The most massive stars are the H-burning O stars and their descendants, the He-burning Wolf-Rayet stars. It has been known for a long time that a significant fraction of these stars lies well outside the boundaries defined by their likely birth places in open clusters and OB associations. In view of the potential confusion in the Galactic plane caused by inaccuracies in distances especially of single stars, some of these stars manifest themselves most clearly by their large separations from the Galactic plane, or their rapid motion. Over 35 years ago, Blaauw (1961) recognized 19 O-type "runaways", i.e. O stars having space motions greater than 40 km s-1, some of whose velocity vectors point back to their origin in recognized clusters or associations. This study was significantly extended later by Gies & Bolton (1986) and Gies (1987). The observed runaway nature of some WR stars was first discussed by Moffat & Isserstedt (1980).
Two plausible theories for the origin of runaways remain, after elimination of the possibility of confusion with hot, low-mass subdwarfs (Gies & Bolton 1986). The first theory is the binary-supernova (SN) scenario (Blaauw 1961), favoured by Stone (1991). The second theory is the cluster ejection scenario (Poveda et al. 1967), in which a star is ejected via dynamical interaction between stars in a young, compact cluster. The second one is favoured by Gies & Bolton (1986) and Leonard & Duncan (1990). In fact, both may be operating.
in which c stands for "compact" companion, a neutron star or black hole, left after the supernova explosion of its progenitor. In this scenario, it is the more massive star that evolves faster at first. Wind mass-loss (possibly assisted by Roche-lobe overflow in the closest massive binaries) makes O stars evolve into lower-mass WR stars, which also evolve along a sequence from cool to hot subtypes within the WN and then WC sequences. This occurs for each component in turn, i.e. at and . At the end of the WR phase, it is assumed that the star explodes as a SN, at both and . If the first SN explosion is symmetric, the binary system will remain bound, since the less massive star explodes, leading in principal to the class of massive X-ray binaries (MXRB). If the first SN is asymmetric, the binary system may disrupt, depending on the magnitude and direction of the extra kick velocity (De Cuyper 1982). The latter disruptive case may be much more common, explaining the origin of some of the high-velocity pulsars and the low frequency of MXRBs among O-type stars in general and high-velocity O stars in particular (van Oijen 1989; see also van den Heuvel & van Paradijs 1997 for the importance of kick velocities; however see also De Cuyper 1982). In either case, the SN explosion is very short compared to the orbital period, so that the stars receive a recoil velocity, i.e. become runaway, with velocities reaching up to 200 km s-1 for the closest, most massive pre-SN binaries. In the case of the second SN, it is the more massive star that explodes, so that the system, if it has not already separated after the first SN, will normally become unbound, producing two high-velocity, single pulsars. In rare cases, the binary can survive this second SN, producing a binary pulsar (De Cuyper 1985).
Our confidence in the binary-SN scenario was considerably enhanced very recently by the discovery of a WR star in the well-known, but highly reddened 4.8-hour period MXRB Cyg X-3 (van Kerkwijk et al. 1992; van Kerkwijk 1993). Previously to this, none of the suspected WR + c systems (Moffat 1982) showed any significant accretion-type X-rays, making their existence questionable. The presence of one clear WR + c system (and possibly one other, HD 197406: Marchenko et al. 1996) is, within the small number statistics, compatible with the relative lifetimes ( 10%) of mainly He-burning WR stars versus the 24 known MXRB main-sequence stars. Possibly, the majority of the runaway WR stars are also single, as for O runaways, having been disrupted in most cases by an asymmetric SN.
On the other hand, the existence of systems (1), (2), (3) and (5) is well established: O + O and WR + O binaries have been known for over half a century, with binary frequencies close to 40% (Garmany, Conti & Massey 1980; Moffat et al. 1986); O + c systems were discovered in the late 1960's from X-ray satellites; runaway neutron stars were discovered not long after the first pulsar was discovered in 1968. Even the association of MXRBs with massive runaways seems now assured (van Oijen 1989), thanks to better statistics, despite some previous doubts (e.g. Gies & Bolton 1986).
The cluster-ejection scenario makes predictions that are somewhat different compared to the binary-SN hypothesis, although both can lead to massive runaways per se. In particular, while the cluster-ejection scenario slightly favors the observed low, but non-zero O + O binary frequency among runaways and accounts better for runaway pairs like AE Aur/µ Col, both scenarios can lead to the large range of runaway space velocities observed, reaching up to some 200 km s-1. However, the binary SN scenario appears to do better in accounting for observed space frequencies, low-mass cut-off, velocity-mass correlation, kinematical ages and presence of MXRBs among the OB runaways (Stone 1991). However, some of Stone's results may be invalidated by his restrictive subsample of main sequence massive close binaries, all having the same initial orbital period and mass ratio. Nevertheless, Blaauw (1993) also notes that the SN scenario allows better for the high frequency of fast rotators and abundance anomalies among O runaways, compared to low-velocity O stars. Indeed, the mere existence of MXRBs proves that SN do take place in massive binaries, which must lead to runaway speeds even if the explosion is symmetric, although most SN are highly asymmetric and lead to disruption and hence single runaways. Nevertheless, a certain fraction of O and WR runaways may have been ejected from compact clusters.
The bulk of these previous studies is based primarily on radial velocities (RV). Proper motions were known, but only for the brighter stars, and with a precision inadequate for viable studies. For WR stars, RVs are virtually useless to determine systemic radial motion; even the more extreme O stars (those with emission lines in their visible spectra) tend to have negatively biased RVs. For these reasons, three independent groups were granted Hipparcos time in 1982 to obtain systematically more precise proper motions of the bulk of the Galactic WR stars down to a feasible magnitude limit, as well as a pre-selected (to keep the numbers reasonable) group of Galactic O stars. This paper presents the merged results of these programs.
The aims are to use these proper motions to:
The tracing back of the origin of motion to parental clusters/associations will be left for a future study, when the proper motions of the clusters/associations themselves have been properly re-assessed using Hipparcos (cf. de Bruijne et al. 1997; Hoogerwerf et al. 1997; de Zeeuw et al. 1997). This must be combined with evolutionary models (e.g. cf. van Rensbergen, Vanbeveren & de Loore 1996).
© European Southern Observatory (ESO) 1998
Online publication: March 3, 1998