Massive hot stars are relevant objects in many areas of astrophysics. They are the main sources of photoionization of their surrounding interstellar medium, and they contribute to the chemical and dynamical evolution of the host galaxy through their stellar winds and supernova explosions. They are also the precursors of Wolf-Rayet stars and LBVs and constitute excellent probes of evolutionary models. For these reasons, advances in our understanding of these objects have an impact on other fields of astrophysics.
In spite of all this, few attempts have been made to analyse massive O stars quantitatively. The most comprehensive study up to now has been that of Herrero et al. (1992, hereafter Paper I), who analysed 24 stars from spectral types B0 to O5. Detailed quantitative analyses of massive stars of early spectral type have been even more scarce. Conti & Frost (1977) first made a systematic analysis of the earliest spectral types, comparing with theoretical predictions by Auer & Mihalas (1972), but only for gravities of = 4.0. Later, Kudritzki (1980) studied HD 93250, an O3 V star, using models similar to those of Auer & Mihalas (1972), and Kudritzki et al. (1983) studied the spectrum of Pup and showed that the star has a low gravity (of the order of = 3.5) and a high helium abundance ( = N(He)/(N(H)+N(He))= 0.14). Kudritzki & Hummer (1990) list 15 stars earlier than spectral type O6 (only three classified as supergiants), with parameters determined using the same methods. Puls et al. (1996) list 22 stars earlier than O6 in the Milky Way and the Magellanic Clouds, still taking advantage of the optical analysis, but already incorporating the effects of sphericity and mass-loss. Pauldrach et al. (1993) analysed Melnick 42 and Pup, using only the UV spectrum, and Taresch et al. (1997) made a very detailed analysis of HD 93 129A using the UV and FUV spectrum. Finally, de Koter (1998) has used the UV O V line at 1371 Å to determine the temperature of very hot stars in R136a.
All these studies have revealed a number of problems in the analyses of these stars, related to our incomplete understanding of these objects. In Paper I, the so-called helium and mass discrepancies, already present in the previous literature (see Kudritzki et al. 1983; Voels et al. 1989; Groenewegen et al. 1989; Herrero et al. 1990) were shown to be systematic. These refer to the discrepancy in the values of the stellar mass and the photospheric helium abundance obtained from the analysis of the spectrum using state-of-the-art model atmospheres and evolutionary models. The explanation of these discrepancies is still unclear (for recent working directions, see Howarth 1998).
In Paper I we already noted the correlation between the mass discrepancy and the distance of the star to the Eddington limit, indicating that the plane-parallel geometry and the hydrostatic equilibrium assumption could be the reason for the low stellar masses derived. However, the use of wind techniques in the same work already indicated that the discrepancy could be reduced, but not solved, by including mass-loss and sphericity effects. This was later confirmed in an analysis of HDE 226 868 (Herrero et al. 1995), where the authors used Unified Model atmospheres to determine the mass of this star, the optical counterpart of Cygnus X-1, by combining the spectroscopic analysis with the orbital data. A similar result has been obtained by Israelian et al. (1999) in an analysis of HD 188 209. The inclusion of mass-loss and sphericity did not seem to have any effect on the helium discrepancy either (Herrero et al. 1995) in spite of the variations that strong winds could introduce in the helium profiles as compared to static, plane-parallel atmospheres (Schaerer & Schmutz 1994). One of the reasons for this weak influence was the fact that the study in Paper I was limited to spectral types of O5 or later, because it was found that above 40 000 K the neutral helium singlet and triplet lines gave different stellar parameters. Herrero (1994) suggested that this was due to the neglect of the so-called line-blocking , the UV background opacity due to metal lines, during the line formation calculations. Also the inclusion of microturbulence in these calculations can reduce this difference (McErlean et al. 1998; Smith & Howarth 1998; Villamariz & Herrero, in prep.).
In this paper we study a few stars of early spectral type in an attempt to cover several objectives. First, we would like to extend our sample from Paper I towards earlier spectral types and thus cover the whole region of interest in the HR diagram with plane-parallel analyses to see their complete behaviour. Although the plane-parallel models will have difficulties in explaining even the optical spectrum of these very hot stars (see Sect. 4), this first step is needed for the subsequent application of more sophisticated models, which will use the experience gained and the parameters obtained as input. An analysis of line-blocking effects is mandatory here, as it has been seen to have an influence for the higher temperatures.
Then, we will repeat the analyses using spherical models with mass-loss, and present the study of some effects that both can influence the determination of stellar parameters and that will help us to gain new insight into the physics of these stars. Having these parameters determined, we can try to establish conclusions with respect to the use of different model atmospheres and techniques for the analysis of very early spectral types. We will compare the results of the spherical models with mass-loss with those from plane-parallel, hydrostatic model atmospheres as well as with the results obtained using the somewhat approximate technique employed by Puls et al. (1996).
In Sect. 2 we present the observations. The effects of line-blocking in plane-parallel models are treated in Sect. 3 and the spectral description and plane-parallel analysis are considered in Sect. 4. Sect. 5 shows the analyses performed with spherical models with mass-loss, while Sect. 6 contains a qualitative study of some effects of interest which explain some difficulties found in the preceding section. Then we present our discussion (Sect. 7) and conclusions (Sect. 8).
© European Southern Observatory (ESO) 2000
Online publication: January 31, 2000