4. The veiling continua
The variable and often quite small equivalent widths of the photospheric absorption lines in TTS can be attributed to an additional continuum and perhaps emission line source. The veiling - usually defined as the ratio of the non-stellar to stellar continuum flux - can be studied either by measuring the equivalent widths of the photospheric lines in a TTS and an appropriate template star directly,
or by calculating the cross-correlation function between the two spectra (normalized by the auto-correlation function of the template) in regions that are free of emission lines and atmospheric absorption lines (for details of this method see, e.g., Guenther & Hessman 1993). The cross-correlation method has the advantage of using essentially all photospheric lines in the spectrum but has to be corrected for the difference in broadening of the two lines. We used both methods for our spectra, and find that both methods are in very good agreement with each other.
One problem in measuring the veiling is the choice of the appropriate template star: the spectral types of TTS are often not very well known. For example, the spectral type of DR Tau in the literature spans from K5IV (Imhoff & Appenzeller 1987) to M0V (Guenther & Hessman 1993). Deriving the correct spectral type is important, because the majority of the photospheric lines of T Tauri stars are stronger in cooler stars. Thus, if the assumed spectral type is too early, the veiling will be overestimated. However, in this case, the veiling will not only be overestimated but will also depend upon the the spectral lines used: the veiling should be independent of the excitation potential for templates star of the correct spectral type. Deriving the veiling by measuring EW (obs)/EW (template) for lines of different excitation potentials and using many different template stars is thus the best way to find the correct template (Basri & Batalha 1990).
We measured the veiling from 4000 to 6500 Å using well-classified template stars spanning a very wide range of spectral types. However, in the following we will not only use the best template star (see Table 2) but other (reasonable) template stars as well to check whether the results depend upon the choice of the template, or not. Errors in the reddening-corrections or flux-calibrations are not serious problems for the veiling measurements, because only the equivalent widths of the spectral lines are of interest.
In principle, it might be possible that variations of the veiling
are caused by variations in the effective temperature of the star
(e.g. extreme star-spot activity), and thus the best template may vary
from spectrum to spectrum. In order to test whether the spectral type
of a T Tauri star is variable, or not, we plotted the excitation
potential of individual lines in both DI Cep and DR Tau
(only 10 lines were available in the spectra of DG Tau) against
the ratio of the equivalent widths in a minimum and maximum veiling
state. The result for DI Cep is shown in Fig. 3. In both cases,
the amount of veiling does not depend on the excitation potential, and
so the spectral types are constant.
The average veiling in DR Tau was
using a K5V star as template and
or using two different
M0V stars. The variations of the veiling are shown in Fig. 4. The
plotted variances are computed from the measurements of the individual
nights, and thus include the errors of the measurement and the
variations. However, the errors of the measurements alone are about
0.21 and thus much smaller than the intrinsic variations, or the
errors due to the difficulty in finding the correct template. The
curves for the veiling derived by using different template stars are
highly similar. In general, the veiling seems to be large when the
equivalent width of H is small. One exception
seems to be JD 2 448 516, where the absolute value of the equivalent
width is low and the veiling is low. However, the variability on this
particular night was extremely large and thus the plotted variance of
H is very large. Since the equivalent width for
an emission line is negative, we find a correlation between the
equivalent width of H and the veiling of 0.56
(first M0V template), and 0.55 (K5V and second M0V
template)(Fig. 5). If the value from the night of JD 2 448 516 is
removed (DR Tau was particularly active on that night), this
correlation increases to 0.82. Thus, there is a significant
anti-correlation between the strength of H
and the veiling in DR Tau (false alarm
probabilities of 7-8%). The very large similarity between the
correlation value derived by using different templates clearly
indicates that this conclusion is not influenced by the choice of the
Although the veiling in DG Tau appears, in
general, to be less than that in DR Tau, the determination is
much more difficult in DG Tau: the very rich emission line
spectrum leaves very few uncontaminated continuum regions
(Fig. 1). Imhoff & Appenzeller (1987) derived a spectral type
of G5IV for DG Tau. In contrast to this, we derive a spectral
type of about K6V by measuring the equivalent widths of photospheric
lines with different excitation potentials. For completeness, we
measured the veiling using a G5IV star, a K0V star, and a K6V star
(see Fig. 4). By far the most prominent photospheric feature is
the LiI 6707 line: it cannot be used to derive the
absolute value of the veiling but it can be used for determine the
relative veiling. We thus derived the relative veiling using this
line, and rescaled it so that the average veiling of the Lithium line
matches the average veiling as derived from the K6V star. Fig. 4
shows the variations of the veiling and the equivalent width in H
. The veiling measured with the K6V template
ranges from about 1-4. We find some correlation between the veiling
and the variations of the equivalent width in H :
0.52 for the G5IV template, 0.53 for the K0V template, and 0.54 for
the K6V template. The correlation between EW(H )
and the veiling as derived from the variations of the Lithium line is
0.71. As in the case of DR Tau, this implies an anti-correlation
between the strength of H and the amount of
veiling (false alarm probabilities of 2-10%). The results are
summarized in Table 2.
Using the high resolution data, we derive a spectral
type for DI Cep of G8IV, compared to G8V listed in the Herbig and
Bell Catalogue (Herbig & Bell 1988). The average veiling is
for the G8IV template, and
to for two G8V
templates). Fig. 4 shows the veiling and the equivalent widths of
H for DI Cep. The similarity between the
two curves is very low: the correlation between EW(H
) and the veiling is 0.18 for all templates.
However, this low correlation is at least partially due to the small
amplitude of the variations.
4.1. Is the veiling just an additional continuum source?
At first glance the photometric measurements seem to indicate that the veiling continuum is not an additional continuum source: the relative variations of the veiling are always larger than the relative broad-band photometric variations measured (See Table 2). However, a closer inspection shows that photometry and measurements of the veiling are in good agreement. First of all, because of the definition of the veiling V, should be compared with , and not . Secondly, the veiling was measured in the 4000-6500 Å region, whereas the photometry in the the I (7600-10 000 Å) band, and it is well known that the photometric variations are larger in the blue than in the red (Hartigan et al. 1994). For example, Herbst (1990) measured the photometric variations of DG Tau and DR Tau on many nights in several years. Computing the variance of the intensity variations in his UBVRI measurements gives 31.9%, 24.2%, 16.9%, 14.0%, and 13.3% for DG Tau. Assuming the same increase from the red to the blue and taking our measurements (variance in the I band 18.7% for DR Tau) we would expect an amplitude of about 22-30% for the photometric variations in the 4000 to 6500 Å regime, which is in excellent agreement with the measured variation in of about 27%. This result is thus in very good agreement with the hypothesis that the veiling is an additional continuum source.
In principle, the veiling might also be created by the filling-in of absorption lines. If this is the case, the filling in would be larger for lines with larger equivalent widths (i.e. highly saturated lines) than for lines with smaller equivalent widths. Thus, the ratio EW (TTS)/EW (template) would be smaller for lines of small equivalent width than for lines of large equivalent width. Similarly, the ratio between the equivalent width of lines of the T Tauri star in two different veiling states would depend on the equivalent width of the line. This test is slightly more sensitive than the previous one, because problems in finding the correct spectral type are eliminated. We find no dependence of this ratio with the equivalent width in our data, indicating that the veiling is unlikely to be produced by filling-in.
The best way to test whether the veiling continuum is predominantly an additional continuum source, is to plot against the absolute fluxes f in the same spectral region. If the veiling is just an additional continuum source, then , and for . Although the nights were not strictly photometric and we did not always have a relative photometric calibration, the slit losses and fluctuations in transparency and seeing were normally small enough, that we can estimate the true fluxes in the 4000-6500 Å band and plot them against . The result is shown in Fig. 6. We find, indeed, that the data are consistent with having , and the for . Thus, we confirm that the veiling effect is really due to an additional continuum source (see, e.g., Hartigan et al. 1989).
© European Southern Observatory (ESO) 1997
Online publication: June 30, 1998