Adopting the effective temperature calibration of Chlebowski & Garmany (1991), the spectral types we have derived yield $T_{\rm eff} = 35\,100 \pm 1000$ K and $T_{\rm eff} = 30\,500 \pm 400$ K for the primary and the secondary respectively. HD149404 has an apparent V magnitude of 5.47 (Drilling 1991). Assuming that the star is a member of the Ara OB1a association (DM = 10.7, Humphreys 1978) and adopting a colour excess E(B-V) = 0.68 (Shull & Van Steenberg 1985) and a spectroscopic luminosity ratio of $0.90 \pm 0.16$ (see above), we derive $\log{(L^{\rm prim}_{\rm bol}/L_{\odot})} = 5.90 \pm 0.08$ and $\log{(L^{\rm sec}_{\rm bol}/L_{\odot})} = 5.78 \pm 0.08$ for the primary and secondary respectively. We caution that the error bars on the luminosities correspond only to the estimated error on the luminosity ratio and an uncertainty of half a spectral class for the bolometric correction. Our error bars account by no means for the (probably large) uncertainties on the distance of HD149404.

The resulting locations of the components of HD149404 in a H-R diagram are shown in Fig. 7 together with the theoretical evolutionary tracks of Schaller et al. (1992). We notice that the secondary star (i.e. the least massive star) seems to be the most evolved component though it lies on an evolutionary track of lower initial mass than the primary. This was already found by Penny et al. (1996), though these authors adopted different spectral types and hence derived different positions for the stars in the H-R diagram. As pointed out in Sect.6, the strong NIII absorptions and the NII emission lines seen in the secondary's spectrum could indicate that nitrogen is slightly enhanced in the atmosphere of this star, which would be a further indication that this star must be an evolved object. A possibility to account for the evolutionary status of the secondary would be to suppose that this star was initially the more massive component of the binary and has lost (or transferred) a considerable fraction of its mass through Roche lobe overflow in the past (see also Penny et al. 1996).

$\begin{figure} \par\includegraphics[width=7.7cm,clip]{MS10412f7.eps} \end{figure}$

Figure 7: Hertzsprung-Russell diagram of HD149404. The open triangle stands for the primary star whilst the filled circle yields the position of the secondary star

The temperatures and luminosities derived here correspond to radii of 24.3 and 28.1 $R_{\odot}$ for the primary and the secondary respectively. Therefore, the radius of the secondary star appears larger than we would expect from the "typical" parameters listed by Howarth & Prinja (1989). Howarth et al. (1997) derived projected rotational velocities of $v_{\rm e}\,\sin{i} = 91$ kms^-1 and 100kms^-1 for the O7.5 and the O9.7 component respectively. Assuming that the stars have the same angular rotational velocity (though the rotation needs not necessarily be synchronous with the orbital motion), we can derive the ratio of the stellar radii from the ratio of the projected rotational velocities. The value obtained in this way $R_{\rm prim}$ / $R_{\rm sec} = 0.91$ is in very good agreement with the ratio of the radii derived above (0.86).

At this point, we need to return to the question of a current RLOF. In fact, the radius of the secondary star derived hereabove (28.1 $R_{\odot}$ ) is only slightly smaller than the radius of its Roche lobe (29 $R_{\odot}$ ) for an inclination of $21^{\circ}$ . Therefore a small error on the (poorly constrained) inclination has a large impact on our conclusion about the current configuration of HD149404. Indeed, it could be that the secondary has just finished RLOF and is now settling down on a "normal" evolutionary track corresponding to its current mass. Alternatively, the secondary could still be filling its critical volume and could be suffering an enhanced mass loss through the inner Lagrangian point.

$\begin{figure} \par\includegraphics[width=15cm,clip]{MS10412f8.eps}\end{figure}$

Figure 8: Schematic view of the wind interaction in the HD149404 system projected on the orbital plane. The changing morphology of the H $\alpha$ line at some specific orbital phases is also sketched as a function of the line of sight. The relative dimensions of the orbit and the stars correspond to an inclination of $i = 21^{\circ }$

7.2 Towards a model for the wind interaction

In Sect.6 we have shown that the H $\alpha$ , H $\beta$ and HeII $\lambda$ 4686 emission lines are not formed in the atmospheres of the stars but must be formed somewhere between the stars. Thaller (1998) suggested that the H $\alpha$ emissions were coming from two focused stellar winds flowing towards a colliding wind interaction between the stars. Though the radius of the secondary star derived in Sect.7.1 supports the possibility of an enhanced mass loss from the secondary focused towards the primary, it is not clear how the model suggested by Thaller (1998) could account for the RV behaviour found in Sect.6. We favour an alternative model where the H $\alpha$ emissions arise in the arms of a colliding wind shock region. Figure8 displays a schematic view of such a wind interaction. As pointed out above, the secondary might actually be suffering a stronger mass loss than would be expected for a "normal" O9.7I star. Therefore, we assume that the two stars have roughly equally strong winds. The Coriolis force bends the arms of the shock region around both stars. In this model, the H $\alpha$ emission forms through recombination in the hatched areas where the density is enhanced due to the shock (Fig.8). The arrows indicate the velocity components of the H $\alpha$ emission regions as inferred from our S-wave analysis. The velocity components result from the combination of the orbital motion and the flow velocity of the plasma in the shock region.

In Fig.8, the projections of the line of sight on the plane of the orbit are indicated for orbital phases near conjunction and quadrature. We see that this rather simple model can at least qualitatively account for the changing morphology and the RV behaviour of the H $\alpha$ emission line in the spectrum of HD149404.

We have analyzed ROSAT observations of HD149404 to search for a signature of the wind interaction in the X-ray domain. HD149404 was observed twice with ROSAT. A first exposure was obtained as part of the ROSAT All Sky Survey (Berghöfer et al. 1996). Another 3ksec PSPC exposure was obtained on JD2449056.813 ( $\phi = 0.70$ ). We have retrieved this latter observation (rp201266) from the ARNIE database at Leicester University. The data were reduced using the XSELECT and XSPEC softwares. The background corrected PSPC count rate is $8.0\,10^{-2}$ ctss^-1. The PSPC spectrum is quite hard, as the ratio of the count rates in the ranges 0.5-2 keV and 0.1-0.5 keV is about 25. Despite the limited S/N ratio, we could fit an absorbed Raymond-Smith model (Raymond & Smith 1977) to the PSPC spectrum fixing the neutral hydrogen column density to its interstellar value $N_{\rm H}^{\rm ISM} = 2.5\,10^{21}$ cm^-2 (Shull & Van Steenberg 1985). The best fit yields an X-ray temperature of $0.84 \pm 0.10$ keV and a luminosity in the 0.1-2.0keV energy range of $L_{\rm X} = 1.6\,10^{32}$ ergs^-1corresponding to a ratio $L_{\rm X}/L_{\rm bol}$ of $3.0\,10^{-8}$ , slightly lower than for typical O stars. This result is in agreement with the value found by Berghöfer et al. (1996). Thus the snapshot ROSAT observations reveal no indication of a strong X-ray excess that could be attributed to a colliding wind emission. This is in line with the results of Chlebowski & Garmany (1991) who found no significant X-ray excess in those O-star binaries that are in a semi-detached or a contact configuration.

7.3 Summary and conclusions

We have used an extensive set of high resolution spectra to derive new orbital elements for the early-type binary HD149404. We have shown that the mask cross-correlation technique developed at the Geneva observatory can be successfully applied to disentangle heavily blended lines in massive binaries. Our results allowed us to propose a new spectral classification for the components. We find that the primary is of spectral type O7.5I(f), while the secondary is most likely an ON9.7I supergiant. The secondary seems to be the most evolved component of the system and its current evolutionary status could best be explained if the system has undergone a Roche lobe overflow episode during the past. The secondary could actually still be rather close to filling its critical volume and this could lead to an enhanced mass loss of the secondary.

The behaviour of the emission lines in the spectrum of HD149404 points towards a wind interaction between the two stars. We propose a simple model where some of the optical emission lines arise in a heavily bended shock region. It seems likely that HD149404 is currently in a rather short-lived evolutionary stage where the wind interaction could be strongly affected by an enhanced mass loss of the secondary star. We caution, however, that this rather simple picture of the interaction could be seriously complicated if the secondary actually filled its critical volume. In this context, we emphasize that some of the crucial parameters of HD149404 such as its distance and the orbital inclination remain poorly constrained. All future observations that could help to determine these parameters would be extremely useful to fully understand this system.

7 Discussion

7.1 The evolutionary status

7.2 Towards a model for the wind interaction

7.3 Summary and conclusions