An automated system to classify stellar spectra — I

Analyses of stellar spectra often begin with the determination of a number of parameters that define a model atmosphere. This work presents a prototype for an automated spectral classification system that uses a 150-Å-wide region around Hβ, and applies to stars of spectral types A–K with normal (scaled solar) chemical composition. The new tool exploits synthetic spectra based on plane-parallel flux-constant model atmospheres. The input data are high signal-to-noise ratio spectra with a resolution greater than approximately 1 Å. The output parameters are forced to agree with an external scale of effective temperatures, based on the infrared flux method. The system is fast — a spectrum is classified in a few seconds — and well suited for implementation on a web server. We estimate upper limits to the 1σ random error in the retrieved effective temperatures, surface gravities and metallicities as 100 K, 0.3 and 0.1 dex, respectively.