next up previous
Next: References Up: On-line determination of stellar Previous: Testing TGMET

Conclusion

The feasibility of a quick determination, at the telescope, of the fundamental atmospheric parameters effective temperature, gravity and metallicity of a star from its spectrum has been demonstrated. This has been done using spectra supplied by the cross-dispersed echelle spectrograph ELODIE, mounted at the coudé focus of the 1.93 m telecope of the Observatoire de Haute-Provence. These spectra have a spectral resolution of 42 000 and cover a large spectral range from 380 to 680 nm. We have been greatly helped by the already installed ELODIE reduction software, supplying on-line the radial velocities and the extracted spectra of the stars.

The method consists in comparing the spectrum of the target star with a library of spectra taken with the same spectrograph, at a typical S/N of 100. Such a library has been assembled , and presently contains the spectra of 211 stars, selected for covering the parameter values encountered in stars belonging to the halo, the thick disk and the old thin disk of the Galaxy. Namely the parameter intervals are [4000K, 6300K], [0.6, 4.7] and [-2.9., 0.35] respectively in $T_\mathrm{eff}$, $\log g$and [Fe/H].

A great amount of care has been put into the elimination of all features not intrinsic to the stars before attempting any comparison between their spectra. In particular we have removed the modulation of the spectra by the blaze of the orders (which still exists even if all spectra are taken with the same spectrograph because of the radial velocity shifts), the telluric lines and the hits by cosmic rays. It was also necessary to bring all spectra to a common resolution to avoid differences due to stellar rotation or macroturbulence or to a slight difference in focus during different observing runs. The degree of resemblance between the spectra of the target and those of the library was quantified by a maximum likelihood approach, involving a weighted least-square fit. The parameters of the target star are determined by taking a weighted mean of the the parameters of the most resembling reference stars.

The accuracy of the method, with the existing library, has been tested. On the average, the accuracy turns out to be 86 K in $T_\mathrm{eff}$, 0.28 dex in $\log g$, and 0.16 dex in [Fe/H] for a target spectrum with S/N = 100. This accuracy critically depends upon the number of spectra in the library and the quality of the parameters collected for each reference object in the literature. A second paper ( Soubiran et al. 1998) deals with this point in more detail, probing the consistency of the library by internal comparison of all the spectra of the library.

Our fundamental stellar parameter extraction software (named TGMET for Temperature, Gravity, METallicity) is available to all ELODIE users at Observatoire de Haute-Provence. It is online, so that these parameters may be derived during observing runs. Its user-guide is also available on the WEB at http://www.obs-hp.fr.

A similar software is planned for the new spectrograph FEROS at La Silla (ESO).

Future improvements are planned and already under way. The most obvious ones are to fill the sparsely populated zones of the library, and in redetermining the fundamental parameters of the stars which have been spotted as doubtful in the consistency check. Another one is to extend the library of reference stars up to F0, at the request of other observers. In particular the asteroseismologists working on the preselection of target stars for the spatial experiment COROT with ELODIE need atmospheric parameters for stars from A to G type. We plan also to improve the algorithm determining the parameters, by taking more rigorously into account the sensitivity of each pixel of the spectrum to a variation of the parameters. This will be particularily useful for very metal-poor stars, for which only very few places in the spectrum contain useful information. An exciting extension to our software would be to derive abundance ratios in addition to metallicity. For example we could try to determine the alpha-elements to iron ratio, which has been shown to present a scatter larger than believed before (King 1997, Carney et al. 1997, Nissen & Schuster 1997). This ratio is particularly important, as the Mg I green triplet and the molecular band MgH are dominant gravity features. Other abundance ratios of interest as [Ba/Fe], [Sr/Fe] or [Na/Fe] are attractive targets.


\begin{acknowledgements}
We are very grateful to Didier Queloz for providing his...
 ...ful comments and suggestions for improving the manuscript.\end{acknowledgements}


next up previous
Next: References Up: On-line determination of stellar Previous: Testing TGMET

9/11/1998