3. Spectroscopy

In addition to the photometric data, our campaign of September 1992 yielded 50 spectrograms distributed over 4 nights with a total useful observing time of about 28 hours and a baseline of 3 days. The observations were performed using a standard Boller & Chivens mod. 31523 grating spectrograph equipped with a CCD detector and attached to the 2.12 m telescope of the Observatorio Astronómico Nacional (San Pedro Martir, BC, Mexico). The adopted configuration and observational procedure gave a wavelength resolution of with a sampling of in the range , covering both H the HeI line, with a signal-to-noise ratio of about 45 dB (i.e., in a linear scale, more than 30000). The images have been processed using the MIDAS package developed by the ESO.

  
Figure 5: Average of our 50 spectrograms normalized to the continuum flux (top) and variance of the corresponding signal as a function of the wavelength (bottom). The height of the main variance peak indicates variations of flux in the H shell nucleus whose standard deviation exceeds the 3% of the continuum level. An enlargement of the HeI region at 6678 Å  is shown on the right

In the mean spectrum, which we show in the top of Fig. 5 (click here) normalized to the continuum flux, it is easy to identify, besides several telluric H ₂ O lines, three components in H (the photospheric line, a circumstellar emission and a shell absorption nucleus). Some circumstellar features (two emission wings and a very weak shell nucleus) seem to affect also the profile of the HeI line, which maintains nevertheless a basically photospheric appearence. Finally, a sharp line visible at 6613 Å cannot originate in the photosphere of a rapid rotating star like 14 Lac. If we identify it with the well-known FeI line, we must assign to its radial velocity a value of about 200 km s ^-1 . The individual normalized spectra are available in electronic form.

In the bottom of the same Fig. 5 (click here) we display, as a function of the wavelenth, the variance of the time series consisting of the successive flux values (normalized to the stellar continuum) registered at each pixel. These variances have been previously purified of the white noise contributions, which have been evaluated pixel by pixel determining, like we did in the analysis of the photometric series, the root-mean-square differences between closely consecutive data. Apparently, the star shows a considerable spectral variability, which affects mainly the circumstellar components of H. Moreover, the figure shows several peaks at the wavelengths of the atmospheric lines, obviously due to air mass and humidity changes, and indicates the presence of minor variations both in the peculiar line at 6613 Å  and in the HeI profile. We can easily realize, observing the enlargement presented on the left in the same figure, that also in this line the variability seems to affect mainly the circumstellar features.

A period analysis has been performed pixel by pixel, using Vanicek's (1971) method and scanning again the frequency interval , in the ranges covered by the visible spectral lines: (except the pixels corresponding to the atmospheric features) and . Following a procedure similar to the one introduced by Gies & Kullavanijaya (1988), in each of these three ranges we averaged the frequency spectra of the series of fluxes which describe the evolution of the line profile. Although our exiguous temporal baseline would not allow us to resolve the frequencies detected in the light curves, the resulting spectra appear consistent with the photometric time scales and lead us to rule out, also in the observed spectral variability, the presence of short period components.

  
Figure 6: Nightly mean profiles in the H region after removal of the telluric lines. The dates are consistent with the Universal Time

  
Figure 7: Decomposition of the H profile adopted for the quantification, displayed in Table 5 (click here), of its observed changes

The changes observed in H are shown in Fig. 6 (click here), in which we can compare our nightly mean profiles. In this picture the atmospheric lines have been removed replacing, in the corresponding pixels, the observed fluxes with values obtained through linear interpolations. Everything, especially in the circumstellar features, appears to change from night to night. In order to quantify these variations, we performed for each spectrogram  a nonlinear least squares fit of the H profile using three gaussian curves to represent, as shown in Fig. 7 (click here), its different components. The nightly mean values of the resulting parameters are displayed in Table 5 (click here). Interested people can ask us for the complete printout. The reliability of the radial velocity data, obviously referred to the Sun, is assured by the adoption, as comparison standards, the above quoted H ₂ O lines.

 

Hel. J.D.   Component Equiv. Width (Å) Radial Vel. (km s ^-1 )

2 448 877.870 photospheric   134 emission

shell nucleus -57

878.821 photospheric   129 emission shell nucleus

879.815 photospheric   134 emission

shell nucleus -58

880.757 photospheric   136 emission

shell nucleus -36

 

The circumstellar components show the expected considerable variations, whereas minor changes in the photospheric profile must be considered dubious: they might be produced by interference phenomena among the different components due to inadequacy of our simple fitting model.

  
Figure 8: Nightly HeI mean profiles (solid lines) compared with a synthetic photospheric profile (dashed lines) corresponding to an intrinsic half-width of 1.1 Å  with a Doppler broadening = 220 km s ^-1

The circumstellar origin of the observed variability is supported by a comparison of the HeI nightly mean profiles with a synthetic photospheric profile ( 220 km s ^-1 ; intrinsic half-width = 1.1 Å; equivalent width = 0.443 Å): the observed changes (see Fig. 8 (click here)) apparently affect only the emission wings and a faint nucleus which seems to show a double core in 3 nights over 4. Our model of photospheric profile has been also combined with one (in the case of Sept. 15) or two gaussian absorption components in a nonlinear least squares fit of the central part of this line, obtaining some meaningful parameters shown in Table 6 (click here).

 

Equivalent Width (Å) Radial Velocity (km s ^-1 ) Hel. J.D.  

Left Core Right Core Photosph. Prof. Left Core Right Core 2 448 877.870 878.821 879.815 880.757

 

  
Figure 9: Radial velocity curve of the sharp line visible at 6613 Å  in the hypothesis of its identification with the FeI line

The poor signal-to-noise ratio does not allow us to perform a profile analysis of the sharp line at 6613 Å: we can produce only its radial velocity curve (Fig. 9 (click here)) and its nightly mean equivalent widths (shown in Table 7 (click here) together with the corresponding mean values of the radial velocity).

 

Hel. J.D.   Equiv. Width (Å) Rad. Vel. (km s ^-1 )   2 448 877.870 878.821 879.815 880.757