1 Introduction

HD108 has stimulated a lot of controversy over the past 75 years. One of the first detailed studies of its spectrum was presented by Andrillat et al. (1973), who also reviewed the line profile changes (from P Cygni to pure emission/absorption lines). Some years later, using the radial velocities of the He II $\lambda \lambda$4200, 4542 absorption lines, Hutchings (1975) derived an orbital period of 4.612 days, but this result was not confirmed by Vreux & Conti (1979). Vreux & Conti (1979) suggested that the radial velocity variations were rather due to wind variability and although they pointed out the possibility of a 1.02 day period, they concluded that the variability of HD108 was most probably not periodic. On the basis of Hutchings's work, Bekenstein (1976) suggested that HD108 had undergone a type II supernova without breakup of the binary. Combining radial velocity data from earlier studies with their own measurements, Aslanov & Barannikov (1989) argued in favor of a 5.7937 days period, again attributed to binarity. But the debate about the real nature of HD108 was not settled. More recently, Underhill (1994) found no periodic variation of the radial velocities, thus rejecting the binary hypothesis. She suggested instead that HD108 was a single star surrounded by a disc and displaying jets that were almost perpendicular to the line of sight. According to Underhill (1994), this particular geometry accounts for the peculiar line profile morphology and could explain the spectral variations seen in HD108. Using V-band polarimetric observations, Fox & Hines (1998) did not find any evidence of a binary nature of HD108, but their results did not allow them to choose between large-scale jets or small-scale blobs as the structures triggering the variability. On the contrary, Barannikov (1999) combining his own recent radial velocity measurements with all older ones, claimed the existence of a longer period of 1627.6 days and concluded again that the star is indeed a binary.

Other properties of HD108 are also intriguing: its UV spectrum suggests a low mass-loss rate (Hutchings & van Heteren 1981), incompatible with the one derived from its optical and IR (Ferrari-Toniolo et al. 1981) spectrum. Moreover, some authors marked HD108 as a runaway star, though they did not agree on the exact value of its peculiar velocity.

In this paper, we will analyse CCD spectra collected over fifteen years. The observations are presented in Sect. 2, and we investigate the main spectral characteristics of HD108 in Sect. 3. In Sect. 4, we will discuss the short- and long-term variations of the radial velocities and search for a periodic behaviour of HD108. In the next sections, we analyse the variability of the equivalent widths and the line profiles. Finally, we examine the photometric data and the proper motion of HD108 before we conclude in Sect. 8.