3. Luminosities and effective temperatures and the uncertainties involved
We used the spectral catalogues by Turnshek et al. (1985) and Allen & Strom (1995) in order to determine the spectral types. As spectral features like the TiO band in late-type stars are changing rapidly with spectral type, the determination of the spectral types should be more accurate than half a subclass. Thus, the typical errors in the effective temperature are less than 30K. Table 3 compares the spectral types derived by us for the binary components with those found in the literature for the unresolved binaries. In general there is a good agreement. Only the early spectral classification of VV Cha and VW Cha by Appenzeller (1977 & 1979) was off by one spectral class due to the veiling of weak photospheric lines in the blue part of the spectra which had been used for spectral classification.
The spectral A index (ratio of the continuum flux at 703.5 nm to the flux in the CaH band at 697.5 nm) for most of our binary components is close to that of dwarf stars (c.f. Kirkpatrick et al. 1991). In general, due to veiling the A index would also be closer to that of dwarf stars and hence could affect a better agreement than there actually is. For a moderate veiling, however, the effect is small and can be neglected for all M-type stars. Thus, the surface gravity of the T Tauri stars in our sample is close to the surface gravity of main-sequence stars and significantly higher than the surface gravity of giants. Walter et al. (1994) tried to derive the intrinsic (R-I) and (V-K) colours of weak-line T Tauri stars by iteratively interpolating between the intrinsic colours of giants and of main-sequence stars. Due to the intrinsic IR excesses of classical T Tauri stars, this method cannot be applied to our sample. Interestingly, Walter et al. (1994) derived intrinsic (R-I) colours for T Tauri stars bluer than those of main-sequence stars. On the other hand, synthetic colours computed from atmospheric models of M dwarfs by Allard & Hauschildt (1995) predict redder intrinsic colours for stars with a somewhat lower surface gravity than main-sequence stars (Allard, priv. comm.). Thus, at the moment it seems appropriate to assume colours and spectral type-effective temperature relations of main-sequence stars for our sample of PMS stars. We used the compilation provided by Hartigan et al. (1994). They thoroughly discussed various sources of errors and uncertainties in estimating luminosities and effective temperatures for PMS binary components and concluded that variability of the stars is the main source of error.
For VW Cha (Sz 24), the brightest star in our sample, more than 300 photometric measurements are available in the literature, half of them in the Bessell/Cousins broad-band photometric system. These measurements yield a variability of VW Cha from V=12:m3 to 13:m0. A colour-magnitude-diagram based on 43 measurements by Bouvier et al. (1988) indicates that the variability is conform with variable extinction (Fig. 4). Variable extinction has been proposed by Grinin (1992) as one possible source for the variability observed in many young stars. It might explain the majority of the photometric variability between 500 nm and 1 µm. In this wavelength region, the largest contribution to the overall spectral energy distribution (SED) usually is from the the stellar photosphere (Bertout et al. 1988; Kenyon & Hartman 1990). On the other hand, among extreme CTTS veiling dominates the SED even at these wavelenghts. For extreme CTTS a similar trend of V vs. V-I may be caused by variations in the veiling (a stronger veiling makes the star appear bluer and brighter). In the blue (U, B band) for almost all T Tauri stars the hot boundary layer, hot spots on the stellar surface, and/or photospheric light scattered by circumstellar material contribute more to the overall SED. Hence, the observed variability in U and B cannot be explained by variable extinction alone.
The scatter of 0:m05 in brightness of the unresolved binary VW Cha around the linear fit in Fig. 4 corresponds to a scatter of 0:m035 per component, which is considerably less than the scatter of 0:m17 assumed by Hartigan et al. (1994) to be typical for T Tauri stars in the H band.
Assuming that variable extinction is the main course of variability for all stars in our sample the uncertainty in the relative photometry of the components in each binary is the largest source of error in estimating the stellar luminosity. It amounts to up to 0:m2 in V for the companion of Sz 24, the binary with the closest separation for which we were able to obtain spatially resolved photometry. In general, however, the errors are considerably smaller (0:m05). We note that this has not to be true for some extreme T Tauri stars like AA Tau, where variations of 1 mag are believed to occur due to hot star spots (Hartigan et al. 1991). The uncertainties resulting from the relative photometry of the binary components are correlated: a fainter secondary means a brighter primary and vice versa.
The highly uncertain distance estimates for the T Tauri stars pose another problem. For Chamaeleon I, e.g., the distance estimates given in the literature range from 115 pc (Thé et al. 1986) to 220 pc (Gauvin & Strom 1992). Similar discrepancies in the distance estimates exist for Chamaeleon II and Ophiuchi. In the following we adopt for Chamaeleon I and Oph a distance of 150 pc. For Chamaeleon II a distance of 200 pc will be assumed (Hughes & Hartigan 1992) as a distance as small as 150 pc would not be conform with the high Lithium abundance observed in the components of Sz 49 and Sz 59 (see below). As both components of a binary star are at the same distance, the error in the absolute distance estimate does not affect the ratio between the estimated luminosities of the two binary components.
As Fig. 4 suggests, extinction values might be variable. Therefore estimates of the extinction derived at different epochs cannot be compared directly. Interestingly, the values derived by us are lower than the values quoted in the literature for all stars. The presence of the secondary causes the unresolved binary to appear redder than the primary actually is and therefore leads to an overestimate of the line of sight extinction. On the other hand, neglecting excess emission due to veiling leads to an underestimate of the foreground extinction (Hartigan et al. 1991). Therefore, the next step will be to determine the typical veiling for the individual components of all the binaries in our sample.
© European Southern Observatory (ESO) 1997
Online publication: June 30, 1998