8 Conclusions

Using a 30 year long campaign of spectroscopic observations, we have analysed the strange behaviour of HD108. The analysis of our RV time series failed to reveal a significant period and our data rule out the periods claimed in previous studies (e.g. Hutchings 1975; Barannikov 1999). However, HD108 presents some outstanding long-term variations: He I and H I lines changed recently from P Cygni profiles to apparently pure absorption lines. A similar situation was observed about 56 years ago. The line profile variations are thus very probably recurrent. We tried to remove the underlying photospheric absorption to better understand some of the variations. The restored Balmer and He I lines consist of rather narrow emission lines overlying a broad (photospheric) absorption line and reveal no evidence for a P Cygni type absorption. The other emission lines either display lower amplitude variations (N III $\lambda \lambda$4634-41, He II $\lambda$4686,...) or are only marginally variable (e.g. the O II emissions, the unidentified $\lambda \lambda$4486, 4504 emissions,...). These various emissions have thus most probably different origins. Several different hypotheses on the wind structure and the nature of HD108 can be considered:

1.

The emission lines in the spectrum of HD108 could be formed in jets as suggested by Underhill (1994). The long-term variations could be due to a strong occultation of the jets by the stellar body. Such an occultation might result from a precession of the jets. In our data, the absorption continues to strengthen after the emission apparently vanished (e.g. He I $\lambda$4713) whereas the EWs should reach a limiting value during the eclipse. One possibility could be that in 2000 we are still in a pre-eclipse phase. However several details remain to be addressed. Where do the rather constant emissions (e.g. He II $\lambda$4686) originate? What triggers the precession of the jets? Moreover, the narrow emission lines indicate that these jets must be seen almost perpendicular to the line of sight and it seems difficult to imagine a geometric configuration that could explain the secular variations without exhibiting a simultaneous modulation of the emission on the (much shorter) time scale of the stellar rotation period;

2.

The star could be surrounded by a disc which undergoes density oscillations (see e.g. the models by Okazaki 1991). The emission lines would be attenuated when these oscillations are eclipsed by the star, provided that the inclination is sufficient. However, the narrow emission lines without any evidence of a double-peaked profile and the spectropolarimetric observations of Harries (2000) suggest that such a disc should be seen face-on. Therefore it seems difficult to explain the observed variations with this model;

3.

Alternatively, the wind of HD108 could be confined by a strong enough magnetic field into a high density cooling disc in the plane of the magnetic equator (see Babel & Montmerle 1997). The emission lines would be formed in this cooling disc. If, for some reason, the strength of the magnetic field changes as a function of time, we expect the density in the disc and hence the emission strength to vary. Again, we would have to explain why there is a dichotomy in the behaviour of the emission lines and what would be the mechanism causing the modulation of the magnetic field strength;

4.

HD108 might be a binary system harboring a compact object which is revolving in a highly eccentric orbit around an Oe star surrounded by a disc seen face-on. Near periastron passage, the X-rays emitted by the accreting compact object (neutron star or black hole) could drastically alter the ionization in the circumstellar enveloppe. The X-rays could completely ionize hydrogen and He  I, and the emission in those lines would thus disappear. As long as the compact companion is far enough from the Oe star, the He II $\lambda$4686 formation zone should not be affected too much. If the compact object follows a highly eccentric orbit, the spectral lines should not be affected during most part of the orbit. As some older data are missing (especially from 1955 to 1968), we cannot check the morphology of the lines at this epoch. On the other hand, the enhanced interaction around periastron should generate an enhanced X-ray emission. In this context, it is worth mentioning that HD108 was detected as a fairly bright source in the ROSAT All Sky Survey (RASS). In fact, Rutledge et al. (2000) quote a ROSAT-PSPC count rate of $0.0614 \pm 0.0124$ctss^-1 for HD108. Assuming that the X-ray emission is produced in an optically thin thermal plasma with kT = 0.5keV and accounting for an interstellar neutral hydrogen column density of $3.4\times 10^{21}$cm^-2 (Diplas & Savage 1994), we derive an unabsorbed flux of $2.3\times 10^{-12}$ergcm^-2s^-1 in the energy range 0.1-2.0keV. Again assuming a distance of 2.51kpc (Gies 1987), we derive an X-ray luminosity of $L_{\rm x} = 1.7\times 10^{33}$ergs^-1 corresponding roughly to $\log{L_{\rm x}/L_{\rm bol}} \sim -6.0$. The ratio between the X-ray and bolometric luminosities in 1990 (i.e. at the time of the RASS) was therefore about six times larger than expected from the "canonical'' relation proposed by Berghöfer et al. (1997). Observations of HD108 with the XMM-Newton satellite might allow to confirm the compact companion hypothesis since they will be obtained when the stars would be close to periastron passage. If this compact companion model applies, then the system should certainly have undergone a supernova. Since the runaway status is not established, another confirmation of this scenario could come from the detection of a supernova remnant (SNR) in the vicinity of the star. No such SNR has been detected so far. In the ambient circumstellar medium ionized by the O-star, the SNR signature would be very difficult to see. Another possibility for a null detection could be that this remnant has already evaporated. Maeder & Meynet (1994) predict lifetimes between $2.6\times 10^6$ and $4.4 \times 10^6$yrs for stars of spectral type O3 to O6, whereas a typical O7.5 star stays $4.6\times 10^6$yrs on the main sequence. Since a SNR takes typically $5\times 10^4$yrs to evaporate, the compact companion hypothesis cannot be ruled out on the grounds of the lack of a detected SNR.

As the behaviour of HD108 seems recurrent, and could reproduce after about 56 years, this star certainly deserves a long-term spectroscopic monitoring. The H I and He I lines could either reach a steady state or continue their evolution towards stronger absorptions, before they gradually recover a P Cygni profile. Photometric observations of this star should also be useful to assess the amplitude of the possible magnitude variations of HD108, which might be in correlation with its line profile variability. Moreover, multiwavelength studies should be undertaken, in order to confirm possible variations of the mass-loss rate or the compact companion hypothesis. Only then, will we finally be able to answer the question about the real nature of HD108.

Acknowledgements

We would like to thank Dr. J. Manfroid for his help in collecting additional spectra in 2000 and Drs. E. Gosset and T. J. Harries for discussion. We thank the referee Dr. O. Stahl for a careful reading of the manuscript and for his valuable suggestions. JMV and GR would like to thank the staff of the Observatoire de Haute-Provence for their technical support during the various observing runs. We are greatly indebted to the Fonds National de la Recherche Scientifique (Belgium) for multiple assistance including the financial support for the rent of the OHP telescope in 1999 and 2000 through contract 1.5.051.00 "Crédit aux Chercheurs'' FNRS. The travels to OHP for the observing runs were supported by the Ministère de l'Enseignement Supérieur et de la Recherche de la Communauté Française. This research is also supported in part by contract P4/05 "Pôle d'Attraction Interuniversitaire'' (SSTC-Belgium) and through the PRODEX XMM-OM and Integral Projects. The SIMBAD database has been consulted for the bibliography.