The polarization is almost constant along the entire spectrum with P = 0.5% and a position angle of at H (see Table 2). On the occasions that we observed HD 76534, both the polarization and the position angle remained essentially constant.
The position angle is in close agreement with the results by Jain & Bhatt (1995), who observed HD 76534 polarimetrically in broad band on two occasions. Their results showed a significant change in the polarization (from 1.26 0.13% to 0.26 0.12% within two months in the R band), while the position angle remained constant within the errors ( and respectively).
If the change in polarization reported by Jain & Bhatt is real, it is reasonable to expect that the bulk of the polarization is circumstellar. This poses the question as to where the circumstellar material responsible for the observed polarization is located. As described by Thé et al. (1985) and Hillenbrand et al. (1992), there is no significant excess emission in the near-infrared which could be attributed to a dense dusty envelope or a circumstellar disk. The IRAS flux densities do not support the presence of much circumstellar matter either. The point source cross-correlation coefficients listed in the IRAS Point Source Catalog (see IRAS Explanatory Supplement, 1985) for this object are less than 100% indicating the presence of an extended infrared source, rather than an unresolved point source. As pointed out by Oudmaijer (1996), such extended IRAS sources are due either to a large extended circumstellar dust shell, or to the star heating up the local interstellar material. Since HD 76534 is located close to a star-forming region, the latter possibility is more likely. In either case, the dust traced by the IRAS detections will not be opaque enough to account for the scattering.
Thus, if the (change of) polarization is due to circumstellar scattering, the agent is not visible in the photometric data. If the polarization were due to a modestly ionized circumstellar disk, where electron scattering is the main source of polarization, it is hard to explain the constant polarization between spectra with and without H emission, as found in Run 2 and 3. Clearly, the question of the origin of the polarization in HD 76534 is not settled yet.
3.2. The spectrum
The radial velocity of HD 76534 was determined by fitting Gaussians through several strong absorption lines in the medium resolution spectrum. This yielded a value of (19 3 km s-1), in agreement with Thackeray, Tritton & Walker (1973) who determined the radial velocity to be 17 km s-1.
In the medium resolution spectra, the H line is double peaked, with a peak-to-peak separation of 125 km s-1. This value is, within the error margins, equal to the v sini of 110 km s-1 of the underlying photosphere (Thackeray, Tritton & Walker, 1973). The central reversal of the emission is located at the stellar rest-velocity.
The full width at the 1% level of the emission line is 740 km s-1 and the full-width-at-half maximum (FWHM) of the line is 350 km s-1, which is larger than the observed correlation between the v sini and the FWHM of the H emission of a Be star would imply (Hanuschik, 1989). It is therefore possible that this surplus broadening may be due to optical depth related effects (e.g. electron scattering). It is also possible that the broadening is in part caused by a radial outflow component.
The H emission line profile is similar to the observation of Thé et al. (1985). They found a strong doubly-peaked emission line, where the maximum of the emission reached 2.5 times the continuum level. From their Fig. 1, we measure an EW of -7 0.5 Å.
3.3. Spectral variations
The spectra of Runs 1, 2A, 2B, 3A and 3B have been continuum normalized and plotted around H in Fig. 1. The upper panel shows the spectrum on 9 January. The line is strongly in emission and has an equivalent width of -5.5 Å. The middle panel shows the low resolution data taken on 11 January. The H line appears much weaker, it effectively fills in the absorption line and hardly exceeds the continuum level. The dotted line represents the spectrum taken only 7 minutes later. The emission decreased between these two observations to the point of disappearing. The lower panel shows the high resolution spectra taken 2 hours after the data taken in the middle panel - H is strongly in emission. The dotted line represents the spectrum half an hour later. Between these two measurements, the EW increased 23%.
The picture that emerges from these data is that between Run 1 on 9 January and Run 2 on January 11, the H line strength had decreased, yet it was strongly in emission only two hours later, still increasing in equivalent width. Clearly, we do not know what the pattern of variation was between the first observation on January 9 and Run 2, 2 days later. The timescale of these variations is an order of magnitude shorter than similar observed behaviour in Be stars like µ Cen.
It is appropriate to investigate whether changes in the continuum flux level could be the reason of the difference in the EW of the emission line. We inspected the number counts in the continuum close to H . Comparisons between Runs 2A and 2B and also between Runs 3A and 3B indicate number count differences in the region of only 3%. This suggests that the continuum level could not have changed enough to account for the H equivalent width changes observed within each of Run 2 and Run 3. We do not have the necessary data to intercompare Runs 1, 2 and 3.
There could be a trivial cause for the variation of the H emission; HD 76534 is a visual binary where the components are separated by , with the secondary 1.3 magnitude fainter in the Gunn z band than HD 76534 (Reipurth & Zinnecker, 1993). In the unlikely event that the AAT tracking would fail on a timescale of minutes, the secondary can dilute (enhance) the spectrum of HD 76534 giving rise to a lower (higher) EW of H . The arguments against this having happened are (i) we have no evidence that the continuum level changed significantly during the observations, (ii) the polarization remained essentially constant, (iii) the position angle of the binary () does not allow both objects to fall within the wide N-S slit, as used during the high resolution observations, at the same time.
As the observed variability is most likely not caused by instrumental or atmospheric effects, we can use the data as timeseries spectra. Each polarization spectrum consists of 4 individual spectra with short integration times. Assuming that the polarization across H is not deviant from the continuum, the low resolution spectra of Runs 2A and 2B constitute a time series of 8 spectra (the sum of the E and O rays) taken within 10 minutes, the medium resolution spectra consist of 8 spectra taken within 41 minutes.
The variation of the EW of H is plotted in Fig. 2, where it decreases from 0.5 to almost 3 Å in 10 minutes in Run 2, while two hours later the emission increases linearly with time; the EW increases from -6.6 to -9.5 Å in 40 minutes, corresponding to 4.35 Å per hour. If the underlying absorption has an equivalent width of 2.8 Å, the maximum observed absorption, we can estimate that the transition from absorption to emission, if steady, would have taken 2.8 hours. Throughout Run 3, the H line has the same peak-to-peak separation and FWHM.
Apart from the H emission, few spectral changes are visible between the spectra of 9 and 11 January. The absorption lines have the same strength within the errors, as illustrated in Table 2, where the EW of the strong He 6678 line is presented. Like the EW, the radial velocity of this line was constant within the errors throughout Run 3. It appears that the spectral type did not change perceptibly between these two spectra.
In the following, we discuss two possible causes for the observed changes, a sudden mass outburst, where the newly ejected ionized material gives rise to emission, or a change in the ionizing continuum from the star which ionizes a different fraction of a steady state circumstellar envelope.
© European Southern Observatory (ESO) 1997
Online publication: July 8, 1998