7 Discussion

The spectrum of HD108 looks pretty much like that of the extreme O8Iaf star HD151804 (Crowther & Bohannan 1997), but exhibits additional emission lines (C  II, O  II,...) and presents spectacular H I and He I line variations.

Underhill (1994) found similarities between the spectrum of HD108 and the spectra of Ofpe/WN9 stars (Bohannan & Walborn 1989) and those of B[e] stars. Comparing with more recent data (e.g. Nota et al. 1996), we notice some marked differences between the spectrum of HD108 and spectra of Ofpe/WNL stars. For example, the He II and N III absorptions are stronger in the spectrum of HD108. Also, the H$\alpha$ emission of HD108 is sharper and the He I lines can evolve into pure absorptions, not only weaker P Cygni. HD108 was also discussed in connection to the Luminous Blue Variable (LBV) stars. However, until now, HD108 has not presented spectacular spectral type variations. The He II $\lambda$4542 line is only slightly variable and the apparent changes in spectral type discussed in Sect.3.1 rather reflect the variations of the wind emission in He I $\lambda$4471. Finally, unlike most LBVs or LBV candidates, no nebular emission was detected around HD108, even if the star has been suggested in the past to be the exciting star of a faint nebula (see Higgs & Ramana 1968). But, in a more recent work, Lozinskaia (1982) excludes the presence of a nebula around HD108.

Andrillat et al. (1982) presented some characteristics of Oe stars. In their sample of stars, H$\alpha$, He I $\lambda$6678 and H I Paschen lines are often present in emission. These emissions are variable, on time scales of several years. They can completely disappear or even become absorption lines. HD108 presents similar characteristics of varying He I and H I lines, and its near IR spectrum is also similar to those of Oe stars. HD108 does not display the Fe II $\lambda \lambda$7515, 7712 and O I $\lambda$7772 lines seen by Andrillat et al. (1982) in most Oe spectra, but these lines are also missing in the spectra of some confirmed Oe/Be stars, as for instance in X Persei.

Moreover, Divan et al. (1983) showed that Oe and Be stars are brighter and redder when the emissions in the Balmer continuum and in H$\alpha$ are stronger. Unfortunately, there are few photometric studies of HD108. Its V magnitude was measured by Plaskett & Pearce (1931), Hiltner & Johnson (1956), Blanco et al. (1968) and Leitherer & Wolf (1984) who all found V=7.40 and by Bouigue et al. (1961), who found V=7.35. The B-V color index is about 0.18 for all articles, except the last one (B-V = 0.14). However, HD108 was sometimes attributed a variable character: in the New Catalogue of Possible Variable Stars (NVS), its V magnitude varies from 7.35 to 7.48. Chilardi et al. (1948) reported magnitude variations up to 0.4 mag, and more recently, Barannikov (1999) showed that there was an abrupt change in the light curve in 1996 and 1997, when He I $\lambda$4471 went into absorption, followed quite soon by H$\gamma$. At that time, the star became suddenly fainter and bluer (in $\Delta(V-R)$). In addition, the Hipparcos satellite has measured a mean $H_{\rm p}$ magnitude of $7.412\pm0.012$ mag with a maximum of 7.39 and a minimum of 7.43 (Hipparcos main catalogue, ESA 1997). The small amplitude of the light curve argues against HD108 being a dormant LBV, but it could be compatible with an Oe/Be behaviour.

HD108 seems thus to share several characteristics of Oe or Be-type stars. However, the spectrum of HD108 displays rather sharp emissions without any indication of a double-peaked morphology that could be attributed to an equatorial disc or flattened wind. Therefore, if HD108 has indeed an equatorial disc, it should be seen under a rather low inclination. In fact, spectropolarimetric observations (Harries 2000) do not exclude the possibility of a disc, provided it would be seen face-on.

Moreover, as mentioned in the introduction, several conflicting determinations of the mass-loss rate of HD108 can be found in the literature. From the unsaturated resonance lines seen in the IUE spectrum of HD108, Hutchings & von Rudloff (1980) and Howarth & Prinja (1989) inferred rather low values of $2\times 10^{-7}$ and $3\times 10^{-6}~M_{\odot}$yr^-1 respectively. Using the H$\alpha$ line flux, Peppel (1984) and Leitherer (1988) obtained $\dot M \sim 10^{-5}~M_{\odot}$yr^-1. Finally, the largest mass-loss rates ( $5\times 10^{-5}~M_{\odot}$yr^-1 comparable to those of Wolf-Rayet stars) were derived from the infrared excess (Ferrari-Toniolo et al. 1981). The various values listed here refer to observations obtained at the same epoch (between 1978 and 1982), thus ruling out any long-term trend as the origin of the discrepancies. These different determinations of $\dot M$ could provide further support for a scenario based on a disc seen face-on. In fact, the UV resonance lines are sensitive to the absorption along the line of sight, whereas the IR continuum probes the emission from the extended disc.

As already quoted in the first section, HD108 has been sometimes classified as a runaway: Bekenstein & Bowers (1974) found a peculiar velocity $v_{\rm p}> 98$ kms^-1 while Underhill (1994) gives $v_{\rm p} = 43$ kms^-1. The Hipparcos satellite has measured the proper motion of HD108 and found $\mu_{\alpha\,\cos(\delta)}= (-5.39\pm 0.66)\times 10^{-3}$arcsecyr^-1 and $\mu_{\delta}= (-0.93\pm 0.54)\times 10^{-3}$arcsecyr^-1. Using the method described by Moffat et al. (1998 and erratum in 1999), we calculate:
$(\mu_{\rm l})_{\rm pec}= (-3.14\pm1.01)\times 10^{-3}$arcsecyr^-1
$(\mu_{\rm b})_{\rm pec}= (0.47\pm0.56)\times 10^{-3}$arcsecyr^-1
yielding $(v_{\rm t})_{\rm pec} = 37.8 \pm 16.5$ kms^-1, assuming a distance from the Sun equal to 2.51 kpc (Gies 1987) and a 30% uncertainty in the distance.

Moffat et al. (1998) define a runaway as a star with $(v_{\rm t})_{\rm pec} > 42 + \sigma_{(v_{\rm t})_{\rm pec}}$. This criterion does not allow us to attribute a definite runaway status to HD108.