3.1. The He ii 4686 line
The profile variability of this line is illustrated in Fig. 1. We notice different kinds of P-Cygni profiles. On several occasions, we can see a double peaked emission with varying intensities of the components. Sometimes both have the same intensity (e.g. JD 2450316.38, 2450638.60,...). On other spectra either the red component (e.g. JD 2449913.54, 2450640.38,...), or the blue component (e.g. JD 2450642.37) is the dominant one. Normal P-Cygni profiles are also observed (e.g. JD 2450318.39, 2450643.60). On the spectra taken during the night JD 2450641, we observe a blue-shifted emission line with weak absorption components on both sides. An even more extreme blueward excursion of the emission component was observed by Underhill (1995a) on 1991 September 24. At that time, the profile consisted of a weak broad emission feature with a very weak absorption component on the longward edge.
We have measured the equivalent widths of the absorption and the emission components of the line. The EWs of the entire profile display some variability, during each night and from night to night. In July 1997, the maximum total EW is observed around JD 2450642.5.
Fig. 2a displays the variability of the equivalent width of the absorption component. This measurement includes the very weak red absorption component that appears when the emission line reaches its most negative radial velocity. One notices that the intensity of the absorption component progressively decreases over five nights before it jumps back to its initial high value (JD 2450643).
An easy means to quantify the line profile variability is through the so-called moments of the line. Let us recall that the nth order moment of a line is defined by (e.g. Castor et al. 1981):
where is the flux of the line and is the corresponding continuum flux. and designate the rest wavelength of the line and the terminal velocity of the wind respectively. In our present analysis, we focus on the first order moment, since it provides a direct measurement of the asymmetry of the line. We adopt a terminal velocity of 2180 km s-1 (Prinja et al. 1990) and we integrate the profile between 4675.8 and 4695.8.
The variations of as a function of time are shown in Fig. 2b. shows no systematic trend except during the blueward excursion of the dominant emission peak when the first order moment becomes negative.
The radial velocities of the dominant peak of the emission component, determined by fitting a gaussian, are displayed in Fig. 2c. We observe variations over a range of some 200 km s-1. Finally, for those spectra on which the emission component appears double-peaked, the radial velocities of both peaks are shown in Fig. 2d. The latter figure indicates that the RV variations of the strongest peak shown in Fig. 2c seemingly result from the alternating intensities of the two peaks.
3.2. Other emission lines
The normalized spectrum of HD 192639 between 4625 and 4660 is illustrated as a function of time in Fig. 3. The relatively intense N iii 4634-41 emission lines do not display the same behaviour as the He ii 4686 line. In particular, the radial velocities of the N iii emission peaks remain nearly constant (see Table 1 below). We notice some intensity variations, that are only slightly correlated to the variations of He ii 4686 (see Fig. 5) and present a much lower relative amplitude than for the latter line.
The formation of the N iii 4634-41 () lines in the atmospheres of O stars has been discussed by Mihalas & Hummer (1973). These authors have shown that in the photosphere of O((f)) and O(f) stars, the level of the ion is overpopulated as a result of dielectronic recombination, leading to photospheric emission of the 4634-41 lines. In the case of HD 192639, the lower amplitude of variation and the different behaviour of the N iii lines compared to the He ii 4686 wind feature probably confirm that the former emission lines arise mainly in the photosphere of the star rather than in the stellar wind.
In order to measure the line profile variability in a quantitative way, we have computed the time variance spectrum (TVS, Fullerton et al. 1996) of our Aurélie observations. For the most variable lines in the spectrum of HD 192639, Table 1 provides the value of the highest peak of over the entire line profile, the velocity interval over which the detected variations are significant at the 99% confidence level, as well as the mean amplitude of the deviation evaluated over . For the N iii 4634-4641 emission lines, the severe blending of the TVS features of these lines and the proximity of the variable C iii 4647-50 absorption lines prevent a reliable evaluation of the mean amplitude. Nevertheless, the TVS parameters given in Table 1 confirm that the N iii 4634-4641 blend displays a lower variability than e.g. the He ii 4686 line.
Table 1. Measured characteristics of the most prominent lines in the blue-violet spectrum of HD 192639. The second column gives the time-averaged radial velocities, whereas the fourth column lists the equivalent widths of the strongest unblended absorption lines. The last three columns provide the results of the TVS analysis for those lines that display significant variability (see text). The amplitude of is expressed as a percentage of the normalized continuum flux. The velocity intervals labelled with a colon are uncertain due to the severe blending of the N iii emission lines.
Brucato (1971) claimed to detect irregular variations in the equivalent width of the N iii emission lines on a time scale of about 20 minutes. Even if the sampling of our time series is not well suited to detect such rapid fluctuations, it appears that the reliability of Brucato's discovery is rather difficult to assess. In fact, Brucato (1971) noticed these rapid variations only during one night, whereas he found only marginal variability during the other nights. Moreover, these results were derived from photographic plates and the author does not provide error estimates for his measurements.
Concerning the two unidentified emission lines at 4486 and 4504, we detect no significant variability and no correlation with the variability of the H line (Fig. 5). We conclude therefore that these lines are probably not produced in the stellar wind, but may arise in the same physical region as the bulk of the N iii emission.
3.3. The absorption lines
All the absorption lines in the observed spectral domain display variations to some extent. Part of the intensity variability may result from uncertainties in the normalization of the spectra. We will therefore focus our attention on the strongest lines that show the most stringent variations, as indicated by the TVS analysis (Table 1), and which are the least sensitive to the uncertainties of the normalization.
Fig. 4 displays the variability of the H absorption line in the spectrum of HD 192639. During the observing run in July 1997, the equivalent width of this line varies over a range of % (1) around the mean value. Larger deviations from the mean equivalent width are observed in August 1996, when the line appears some 18-26% weaker than on the average.
Besides the variations of its intensity, the H line displays also important profile variability (Fig. 4). The core of the line changes from a symmetric nearly gaussian shape to a rather sharp asymmetric minimum, that appears either red-shifted or blue-shifted. Occasionally, we also observe a flat minimum.
We have constructed a dynamical spectrum by assembling quotient spectra obtained after division of the individual data by the mean spectrum. A close inspection of this dynamical spectrum reveals a tight resemblance between the deformation pattern of the H, H, He i 4471 and He ii 4686 lines, in the sense that the stronger is the He ii 4686 emission component, the weaker are the absorption lines.
To quantify this impression, we have used the local pattern cross-correlation technique discussed by Vreux et al. (1992) and by Gosset et al. (1994). Some results obtained using the H line as a reference are shown in Fig. 5. This figure underlines the strong correlation between the mean deformation pattern of the different lines in the spectrum of HD 192639.
In Fig. 3 we notice the strong changes in the visibility of the C iii 4647-50 absorption lines. On JD 2450318.39 and JD 2450643.60, these lines have nearly disappeared, whereas they present their maximum intensity on JD 2449913.54 and JD 2450640.59. The cross-correlation method indicates that the deformation pattern of these lines is correlated to the one of the H line (Fig. 5).
The correlations between the profile variations of the absorption lines and those of the He ii 4686 emission suggest that the bulk of the variability of the absorption line profiles in the spectrum of HD 192639 is produced by a variable stellar wind emission rather than by a genuine photospheric phenomenon.
Our results are in good agreement with the conclusions of previous studies. Underhill (1995a) noticed that the leading members of the He i series are displaced shortwards by some 11 km s-1 indicating that the cores of the strong He i lines are formed in the wind. The same holds true for the Balmer lines H and H. Herrero et al. (1992) also noticed a strong dilution effect of the He i 4471, H and H lines due to a stellar wind. In the same context, Fullerton (1990) noticed the strong resemblance between the variations of the C iv 5801, 5812 absorption lines and those of the absorption trough of the He i 5876 P-Cygni profile. From these results, Fullerton et al. (1996) conclude that the variability of the C iv doublet in the spectrum of HD 192639 is probably not consistent with a purely photospheric origin.
We have measured the equivalent widths of several strong, unblended lines in the spectrum of HD 192639. The average values and the standard deviations are listed in Table 1. Of particular interest are the classification lines He i 4471 and He ii 4542. During the 1997 observing campaign, the standard deviations of the EWs of these lines amount respectively to 7% and 5% of their intensity (Table 1). As for most of the absorption lines, the EWs of the He i 4471 and He ii 4542 lines are the weakest in August 1996, when the intensities deviate by some 14% from their mean values. Considering our whole dataset, the O-star classification criterion (Conti 1973) varies between 0.066 and 0.173, i.e. around the border line between spectral types O7.5 and O8, with an average value of corresponding to spectral type O8. However, one has to bear in mind that at least the He i 4471 line and possibly also the He ii 4542 line are partially filled in by emission from the stellar wind and their EWs therefore do not depend on the temperature of the photosphere alone.
Since the cores of most of the absorption lines are formed in the wind, one has also to be careful when deriving the chemical composition of the star using a plane parallel model atmosphere. As a matter of fact, the optical depth of the wind for different transitions not only depends on the abundances of the corresponding ions, but also on the physical conditions in the wind.
In order to search for long-term changes, we have compared the equivalent widths listed in Table 1 to the average values reported in the literature (Oke 1954, Mannino & Humblet 1955, Underhill 1995b). We find a very good agreement between our values and the EWs of Underhill (1995b). On the other hand, the older determinations yield systematically larger EWs than our data, although the various datasets usually agree within the limits set by the estimated errors and the observed range of variability. The most outstanding differences concern the He i 4471, 4713 and He ii 4542 lines that appeared some 15-40% stronger during the older observations (Oke 1954, Mannino & Humblet 1955). Whether or not these differences reflect a genuine long-term change is however not clear since a reliable estimate of the error bars on the oldest photographic EW determinations is lacking.
3.4. Radial velocities
Conti et al. (1977) found that the radial velocity of HD 192639 was probably variable over a small range. Later on, Underhill (1995a) suggested that HD 192639 could be a single-lined spectroscopic binary with an orbital period of a few days.
Binarity could be a possible interpretation of the observed deformations of the He ii 4686 line. Strong line profile variability in this line is indeed expected to occur as a consequence of colliding winds in early-type binary systems (Rauw 1997). However, the small range of radial velocity variations would probably indicate a low orbital inclination and the expected line profile variability drastically decreases for lower inclinations. Therefore, it seems unlikely that wind interactions in a binary system could account for the deformations of the He ii 4686 line observed in the spectrum of HD 192639.
We have measured the radial velocities of the most prominent absorption lines in our spectra. The central wavelengths of the absorption lines are determined by fitting Gaussian profiles and the radial velocities are computed with respect to the effective wavelengths tabulated by Conti et al. (1977). The results are given in Table 1 and the mean radial velocity of ten absorption lines is shown as a function of time in Fig. 6. This mean radial velocity has an average value and a standard deviation of respectively -8.1 km s-1 and 5.0 km s-1. The range of radial velocity variability of the different lines, quoted in Table 1, is in good agreement with the results of Fullerton (1990). Fig. 6 shows that the radial velocity of the absorption lines is slightly more positive during two nights (i.e. JD 2450641-642) than on the average. This more positive velocity occurs simultaneously with the minimum of the first order moment of the He ii 4686 line, i.e. when the emission component of the latter line reaches its maximum blueward excursion. However, it seems rather unlikely that the effect seen in Fig. 6 reflects a true velocity variation in a real binary system. In fact, in most of the absorption lines, at least part of the radial velocity variations result from the line profile variability. As discussed above, this variability is probably due to a changing stellar wind emission. On the other hand, the lower panel of Fig. 6 illustrates the radial velocity of the He ii 4542 line as a function of time. This line presents a slightly lower line profile variability and is probably less affected by emission from the stellar wind (Herrero et al. 1992) than some of the other lines used to compute the mean velocity. One can see on this figure that the amplitude of velocity variation is also smaller than for the mean of the RVs.
We conclude therefore that the present analysis does not provide any clear indication in favour of a binary nature of HD 192639.
© European Southern Observatory (ESO) 1998
Online publication: June 26, 1998