2006, \aj,
2779:   131, 1163
2780: \bibitem[Sloan et al.(2006a)]{slo06a} Sloan, G.~C., Devost, D., Bernard-Salas,
2781:   J., Wood, P.~R., \& Houck, J.~R.\ 2006, \apj, 638, 472
2782: \bibitem[Sloan \& Egan(1995)]{se95} Sloan, G.~C. \& Egan, M.~P.\ 1995, \apj,
2783:   444, 452
2784: \bibitem[Sloan et al.(2007)]{slo07} Sloan, G.C., Jura, M., Duley, W.~W.,
2785:   Kraemer, K.~E., Bernard-Salas, J., Forrest, W.~J., Sargent, B., Li, A.,
2786:   Barry, D.~J., Bohac, C.~J., Watson, D.~M., \& Houck, J.~R.\ 2007, \apj, 
2787:   664, 1144
2788: \bibitem[Sloan et al.(2006b)]{slo06b} Sloan, G.~C., Kraemer, K.~E., Matsuura, 
2789:   M., Wood, P.~R., Price, S.~D., \& Egan, M.~P.\ 2006, \apj, 645, 1118
2790: \bibitem[Sloan et al.(2003)]{slo03} Sloan, G.~C., Kraemer, K.~E., Price, 
2791:   S.~D., \& Shipman, R.~F.\ 2003, \apjs, 147, 379
2792: \bibitem[Sloan et al.(1996)]{slo96} Sloan, G. C., LeVan, P. D., \&
2793:   Little-Marenin, I. R. 1996, \apj, 463, 310
2794: \bibitem[Sloan \& Price(1995)]{sp95} Sloan, G.~C. \& Price, S.D.\ 1995, \apj,
2795:   451, 758
2796: \bibitem[Sloan \& Price(1998)]{sp98} Sloan, G.~C. \& Price, S.D.\ 1998, \apjs,
2797:   119, 141
2798: \bibitem[Sloan et al.(2008a)]{slo08a} Sloan, G.~C., Zijlstra, A.~A., Kraemer,
2799:   K.~E., Markwick-Kemper, F., \& Leisenring, J.~M.\ 2008, in The Biggest, 
2800:   Baddest, Coolest Stars, ed. R.~E.\ Stencel \& B.~J.\ Smith
2801: \bibitem[Sloan et al.(2008b)]{slo08b} Sloan, G.~C., et al.\ 2008, \mnras,
2802:   in preparation  % 08b - sgal - 13-14 um feature
2803: \bibitem[Sloan et al.(2008c)]{slo08c} Sloan, G.~C., et al.\ 2008, \aj, 
2804:   in preparation  % 08c - speccal, A dwarfs and K giants
2805: \bibitem[Sloan et al.(2008d)]{slo08d} Sloan, G.~C., et al.\ 2008, \apj,
2806:   in preparation  % 08d - M giants
2807: \bibitem[Smith et al.(1995)]{smi95} Smith, V.~V., Plez, B., Lambert, D.~L., 
2808:   Lubowich, D.~A.\ 1995, \apj, 441, 735
2809: \bibitem[Speck et al.(2006)]{spe06} Speck, A.~K., Cami, J., Markwick-Kemper,
2810:   C., Leisenring, J., Szcerba, R., Dijkstra, C., Van Dyk, S., \& Meixner, M.\
2811:   2006, \apj, 650, 892
2812: \bibitem[Spinrad \& Wing(1969)]{sw69} Spinrad, H.\ \& Wing, R.~F.\ 1969,
2813:   \araa, 7, 249
2814: \bibitem[Stecher \& Donn(1965)]{sd65} Stecher, T.~P.\ \& Donn, B.\ 1965, \apj,
2815:   142, 1681
2816: \bibitem[Stencel et al.(1990)]{ste90} Stencel, R.~E., Nuth, J.~A.,
2817:   Little-Marenin, I.~R., \& Little, S.~J.\ 1990, \apj, 350, L45.
2818: \bibitem[Sugerman et al.(2006)]{sug06} Sugerman, B.~E.~K., et al.\ 2006,
2819:   Science, 313, 196
2820: \bibitem[Suto et al.(2006)]{sut06} Suto, H., Sogawa, H., Tachibana, S., Koike, 
2821:   C., Karoji, H., Tsuchiyama, A., Chihara, H., Mizutani, K., Akedo, J., Ogiso, 
2822:   K., Fukui, T., \& Ohara, S.\ 2006, \mnras, 370, 1599
2823: \bibitem[Sylvester et al.(1998)]{syl98} Sylvester, R.~J., Skinner, C.~J., \& 
2824:   Barlow, M.~J. 1997, \mnras, 301 1083
2825: \bibitem[Szym\'{a}nski(2005)]{ogle05} Szyma\'{n}ski, M.~K.\ 2005, Acta Astron.,
2826:   55, 43
2827: %\bibitem[Teplitz et al.(2006)]{tep06} Teplitz, H.~I., Armus, L., Soifer, B.~T.,
2828: %  Charmandaris, V., Marshall, J.~A., Spoon, H., Lawrence, C., Hao, L.,
2829: %  Higdon, S., Wu, Y., Lacy, M., Eisenhardt, P.~R., Herter, T., \& Houck, J.~R.\
2830: %  2006, \apj, 638, L1
2831: \bibitem[Teplitz et al.(2007)]{tep07} Teplitz, H.~I., Desai, V., Armus, L.,
2832:   Chary, R., Marshall, J.~A., Colbert, J.~W., Frayer, D.~T., Pope, A., Blain, 
2833:   A., Spoon, H.~W.~W., Charmandaris, V., \& Scott, D.\ 2007, \apj, 659, 941
2834: \bibitem[Tielens(1990)]{tie90} Tielens, A.~G.~G.~M.\ 1990, in From Miras
2835:   to Planetary Nebulae:  Which Path for Stellar Evolution?, ed. M.~O.\
2836:   Mennessier \& A.\ Omont (Gif-sur-Evette:  Editions Fronti\`{e}res), 186
2837: \bibitem[Trams et al.(1999)]{tra99} Trams, N.~R., van Loon, J.~Th., Waters,
2838:   L.~B.~F.~M., Zijlstra, A.~A., Loup, C., Whitelock, P.~A., Groenewegen,
2839:   M.~A.~T., Blommaert, J.~A.~D.~L., Siebenmorgen, R., Heske, A., \& Feast,
2840:   M.~W.\ 1999, \aap, 346, 843
2841: \bibitem[Udalski et al.(1997)]{ogle97} Udalski, A., Kubiak, M., \&
2842:   Szym\'{a}nski, M.\ 1997, Acta Astron., 47, 319
2843: \bibitem[Valenti et al.(1998)]{val98} Valenti, J.~A., Piskunov, N., 
2844:   Johns-Krull, C.~M.\ 1998, \apj, 498, 851
2845: \bibitem[van Loon et al.(1999)]{vl99} van Loon, J.~Th., Groenewegen, M.~A.~T.,
2846:   de Koter, A., Trams, N.~R., Waters, L.~B.~F.~M., Zijlstra, A.~A., Whitelock,
2847:   P.~A., \& Loup, C.\ 1999, \aap, 559, 572  
2848: \bibitem[van Loon et al.(2001)]{vl01} van Loon, J.~Th., Zijlstra, A.~A., 
2849:   Bujarrabal, V., \& Nyman, L.-\AA.\ 2001, \aap, 368, 950 
2850: \bibitem[van Loon et al.(2005a)]{vl05a} van Loon, J.~Th., Cioni, M.-R.~L., 
2851:   Zijlstra, A.~A., \& Loup, C.\ 2005, \aap, 438, 273
2852: \bibitem[van Loon et al.(2005b)]{vl05b} van Loon, J.~Th., Oliveira, J.~M., 
2853:   Wood, P.~R., Zijlstra, A.~A., Sloan, G.~C., Matsuura, M., Whitelock, P.~A.,
2854:   Groenewegen, M.~A.~T., Blommaert, J.~A.~D.~L., Cioni, M.-R.~L., Hony, S.,
2855:   Loup, C., \& Waters, L.~B.~F.~M.\ 2005, \mnras, 364, L71
2856: \bibitem[van Loon et al.(2006)]{vl06} van Loon, J.~Th., Marshall, J.~R., 
2857:   Cohen, M., Matsuura, M., Wood, P.~R., Yamamura, I., \& Zijlstra, A.~A.\
2858:   2006, \aap, 447, 971
2859: \bibitem[van Loon et al.(2008)]{vl08} van Loon, J.~Th., Cohen, M., Oliveira,
2860:   J.~M., Matsuura, M., McDonald, I., Sloan, G.~C., Wood, P.~R., \& Zijlstra,
2861:   A.~A.\ 2008, \mnras, submitted
2862: \bibitem[Vassiliadis \& Wood(1993)]{vw93} Vassiliadis, E. \& Wood P.~R. 1993,
2863:   \apj, 413, 641
2864: \bibitem[Werner et al.(2004a)]{wer04a} Werner, M.~W., et al.\ 2004a,
2865:   \apjs, 154, 1
2866: \bibitem[Werner et al.(2004b)]{wer04b} Werner, M.~W., Uchida, K.~I., Sellgren, 
2867:   K., Marengo, M., Gordon, K.~D., Morris, P.~W., Houck, J.R., \& Stansberry,
2868:   J.~A.\ 2004b, \apjs, 154, 309
2869: \bibitem[Wheeler et al.(1989)]{whe89} Wheeler, J.~C., Sneden, C., \& Truran,
2870:   J.~W., Jr.\ 1989, \araa, 27, 279
2871: \bibitem[Whitelock \& Feast(2000)]{wf00} Whitelock, P. \& Feast, M. 2000,
2872:   Mem.\ Soc.\ Astron.\ Ital.\ 71, 601
2873: \bibitem[Whitelock et al.(2003)]{whi03} Whitelock, P.~A., Feast, M.~W., 
2874:   van Loon, J.~Th., \& Zijlstra, A.~A.\ 2003, \mnras, 342, 86
2875: \bibitem[Whitney et al.(2003a)]{whi03a} Whitney, B.~A., Wood, K., Bjorkman, 
2876:   J.~E., \& Wolff, M.~J.\ 2003a, \apj, 591, 1049
2877: \bibitem[Whitney et al.(2003b)]{whi03b} Whitney, B.~A., Wood, K., Bjorkman, 
2878:   J.~E., \& Cohen, M.\ 2003b, \apj, 598, 1079
2879: \bibitem[Woitke(2006)]{woi06} Woitke, P. 2006, \aap, 452, 537
2880: \bibitem[Wood(1981)]{woo81} Wood, P.~R. 1981, in Physical Processes in
2881:   Red Giants, ed. I. Iben \& A. Renzini, (Dordrecht:  Reidel), 135
2882: \bibitem[Wood et al.(1983)]{wbf83} Wood, P.~R., Bessell, M.~S., \& Fox, M.~W.\
2883:   1983, \apj, 272, 99
2884: \bibitem[Wood et al.(1985)]{wbp85} Wood, P.~R., Bessell, M.~S., \& Paltoglou,
2885:   G.\ 1985, \apj, 290, 477
2886: \bibitem[Wood et al.(1992)]{woo92} Wood, P.~R., Whiteoak, J.~B., Hughes, 
2887:   S.~M.~G., Bessell, M.~S., Gardner, F.~F., \& Hyland, A.~R.\ 1992, \apj, 
2888:   397, 552
2889: \bibitem[Wu et al.(2006)]{wu06} Wu, Y., Charmandaris, V., Hao, L., Brandl, 
2890:   B.~R., Bernard-Salas, J., Spoon, H.~W.~W.\ Houck, J.~R.\ 2006, \apj, 639, 157
2891: \bibitem[Wyckoff \& Clegg(1978)]{wc78} Wyckoff, S.\ \& Clegg, R.~E.~S.\
2892:   1979, \mnras, 184, 127
2893: \bibitem[Zijlstra et al.(2006)]{zij06} Zijlstra, A.~A., et al.\ 2006, \mnras,
2894:   370, 1961
2895: 
2896: \end{thebibliography}
2897: 
2898: 
2899: \clearpage
2900: 
2901: \clearpage
2902: \thispagestyle{empty}
2903: \setlength{\voffset}{15mm}
2904: \clearpage
2905: \setlength{\voffset}{0mm}
2906: 
2907: \clearpage
2908: 
2909: \clearpage
2910: 
2911: \end{document}
2912: