A&A 377, 945-954 (2001)
M.-R. L. Cioni1 -
J.-B. Marquette2 -
C. Loup2 -
M. Azzopardi5 -
H. J. Habing1 -
T. Lasserre4,3 - E. Lesquoy4,2
1 - Sterrewacht Leiden, Postbus 9513, 2300 RA Leiden, The Netherlands
2 - Institute d'Astrophysique de Paris, CNRS, 98bis Bld. Arago, 75014 Paris, France
3 - Max-Planck-Institut für Kernphysik, Postfach 10 39 80, 69029 Heidelberg, Germany
4 - Departement d'Astrophysique, Physique des Particules, Physique Nucleaire et d'Instrumentation associée (DAPNIA), Service de Physique des Particules (SPP), CEN Saclay, 91191 Gif-sur-Yvette Cedex, France
5 - Observatoire de Marseille, LAM, 2 place Le Verrier, 13248 Marseille Cedex 4, France
Received 5 April 2001 / Accepted 6 August 2001
We present the first cross-identifications of sources in the near-infrared DENIS survey and in the micro-lensing EROS survey in a field of about 0.5 square degrees around the optical center (OC) of the Large Magellanic Cloud. We analyze the photometric history of these stars in the EROS data base and obtain light-curves for about 800 variables. Most of the stars are long period variables (Miras and Semi-Regulars); a few Cepheids are also present. We also present new spectroscopic data on 126 asymptotic giant branch stars in the OC field, 30 previously known and 96 newly discovered by the DENIS survey. The visible spectra are used to assign a carbon- (C-) or oxygen-rich (O-rich) nature to the observed stars on the basis of the presence of molecular bands of TiO, VO, CN, C2. For the remaining of the stars we used the ( ) color to determine whether they are O-rich or C-rich. Plotting versus we find three very distinct period-luminosity relations, mainly populated by Semi-Regular of type a (SRa), b (SRb) and Mira variables. Carbon-rich stars occupy mostly the upper part of these relations. We find that of the asymptotic giant branch population are long period variables (LPVs).
Key words: stars: evolution - late-type - variables: general - Magellanic Clouds
In this study we inspected about 800 EROS light-curves in a region containing the optical center field (OC) in the Bar of the Large Magellanic Cloud (LMC); it is one of the fields searched for AGB stars in the late 1970s by Blanco et al. (1980). Combining these data with the DENIS data we derive three period-luminosity relations. We discuss the properties of the stars in the -magnitude diagram and in the color-magnitude diagram ( , ) with the additional information whether the star is C-rich or O-rich. Wood (1999, 2000) found five linear relations, we find only three because of lower sensitivity in our survey data.
Section 2 describes the DENIS data, the EROS data, the cross-identification between these two data sets and the spectroscopic observations. Section 3 describes the classification criteria applied to determine the variability and the spectral type of the sources. Section 4 discusses the period-luminosity relation(s) and the position of different types of variables in the ( , ) color-magnitude diagram. Finally Sect. 5 summarizes the contents of this study.
|Figure 1: Map of the analyzed area. The big circle limits the OC field and the rectangle marks the area where multi-object spectroscopy was performed. Small circles identify the classified variables and crosses identify variable stars without assigned classification. Small dots represent all the DCMC sources in the field. The upper right empty area coincides with the position of the out of use CCD. Numbers denote the squared CCDs on the EROS mosaic. The inhomogeneous distribution of the variable stars among the CCD quadrants is due to the incompleteness of the EROS data.|
|Open with DEXTER|
We used the Multi-Object-Spectroscopy mode of the EMMI instrument. We observed six overlapping fields of with on average 14 slits punched per mask. The CCD frames were corrected for the bias level introduced by detector electronics (a set of 9 frames per night were observed and then combined to calculate the statistical median bias value for the whole period), for the dark current (no dark frames were observed during the same period so we used a value of determined during another observing campaign; the stability and the almost negligible value of the dark signal is confirmed by the NTT - team) and for the relative pixel response (flat-field - three frames were observed and mediated per mask). The sky-subtracted spectra were then wavelength calibrated and corrected for the atmospheric extinction. The data reduction process was done using the MIDAS package.
Five of the observed spectra are of poor quality, but good enough to determine the C-rich or O-rich nature of the corresponding source. 14 of the observed spectra are useless because the star was too faint or because of electronic interferences. In one case the wrong star was observed and in two cases the slit was in between two blended sources. In the end we confirmed the spectral type of 30 objects and assigned the spectral type to 96 objects of the original sample (Sect. 4). Note that at the position of BMB-O 46 (Blanco et al. 1980) we detected a star of C type while Blanco et al. assign it a late M type.
Frequent sampling is essential to discover irregularities or to confirm regularities in the shape of the light curve. For example the distinction between SRa and SRb focuses on the regularity of the periodicity and on the presence of more than one period, which can only be disentangled when the sampling of the light-curve is frequent enough. These better sampled light curves allow a much better application of the classification criteria introduced in the GCVS to a much larger sample of stars.
So far the detection of small-amplitude variables in the Magellanic Clouds has been limited. The large survey by Hughes (1989) detected only variables with mag. Detailed light curves from observations of variables in our Galaxy have been preferentially determined for large amplitude variable stars or stars with infrared excess.
There are many more variable stars observed by EROS that can be classified with refined criteria and considerably better statistical completeness. Such studies are outside of the scope of this study.
|Figure 2: Light-curves in the red EROS filter over the measured time range (left) and folded on the major period (right) for: a) cepheids, b) Miras, c) SRas and d) SRbs. Origin of days for left panels is January 1st, 1990 (JD 2447892.5).|
|Open with DEXTER|
We classify 43 sources as Mira variables adopting the criterion that
these sources amounting to
of the AGB stars (Fig. 2b).
Their light curve may show the following additional characteristics:
- maxima variable by up to 1 mag - small variation of the period from
cycle to cycle - presence of a bump in the ascending or descending
branch - variation of the phase of minimum. Variables with smaller
amplitude of variation have been classified as SR. We identify
of AGB stars as SRa (Fig. 2c) and SRb(Fig. 2d), respectively. SRbs have more than one
pulsation period. Six
of the AGB stars have a light-curve with
too much structure to identify them with either of the two
sub-classes of SRs. Among the acceptable-quality light curves
there are sources which can be classified as "irregular variables'':
clearly variable but irregular in amplitude, shape and size of the
individual peaks, therefore without a main period. Finally there are
also sources with a light-curve of too poor a S/N ratio or a
light-curve which slowly increases or decreases in the whole range of
time. In the latter case the monitoring period was not long enough to
detect periodicity; more data from the EROS or MACHO (Alcock et al. 1997) projects are certainly required. These last two
groups of acceptable quality light-curves, because they indicate that the
sources are variable but the light-curve is not good enough to assign
a class of variability, correspond to about of AGB stars. Table 1 summarizes the percentages of the
different types of AGB stars found in this work and estimated as
described in Sect. 5.
|Figure 3: Color-magnitude diagram of the spectroscopically observed DENIS candidates. Filled circles represent C-rich stars, empty circles represent O-rich stars and crosses represent S stars. The vertical dashed lines discriminate between the regions dominated by C-rich, O-rich stars and obscured stars, as derived in Sect. 4.|
|Open with DEXTER|
For a detailed description of the features characterizing these stars and their variation as a function of the spectral sub-type we refer the reader to the Atlas of Digital Spectra of Cool Stars (Turnshek et al. 1985).
|Figure 4: Histogram distribution of the amplitude of a) all the variable sources. For good-quality light-curves both b) the distribution of amplitudes and c) periods are shown as a function of O-rich or C-rich type. d) shows how the amplitudes distribute as a function of , the horizontal line indicates the applied selection criterion to distinguish Miras from SRs.|
|Open with DEXTER|
The location of Miras is consistent with the LMC Mira sequence of Feast (1989). The distribution of SRas is also consistent with what has been found in the Galaxy (Bedding & Zijlstra 1998) and previously in the Magellanic Clouds (Wood & Sebo 1996). The longer period of the SRbs gives rise to a new relation so far only discussed by Wood (1999, 2000), his sequence D.
There are few more objects in Fig. 5 with short periods that correspond to a sequence found by Wood and labelled by him as A. Both Wood's sequences A and B are densely populated at fainter magnitudes, while our sequences are better populated at brighter magnitudes. We attribute this difference to the limiting magnitude of DENIS (Cioni et al. 2000) and to the incompleteness of EROS in detecting small amplitude variables. The sequence of shorter periods of SRbs, confirmed to be mostly on sequence B, coincides also with the short period stars found in the Baade's window by the ISOGAL and MACHO collaboration (Alard et al. 2000). Wood's additional sequence E of eclipsing binaries is definitely below the DENIS limiting magnitudes.
It is important to emphasize that the magnitude of DENIS is a single epoch measurement and thus the -magnitude diagram is affected by the scatter due to the amplitude of variation of each source; this effect is larger for Miras than for Semi-Regulars.
|Figure 5: Period-magnitude diagram for the good-quality light-curve variable stars; the period (P) is in days. Symbols refer to different types of variable sources and are explained in the plot. Only the long period of SRbs is plotted in this figure, see Fig. 6 for the location of the two derived periods. Filled symbols are C-rich objects ( ) and empty symbols are O-rich objects. The position of the tip of the red giant branch (TRGB) as derived by Cioni et al. (2000a) is indicated. The tilted line represents the period-luminosity relation derived by Feast (1989) for Miras in the Galaxy translated in the LMC at (m-M)=18.55. Capital letters identify the sequences as in Wood et al. (1999).|
|Open with DEXTER|
|Figure 6: Period-magnitude diagram of SRbs. Filled symbols show the long period and empty symbols show the short period for the same source. Horizontal lines connect possible eclipsing binaries. The period (P) is in days.|
|Open with DEXTER|
Each period-luminosity relation corresponds, theore-tically, to a region of instability which is normally associated to a certain mode of pulsation; the presence of multi-periods indicates the coexistence of two pulsation modes. Figure 5 of Wood & Sebo (1996) indicates that the long period in SRbs results from radial fundamental mode pulsation, while Miras are first overtone pulsators and other Semi-Regulars are mostly high radial overtone pulsators; the predicted gap between fundamental and first overtone sequences is in very good agreement. This would also agree with the pulsation mode derived by Whitelock & Feast (2000) from measured diameters (Hipparcos data) of Miras in the Milky Way. On the other hand if Miras are fundamental mode pulsators, as suggested by Wood et al. (1999), then the long period sequence of SRbs requires another explanation, because theoretically there exists no pulsation mode at a longer period than that of the fundamental mode.
The presence of two simultaneous periods in SRbs may be explained by the coupling between dust formation and star pulsation (Winters et al. 1994) because of a pulsation of longer period produced by the dynamical structure of the circumstellar dust shells. As shown by Fleischer et al. (1992) for large carbon to oxygen ratio ( ) the periodic ejection of a dust shell occurs exactly with the pulsational period while for , as the amount of dust that can condense is smaller, shell ejection takes place at longer intervals than one period (Fleischer et al. 1999): the pulsation corresponds to the shorter of the two periods and the longer is due to obscuration by dust. It is however questionable whether such a model applies to stars with silicate type dust and lower mass-loss rates (10-6 to yr-1) such as those derived from CO observations (Loup et al. 1993), but preliminary calculations support that the model does apply (Winters, private communication). If there is circumstellar dust the star is expected to have redder ( ) than those observed by DENIS in this sample. On the other hand if at these colours the molecular blanketing effect (in the J band due to C2 and CN) dominates the dust obscuration effect the blue ( ) colour is explained.
Wood et al. (1999, 2000) suggest that sequence D is populated by the longer period of a binary system. Such a scenario is also poorly constrained. We found in our sample only three possible eclipsing binaries (Fig. 6). Moreover the number of objects populating this sequence is too high with respect to theoretical predictions. Whitelock (1986) shows, from measurements in globular clusters, that the evolutionary path of long period variables goes from SRa to Mira following a line of shallower slope, in agreement with the theoretical calculations by Vassiliadis & Wood (1993), than the period-magnitude relation. The evolutionary path thus shows that the stars pass through a variable phase, then the pulsation goes away and they slowly move into the next sequence. Gaps between the sequences reflect a "phase'' of some stability which could be represented by the of non-variable AGB stars. However because of the existing gap between the starting of the core Helium burning (TRGB) and the beginning of the TP-AGB phase (Iben & Renzini 1983) some of the AGB stars in our sample might be in an earlier evolutionary phase (E-AGB) or they are TP-AGB stars in between thermal pulses that esperience a significant decline of luminosity (Wood et al. 1999). Little evidence of pulsation has been found in AGB stars just above the horizontal branch (Feast & Whitelock 1987), these E-AGB stars might be of the same type as those AGBs we do not detect, within the completness limits of the sample, as variables. Alard et al. (2000) found that of the stars in their sample are variable. These stars have been selected to have a detection both at m and m (ISO bands). Because of the detection at m, that probes the existence of a circumstellar dust envelope (Omont et al. 1999), we think that Alard et al. sample is formed by TP-AGB stars which are therefore more evolved than our sample of AGB stars selected on the basis of near-infrared colours. It is thus not surprising that we find a lower percentage of AGB variables. However it is remarkable that in our sample only of the high mass-loss stars ( ) are variable, because mass-loss and variability are believed to be connected. The result may be due to incompleteness of the sample (Sect. 5.1); note that the sample size for stars with ( ) is much smaller because DENIS does not detect very obscured AGB stars.
The color-magnitude diagrams (
Fig. 7 characterize the location of the different types of
variable. Miras cover uniformly the region where O-rich, C-rich and
obscured stars are located (Fig. 3). Comparing the four
diagrams, the reddest objects are all of Mira type. Not all Miras,
however, have red colors. This may indicate a difference in initial
mass. The initial "position'' (luminosity, period, color) of the
stars in the TP-AGB phase is given when they experience the first
TP. All TP-AGB stars loose mass at a rate that varies from 10-8to
yr-1 and will end as a
white dwarf. Because the time on the AGB of the less massive stars is
much longer than that of the more massive stars, the mass loss rate of
a low mass star can be much lower. As a consequence in the last stages
of the AGB (Mira) low mass stars do not become much redder because of
mass loss, but the more massive stars will. The AGB phase is so brief
(Vassiliadis & Wood 1993) that there is very little increase
in luminosity. Stars that are too light will remain O-rich because
the TPs have too little effect. Stars that are too massive will remain
O-rich because of hot-bottom burning. Only stars in a well-defined
mass interval will become C-rich. In the LMC this explains why
C-rich stars have almost all the same luminosity (Groenewegen
1994). SRas are distributed in both branches of O-rich and
C-rich stars while SRbs are mostly concentrated only in the O-rich
branch (Fig. 7). Sources with good or acceptable quality
light-curves, but no classification, are distributed in different
regions of the diagram and their type can probably be derived from the
diagrams of the classified sources.
|Figure 7: Color-magnitude diagrams ( , ) of the detected variables. Symbols are as in Fig. 5 and crosses indicate sources of acceptable-quality light-curve; Cepheids are excluded from these plots.|
|Open with DEXTER|
SRbs might be in a transition phase between O-rich SRas and Miras. At the beginning and at the end of the TP-AGB phase the amount of material available to form dust is high, in the first case because the excess of oxygen is high and in the other one because the excess of carbon is high. We expect such objects to have light-curves like SRas; dust would obscure the star at intervals approximately equal to the pulsation period. In the transition from O-rich to C-rich stars the amount of dust reduces, therefore these objects would behave like SRbs. Another possibility is that SRbs are less evolved than O-rich SRas and have therefore have less matter that can condense into dust, main requirement to have a multi-periodic light-curve such as those calculated by Winters et al. (1994). The monitoring of dust features for SRas and SRbs of comparable near-infrared properties will certainly put constraints on the proposed scenarios and allow to discriminate against the existence of a binary system; such a program is currently under development.
Comparing the -magnitude diagram with the models by Vassiliadis & Wood (1993) we find that our sample contains AGB stars with masses in the range from 0.9 to and age range from 0.1 to 10 Gyr. An intermediate/old population is definitely present in the LMC.
It is interesting to notice that there are some Semi-Regulars and few Miras located below the TRGB. These sources could be AGB stars in the early evolutionary phase (burning He in a shell) or TP-AGB (burning H and He into concentric shells) between thermal pulses or variables for the first time on the red giant branch. We are going to investigate spectroscopically the nature of these variable objects.
The access to the EROS database has been kindly granted by the EROS collaboration. The GAIA and CURSA packages have been available thanks to the work of the Starlink project, UK. We thank Jean-Philippe Beaulieu, Sylvain Lupone and Gregoor Verschuren for valuable assistance during data treatment, Ariane Lançon, Yvonne Simis, Peter Wood, Jan-Martin Winters Joris Blommaert and Patricia Whitelock for useful discussions.
|13||Mira with bump|
|16||SRa with bump|
Table A.2 listing the DENIS/EROS sources (Fig. 5) with a good light-curve quality, classified in the present work with observed spectra is electronically available at the CDS. It contains the following items: the DCMC source identifier, the EROS source number, the period, the light-curve quality factor, the classification as described in Table A.1, the EROS quadrant where the source was observed and notes that could be the Blanco et al. (1980) counterpart or the longer period in the case of SRbs.
Also, a Table A.3 listing the spectral nature of the newly observed DENIS sources, not included in Fig. 5 and in the previous table, is available at the same place. A more detailed analysis with the aim to assign the spectral type to these sources will be the subject of a forthcoming paper, together with equivalent observations in other locations in the LMC and in the Small Magellanic Cloud.