3.1. Short description of the spectrum
Apart from the strong H emission (EW = -8 ) and the filled-in H absorption, no apparent emission lines are found. Strong HeI lines at 4921, 5047, 5876 and 6678 show absorption and allow for a determination of the stellar velocity, which is found to be 14 3 km s-1 (heliocentric). This is consistent with other determinations (see Oudmaijer & Drew 1997). Within the observational error-bars, we do not find evidence for radial velocity variations.
The continuum corrected spectra around the H and H lines are shown in Fig. 1. H is a strong doubly-peaked emission line, while the doubly-peaked H emission hardly reaches above the local continuum. The peak separation in the H line is 242 5 km s-1 which is almost twice as large as observed in the H line (143 5 km s-1). The larger peak separation in H is consistent with the fact that both lines are formed in a rotating Keplerian type disk. If both lines, or H alone, are optically thick, the H forming region is larger than the H forming region, and will thus trace lower rotational velocities.
3.2. Variability on short time scales
Visual inspection of the 30 individual spectra did not reveal any obvious variability in the H line, so instead, we adopt the simple statistical formalism presented by Henrichs et al. (1994), and already exploited by Oudmaijer & Bakker (1994) for a similar experiment for the post-AGB star HD 56126. This method was devised by Henrichs et al. to spot the regions of interest in their multi-epoch data on hot stars. The variability can be expressed in a temporal variance spectrum (TVS):
where N is the number of spectra, represents the constructed average spectrum, the individual spectra, and = /(SNR) of each individual pixel of the spectra.
Then, the so-called temporal sigma spectrum, TSS = ), is calculated. This quantity represents approximately (). traces the standard deviation of the variations of the individual spectra with respect to the average spectrum, while represents the standard deviation of the average spectrum. If no significant variations are present in the spectra, the ratio of these two numbers, , will be close to one, deviations are directly represented in units of the noise level, that is to say a peak `Temporal Sigma Spectrum' of three corresponds to a variability at a 3 level.
During the extraction process, IRAF provides a SNR spectrum based on the photon-statistics of the data. This is very convenient, as the SNR changes strongly over a given order because the blaze of the spectrograph results in lower count-rates and thus lower SNR at its edges, while, of course, the countrates and SNR also change across absorption and emission lines compared to the local continuum. As a check, we measured the SNR in several wavelength intervals. The IRAF-extracted SNR were scaled up by 40% to bring the measured and the IRAF SNR in agreement. In the remainder of this exercise, we will use these SNR spectra as input for Eq. 1.
The average spectrum was constructed by summing all individual spectra. After this, the spectra were continuum rectified as input for Eq. 1. In the case of the H line, the wavelength region used for the fit were the blue and red continua beyond 13 from the center of the line. The spectrum was computed by dividing the individual continuum rectified spectra by their respective SNR spectra.
Fig. 2 shows an example of the usefulness of the method. In the middle panel the total spectrum in the order around the telluric absorption bands at 6870 are shown, the lower panel shows the derived TSS. The continuum shows a trend from TSS 0.9 to 1.1, indicating a slight variation in the continuum of the extracted spectrum. The telluric lines vary however at the 2-4 level. This is due to the changing airmass during the observations: as the airmass increases from 1.30 (zenith distance 39o) to 2.0 (59o) the telluric absorption becomes stronger. This is visible in the upper panel which shows an overplot of the first and last spectrum taken. The changes, which are only a few percent of the continuum level, are real.
The fact that a slight variability is traced in the continuum illustrates a very important caveat of the method. The TSS only depends on photon-statistics, and is insensitive to any systematic errors that may be present. In particular, a variable response curve (`blaze') of the echelle, can show up as variability, while in reality such variations are purely systematic rather than intrinsic. In the case of Fig. 2 this is not so important, as the entire order can be used for the continuum rectification, effectively removing this effect. In the case of H however, a large part of the spectrum can not be used for the continuum rectification as, of course, it is covered by the H line itself.
3.2.2. The H line
Fig. 3 shows the resulting TSS for the H order. The telluric lines show variability at the 2.5 level. This is smaller than the variability observed for telluric lines in Fig. 2, but can be explained by the fact that these lines and their changes are weaker. It is nevertheless an important check to note that the method also works in the H order.
H itself hardly shows any variability. In fact, the most significant variability is due to the telluric absorption in the central minimum of the line. The low 1.5 variability observed across the line is statistically not significant. Nevertheless, we investigated the possible cause of these marginal variations. This was done bearing in mind the fact that a heavily rebinned spectrum has a much larger SNR, and thus any variations should show up with greater significance.
Unfortunately, the echelle orders' wavelength ranges are rather limited (about 67 or 3000 km s-1 in the H order) compared to the extent of the line itself (full-width at zero intensity 1000 km s-1). It is thus possible that systematic variations in the continuum interpolated underneath H may be mis-interpreted as revealing intrinsic variations in the line. Indeed, by dividing all individual spectra around H by the same (rescaled) continuum fit, it became clear that the curvature of the spectra varies in time on a level less than a few%, having biggest impact on the red end of the spectrum. This is probably related to a well-known varying blaze due to the de-rotator optics in UCLES.
In order to check whether the line may be intrinsically variable, we performed some tests. The main reasoning behind these tests is that if the response curve of the echelle is variable in time, the adjacent (line-free) orders should show a similar variability. We therefore investigated the two orders next to the H order in the echellogram, and continuum rectified these using the same pixel range as the H order, i.e. not using the 25 around the center of H, and looked for evidence of variability.
We measured the EW of a fiducial line over the same pixel-range in these orders (corresponding to 26 ) as H. The measured EW in both orders is close to 0 , but has a scatter of 0.12 . This is to be compared with the scatter in the EW of the H line of 0.21 . Based on the variations of the EW of the fiducial lines in the continuum of the adjacent echelle orders and the mean height of H line over the measured interval, we would expect a scatter of 0.16 in the EW of H. The scatter of the EW of H is slightly larger than this, corresponding to variability at a 1.3 level.
Having established that the total H EW is hardly variable, the question is whether this is because the total line is not variable at all, or whether the line-profile changes in such a way that the total line-flux is nearly constant. Checks on the individual spectra show that the small changes in the line-profile are in phase with each other, and more importantly, in phase with changes in the red continuum flux. This indicates that the line-profile as such does not vary, while it traces the changes in the continuum level. Hence the insignificant variability in the EW is not due to changes in the line-profile.
Based on the facts that the EW of the H line changes at a similar amplitude as the EW in the same pixel-range of the adjacent orders, and that the `changes' traced by the TSS spectrum are in phase with the red continuum, we conclude that during the two hours of these observations, no significant variability was present in the H line of HD 76534.
3.3. Hipparcos photometry re-visited
The Hipparcos satellite observed HD 76534 125 times between 1989 and 1993 photometrically in a passband similar to the V band (ESA 1997). These observations were reported on in the paper on Herbig Ae/Be stars by van den Ancker, de Winter & Tjin A Djie (1998). These authors mentioned that the star is probably photometrically variable, which they based on the fact that the variance of the individual photometric points is larger than the observational errors. No lightcurves were provided however.
Here we look at the data provided by the Hipparcos catalogue into more detail. The photometry is plotted as function of Julian Date in Fig. 4. In the first year of the mission, HD 76534 was constant within the errorbars until the object brightened by about 0.1 magnitude, reaching a maximum around May 1991. The period of brightening and fading lasted about 100 days. Afterwards the object `flickers' around a mean value close to what was measured before the maximum.
Could this rise in brightness be associated with the spectral behaviour of the object? Mennickent et al. (1998) published 11 year long photometric monitoring of Eri, and found several similar changes in the Strömgren photometry of the object. A period search revealed that rises in brightness of 0.1 mag. occurred with a period of 486 days, while the rising and fading of the object lasted about 100 days. From the colour changes, they found that the brightness maxima correspond to slight increases in effective temperature of the star. Although the overlap between spectroscopic and photometric data is not very complete, Mennickent et al. find a rough correlation between the jumps in brightness and periods of H emission in the star.
© European Southern Observatory (ESO) 1999
Online publication: October 4, 1999