6 Spectrum variability

We have already mentioned in the introduction that Gravina (1982) found variations in the different emission features over four nights which can be up to a factor of five. Since her plates were not calibrated, the indications point toward a strong variability. We can quantify this with the lines which are present in our spectra taken at different dates. Table 4 presents the results which are plotted in Figs. 2 and 3. Let us remark that we have only considered lines which on the two plates possess the same identification and which have been measured as isolated features. The fact is that on the two pairs of plates often the lines and their identifications are not the same.

  

Figure 2: Equivalent widths of lines in the spectrum from the 29.01.98 plotted against equivalent widths in the spectrum from the 17.08.91. Crosses represent Paschen lines

  

Figure 3: Equivalent widths of lines in the spectrum from the 23.07.93 plotted against equivalent widths in the spectrum from the 29.01.98. Crosses represent Paschen lines

 

Table 4: Comparison of equivalent widths of lines at different epochs

 

Note: The first column provides the wavelength of the line in Å; the second and the third, the equivalent width of the line at the dates given at the heading of the table, also in Å. The fourth column provides the line identification.

In Fig. 2 one can see clearly that on the average the equivalent widths have changed by a factor of 2, with some individual deviations from the average relation. Figure 3 shows a much more scattered relation, which also holds between the limits of a factor 2. The only exception is the N I line $\lambda$8703 which deviates more than that.

The variations found here are in line with what we found in other stars of this type-typically the variations in equivalent width are of the order of a factor 2. Due to the lack of time coverage we do not know if these variations are of daily, monthly or yearly character. If we take the example of other stars of this type examined in this series of papers, the variations correspond to time scales of the order of days. In favor of this latter assumption speaks the fact that among the few lines which overlap in the interval 6270-6406 on January 28 and 29, 1998 one finds variations of the order of $\pm$20$\%$.

The variations are also evidenced by the H$\alpha$ measurements carried out by Swings & Andrillat (1981) and the present paper.