5. Discussion

Despite the selection criteria we use, there are 6 stars of group 3 in Table 1 (click here), and 3 in Table 2 (click here). All these sources have S₁₀₀ larger than S₆₀, or a very large S₆₀/S₂₅ ratio, which is normally characteristic of HII regions and galaxies. A look at the IRAS maps show that these sources are located in regions with a high background at 100 and , and that S₁₀₀ and/or S₆₀ are contaminated. Three sources, LI-LMC 932, 1107, and 528, are also found by Reid et al. (1990). They do not give any flux, and find no or a fainter flux. Using the fluxes determined by Reid et al., these sources would belong to group 1 or 2 (see Table 8 (click here)). We expect that probably a few good AGB candidates have been ruled out of our list for similar reasons, rejected because of a high 100 or/and flux due to contamination. There is however no systematic way to pick up such sources, the only way would be a careful examination of IRAS maps for all of them. The case of planetary nebulae is more marginal. In Table 3 (click here), 9 sources belong to group 3. However, they are normally very cold objects and their IRAS flux ratios are close to the limits of our selection criteria.

Conversely there are several rejected sources in Table 7 (click here) belonging to group 1. Most of them are, however, associated with star clusters, or hot stars, which is not contradictory with belonging to group 1. There are a few very interesting sources, blue luminous variables and Wolf-Rayet stars which deserve more studies. The few sources of group 1 associated with HII regions all have very cold IRAS colors, close to the limits of our selection criteria.

In Fig. 2 (click here), we have plotted the location of optically known stars (Table 1 (click here), Fig. 2 (click here)a), obscured AGB stars (Table 2 (click here), Fig. 2 (click here)b), and planetary nebulae (Table 3 (click here), Fig. 2 (click here)c), in the same kind of diagram as the one presented in Fig. 1 (click here) for galactic sources and foreground stars (Table 6 (click here)). One would expect that optical stars have C₂₁ smaller than -0.3 (), however in Fig. 2 (click here)a 28% of the sources have C₂₁>-0.3. Most stars in Table 1 (click here) are M supergiants. In our Galaxy, it is known that some M supergiants have a "cold'' circumstellar envelope though not optically thick. To model their energy

Figure 2: Location of sources found in Table 1 (click here) (Fig. 2 (click here)a), Table 2 (click here) (Fig. 2 (click here)b), and Table 3 (click here) (Fig. 2 (click here)c) in the [C₂₁, S₁₂] diagram. Full dots correspond to the IRAS fluxes determined by Schwering & Israel, open triangles correspond to IRAS fluxes determined by Reid et al. One can see that the disagreement between both determinations is often much larger than the "standard'' adopted calibration uncertainty of 15% on IRAS fluxes (corresponding to on C₂₁). On average, taking into account the large uncertainty on IRAS fluxes (see Sect. 5), obscured AGB stars are colder than sources with optical counterparts, as expected

distribution, Rowan-Robinson & Harris (1983) had to use a dust temperature at the inner radius of the dust shell of only typically 500 K, far below the dust condensation temperature. This would reflect either a peculiar mass-loss history, or the fact that the dust condensates much further from the star than in M giants. It might be that the same phenomenon occurs in some M supergiants of the LMC.

However, here we work with fluxes close to the detection limit and we should first invoke the uncertainty of these fluxes. Both Schwering & Israel and Reid et al. give a typical calibration uncertainty of 15%, leading to an uncertainty of on C₂₁, though Reid et al. note that this uncertainty can be much larger for some sources. In Table 8 (click here), for sources in common to Schwering & Israel and Reid et al., we present a comparison of the various determinations of IRAS fluxes. Clearly, the disagreement between Schwering & Israel and Reid et al. is often much larger than 15%. In fact, as noted by Reid et al., there is a systematic disagreement for S₁₂ and S₂₅. Reid et al. underestimate S₁₂ and S₂₅ by typically 25% compared to Schwering & Israel, and by 10% compared to the PSC. Reid et al. say that 10% is acceptable as it is inside the admitted 15% uncertainty. This is correct when one considers one source, but in a statistical sample they should not find any systematic deviations. Clearly, IRAS fluxes have not been determined with the same method by Schwering & Israel and by Reid et al. As a consequence, there is also a systematic deviation in the value of C₂₁ which is generally underestimated by Reid et al. compared to Schwering & Israel. The typical uncertainty on C₂₁ that we derive from Table 8 (click here) is , and hence only a few sources in Table 1 (click here) really have a cold C₂₁ colour (Fig. 2 (click here)a). For them, one could also doubt that the IRAS source is actually associated with the optically identified star.

With such uncertainties in IRAS fluxes, the list of obscured AGB star candidates given in Table 4 (click here) probably misses a few objects whose 60 and/or fluxes are contaminated. Conversely, a few sources in Table 4 (click here) might be associated with HII regions. However, we think that this list is the most complete one can make in a systematic way, and is quite reliable as it can be seen in Table 1 (click here) and 2 for identified objects. This list contains 6 times more sources than known obscured AGB stars listed in Table 2 (click here). Therefore further studies should be performed to clearly identify them. We expect that many of them will be identified through the IJK' observations of the DEep Near Infrared Survey of the southern sky (DENIS).

Acknowledgements

We are very grateful to P. Whitelock and B.E Westerlund for their helpful comments. This research has made use of the Simbad database, operated at CDS, Strasbourg, France.

tex2html_wrap1898
Table 1: Optically known M and C stars