The grid of 254 computed spectra is listed in Table 1. The range in temperature extends from 3 500 to 7 500 K, with a 250 K step. The models have been computed for metallicities  = -2.5, -1.5, -0.5, 0.0 and +0.5. For each ,  pair, three spectra at different surface gravities are given which roughly correspond to luminosity classes V, III, and I in the MKK classification system. The surface gravities are log g = 4.5, 2.0 and 1.0 for 3500 (K) 5 000, and log g = 4.5, 3.0 and 2.0 for the hotter models. Only the spectrum corresponding to = 3 500 K,  = -2.5 and log g = 4.5 is missing in this scheme (cf. Table 1 and Fig. 5), because we have not been able to obtain the constancy of flux with depth during the computation of this model. There are also a few models which have, in the upper layers, a percentage error that it too large in the flux derivative dH/d , indicating the failure of the condition of the constancy of the flux with depth. These models are marked with a slanted entry in Table 1. In some cases, the problem is due to the overcoming of the temperature limit T = 2 089 K of the opacity distribution function tables (ODFs) by the temperatures of the uppermost layers of the model. The corresponding synthetic spectra may be of lower quality.
As an example of the spectra which are electronically retrievable, the spectrum for the full 7650-8750 Å interval and = 5500 K,  = -0.5, log g = 3.0 parameters is presented in Fig. 1. Remaining Figs. 2-93 will focus on the GAIA 8500-8750 Å interval only. The effects caused by changing the temperature, the metallicity and the gravity are illustrated in Figs. 2, 3 and 4 respectively.
The synthetic spectra are based on the grids of model atmospheres computed by Castelli (1999) with an updated version of the ATLAS9 code (Kurucz 1993a). These models differ from Kurucz's models only for the convection. They are available at Kurucz's website (http://cfaku5.harvard.edu).
Our synthetic spectra follow the K step of the grid of available model atmospheres, but only a sub-set of the available surface gravities have been explored. Furthermore, only the microturbulent velocity = 2 km s-1 was considered.
Synthetic spectra were computed with the SYNTHE code of Kurucz (1993b) at a 500 000 resolving power. They were then degraded to the resolving power of the spectra in Paper I by adopting a Gaussian instrumental profile. No rotational velocity and macroturbulent velocity were assumed in our simulations.
In computing the spectra we adopted the solar abundances given by Pagel (1997), while the model atmospheres (which are based on Kurucz' ODFs), use the Anders & Grevesse (1989) solar abundances. Elements for which the abundances in model atmospheres and synthetic spectra differ are N (-0.08) O (-0.06), F (-0.04), S (+0.09), Ar (-0.06), K (-0.12), Sc (+0.08), Ti (+0.04), Fe (-0.15), Sr (+0.07), Zr (+0.10), La (-0.05), Ce (+0.03). The numbers in parenthesis are the logarithmic differences between the abundances from Anders & Grevesse and Pagel.
Kurucz (1995a) has been the source for the atomic lines, Kurucz (1993b) for the molecular lines of C2, CN, CO and hydrides (CH, NH, OH, MgH, SiH), and Kurucz (1999) for TiO data. More details about atomic and molecular lines can be found in Kurucz (1995b) and in the Kurucz website. Molecular lines were considered for the models with 6500 K. The TiO contribution is inessential at all temperatures for  = -2.5, while it becomes signficant at = 3750 K for  = -1.5, = 4000 K for  = -0.5, = 4250 K for  = 0.0, and = 4500 K for  = +0.5. The CN is the molecule with the strongest lines in the GAIA 8500-8750 Å region for the models with 4500 K.
To assess the usefulness of the computed spectra we compare three of them with real spectra in Figs. 90, 91 and 92. Figure 90 shows the comparison between the = 5750 K,  = 0.0, synthetic spectrum and the Sun observed spectrum taken from Kurucz et al. (1984) atlas (degraded to 20 000 resolving power). Figures 91 and 92 compare the observed spectra for G5 V and K5 V stars from Paper I with the closest synthetic spectra here computed.
In the comparison against the Sun, the larger wings for the Ca II lines are mostly related to the structure of the ATLAS9 models. When the Holweger-Müller (1974) solar empirical model is used the agreement with the observed spectrum improves considerably. The same effect was shown by Cayrel et al. (1996) for the solar Ca I triplet at 6102, 6122, and 6162 Å. About the comparison with stars other than our Sun, the reader should bear in mind that our spectra are not intended to match any star in particular (which will have non zero rotational and macroturbolent velocities, a microturbolent velocity probably different from the = 2 km s-1 here adopted, and a quite probable non-solar partition of the metal abundances). The present atlas instead is intended to explore the effects of changing the basic parameters describing the spectrum of a star in order to provide the simulations of GAIA observations with a realistic and complete set of input spectra.
The comparisons in Figs. 90-92 indicate that the synthetic spectra well represent the real spectra. They therefore constitute a useful input databank for simulations of GAIA observations as well as an aid to ground-observer spectroscopists working at moderately high resolving power ( ) in the near-IR region of the spectrum. The original computation at a 500 000 resolving power will allow us to re-map the whole grid of synthetic spectra at any resolution that the industrial designing of GAIA should require other than the currently baselined .
The interested reader can evaluate the effects of changing the microturbolent velocity by inspecting Fig. 93 where a zoomed portion of three spectra with the same = 5750 K, log g = 2.0 and  = 0.0 but different microturbolent velocities ( = 1, 2 and 4 km s-1) are compared. The comparison is intentionally carried out for a gravity typical of a supergiant because the effect of the microturbulent velocity increases with decreasing gravity.
The 254 spectra are presented in Figs. 5 to 89 (available only electronically) following the scheme in Table 1. Each figure is devoted to a triplet of spectra characterized by the same temperature and metallicity but with different surface gravities. This arrangement is quite similar to that of Paper I where the spectra were arranged into luminosity sequences at a given spectral type, and should facilitate the intercomparison between observed and synthetic spectra.
Only the 8490-8750 Å interval of interest to the baseline configuration
of GAIA is presented in the figures. The remaining 7650-8490 Å interval
is available only via the spectra in electronic form. The most relevant
lines in the 8490-8750 Å range are tabulated in Table 2.
Copyright The European Southern Observatory (ESO)