V. Malyuto et al., QUANTITATIVE SPECTRAL CLASSIFICATION OF GALACTIC DISC K-M STARS FROM SPECTROPHOTOMETRIC MEASUREMENTS, Monthly Notices of the Royal Astronomical Society, 286(2), 1997, pp. 500-512
New spectral observations for 47 southern galactic red supergiants obt
ained with the new RUBIKON spectrophotometer (developed at the Astrono
misches Institut der Ruhr-Universitat Bochum) at the Bochum 61-cm tele
scope on La Silla are presented. The spectra range from 4800 to 7700 A
ngstrom and their resolution is 10 Angstrom. The mean error of absolut
e fluxes is 0.028 mag and that of relative fluxes 0.021 mag. The spect
ra will be available at the Strasbourg Stellar Database (CDS). Togethe
r with data taken from recently published spectral catalogues, the new
observations have been used to define spectral indices as measures of
the strengths of the following features: Fe I + TiO alpha(1), Mgb + T
iO alpha(0), NaD + TiO gamma'(1), TiO gamma'(0) and TiO gamma(1) syste
ms. The indices have been checked against errors introduced by reducti
ons, interstellar reddening and different resolutions of different spe
ctral catalogues, and have been found to be very insensitive to all th
ese effects. Therefore, different catalogues may be combined without a
ny loss of accuracy and homogeneity. The mean error of a single index
has been found to be 0.011 mag. For stars from K4 to M7, a strong temp
erature dependence is found for all indices. For the Fe I + TiO and es
pecially the Mgb + TiO features, a strong dependence on luminosity has
also been observed. These indices therefore have been combined to for
m a luminosity index, while the others together form a spectral index.
The combined indices have been calibrated in terms of MK data using t
he stepwise linear regression technique, and may be used for quantitat
ive two-dimensional spectral classification of late K- and M-type star
s. The mean error of the classification is 0.6 of spectral subtype and
0.8 of luminosity class, which is much higher than would be expected
from the uncertainty of the indices alone (which, e.g., for an M4 gian
t correspond to an uncertainty of 0.1 of spectral subtype and 0.3 of l
uminosity class). This may be explained by the uncertainty of the orig
inal MK classifications and the variability of some programme stars.