The practical implementation of spectral subtraction can also have problems, particularly for binary systems such as those studied in this work. In all cases it is necessary to choose standards which closely match the spectral type, surface gravity and chemical composition of the active star. A mismatch in spectral types between the inactive and active stars would manifest itself as a concavity or convexity in lines of interest due to the dependence of line wings. However at medium resolution it has been demonstrated that the strength of H is relatively constant for the range of spectral types from G2 to K5 (Huenemoerder & Ramsey 1984). The line profiles in many active stars are dominated by Doppler broadening with high rotation rates so standard spectra have to be artificially broadened and shifted to match the velocities of the active star. For the case of binary stars the situation is more complex since the lines from individual components may be highly blended close to conjunctions.
The implementation of the spectral subtraction technique for this work is similar to that discussed by Barden (1984). However we do not use an iterative matching technique to derive radial and rotational velocities and intensity weights due to the blending of lines. We recently checked the validity of the ephemeris and velocity semi-amplitudes of ER Vul (Gunn et al. 1996) so that accurate model velocities could be computed close to conjunction. These along with rotational velocities taken from Barden (1984) were used in the spectral synthesis.
Traditionally intensity weights for spectral subtraction have been obtained from photometric light-curve observations. During eclipse these are not representative of the light ratio for the system hence intensity weights are computed using the assumed component radii for each star and the values of the Planck function for each effective temperature computed at the spectral order central wavelength. The effective temperatures are computed using the standard spectral type-temperature relations given by Schmidt-Kaler (1982). During eclipses an obscuration correction is applied to the intensity weights. This simple geometric model does not take account any inhomogeneities in the surface brightness of the eclipsed star which will arise from active regions and limb-darkening effects. Details of how standard spectra are combined to form the synthetic spectrum are given by Barden (1984). The software used for this work is called CORREL (Gunn 1995) and is an interactive command-line driven package allowing complete automation of the analysis procedure. The final subtracted spectra were analysed by measuring excess absorption or emission features in the active lines of the target star.
© European Southern Observatory (ESO) 1997
Online publication: July 8, 1998