1: \nofiles

2: \documentclass[dvips]{article}

3: \usepackage{supertabular,lscape,epsfig}

4: \usepackage{amssymb}

5: \usepackage{amsmath}

6: \DeclareSymbolFont{ppa}{OT1}{ppl}{m}{it}

7: \DeclareMathSymbol{\vv}{\mathalpha}{ppa}{'166}

8: 

9: \thinmuskip = 2mu plus 0mu minus 2mu

10: \medmuskip = 2.3mu plus 1mu minus 2.3mu

11: \thickmuskip = 2.6mu plus 2mu minus 2.6mu

12: 

13: \begin{document}

14: 

15: \newcommand{\dd}{\,{\rm d}}

16: \newcommand{\ie}{{\it i.e.},\,}

17: \newcommand{\etal}{{\it et al.\ }}

18: \newcommand{\eg}{{\it e.g.},\,}

19: \newcommand{\cf}{{\it cf.\ }}

20: \newcommand{\vs}{{\it vs.\ }}

21: \newcommand{\zdot}{\makebox[0pt][l]{.}}

22: \newcommand{\up}[1]{\ifmmode^{\rm #1}\else$^{\rm #1}$\fi}

23: \newcommand{\dn}[1]{\ifmmode_{\rm #1}\else$_{\rm #1}$\fi}

24: \newcommand{\upd}{\up{d}}

25: \newcommand{\uph}{\up{h}}

26: \newcommand{\upm}{\up{m}}

27: \newcommand{\ups}{\up{s}}

28: \newcommand{\arcd}{\ifmmode^{\circ}\else$^{\circ}$\fi}

29: \newcommand{\arcm}{\ifmmode{'}\else$'$\fi}

30: \newcommand{\arcs}{\ifmmode{''}\else$''$\fi}

31: \newcommand{\MS}{{\rm M}\ifmmode_{\odot}\else$_{\odot}$\fi}

32: \newcommand{\RS}{{\rm R}\ifmmode_{\odot}\else$_{\odot}$\fi}

33: \newcommand{\LS}{{\rm L}\ifmmode_{\odot}\else$_{\odot}$\fi}

34: 

35: \newcommand{\Abstract}[2]{{\footnotesize\begin{center}ABSTRACT\end{center}

36: \vspace{1mm}\par#1\par

37: \noindent

38: {~}{\it #2}}}

39: 

40: \newcommand{\TabCap}[2]{\begin{center}\parbox[t]{#1}{\begin{center}

41:   \small {\spaceskip 2pt plus 1pt minus 1pt T a b l e}   

42:   \refstepcounter{table}\thetable \\[2mm]

43:   \footnotesize #2 \end{center}}\end{center}}

44: 

45: \newcommand{\TableSep}[2]{\begin{table}[p]\vspace{#1}

46: \TabCap{#2}\end{table}}

47: 

48: \newcommand{\FigCap}[1]{\footnotesize\par\noindent Fig.\  %

49:   \refstepcounter{figure}\thefigure. #1\par}

50: 

51: \newcommand{\TableFont}{\footnotesize}

52: \newcommand{\TableFontIt}{\ttit}

53: \newcommand{\SetTableFont}[1]{\renewcommand{\TableFont}{#1}}

54: 

55: \newcommand{\MakeTable}[4]{\begin{table}[p]\TabCap{#2}{#3}

56:   \begin{center} \TableFont \begin{tabular}{#1} #4 

57:   \end{tabular}\end{center}\end{table}}

58: 

59: \newcommand{\MakeTableTop}[4]{\begin{table}[t]\TabCap{#2}{#3}

60:   \begin{center} \TableFont \begin{tabular}{#1} #4 

61:   \end{tabular}\end{center}\end{table}}

62: 

63: \newcommand{\MakeTableSep}[4]{\begin{table}[p]\TabCap{#2}{#3}

64:   \begin{center} \TableFont \begin{tabular}{#1} #4

65:   \end{tabular}\end{center}\end{table}}

66: \newcommand{\TabCapp}[2]{\begin{center}\parbox[t]{#1}{\centerline{

67:   \small {\spaceskip 2pt plus 1pt minus 1pt T a b l e}

68:   \refstepcounter{table}\thetable}

69:   \vskip2mm

70:   \centerline{\footnotesize #2}}

71:   \vskip3mm

72: \end{center}}

73: 

74: \newcommand{\MakeTableSepp}[4]{\begin{table}[p]\TabCapp{#2}{#3}\vspace*{-.7cm}

75:   \begin{center} \TableFont \begin{tabular}{#1} #4 

76:   \end{tabular}\end{center}\end{table}}

77: 

78: \newfont{\bb}{ptmbi8t at 12pt}

79: \newfont{\bbb}{cmbxti10}

80: \newfont{\bbbb}{cmbxti10 at 9pt}

81: \newcommand{\uprule}{\rule{0pt}{2.5ex}}

82: \newcommand{\douprule}{\rule[-2ex]{0pt}{4.5ex}}

83: \newcommand{\dorule}{\rule[-2ex]{0pt}{2ex}}

84: \def\thefootnote{\fnsymbol{footnote}}

85: 

86: \newenvironment{references}%

87: {

88: \footnotesize \frenchspacing

89: \renewcommand{\thesection}{}

90: \renewcommand{\in}{{\rm in }}

91: \renewcommand{\AA}{Astron.\ Astrophys.}

92: \newcommand{\AAS}{Astron.~Astrophys.~Suppl.~Ser.}

93: \newcommand{\ApJ}{Astrophys.\ J.}

94: \newcommand{\ApJS}{Astrophys.\ J.~Suppl.~Ser.}

95: \newcommand{\ApJL}{Astrophys.\ J.~Letters}

96: \newcommand{\AJ}{Astron.\ J.}

97: \newcommand{\IBVS}{IBVS}

98: \newcommand{\PASP}{P.A.S.P.}

99: \newcommand{\Acta}{Acta Astron.}

100: \newcommand{\MNRAS}{MNRAS}

101: \renewcommand{\and}{{\rm and }}

102: \section{{\rm REFERENCES}}

103: \sloppy \hyphenpenalty10000

104: \begin{list}{}{\leftmargin1cm\listparindent-1cm

105: \itemindent\listparindent\parsep0pt\itemsep0pt}}%

106: {\end{list}\vspace{2mm}}

107: 

108: \def\TYLDA{~}

109: \newlength{\DW}

110: \settowidth{\DW}{0}

111: \newcommand{\dw}{\hspace{\DW}}

112: 

113: \newcommand{\refitem}[5]{\item[]{#1} #2%

114: \def\REFARG{#3}\ifx\REFARG\TYLDA\else, {\it#3}\fi

115: \def\REFARG{#4}\ifx\REFARG\TYLDA\else, {\bf#4}\fi

116: \def\REFARG{#5}\ifx\REFARG\TYLDA\else, {#5}\fi.}

117: 

118: \newcommand{\Section}[1]{\section{\hskip-6mm.\hskip3mm#1}}

119: \newcommand{\Subsection}[1]{\subsection{#1}}

120: \newcommand{\Acknow}[1]{\par\vspace{5mm}{\bf Acknowledgements.} #1}

121: \pagestyle{myheadings}

122: 

123: \newcommand{\xrule}{\rule{0pt}{2.5ex}}

124: \newcommand{\xxrule}{\rule[-1.8ex]{0pt}{4.5ex}}

125: \def\thefootnote{\fnsymbol{footnote}}

126: \hyphenation{OSARGs}

127: 

128: \begin{center}

129: {\Large\bf The Optical Gravitational Lensing Experiment.\\

130: \vskip3pt

131: Period--Luminosity Relations\\

132: \vskip6pt

133: of Variable Red Giant Stars\footnote{Based on

134: observations obtained with the 1.3-m Warsaw telescope at the Las Campanas

135: Observatory of the Carnegie Institution of Washington.}}

136: \vskip1.0cm

137: {\bf I.~~S~o~s~z~y~\'n~s~k~i$^1$,~~W.\,A.~~D~z~i~e~m~b~o~w~s~k~i$^1$,

138: ~~A.~~U~d~a~l~s~k~i$^1$,\\

139: M.~~K~u~b~i~a~k$^1$,~~M.\,K.~~S~z~y~m~a~{\'n}~s~k~i$^1$,~~G.~~P~i~e~t~r~z~y~\'n~s~k~i$^{1,2}$,\\

140: \L.~~W~y~r~z~y~k~o~w~s~k~i$^{1,3}$,~~O.~~S~z~e~w~c~z~y~k$^{1,2}$,~~and~~K.~~U~l~a~c~z~y~k$^1$}

141: \vskip8mm

142: {$^1$Warsaw University Observatory, Al.~Ujazdowskie~4, 00-478~Warszawa,

143: Poland\\

144: {\small e-mail: 

145: (soszynsk,wd,udalski,mk,msz,pietrzyn,wyrzykow,szewczyk,kulaczyk)@astrouw.edu.pl}\\

146: $^2$ Universidad de Concepci{\'o}n, Departamento de Fisica, Casilla 160--C,

147: Concepci{\'o}n, Chile\\ 

148: $^3$ Institute of Astronomy, University of Cambridge, Madingley Road,

149: Cambridge CB3 0HA, UK}

150: \end{center}

151: 

152: \vskip1.2cm

153: 

154: \Abstract{Period--luminosity (PL) relations of variable red giants in the

155: Large (LMC) and Small Magellanic Clouds (SMC) are presented. The PL

156: diagrams are plotted in three planes: $\log P$--$K_S$, $\log P$--$W_{JK}$,

157: and $\log P$--$W_I$, where $W_{JK}$ and $W_I$ are reddening free Wesenheit

158: indices. Fourteen {\it PL} sequences are distinguishable, and some of them

159: consist of three closely spaced ridges. Each of the sequences is fitted

160: with a linear or quadratic function. The similarities and differences

161: between the {\it PL} relations in both galaxies are discussed for four

162: types of red giant variability: OGLE Small Amplitude Red Giants (OSARGs),

163: Miras and Semiregular Variables (SRVs), Long Secondary Periods (LSPs) and

164: ellipsoidal variables.

165: 

166: We propose a new method of separating OSARGs from non-variable stars and

167: SRVs. The method employs the position in the reddening-free {\it PL}

168: diagrams and the characteristic period ratios of these multiperiodic

169: variables. The {\it PL} relations for the LMC OSARG are compared with the

170: calculated relations for RGB models along isochrones of relevant ages and

171: metallicities. We also compare measured periods and amplitudes of the

172: OSARGs with predictions based on the relations valid for less luminous

173: solar-like pulsators.

174:   

175: Miras and SRVs seem to follow {\it PL} relation of the same slopes in the

176: LMC and SMC, while for LSP and ellipsoidal variables slopes in both

177: galaxies are different. The {\it PL} sequences defined by LSP variables and

178: binary systems overlap in the whole range of analyzed wavebands. We put

179: forward new arguments for the binary star scenario as an explanation of the

180: LSP variability and elaborate on it further. The measured pulsation to

181: orbital period ratio implies nearly constant ratio of the star radius to

182: orbital distance, $R/A\approx0.4$, as we find. Combined effect of tidal

183: friction and mass loss enhanced by the low-mass companion may explain why

184: such a value is preferred.}{Stars: AGB and post-AGB -- Stars: late-type --

185: Stars: oscillations -- Magellanic Clouds}

186: 

187: \Section{Introduction}

188: The first attempt to determine period--luminosity (PL) relation for long

189: period variables (LPVs) was made by Gerasimovi\v{c} (1928), who noticed

190: that Mira stars with longer periods are on average fainter at visual

191: wavelengths. This result has been confirmed by subsequent studies (\eg

192: Wilson and Merrill 1942, Osvalds and Risley 1961, Clayton and Feast 1969),

193: however the scatter of this period--luminosity dependence turned out to be

194: very large.

195: 

196: First tight {\it PL} relation for LPVs was discovered for Mira stars at

197: near infrared (NIR) wavebands (Glass and Lloyd Evans 1981). This {\it PL}

198: law, based on only 11 Miras in the LMC, was refined by extensive studies of

199: Feast \etal (1989) and Hughes and Wood (1990). The second, parallel {\it

200: PL} sequence, occupied by semiregular variables (SRVs), was identified by

201: Wood and Sebo (1996). This sequence was shifted relative to the Miras'

202: ridge toward shorter periods by a factor of two.

203: 

204: However, the subject of {\it PL} distribution of LPVs has progressed

205: rapidly over recent years when large microlensing surveys (MACHO, OGLE,

206: EROS, MOA) published long-term photometry of huge number of stars. Complex

207: structure of the {\it PL} distribution was demonstrated for the first time

208: by Cook \etal (1997), who published {\it PL} diagram for variable stars

209: detected during the MACHO survey in the LMC. A series of three or four {\it

210: PL} sequences defined by LPVs can be distinguished in that diagram.

211: 

212: Sharper picture was presented by Wood \etal (1999), who distinguished and

213: described five {\it PL} sequences (denoted as A--E) in the

214: period--Wesenheit index plane. Wood (2000) showed similar distribution in

215: the $\log P$--$K$ diagram. These results were then confirmed by many

216: studies based on observations originated in various sources (Cioni \etal

217: 2001, 2003, Noda \etal 2002, Lebzelter \etal 2002, Ita \etal 2004,

218: Groenewegen 2004, Fraser \etal 2005).

219: 

220: Kiss and Bedding (2003, 2004) used OGLE data to reveal new features in the

221: {\it PL} distribution. They noticed that sequence B consists of two closely

222: spaced parallel ridges (Ita \etal 2004 denoted the additional sequence as

223: C$'$). Below the tip of the red giant branch (TRGB) Kiss and Bedding (2003)

224: found three sequences shifted in $\log P$ relative to stars brighter than

225: TRGB. It was the definitive proof that stars in the first ascent Red Giant

226: Branch (RGB) pulsate similarly to objects being in the Asymptotic

227: Giant Branch (AGB) phase.

228: 

229: The Optical Gravitational Lensing Experiment (OGLE) collected unprecedented

230: amount of photometric data of stars in the Large and Small Magellanic

231: Clouds. Both galaxies have been constantly monitored since 1997 and at

232: present time this is the best and longest available photometric dataset for

233: analyzing huge number of variable red giants. Our studies on LPVs resulted

234: in many discoveries, regarding also the {\it PL} relations.

235: 

236: Soszy{\'n}ski \etal (2004a) showed that OGLE Small Amplitude Red Giants

237: (OS\-ARGs) constitute separate class of variable stars, with different

238: structure in the {\it PL} plane than ``classical'' SRVs and Miras. We

239: indicated two previously overlooked {\it PL} relations -- the longest

240: (${\rm a}_1$) and the shortest (${\rm a}_4$) period sequences followed by

241: AGB OSARGs. We also suggested a method of empirical division between RGB

242: and AGB OSARGs fainter than TRGB.

243: 

244: Red giants revealing ellipsoidal modulation caused by binarity were

245: analyzed by Soszy{\'n}ski \etal (2004b). It was shown that, if true orbital

246: periods are considered, the {\it PL} relation of ellipsoidal variables

247: (sequence~E) is a direct continuation of sequence~D occupied by mysterious

248: Long Secondary Period (LSP) variables. This is a hint that the LSP

249: phenomenon may be related to binarity, but taking into account available

250: radial velocity measurements, the secondary component usually must be a low

251: mass object, possibly former planet. This idea was supported by

252: Soszy{\'n}ski (2007) who discovered in some LSP variables ellipsoidal-like

253: and eclipsing-like modulations with periods equal to LSPs.

254: 

255: In Soszy{\'n}ski \etal (2005) we again increased complexity of the PL

256: distribution of LPVs. Each of the NIR {\it PL} sequences C$'$, C and D in

257: the LMC (occupied by SRV, Miras and LSP variables) split into two separate

258: ridges in the period -- optical Wesenheit index plane, what corresponds to

259: the spectral division into oxygen-rich (O-rich) and carbon-rich (C-rich)

260: AGB stars. Thus, we found a new photometric method of distinguishing between

261: these two populations.

262: 

263: In this paper we describe in details the {\it PL} relations of variable red

264: giants in both Magellanic Clouds. We show new details in the {\it PL} plane

265: and compare {\it PL} distribution in the LMC and SMC. The paper is

266: organized as follows. Section~2 gives details of the observations and data

267: reduction. In Section~3 the {\it PL} relations are presented with a

268: description of their derivation. A discussion about four types of red giant

269: variability -- OSARGs, Miras/SRVs, LSPs and ellipsoidal modulation -- is

270: given in Sections~4--7. Section~8 summarizes and concludes the paper.

271: 

272: \Section{Observations and Data Reduction}

273: Our analysis is based on the photometry of stars in the LMC and SMC

274: observed during the second and third phases of the OGLE survey (OGLE-II and

275: OGLE-III). Both galaxies have been monitored since 1997 with the 1.3-m

276: Warsaw telescope at the Las Campanas Observatory, Chile, which is operated

277: by the Carnegie Institution of Washington. During the OGLE-II survey

278: (1997--2000) the telescope was equipped with the ``first generation''

279: camera with the SITe ${2048\times2048}$ CCD detector. Details of the

280: instrumentation setup can be found in Udalski \etal (1997). In 2001 the

281: telescope equipment was upgraded to a~wide field ${8192\times8192}$ mosaic

282: camera consisting of eight SITe CCD detectors (Udalski 2003).

283: 

284: In this work we used OGLE-II observations supplemented by the OGLE-III

285: data. The OGLE-II fields cover about 4.5 square degrees in the central

286: regions of the LMC and 2.4 square degrees in the SMC. The last observations

287: used in this analysis were collected in February 2007, so our photometry

288: spans 10 years. The data have been obtained in two standard {\it I} and

289: {\it V} filters, but majority of the observations (typically 700--900

290: points per star) were carried out in the former bandpass. Saturation

291: occurred for stars brighter than $I=12.5$~mag. In the {\it V}-band we

292: collected about 50 points per star.

293: 

294: The data were reduced using the standard OGLE pipeline, as described by

295: Udalski (2003). For details about the transformation of the instrumental

296: photometry to the standard system and the determination of the equatorial

297: coordinates of stars the reader is referred to Udalski \etal (2000). The

298: well calibrated OGLE-II photometry and OGLE-III observations were tied

299: using magnitudes of several dozen constant stars in the closest

300: neighborhood around each object.

301: 

302: The $K_S$ and $J$ NIR photometry used in this analysis was taken from 2MASS

303: All-Sky Catalog of Point Sources (Cutri \etal 2003). We performed

304: cross-identifica\-tion between OGLE and 2MASS sources with 1~arcsec search

305: radius, but earlier small shifts to RA and DEC had been added to compensate

306: for systematic differences between both coordinate systems. We found

307: confident 2MASS counterparts for 98\% of red stars observed by OGLE. For

308: further analysis, we included only stars with all four -- $V$, $I$, $J$ and

309: $K_S$ -- bands available.

310: \vspace*{9pt}

311: \Section{Period--Luminosity Relations for Variable Red Giants}

312: \vspace*{4pt}

313: A period analysis for all stars with $(V-I)>0.5$~mag and $I<17$~mag (in the

314: SMC with $I<17.5$~mag) was performed with program {\sc Fnpeaks} developed

315: by Z.~Ko{\l}aczkowski (private communication). The frequency range searched

316: was from 0.0 to 0.9 cycles per day. For each light curve we determined the

317: highest peak in the Fourier spectrum, then the third order Fourier series

318: was fitted to the observations and subtracted from it. The procedure was

319: repeated on the residual data until fifteen periods per star were derived.

320: Moreover, we recorded the amplitudes and signal-to-noise (S/N) parameters

321: associated with each period.

322: 

323: We constructed {\it PL} diagrams in three variants for both galaxies: $\log

324: P$--$K_S$, $\log P$--$W_{JK}$, and $\log P$--$W_I$. $K_S$ is the 2MASS

325: single epoch magnitude with no compensation applied for interstellar

326: reddening. The single epoch random phase observations cause additional

327: spread in the {\it PL} relations for large amplitude variables. This effect

328: has been avoided for Miras and SRVs thanks to converting single epoch

329: observations into mean magnitudes using complete {\it I}-band light

330: curves. Details of this method were described by Soszy{\'n}ski \etal

331: (2005).

332: 

333: $W_{JK}$ and $W_I$ are NIR and optical Wesenheit indices (Madore 1982), \ie

334: red\-dening free quantities being a linear combinations of selected

335: magnitudes and colors:

336: \begin{eqnarray}

337: W_{JK}&=&K_S-0.686(J-K_S)\\

338: \noalign{\vskip6pt}

339: W_I&=&I-1.55(V-I)

340: \end{eqnarray}

341: 

342: \newpage

343: \renewcommand{\TableFont}{\scriptsize}

344: \renewcommand{\arraystretch}{1.05}

345: \MakeTable{l@{\hspace{4pt}}c@{\hspace{4pt}}l@{\hspace{8pt}}r@{\hspace{4pt}}r@{\hspace{8pt}}r@{\hspace{4pt}}r@{\hspace{8pt}}r@{\hspace{4pt}}r@{\hspace{4pt}}r@{\hspace{8pt}}}{12.5cm}

346: {Period--Luminosity Relations of LPVs in the LMC}

347: {\hline

348: \noalign{\vskip3pt}

349:  &&& \multicolumn{2}{c}{$K_S=\alpha(\log{P}-2.0)+\beta$} & \multicolumn{2}{c}{$W_{JK}=\alpha(\log{P}-2.0)+\beta$} & \multicolumn{3}{c}{$W_I=\alpha\log^2{P}+\beta\log{P}+\gamma$} \\

350: \multicolumn{3}{c}{\raisebox{1.5ex}{Sequence}} & \multicolumn{1}{c}{$\alpha$} & \multicolumn{1}{c}{$\beta$} & \multicolumn{1}{c}{$\alpha$} & \multicolumn{1}{c}{$\beta$} & \multicolumn{1}{c}{$\alpha$} & \multicolumn{1}{c}{$\beta$} & \multicolumn{1}{c}{$\gamma$} \\

351: \noalign{\vskip3pt}

352: \hline

353: \noalign{\vskip3pt}

354:        &     & ${\rm b}_1$ & $-3.34\pm0.04$ & $11.51\pm0.02$ & $-3.81\pm0.05$ & $10.64\pm0.02$ & $-2.648$ & $3.780$ & $13.098$ \\

355:        & RGB & ${\rm b}_2$ & $-3.58\pm0.04$ & $10.94\pm0.02$ & $-3.96\pm0.04$ & $10.03\pm0.03$ & $-3.396$ & $5.118$ & $12.182$ \\

356:        &     & ${\rm b}_3$ & $-3.72\pm0.05$ & $10.22\pm0.03$ & $-4.02\pm0.05$ &  $9.32\pm0.02$ & $-3.564$ & $4.473$ & $12.592$ \\

357: OSARGs &     & ${\rm a}_1$ & $-3.69\pm0.04$ & $11.13\pm0.02$ & $-3.92\pm0.04$ & $10.31\pm0.02$ & $-1.604$ & $0.254$ & $15.782$ \\

358:        &     & ${\rm a}_2$ & $-3.70\pm0.03$ & $10.59\pm0.02$ & $-3.99\pm0.03$ &  $9.70\pm0.02$ & $-1.669$ & $-0.106$ & $15.731$ \\[-1ex]

359:        &\raisebox{1.5ex}{AGB} & ${\rm a}_3$ & $-3.80\pm0.04$ & $9.98\pm0.03$ & $-4.04\pm0.04$ & $9.08\pm0.03$ & $-1.556$ & $-0.905$ & 15.878 \\

360:        &     & ${\rm a}_4$ & $-4.01\pm0.04$ & $9.40\pm0.04$ & $-4.18\pm0.04$ & $8.51\pm0.04$ & $-1.382$ & $-1.839$ & 16.235 \\

361: \noalign{\vskip2pt}

362: \hline

363: \noalign{\vskip3pt}

364:          &  & C$_{\rm O}$ & $-4.17\pm0.08$ & $12.59\pm0.04$ & $-4.34\pm0.09$ & $11.84\pm0.04$ & $-9.803$ & $33.672$ & $-16.149$ \\[-1ex]

365: Miras    & \raisebox{1.5ex}{O-rich} & C$'_{\rm O}$ & $-4.35\pm0.07$ & $11.25\pm0.02$ & $-4.67\pm0.07$ & $10.39\pm0.02$ & $-9.169$ & $25.589$ & $-5.158$ \\

366: and SRVs &  & C$_{\rm C}$ & $-4.07\pm0.11$ & $12.71\pm0.06$ & $-5.19\pm0.11$ & $12.01\pm0.06$ & $-6.618$ & $25.468$ & $-12.522$ \\[-1ex]

367:          & \raisebox{1.5ex}{C-rich} & C$'_{\rm C}$ & $-4.06\pm0.08$ & $11.39\pm0.04$ & $-4.89\pm0.08$ & $10.39\pm0.04$ & $-3.133$ & $7.278$ & $8.830$ \\

368: \noalign{\vskip3pt}

369: \hline

370: \noalign{\vskip3pt}

371:  & O-rich & D$_{\rm O}$ & $-4.41\pm0.07$ & $15.05\pm0.05$ & $-4.64\pm0.08$ & $14.43\pm0.05$ & $-5.882$ & $23.707$ & $-10.128$ \\[-1ex]

372: \raisebox{1.5ex}{LSPs} & C-rich & D$_{\rm C}$ & $-4.38\pm0.25$ & $15.26\pm0.23$ & $-5.09\pm0.28$ & $14.92\pm0.26$ & $0.0$ & $-7.71$ & $33.44$ \\

373: \noalign{\vskip3pt}

374: \hline

375: \noalign{\vskip3pt}

376: \multicolumn{2}{l}{Ell.} & E & $-3.41\pm0.10$ & $14.48\pm0.06$ & $-3.79\pm0.11$ & $13.92\pm0.06$ & $0.0$ & $-3.94$ & $22.20$ \\

377: \noalign{\vskip3pt}

378: \hline

379: }

380: \MakeTable{l@{\hspace{4pt}}c@{\hspace{4pt}}l@{\hspace{8pt}}r@{\hspace{4pt}}r@{\hspace{8pt}}r@{\hspace{4pt}}r@{\hspace{8pt}}r@{\hspace{4pt}}r@{\hspace{4pt}}r@{\hspace{8pt}}}{12.5cm}

381: {Period--Luminosity Relations of LPVs in the SMC}

382: {\hline

383: \noalign{\vskip3pt}

384:  &&& \multicolumn{2}{c}{$K_S=\alpha(\log{P}-2.0)+\beta$} & \multicolumn{2}{c}{$W_{JK}=\alpha(\log{P}-2.0)+\beta$} & \multicolumn{3}{c}{$W_I=\alpha\log^2{P}+\beta\log{P}+\gamma$} \\

385: \multicolumn{3}{c}{\raisebox{1.5ex}{Sequence}} & \multicolumn1{c}{$\alpha$} & \multicolumn1{c}{$\beta$} & \multicolumn1{c}{$\alpha$} & \multicolumn1{c}{$\beta$} & \multicolumn1{c}{$\alpha$} & \multicolumn1{c}{$\beta$} & \multicolumn1{c}{$\gamma$} \\

386: \noalign{\vskip3pt}

387: \hline

388: \noalign{\vskip3pt}

389:        &     & ${\rm b}_1$ & $-3.56\pm0.07$ & $11.83\pm0.04$ & $-3.84\pm0.08$ & $11.05\pm0.04$ & $-1.448$ & $0.655$ & $15.482$ \\

390:        & RGB & ${\rm b}_2$ & $-3.88\pm0.07$ & $11.21\pm0.04$ & $-4.25\pm0.08$ & $10.33\pm0.04$ & $-2.844$ & $4.134$ & $12.874$ \\

391:        &     & ${\rm b}_3$ & $-4.40\pm0.07$ & $10.14\pm0.05$ & $-4.48\pm0.09$ &  $9.39\pm0.05$ & $-3.583$ & $4.782$ & $12.639$ \\

392: OSARGs &     & ${\rm a}_1$ & $-3.66\pm0.04$ & $11.55\pm0.02$ & $-4.02\pm0.05$ & $10.74\pm0.03$ & $-0.520$ & $-2.574$ & $18.041$ \\

393:        &     & ${\rm a}_2$ & $-3.89\pm0.04$ & $10.85\pm0.02$ & $-4.11\pm0.04$ & $10.06\pm0.03$ & $-1.111$ & $-0.912$ & $16.248$ \\[-1ex]

394:        & \raisebox{1.5ex}{AGB} & ${\rm a}_3$ & $-4.04\pm0.05$ & $10.19\pm0.04$ & $-4.40\pm0.05$ & $9.25\pm0.04$ & $-1.179$ & $-1.090$ & 16.001 \\

395:        &     & ${\rm a}_4$ & $-4.29\pm0.04$ & $9.52\pm0.03$ & $-4.49\pm0.04$ & $8.89\pm0.04$ & $-0.688$ & $-2.886$ & 16.997 \\

396: \noalign{\vskip2pt}

397: \hline

398: \noalign{\vskip3pt}

399:          &  & C$_{\rm O}$ & $-4.14\pm0.17$ & $12.90\pm0.05$ & $-4.39\pm0.19$ & $12.14\pm0.05$ & $-6.900$ & $23.341$ & $-6.495$ \\[-1ex]

400: Miras    & \raisebox{1.5ex}{O-rich} & C$'_{\rm O}$ & $-4.47\pm0.15$ & $11.62\pm0.03$ & $-4.94\pm0.15$ & $10.81\pm0.03$ & $-3.816$ & $8.379$ & $9.106$ \\

401: and SRVs &  & C$_{\rm C}$ & $-4.22\pm0.16$ & $13.11\pm0.07$ & $-5.26\pm0.14$ & $12.43\pm0.06$ & $-3.991$ & $12.486$ & $3.605$ \\[-1ex]

402:          & \raisebox{1.5ex}{C-rich} & C$'_{\rm C}$ & $-4.40\pm0.15$ & $11.88\pm0.04$ & $-5.05\pm0.13$ & $10.90\pm0.04$ & $-3.482$ & $7.933$ & $9.362$ \\

403: \noalign{\vskip3pt}

404: \hline

405: \noalign{\vskip3pt}

406:  & O-rich & D$_{\rm O}$ & $-4.13\pm0.11$ & $15.16\pm0.08$ & $-4.40\pm0.11$ & $14.61\pm0.08$ & $-4.752$ & $19.487$ & $-6.363$ \\[-1ex]

407: \raisebox{1.5ex}{LSPs} & C-rich & D$_{\rm C}$ & $-4.11\pm0.30$ & $15.39\pm0.30$ & $-4.46\pm0.33$ & $14.77\pm0.30$ & $0.0$ & $-7.00$ & $31.40$ \\

408: \noalign{\vskip3pt}

409: \hline

410: \noalign{\vskip3pt}

411: \multicolumn{2}{l}{Ell.} & E & $-3.15\pm0.12$ & $14.70\pm0.07$ & $-3.32\pm0.16$ & $14.10\pm0.07$ & $0.0$ & $-3.90$ & $22.43$ \\

412: \noalign{\vskip3pt}

413: \hline

414: }

415: 

416: \begin{figure}[p]

417: \hglue-11mm{\includegraphics[width=14.5cm]{fig1.ps}}

418: \vspace{-1.1cm}

419: \FigCap{Period--luminosity diagrams of variable red giants in the

420: LMC. OSARG variables are shown as blue points (RGB as light blue, AGB as

421: dark blue). Miras and SRVs are marked with pink (O-rich) and red (C-rich)

422: points. Light and dark green points refer to O-rich and C-rich LSP

423: variables, respectively. Yellow points indicate ellipsoidal red giants.}

424: \end{figure}

425: \begin{figure}[p]

426: \hglue-11mm{\includegraphics[width=14.5cm]{fig2.ps}}

427: \vspace{-0.8cm}

428: \FigCap{Period--luminosity relations for red giants in the LMC fitted to the sequences shown in

429: Fig.~1. The colors of lines indicate the same types of stars as in Fig.~1.}

430: \end{figure}

431: \begin{figure}[p]

432: \hglue-11mm{\includegraphics[width=14.5cm]{fig3.ps}}

433: \vspace{-0.8cm}

434: \FigCap{Period--luminosity diagrams of variable red giants in the

435: SMC. The colors represent the same types of stars as in Fig.~1.}

436: \end{figure}

437: \begin{figure}[p]

438: \hglue-11mm{\includegraphics[width=14.5cm]{fig4.ps}}

439: \vspace{-0.8cm}

440: \FigCap{Period--luminosity relations for red giants in the SMC fitted to the sequences shown in

441: Fig.~3. The colors of lines indicate the same types of stars as in

442: Fig.~1.}

443: \end{figure}

444: 

445: The {\it PL} diagrams for LMC and SMC are shown in Fig.~1 and Fig.~3,

446: respectively. We present here only the stars that have been used to

447: determine the {\it PL} relations. For Miras and SRVs, LSP and ellipsoidal

448: variables only one point per star is shown, while for OSARGs more periods

449: per star may by displayed. Detailed description of the criteria used for

450: the selection of variables are given in the next four Sections. The LMC PL

451: sequences appear better defined which may be explained by much larger

452: sample of objects and by much smaller depth in the line of sight of the LMC

453: than the SMC.

454: 

455: To derive possibly most precise regression lines we first converted the

456: {\it PL} diagrams into density maps and found points along the given {\it

457: PL} sequence where the density of points is the largest. Then, we performed

458: linear or quadratic least square fit to these points. Note that the

459: described method produces regression functions in general different than

460: ordinary least square fits to the {\it PL} points, \ie we usually obtained

461: steeper lines.  Figs.~2 and~4 shows the functions fitted to the {\it PL}

462: sequences. The same functions are given in Tables~1 and~2.

463: 

464: At $K_S$ and $W_{JK}$ the {\it PL} relations have been approximated by

465: linear functions, while $\log P$--$W_I$ laws have usually been fitted with

466: quadratic functions, because of distinct curvature of these relations. This

467: non-linearity is associated curvature of the distribution visible in the

468: color--magnitude $(V-I)$--$I$ diagram. Optical and NIR color--magnitude

469: diagrams for stars included in Figs.~1 and~3 are shown in Fig.~5. One can

470: notice different features of $(V-I)$--$I$ and $(J-K)$--$K$ distributions in

471: both Clouds. Note, that the coefficients of the parabolas fitted to the

472: period--$W_I$ relations (Tables~1 and~2) have considerable

473: errors. Nevertheless, one may find these cofficients useful.

474: 

475: \begin{figure}[t!]

476: \centerline{\includegraphics[width=14cm]{fig5.ps}}

477: \vspace{-0.5cm}

478: \FigCap{Color--magnitude diagrams for LMC ({\it upper panels}) and SMC ({\it lower

479: panels}) variable red giants. The colors represent the same types of stars

480: as in Fig.~1.}

481: \end{figure}

482: 

483: In the next Sections we describe in more details four classes of variable red

484: giants visible in the {\it PL} diagrams.

485: 

486: \Section{OGLE Small Amplitude Red Giants}

487: The name ``OGLE Small Amplitude Red Giants'' (OSARGs) has been proposed by

488: Wray \etal (2004), who analyzed about 18\,000 of such objects in the

489: Galactic bulge using OGLE-II photometry. The term ``OSARG'' is not fully

490: synonymous with ``Small Amplitude Red Variable'' (SARV) from the

491: classification introduced by Eggen (1977, and references within), \ie stars

492: with visual amplitudes smaller than 0.5~mag. As shown by Soszy{\'n}ski

493: \etal (2004a) the OSARG variables cannot be distinguished from

494: ``classical'' SRVs (hereafter just SRVs) using solely amplitudes of

495: variations, because this is an overlapping property of both

496: groups. Moreover, Eggen's SARVs include wide variety of stellar types,

497: while OSARG variables constitute a separate class of variable stars

498: (Soszy{\'n}ski \etal 2004a).

499: 

500: It is known that all red giants of type K5 and cooler are variable in

501: brightness, and amplitude of variations increases with decreasing

502: temperature of the stars (Edmonds and Gilliland 1996, Henry \etal 2000).

503: Using OGLE photometry we can detect stars with {\it I}-band peak-to-peak

504: amplitudes as small as 4~mmag.

505: 

506: \Subsection{Selection of OSARGs}

507: Since OSARGs with the smallest detectable amplitudes are mixed with

508: constant stars (strictly speaking, with stars having amplitudes too small

509: to be visible in our data), it is not an easy task to separate both groups.

510: Likewise, the longest period sequences of OSARGs overlap with {\it PL}

511: ridge of SRVs, thus a separation of the two populations is in general

512: problematic. Any of simple diagrams that can be prepared using available

513: parameters of stars (periods, magnitudes, colors, amplitudes) did not

514: enable us to precisely separate long-period OSARGs from SRVs, and faintest

515: OSARGs from non-variable stars.

516: 

517: To solve this problem we worked out a new method of OSARG variables

518: detection. Our algorithm takes into account a star position on the PL

519: diagrams and characteristic period ratios of these multiperiodic variables.

520: Our basic tool is the diagram $\Delta\log P_L$--$P_S/P_L$, where $P_S$ and

521: $P_L$ is any pair of periods (shorter and longer, respectively) selected

522: from the most significant periodicities of the given star, and $\Delta\log

523: P_L$ is a horizontal distance of $\log P_L$ from selected sequence in the

524: selected {\it PL} plane (in our analysis from sequence~A).

525: 

526: \begin{figure}[htb]

527: \centerline{\includegraphics[width=12.2cm]{fig6.ps}}

528: %\vglue-3mm

529: \FigCap{Diagram $\Delta\log P_L$--$P_S/P_L$ for OSARG variables in the

530: LMC. $P_S$ and $P_L$ are, respectively, any shorter and longer periods of

531: a given star. $\Delta\log{P_L}$ is a horizontal distance from sequence~A

532: in $\log P$--$W_I$ plane. See text for detailed description.}

533: \end{figure}

534: Fig.~6a shows such a diagram for OSARGs tentatively selected from our data.

535: $\Delta\log P_L$ in this diagram were measured in the $\log P_L$--$W_I$

536: plane, but very similar picture was independently obtained using $\log

537: P_L$--$W_{JK}$ diagram. Distinct groups of points are fingerprints of the

538: OSARG variability. In order to utilize this distribution for selecting

539: OSARGs, we smoothed it with a Gaussian filter and derived density map

540: visualized in Fig.~6b.

541: 

542: Then, we fitted two-dimensional Gauss function to every group of points

543: visible in Fig.~6b, and prepared artificial density map which is shown in

544: Fig.~6c. In this way we rejected points which occurred in the $\Delta\log

545: P_L$--$P_S/P_L$ diagrams by chance from spuriously detected periods. The

546: last step of our procedure was normalization of our map, \ie setting up the

547: maxima of all fitted Gaussian at the same level (Fig.~6d). Thanks to that

548: every pair of periods was treated with the same weight, regardless of which

549: {\it PL} relations corresponded to these periods. Two such maps as shown in

550: Fig.~6d have been prepared -- one with $\Delta\log P_L$ measured in the

551: $\log P_L$--$W_I$ plane, and the second in the $\log P_L$--$W_{JK}$

552: diagram.

553: 

554: The procedure of selecting OSARG variables was performed as follows. For

555: each star we examined each pair of periods (selected from 15 derived

556: periods) for which S/N parameter was larger than 3.0. We derived ratios of

557: both periods and $\Delta\log P_L$ parameters in the $\log P_L$--$W_I$ and

558: $\log P_L$--$W_{JK}$ planes. Then, for subsequent Gaussians shown in

559: Fig.~6d we calculated values in respective ($\Delta\log P_L$, $P_S/P_L$)

560: points, thus only points close to the maximum of the given Gaussian

561: produced values significantly larger than zero. Additionally, we took the

562: product of the values obtained with the two Wesenheit indices so that the

563: values for accidental points were close to zero.

564: 

565: We co-added every such derived value for each Gaussian and each pair of

566: periods. Resulting sum is an indicator of the OSARG variability. For the

567: best candidates for OSARG variables this quantity is larger than 10. For

568: our purposes we set the threshold of the OSARG indicator to be larger than

569: 3. Our method allows effective discrimination of OSARGs from non-variable

570: stars, as well as from SRVs.

571: 

572: \subsection{Period--Luminosity Relations for OSARGs}

573: The OSARGs constitute probably the most numerous class of variable stars in

574: the Magellanic Clouds. We selected about 17\,000 such objects in the LMC

575: and about 3500 in the SMC, which is more than a number of Cepheids, RR~Lyr

576: stars, eclipsing binaries and other types of variable red giants. OSARG

577: variables brighter than TRGB ($K_S=12.05$~mag in the LMC, $K_S=12.7$~mag in

578: the SMC) were obviously recognized as AGB stars, while below TRGB we

579: separated RGB and AGB stars using the feature noticed by Soszy{\'n}ski

580: \etal (2004a). If any of the periods falls on the shortest-period sequence

581: ${\rm a}_4$, the star was recognized as AGB object. Otherwise we marked the

582: OSARG as RGB star, although small contribution of AGB may still exist among

583: these group. In Figs.~1--5 RGB OSARGs are marked with light blue points and

584: lines, while AGB OSARGs are drawn in dark blue.

585: 

586: As shown by Soszy{\'n}ski \etal (2004a) OSARGs follow a series of three

587: (RGB stars) or four (AGB stars) narrow sequences, spreading over periods

588: ranging from 8 to 160 days. In this paper we use labels introduced by

589: Soszy{\'n}ski \etal (2004a): ${\rm a}_1$~--~${\rm a}_4$ for AGB and ${\rm

590: b}_1$~--~${\rm b}_3$ for RGB OSARGs. The {\it PL} relations of RGB OSARGs

591: are shifted in $\log P$ relative to AGB OSARGs, what can be explained by

592: the temperature difference of both populations at the same luminosity (Kiss

593: and Bedding 2003). Soszy{\'n}ski \etal (2004a) also showed that the third

594: sequence (\ie sequences~${\rm a}_3$ and ${\rm b}_3$ or Wood's sequence~A)

595: is split into three ridges, what manifests itself in the period -- period

596: ratio diagram with distinct groups of points around $P_S/P_L\approx0.9$ and

597: $P_S/P_L\approx0.95$. We marked these two additional {\it PL} relations

598: with dotted lines in Figs.~2 and~4. We observe this feature for RGB and AGB

599: stars located below the TRGB and for some AGB objects brighter than the

600: TRGB. For the brightest stars this split is not visible in our data.

601: 

602: Our new analysis revealed that the same feature may also concern other {\it

603: PL} sequences of OSARG variables. Similar clouds for period ratios 0.9 and

604: 0.95 are clearly visible for the second sequence (${\rm a}_2$ and ${\rm

605: b}_2$). For the fourth sequence (${\rm a}_4$) this property is marginally

606: visible, what can be an effect of smaller number of stars that follow this

607: shortest period relation. Nevertheless, we also drew the additional dotted

608: lines in the {\it PL} diagrams for this ridge. For the first, longest

609: period sequences (${\rm a}_1$ and ${\rm b}_1$) we cannot detect any

610: distinct group of points close to the period ratio 0.9 or~0.95.

611: 

612: It seems that {\it PL} sequences for OSARG variables are best separated in

613: the $\log P-W_I$ diagram. This plane also reveals the largest differences

614: between red giants in the LMC and SMC. The {\it PL} sequences in the LMC

615: are significantly steeper and are more curved than those in the SMC. This

616: difference may be directly related to different distribution of points in

617: the color--magnitude ($V-I$, $V$) diagrams given in Fig.~5, what may

618: reflect differences in ages and metallicity of red giants between the SMC

619: and LMC. In both galaxies, the slopes of the OSARG {\it PL} relations

620: increase with decreasing periods.

621: 

622: It is natural to associate the {\it PL} sequences with individual modes of

623: stellar oscillations. Identification of the modes would yield new

624: constraints on models of red giants as well as on star formation in the

625: Magellanic Clouds. As an illustration, we considered the ${\rm b}_1$, ${\rm

626: b}_2$, and ${\rm b}_3$ sequences in the LMC, assuming that they correspond

627: to first three radial modes.

628: 

629: Salaris and Girardi (2005) showed that the upper part of the RGB in the LMC

630: is composed of stars with heavy element abundance in the wide range, from

631: ${\rm [Fe/H]}=-1$ to $-0.3$, which corresponds to age between 1 and

632: 11~Gy. However, for most of the objects ${\rm [Fe/H]}=-0.65\pm0.1$ and

633: there is a gap in star formation between 7~Gy and 9~Gy.

634: 

635: \begin{figure}[htb]

636: \centerline{\includegraphics[width=12cm]{fig7.ps}} 

637: \FigCap{Sequences ${\rm b}_1$, ${\rm b}_2$, and ${\rm b}_3$ of the 

638: {\it PL} relation for the RGB OSARGs of the LMC (see Fig.~1) compared with

639: relations calculated on the basis of the selected BaSTI isochrones for the

640: first three radial modes. Ages and metal abundance parameters are given in

641: legend. The initial masses in the order of increasing age are equal

642: 1.26~\MS, 1.02~\MS, and 0.85~\MS. Mass loss in the RGB is included in the

643: model calculations and the corresponding masses at the RGB-tip are 1.13~\MS,

644: 0.94~\MS, and 0.79~\MS.}

645: \end{figure}

646: Our theoretical {\it PL} relations were based on the BaSTI (Pietrinferni

647: \etal 2004) isochrones. For each model, we recalculated envelope structure

648: and determined the radial mode periods. The $W_I$ index was calculated from

649: $V$ and $I$ data from BaSTI assuming 18.45~mag for the distance modulus. A

650: comparison of the selected theoretical and observed sequences is shown in

651: Fig.~7. Note that fitting requires younger and more metal abundant objects

652: at higher luminosities. The agreement between observed and calculated

653: relations, though not perfect seems encouraging. Detailed comparison

654: requires simulation of the the LMC RGB population. The result will

655: certainly differ somewhat from simulation based in the CMD data presented

656: in Fig.~5 of Salaris and Girardi (2005) and reconciling the results will

657: require new modeling. It is likely that the four ${\rm a}_k$ sequences

658: correspond to first four radial modes.

659: 

660: \begin{figure}[p]

661: \centerline{\includegraphics[width=13cm]{fig8.ps}}

662: \vspace*{-4mm}

663: \FigCap{The {\it upper panel} shows the same data as in Fig.~7 but the 

664: lines refer to periods ($P_{\rm max}=\nu_{\rm max}^{-1}$) corresponding to

665: maximum amplitude of solar-like oscillations calculated with Eq.~(3) for

666: the same isochrones as in Fig.~7. Dots in the {\it lower panel} shows the

667: mode amplitudes for the RGB OSARGs of the LMC. The lines were calculated

668: with Eq.~(4) for the same isochrones as in the {\it upper panel}.}

669: \end{figure}

670: 

671: \subsection{OSARGs as Solar-Like Pulsators}

672: It is not yet known how pulsation observed in RGB and AGB stars are

673: excited. Do they result from an intrinsic mode instability or rather, like

674: in the sun, from a stochastic energy input to modes which are intrinsically

675: stable. Both options have been considered Dziembowski \etal (2001) and by

676: Bedding (2003). In both, there is a crucial role of convection which makes

677: the problem difficult for theory. Phenomenological arguments, based on

678: structure of power spectra, in favor for the solar-like option in the case

679: of the SRVs were first given by Christensen-Dalsgaard \etal (2001) and

680: supported by Bedding (2003). Power spectra of OSARGs resemble those of

681: SRVs. Only the amplitudes are smaller. Thus, the solar-like option appears

682: even more probable for OSARGs.

683: 

684: Stello \etal (2007) reported recently results of a multi-site search for

685: solar-like oscillations in red giants in the open cluster M67. They did not

686: report individual mode frequencies but only excess of power in some

687: individual objects near location predicted by the scaling relation based on

688: the solar data,

689: $$\nu_{\rm max}=\frac{(M/\MS)(T_{\rm eff}/{\rm T}_{{\rm eff}\odot})^{3.5}}

690: {(L/\LS)}\times3050~\mu{\rm Hz}\eqno(3)$$

691: 

692: Bedding and Kjeldsen (2003) showed that this relation approximately

693: describes location of the highest peaks in oscillation spectra of nearby

694: dwarfs, subgiants and giants. Stello \etal (2007) also compared the

695: measured excesses of power with the scaling relation for peak amplitudes

696: $$A_\lambda=\frac{L/\LS(5.1\pm0.3)~\mu{\rm mag}}{(\lambda/550~{\rm

697: nm})(T_{\rm eff}/{\rm T}_{{\rm eff}\odot})^2(M/\MS)}\eqno(4)$$

698: derived by Kjeldsen and Bedding (1995). Only for the objects lying in the

699: lower RGB ($L/\LS<30$) they were able to find measurable excesses. The

700: values are crudely consistent with the prediction. For brighter object they

701: could derive only upper limits.

702: 

703: The RGB OSARGs are considerably brighter than the M67 red giants for which

704: the excess have been measured. At $W_I=12$~mag their luminosity is between

705: 1.5 and $2\times10^3\LS$. How the two scaling relations work at these high

706: luminosities is shown in Fig.~8. We used the same isochrones as adopted in

707: the {\it PL} relations shown in the previous Section, to calculate $P_{\rm

708: max}=1/\nu_{\rm max}$ using Eq.~(3) and $A_I$ using Eq.~(4). In the latter,

709: we adopted $\lambda=810$~nm.

710: 

711: In the upper plot we may see that maximum of power corresponds to the first

712: and second overtone of radial modes. We may conclude that there is a crude

713: agreement with data. At lower luminosity the calculated lines run well

714: above the ${\rm b}_1$ sequence. However, we do not take it as an evidence

715: against the solar-like interpretation of the OSARGs, because in all

716: solar-like pulsators modes are seen in a rather wide range of

717: frequencies. On the other hand, since in various stars low-order modes are

718: excited by the other type of driving, even a very good agreement could not

719: be regarded as a proof that the OSARGs are indeed solar-like pulsators.

720: 

721: As for the amplitude, the agreement is worse, as the lower panel of Fig.~8

722: shows. Again, the agreement should not be seen as evidence against the

723: solar-like nature of the OSARGs. Eq.~(4) is not based on any solid

724: theory. In fact, a more recent study of stochastic excitation by Samadi

725: \etal (2005) suggests the $(L/M)^{0.7}$ amplitude scaling, instead of $(L/M)$

726: adopted in Eq.~(4). Slower amplitude rise is indeed more consistent with

727: data. At this point, we just want to stress that the OSARGs are likely to

728: provide the best constraints on models of stochastic excitation over a wide

729: range of stellar parameters.

730: 

731: \Section{Miras and Semiregular Variables}

732: 

733: Mira stars and SRVs are regarded as two different types of LPVs (Kholopov

734: \etal 1985), although it seems that both groups represent a continuum. In

735: the period--NIR luminosity diagrams these objects obey two parallel PL

736: relations (sequences~C and~C$'$) which are thought to represent fundamental

737: and first overtone modes of pulsations. Soszy{\'n}ski \etal (2005) studied

738: 3200 Miras and SRVs in the LMC and noticed important feature: in the

739: period--$W_I$ diagram each {\it PL} sequence splits into two well separated

740: ridges what corresponds to spectral division into O-rich and C-rich AGB

741: stars.

742: 

743: In this paper we used the sample of Miras and SRVs in the LMC selected by

744: Soszy{\'n}ski \etal (2005). We slightly corrected this list by applying the

745: OSARG indicator described in Section~4.1. This time we left on the list

746: only variables with very small value of this parameter, smaller than

747: 0.5. In the SMC we selected Miras and SRVs in the same manner as in the LMC

748: and found in total 630 objects. From our fit we excluded stars redder than

749: $(J-K_S)=2.3$~mag, \ie objects obscured by dust, and stars populating dim

750: sequence between C and C$'$ (see discussion below).

751: 

752: In our diagrams Miras and SRVs are indicated with pink (O-rich) and red

753: (C-rich) points and lines. We plotted only one point per star corresponding

754: to the most significant period. In the LMC the division into O- and C-rich

755: variables was done in the period--$W_I$ diagram, where both populations can

756: be easily distinguished. However, the most striking difference between {\it

757: PL} relations of LPVs in the LMC and SMC again occurs in the period--$W_I$

758: plane. In the SMC O-rich and C-rich variables are much worse separated than

759: in the LMC. Besides, there appears to be almost no bright O-rich Miras and

760: SRVs in the SMC, what is a well known effect of lower

761: metallicity. Therefore, in the SMC we used another widely used tool for

762: distinguishing between O-rich and C-rich red giants -- $(J-K_S)$

763: colors. Stars bluer than $(J-K_S)=1.3$~mag have been recognized as O-rich

764: giants, while the remaining objects as C-rich stars.

765: 

766: Let us focus on sequence C$'$ occupied by probable first overtone SRVs. At

767: the NIR domain this sequence is roughly parallel to OSARG {\it PL}

768: relations, what was an argument for recognizing OSARGs as the same type of

769: variables as Miras and SRVs, but pulsating in higher overtones (Wood \etal

770: 1999).  However, the same stars plotted in the $\log P$--$W_I$ diagram

771: reveal completely different behavior. The {\it PL} law of O-rich SRVs

772: crosses the OSARG sequences, while C-rich variables are in different

773: position relative to OSARGs. This discrepancy between NIR and optical {\it

774: PL} diagrams suggests that OSARGs represent a different, presumably less

775: evolved, population of stars than SRVs and Miras. To check this, we looked

776: for OSARGs showing modes belonging to sequence~C, which is defined by

777: former types of variables.

778: 

779: Among 17\,000 of stars in the LMC classified as OSARGs, we found only a few

780: objects with distinct periodicity associated with sequence~C. However, for

781: these objects our classification as OSARGs is uncertain, because none of

782: the secondary periods fall on sequence~A, which belongs solely to

783: OSARGs. It is probable that the OSARG signature in these stars is caused

784: either by spurious periods or periods are real, but in fact associated with

785: higher overtones of more evolved objects. In conclusion, we did not find

786: any convincing examples of variables belonging to both classes: SRVs and

787: OSARGs.

788: 

789: We noted that the secondary periods for some SRVs in both Clouds seem to

790: follow an additional {\it PL} sequence located between sequences C and

791: C$'$. Moreover, Soszy{\'n}ski \etal (2005) mentioned about a dim sequence

792: located in the same place, but defined by the primary periods of some small

793: amplitude SRVs (these stars are not shown in Figs.~1 and~3). If this

794: sequence is real and corresponds to a radial mode, then sequence~C cannot

795: correspond to the fundamental mode. This would imply a revision of the mode

796: identification in Miras and SRVs. In such a case one can expect that the

797: fundamental-mode oscillations would be observed for periods by a factor of

798: about two longer than those associated with sequence~C. Indeed, we have

799: found a number of SRVs with secondary periods in this region of the PL

800: diagram, \ie between sequences C and D. Unfortunately, binary systems

801: (sequence E with halved periods) are also expected to populate this region,

802: and we would not be able to distinguish between pulsation and ellipsoidal

803: variations of SRVs in binary systems.

804: 

805: There is also a number of SRVs -- mostly carbon stars -- with secondary

806: periods shorter than those from sequence C$'$. These periods could be

807: associated with pulsation modes higher than the first overtone, however

808: this problem needs further exploration.

809: 

810: We have already noted that the largest differences in the {\it PL}

811: relations between the two Clouds are seen in the $\log P$--$W_I$

812: plane. This conclusion concerns especially O-rich stars which populate

813: different range of magnitudes and the {\it PL} relations differ in

814: slopes. The sequences in $K_S$ and $W_{JK}$ are more similar. In

815: particular, the slopes of the {\it PL} relations are the same within the

816: error limits in both environments. To derive the zero-point differences

817: between {\it PL} laws, we assumed the same slope for the LMC and the SMC

818: stars. For the whole sequence~C and for O-rich sequence C$'$, we obtained

819: similar vertical distances of about 0.35~mag in $K_S$ band and about

820: 0.4~mag in $W_{JK}$. It is in agreement with the previously derived values:

821: \eg 0.4~mag for dereddened $K_0$ magnitudes obtained by Wood (1995) or

822: 0.36~mag derived by Ita \etal (2004) for sequence~C in $K_S$ (not corrected

823: for extinction). For carbon stars of sequence C$'$ the zero-point

824: difference is considerably larger, that is 0.43~mag in $K_S$ and 0.47~mag

825: in $W_{JK}$. Note that the last value is very close to the estimated

826: difference between distance moduli of the LMC and SMC (\eg Udalski \etal

827: 1999).

828: 

829: \Section{Long Secondary Periods}

830: The longest period sequence in our {\it PL} pattern (sequence~D) is

831: associated with the most mysterious phenomenon connected with red giants

832: variability.  Long Secondary Periods (LSPs) represent the only unexplained

833: type of large amplitude stellar variability known today. Wood \etal (2004)

834: discussed possible explanations of the LSPs and concluded that most

835: plausible mechanism that may cause LSPs is gravity mode

836: excitation. However, recently Soszy{\'n}ski (2007) presented observational

837: arguments for binary origin of the LSPs. In this scenario dust and gas

838: originated in stellar wind form a cloud orbiting the red giant causing a

839: periodic obscuration. Radial velocity measurements (Hinkle \etal 2003, Wood

840: \etal 2004) are consistent with both low-mass companion and non-radial

841: oscillations. However, our photometric data contain new hints supporting

842: the former option.

843: 

844: Our samples of LSP variables were initially selected using their position

845: in the $\log P$--$W_{JK}$ diagrams. After a visual inspection, we chose

846: about 1600 LMC and 300 SMC distinct LSP stars with characteristic

847: ``eclipsing-like'' light curves. The division into O-rich and C-rich stars

848: in the LMC was based on the position in the $\log P$--$W_I$ diagram. For

849: the SMC stars, just like in the case of Miras and SRVs, we relied only on

850: the $(J-K_S)$ colors.

851: 

852: The strong argument for a binary scenario is that the sequence~D overlaps

853: with sequence~E, formed by red giants in binary systems, and seems to be a

854: continuation of this ridge toward brighter stars (Soszy{\'n}ski

855: \etal 2004b). It is striking that this behavior appears in all analyzed PL

856: planes: $\log P$--$K_S$ (Derekas \etal 2006), $\log P$--$W_I$, $\log

857: P$--$W_{JK}$ for both Clouds. Moreover, we found that both sequences seem

858: to match each other at other bandpasses: $J$, $I$ and even $V$. If the LSP

859: phenomenon were caused by a stellar oscillation, it would have to be an

860: unlikely coincidence.

861: 

862: In contrast to sequences~C and~C$'$, the slopes of sequence~D in the NIR

863: domain are different in both galaxies. The LSP variables in the SMC define

864: ridge of shallower slope than in the LMC. It is interesting that the same

865: behavior exhibits sequence~E defined by close binaries. We take this fact

866: as another argument supporting binary explanation of the LSP phenomenon.

867: 

868: The name ``long secondary periods'' is somewhat misleading, because quite

869: often amplitude of the LSP variability is higher than that associated with

870: the short periods. Most of LSP stars are OSARGs both of RGB and AGB

871: type. However, in the latter case the amplitudes are much higher. Thus our

872: sample of objects showing LSP consist predominantly of the AGB OSARGs.

873: 

874: OSARG variables with the LSP modulation do not follow exactly the same {\it

875: PL} relations as non-LSP stars. The larger LSP amplitudes, the larger

876: discrepancy between {\it PL} sequences. Stars with the largest amplitudes

877: of the LSP modulations have distinctly smaller mean luminosities compared

878: to their non-LSP counterparts. Besides, there appears to be a correlation

879: between the LSP amplitudes and the minimum periods of OSARG oscillations.

880: 

881: An interesting issue concerning LSPs is a relative number of red giants

882: exhibiting this phenomenon. Typical values found in the literature are

883: 25--30\%. However, our examination shows that this proportion depends on

884: the lower amplitude of the LSP variability concerned in analysis. Derekas

885: \etal (2006) indicated that maximum possible amplitudes of the LSP

886: modulation increase with luminosity of stars, but the smallest amplitudes

887: are at the detection limit along the whole range of magnitudes. If the

888: minimum detected amplitudes are around 10~mmag, than the LSPs appear for

889: about 30\% of LSPs. If the limit for amplitude is decreased to 5~mmag, the

890: proportion of stars with LSPs reaches 50\%. We can also consider the stars

891: with LSP variations of amplitudes larger than pulsational (OSARG, SRV)

892: amplitudes. Then, we detect the LSP modulation in about 30\% of the whole

893: sample.

894: 

895: Adopting the binary star scenario, we have to explain why the {\it PL}

896: relation for the LSP is nearly parallel to {\it PL} relations for radial

897: pulsation modes and why the phenomenon is so common. The near parallel run

898: of the {\it PL} relations, translates to almost constant ratio of star to

899: orbit radii, $R/A$. Combining the Kepler law with the expression for radial

900: mode periods, we obtain

901: $$\frac{R}{A}=0.24(1+q)^{-1/3}Q_k^{-2/3}\left(\frac{P_k}{P_{\rm

902: orb}}\right)^{2/3}\eqno(5)$$ 

903: where $q$, which we will neglect, is the secondary companion to total mass

904: ratio and $Q_k$ is the pulsation constant. In Section~4.2, we associated

905: the sequences ${\rm b}_k$ ($k=1,2,3$) of the {\it PL} relations for the RGB

906: OSARGs with the first three radial modes and suggested the same association

907: with sequences ${\rm a}_k$ ($k=1,2,3$) for the AGB OSARGs. We follow it

908: here and, in agreement with the adopted scenario, we set $P_{\rm

909: orb}=P_{\rm LSP}$. The {\it PL} relation for LSP variables is defined

910: primarily by the AGB OSARGs, thus we should use periods from the ${\rm

911: a}_k$ sequences. Unfortunately, we do not have the $Q_k$ values for the AGB

912: star. However, we noted in our RGB star models that $Q_2$ values change

913: only little along the isochrones. In the considered models, the values rise

914: from 0.024 to 0.025. We adopted the upper value and in the Petersen diagram

915: identified the $P_S/P_L=0.05$ as the $P_3/P_{\rm orb}$ ratio. With these

916: numbers, Eq.~(5) yields $R/A=0.4$.

917: 

918: What may keep the hypothetical low-mass companion in a narrow range of

919: fractional distances away of stellar surface? Stars expand through the

920: whole RGB and most of AGB phase. However, mass loss causes that orbits of

921: their low-mass companions expand too. At typically adopted mass loss rates,

922: the $R/A$ ratio increases. As long as its value is well below one, tidal

923: effects are negligible and changes of the orbital radius are determined

924: only by mass loss from the star (we ignore here possible interactions with

925: other companions). The rate of $A$ decrease due to tidal friction is

926: proportional to $(R/A)^7$ (\eg Zahn 1977), so that once tidal effects

927: become important, the sign of the temporal derivative of $A$ may quickly

928: reverse and it is usually assumed (\eg Villaver and Livio 2007) that the

929: companion is engulfed into the red giant envelope. However, the tidal

930: interaction affects also the star and it does not seem unreasonable to

931: assume that it enhances mass loss. If the mass loss rate becomes

932: sufficiently high the orbital distance my start to increase again and the

933: companion migration toward the stellar surface may be halted.  Stellar

934: radius will then continue to increase but the $R/A$ ratio will stay nearly

935: constant at the value, which we estimated to be about 0.4.

936: 

937: Since about 30--50\% of red giants exhibits the LSP variability and

938: certainly there must be a range of unfavorable inclinations for its

939: detection, we must postulate that nearly all red giants have their low mass

940: companions. This would have far reaching consequences for our views on

941: frequency planetary systems in the universe. We are aware of problems that

942: the proposed scenario poses. However, the problems, with an alternative

943: explanation of the LSP that postulates g-mode excitation seem even harder

944: to overcome.

945: 

946: \Section{Ellipsoidal and Eclipsing Red Giants}

947: 

948: The last class of LPVs represented in the rich structure of the {\it PL}

949: distribution are red giants in the binary systems -- ellipsoidal and

950: eclipsing variables. Wood \etal (1999) suggested that sequence~E is made up

951: of contact binaries. It was confirmed by Ruci{\'n}ski and Maceroni (2001),

952: who analyzed long-period eclipsing binaries detected in the OGLE SMC fields

953: and indicated a distinct period--luminosity--color relation.

954: 

955: Soszy{\'n}ski \etal (2004b) selected and studied a sample of 1660

956: ellipsoidal and eclipsing binaries with a red giant as one of the

957: components. They showed that the scatter of the {\it PL} relation strongly

958: depends on the amplitude of ellipsoidal variability -- the larger

959: amplitudes the tighter {\it PL} strip. It is understandable, because

960: variables with the largest amplitudes must lay close to the line defined in

961: the {\it PL} plane by systems with red giants filling up their Roche

962: lobes. Obviously this theoretical line indicates the short period edge of

963: the {\it PL} distribution, so the widening of the sequence must proceed

964: toward longer periods.

965: 

966: Consequently, the determination of the {\it PL} relation depends on

967: amplitudes of ellipsoidal variables -- larger amplitude variables have on

968: average shorter periods. To get consistent result in both galaxies, we cut

969: the sample of Soszy{\'n}ski \etal (2004b) at $A(I)=0.03$~mag. We excluded

970: also eclipsing variables, because eclipses of various depths randomly

971: change the mean magnitudes of the stars. Finally, Soszy{\'n}ski \etal

972: (2004b) showed that about 10\% of ellipsoidal red giants reveal clear

973: deformation caused by the eccentricity of their orbits. Periods of these

974: systems are systematically longer than periods of ellipsoidal variables

975: with circular orbits, therefore we removed the former group from our

976: sample. In the SMC we selected ellipsoidal red giants in the same manner as

977: in the LMC, and we found in total 440 objects.

978: 

979: Sequence~E stars are marked in yellow in Figs.~1--5. Note, that in Wood

980: \etal (1999) and in many subsequent papers this sequence was shifted toward

981: shorter periods by a factor of two, because half the orbital periods were

982: presented. The spread of sequence E in the NIR domain is considerably

983: larger than in the period--$W_I$ plane. It can be partly explained by

984: errors of the 2MASS photometry increasing quickly for fainter stars.

985: 

986: As it was noted in the previous Section, ellipsoidal variables in the LMC

987: follow at NIR passbands somewhat steeper {\it PL} relation than in the

988: SMC. Since the {\it PL} relation of ellipsoidal variables is a projection

989: of radius--luminosity dependence, this fact can be utilized for studying

990: parameters of red giant stars in environments of different metallicities.

991: 

992: \Section{Conclusions}

993: In this paper we showed the most complex structure of the {\it PL}

994: distribution presented so far. The Wood's five ridges turn out to be an

995: overlap of fourteen sequences (if consider closely spaced {\it PL}

996: relations of OSARGs, the number of sequences exceeds twenty). In order to

997: help recognizing {\it PL} relations with published {\it PL} laws we provide

998: Table~3 with appropriate identifications.

999: 

1000: Diagrams employing period ratios (similar to the Petersen diagram) were

1001: used as a tool for discriminating OGLE Small Amplitude Red Giants

1002: (OSARGs). We compared three sequences of the {\it PL} relations for the LMC

1003: OSARGs in the RGB phase with the calculated relation for first three radial

1004: modes using isochrone calculations. We found an essential agreement with

1005: our knowledge about metallicities and ages of red giants population in the

1006: LMC. However, there are also discrepancies which may suggest the necessity

1007: of refinement of the knowledge and/or stellar models. This is a potential

1008: application of the OSARGs.

1009: 

1010: \MakeTableTop{l@{\hspace{4pt}}c@{\hspace{4pt}} l@{\hspace{8pt}} c@{\hspace{8pt}} c@{\hspace{8pt}} c@{\hspace{8pt}}}{12.5cm}

1011: {Labels of the {\it PL} relations in this and previous papers}

1012: {\hline

1013: \noalign{\vskip3pt}

1014: \multicolumn{3}{c}{this paper} & Wood \etal 1999 & Kiss and Bedding 2003 & Ita \etal 2004 \\

1015: \noalign{\vskip3pt}

1016: \hline

1017: \noalign{\vskip3pt}

1018:        &     & ${\rm b}_1$ &   & $R_1$ &                  \\

1019:        & RGB & ${\rm b}_2$ & B & $R_2$ & $B^-$ \\

1020:        &     & ${\rm b}_3$ & A & $R_3$ & $A^-$ \\

1021: OSARGs &     & ${\rm a}_1$ &   &                           &                  \\

1022:        &     & ${\rm a}_2$ & B & 2O                        & $B^+$ \\[-1ex]

1023:        & \raisebox{1.5ex}{AGB} & ${\rm a}_3$ & A & 3O      & $A^+$ \\

1024:        &     & ${\rm a}_4$ &   &                           &                  \\

1025: \noalign{\vskip3pt}

1026: \hline

1027: \noalign{\vskip3pt}

1028:          &  & C$_{\rm O}$ & C & F  & C \\[-1ex]

1029: Miras    & \raisebox{1.5ex}{O-rich} & C$'_{\rm O}$ & B & 1O & C$'$ \\

1030: and SRVs &  & C$_{\rm C}$ & C & F  & C \\[-1ex]

1031:          & \raisebox{1.5ex}{C-rich} & C$'_{\rm C}$ & B & 1O & C$'$ \\

1032: \noalign{\vskip3pt}

1033: \hline

1034: \noalign{\vskip3pt}

1035:  & O-rich & D$_{\rm O}$ & D & $L_2$ & D \\[-1ex]

1036: \raisebox{1.5ex}{LSPs} & C-rich & D$_{\rm C}$ & D & $L_2$ & D \\

1037: \noalign{\vskip3pt}

1038: \hline

1039: \noalign{\vskip3pt}

1040: \multicolumn{2}{l}{Ell. and Ecl.} & E & E$^*$ & $L_1^*$ & E$^*$ \\

1041: \noalign{\vskip3pt}

1042: \hline

1043: \multicolumn{6}{l}{$^*$ -- the sequence is shifted due to halving the

1044: orbital periods.}

1045: }

1046: 

1047: Most likely mechanism responsible for pulsation in the OSARGs is the

1048: stochastic excitation. Therefore we looked at these objects as solar-like

1049: pulsators. The range of periods agrees with predictions based on

1050: extrapolation of the relations found for much less luminous stars. The

1051: amplitudes are lower than predicted by the Kjeldsen and Bedding (1995)

1052: formula, where the the amplitude rises linearly with the luminosity-to-mass

1053: ratio. However, there are calculations predicting a slower amplitude

1054: rise. Testing theory of stochastic excitation is another possible

1055: application of OSARGs.

1056: 

1057: We would like to bring the reader's attention to NIR Wesenheit index

1058: ($W_{JK}$). The sequences in the period--$W_{JK}$ plane are generally

1059: better defined than those at $K_S$ magnitudes. Wesenheit index is a

1060: reddening independent quantity, so even heavy reddened Miras fall on the

1061: {\it PL} sequence. Therefore, the period--$W_{JK}$ relations can be used as

1062: a distance indicators without correcting them for reddening. Moreover, 

1063: O- and C-rich Miras and SRV obey very similar relations in the

1064: period--$W_{JK}$ plane, while at $K_S$ the {\it PL} relations are

1065: significantly different for both populations. We conclude that $W_{JK}$

1066: index can be a useful tool for studying LPVs.

1067: 

1068: Comparison of the {\it PL} relations in NIR supports the binary star

1069: scenario as the explanation of the LSP variability. We further develop this

1070: scenario taking into account data on the ratios of the short period

1071: (pulsational) and LSP variability. For OSARGs these ratio is nearly

1072: constant which translates into nearly constant ratio of the stellar to

1073: orbital radius the value of about 0.4. We proposed that the small mass

1074: companion position is determined by the balance between mass loss and tidal

1075: effects. This scenario requires that the proximity of the companion

1076: enhances mass loss.  Moreover, it requires that a substantial fraction

1077: (majority) of red giants have such companions and that their mass cannot be

1078: too small.

1079: 

1080: \vspace*{9pt}

1081: 

1082: \Acknow{The paper was supported by the Foundation for Polish Science 

1083: through the Homing Program and by MNiSW grants: 1P03D01130 and

1084: N20303032/4275.

1085: 

1086: This publication makes use of data products from the Two Micron All Sky

1087: Survey, which is a joint project of the University of Massachusetts and the

1088: Infrared Processing and Analysis Center/California Institute of Technology,

1089: funded by the National Aeronautics and Space Administration and the

1090: National Science Foundation.}

1091: 

1092: \vspace*{9pt}

1093: 

1094: \begin{references}

1095: \refitem{Bedding, T.~R.}{2003}{Astrophys. and Space Sci.}{284}{61}

1096: \refitem{Bedding, T.~R., and Kjeldsen, H.}{2003}{Publ. Astron. Soc. Aust.}{20}{203}

1097: \refitem{Christensen-Dalsgaard, J., Kjeldsen, H., and Mattei, J.A.}{2001}{\ApJ}{562}{141}

1098: \refitem{Cioni, M.-R.L., Marquette, J.-B., Loup, C., Azzopardi, M., Habing,

1099: H.J., Lasserre, T., and Lesquoy, E.}{2001}{\AA}{377}{945}

1100: \refitem{Cioni, M.-R.L., \etal}{2003}{\AA}{406}{51}

1101: \refitem{Clayton, M.L., and Feast, M.W.}{1969}{\MNRAS}{146}{411}

1102: \refitem{Cutri, R.M., \etal}{2003}{~}{~}{``2MASS All-Sky Catalog of Point Sources''}

1103: \refitem{Derekas, A., Kiss, L.L., Bedding, T.R., Kjeldsen, H., Lah, P., and

1104: Szab{\'o}, G.M.}{2006}{\ApJL}{650}{L55}

1105: \refitem{Dziembowski, W.A., Gough, D.O., Houdek, G., and Sienkiewicz, R.}{2001}{\MNRAS}{328}{601}

1106: \refitem{Edmonds, P.D., and Gilliland, R.L.}{1996}{\ApJL}{464}{L157}

1107: \refitem{Eggen, O.J.}{1977}{\ApJ}{213}{767}

1108: \refitem{Feast, M.W., Glass, I.S., Whitelock, P.A., and Catchpole, R.M.}{1989}{\MNRAS}{241}{375}

1109: \refitem{Fraser, O.J., Hawley, S.L., Cook, K.H., and Keller, S.C.}{2005}{\AJ}{129}{768}

1110: \refitem{Gerasimovi\v{c}, B.P.}{1928}{Proceedings of the National Academy of Science}{14}{963}

1111: \refitem{Glass, I.S., and Lloyd Evans, T.}{1981}{Nature}{291}{303}

1112: \refitem{Groenewegen, M.A.T.}{2004}{\AA}{425}{595}

1113: \refitem{Henry, G.W., Fekel, F.C., Henry, S.M., and Hall, D.S.}{2000}{\ApJS}{130}{201}

1114: \refitem{Hinkle, K.H., Lebzelter, T., Joyce, R.R., and Fekel, F.C.}{2002}{\AJ}{123}{1002}

1115: \refitem{Hughes, S.M.G., and Wood, P.R.}{1990}{\AJ}{99}{784}

1116: \refitem{Ita, Y., Tanab{\'e}, T., \etal}{2004}{\MNRAS}{347}{720}

1117: \refitem{Kholopov, P.N. \etal}{1985}{~}{~}{``General Catalog of Variable

1118: Stars'', The Fourth Edition, Nauka, Moscow}

1119: \refitem{Kiss, L.L., and Bedding, T.R.}{2003}{\MNRAS}{343}{L79}

1120: \refitem{Kiss, L.L., and Bedding, T.R.}{2004}{\MNRAS}{347}{L83}

1121: \refitem{Kjeldsen, H., and Bedding, T.R.}{1995}{\AA}{293}{87}

1122: \refitem{Lebzelter, T., Schultheis, M., and Melchior, A.L.}{2002}{\AA}{393}{573}

1123: \refitem{Madore, B.F.}{1982}{\ApJ}{253}{575}

1124: \refitem{Noda, S., \etal}{2002}{\MNRAS}{330}{137}

1125: \refitem{Osvalds, V., and Risley, A.~M.}{1961}{Publ. Leander McCormick Obs.}{11}{147}

1126: \refitem{Pietrinferni, A., Cassisi, S., Salaris, M., and Castelli, F.}{2004}{\ApJ}{612}{168}

1127: \refitem{Rucinski, S.M., and Maceroni, C.}{2001}{\AJ}{121}{254}

1128: \refitem{Salaris, M., and Girardi, L.}{2005}{\MNRAS}{357}{669}

1129: \refitem{Samadi, R., Goupil, M.-J., Alecian, E., Baudin, F., Georgobiani,

1130: D., Trampedach, R., Stein, R., and Nordlund, A.}{2005}{Journal of

1131: Astrophysics and Astronomy}{26}{171}

1132: \refitem{Soszy\'nski, I., Udalski, A., Kubiak, M., Szyma\'nski, M.,

1133: Pietrzy\'nski, G., \.Zebru\'n, K., Szewczyk,~O., and Wyrzykowski, {\L}.}{2004a}{\Acta}{54}{129}

1134: \refitem{Soszy\'nski, I., Udalski, A., Kubiak, M., Szyma\'nski, M., 

1135: Pietrzy\'nski, G., \.Zebru\'n, K., Szewczyk,~O., Wyrzykowski, {\L}., 

1136: and Dziembowski, W.A.}{2004b}{\Acta}{54}{347}

1137: \refitem{Soszy\'nski, I., Udalski, A., Kubiak, M., Szyma\'nski, M., 

1138: Pietrzy\'nski, G., \.Zebru\'n, K., Szewczyk, O., Wyrzykowski, {\L}., and 

1139: Ulaczyk, K.}{2005}{\Acta}{55}{331}

1140: \refitem{Soszy\'nski, I.}{2007}{\ApJ}{660}{1486}

1141: \refitem{Stello, D., et al.}{2007}{\MNRAS}{377}{584}

1142: \refitem{Udalski, A.}{2003}{\Acta}{53}{291}

1143: \refitem{Udalski, A., Kubiak, M., and Szyma\'nski, M.}{1997}{\Acta}{47}{319}

1144: \refitem{Udalski, A., Szyma\'nski, M., Kubiak, M., Pietrzy\'nski, G., 

1145: Soszy\'nski, I., Wo\'zniak, P., and \.Zebru\'n, K.}{1999}{\Acta}{49}{223}

1146: \refitem{Udalski, A., Szyma\'nski, M., Kubiak, M., Pietrzy\'nski, G., 

1147: Soszy\'nski, I., Wo\'zniak, P., and \.Zebru\'n, K.}{2000}{\Acta}{50}{307}

1148: \refitem{Villaver, E., and Livio, M.}{2007}{\ApJ}{661}{1192}

1149: \refitem{Wilson, R.E., and Merrill, P.W.}{1942}{\ApJ}{95}{248}

1150: \refitem{Wood, P.~R.}{1995}{~}{~}{{\it IAU Colloq.~155}: ``Astrophysical 

1151: Applications of Stellar Pulsation'', {\bf 83}, 127}

1152: \refitem{Wood, P.R.}{2000}{Publ. Astron. Soc. Aust.}{17}{18}

1153: \refitem{Wood, P.R., Olivier, E.A., and Kawaler, S.D.}{2004}{\ApJ}{604}{800}

1154: \refitem{Wood, P.R., and Sebo, K.M.}{1996}{\MNRAS}{282}{958}

1155: \refitem{Wood, P.R., \etal (MACHO team)}{1999}{~}{~}{in: {\it IAU Symp.}

1156: {\bf 191}, ``Asymptotic Giant Branch Stars'', Ed. T.~Le~Bertre, A.~L\'ebre, 

1157: and C.~Waelkens (San Francisco: ASP), 151}

1158: \refitem{Zahn, J.-P.}{1977}{\AA}{57}{383}

1159: \end{references}

1160: \end{document}

1161: