2 Stellar content and data

2.1 Physical parameters of the stars

The present sample was assembled by merging the stellar library of F, G, K stars (Soubiran et al. 1998) serving as a basis for the TGMET program with additional spectra taken from other observing programs and from the near-to-line ELODIE archive (http://www.obs-hp.fr). The spectra of the database correspond to stars which have been selected because they have published values of $T_{\rm eff}$ or ( $T_{\rm eff}$, $\log~g$, [Fe/H]), or reliable estimations of the absolute magnitude M_V, deduced from Hipparcos parallaxes. The main source of atmospheric parameters is the catalogue of [Fe/H] determinations, 1996 and 2001 versions (Cayrel de Strobel et al. 1997 and 2001) but two other sources have been also used: Carney et al. (1994) and Thévenin (1998). Effective temperatures were also taken in the lists published by Blackwell & Lynas-Gray (1998), Alonso et al. (1996a) and Alonso et al. (1999a), or calculated from the colour indices V-K or b-y following the relations $\mbox{$T_{\rm eff}$ }=f$(colour, [Fe/H]) established by Alonso et al. (1996b) and (1999b) for dwarfs and giants respectively. Each star had ( $T_{\rm eff}$, $\log~g$, [Fe/H]) estimated by averaging the determinations found in the literature, giving a half weight to old references and Strömgren photometry and a double weight to $T_{\rm eff}$ determined by Blackwell & Lynas-Gray (1998). As in the TGMET library, the parameters were then labelled according to their reliability which was estimated from the number of determinations and their standard deviation around the mean. The reliability scale runs from 0 to 4, with the highest reliability 4 corresponding to uncertainties lower than 80 K in $T_{\rm eff}$, and 0.06 dex in [Fe/H]. These uncertainties can reach 115 K and 0.09 dex for reliability 3, and 150 K and 0.11 dex for the reliability 2. Reliability 1 is attributed for parameters based on old determinations or photometry, or presenting large discrepancies between references. Reliability 0 means that no reference on atmospheric parameters was found in the literature. The database consists of 908 spectra corresponding to 709 different stars which are shown in the plane ($\log$ $T_{\rm eff}$, $\log~g$) in Fig. 1 and in the plane ($\log$ $T_{\rm eff}$, [Fe/H]) in Fig. 2. $T_{\rm eff}$ ranges from 3700 K to 13600 K, $\log~g$ ranges from 0.03 to 5.86 and [Fe/H] ranges from -2.8 to +0.7. In these two plots, new estimations of ( $T_{\rm eff}$, $\log~g$, [Fe/H]) were used instead of those from the literature, which are incomplete. These new estimations were obtained with the current version of TGMET which is still under development (a paper presenting this new version is in preparation and a description is given in the electronic version of the archive). Atmospheric parameters from the literature together with new estimations and measured line indices from the spectra presented here are given in Table 1, available only in electronic form (see Sect. 6).

Figure 1: $T_{\rm eff}$ vs. $\log~g$ for the 709 stars of the archive. For the sake of homogeneity, the atmospheric parameters presented here are not from the literature, which is incomplete, but were estimated with the current version of TGMET (Sect. 5.3.3)

More information on each star is available in the header of the FITS spectra. It includes the absolute magnitude M_V from Hipparcos parallaxes when available. A scale of reliability was also established from the precision of the parallax and V magnitude from Tycho-2 (Høg et al. 2000), 4 corresponding to parallaxes with a precision better than 10%. Also available in the FITS headers are the spectral types, (V, B-V) taken from the Tycho-2 catalogue (converted from the Tycho-2 $B_{\rm T}$ and $V_{\rm T}$) and radial velocities usually measured at the telescope. The on-line cross-correlation technique was used to measure the radial velocity of strong-lined spectra, corresponding to moderate effective temperatures up to 6500 K but for hotter stars, the radial velocity was estimated using a least-squares deconvolution technique (Donati et al. 1997).

The database includes several stars with line profiles broadened by rotation, macroturbulence or binarity and a few stars with spectral peculiarities (Ap, Am, emission line stars...). These pecularities are indicated in the headers.

Figure 2: Distribution of the 709 stars of the archive in the plane $T_{\rm eff}$ vs. [Fe/H] estimated with the current version of TGMET (Sect. 5.3.3)

2.2 ELODIE spectra

ELODIE provides a spectral range of 390-680 nm recorded in a single exposure as 67 orders on a 1K CCD at a mean resolving power of 42000. Optimal extraction and wavelength calibration are automatically performed by the on-line reduction software TACOS. Complete information on ELODIE can be found in Baranne et al. (1996) and in the TACOS user manual by Queloz (1996). The global efficiency of the instrument drops in the blue by a factor 5 at 440 nm for a solar-type star and by a factor 50 at 390 nm. Hence we decided to limit the spectral range to $\lambda\lambda = 410{-}680$ nm.

The typical signal-to-noise (S/N) ratio per pixel at 550 nm is 150 (it is higher than 80 for more than 90% of the spectra). Several spectra with a lower S/N, down to 35, correspond generally to faint deficient stars which have been included in the database for a better sampling of the parameter space.

For the present purpose, the 908 spectra of the database were processed to provide 1D spectra rebinned in wavelength and calibrated in absolute fluxes. Since the main goal of the archive was to construct a library of spectra representative of the various stellar types and an interpolated grid of spectra covering the parameter space, all the spectra were reduced to the rest-frame. The archive was prepared from the flat-fielded and order-extracted TACOS spectra in two main steps. (1) The orders were de-blazed, the spikes due to cosmic rays and telluric lines were masked, all the orders were connected in one spectrum, and the pseudo-continuum normalization was computed. (2) The spectra were calibrated into "physical'' flux, i.e. above the atmosphere.