Derivation of veiling, visual extinction and excess flux from spectra of T Tauri stars

This work aims to analyse within a rigorous framework, veiling, visual extinction and excess extraction from the spectra of T Tauri stars. We investigate further the method of Hartigan et al. (1989) for veiling estimate from small spectral bandwidth of a few tens of Angstroms. The calculated veiling value is sensitive to the estimated noise ratio and to spectral mismatches between the object and the reference. We show that an incorrect input noise ratio together with low contrast spectra in noise units can lead to important biasses and we propose solutions to minimize this problem. In case of spectral mismatches and for large contrast spectra compared to the residual of the veiling equation, the relative veiling bias is dominated by the apparent veiling of the reference with respect to the correct underlying object stellar spectrum. The veiling error is found to be proportional to the square of the veiling when the latter becomes larger than unity. If we are limited by the statistical noise, it is little dependent on the spectral resolution. Because of systematic errors, however, it will be difficult to estimate the veiling in a very small bandwidth at spectral resolutions of a few hundreds. For visual extinction and excess estimates, we generalize the discrete method of Gullbring et al. (1998) by a continuous approach. This new approach, which uses the spectra as a whole through a continuous modelling, has been successfully tested on simulated data. The visual extinction error is proportional to the veiling when the veiling becomes larger than unity and to a function which depends on the input reference spectrum. This function decreases with increasing spectrum contrast, which means going from earlier to later spectral types. If we are limited by the statistical noise, it is, like the veiling, little dependent on the spectral resolution. For very active T Tauri stars or when the excess is dominated by emission lines, however, it will be difficult to handle very low spectral resolutions, because of systematic errors. The real sensitivity to biasses and the performances of the algorithm are to be studied experimentally. Nevertheless, an efficient use of all the information contained in the spectra through the proposed “continuous” approach, together with a better understanding of the sources of bias, can greatly help to derive the visual extinction and the excess on objects much fainter than those so far studied. Send offprint requests to: A. Chelli