Astron. Astrophys. 326, 143-154 (1997)

6. Discussions

6.1. A word of caution about the colors

The powerful deconvolution method decribed in Sect. 2.1 has allowed us to resolve the stars populating the core of the LMC OB association LH 39 in more than 30 components. More especially, it has shown the presence of two previously unknown stars, #7 and #21, which are the closest to R 84 at a resolution of 0 .19 (FWHM) in the optical domain. Furthermore, this deconvolution code has enabled us for the first time to perform the photometry of the components. However, we should underline that this photometry is relative for a number of reasons which have nothing to do with the limitations of the code. We have used Stahl et al.'s (1984 ) observations of R 84 to calibrate our UBVR observations carried out at epochs different from theirs. Moreover, since the calibration is based on only one "standard star", color corrections have not been possible. This shortcoming particularly affects the colors, and we have therefore taken care not to over-interpret them. Another point, as discussed in Sect. 5, is the slight color variability of R 84. However, this variability, of the order of 0.05 mag, is smaller than our inaccuracies. We envisage high resolution photometric observations including standard stars to improve the colors and use them for further study of R 84.

6.2. The late-type component

R 84 is unique since it shows the features of both a transition Ofpe/WN9 type star and a late supergiant. No other star of this family is known to be associated with an evolved late-type star. The very large IR excess observed towards R 84 was attributed by Allen & Glass (1976 ) to the presence of a late-type supergiant component that provides the near-IR flux. This component was spectroscopically detected by Cowley & Hutchings (1978 ) who classified it as M2 on the basis of the relative strengths of TiO bands at 5167, 5448, and 6159. Later, Wolf et al. (1987 ) published a CASPEC spectrum containing several absorption lines mainly of neutral elements and TiO bands and confirmed the spectral type M2 Ia. Also, McGregor et al. (1988 ) interpreted the CO absorption bands in the 2.0-2.4 µm region of the spectrum of R 84 to be due to a cool supergiant companion star.

Cowley & Hutchings (1978 ), using a K magnitude of 8.49 (Allen & Glass 1976 ) and Johnson's calibration for M0 stars, estimated an absolute magnitude of = -6.7 for the supergiant component. However, this is certainly an overestimate, since it assumes that the W-R star has no important contribution to the K flux. Recent works on Ofpe/WN9 stars, LBVs, and related objects have shown the presence of extended envelopes around these stars that produce a strong IR excess in their spectrum. Moreover, in the particular case of R 84, Stahl et al. (1984 ) provide evidence for the existence of a circumstellar dust shell especially on the basis of a large K - L color that cannot be explained by a late-type star alone.

From an absolute magnitude of = -5.6, expected for an M2 Iab star (Lang 1992 ) and assuming that the putative late-type supergiant is a foreground star, we derive a V magnitude of 13.2. This would be the brightest star in the field of view after R 84. However, no star as bright is disclosed by our photometry using resolutions of 0 .12 (in H band) and 0 .19 (in V) in a field of 15 centered on R 84. The second brightest star of Table 2 with red colors, i.e. #34 with V = 14.18, is too far apart to fall into the large apertures (diameters of order 15 ) used in the classical photometry or to contribute to the spectrum of R 84. The best candidate for the late-type companion seems therefore to be star #7, lying at 0 .46 N, 1 .60 W of R 84. It is the brightest and the reddest star of the near-IR sample and also the closest to R 84 (Table 3, Figs. 2, 3, 6). Moreover, its H - K = 0.22 mag, although suffering from a rather large uncertainty, is compatible with that expected for an M2 type (Koornneef 1983 ). However, with its V = 17.47 0.2 or K = 14.10 0.22, it is extremely weak, unless it varies strongly. We will see below (Sect. 6.3) that star #7 is probably a line-of-sight object rather than a genuine association member.

R 84 is known to be variable. Its V brightness has been reported to change by 0.2 mag between 1972 and 1984 (see Stahl et al. 1984 and references therein). However, we do not know whether there is an offset in the zero points used by various observers. Anyhow, Stahl et al. (1984 ), using a homogeneous set of observations, found a magnitude variation of V = 0.08 between 1983 and 1984. The variations in the near-IR magnitudes are also comparable to those in the optical domain (Allen & Glass 1976 , Stahl et al. 1984 ). From the similarity of the color variations of R 84 with those of R 85 and R 99 Stahl et al. (1984 ) suggest that they might be due to the blue star. However, since in the reported classical photometries rather large apertures are used (15 ) we cannot a priori exclude the possibility that this similarity may be a coincidence and that part of the observed variations may be due to the other cluster stars falling in the aperture. However, in order to get a variation of 0.08 mag in the integrated V magnitude, star #7 or another star of the field should undergo rather unrealistic variations.

Although we cannot preclude the presence of a line of sight companion lying closer than 0 .12 to R 84, the possibility of a binary system as suggested by McGregor et al. (1988 ) is very appealing. In order to investigate the possibility of a binary system, one has to measure the actual radial velocity of R 84. However, this is not an easy task, since Crowther et al. (1995 ) have shown that even the He II 4542 absorption is not a pure photospheric line, but is already affected by the stellar wind. Therefore, one has to rely on the radial velocities of pure emission lines. Cowley & Hutchings (1978 ) measured a radial velocity of km s^-1 for the N iii, Si iv, He II and C iii emission lines on their spectrum obtained in Nov. 1977. Moffat (1989 ) found no indication of variations on time scales from a day to a year in his He II 4686 data obtained in 1978 and 1980. He derived a mean radial velocity of 208 km s^-1 with a standard deviation of 5 km s^-1. Nota et al. (1996 ) measured a radial velocity of 235 km s^-1 in 1991 September, whereas our spectrum obtained in 1989 September yields a value of 222 km s^-1. Considering the mean radial velocity of the Of emission lines He II 4686 and N iii 4634-40, we derive a mean velocity of 205 km s^-1 from Moffat's (1989 ) data set, whereas our spectrum yields a value of 226 10 km s^-1 and Nota et al. (1996 ) find 246 13 km s^-1 including also the H and He i 6678 lines in their mean. We have also measured the radial velocities on the AAT spectrum of Smith et al. (1995 ) kindly provided by Dr. P. Crowther. We obtain a velocity of 269 km s^-1 for the He II 4686 line and 256 km s^-1 for the mean of the Of emission lines. These data are too scarce to draw any firm conclusion on the binary status of R 84, but they clearly indicate the presence of radial velocity variations and are not incompatible with the possibility of a long-period low-inclination orbit.

6.3. Cluster membership

The color-magnitude diagram in the V - R, V plane for stars with the deconvolution photometry is shown in Fig. 6. Two distinct groupings appear, a vertical, blue one lying at V - R 0.1 and a smaller, red population with V - R 0.5 mag. However, the apparent "main sequence" may be contaminated by evolved, foreground stars less affected by reddening. We must therefore check the effects of reddening upon the abscissa. A two-color U - B, B - V diagram turns out to be a useful complement to separate the foreground stars from the association members. Unfortunately, in the field of R 84 we have only 11 stars with measured U, B, and V photometry. Most of them appear to have a reddening-free index (Massey et al. 1995 ) between and and lie on reddening lines between spectral types B5 and B0. These stars (#2, 3, 4, 5, 14, 36, 38) are therefore very likely members of the OB association. However, photometry alone is not sufficient and as long as spectral types are not derived from spectroscopy we should be very cautious about these results. Three stars, #7, #21, and #37, stand out of the main group. These stars are subject to large uncertainties (Sect. 2.1.3). Star #37 is particularly faint in U and the fact that the U image could not be flat-fielded introduced a large error in this band. The error bars on the colors of these stars are therefore 0.6, 0.4, and 0.3 mag respectively. Whereas star #37 is most likely a red foreground star, the situation is less clear for star #21 for which more accurate photometry is needed. The same holds true for star #7, although this star is more likely a line-of-sight component.

Fig. 6 represents a small group of stars concentrated in the core of the association, whereas the color-magnitude diagram of the full SUSI field is displayed in Fig. 7. The magnitudes were obtained using S-Extractor, an efficient package for aperture photometry of large numbers of sources in large fields (Bertin & Arnouts 1996 ). The code has allowed us to derive the photometry of the brightest, non-blended stars for comparison with the results obtained using deconvolution in the smaller field. Since R 84 is saturated in the R filter, the flux calibration in this band was done using the magnitude of the relatively bright star #14, given by deconvolution (Table 2), whereas the V band results are based on the magnitude of R 84. A remarkable feature of the color-magnitude diagram is the good agreement between the characteristics of the different groupings obtained by the deconvolution code applied to the crowded central part of the field on the one side and the aperture method applied to isolated stars in the whole SUSI field on the other side. Whereas the sequence with essentially traces the OB association, the redder group with is due to LMC field stars. This field star grouping may be a mix of several populations, as Walker (1995 ) has shown in the case of the LMC cluster NGC 1866.

Fig. 7. Color-magnitude diagram of a total of 212 stars in the full .5 .8 SUSI field. The asterisks represent the deconvolution magnitudes (16 .4 16 .4 crowded field around R 84) and the black dots are the rather isolated stars for which we get aperture photometry.

6.4. Radial velocities

We have measured the radial velocities of 15 known members of the LH 39 association (Schild 1987 ) on the spectra kindly provided by Dr. H. Schild. Since these spectra have a low FWHM resolution of 6.8 Å, they can only provide a rough estimate of the radial velocities. Our measurements span a rather wide range between 215 and 440 km s^-1. The above derived radial velocities of stars #35 and #36 are therefore fully consistent with the mean value of the LH 39 association, whereas the velocity of R 84 lies towards the lower end of our range.

© European Southern Observatory (ESO) 1997

Online publication: April 20, 1998
helpdesk.link@springer.de