1: \documentclass{emulateapj}
2:
3:
4: \shorttitle{Long Secondary Periods and Binarity in Red Giant Stars}
5: \shortauthors{Soszy{\'n}ski}
6:
7: \begin{document}
8:
9: \title{Long Secondary Periods and Binarity in Red Giant Stars}
10:
11: \author{I. Soszy{\'n}ski}
12: \affil{Warsaw University Observatory, Al.~Ujazdowskie~4, 00-478~Warszawa, Poland}
13: \email{soszynsk@astrouw.edu.pl}
14:
15: \begin{abstract}
16: Observational arguments supporting the binary explanation of the long
17: secondary periods (LSP) phenomenon in red giants are presented.
18: Photometry of about 1200 semiregular variables with the LSP in the Large
19: Magellanic Cloud are analyzed using the MACHO and OGLE photometry. For
20: about 5\% of these objects additional ellipsoidal-like or eclipsing-like
21: modulation with the same periods as the LSP is detectable. These
22: double-humped variations are usually shifted in phase comparing to the LSP
23: light curves. I discuss the model of binary system with a red giant as the
24: primary component and a low-mass object as the secondary one. The mass lost
25: by the red giant through the wind follows the spiral pattern in the orbit
26: around the primary star and obscures it causing the LSP variations.
27: \end{abstract}
28: \keywords{stars: binaries: close --- planetary systems --- stars: AGB and
29: post-AGB}
30:
31: \vspace{0.3cm}
32: \section{Introduction}
33:
34: Among numerous classes of variable stars only one type of large-amplitude
35: stellar variability remains completely unexplained. This is the long
36: secondary periods (LSP) observed in luminous red giant stars. The LSP
37: variability with periods between 200 and 1500 days and with $V$-band
38: amplitudes up to 1~mag occurs in $\sim30$\% of Semiregular Variables (SRVs)
39: and OGLE Small Amplitude Red Giants (OSARGs). This phenomenon has been
40: known for decades \citep{pg54,hou63}, but an interest in the stars with the
41: LSP has been renewed since \citet{woo99} showed that these objects follow a
42: period--luminosity (PL) relation (sequence D).
43:
44: In recent years our knowledge about the LSP phenomenon has significantly
45: increased, although its origin remains a mystery. \citet{hin02} and
46: \citet{woo04} studied spectral features of several Galactic stars with the
47: LSP and detected radial velocity variations with amplitudes of a few km/s
48: which agree with photometric long-period variations. \citet{sos04} used
49: Optical Gravitational Lensing Experiment (OGLE) data to select and analyze
50: close binary systems in the Large Magellanic Cloud (LMC) with a red giant
51: as one of the components. They noticed that the sequence D in the PL
52: diagram overlaps and is a direct continuation of the PL sequence of
53: ellipsoidal and eclipsing red giants\footnote{if the real orbital periods
54: of binary variables are considered, i.e. periods two times longer than
55: obtained automatically.} (sequence E), suggesting that the LSP phenomenon
56: is related to binarity. \citet{sos04} also showed that some evident
57: ellipsoidal red giants exhibit simultaneously OSARG-type variability, thus,
58: it was directly demonstrated that in some cases the binary explanation of
59: the LSP is true. Nevertheless, there is no doubt that the bulk of the LSP
60: variables are not typical ellipsoidal or eclipsing binaries.
61:
62: \citet{sos05} discovered that the sequence D split into two ridges in the
63: period -- Wesenheit index ($W_I=I-1.55(V-I)$) plane, which corresponds to
64: the spectral division into oxygen-rich and carbon-rich AGB stars. The same
65: feature was noticed for Miras and SRVs (sequences C and C$'$), and the
66: sequence D contains relatively much smaller number of carbon-rich stars.
67:
68: Recently \citet{der06} presented a period--luminosity--amplitude analysis
69: of variable red giants in the LMC. They examined amplitudes of ellipsoidal,
70: LSP and Mira-like variables using MACHO red ($R_M$) and blue ($B_M$)
71: photometry. The amplitude distribution for the LSP stars turned out to be
72: different than for ellipsoidal and pulsating variables, but blue-to-red
73: amplitude ratios of the LSP stars (typically 1.3) is more similar to this
74: quantity in pulsating variables ($\sim1.4$) than to ellipsoidal/eclipsing
75: binaries ($\sim1.1$). This last feature is used in present work for
76: separation of the LSP and ellipsoidal/eclipsing variability in the same
77: light curves.
78:
79: Various hypotheses have been proposed to explain the origin of the LSP
80: variability: rotation of a spotted star, episodic dust ejections, a radial
81: and non-radial pulsation and binary companions including planets or brown
82: dwarfs. \citet{woo99} suggested that the sequence D stars are components of
83: semidetached binary systems. The matter lost by the AGB star forms a dusty
84: cloud around the companion, and regularly obscures the primary component
85: causing the LSP variability. \citet{hin02} and then \citet{ow03} mentioned
86: that the radial velocity measurements are consistent only with the binary
87: or pulsation explanations of the LSP. \citet{woo04} ruled out the binary
88: hypothesis, because a short ($\sim1000$ yr) timescale on which the
89: companion should merge with the red giant. They suggested that the most
90: likely explanation of the LSP are low degree g$^+$ pulsation modes trapped
91: in the outer radiative layers of the star.
92:
93: The main goal of this paper is to find observational evidences for or
94: against the binary explanation of the LSP. Since the PL sequence populated
95: by the ellipsoidal and eclipsing variables overlaps with the LSP sequence
96: \citep{sos04}, it should be possible to find stars revealing simultaneously
97: both types of variability. If periods are the same, the LSP phenomenon must
98: be related to binarity. If not, the LSP is presumably caused by another
99: reason.
100:
101: \vspace{0.5cm}
102: \section{Data Analysis}
103:
104: Since the LSP are sometimes as long as 1500 days, presented analysis is
105: based on observational data originated in two sources: MACHO and OGLE
106: surveys. Merged light curves from both projects covered 15 years of
107: observations: from 1992 to 2006. The sequence D stars selected by
108: \citet{sos04} in the OGLE database were cross-identified with objects
109: collected in the MACHO
110: archive\footnote{http://wwwmacho.mcmaster.ca/Data/MachoData.html}. I found
111: about 1200 counterparts of the previously selected variables. Then,
112: $R_M$-band MACHO observations have been merged with the OGLE points by
113: scaling amplitudes and shifting zero points of the photometry. The
114: parameters of this transformation have been found by the least square
115: fitting to the measurements obtained between 1997 and 2000, i.e., when both
116: projects observed the LMC fields at the same time.
117:
118: \vspace{0.3cm}
119: \subsection{Searching for orbital periods different than the LSP}
120:
121: Searching for ellipsoidal or eclipsing variability with different periods
122: than the LSP was a relatively easy procedure. For each object a third
123: order Fourier series was fitted to the LSP light curve and subtracted from
124: the points. Then, the period search was performed for the residual
125: data. This procedure was repeated until four additional periods per star
126: have been found.
127:
128: \begin{figure}
129: \includegraphics[width=8.7cm]{f1.ps}
130: \vspace{-0.5cm}
131: \caption{Light curves of LSP stars with secondary periods in the binarity
132: range. Left column shows light curves folded with the LSP, and right
133: column presents these data after subtracting the LSP variability and
134: folded with different periods.}
135: \end{figure}
136:
137: From these results I selected and visually inspected light curves with
138: periods lying between sequences C and D, i.e., where one can expect
139: ellipsoidal or eclipsing variables (automatic procedure gives a half of
140: orbital periods). The best seven light curves of that type are plotted in
141: Fig.~1. Phased LSP light curves are shown in the left column, the right
142: column presents the light curves of these objects after subtracting the LSP
143: variations and phased with periods two times longer than obtained
144: automatically.
145:
146: As one can see none of these secondary modulations is distinct,
147: indisputable eclipsing or ellipsoidal light curve. The most likely
148: explanation of these modulations are variations of phases or amplitudes of
149: the LSP variability which produce such ``artificial'' variability in
150: residual light curves. A very similar behavior occurred in eclipsing binary
151: described in Section~3.
152:
153: \vspace{0.3cm}
154: \subsection{Searching for orbital periods the same as the LSP}
155:
156: If the LSP is related to binarity, one can expect that a number of objects
157: shows ellipsoidal or eclipsing variability of the same period as the LSP.
158: Unfortunately, detecting such a modulation in the LSP light curves is not
159: an easy task, because (i) the ellipsoidal light curves have usually much
160: smaller amplitudes than LSP, (ii) shorter semiregular variability is
161: superimposed, (iii) the LSP light curves often change amplitudes, phases
162: and periods. A careful investigation of the LSP light curves reveals that
163: some of them show shallow secondary minima, what may be a sign of
164: ellipsoidal variations superimposed on the LSP. However, since the LSP
165: light curves appear in several variants, it is possible that such behavior
166: is not related to binarity.
167:
168: To separate LSP and possible ellipsoidal/eclipsing variations I used a
169: feature noticed by \citet{der06}. They studied MACHO $B_M$ and $R_M$
170: photometry of long period variables in the LMC and found that blue-to-red
171: amplitude ratios are different for ellipsoidal and LSP variables. In the
172: ellipsoidal red giants, where the variability is dominated by the geometric
173: changes, amplitudes in different filters are very similar. The median value
174: of $A(B_M)/A(R_M)$ is equal to 1.1. For the LSP stars $A(B_M)/A(R_M)$ is
175: more similar to pulsating variables and equal on average to 1.3. It means
176: that if the amplitudes of the $R_M$-band light curves are scaled by a
177: factor 1.3 and the $B_M$ observations are subtracted, the LSP variations
178: will be canceled, but not the possible ellipsoidal/eclipsing modulation.
179:
180: \begin{figure*}
181: \hspace{1cm}
182: \includegraphics[width=16cm]{f2.ps}
183: \vspace{-2.5cm}
184: \caption{Six LSP star exhibiting additional double-humped
185: variations. Upper diagram in each panel shows MACHO $R_M$-band
186: photometry, middle diagram contains $B_M$-band data, lower diagram
187: presents the residual data after scaling $R_M$ and subtracting $B_M$
188: light curves. The dashed lines show the phases of minimum brightness
189: during the LSP variations. Note that each light curve of given star is
190: folded with the same period.}
191: \end{figure*}
192:
193: The procedure was as follows. The $B_M$ and $R_M$-band amplitudes were
194: determined by fitting the spline function to the folded light curves. Then,
195: the $R_M$ magnitudes were converted into flux and linearly scaled to obtain
196: the same amplitude (in magnitudes) as in the $B_M$ bandpass. The flux was
197: again converted into magnitudes, and $B_M$ measurements were subtracted
198: point by point from scaled $R_M$ data. The final step of the procedure was
199: fitting a sinusoid of a period 1 year, and subtracting the function from
200: the residual data. This way I removed a differential refraction effect
201: distinctly visible in MACHO data.
202:
203: For the vast majority of light curves the LSP and pulsation (semiregular)
204: variability subtracted very well. No significant long-period variability
205: was detectable in the residual points. However, for about 5\% of stars I
206: found clear periods equal to half the LSP, i.e. I noticed double-humped
207: curves with periods the same as of the LSP. All these residual data are
208: available from the electronic edition of the Astrophysical Journal.
209: The most prominent curves of this type are presented in Fig.~2. In these
210: cases the residual data seem to arrange in eclipsing-like or
211: ellipsoidal-like light curves.
212:
213: Fig.~3 shows period--$W_I$ diagram (where $W_I=I-1.55(V-I)$ is the
214: reddening-free Wesenheit index) for stars from sequences D and E. Full
215: circles show the position of the LSP stars with the double-humped
216: curves. Note that these objects appear along the complete length of
217: sequence~D, although there is an overdensity for shorter-period stars,
218: i.e. where sequences D and E overlap. Note also that double-humped
219: objects populate both, O-rich and C-rich, sequences of the LSP stars
220: \citep{sos05}.
221:
222: \begin{figure}
223: \vspace{-0.6cm}
224: \hspace{-2.0cm}
225: \includegraphics[width=12cm]{f3.ps}
226: \vspace{-8cm}
227: \caption{Period--$W_I$ diagram for the LSP stars and ellipsoidal red giants
228: in the LMC. Crosses show the LSP stars (sequence D), empty circles
229: indicate ellipsoidal red giants (sequence E), filled circles mark the LSP
230: stars with double-humped curves visible in residual data.}
231: \end{figure}
232:
233: It is worth mentioning that the described procedure does not demand any
234: assumptions concerning periods or light curve shapes. It is a simple
235: transformation, the same for each observing point. The only parameter of
236: this transformation is a blue-to-red amplitude ratio of the LSP
237: variability, but I checked that the double-humped residual light curves are
238: visible for relatively wide range of this parameter, so it is not very
239: important to measure amplitudes very accurately.
240:
241:
242: \vspace{0.3cm}
243: \section{Discussion}
244:
245: I have shown that about 5\% of sequence D objects have ellipsoidal or
246: eclipsing-like modulation with periods the same as the LSP. In the sample
247: of 1200 sequence D stars I did not find any distinct ellipsoidal or
248: eclipsing binary with period different than the LSP. Although various
249: scenarios are still possible, it is justified to re-invoke the binary
250: hypothesis proposed by \citet{woo99}. In this explanation the AGB star in a
251: close binary system losses mass to the secondary component. The matter
252: forms a dusty cloud around and behind the companion and regularly obscures
253: the primary star.
254:
255: An argument for this hypothesis is a phase lag between
256: ellipsoidal/eclipsing and the LSP variations clearly visible in
257: Fig.~2. Only one object of six -- the star with the shortest period -- does
258: not exhibit such behavior. I checked that it might be a rule for the
259: shortest period sequence D variables. For longer periods the minima of the
260: LSP brightness variation occur about 0.05--0.10 of a cycle after the minima
261: of the ellipsoidal/eclipsing light curves. Consistent results were
262: presented by \citet{woo01}, who discovered a phase lag of $\sim$1/8 between
263: LSP light curves and radial velocity curves. Such phase offsets between
264: ellipsoidal/eclipsing and LSP minima agree very well with hydrodynamical
265: simulations of a wind driven accretion flow in binary systems
266: \citep{tj93,mm98,nag04}. These models predict that a matter lost by a red
267: giant in a binary system follows the spiral pattern with maximum density
268: located behind the secondary component.
269:
270: An exhaustive discussion about possible origins of the LSP was done by
271: \citet{woo04}. They argued that the merger timescale for red giants and its
272: companion in close binary systems is of the order of $10^3$ years, while
273: the lifetime in the AGB phase is two orders of magnitude longer. Thus, 30\%
274: of SRVs showing the LSP is highly inconsistent with the binary scenario.
275:
276: However, the mass transfer in a binary system may be due to the Roche-lobe
277: overflow \citep[e.g.][]{pac71}, or through the stellar wind
278: \citep[e.g.][]{abm87,tj93}. The former process tends to circularize the
279: orbits and to synchronize the spin of the stars with the orbital rotation,
280: which results in shrinking the orbits and finally in merging the
281: components. The latter phenomenon increases the eccentricity of the orbits
282: and, if the bulk of the mass lost by the red giants escapes from the system,
283: the distance between components may even increase. Thus, the main argument
284: of \citet{woo04} against the binary explanation may be not valid, if we
285: assume that the red giants in the binary systems do not fully fill their
286: Roche lobe, and the bulk of mass transfer is driven by the stellar
287: wind. The non-sinusoidal radial velocity curves observed by \citet{hin02}
288: and \citet{woo04}, which can be explained by the eccentricity of the
289: orbits, are in good agreement with this scenario.
290:
291: \begin{figure*}
292: \includegraphics[width=17cm]{f4.ps}
293: \vspace{-0.5cm}
294: \caption{Three upper panels show LSP stars with variable amplitudes. In the
295: lower panel the probable eclipsing binary is presented. Left and right
296: diagrams contain unfolded and folded light curves, respectively.}
297: \end{figure*}
298:
299: Presented hypothesis nevertheless requires that at least 30\% of AGB stars
300: exist in close binary systems. Moreover, while the studied LSP variables
301: have velocity amplitudes of only a few km/s \citep{hin02,woo04}, the
302: confirmed binary systems (sequence~E stars) appear to have velocity
303: amplitudes about ten times larger \citep{awc06}. The radial velocity
304: measurements suggest that the second component in the LSP stars may be a
305: brown dwarf. \citet{ret05} proposed that the Jupiter-like planets may
306: accrete the matter from its host star and increase mass into the brown
307: dwarfs range. This hypothesis would explain such large number of the LSP
308: cases among AGB stars, if we assume that planets with a separation of 1--5
309: AU are common. To test the binary hypothesis it would be interesting to
310: measure the radial velocity changes for the LSP stars with the
311: double-humped variations. If these stars are ellipsoidal or eclipsing
312: variables indeed, their velocity amplitudes should be similar to these
313: observed in sequence E stars, i.e. significantly larger than for the
314: remaining LSP stars.
315:
316: The brown dwarf scenario can also explain why only 5\% of our sample show
317: the double-humped variability. For low-mass secondary components the
318: ellipsoidal and eclipsing variations have too small amplitudes to be
319: detected in our procedure. One should remember that the residual data
320: obtained by scaling $R_M$ and subtracting $B_M$ magnitudes can show
321: variability with amplitudes of about 0.2 of the original ellipsoidal
322: amplitudes ($R_M$ were scaled by a factor of 1.3 and $A(B_M)/A(R_M)$ is on
323: average 1.1 for ellipsoidal variables). Moreover, presented model does not
324: assume that the red giant fills entirely the Roche lobe, because the bulk
325: of the mass flow is through a stellar wind, so the separation between the
326: components can be too large to cause significant ellipsoidal variability.
327: Note also that observed number of possible ellipsoidal and eclipsing
328: variables among the LSP variables is in agreement with relative number of
329: sequence~E stars. The LSP modulation is observed for about 30\% of the AGB
330: stars, so 5\% of these objects gives about 1-2\% of the whole
331: population. Exactly the same relative number of ellipsoidal or eclipsing
332: variables is observed for fainter red giants \citep{sos04}, so one should
333: not expect to detect many more such objects among brighter stars.
334:
335: Of course, it cannot be absolutely excluded that the double-humped
336: variations have different explanation than the binarity. However, the
337: non-binarity model of the LSP phenomenon has to explain why produced
338: residual light curves have ellipsoidal or eclipsing-like shapes, why it
339: appears only in a few percent of the sequence D stars, and why there is
340: phase lag between LSP and double-humped variations. All these facts can be
341: explained by the binary scenario.
342:
343: Additional clues on the origin of the LSP can be given by an analysis of
344: typical and peculiar cases of the light curves. A common feature of the LSP
345: variations is the modulation of the amplitudes and phases. A few such light
346: curves are shown in in Fig.~4. It seems that the phase shifts are
347: correlated with the depth of minima -- the deeper minimum occurs earlier
348: than the shallower one (see the folded light curve in Fig.~4a). Sometimes
349: the LSP variations completely disappear or appear in red giants which did
350: not show such modulation before (Fig.~4b). The conclusion that can be drawn
351: from these cases is that the regular (with no LSP) apparent luminosity of
352: the red giants is the same as in the maximum apparent luminosity of the LSP
353: variations, i.e. the LSP phenomenon decreases the total luminosity of the
354: star. This fact must be taken into consideration by any theory of the LSP
355: variability.
356:
357: The hypothesis of a binary system with a mass-losing AGB star seems to agree
358: with these results, because the obscuration by a cloud of matter reduces
359: the total luminosity of the star. The amplitude and phase changes are
360: likely connected with the variable mass loss rate. An interesting LSP light
361: curve is presented in Fig.~4c. Starting from the beginning of the
362: observations the depth and width of the minima were increasing. After
363: a few cycles this process affected the maximum of the light curve, and the
364: total apparent optical luminosity of the star dropped down. Then, the
365: object returned, more or less, to the previous stage. This behavior can be
366: interpreted as a sudden rise of a mass loss rate, which caused the whole
367: system to be hidden in the cloud of matter.
368:
369: In Fig.~4d I show a probable eclipsing binary system with the AGB star as
370: one of the components. This object presents striking similarities to the
371: LSP stars. First, it is located in the sequence D in the PL
372: diagram. Second, the light curve changes the amplitudes and phases of
373: variations. Third, after subtracting a function fitted to the primary
374: (eclipsing) variability I obtained secondary long-period variations caused
375: probably by modulation of the primary period. The same behavior I observed
376: for the LSP stars presented in Fig.~1.
377:
378: \vspace{0.3cm}
379: \section{Summary and Conclusions}
380:
381: In this paper I show that careful analysis of available data may shed new
382: light on the nature of the last unexplained type of stellar
383: variability. Arguments in favor of the binary explanation of the LSP
384: phenomenon are as follows:
385: \begin{enumerate}
386: \item There are no reliable examples of the LSP stars with ellipsoidal or
387: eclipsing variations with different periods.
388: \item At least 5\% of the sequence D stars exhibit ellipsoidal or
389: eclipsing-like variability with the same period as the LSP.
390: \item Phase lag between ellipsoidal and LSP variations agrees well with
391: models of wind accretion in binary systems.
392: \item The LSP light curves with variable amplitudes can be explained by a
393: variable mass-loss rate in binary systems.
394: \end{enumerate}
395:
396: The binary scenario can explain many features of the LSP variables. The PL
397: relation (which is a direct continuation of ``binary'' sequence E) is a
398: projection of a radius--luminosity relation for red giants. Positive
399: correlation between amplitudes of the LSP variability and mean luminosity
400: \citep{sos04,der06} may be caused by the increasing mass loss rate with
401: the luminosity. The dimming during the LSP minima are consistent with
402: the obscuration by a dusty cloud of matter orbiting the red giant. Finally,
403: the radial velocity variations are in agreement with eccentric motion of
404: the low-mass companion. The long-term project of radial velocity
405: measurements of selected sequence D stars have been recently finished
406: (P.~Wood, private communication). I expect that these data will
407: definitively solve the LSP problem.
408:
409: \acknowledgments
410: { I am deeply grateful to Professors W.~D.~Dziembowski, B.~Paczy{\'n}ski,
411: A.~Udalski, and Dr.~Z.~Ko{\l}aczkowski for the careful and critical
412: reading of the manuscript and many useful discussions. I thank the
413: anonymous referee for helpful comments. The paper was supported by the
414: Foundation for Polish Science through the ``Homing Programme''.
415:
416: This paper utilizes public domain data obtained by the MACHO Project,
417: jointly funded by the US Department of Energy through the University of
418: California, Lawrence Livermore National Laboratory under contract
419: No. W-7405-Eng-48, by the National Science Foundation through the Center
420: for Particle Astrophysics of the University of California under
421: cooperative agreement AST-8809616, and by the Mount Stromlo and Siding
422: Spring Observatory, part of the Australian National University.
423:
424: This publication is partly based on the OGLE observations obtained with
425: the Warsaw Telecope at the Las Campanas Observatory, Chile, operated by
426: the Carnegie Institution of Washington.}
427:
428:
429: \begin{thebibliography}{}
430: \bibitem[Adams et al.(2006)]{awc06} Adams, E., Wood, P.~R., \& Cioni, M.-R.\ 2006, Memorie della Societa Astronomica Italiana, 77, 537
431: \bibitem[Anzer et al.(1987)]{abm87} Anzer, U., Boerner, G., \& Monaghan, J.~J.\ 1987, \aap, 176, 235
432: \bibitem[Derekas et al.(2006)]{der06} Derekas, A., Kiss, L.~L., Bedding, T.~R., Kjeldsen, H., Lah, P., \& Szab{\'o}, G.~M.\ 2006, \apjl, 650, L55
433: \bibitem[Hinkle et al.(2002)]{hin02} Hinkle, K.~H., Lebzelter, T., Joyce, R.~R., \& Fekel, F.~C.\ 2002, \aj, 123, 1002
434: \bibitem[Houk(1963)]{hou63} Houk, N.\ 1963, \aj, 68, 253
435: \bibitem[Mastrodemos \& Morris(1998)]{mm98} Mastrodemos, N., \& Morris, M.\ 1998, \apj, 497, 303
436: \bibitem[Nagae et al.(2004)]{nag04} Nagae, T., Oka, K., Matsuda, T., Fujiwara, H., Hachisu, I., \& Boffin, H.~M.~J.\ 2004, \aap, 419, 335
437: \bibitem[Olivier \& Wood(2003)]{ow03} Olivier, E.~A., \& Wood, P.~R.\ 2003, \apj, 584, 1035
438: \bibitem[Paczy{\'n}ski(1971)]{pac71} Paczy{\'n}ski, B.\ 1971, \araa, 9, 183
439: \bibitem[Payne-Gaposhkin(1954)]{pg54} Payne-Gaposhkin, C.\ 1954, Harvard Annals, 113, No. 4
440: \bibitem[Retter(2005)]{ret05} Retter, A.\ 2005, Bulletin of the American Astronomical Society, 37, 1487
441: \bibitem[Soszy{\'n}ski et al.(2004)]{sos04} Soszy{\'n}ski, I., Udalski, A.,
442: Kubiak, M., Szyma{\'n}ski, M.~K., Pietrzy{\'n}ski, G., {\.Z}ebru{\'n}, K.,
443: Szewczyk, O., Wyrzykowski, {\L}., \& Dziembowski, W.~A.\ 2004, Acta
444: Astron., 54, 347
445: \bibitem[Soszy{\'n}ski et al.(2005)]{sos05} Soszy{\'n}ski, I., Udalski, A., Kubiak, M., Szyma{\'n}ski, M.~K., Pietrzy{\'n}ski, G., {\.Z}ebru{\'n}, K., Szewczyk, O., Wyrzykowski, {\L}., \& Ulaczyk, K.\ 2005, Acta Astron., 55, 331
446: \bibitem[Theuns \& Jorissen(1993)]{tj93} Theuns, T., \& Jorissen, A.\ 1993, \mnras, 265, 946
447: \bibitem[Wood et al.(1999)]{woo99} Wood, P.~R., et al. 1999, in IAU Symp.~191: Asymptotic Giant Branch Stars, eds. T.~Le~Bertre, A.~L{\'e}bre, and C.~Waelkens (San Francisco: ASP), 151
448: \bibitem[Wood et al.(2001)]{woo01} Wood, P.~R., Axelrod, T.~S., \& Welch, D.~L.\ 2001, ASP Conf.~Ser.~229: Evolution of Binary and Multiple Star Systems, eds. Ph.~Podsiadlowski, S.~Rappaport, A.~R.~King, F.~D'Antona, and L.~Burder, 233
449: \bibitem[Wood et al.(2004)]{woo04} Wood, P.~R., Olivier, E.~A., \& Kawaler, S.~D.\ 2004, \apj, 604, 800
450:
451: \end{thebibliography}
452:
453: \end{document}
454:
455: