Astron. Astrophys. 363, 995-1004 (2000)

4. Discussion and conclusions

For the first time an extended data set of high-quality spectra with high resolution in wavelength and time was recorded for HD199478. The spectra are analysed in terms of line-profile variability using contemporary techniques of time-series analysis, such as Temporal Variance Spectrum and the 2d-Discrete Fourier Transform. We found that the profile of the star consists of a highly variable emission core superimposed on almost constant emission wings. The wings are weak (about 4 to 5% above the continuum) and spread to velocities much higher than the terminal velocity of the wind. The presence of such wings appears to be a common characteristic of the profiles of most BA-type supergiants (Kaufer et al. 1996) and is likely due to electron-scattering in the deep atmospheric layers (Hubeny & Leitherer 1989).

Fourier analysis of the two longest time series shows that the profile is periodically variable. The periodic component consists of variations in velocity and intensity of blue- and red-shifted emission peaks of the line, which result in drastic alterations in the shape of the profile from almost symmetric and unshifted emission, with respect to the stellar rest frame, through blue- or red-shifted asymmetric emission to double-peaked emission or a reverse P Cygni-type profile. Significant variations in total emissivity (i.e. EW) of the line are also established but these variations do not seem to be obviously linked to changes in the shape of the profile. The source of variability is not clear at present. On the one hand, the TVS of the line clearly indicates that the variations are linked to processes in the wind: the velocity range over which significant variations occur, V=430 km s^-1, is larger than 2sini=90 km s^-1. On the other hand, the detection of continuous absorption lpv together with the result that the emissivity tends to anticorrelate with variations in the strength (i.e. EW) of the 6582 photospheric line suggest that the variability might be assigned (at least partially) to changes in the stellar photosphere. Summarising, we conclude that the variability of HD199478 is more likely a result of an interplay between variable wind emission superimposed on variable photospheric absorption.

Absorption line-profile variability consisting of variations in velocity of HI, HeI and metal lines (such as MgII, SiII and FeII) of HD199478 was first established by Denizman & Hack (1988) through a study of 12 photographic spectra obtained in 1970 and 1986. Our analysis confirmed the presence of continuous radial-velocity variability (typical dispersion of =5 km s^-1) in photospheric lines and furthermore revealed the existence of significant line-strength (i.e. EW) variations (up to 13% of the mean) for these lines too. The simultaneous appearance of radial-velocity and line-strength variations implies that the variability observed is likely connected to changes in velocity and temperature structures of the stellar photosphere. Unfortunately, the present data do not enable us to perform a detailed study of the phenomenon and to specify its main properties. The relationship (if any) between photospheric line-profile variability and wind variability (traced by ), is also unknown. Long-term, time-resolved, large spectral window observations are needed to adequately resolve the problem of photospheric line-profile variability of HD199478.

Although our data are insufficient to provide deep insight into the nature of the variability of HD199478, some knowledge about the properties of the wind can still be obtained. For example, close inspection of differently shaped profiles shows that the envelope of this late-type B supergiant is likely axially-symmetric and disturbed. Indeed, it is obvious that neither blue-shifted emission nor red-shifted absorption can originate from density variations in a spherically symmetric uniformly-distributed stellar wind. On the other hand, double-peaked emission with cyclic V/R variations can be readily interpreted in terms of axial symmetry, e.g. perturbed stellar disks, similar to those in Be stars (Okazaki 1996). However, the appearance of blue-/red-shifted and unshifted single-peaked emission as well as a reverse P Cygni-type profile implies that the phenomenon detected in HD199478 does not appear to be identical to that observed in Be stars. The time-scales of the two are also quite different - about 4 to 5 weeks for HD199478 and of the order of years and decades for Be stars, making it less likely they are caused by the same mechanism. Thus we conclude that the wind of HD199478 is more likely axially-symmetric and perturbed, and that the source of wind variability is different from that operating in Be-stars winds. It is worth noting in addition that the detection of blue-shifted absorption components in the UV wavelengths (Bates, Halliwell, Brown-Kerr 1986) also indicates the presence of large-scale structures in the wind of this late-B type supergiant. It is therefore interesting to ascertain if the structures responsible for variations in are in some way related to those producing DACs in UV lines, as has been found previously for O-stars winds (Kaper et al. 1997).

Some knowledge of the origin of variability could be obtained by analysing the time-scales of the phenomenon observed. For example, the periods derived by Fourier analysis are found to be a factor of twenty longer than the radial flow time of the wind, ((HD199478) = 1.68 days). This result suggests that the variations are not intrinsic to the wind. From the fact that similar patterns of line-profile variability were recorded twice over observations separated by about 5 months (about 90), one can furthermore suggest that the variations are not transient feature of the wind, but persist for many flow times and must be therefore maintained by photospheric processes. There is some evidence to suggest in addition that the wind of HD199478 might be rotationally modulated . On the one hand, the periods of variability determined by Fourier analysis are a factor of 3 to 5 longer than the radial fundamental pulsation period =7.8 days. (The later was computed for a pulsation constant log Q = -1.4 used by Kaufer et al. (1996) for BA-type supergiants.) On the other hand, these periods fall just between the rotational periods as estimated from the break-up velocity =167.6 km s^-1 and vsini, i.e. between 25 and 88 days. These findings imply that rotational modulation could be a possible source of variation in the lower region of the wind, where the -line forms.

Absorption line-profile variability consisting of radial-velocity and EW variations as well as profiles with V/R variations have been observed by Kaufer et al. (1996, 1997) in a sample of 3 late-type B and 3 early-type A supergiants. The former was interpreted as due to pulsation (both radial and non-radial) in the stellar photospheres while the latter was attributed to rotation. A comparison of our results with those published by Kaufer et al. shows that the variability of HD199478 is in many aspects similar to that established in BA-type supergiants, suggesting that the same mechanism is likely responsible for the phenomena observed.

Recent theoretical computations (Pamyatnykh 1998) have predicted a significant extension of the high-order g-mode instability, which causes the variability of Slowly Pulsating B-stars (SPBs) and stars, to higher luminosity. This prediction was supported by Hipparcos observations, that revealed 72 newly discovered SPBs and 4 stars, whose position in the HR diagram is fully consistent with the theoretically determined instability domains (Waelkens et al. 1998). In addition, these observations revealed 32 new supergiants with Cyg-type variations, which inhabit a region extending from the instability strip for g-mode oscillations towards variable B-type supergiants and the instability strip for strange-mode oscillations (Kiriakidis et al. 1993) predicted for more massive stars (e.g. LBVs). These results strongly suggest that all early type stars are likely pulsationally unstable. Unfortunately, none of the stars studied by Kaufer et al. nor that studied by us has been ever observed systematically in terms of photometric variation. Thus although they all are known to be photometrically variable neither typical timescales nor behaviour pattern of variability are known at present. Simultaneous photometric and spectral observations are needed to convincingly prove the pulsation origin of the photospheric variability of these stars.

The data presented in this study are obviously insufficient and do not enable the nature and the origin of variability of HD199478 to be investigated completely. The consideration given above concerning the structure and geometry of the wind based on the profiles does not take into account the impact of the underlying photospheric profile. A detailed modelling of the presented profiles, that takes into account the variability of both the stellar photosphere and the wind, is needed to clarify the picture of variability of HD199478. We are beginning a large ground-based observational program (including spectral and photometric observations) designed to determine the properties of the photospheric line-profile variability and to ascertain whether this variability is related to wind variability (traced by ). The performance of simultaneous photometric observations is very important since it may give information on the nature of the processes that governs this variability.

© European Southern Observatory (ESO) 2000

Online publication: December 5, 2000
helpdesk@link.springer.de