1 Introduction

Spectral libraries covering the HR diagram with medium to high resolution and large spectral range are essential tools in astronomy. We have identified two areas where there will be a special need of such libraries in the coming years: the automated parameterization of stellar spectra and the spectral synthesis of stellar populations of galaxies. New multi-object spectrographs on large telescopes, such as Giraffe on the VLT, will soon make it possible to observe hundreds of objects during the same exposure with an unprecedented resolution and spectral coverage. The challenge will be to extract the maximal information in a reasonable time.

For stars, medium to high resolution spectra are the basis for determining radial velocities, atmospheric parameters ( $T_{\rm eff}$, $\log~g$, [Fe/H]) and eventually projected rotational velocities, $v\sin i$. A new method, called TGMET, is emerging to make the parameterization of stellar spectra in terms of temperature, gravity and metallicity fast and automatic (Katz et al. 1998). It relies on the direct comparison of a target spectrum to a library of reference spectra with known atmospheric parameters (Soubiran et al. 1998). TGMET was originally dedicated to the measurement of F, G, K stars observed with ELODIE. The present paper continues the TGMET project. The two major points addressed are (1) the extension of the reference library to all spectroscopic types and luminosity classes and its densification, and (2) to resample the spectra in wavelength (i.e. remove the instrumental signature in ELODIE spectra and connect the echelle orders together) in order to enable comparison of spectra of any origin.

The TGMET project appeared to be also very useful for extragalactic studies where medium resolution libraries with large coverage in stellar parameters are needed. Stellar libraries, calibrated in flux, are used to model composite spectra through stellar population synthesis (e.g. the PEGASE program, Fioc & Rocca-Volmerange 1997). An important issue is to disentangle the effects of age and metallicity and it has been demonstrated that the spectral resolution is a key factor (Worthey & Ottviani 1997, hereafter WO97; Vazdekis 1999). To benefit from a resolution higher than about R=2000both the internal kinematics and the characteristics of the stellar populations must be fitted simultaneously. Making available an extended library of stellar spectra is the first step in this direction.

Current libraries are based on low resolution spectroscopy (e.g. Jacoby et al. 1984; Serote Roos et al. 1996; Pickles 1998) or they are restricted to a limited area of the HR diagram (e.g. Montes et al. 2000 and Soubiran et al. 1998 for F, G, K stars). Here we present an homogeneous dataset of spectra which fills the deficiency of such libraries. The sample of stars and associated data are presented in Sect. 2. Sections 3 and 4 describe the two phases of the data processing. In Sect. 5 we evaluate the quality of the archive. The access to the data through HYPERCAT or at the Centre de Données astronomiques de Strasbourg are described in Sect. 6.