The length and coverage of our datasets, combined with known cyclical wind effects limit the frequency range in which we can detect pulsation periods from about 1.5 to 10 c d-1. Of course we cannot rule out the presence of other modes in these stars. In fact it is likely that more modes will be found with better datasets like for example in the case of the O9.5V star Oph (Kambe et al. 1993, Reid et al. 1993, Kambe et al. 1997, Jankov et al. 1998) where increasingly higher-quality spectra and denser coverage revealed a larger number of modes, up to = 18.
The presence of NRP in Per was already suspected by Gies & Bolton (1986) on the basis of radial velocity variations, which they assumed to be due to NRP, but no significant period or mode could be identified from their sample of 38 photographic spectra. Significant profile variability in many lines (except in 4713 Å!) was also found by Fullerton (1990), but no mode or period could be established.
A preliminary analysis of the data of the present paper on Cep was given by Henrichs (1991), who found the = 5 NRP mode with a 6.5 h period, and mentioned the possible presence of the = 3 mode at 12.5 h. Walker (1991) reported a likely = 5 or 6 mode (no period) for this star, but without further details we cannot compare his results with ours.
We also find small EW variations in this star, and some frequencies do coincide approximately with the pulsation frequencies. They could in principle be to be due to temperature effects. Higher-quality data are needed to confirm this.
Marchenko et al. (1998) report the presence of a strong 0.63 day period in the HIPPARCOS photometry of Cep (but none in Per) over the 3.5 year lifetime of the satellite, which partly covered our campaign. This period is not compatible with any of the NRP or wind periods, and its origin is unclear.
We now return to the question set out at the beginning of this research whether non-radial pulsation can be the origin of the cyclical wind variability. If the mode we found is the only mode in Per, it appears that the pattern speed of the waves running around the star superposed on the rotation (see Sect. 3.1) is too high to be compatible with the observed wind periods (1, 2, 4 days). This is probably true for most short-period single mode pulsations, since for most O stars the rotation rate is much lower. An exception is the O4 star Pup, for which Reid and Howarth (1996) report evidence for a direct connection between wind features and an NRP mode. They found both in H up to 0.3 and in HeII 5411 the same period of 8.54h. However, they consider the 19.6h period found in IUE observations incompatible with this NRP period.
The presence of multimodes, however, has possibly interesting consequences for the origin of cyclical wind variability (see also Rivinius et al. (1998) for the Be star µ Cen). Consider for example a case with two different prograde sectoral modes, traveling around the star with different frequencies. A given crest of the faster wave will at a certain moment overtake a crest of the slower wave, which means an enhancement of the total amplitude at a certain longitude. The next enhancement will be when a different crest will overtake, but this will be at a different longitude. This will give rise to cyclical surface amplitude enhancements which may cause wind perturbations that are related to the relative traveling speeds and the m values of the NRP waves. The simultaneous presence of more than two modes will increase the complexity of this beating effect. Observations show, however, that the periods of cyclical wind variability of O (and B) stars scale with the rotation period of the stars (Prinja 1988, Henrichs et al. 1988), and it is difficult to understand how this could be related to the above described effect of beating NRP modes.
We therefore think that the best candidate for the cause of the cyclical wind variability still remains the presence of weak magnetic fields on the surface, corotating with the star. A proof has to wait for a systematic deep survey of these fields. A preliminary upper limit of 70 G on the longitudinal component of the magnetic field strength of Per was presented by Henrichs et al. (1998a).
In conclusion, if our interpretation is correct, the number of confirmed O stars with NRP is now about 6 (see Fullerton et al. 1996, including Pup). In the light of asteroseimological applications, we note that 5 of these are runaways. Although the statistics are limited, OB runaway stars tend to rotate rapidly and to have an enhanced surface He abundance (Blaauw 1993). These factors might play a role in exciting the NRP. The exception is the pulsator 10 Lac, which is not classified as a runaway star, but this star is associated with a bow shock detected on IRAS 60 micron maps (van Buren et al. 1995), indicating a post binary mass-transfer history, and hence a different internal structure is likely, perhaps favouring the existence of NRP modes.
Because of the known ubiquity of line-profile variables among O stars it one can expect that with a concentrated observational effort with sufficient S/N, coverage and time resolution, more pulsation-mode identifications in O stars are likely to follow. Higher spectral resolution is needed to find possible multimodes in these stars. The possible consequence of such multimodes for cyclical wind behavior needs to be investigated.
© European Southern Observatory (ESO) 1999
Online publication: April 12, 1999