9 Conclusions

We derived the parameters $T_{\rm eff}$ and $\log\,g$ from the whole IUE energy distribution for 29 out of the 31 metal-poor A-type stars studied by Kinman et al. (2000) (KCC). We used both the same models adopted by KCC and new models computed with ODFs which take into account the H-H and H-H⁺ semi-molecular absorptions at 1400 Å and 1600 Å, whose intensity strongly depends on $T_{\rm eff}$ and $\log\,g$.

For most stars, the new-ODF models and the old-ODF models lead pratically to the same values of $T_{\rm eff}$ and $\log\,g$, when the fit is performed in the UV range 1200-3300 Å; the difference is within the uncertainty of the fit of the observations to the models, namely 50 K in $T_{\rm eff}$ and 0.1 dex in $\log\,g$. Larger differences in $\log\,g$ on the order of 0.2 dex or more have been found only for few stars. However, a detailed inspection of the figures in Appendix A points out the better fit provided by the new-ODF models, especially for the coolest stars in the sample, i.e. those from HD 8376 (Fig. A.11) to HD 202759 (Fig. A.15). For all the stars, the computed energy distribution shortward of 1600 Å is no longer brighter than the models, so that the shortcoming pointed out by Huenemorder et al. (1984) has been completely overcome by the new models.

HD 130201 is the only star in the sample with both parameters from the whole ultraviolet energy distribution in clear disagreement with those from KCC ($\Delta$ $T_{\rm eff}$=250 K, $\Delta$$\log\,g$=1.0). For all the other stars hotter than about 8700 K the parameters from KCC agree with those from the whole IUE spectra within 150 K in $T_{\rm eff}$ and 0.2 dex in $\log\,g$. We did not observe any trend in $\Delta$ $T_{\rm eff}$ or $\Delta$$\log\,g$ as a function of $T_{\rm eff}$ for these stars. Also for stars cooler than about 8700 K the differences in $T_{\rm eff}$ are not larger than 150 K (except for HD 60788 and HD 78913, with $\Delta$ $T_{\rm eff}$=200 K), whereas the gravities derived from fitting the UV energy distributions to the models are systematically smaller than the gravities obtained from photometric and spectrophotometric data in the visual range. The average difference is about 0.3 dex. This discrepancy in $\log\,g$ is insensitive to reddening, microturbulent velocity, metallicity, and mixing-length parameter for the treatment of the convection. The comparison of the parameters derived in this paper with those derived from two independent data sets and analyses in the visual range (Wilhelm et al. 1999; Gray et al. 1996) does not allow to decide whether the problem resides in the UV or in the visual range, since our parameters from UV agree with Wilhelm et al.'s (1999) and do not agree with Gray et al.'s (1996).

On the other hand, a comparison with ZAHB models (based on Sweigart & Catelan 1998) at metallicity Z=0.002 and various values of primordial helium content indicates that the gravities derived from UV data are too low for stars cooler than about 9000 K.

We conclude that fitting the most recent model atmospheres to IUE ultraviolet energy distributions yields reliable values of $T_{\rm eff}$ and $\log\,g$ for HB A-type stars hotter than about 8700 K, whereas for cooler stars only $T_{\rm eff}$ is acceptable. Further investigations are needed in order to understand why the present model atmospheres yield such discrepant results on gravities in the UV and visual ranges at $T_{\rm eff}$ lower than about 8700 K. A possibility is that the classical LTE models are inadequate to represent the atmospheres of the cooler HB A-type stars. For instance, both NLTE and convection can play an important role in the modelisation of the metal-poor, low gravity A-type stars investigated in this paper. Also, missing lines in the ultraviolet, as well as different elemental abundances from those adopted by us (with the exception of Mg, Ti, and Fe) could be the cause for the inconsistency in the gravities. New spectrophotometric observations in the visible and high-resolution spectroscopic observations, mostly in the UV, would be very useful in order to better define possible deficiencies of the models for the HB A-type stars.

Acknowledgements

We wish to thank T. Kinman and M. Catelan for useful comments on the manuscript, and M. Catelan for providing us with the ZAHB unpublished models discussed by Sweigart & Catelan (1998).