4 Veiling versus brightness

Figure 11: Veiling versus brightness and colour

Figure 12: Two spectra with very different veiling, but with the same brightness of the star. Lower: the spectrum of $\gamma$ Cep artificially veiled by a factor 2

Figure 13: Solid lines: average spectra of RW Aur A at different levels of veiling. Thin lines: comparison spectra of 40 Peg (G8II), $\gamma$ Cep (K1III-IV) and 61 Cyg A (K5V), artificially spun up to $v\,\sin\,i=30$kms-1 and veiled to the same levels as RW Aur A

The veiling of the photospheric spectrum was determined in several spectral regions: 5556-5610Å (Fe I, Ca I), 5719-5751Å (V I, Fe I), 5855-5870Å (Ca I), 6010-6045Å (Mn I, Fe I). In the blue region of the spectrum, the broad emission lines are blended, and there are no continuum windows. Therefore we could not select photospheric lines useful for veiling determinations shortward of 5000Å.

The veiling was derived using two different methods. The first one is based on the method described by Hartigan et al. (1989). This method consists of a comparison, within a small wavelength band of a few tens of Å, between the spectra of the T Tauri star with that of a template star. The latter is chosen so that its spectrum accurately matches the unveiled photospheric spectrum of the target. The code implemented in Porto also provides the radial velocity and the rotational broadening of the target star when compared with the template star. Normally the code adjusts simultaneously four parameters: the radial velocity, $v\,\sin i$ and two parameters related to the continuum excess. The code also provides mechanisms for constraints on parameter values, useful when some free parameters have been measured previously using a different method.

For RW Aur A we used $\gamma$ Cep (K1III-IV) and 61 Cyg A (K5V) as template stars. Both templates give similar results within the uncertainties.

The second method is based on measurements of the equivalent widths of selected absorption lines in spectra of the target and the template star: $\rm veiling=EW$(template)/EW(TTS)-1. The template spectrum was rotationally broadened to $v\,\sin i=25$kms^-1 in order to get the same blends of lines as in the spectrum of RW Aur A. In this method we assume that the line width ($v\,\sin i$) is the same in all the spectra, and that the EW of a line varies due to the veiling variations. The errors come mainly from the uncertainty in the continuum level, especially in the spectra with stronger emission lines.

There is a correlation in the time variation of the veiling derived from different spectral regions. However, the absolute value of the veiling can differ significantly from region to region. In Table 1 we give the average values of the veiling obtained by the two methods derived from the Fe I lines in the region of 5556-5610Å. These are the strongest lines of higher excitation potentials, and they are less sensitive to temperature, which makes the selection of the template less critical.

The most unexpected result of this study is that no correlation was found between the veiling and the brightness of the star, although the brightness varied by one magnitude, and the veiling varied over a wide range (Fig. 11). Moreover, quite different veiling factors were found in spectra taken at the same brightness of the star, for example, in the nights HJD 2451124 and 1125, when photometry was made at two telescopes. Another example (HJD 2450382 and 0384) is given in Fig. 12 showing different veilings (1.5 and 7.0) but no significant difference in brightness. The veiling was determined in the spectral region of the V band. If the veiling were caused by additional continuum radiation in this spectral region, the brightness difference in V would be more than one magnitude!

Another oddity is the variability of "$v\,\sin i$'' from night to night in the range 16 to 40kms^-1. The effect is real, not instrumental, e.g. the sky line at 5577Å remains perfectly narrow in the spectra which show broadened photospheric lines. The minimal value of $v\,\sin i$of 16kms^-1 can be considered as due to stellar rotation. No periodicity in the variations of the "$v\,\sin i$'' parameter was found. We suggest that the broader lines are contaminated by an additional absorption from layers above the photosphere. An argument in favour of this suggestion can be seen in the variability of the Ba II 5853Å line. The line was enhanced in some nights, that is the ratios of Ba II/Ca I or Ba II/Fe I were too large for a dwarf. This strong line, indicating $\log\,g\leq2$ like in supergiants, was observed mostly in spectra with low veiling (Fig. 13).

We conclude that Ba II is formed in a shell above the photosphere. The same shell must contribute also to absorption in many other lines, which make them deeper and broader, thus simulating low veiling and variable $v\,\sin i$. The spectral features of a "warm shell'' in RW Aur were also found by Herbig & Soderblom (1980). More information about the shell lines is given in Sect. 7.

Hence, we conclude that the variations of the veiling we observed in RW Aur A are caused by at least two different processes: 1) the "true'' veiling due to continuum + line emission, and 2) absorption in the shell which imitates low veiling. This may partly explain why the observed "veiling'' is not correlated with brightness, although the considerable increase in veiling at constant brightness remains a mystery.