3 The models

3.1 The old-ODF models

The old-ODF models and the old-ODF fluxes used in this paper are those computed by Castelli (1999) and adopted by KCC. They were computed with the ATLAS9 code by using ODFs from Kurucz (1990). The solar abundances adopted for the old-ODFs, the old-ODF models and the old-ODF fluxes are those from Anders & Grevesse (1989), except for iron when the $\alpha$-enhanced models are considered. For $\alpha$-enhanced ODFs, models, and fluxes, the iron solar abundance was assumed equal to $\log(N_{\rm Fe}/N_{\rm tot})=-4.53$, according to Holweger et al. (1995).

3.2 The new-ODF models

New ODFs were computed for metallicities covering the values of the sample, i.e. $\rm [M/H]=-1.0$, -1.0a, -1.5, -1.5a, -2.0, -2.0a, -2.5, -2.5a, and -3.0a. Furthermore, new-ODFs for $\rm [M/H]=-1.25$a, -1.75a, and -2.25a were obtained by interpolation. The symbol "a'' near the metallicity indicates an enhancement [ $\alpha/\alpha_{\odot}]=+0.4$ dex for the $\alpha$-elements O, Ne, Mg, Si, S, Ar, Ca, and Ti. For each metallicity, ODFs were computed for microturbulent velocities $\xi=0$, 1, 2, 4, and 8 km s^-1, in analogy with the old-ODFs computed by Kurucz (1990).

In the new-ODFs the Lyman-$\alpha$ H-H and H-H⁺ quasi-molecular absorptions near 1600 Å and 1400 Å are considered and they are computed according to Allard et al. (1998). The solar abundances adopted for the new-ODFs, the new-ODF models and new-ODF fluxes are those from Grevesse et al. (1996). Modifications in the treatment of the overlapping lines at the end of the term series have slightly changed the shape of the flux computed just shortward of the Balmer discontinuity. More details about the new-ODFs can be found in Castelli & Kurucz (2001).

Small grids of ATLAS9 models and fluxes were generated by using the new-ODFs in order to derive the stellar parameters from the fit of the IUE observations to the grids of synthetic fluxes. Also the final fitting model was directly computed with the ATLAS9 code. All the adopted models were computed with the option for the convection switched on, but with the option for the approximate overshooting switched off. The convection is treated with the mixing-length theory. The mixing-length to the pressure scale height ratio $L/H_{\rm p}$was assumed to be 1.25. The computed convective flux decreases with increasing $T_{\rm eff}$, so that it becomes either negligible or equal to zero in the hottest models considered in this paper.

Figures 1 and 2 compare fluxes $F_{\lambda }$ computed from old-ODF models and new-ODF models. Figure 1a shows fluxes computed for $\log\,g$=3.0, $\rm [M/H]=-1.50$a, and different $T_{\rm eff}$ equal to 9000 K, 8500 K, and 8000 K. Figure 1b shows fluxes computed for $T_{\rm eff}$=8000 K, $\rm [M/H]=-1.5$a, and different $\log\,g$ equal to 4.00 dex, 3.00 dex, and 2.00 dex. Figure 2 shows fluxes computed for $T_{\rm eff}$=8000 K, $\log\,g$=3.0, and different metallicities [M/H] equal to -1.00a, -1.50a, and -2.00a. All the models displayed in Figs. 1 and 2 are computed with ODFs corresponding to a microturbulent velocity $\xi =2$ km s^-1. The differences between the old-ODF fluxes and new-ODF fluxes shortward 1600 Å increase with decreasing $T_{\rm eff}$, increasing gravity, and decreasing metallicity. This behaviour is well manifest in the IUE spectra showed in Figs. A.1-A.15 of Appendix A, where the stars are ordered by decreasing $T_{\rm eff}$.