Read 2300827.DVI text version

Astron. Astrophys. 341, 827­841 (1999)

ASTRONOMY AND ASTROPHYSICS

Luminous pre-main sequence stars in the LMC?

H.J.G.L.M. Lamers1,2 , J.P. Beaulieu3 , and W.J. de Wit1

1 2 3

Astronomical Institute, Princetonplein 5, 3584 CC Utrecht, The Netherlands (e-mail: [email protected]; [email protected]) SRON Laboratory for Space Research, Sorbonnelaan 2, 3584 CA Utrecht, The Netherlands Kapteyn Laboratorium, Postbus 800, 9700 AV Groningen, The Nederlands (e-mail: [email protected])

Received 4 June 1998 / Accepted 16 October 1998

Abstract. We report the serendipitous discovery of seven luminous irregular variables in the LMC by the EROS project. The 0m , and vary by stars have V 15m to 17m , (BE - RE ) m about 0.4 on timescales of 10 to 40 days. The variations in BE and RE are about equal but the stars are slightly bluer when they are fainter. Medium resolution optical spectra show strong H emission, weak H emission and absorption lines of the higher Balmer lines. Absorption lines of He i, C iii, O ii and Si iv are present in the spectra of some of the stars. The spectral types derived from the presence or absence of the lines are between late-O and late-B. Effective temperatures and luminosities were derived from the energy distributions and independently from the EROS photometry combined with the approximate spectral types. The luminosity, derived from V , E(B - V ), and the distance of the LMC, are in the range of 103 to 104 L . The strong H emission, the spectral types, the luminosities and the irregular variations are similar (but not the same) to those of the Galactic pre-main sequence HAeBe stars (i.e. the Herbig Ae/Be stars). However the stars are about a factor 10 more luminous than the luminosity upper limit for Galactic HAeBe stars. This might be due to a shorter accretion timescale (accr = M /M ), or to the smaller dust-to-gas ratio in the LMC. Key words: stars: variables: general ­ stars: emission-line, be ­ stars: pre-main sequence ­ stars: formation ­ ISM: reflection nebulae ­ galaxies: Magellanic Clouds

1. Introduction This paper deals with the study of seven possible luminous premain sequence stars in the Large Magellanic Cloud (LMC). The stars were found serendipitously with the EROS experiment in a search for quasars behind the LMC. The discovery and a first description of their photometric properties was published by Beaulieu et al., 1996 (Paper I) and Beaulieu & Lamers (1997). Here we give a full account of the observations and the criteria that led to the discovery of these stars. We present a description of their photometric and spectroscopic properties, based on medium resolution optical spectra. We list the arguments which

Send offprint requests to: H.J.G.L.M. Lamers

suggest that the stars are massive pre-main sequence stars. This is the first combined photometric/spectroscopic study of extragalactic pre-main sequence stars. Panagia et al. (1998) reported the detection of pre-main sequence stars in the LMC based on strong H excesses as determined from a comparison of narrow band and broad band photometry of the entire field around SN 1987A. The group of luminous pre-main sequence stars, known as Herbig Ae/Be stars or HAeBe stars 1 was first identified by Herbig (1960) on the basis of their presence in star forming regions. The stars are of spectral type A or earlier, lie in obscured regions and illuminate bright nebulosities in their immediate vicinities. By definition these stars have the H line in emission. The HAeBe stars show different types of photometric variability: about 25% of them show irregular brightness variations up to V 3m . These are mainly stars of spectral type A0 or later. There is also a group of of HAeBe stars that show irregular rapid variations of smaller amplitude, typically 0.1m . These stars are usually of spectral type B or A (Finkenzeller & Mundt 1984; Hillenbrand et al. 1992). For reviews see Th´ et al. (1994). e The Galactic HAeBe stars have luminosities between 1.0 < log L/L < 3.2, temperatures between about 8000 Teff 30 000 K and masses in the range of 1.5 to 15 or 9 M (Strom et al. 1972, Berrilli et al. 1992). They are the more massive counterparts of the T Tauri stars. The distribution of the Herbig Ae/Be stars in the HR-diagram has an upper limit which slopes down from about log L 3.2 at log Teff = 4.30 to log L 1.0 at log Teff =3.90. This upper limit indicates the location in the HRdiagram where the pre-main sequence star become detectable at optical wavelengths. It has been suggested that the observed upper limit may coincide with the predicted "birthline" that marks the end of the fast accretion phase (Palla & Stahler, 1993; 1994) or with the path followed by a continously accreting star (Bernasconi & Maeder, 1996). One may expect that in galaxies with a low metallicity, and hence with a smaller dust-to-gas ratio, the pre-main sequence evolution might be different from that of Galactic stars. In this paper we present the observations of irregular photometric variables with strong H emission in the LMC that might

Th´ (1994) has suggested the name "HAeBe" stars for the Herbig e Ae/Be stars and we will use this name.

1

828

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

be the counterparts of the Galactic HAeBe stars. In Sect. 2 we discuss the photometry of the EROS experiment and the selection criteria that were used to find quasars but that turned up the seven HAeBe candidates in the LMC. In Sect. 3 we discuss the photometry of the program stars and their variations in magnitudes and in colour. In Sect. 4 we describe the medium resolution optical spectra of the program stars, we derive the spectral types and we present evidence that all seven stars are in H ii regions. Sect. 5 deals with the extinction, the effective temperatures and the luminosities of the stars and their location in the Hertzsprung-Russell diagram (HR-diagram). In Sect. 6 we present the evidence that the stars are probably pre-main sequence objects, similar to the Galactic HAeBe stars. Therefore we will refer to the program stars as EROS-LMC-HAeBecandidates, or ELHCs. In Sect. 8 we discuss the possible reason for the difference in photometric behaviour between the Galactic HAeBe stars and the ELHCs. In the last section we discuss the results of this study and we point to some of the problems that exist with identification of the stars as HAeBe stars. 2. The observations and the program stars 2.1. The photometric observations The photometric observations were made in the 1991/1992 campaign of the EROS project. EROS (Exp´ rience de Recherche e d'Objects Sombres) is a collaboration between astronomers and particle physicists to search for dark matter in the form of compact objects through microlensing effects on stars in the LMC (Aubourg et al., 1993 and 1995). The observations were obtained using a 0.4 meter f/10 reflecting telescope mounted on top of the Great Prism Objective telescope at ESO La Silla and a 16 CCD camera (Arnaud et al., 1994a and 1994b). Each CCD (579 x 400 pixels) covers a field of 11.6 x 8 arcmin with pixels of 1.25 arcsec. The EROS photometric system consists of a blue and a red band, BE and RE , and was described by Grison et al. (1995). The BE bandpass is close to that Johnson V band (VJ ), but it is broader. The RE bandpass is intermediate between the bands of Cousins RC and IC . The magnitude scale is defined such that an A0 star has BE - RE = 0 with BE VJ and RE RC . The EROS magnitudes and colours can be converted into the V and B - V values of the Johnson photometric system by means of the transformation relations. Grison et al. (1995) derived the transformation relations for stars with a wide range of colours. We have improved the calibration by selecting 24 stars with accurate Johnson photometry in a range (B - V ) and V of our stars VJ = BE (±0.05) - 0.47(±0.1)(BE - RE ) (1)

2.2. The search for variability in the EROS database We have searched for variability in the 80 000 stars of the 19911992 data set in a systematic way. 1. (A) Periodic variability: We have used the modified periodogram method (MPG), which is efficient to search for periodicity of arbitrary shapes in time series by the use of an orthogonal combination of Fourier harmonics. For details of the method, see Grison (1994). For each variable star the following parameters are calculated: (a) the most probable period, (b) the confidence level (i.e. the probability that the observed fluctuation is not due to Gaussian noise), and (c) the first five Fourier coefficients of the best periodic function fitted to the measurements. With this method we are able to detect periodic fluctuations with amplitudes on the order of the photometric precision. The analysis and the first results on periodic variables detected by EROS, i.e. eclipsing binaries, Cepheids, and RR Lyrae are described in Grison et al. (1995), Beaulieu et al. (1995), and Hill & Beaulieu (1996, 1997) respectively. 2. (B) Non-periodic variablity: For non-periodic variability, we use a set of criteria similar to the ones to search the microlensing candidates. To summarize, we search for bumps in observing windows of different sizes. 2.3. The selection criteria to search for quasars Our program stars were serendipitously discovered in the EROS database while searching for QSOs on the basis of optical variability criteria. The detection of QSOs behind the LMC or the SMC would provide background sources for the study of the column density of absorption lines due to the Magellanic Cloud interstellar medium. In the field monitored in 1991/1992 by EROS, about 10 QSOs are expected of which one third should be optically variable. Since it is not possible to isolate QSOs on the basis of two-colour photometry only, it was decided to search for QSOcandidates on the basis of their light curve properties using the following criteria: (1) The objects should be blue: BE - RE < 0.5 mag. This is to avoid the detection of many red irregular variable stars. (2) The objects should be non-periodic variables. If variability is detected with the MPG method, the most probable period has to be greater than half the observing campaign (i.e. P > 60 days). Following these two selection criteria seven blue quasar candidates with irregular variability were found. We obtained optical spectra for all of them. These spectra (which will be discussed in Sect. 4) showed that the objects have stellar spectra and H emission with a radial velocity shift of +270 km s-1 . This is the radial velocity of the LMC. It demonstates that the objects are not quasars but stars or star-like objects in the LMC. They are the program stars of this paper.

(B - V )J = -0.06(±0.02) + 0.91(±0.06) (BE - RE ) (2) During the 117 days of the 1991-1992 campaign, about 2500 CCD images of a field of 1 x 0.4 degrees centered on the bar of the LMC were taken. Observations were taken alternatively in the blue (BE ) and the red (RE ).

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

829

Fig. 1. The area of the Bar of the LMC that was studied by EROS is shown. It was covered by sixteen CCD-frames. We searched for blue irregular variables in eight CCDs in the right half of the EROS-field. (see Fig. 2.)

Fig. 2. The location of the program stars in the EROS field, covered by eight CCDs. Each CCD is 400×579 pixels, corresponding to 8.0×11.6 arcmin. The stars are indicated by their number. Notice that they are found in two groups: one near the lower middle part (CCDs 1 and 2) and one in the upper right hand corner (CCD 8)

2.4. The program stars The coordinates of the program stars, which we call ELHC1 through 7, are listed in Table 1. Coordinates are accurate to within 3 arcsec using the procedure discribed by Grison et al. (1995). For completeness we list the name of the stars in the EROS project. The table also gives the number of observations and the epoch of the observations expressed in days. The numbering of the dates starts at January 1 1990, so 2447862.5 should be added to obtain the Julian Day number. The observations are from the first half of 1992 and cover a total period of 117 days. Each star was observed about 9 times per night in both filters. This results in a total of about 1000 observations per object per filter.

Fig. 1 shows the area of the Bar of the LMC that was observed with EROS. Fig. 2 shows the eight CCD frames that were searched for irregular blue variables. The location of the stars with respect to one another is indicated. Although the selection criteria were applied to all eight EROS CCD-fields in the LMC, the blue irregular variables (quasar candidates) were found in three CCD-frames only: viz. in EROS CCD-frames: nrs 1, 2 and 8. The stars are clustered in two groups. One group contains stars nrs 1, 2, 5 and 6. This group extends over about 11 arcmin = 160 pc. The other group contains stars 3, 4 and 7, in an area with a diameter of about 7 arcmin = 100 pc. Finding charts for the seven stars are available on the Web: http://www.iap.fr/ beaulieu.html.

830 Table 1. The program stars Name1 ELHC1 ELHC2 ELHC3 ELHC4 ELHC5 ELHC6 ELHC7 EROS name 1603-1 111-2 830-8 1335-8 1414-1 1457-1 1005-8 RA(2000) (h,m,s) 05 18 15.3 05 18 33.6 05 16 29.5 05 17 18.3 05 18 21.3 05 18 18.7 05 16 39.5

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

Dec(2000) (0 , , ) -69 30 15 -69 39 42 -69 17 17 -69 21 38 -69 32 39 -69 35 30 -69 20 49

Nr obs. 1087 952 1026 1026 1086 982 1026

Dates2 718-833 717-833 717-829 717-829 717-833 718-833 717-829

1) ELHC = EROS LMC HAeBe Candidate 2) Date since Jan. 1 1990 = JD 2447862.5

3. The photometric variations 3.1. The light curves We have derived the daily average magnitudes for each star and each filter. Since each star was observed on the average 9 times per night, the daily average is the mean value of about 9 observations. The daily average magnitudes have an accuracy of typically 0.02m in BE and RE and about 0.03m in (BE - RE ). Figs. 4 and 3 show the light curves of the seven EROS-LMCHAeBe-candidates in BE , RE and (BE - RE ). The stars are variable in an irregular way with a full amplitude of 0.14m to 0.45m in BE and RE and 0.07m to 0.15m in (BE - RE ). All stars show a light curve consisting of dips and bumps with a duration of about 10 to 20 days. Stars nr 1 and 4 also show evidence for a slow dimming of the star, lasting about 30 to 40 days followed by a rapid brightening in about 10 days. Star nr 7 shows the most pronounced dip of 0.40m in BE and RE with a duration of about 50 days. The mean magnitudes of the program stars in BE , RE and (BE - RE ) over the full period of 117 days are listed in Table 2, together with the maximum and minimum values and the amplitudes. Notice that all stars have a (BE - RE ) colour of about 0.0m . 3.2. The colour-magnitude variations The data in Figs. 4 and 3 and in Table 2 show that the (BE - RE ) colour variations are small. The (BE - RE ) colour curves of stars nrs 1, 2, 3, 4 and 5 show the same type of variation as the light curves in BE or RE but with smaller amplitudes. These stars are bluer when thay are fainter. Stars nr 6 and 7 show some evidence for an opposite behaviour: their (BE - RE )-curve is slightly anticorrelated with the light curve. These stars are redder when they are fainter. The plots of colour versus magnitudes, which are not shown here, confirm these conclusions. We determined the slope of the colour-magnitude variation from a weighted least square fit of the daily averages. The value of the slope d(BE - RE )/dRE with its uncertainty is given in the last column of Table 2. The gradient is between -0.14 and -0.32 for stars ELHC1 through 5. Stars ELHC6 and ELHC7 show a small positive slope with a value of only 0.07 ±0.03. The

Fig. 3. The light curves and the colour curve of star ELHC1 with daily averages. Dates since Jan. 1 1990 (JD =2447862.5)

small positive values of the gradient and its uncertainty show that they do not share the negative colour-magnitude relation of the other stars. We will discuss the nature of the different colour in Sect. 8. 4. The optical spectra 4.1. The observations The spectroscopic observations were performed at ESO La Silla using the ESO 1.5m telescope equiped with a Boller and Chivens spectropgraph in August 1994 for stars EHLC2 to EHCL7. Star EHLC1 has been observed with EFOSC1 in August 1994 in very bad conditions, and a poor spectrum has been obtained. Therefore star ELHC1 was excluded from the rest of this study. The instrumental response has been derived by observing standard stars before and after the spectrum of each program ° star. This give a flux accuracy of about 5% for > 4000 A. The flux calibration at shorter wavelength is much more uncertain. The wavelength scale of the spectra was determined with the He-Ar reference spectrum and the emission skylines ° ° of NI (5201.7 and 5228.9 A) and OI (5578, 6301 and 6364 A), ° at > 4200 A, which cor° to an accuracy of about 0.2 to 0.4 A responds to v 25 km s-1 . For shorter wavelengths we used the emission lines of the very extended H ii region near star ELHC2. This gives a reasonably accurate wavelength scale ° ° down to 4000 A with an accuracy of about 0.8 A or 60 -1 ° the wavelength scale is uncertain by km s . For < 4000 A ° a few A, which corresponds to about 200 km s-1 . The resolution of the spectra of stars nr 2, 4, 7, which were taken with grating nr 4 (4600 groves mm-1 , blaze wavelength ° ° 6825 A, 116 A mm-1 ), has been determined from the FWHM

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

831

Fig. 4. The light curves and the colour curve of stars ELHC1 to ELHC6 with daily averages. Dates since Jan. 1 1990 (JD =2447862.5)

832 Table 2. Magnitudes and colours Star ELHC1 ELHC2 ELHC3 ELHC4 ELHC5 ELHC6 ELHC7 Mean 15.35 15.07 15.66 14.79 15.63 15.84 16.95 BE Min 15.22 14.92 15.52 14.63 15.54 15.77 16.76

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

Max 15.41 15.25 15.77 14.88 15.68 15.99 17.21

Ampl 0.19 0.33 0.25 0.26 0.14 0.22 0.45

Mean 15.30 15.05 15.67 14.70 15.64 15.72 16.90

RE Min 15.15 14.87 15.52 14.53 15.54 15.67 16.72

Max 15.37 15.30 15.78 14.84 15.72 15.86 17.14

Ampl 0.22 0.43 0.26 0.31 0.18 0.19 0.41

Mean +0.05 +0.02 -0.01 +0.08 -0.01 +0.11 +0.06

(BE - RE ) Min +0.03 -0.06 -0.05 +0.04 -0.07 +0.08 +0.02

Max +0.09 +0.09 +0.02 +0.13 +0.03 +0.15 +0.13

Ampl 0.07 0.15 0.08 0.09 0.11 0.07 0.12

d(BE -RE ) dRE

-0.19 ± 0.04 -0.32 ± 0.03 -0.14 ± 0.03 -0.22 ± 0.05 -0.26 ± 0.04 +0.07 ± 0.03 +0.07 ± 0.03

The values of d((BE - RE ))/dRE were derived from a linear least square fit of the magnitudes. Therefore the listed value is different from the ratio of the amplitudes.

of the emission lines of the H ii region near star ELHC2. The mean value of the FWHM of the emission of H, H, H and the [N ii], [S ii] and [O iii] lines corresponds to 230 ± 40 km s-1 . The resolution of the spectra of stars nr 3, 5 and 6, which were taken with grating nr 2 (2300 groves mm-1 , blaze wavelength ° ° 4550 A, 224 A mm-1 ), was determined from three forbidden emission lines of [N ii] and [S ii] near star nr 6. This indicates a resolution of better 150 km s-1 . The optical spectra are shown in Figs. 5 through 10. The wavelengths of the H i, He i and other classification lines (see below) are indicated. The spectra are corrected for the radial velocity of +270 km s-1 of the LMC but not for the motion of the earth around the sun. In order to reduce the noise, we have smoothed the spectra with a rectangular boxcar profile ° with a width of 15 A, which degrades the resolution to about 1000 km s-1 . The smoothed spectra have a S/N ratio at > ° 4200 A of about 20 for star 2, 50 for stars 3 and 4, 75 for stars 5 and 6, and 15 for star 7. At shorter wavelengths the S/N is ° smaller. With a resolution of 15 A we can only detect lines with ° ° an equivalent width larger than about 2 A for a S/N=25 and 1 A for S/N=50. The spectra of B-type stars in the LMC do not show many lines with such large equivalent widths, except those of H i and He i. 4.2. A discussion of the spectra In order to estimate the spectral types, we searched all spectra for the presence or absence of typical classification lines. These ° are the H i and the He i lines and the lines of He ii (4686 A), ° O ii (4642, 4649, 4651 and 4662 A), ° C iii (4647 and 4650 A), ° Si ii (4128 and 4131 A) and Si iv (4089 and ° Mg ii (4481 A), ° ° 4116 A). The lines of O ii and C iii near 4650 A, if present, will blend into one absorption component. The spectra are shown in Figs. 5 through 10. The location of the lines which are probably present is shown by vertical lines: full lines for H i, dotted lines for He i and dashed lines for the classification lines (mainly Si iv and C iii/O ii). We briefly describe the spectrum of each of the program stars, Based on the presence or absence of the spectral features and the classification criteria for spectral types (Jaschek & Jaschek, 1987; Lang 1980, p 566) we have assigned approximate spectral types to the stars. These spectral types are only indicative. The

criteria that we used are derived for Galactic stars. The LMC stars have a lower metallicity than the Galactic stars (ZLMC 0.3ZGal ; Russell & Dopita, 1990) so the strength of the metal lines and the ratio of metal lines relative to H and He lines do not easily fit the spectral type criteria for Galactic stars. Moreover, because of the low resolution and low S/N ratio it is difficult to determine line ratios. Therefore we can only use the presence or absence of the lines as a criteria instead of the line ratios. The spectrum of each star is described.

­ ELHC2 All Balmer lines from H to H8 have emission components. Since the emission lines of [N ii], [O ii] and [S ii] are also strong, the emission components are probably formed in an H ii region surrounding the star. The FWHM of H is similar to the resolution and the equivalent width ° of 104 A is much larger than that of the other stars (Table 3). This suggests that most of the H emission comes from the nebula. Apart from the emission components, the spectrum shows absorption components of H, H and possibly also ° of the higher Balmer lines. The He i lines at 6678 A and ° are absent or weakly in absorption. The feature at 5876 A ° 4650 A due to O ii or C iii is clearly present and there are two absorptions at the wavelengts of the Si iv lines. The Balmer jump is small. These characteristics suggest a spectral type early-B or O. ­ ELHC3 H is in emission, H is filled in with emission, but the other Balmer lines are in absorption. The spectrum does not show a Balmer jump, because the flux level at ° ° about 3600 A is the same as at 3900 A. This star shows the best evidence for the presence of He i absorption lines. The 6678 and 5876 lines are absent, but there are absorptions at the wavelengths of many of the other He i line, in particular 4471, 4388, 4169, 4026, 3965 and 3927. There is no evidence for the presence of other lines. This suggests an early-B spectral type. ­ ELHC4 H is in emission, but the other Balmer lines are in absorption, except H which appears to be filled in. The ° He i lines are weak or absent. There is a feature at 4650 A possibly due to C iii/O ii, but the Si iv lines are absent. There is a small Balmer jump. This suggests a spectral type mid to early-B, with a significant uncertainty.

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

833

° Fig. 5. The spectrum of star ELHC2 at a resolution of 15 A. Notice the strong H and H emission, and the emission lines of [N ii] at 6548 ° ° ° and 6583 A, [O ii] at 3726/3728 A and [S ii] at 6716 and 6730 A of ° the H ii region. The Si iv lines and the O ii/C iii lines at 4650 A are in absorption.

° Fig. 8. The spectrum of star ELHC5 at a resolution of 15 A. H is strongly in emission, H is partly filled-in and the other Balmer lines are in absorption. He i is weak or absent, but the O ii/C iii blend is strong.

° Fig. 6. The spectrum of star ELHC3 at a resolution of 15 A. H and H are in emission. The other Balmer lines and many He i lines are strongly in absorption. Other lines are absent.

° Fig. 9. The spectrum of star ELHC6 at a resolution of 15 A. H is strongly in emission, H is partly filled-in and the other Balmer lines are in absorption. He i is weak or absent, but the O ii/C iii blend is strong.

° Fig. 7. The spectrum of star ELHC4 at a resolution of 15 A. H and H are in emission, but the other Balmer lines are in absorption. The He i lines are weak or absent, and so are the O ii/C iii and Si iv lines.

° Fig. 10. The spectrum of star ELHC7 at a resolution of 15 A. H and H are in emission, and H is partly filled-in. He i may be present in absorption, but the S/N ratio of this spectrum is small.

834

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

­ ELHC5 H is strongly in emission, H is partly filled-in and the other Balmer lines are strongly in absorption. He i is weak or absent, but the C iii/O ii blend is clearly present. This star has the strong Balmer jump of about 0.7m , which suggests a spectral type of very late-B or even A0. ­ ELHC6 H is strongly in emission, H is partly filledin and the other Balmer lines are clearly in absorption up to H16. He i is weak or absent, but the C iii/O ii blend is present. There is a strong unidentified absorption near ° 4726 A. This line is two pixels wide and its reality has to be checked independently. There is a clear Balmer jump. This suggests a spectral type of late-B. ­ ELHC7 The spectrum of this star has a low S/N ratio of about 15. H is strongly in emission and H is weakly in emission. The star has the largest H equivalent width, ° 130 A, of the ELHC stars. Its FWHM is higher than the spectral resolution, suggesting that a considerable part of the emission is of stellar origin rather than nebular. The [S ii] lines are clearly in emission but the [N ii] lines are weak or absent. H is partly filled-in and the higher Balmer lines are in absorption. He i may be present in absorption, but the S/N ratio of this spectrum is too small to be sure. There is evidence for a strong Balmer jump, but this depends strongly on the low-flux calibration. The spectral type is uncertain, but it is possibly mid-B to late-B. The information about the spectra is summarized in Table 3. The information about H was derived from the original unsmoothed spectra. Notice that for all stars, except ELHC 2 and 4, the FWHM of the H emission is considerably wider than than the spectral resolution. This shows that for these stars the H emission is mainly of stellar origin, and not from a surrounding H ii region. For star ELHC2 both the strong forbidden nebular lines and the small FWHM of H suggest that the emission lines are mainly of nebular origin. Star ELHC4 does not show nebular lines, so the H emission is probably of stellar origin, and its true FWHM is less than the spectral resolution of 230 km s-1 . Star ELHC7 has strong nebular lines, but the FWHM of H is larger than the spectral resolution. This suggests that H consists partly of nebular emission but has a significant contribution from a stellar component. The spectral types of the ELHC stars are between spectral types early-B or O and late-B. Although these spectral types are only indicative, they agree rather well with the effective temperature derived below from the energy distribution. 5. Extinction, temperatures and luminosities We can derive the extinction and the stellar parameters in two independent ways: (a) from the calibrated energy distributions of the spectra. (b) from the spectral types and the EROS photometry. 5.1. The stellar parameters derived from the energy distributions The energy distributions from the calibrated optical spectra be° tween 3600 and 7000 A were compared with those from model

atmospheres. We have used the large grid of model atmospheres from Kurucz (1979). The models for the solar metallicity were adopted for the LMC stars, because the grid of solar metallicity models is larger than for lower metallicities and the energy distribution of models with 30 000 < Teff < 10 000 K depends very little on the metallicity. The observed energy distributions were corrected for interstellar extinction with E(B - V ) = 0.17, consisting of the galactic foreground with E(B-V )GAL = 0.07 and of the mean extiction in the Bar of the LMC with E(B - V )LM C = 0.10 (Fitzpatrick, 1986). For both contributions we adopted the extinction curves derived by Fitzpatrick. In addition to the interstellar extinction, the ELHC stars may also suffer from circumstellar extinction. The amount of circumstellar extinction is unknown. Galactic HAeBe stars can have extinctions up to 2 mag. Since the observed colours of our (B - V )J ELHC stars are on the order of (BE - RE ) 0.0m to +0.10m (see Table 5, third column) and the interstellar reddening is already 0.17m the circumstellar reddening cannot be very large. In fact, if we adopt a first estimate for the intrinsic colour of (B - V )0 -0.30 for the early-B/O star nr 2; (B - V )0 -0.20 for the early-B stars nr 3 and 4; and (B - V )0 -0.10 for the late-B stars nr 5 and 6, we find a first estimate of the circumstellar reddening of E(B - V )CS 0.11 for star 2 and E(B - V )CS < 0.06 for the other stars. This indicates that the circumstellar extinction is small and less than about 0.1m in E(B - V ). We have corrected the observed energy distributions for the interstellar extinction of E(B -V )IS = 0.17 and for additional circumstellar extinction with 0.0m < E(B -V )CS < 0.15m in steps of 0.03m . The circumstellar extinction curve was assumed to have the same shape as the interstellar extinction curve of the LMC. The model atmosphere predictions were fitted to the observed energy distributions corrected for the interstellar extinction and for the circumstellar extinction with different values of E(B - V )CS . The accuracy of the fit was expressed in terms model /F . of the standard deviation of the values of log F Table 4 gives the results of this fit. We could find a good fit for each adopted value of E(B -V )CS with only minor differences in the quality of the fit. Therefore we list the best fits for the two extremes of E(B - V )CS =0.0m and 0.15m , together with the model /F ). The best fit for star nr value of and the ratio log (F 2 is reached at a high temperature of 40 000 K. However, at those high temperatures the optical energy distribution is insensitive to Teff . We show this by also giving the fit for Teff =30 000 K, which is only marginally worse. Therefore we adopt a mean value of Teff =35 000 K. The adopted values (Table 4, column 7) are in between the extremes for the two values of E(B-V )CS . Except for star 2, we find temperatures in the range of about 10 000 to 25 000 K, i.e. in the Teff -range of the B-type stars. The effective temperatures have a typical uncertainty of a few thousand K except for the two hottest stars, nrs 2 and 3, which have a larger uncertainty. Star 7, for which we could not derive a spectral type has a temperature of about 14 000 K which suggests about a mid-B spectral type. The temperatures in Table 4, derived from

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC? Table 3. Summary of spectral characteristics Star ELHC 2 3 4 5 6 7 Resol. FWHM (km s-1 ) 230 150 230 150 150 230 H FWHM (km s-1 ) 225 420 240 370 670 450 H W ° (A) 104 15 16 11 20 130 H Balmer jump (mag.) 0.2 0.0 0.4 0.7 0.3 0.6: He i C iii,O ii ° 4650 A yes weak weak strong strong ? Si iv Type

835

emis weak emis weak emis filled-in filled-in emis

no yes weak or absent weak or absent weak or absent present?

yes no no no no ?

early B or O early B mid to early B? late B late B late B?

The information about H was derived from the original spectra.

the energy distribution of the calibrated optical spectra, agree reasonably well with the estimates of the spectral types, derived from the presence or absence of absorption lines. This gives confidence in both determinations of Teff . (Since the predicted energy distributions are not very sensitive to gravity, the values of log g of the best fitting models are not reliable). model /F ) can be used to derive the The values of log (F angular diameter of the stars. Since the distance of the LMC is known (51 kpc; Panagia et al. 1998), the radius of the stars are known. The radius together with the value of Teff gives the luminosity. We derived a radius and a luminosity for the two best fits with minimum or maximum circumstellar extinction, and adopted the mean value for R and log L . These values are also given in Table 4, in columns 8 and 9. The values of R are quite accurate because they are derived directly from the angular diameter, i.e. from the ratio between the predicted and the observed flux. The resulting luminosities have an uncertainty of 0.2 to 0.3 dex. 5.2. The stellar parameters derived from the EROS photometry and the spectral types The stellar parameters can be derived independently if we adopt intrinsic colours from the spectral types derived in Sect. 4 and listed in Table 3. The EROS photometry, combined with these intrinsic colours gives the value of the extinction and the stellar radius. We will argue below in Sect. 8 that the photometric variations of the ELHC stars are mainly due to variable extinction by dust clouds in the vicinity of the stars. The same mechanism explains the photometric variations of Galactic HAeBe stars. Therefore we use the photometry when the ELHC stars are visually brightest to derive the stellar parameters. The magnitudes of the program stars when they are brightest, listed in Table 2, and the EROS colour (BE - RE ) at that epoch are derived from the light curves. These values can be converted into VJ and (B - V )J of the Johnson system by means of the transformation Eqs. (1) to (3). The results are listed in Table 5. The intrinsic colour (B - V )0 is derived from the approximate spectral types, using the relations between spectral type and colour by Schmidt-Kaler (1982), taking into account the uncertainty of the spectral types. This method works only

for stars nr 2 through 6, because the spectral types of star 1 and 7 are unknown. The values of (B - V )0 and the resulting values of the total reddening E(B - V )tot are also listed in Table 5. The interstellar reddening to the Bar of the LMC is E(B -V )IS is 0.17 m (see Sect. 5.1) and the remaining reddening is due to the circumstellar dust. The resulting value of E(B - V )CS is listed in Table 5 column 6. This reddening is between 0.0m and 0.11m . Star 5 has a total reddening of 0.01m which is lower than the value of the interstellar reddening in the Galaxy and the LMC. This means that the circumstellar extinction is negative and about E(B - V )CS = -016 ± 0.06. We will show below (Sect. 8) that stars with a flattened scattering region can have a positive CS reddening if the disk is seen edge-on and a negative CS reddening if the disk is seen pole-on. In the latter case the contribution of the scattered light, which is strongest at shorter wavelengths, dominates over the absorption. The visual extinction AV can be found from E(B - V ) by assuming a ratio of AV /E(B - V ) = 3.0. This is the mean of the value of 3.05 for the Galactic extinction and 2.90 for the LMC (Fitzpatrick 1986). The resulting absolute visual magnitude MV , derived by assuming a distance modulus of 18.60 ± 0.04 (Panagia et al. 1998) is listed in Table 5 column 8. The uncertainty is the same as for AV . The program stars have absolute visual magnitudes between -3.3 and -4.8. The Bolometric Correction of the stars can be estimated from their assigned spectral type. The value will have a significant uncertainty because of the uncertainty in the spectral type. They are derived from the relation between BC and spectral type from Schmidt-Kaler (1982) and are listed in Table 5 column 9. We adopted the relation for giants since the values of MV suggests luminosity class II or III for our stars. The absolute visual magnitudes, the bolometric magnitudes, log L and Teff derived from the assigned spectral types are listed in the last columns of Table 5. (If we had used the mean magnitudes and colours, listed in Table 2, rather than the data at maximum brightness we would have found luminosities which are only 0.1 dex fainter.) The stellar parameters listed in Table 5 are derived completely independent from those in Sect. 5.1 and listed in Table 4. The agreement between the luminosities and temperatures derived in both ways supports the reliablity of the results and

836

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

Table 4. Stellar parameters derived from the observed energy distributions Star ELHC 2 E(B - V )CS 0.00 0.00 0.15 0.15 0.00 0.15 0.00 0.15 0.00 0.15 0.00 0.15 0.00 0.15 Teff (K) 40 000 30 000 40 000 30 000 18 000 25 000 14 000 18 000 9 000 11 000 12 000 12 000 13 000 15 000 log g 4.5 4.0 4.5 4.0 2.5 4.0 2.0 2.5 4.0 4.0 2.0 2.0 4.0 4.0 0.049 0.051 0.063 0.069 0.018 0.012 0.022 0.021 0.068 0.064 0.079 0.064 0.020 0.027

model log (F /F )

Teff (103 K) 35 ± 10

R /R 4.5 ± 1.0

log L (L ) 4.4 ± 0.3

3 4 5 6 7

-23.177 -22.963 -22.915 -22.701 -22.363 -22.340 -21.890 -21.848 -21.850 -21.830 -22.118 -21.853 -22.678 -22.535

22 ± 4 16 ± 2 10 ± 1 12 ± 1 14 ± 1

8±1 14± 1 15± 1 13± 2 6±1

4.2 ± 0.3 4.1 ± 0.3 3.3 ± 0.2 3.5 ± 0.2 3.1 ± 0.2

Table 5. Stellar parameters derived from the EROS-photometry and spectral types Star VJ ELHC EROS 2 3 4 5 6

(B - V )J EROS -0.02±0.03 -0.06±0.03 +0.03±0.03 -0.06±0.03 +0.03±0.03

(B - V )0 spectr. -0.30±0.03 -0.26±0.03 -0.20±0.06 -0.07±0.04 -0.11±0.04

E(B - V ) total

E(B - V ) CS

AV

MV -4.5 -3.7 -4.7 -3.6 -3.4

BC spectr. 2.9±0.6 2.4±0.5 1.6±0.4 0.7±0.3 0.8±0.3

Mbol -7.4±0.7 -6.2±0.6 -6.3±0.5 -4.3±0.4 -4.2±0.4

log L 4.9±0.4 4.4±0.3 4.4±0.2 3.6±0.2 3.6±0.2

Teff /103 spectr. 30±7 25±4 17±3 11±3 13±3

14.90 15.52 14.58 15.54 15.72

0.28±0.04 0.11±0.05 0.84±0.12 0.20±0.04 0.03±0.05 0.60±0.12 0.23±0.07 0.06±0.07 0.69±0.21 0.01±0.05 -0.16±0.06 0.51±0.15 0.14±0.05 -0.03±0.06 0.51±0.15

For star 5 and star 6 we adopted the extinction of the LMC Bar: E(B - V ) = 0.17 Table 6. Predicted and observed radii of H ii regions Star ELHC Teff K 35 000 22 000 16 000 10 000 12 000 14 000 R R 4.5 8 14 15 13 6 rs (pc) n = 1 cm-3 pred 32.0 2.8 0.8 <0.5 <0.5 <0.5 rs (pc) obs 1.2 0.8 1.2 ­ <0.3 0.3

indicates that the estimated spectral types are approximately correct. 6. The H ii regions surrounding the ELHC stars The spectroscopic observations, made with a long slit of 400 pixels = 57", of the stars show the presence of emission lines on both sides of the stellar spectrum of stars nr 2, 3, 4, 6 and 7. If star 5 is in an H ii region, its diameter is smaller than 1 arcsec. We should point out here that the H emission, measured in the stellar spectrum and shown in Figs. 5 to 10 is much stronger than the nebular emission. So most of the emission comes from either the star itself or from a region within 0.5 arcsec (0.12 pc) from the star. We can compare the observed sizes of the H ii regions with predicted sizes as a function of Teff . The number of photons in the Lyman continuum has been calculated by Panagia (1973) and by Vacca et al. (1996) for O-stars with solar metalicity. The values given by Vacca et al. are based on a new calibration of MV and Teff as a function of spectral types for O-stars. Although these later authors find a large difference with the earlier data by Panagia for the same spectral types, the values are about the same as a function of Teff . We use the values by Vacca et al. for the O-stars down to B0.5 and those by Panagia to lower temperatures down to Teff = 17 000 K. The predicted radii of the

2 3 4 5 6 7

H ii regions for the ELHC stars based on the predicted number of Lyman continuum photons by Vacca et al. and Panagia, but scaled to the effective temperatures and radii of the ELHC stars as given in Table 4 are listed in Table 6. We adopted the predicted Lyman fluxes for luminosity class III stars and an ambient o density of 1 atom cm-3 . For a given value of Teff the Str¨ mgren radius scales with (R n)2/3 . ­ ELHC2 This star has the largest H ii region which has a diameter of about 10" (radius 1.2 pc) with strong emission lines (see below). This is in agreement with the fact that 35 000 K this is the hottest star in our sample. For Teff we expect a Str¨ mgren radius of 32 pc and for 25 000 K we o

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

837

­

­

­

­

­

expect 7 pc. The observed radius is much smaller, possibly due to a higher density. The H ii region near star 2 has a complicated structure and seems to consist of several H ii regions close together. The star is located in a huge H i complex, in between two H ii concentrations. (The strength of the emission lines from this nebula is given in Appendix A.) ELHC3 The H ii region has a diameter of 6" (radius 0.8 pc). o For Teff 22 000 K we expect a Str¨ mgren radius of about 2.8 pc. The star is slightly off-center from the H ii region. ELHC4 The H ii region has an observed diameter of 10" (radius 1.2 pc) and a predicted radius of 0.8 pc. The star is strongly off-center from the H ii region. This might indicate that either the distribution of gas in the surrounding of this star is asymmetric, or there is another hot star nearby that is responsible for the ionization. ELHC5 The star is in a crowded region with a star at either side at a distance of 1.4". Therefore we have no information about an H ii region near this star. At a temperature of Teff 10 000 K we do not expect an observable H ii region. ELHC6 This star has a very faint H ii region. We could not measure its size. The faintness of the H ii region is qualitatively in agreement with the late-B spectral type. ELHC7 The star has a very small H ii region with a diameter of only 3" (radius 0.3 pc). For Teff 14 000 K we expect a Str¨ mgren radius of < 0.5 pc. o

evolutionary tracks of massive stars are. (The evolved B and A type giants are also in this region). Based on these arguments we conclude that our program stars are most likely pre-main sequence stars. The spectral characteristics of our program stars also show similarities to those of Galactic normal Be stars. The H emis° sion of normal Be stars is usually in the range of 10 to 100 A, which is similar to the values of our program stars (Schild 1978). Many normal Be stars show photometric variations, but the large majority varies by less than 0.1m in V (Feinstein & Marraco 1979). Most of these variations are on timescales of hours to days (Dachs 1982). A minority of the Be stars vary by more than about 0.1m in V . This variation occurs on timescales of more than about one month, and mostly on timescales of years. These stars are usually redder when they are brighter, with a gradient (B - V )/V -0.1 to -0.3 (Dachs 1982). Several of the characteristics of the ELHC stars are similar to those of a small subset of the Be stars: e.g. the amplitude of the photometric variations and the fact that they get redder when brighter. However, the timescale of the variations and the light curves are different from those of the Be stars: the ELHC variations have amplitudes of 0.15 to 0.40 in VJ on timescales of 10 days to months, whereas the medium term variations of Be stars have amplitudes between 0.1 and 0.2 mag. and periods between 5 days and tens of days (Stefl 1998). The observed photometric variations of the ELHC stars more strongly resemble the irregular variations seen in Galactic HAeBe stars. Therefore we conclude that the ELHC stars are most likely pre-main sequence objects, similar to the Galactic HAeBe stars. A recent study of stellar populations near the ELHC stars confirms that the stas are in star forming regions, see Sect. 9. (If the stars are not pre-main sequence stars, they are some kind Be-type stars with peculiar variability). 7.2. The location in the HR-diagram Fig. 11 shows the location of our program stars and of the Galactic pre-main sequence stars in the HR-diagram. The values of Teff and log L are derived from the energy distributions as listed in Table 4. The data from the EROS photometry, listed in Table 5 are quite similar and approximately within the error bars of the figure. The location of the Galactic HAeBe stars (Berrilli et al. 1992) is also given. The HAeBe stars with the most uncertain parameters, indicated in Fig. 5 of Berrilli et al., are omitted. The luminosities derived by Hillenbrand et al. (1992) are generally lower because they assumed that part of the luminosity is from the accretion disk. The figure shows clearly that the ELHC stars are more luminous than the vast majority of the Galactic HAeBe stars by about a factor 10 to 30. The luminosity of the ELHC stars, plotted in Fig. 11 and a comparison with the pre-main sequence evolutionary tracks from Iben (1965) (only the main-sequence position is shown in the figure) suggests that the ELHC-stars have masses in the range of 8 to 20 M . This is higher than the mass range for the Galactic HAeBe stars of 1.5 to 15 M (Strom et al. 1972).

7. The nature of the ELHC-stars: pre-main sequence stars? 7.1. Comparison with galactic HAeBe stars We suggest that our program stars are pre-main sequence stars, quite similar to the Galactic HAeBe stars. The arguments for this suggestion are: 1. The strong H emission with equivalent widths of 10 to ° 130 A is quite similar to the H emission of HAeBe stars ° which have W 5 to 130 A (Finkenzeller & Mundt 1984). 2. Irregular and unpredictable photometric variability is characteristic for Galactic HAeBe stars. The time scale of the variations of the HAeBe stars is typically days to tens of days (e.g. Th´ 1994; Herbst 1994; Shevchenko et al. 1994). e The time scale for photometric variations of our program stars extends to longer periods of about 10 to 40 days. The photometric variations of the Galactic HAeBe stars are typically less than about 0.1m for stars of types earlier than B7 and of 1 to 2m for stars later than A0 (Finkenzeller & Mundt 1984; Th´ 1994). The ELHC stars vary by a few tenths of e magnitudes. (We return to this in Sect. 8.) 3. HAeBe stars are associated with nebulosities and are found in star forming regions. All our program stars are in H ii regions, either due to the stars themselves or due to nearby hot stars. This indicates that the stars are in young, possibly star forming regions. 4. The stars are located in the HR-diagram to the right of the main sequence (see below) where the pre-main sequence

838

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

Fig. 11. The comparison of the ELHC stars (large dots) with Galactic pre-main sequence stars (small open circles) from Berrilli et al. (1992) in the HR-diagram. The dashed line is the predicted "birthline" for an accretion rate of 10-5 M yr-1 and the dotted line is for 10-4 M yr-1 from Palla & Stahler (1993; 1994). The full line is the zero age main sequence, with the masses indicated. The ELHC stars are from left to right: nr 2, 3, 4, 7, 6 and 5.

The dashed line in Fig. 11 is the "birthline" of the Galactic HAebe stars for a fast accretion timescale of accr M /M with an accretion rate of M = 1 × 10-5 M yr-1 calculated by Palla & Stahler (1993). This is the predicted location where the fast protostellar accretion ends and the slow evolution to wards the main sequence starts. (Note that M is not the present accretion rate, which might be much smaller.) The observed upper limit for the Galactic HAeBe stars agrees approximately with the predicted birthline for a fast accretion rate of 10-5 M yr-1 . However, the ELHC stars are above this birthline by about a factor 3 or 10 in luminosity. The luminosity of the predicted birthline depends on the accretion rate. A higher accretion rate implies that the accretion ends at an earlier phase, when the star has a lower value Teff . The predicted birthline for a fast accretion rate of 10-4 M yr-1 is also plotted in Fig. 11 (from Palla & Stahler 1993). This line agrees better with the observed location of the ELHC stars. The reason for the higher accretion rate in the LMC is not known, but it is possibly related to the lower metallicity of the LMC compared to the Galaxy. 8. The photometric variability of the ELHC stars The program stars are irregular photometric variables. (In fact, this was one of the selection criteria for the quasar candidates.)

The Galactic HAeBe stars are also irregular variables. However there is an important difference between the variations of the HAeBe stars and the ELHC stars: the Galactic HAeBe stars are usually redder when they are fainter (with a few exceptions), but the ELHC stars are usually slightly bluer when they are fainter (with ELHC6 and 7 as exceptions). In this section we suggest an explanation for this difference. The photometric variations of Galactic HAeBe stars are explained in terms of variable circumstellar (CS) extinction, due to irregularities in the outer part of the CS disk. This explanation is supported by the detection of variable linear polarization (Grinin et al. 1988, Grinin 1994). There are a few special exceptions: some stars show a "blueing" effect at a deep photometric minimum. This is explained by the presence of a scattering region around the star. Since the scattering is more efficient at short wavelengths than at long wavelengths, the scattered radiation will be brighter in the blue than in the red. If the obscuring dust cloud is smaller in projection than the extent of the scattering region and obscures the star efficiently (by a few magnitudes), the radiation from the scattering region will dominate the radiation observed in the aperture and the star will be fainter but bluer. This explains why the blueing effect is only seen during deep photometric minima for Galactic HAeBe stars observed with a small aperture, i.e. for a small contribution by the scattering region (Th´ 1994, Grinin 1994). e Why would the majority of our program stars show this blueing effect at all phases? We suggest that this is because we are observing the stars from a distance of 50 kpc. Therefore the scattering nebula, associated with the HAeBe stars, will be unresolved and will contribute significantly to the observed brightness of the object. For instance, in the case of the Galactic HAeBe object V380 Ori, which has a large nebula of about 1 arcmin in diameter, the projected nebula at the distance of the LMC would spread over 0.6 arcsec only and would not be resolved in the EROS photometry. So the scattered light from the nebula will contribute significantly to the EROS photometry. Fig. 12 shows three geometrical configurations to explain the different colour-magnitude variations of the ELHC stars in the LMC. Each model consists of a star with a scattering region and a passing obscuring dust cloud. The EROS photometry of the observed object includes the radiation from the star as well as from the scattering nebula. We express the effect in terms of BE and RE variations when an obscuring dust cloud passes in front of the scattering region (position 1) or in front of the star (position 2). I: a spherical scattering region In case I the star is embedded in a spherical scattering cloud. If there is no obscuration, the colour of the star will not be affected by the presence of the scattering cloud, because the decrease of the short wavelength flux from the star is exactly compensated by the radiation from the scattering region. If the dust cloud is in position 1 the scattered blue radiation will be reduced more than the red radiation and so the object will be (slightly) fainter and redder. If the cloud is in position 2 the object will also appear

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

839

Fig. 12. Three configurations of a scattering nebula around a HAeBe star: (I) a spherical nebula, (II) a flat edge-on nebula, (III) a flat pole-on nebula. The position of an obscuring cloud next to the star (position 1) and in front of the star (positon 2) is indicated. The effect on the magnitude and colour variations are described in Sect. 8.

on the brightness of the scattering disk compared to the star and on the size of the dust cloud. In position 2 the object will appear fainter and bluer, with the blueing depending on the brightness of the scattering disk. We see that obscuration by a dust cloud can make the object either redder or bluer, depending on the geometry of the scattering region. There are two cases when the object will get bluer when it is fainter: either a spherical scattering region, or a scattering disk that is seen pole-on (or better: not edge-on). Since these two are the most likely configurations anyway, it is not surprising that five of our program stars show this blueing effect. Only two of our program stars, ELHC6 and ELHC7, are slightly redder when they are fainter. These stars either have a flat scattering region that is seen edge-on, or they have a scattering region that is small. If the scattering region is flat, the colour of the object can be either bluer (pole-on) or redder (edge-on) than the intrinsic colour of the star, even in the absence of an obscuring cloud. This effect might be the reason why the value of E(B - V ) of star ELHC5 is smaller than that expected for stars in the Bar of the LMC. If this effect is important, it casts some doubt on the accuracy of the effective temperature determinations based on the EROS photometry or the spectrophotometry. However, the reasonably good agreement between the spectral types and the temperatures determined from the energy distributions suggests that the effect of a flat scattering disk is not very important for most ELHC stars. It can be checked by better spectral type determinations based on higher resolution spectra. 9. Discussion and conclusions We discussed the properties of a group of seven irregular blue variables in the LMC, discovered in the EROS database in a search for quasars behind the LMC. The stars are variable with an amplitude of about 0.15m to 0.45m in BE and RE . Five of them are slightly bluer when they are fainter and two are slightly redder when they are fainter. Low resolution spectra show that the stars have strong H emission. The equivalent ° width of the H emission is between 11 and 130 A, and the FWHM is between 225 and 670 km s-1 . Most stars also have H in emission or filled-in with emission, but the higher Balmer lines are usually in absorption. From the presence or absence of spectral features, we try to estimate spectral types. This is hampered by the low resolution of about 250 km s-1 and by the fact that the metal lines in the spectra of LMC stars are weaker than in Galactic stars, due to the lower metallicity. We find that the stars are of spectral types between early-B or O (stars ELHC2 and ELHC3), early to mid-B (ELHC4) and lateB (ELHC5, ELHC6 and ELHC7). We have derived the luminosities of the stars in two independent ways. (a) In the first method we use the EROS photometry (colour and magnitude) together with the intrinsic colour (B - V )0 from the spectral types to derive the total extinction, i.e. the interstellar plus circumstellar extinction. Assuming a ratio of AV /E(B - V ) = 3.0 we derived the visual extinction and the

fainter and redder, but more so than with the cloud in position 1. If the stellar radiation is obscured to a level where the scattered radiation becomes dominant, the object will be fainter and bluer. II: a scattering disk seen edge-on In this case the object will always be redder than the star itself, even in the absence of a dust cloud, because a large fraction of the scattered radiation (perpendicular to the disk) will not be observed. This is predominantly blue radiation. If the obscuring dust cloud is in front of the scattering disk, the brightness and the colour of the object will hardly change because the contribution by the scattered light in the plane of the disk is small. If the cloud is in position 2 in front of the star the object will be fainter and redder. III: a scattering disk seen pole-on In this case the object will always appear bluer than the star itself, because of the significant contribution by the scattered blue radiation and the small or lack of reddening of the star by the scattering disk. This implies that the object appears to have a negative extinction. If the cloud is in position 1 the object will appear redder and fainter. The amplitude of the change depends

840

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC?

absolute magnitudes MV . Adopting the bolometric corrections from the spectral types, we find the bolometric magnitudes and L . (b) In the second method we used the absolute calibrated spectrophotometry of 6 stars. The observed energy distribution was compared to that expected from model atmospheres of different values of Teff and different values of E(B - V ). We adopted the galactic extinction law for the galactic foreground extinction with E(B - V ) = 0.17 and the extinction law of the LMC for the unknown circumstellar extinction. The best fit between the observed and predicted energy distributions gives the value of Teff as well as the extinction. This method does not give a unique solution because we could always find several good fits for different values of the E(B - V )CS . Therefore we adopted the range of 0 < E(B -V )CS < 0.15 mag., which is suggested by the EROS photometry. We then adopted the mean value of the two extremes, with the corresponding uncertainty limits. The results of the two independent determinations of L agree reasonably well with one another (Tables 4 and 5). The values of Teff derived from the energy distributions in method (b) agree approximately with the spectral types. This gives confidence in our estimated spectral types and in the determinations of L . It also shows that most of the luminosity is due to the star and not to an accretion disk. The luminosities are in the range of 3.1 < log(L /L ) < 4.9. In this analysis we have assumed that the circumstellar extinction curve has the same shape as the interstellar extinction curve in the LMC. This is not necessarily the case. If the circumstellar extinction is due to scattering by large particles, it may have a gray component. In that case we may have underestimated the extinction and the luminosities might be even higher. On the other hand, our observed energy distribution always includes that of the scattering nebula because of the large distance of the LMC. This means that scattered radiation is included in our measurements. So, if the scattering nebula is more or less spherical, gray scattering will not affect our results. (The effect of a circumstellar scattering nebula on the colours and the colour variations of the ELHC stars is discussed in more detail in a forthcoming paper by Beaulieu et al.) The stars are located in the HR-diagram to the right of the main sequence (see Fig. 11). The spectral types, the H emission, the photometric variability, the location of the stars in the HR-diagram and their concentrated location in small groups in the Bar of the LMC suggest that the stars are massive pre-main sequence stars, quite similar to the Galactic HAeBe stars. The ELHC stars are brighter than most of the Galactic HAeBe stars and they are located above the birthline of the Galactic pre-main sequence stars in the HR-diagram. This might be due to either (a) a shorter accretion timescale for Galactic HAeBe stars due to the lower metallicity in the LMC, so that the high-accretion phase ends earlier or (b) a lower dust-to-gas ratio in the LMC, again due to the lower metallicity.

We discuss these two possibilities. (a) The birthline depends on the accretion rate in the early evolution of the HAeBe stars (i.e. before they become visible) because the stars only become optically visible before they reach the main sequence, if the accretion timescale, accr M /M , is shorter than the Kelvin Helmholtz timescale. So a high accretion rate implies that more massive stars can be observed before they reach the main sequence. The data roughly suggest an accretion rate of about 10-4 M yr-1 (Fig. 11). The present accretion rate must be much smaller, because the spectrum shows that the energy distribution is dominated by a stellar contribution, rather than by radiation from an accretion disk. Bernasconi & Maeder (1996) have calculated the pre-main sequence evolution of stars for a metallicity of Z = 0.020 and Z = 0.001. They find that the contraction time is shorter in the low-Z models by about a factor 2 in the mass range of 5 to 25 M . This agrees qualitatively with the higher birthline for the LMC than for the Galaxy. (b) If the low optical brightness of a pre-main sequence object during the accretion phase is (partly) due to circumstellar extinction, the low dust-to-gas ratio of the LMC will also help to make the stars appear optically visible in an earlier phase than their Galactic counterparts. If metallicity plays a crucial role in either the accretion rate or in the extinction by dust in the accretion phase, we may expect that the birthline of HAeBe stars in the SMC is even higher in the HR-diagram. We are presently involved in a study of pre-main sequence stars in the SMC. The identification of the stars as extragalactic HAeBe candidates depends on a combination of properties: spectroscopic, photometric and their location in the LMC. The identification can be confirmed by a study of higher resolution spectra, which will result in better spectral type determinations and may show evidence for circumstellar disks. If the stars are indeed pre-main sequence objects, we expect them to occur in star forming regions with other pre-main sequence stars. A study of the stellar populations in terms of colour-magnitude diagrams around the ELHC stars would be very useful. We have started such a study and the first results are encouraging in the sense that we find a significant population of young objects near most of the ELHC stars. The location of all seven ELHC stars in the HR-diagram above the majority of the Galactic HAeBe stars could be due to a selection effect. In the selection criteria (Sect. 2.3) that were used to find these objects we did not set a lower limit to the brightness. However, the detection of the photometric variations depends of course on the photometric accuracy of the data, which in turn depends on the brightness of the stars. So the brighter objects which satisfied the selection criteria will be found more easily than the possible fainter objects. This implies that the ELHC stars studied in this paper may be only the brightest ones, i.e. the tip of the iceberg. We are presently making a statistical study based on a larger sample of candidate pre-main sequence stars in the LMC and the SMC.

H.J.G.L.M. Lamers et al.: Luminous pre-main sequence stars in the LMC? Table A1. Emission lines of the H ii region near ELHC2 obs 6730 6715 6582 6562 6546 5006 4860 4405 4338 4306 4101 3765 3728 theor 6731 6717 6584 6562 6548 5007 4861 4340 4102 3727 Ion [SII] [SII] [NII] H [NII] [OIII] H H H [OII] FWHM 4.5 4.3 5.0 4.9 3.7 4.1 4.0 3.2 3.9 3.7 2.1: 3.6 5.2 W /W (H) 0.12 0.15 0.18 1.00 0.06 0.03 0.07 0.03 0.02 0.04 0.001: 0.03 0.17

841

In this study we have adopted a distance modulus of 18.60 mag. for the LMC. If we had adopted 18.3 mag. the stars would be fainter by 0.3 mag. (Table 5) but they would still be above the Galactic birthline.

Acknowledgements. We thank Dolf de Winter, Vladimir Grinin, Andre Maeder, Gwendolyn Meeus, Stanislav Stefl and Paul Wesselius for useful discussions and Nino Panagia for constructive comments on the manuscript. JPB gratefully acknowledges the financial support of the French Acad´ mie des Sciences. e

Appendix A: the H ii region around ELHC2 The H ii region around star nr 2 has the largest number of emission lines. Table A1 gives the list of the lines, their FWHM and the relative intensities. The equivalent width of the lines were measured at a distance of about 4" from the star. The intensities are expressed relative to the equivalent width of H. References

Arnaud M., Aubourg, E., Bareyre P., et al., 1994a, Exp. Astron 4, 265 Arnaud M., Aubourg, E., Bareyre P., et al., 1994b, Exp. Astron 4, 279 Aubourg E., Bareyre P., Br´ hin S., et al., 1993, Nat 363, 623 e Aubourg E., Bareyre P., Br´ hin S., et al., 1995, A&A 301, 1 e Beaulieu J.P., Grison P., Tobin W., et al., 1995, A&A 303, 137 Beaulieu J.P., Lamers H.J.G.L.M., Grison P., et al., 1996, Sci 272, 995 (Paper I) Beaulieu J.P., Lamers H.J.G.L.M., 1997, In: Ferlet R., Maillard J.P. (eds.) Variable stars and the astrophysical return of microlensing surveys. Editions Frontieres, p. 319

Berrilli F., Corciulo G, Ingrosso G., et al., 1992, ApJ 398, 254 Bernasconi P., Maeder A., 1996, A&A 307, 829 Dachs J., 1982, In: Jaschek M., Groth H.G. (eds.) Be stars. Reidel, Dordrecht, p. 19 Feinstein A., Marraco H.G., 1979, AJ 84, 1713 Finkenzeller U., Mundt R., 1984, A&AS 55, 109 Fitzpatrick E.L., 1986, AJ 92, 1068 Grinin V.P., 1994, In: Th´ P.S., Perez M.R., van den Heuvel E.P.J. (eds.) e Nature and evolutionary status of Herbig Ae/Be stars. ASP Conf. Ser. Vol. 62, p. 63 Grinin V.P., Kiselev N.N., Minikhulov N.H., Voshchinnikov N.V., 1988, Soviet Astr. 14, 219 Grison P., 1994, A&A 289, 404 Grison P., Beaulieu J.P., Pritchard J.D., et al., 1995, A&AS 109, 447 Herbig G.H., 1960, ApJS 4, 337 Herbig G.H., 1994, In: Th´ P.S., Perez M.R., van den Heuvel E.P.J. e (eds.) Nature and evolutionary status of Herbig Ae/Be stars. ASP Conf. Ser. Vol. 62, p. 3 Herbst W., 1994, In: Th´ P.S., Perez M.R., van den Heuvel E.P.J. (eds.) e Nature and evolutionary status of Herbig Ae/Be stars. ASP Conf. Ser. Vol. 62, p. 35 Hill V., Beaulieu J.P., 1997, In: Ferlet R., Maillard J.P. (eds.) Variable stars and the astrophysical return of microlensing surveys. Editions Frontieres, p. 267 Hillenbrand L.A., Strom S.E., Vrba F.J., Keene J., 1992, ApJ 397, 613 Iben I., 1965, ApJ 141, 993 Jaschek C., Jaschek M., 1987, In: The classification of stars. Cambridge Univ. Press Kurucz R.L., 1979, ApJS 40, 1 Lang K.R., 1980, Astrophysical Formulae. Springer, Berlin, p. 566 Bernasconi P.A., Maeder A., 1996, A&A 307, 829 Panagia N., 1973, AJ 78, 929 Panagia N., Romaniello M., Scuderi S., Kirshner R.P., 1998, ApJ, submitted Palla F., Stahler S.W., 1993, ApJ 418, 414 Palla F., Stahler S.W., 1994, In: Th´ P.S., Perez M.R., van den Heuvel e E.P.J. (eds.) Nature and evolutionary status of Herbig Ae/Be stars. ASP Conf. Ser. Vol. 62, p. 391 Rusell S.C., Dopita M.A., 1990, ApJS 74, 93 Schild R.E., 1978, ApJS 37, 77 Schmidt-Kaler Th., 1982, In: Landolt Bornstein Vol. 2, Subvol. B, Springer Verlag, Berlin Shevchenko V.S., Grankin K.N., Ibragimov M.A. et al., 1994, In: Th´ e P.S., Perez M.R., van den Heuvel E.P.J. (eds.) Nature and evolutionary status of Herbig Ae/Be stars. ASP Conf. Ser. Vol. 62, p. 47 Strom S.E., Strom K.M., Yost J., et al., 1972, ApJ 173, 353 Th´ P.S., 1994, In: Th´ P.S., Perez M.R., van den Heuvel E.P.J. (eds.) e e Nature and evolutionary status of Herbig Ae/Be stars. ASP Conf. Ser. Vol. 62, p. 23 Vacca W.D., Garmany C.D., Shull J.M., 1996, ApJ 460, 914

Information

2300827.DVI

15 pages

Find more like this

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate

1137274


You might also be interested in

BETA
2300827.DVI