1: \documentclass[12pt,preprint]{aastex}
2: %% manuscript produces a one-column, double-spaced document:
3: %% \documentclass[manuscript]{aastex}
4: %% preprint2 produces a double-column, single-spaced document:
5: %% \documentclass[preprint2]{aastex}
6: %% Sometimes a paper's abstract is too long to fit on the
7: %% title page in preprint2 mode. When that is the case,
8: %% use the longabstract style option.
9: %% \documentclass[preprint2,longabstract]{aastex}
10: \newcommand{\ltsimeq}{\raisebox{-0.6ex}{$\,\stackrel
11: {\raisebox{-.2ex}{$\textstyle <$}}{\sim}\,$}}
12: \newcommand{\gtsimeq}{\raisebox{-0.6ex}{$\,\stackrel
13: {\raisebox{-.2ex}{$\textstyle >$}}{\sim}\,$}}
14: \newcommand{\prsimeq}{\raisebox{-0.6ex}{$\,\stackrel
15: {\raisebox{-.2ex}{$\propto$}}{\sim}\,$}}
16: %% \slugcomment{Submitted to ApJL}
17: \shorttitle{NGC\,6791}
18: \shortauthors{van Loon et al.}
19: \begin{document}
20: \title{{\it Spitzer} Space Telescope evidence in NGC\,6791: no super-mass-loss
21: at super-solar metallicity to explain helium white dwarfs?\altaffilmark{1}}
22: \author{Jacco Th. van Loon\altaffilmark{2},
23: Martha L. Boyer\altaffilmark{3},
24: Iain McDonald\altaffilmark{2}
25: }
26: \altaffiltext{1}{This work is based on observations made with the {\it Spitzer
27: Space Telescope}, which is operated by the Jet Propulsion Laboratory,
28: California Institute of Technology under a contract with NASA.}
29: \altaffiltext{2}{Astrophysics Group, Lennard-Jones Laboratories, Keele
30: University, Staffordshire, ST5 5BG, UK; jacco@astro.keele.ac.uk,
31: iain@astro.keele.ac.uk.}
32: \altaffiltext{3}{Dept.\ of Astronomy, School of Physics \& Astronomy, 116
33: Church Street S.E., University of Minnesota, Minneapolis, MN 55455;
34: mboyer@astro.umn.edu.}
35: \begin{abstract}
36: We use archival {\it Spitzer} Space Telescope photometry of the old,
37: super-solar metallicity massive open cluster NGC\,6791 to look for evidence of
38: enhanced mass loss, which has been postulated to explain the optical
39: luminosity function and low white dwarf masses in this benchmark cluster. We
40: find a conspicuous lack of evidence for prolificacy of circumstellar dust
41: production that would have been expected to accompany such mass loss. We also
42: construct the optical and infrared luminosity functions, and demonstrate that
43: these fully agree with theoretical expectations. We thus conclude that there
44: is no evidence for the mass loss of super-solar metallicity red giants to be
45: sufficiently high that they can avoid the helium flash at the tip of the red
46: giant branch.
47: \end{abstract}
48: \keywords{stars: evolution --- stars: luminosity function, mass function ---
49: stars: mass loss --- stars: Population II --- open clusters and associations:
50: individual (NGC\,6791) --- infrared: stars}
51: %% \object{NGC\,6791}
52: %=========================================================================== 1
53: \section{Introduction}
54:
55: Low-mass stars in galactic globular clusters lose $\sim20$\% of their mass
56: during the ascent of the first red giant branch
57: \citep[RGB;][]{Dupree07,Origlia07,Mcdonald07}, offering a partial explanation
58: for the blue extent of the core-helium burning horizontal branch
59: \citep{Rood73}. It has been argued recently that in the old (8 Gyr),
60: super-solar metallicity ($Z\simeq0.05$) massive open cluster NGC\,6791, at
61: least 40\% of RGB stars experience higher rates of mass loss; this would lead
62: to their departure from the RGB before the helium flash, resulting in
63: undermassive helium white dwarfs \citep{Castellani93,Hansen05,Kalirai07}. If
64: true, this would have important implications for the white dwarf populations
65: produced in metal-rich stellar systems, and for the mass return into the
66: interstellar medium of such systems --- which include giant elliptical
67: galaxies and the bulges of spiral galaxies. \citet{Bedin08}, on the other
68: hand, propose an alternative explanation for the white dwarf properties in
69: NGC\,6791 without the need for enhanced RGB mass loss.
70:
71: We investigate the evidence for super-mass-loss on the RGB of NGC\,6791 by
72: looking for circumstellar dust that may form in the winds from cool RGB stars
73: \citep[e.g.,][]{Gehrz71,Origlia07} in archival {\it Spitzer} Space Telescope
74: infrared (IR) images, and by comparing the IR and optical luminosity functions
75: with model predictions.
76:
77: %=========================================================================== 2
78: \section{{\it Spitzer} photometry of NGC\,6791}
79:
80: The observations of NGC\,6791 presented here were intended for {\it James
81: Webb} Space Telescope photometric and astrometric calibration
82: \citep{Diazmiller07}; we obtained the data from the {\it Spitzer} Space
83: Telescope \citep[hereafter {\it Spitzer};][]{Werner04} public archive. The
84: observations were made with the {\it Spitzer} Infrared Array Camera
85: \citep[IRAC;][]{Fazio04} on 2005 June 14 UT. The image maps cover
86: $\approx20^\prime\times15^\prime$, with 3.6 and 5.8 $\mu$m mosaics offset to
87: the northwest from the cluster center by $\approx8^\prime$ and by
88: $\approx3^\prime$ at 4.5 and 8 $\mu$m. The region around the cluster is
89: displayed in Fig.\ \ref{f1}. High dynamic range mode had been implemented to
90: prevent saturation of brighter sources (1 s exposures) while maintaining a
91: high signal-to-noise on fainter sources (26.8 s exposures), and observations
92: had been made with an eight position cycling dither pattern to build
93: redundancy against outliers and artefacts.
94:
95: Raw data were processed with {\it Spitzer} Science Center (SSC) pipeline
96: version S14.0.0. We corrected the Basic Calibrated Data images for array
97: distortions and mosaicked the resulting images with the SSC Legacy MOPEX
98: software v16.2.1 \citep{Makovoz05}, which includes background-matching to
99: minimize pixel offsets in overlapping regions of the mosaics and eliminates
100: cosmic rays and other outliers.
101:
102: We carried out point-spread function photometry using the DAOphot II
103: photometry package \citep{Stetson87}. Sources at least 4$\sigma$ above the
104: background were extracted. A sharpness cut-off helped eliminate extended
105: sources and remaining outliers. A pixel-phase-dependent photometric correction
106: was applied to the 3.6 $\mu$m fluxes \citep{Reach05}. The flux estimates are
107: color corrected using a 5000 K blackbody, as specified in the IRAC Data
108: Handbook, version 3.0. Zero magnitudes were taken from the same reference, and
109: flux uncertainties include both the DAOphot errors and calibration errors
110: \citep{Reach05}. False star tests indicate that the catalogue is complete to
111: at least 90\% in the core of the cluster down to $[3.6]=17.5$ mag and
112: $[8]=16.5$ mag.
113:
114: %=========================================================================== 3
115: \section{Cluster membership}
116:
117: Proper motions have been measured for many stars in and around NGC\,6791
118: \citep{Monet03,Zacharias04,Dias06}. We designate a star a proper motion
119: non-member if the absolute proper motion is $>20$ mas yr$^{-1}$ in the
120: \citet{Monet03} catalogue, $>14$ mas yr$^{-1}$ in the \citet{Zacharias04}
121: catalogue, or its membership probability was set zero in the \citet{Dias06}
122: catalogue. Fig.\ \ref{f2} displays the resulting selection of possible members
123: (top row) and proper motion non-members (bottom row). The figure illustrates
124: the choice of selection criteria for the \citet{Monet03} and
125: \citet{Zacharias04} catalogues, which are based on the clearly defined clump
126: of stars with small proper motions (possible members) compared to the much
127: more widely distributed foreground population. The \cite{Dias06} selection
128: criteria include a positional element. Note that if a star was excluded as a
129: member by any one of these works it is labelled a non-member in all three
130: panels of Fig.\ \ref{f2}.
131:
132: The proper motion non-members are identified with crosses in the {\it Spitzer}
133: [3.6], [3.6]--[8] color-magnitude diagram (Fig.\ \ref{f3}). This eliminates
134: for instance the very brightest star \citep[\#2240 in][]{Stetson03} and the
135: bright very red star at [3.6]=9.7 mag, [3.6]--[8]=2.4 mag \citep[\#2349
136: in][]{Stetson03}. Objects in the faint, red corner of the diagram are
137: background galaxies; slightly extended objects with similar colors are
138: prominent on the {\it Spitzer} images. As expected for galaxies, these objects
139: are spread uniformly across the images.
140:
141: To eliminate field stars with no or ambiguous proper motion information, a
142: further criterion is applied on the basis of the optical V, B--V diagram
143: \citep[Fig.\ \ref{f4}, using data from][]{Stetson03}: the main sequence, RGB
144: and red clump (around $V=14.6$ mag, $B-V\simeq1.35$ mag) of NGC\,6791 are well
145: separated from the bulk of the foreground main-sequence stars. In our
146: subsequent analysis we keep all remaining possible cluster members that have
147: proper motion information, photometry in all {\it Spitzer} IRAC bands and the
148: V-band, and that are located within $0.1^\circ$ projected distance from the
149: cluster centre (290.23, +37.77).
150:
151: %=========================================================================== 4
152: \section{Evidence for circumstellar dust emission}
153:
154: In the absence of longer-wavelength data, [3.6]--[8] is the most sensitive
155: {\it Spitzer} color to the presence of circumstellar dust. Dust causes a
156: reddening of this color, firstly, due to emission at 8 $\mu$m, and secondly,
157: due to extinction at 3.6 $\mu$m if the dust envelope is very optically thick.
158: The RGB in NGC\,6791 is surprisingly narrow in [3.6]--[8] color (Fig.\
159: \ref{f3}) down to at least $[3.6]\simeq13.5$ mag, i.e.\ more than four
160: magnitudes below the RGB tip. Compared to the distribution mirrored around the
161: peak (Fig.\ \ref{f5}), at $[3.6]-[8]=-0.07$ mag, there are 14 out of 72 stars
162: ($19\pm5$\%) that are redder than expected from a distribution which is
163: symmetric in color. Not only is this but a small fraction, the colors are also
164: only marginally red, with $\Delta([3.6]-[8])<0.1$ mag for all but one with
165: $\Delta([3.6]-[8])\simeq0.22$ mag. The latter is at the faint end of this RGB
166: selection; indeed, there is no sign of the brightest stars being more likely
167: to show an IR excess as it is observed, e.g., in $\omega$\,Centauri
168: \citep{Boyer08}.
169:
170: The fraction of dusty stars along the upper 4.5 mag of the RGB in NGC\,6791
171: may be compared with that in the globular cluster 47\,Tuc \citep{Origlia07}.
172: There, $\sim340$ stars are counted in the central part of 47\,Tuc, amongst
173: which 90 are dusty (26\%). In the outskirts of 47\,Tuc, there are $\sim160$
174: stars, amongst which 9 are dusty (6\%). These statistics may be affected by
175: blending, in the central part, and field star contamination, in the outskirts.
176: Nonetheless, there is no evidence for an abnormally high fraction of dusty
177: stars in NGC\,6791 despite an order of magnitude higher metal content than in
178: 47\,Tuc.
179:
180: Following \citet{Groenewegen06}, $\Delta([3.6]-[8])=0.1$ mag corresponds to a
181: mass-loss rate of $1-2\times10^{-8}$ M$_\odot$ yr$^{-1}$ for an M-type star
182: with aluminium-oxide or silicate dust and a luminosity $L=3,000$ L$_\odot$,
183: wind speed $v=10$ km s$^{-1}$, and dust-to-gas mass ratio $\psi=0.005$. In the
184: extreme case of an M10-type star with pure aluminium-oxide dust the
185: corresponding mass-loss rate is nearly $5\times10^{-8}$ M$_\odot$ yr$^{-1}$,
186: but to our knowledge no such RGB star exists. The mass-loss rate scales
187: approximately as $\dot{M}\propto v \sqrt{L}/\psi$ \citep{Vanloon00}. For a
188: dust-driven wind one expects the dust-to-gas ratio in the metal-rich stars of
189: NGC\,6791 to be higher by a factor $\sim2.5$, and the wind speed by a factor
190: $\sim1.6$ \citep{Marshall04}. The luminosity range along this part of the RGB
191: is $\sim10^{2-3}$ L$_\odot$, which lowers the Groenewegen value for $\dot{M}$
192: by up to a factor 5. If the winds are as fast as sometimes seen in
193: chromospherical lines of warmer RGB stars \citep{Mauas06,Mcdonald07} then
194: $\dot{M}$ could be higher. We thus settle on a conservative estimate of
195: $\dot{M}\ltsimeq10^{-8}$ M$_\odot$ yr$^{-1}$. The mass-loss rate may still be
196: higher if the dust-to-gas ratio is lower than assumed --- in spite of the high
197: abundance of condensable material. However, our estimate compares favorably
198: with the empirical relationships of \citet{Reimers77} and \citet{Schroeder05},
199: which yield $\dot{M}\sim10^{-8}$ and a few $10^{-9}$ M$_\odot$ yr$^{-1}$,
200: respectively, for a typical RGB star of 1 M$_\odot$, 4000 K and 500 L$_\odot$.
201: This suggests that the basic assumptions we made are not unreasonable.
202:
203: If such mass loss is sustained over 19\% (the observed fraction of dusty
204: stars) of the time it takes to evolve along this part of the RGB,
205: $t\sim8\times10^7$ yr \citep{Marigo08}, then they will lose
206: $\Delta M\ltsimeq0.2$ M$_\odot$ --- insufficient to avoid the helium flash.
207: These stars were born with a mass $M_{\rm initial}\simeq1.1$ M$_\odot$, and
208: would need to shed $\Delta M_{\rm implied}\simeq0.7$ M$_\odot$ to produce
209: white dwarfs of the masses measured by \citet{Kalirai07}. Both
210: \citet{Kalirai07} and \citet{Hansen05} point out that not all RGB stars may
211: avoid the helium flash. Given the above estimates of $\dot{M}$, fraction of
212: dusty stars and RGB lifetime, the {\it Spitzer} data suggest that no more than
213: $\dot{M} t/\Delta M_{\rm implied} \times 19$\% or $\ltsimeq22$\% of RGB stars
214: in NGC\,6791 can sustain the required heavy mass loss.
215:
216: There is a possibility that we have missed a star at the very tip of the RGB
217: losing mass at a high rate, but only briefly before it leaves the RGB. Indeed,
218: dust-accompanied mass loss, at rates as high as $10^{-6}$ M$_\odot$ yr$^{-1}$,
219: is seen to happen predominantly near the tip of the RGB
220: \citep[][]{Vanloon06,Origlia07,Boyer08}. It is unclear, however, whether such
221: star has sufficient time to lose enough mass to avoid the helium flash. In any
222: case, despite the important effect on the remaining mantle mass and its
223: subsequent appearance on the horizontal branch, the white dwarf descendant
224: would have a mass that is indistinguishable from that of a CO core following
225: the helium flash, as the luminosity on the RGB is a direct measure of the
226: helium core's mass \citep{Castellani93}.
227:
228: %=========================================================================== 5
229: \section{The RGB luminosity function}
230:
231: To explain white dwarf masses 0.1 M$_\odot$ lower than the core mass at the
232: tip of the RGB, the luminosity function must be depleted over 0.6 dex in $\log
233: {\rm L}$ \citep[Table 2 in][]{Castellani93}, corresponding to 1.5 mag. We
234: demonstrate below that such depletion in NGC\,6791 is not corroborated by the
235: measurements.
236:
237: We constructed cumulative luminosity functions (LFs), in both the {\it
238: Spitzer} IRAC bands and the V-band (Fig.\ \ref{f6}). The IR and V-band LFs
239: were aligned such that the very obvious bump due to the red clump coincides
240: --- this required a shift towards brighter values of the V-band LF by 3 mag.
241: This difference, and the much steeper bright end to the V-band LF compared to
242: the IR LFs, are due mainly to the bolometric corrections to the V-band
243: (BC$_{\rm V}$) of the increasingly cooler RGB stars as they approach the RGB
244: tip \citep[a small contribution is due to the extinction,
245: $E(B-V)\simeq0.14$;][]{Kalirai07}.
246:
247: The observed LFs compare favorably with model LFs from the Padua group
248: \citep{Marigo08}, displayed in the lower panel of Fig.\ \ref{f6}. The models,
249: displayed for a distance modulus of 13.0 mag, show the same shape and location
250: of the red clump and main sequence (the rise in the V-band LF at the faint end
251: of the scale). The models, which do not include mass loss along the RGB,
252: reproduce the observed difference at the bright end between the IR and V-band
253: LFs rather well. The most metal-rich Padua model is with $Z=0.03$ not quite as
254: metal-rich as NGC\,6791 ($Z=0.05$), and if compared to the $Z=0.02$ Padua
255: model it is evident that a $Z=0.05$ model LF would be truncated at even
256: fainter magnitudes, in close agreement with the observed LF. This is
257: corroborated by the good fit to the optical colour-magnitude diagram of a
258: $Z=0.06$ isochrone using the \citet{Reimers77} mass-loss prescription
259: \citep{Claret07}. \citet{Kalirai07} find that the V-band LF of NGC\,6791 is
260: depleted with respect to the (sub-)solar metallicity clusters Berkeley\,17 and
261: M\,67, over the top magnitude of the RGB. They do not, however, consider that
262: BC$_{\rm V}$ also differs by $>1$ mag \citep{Marigo08}, nor the effects from
263: contamination by field stars.
264:
265: With regard to the IR LF, which more accurately measures the bolometric output
266: of these stars, the difference with the models can be entirely explained by
267: stochastic effects. The NGC\,6791 cluster is not populous enough to harbor
268: many stars near the RGB tip; while there are few tip-RGB stars present now, it
269: is conceivable that one or two more may have been present several million
270: years ago or will be present several million years from now. It can also not
271: be excluded that a tip-RGB star was missed, for instance, because it had
272: fallen outside the {\it Spitzer} coverage, or because of an inaccuracy in the
273: proper motion measurement, or for other observational reasons. To demonstrate
274: the feebleness of any indication of a prematurely truncated RGB, the 8-$\mu$m
275: LF is shown again after adding a mere two stars at the tip of the RGB. This
276: brings the observed LF in excellent agreement with the theoretical prediction.
277: Indeed, the relative numbers of post-helium-flash red clump stars compared to
278: those at similar luminosities on the RGB are well reproduced (Fig.\ \ref{f6}),
279: suggesting that the vast majority of RGB stars do go on to undergo the helium
280: flash.
281:
282: %=========================================================================== 6
283: \section{Conclusions}
284:
285: We conclude that there is good agreement between the observed and expected
286: optical and infrared luminosity functions along the RGB in NGC\,6791, and that
287: there is little circumstellar dust observed around RGB stars in that old
288: metal-rich massive open cluster. Hence there is no direct evidence supporting
289: the suggestion that metal-rich stars avoid the helium flash at the tip of the
290: RGB and become undermassive helium-core white dwarfs as a result of
291: particularly strong stellar winds.
292:
293: \acknowledgments
294:
295: We thank both referees for their thorough reviews. MLB is supported by the
296: University of Minnesota Louise T.\ Dosdall Fellowship. IM acknowledges an STFC
297: studentship.
298:
299: %% {\it Facilities:} \facility{Spitzer:IRAC}
300:
301: \begin{thebibliography}{}
302: \bibitem[Bedin et al.(2008)]{Bedin08} Bedin, L. R., Salaris, M., Piotto, G.,
303: Cassisi, S., Milone, A. P., Anderson, J., \& King, I. R.\ 2008, ApJ Letters in
304: press (arXiv:0804.1792)
305: \bibitem[Boyer et al.(2008)]{Boyer08} Boyer, M. L., McDonald, I., van Loon, J.
306: Th., Woodward, C. E., Gehrz, R. D., Evans, A., \& Dupree, A. K.\ 2008, AJ,
307: 135, 1395
308: \bibitem[Castellani \& Castellani (1993)]{Castellani93} Castellani, M., \&
309: Castellani, V.\ 1993, ApJ, 407, 649
310: \bibitem[Claret (2007)]{Claret07} Claret, A.\ 2007, A\&A, 467, 1389
311: \bibitem[Dias et al.(2006)]{Dias06} Dias, W. S., Assafin, M., Fl\'orio, V.,
312: Alessi, B. S., \& L\'{\i}bero, V.\ 2006, A\&A, 446, 949
313: \bibitem[Diaz-Miller (2007)]{Diazmiller07} Diaz-Miller, R. I.\ 2007, in: The
314: Future of Photometric, Spectrophotometric and Polarimetric Standardization,
315: ed.\ C. Sterken, ASPC, 364, p81
316: \bibitem[Dupree et al.(2007)]{Dupree07} Dupree, A. K., Li, T. Q., \& Smith, G.
317: H.\ 2007, AJ, 134, 1348
318: \bibitem[Fazio et al.(2004)]{Fazio04} Fazio, G. G., et al.\ 2004, ApJS, 154,
319: 10
320: \bibitem[Gehrz \& Woolf (1971)]{Gehrz71} Gehrz, R. D., \& Woolf, N. J.\ 1971,
321: ApJ, 165, 285
322: \bibitem[Groenewegen (2006)]{Groenewegen06} Groenewegen, M. A. T.\ 2006, A\&A,
323: 448, 181
324: \bibitem[Hansen (2005)]{Hansen05} Hansen, B. M. S.\ 2005, ApJ, 635, 522
325: \bibitem[Kalirai et al.(2007)]{Kalirai07} Kalirai, J. S., et al.\ 2007, ApJ,
326: 671, 748
327: \bibitem[Makovoz \& Marleau (2005)]{Makovoz05} Makovoz, D., \& Marleau, F. R.\
328: 2005, PASP, 117, 1113
329: \bibitem[Marigo et al.(2008)]{Marigo08} Marigo, P., Girardi, L., Bressan, A.,
330: Groenewegen, M. A. T., Silva, L., \& Granato, G. L.\ 2008, A\&A in press
331: (arXiv:0711.4922)
332: \bibitem[Marshall et al.(2004)]{Marshall04} Marshall, J. R., van Loon, J. Th.,
333: Matsuura, M., Wood, P. R., Zijlstra, A. A., \& Whitelock, P. A.\ 2004, MNRAS,
334: 355, 1348
335: \bibitem[Mauas et al.(2006)]{Mauas06} Mauas, P. J. D., Cacciari, C., \&
336: Pasquini, L.\ 2006, A\&A, 454, 609
337: \bibitem[McDonald \& van Loon (2007)]{Mcdonald07} McDonald, I., \& van Loon,
338: J. Th.\ 2007, A\&A, 476, 1261
339: \bibitem[Monet et al.(2003)]{Monet03} Monet, D. G., et al.\ 2003, AJ, 125, 984
340: \bibitem[Origlia et al.(2007)]{Origlia07} Origlia, L., Rood, R. T., Fabbri,
341: S., Ferraro, F. R., Fusi Pecci, F., \& Rich, R. M.\ 2007, ApJ, 667, L85
342: \bibitem[Reach et al.(2005)]{Reach05} Reach, W. T., et al.\ 2005, PASP, 117,
343: 978
344: \bibitem[Reimers (1977)]{Reimers77} Reimers, D.\ 1977, A\&A, 61, 217
345: \bibitem[Rood (1973)]{Rood73} Rood, R. T.\ 1973, ApJ, 184, 815
346: \bibitem[Schr\"oder \& Cuntz (2005)]{Schroeder05} Schr\"oder, K.-P., \& Cuntz,
347: M.\ 2005, ApJ, 630, L73
348: \bibitem[Stetson (1987)]{Stetson87} Stetson, P. B.\ 1987, PASP, 99, 191
349: \bibitem[Stetson et al.(2003)]{Stetson03} Stetson, P. B., Bruntt, H., \&
350: Grundahl, F.\ 2003, PASP, 115, 413
351: \bibitem[van Loon (2000)]{Vanloon00} van Loon, J. Th.\ 2000, A\&A, 354, 125
352: \bibitem[van Loon et al.(2006)]{Vanloon06} van Loon, J. Th., McDonald, I.,
353: Oliveira, J. M., Evans, A., Boyer, M. L., Gehrz, R. D., Polomski, E., \&
354: Woodward, C. E.\ 2006, A\&A, 450, 339
355: \bibitem[Werner et al.(2004)]{Werner04} Werner, M. W., et al.\ 2004, ApJS,
356: 154, 1
357: \bibitem[Zacharias et al.(2004)]{Zacharias04} Zacharias, N., Urban, S. E.,
358: Zacharias, M. I., Wycoff, G. L., Hall, D. M., Monet, D. G., \& Rafferty, T.
359: J.\ 2004, AJ, 127, 3043
360: \end{thebibliography}
361:
362: \clearpage
363:
364: \begin{figure}
365: \plotone{f1.eps}
366: \caption{{\it Spitzer} IRAC composite of the region around NGC\,6791
367: (indicated with the cross). Red, yellow-green, green-cyan, and blue are 8,
368: 5.8, 4.5, and 3.6 $\mu$m, respectively.}
369: \label{f1}
370: \end{figure}
371:
372: \begin{figure}
373: \plotone{f2.eps}
374: \caption{Proper motion maps based on data from \citet{Monet03},
375: \citet{Zacharias04} and \citet{Dias06}, which are used to select against
376: non-members of NGC\,6791 (bottom row of panels).}
377: \label{f2}
378: \end{figure}
379:
380: \begin{figure}
381: \plotone{f3.eps}
382: \caption{{\it Spitzer} IRAC color-magnitude diagram of NGC\,6791. The RGB is
383: well populated until the tip at [3.6]$\simeq9$ mag, and the red clump is
384: visible around [3.6]=11.5 mag. Hardly any star shows evidence for reddening
385: due to a contribution of circumstellar dust emission at 8 $\mu$m.}
386: \label{f3}
387: \end{figure}
388:
389: \begin{figure}
390: \plotone{f4.eps}
391: \caption{Optical color-magnitude diagram \citep{Stetson03}, used to exclude
392: remaining field stars left of the dividing line.}
393: \label{f4}
394: \end{figure}
395:
396: \begin{figure}
397: \plotone{f5.eps}
398: \caption{{\it Spitzer} IRAC color distribution along the RGB in NGC\,6791.
399: Overplotted (dotted) is the same histogram, but mirrored with respect to its
400: peak. This more clearly shows which stars are redder than the bulk of RGB
401: stars.}
402: \label{f5}
403: \end{figure}
404:
405: \begin{figure}
406: \plotone{f6.eps}
407: \caption{{\it Spitzer} IRAC and optical V-band \citep{Stetson03} luminosity
408: functions in NGC\,6791. The 8-$\mu$m luminosity function is also plotted after
409: adding just two more stars near the tip of the RGB. A comparison with 8-Gyr
410: Padua models \citep{Marigo08} shows good agreement, especially considering
411: that NGC\,6791 is more metal-rich. Stochastics introduced by sampling the
412: small number of stars on the upper RGB can fully explain the discrepancy
413: between the data and models.}
414: \label{f6}
415: \end{figure}
416:
417: \end{document}
418: