3. Observational determination of the stellar fundamental parameters
We study an homogeneous sample of 114 late-type nearby stars of spectral types in the range from F to late K, carefully selected by M.-N. Perrin. They are part of the proposal 132 (M.-N. Perrin), accepted as an Hipparcos program in 1982. Later on, the proposal was updated as the INCA proposal 011 by A. Baglin, M.-N. Perrin, Y. Lebreton and R. and G. Cayrel. These stars are closer than about 25 parsecs which ensures an excellent accuracy of their parallax determination by Hipparcos.
Among these, we have retained the stars which have been submitted to detailed spectroscopic analysis from the ground and which appear in the last version of the Catalogue of [Fe/H] determinations: 1996 edition by G. Cayrel de Strobel et al. (1997). Their metal to hydrogen ratio [Fe/H], i.e. the logarithm of the iron to hydrogen ratio (by number of atoms) relative to the solar value, ranges from about -1.0 to about +0.3 dex corresponding to Population I and thick-disk population. For each star we assign an averaged abundance determined from the above catalogue (for more details see the caption of Table 1). The mean internal error on [Fe/H] is of the order of 0.1 dex. This does not include systematic errors, as the fact that all abundances in the catalogue are not corrected for NLTE effects. In determinations based on spectra taken with solid state detectors, with high quantum efficiency, the random error coming from equivalent width measurements enters for about 0.03 dex only, the remaining part coming from uncertainties in the fundamental parameters and .
Table 1. Observational parameters for Sample 1. Parallax , and relative error on it come from the Hipparcos main catalogue. The apparent magnitude V is from the Hipparcos Input Catalogue. and were derived from the bolometric fluxes of Alonso et al. (1995, 1996a).Note that the zero-point of bolometric magnitudes is not that used by Alonso, but is such that =4.75 for the Sun. [Fe/H] values are a weighted average of individual values from spectroscopic analyses taken in Cayrel de Strobel et al. (1997). A quality index "qlt" was attributed to the adopted value of [Fe/H] according to the following code: 4: average of at least six determinations obtained with recent high S/N spectra
We eliminated the suspected unresolved binaries and only kept stars with parallaxes determined with an accuracy better than 5 per cent. Then, among the remaining stars, we extracted three homogeneous and independent subsamples.
Sample 1 is constituted of 33 stars with directly determined bolometric fluxes, so the problem of the value of the bolometric correction is eliminated. The relevant data for Sample 1 are listed in Table 1. The fluxes were derived by Alonso et al. (1995) with an accuracy of about 2 per cent by integrating UBVRIJHK photometry. The average absolute error on bolometric magnitude resulting from the parallax and bolometric flux uncertainty is of about 0.03 to 0.06 magnitude. The effective temperature was obtained by Alonso et al. (1996a) from the bolometric flux by the Infra-Red Flux Method (IRFM) (Blackwell & Shallis 1977). The method makes use of a grid of theoretical model line-blanketed flux distributions (Kurucz 1991). The mean internal error on effective temperature is about 80 K for temperatures greater than 4200 K (Alonso et al. 1996a). Fig. 1 shows the resulting positions of the stars in the (log , ) diagram with their individual error bars. The striking feature of Fig. 1 is the lack of clear separation between stars having a solar metallicity, and those in the metallicity range [-1.0;-0.5], confirming with high quality data the result already obtained by Perrin et al. (1977), on less accurate data. In Fig. 2 the sample of stars plotted in Fig. 1 is split into 2 subsamples corresponding 1) to stars of solar metallicity, plotted in Fig. 2a and 2) to moderately deficient stars, plotted in Fig. 2b. Theoretical isochrones are also plotted in Fig. 2 but this will be discussed later in Sect. 5.
Sample 2 is constituted of 64 stars with effective temperatures derived from detailed spectroscopic analyses. For each star at least one determination of the effective temperature and of [Fe/H] was made with modern detectors (CCD or Reticon). We adopt a mean internal uncertainty on effective temperature of 150 K. Bolometric magnitudes were obtained using bolometric corrections from Alonso et al. (1996a, 1996b), with a correction of the zero point giving a bolometric magnitude of 4.75 to the Sun. The resulting positions of the stars in the (log , ) diagram are plotted in Fig. 3 and will be discussed later in Sect. 5.
Sample 3 is constituted of 15 stars with effective temperatures derived from the () color index and from [Fe/H] according to the results from Carney et al. (1994), Johnson et al. (1966), Koorneef (1983). The mean internal uncertainty on effective temperature is of about 75 K. The bolometric magnitudes are derived from the V magnitude given in the Hipparcos input catalogue, using bolometric corrections from Alonso et al. (1996a, 1996b), with the same zero-point. Fig. 4 shows the resulting positions of the stars in the (log , ) diagram with their error bars and will be discussed in Sect. 5. Very clearly Fig. 3 has a much larger scatter than Figs. 1, 2 and 4, showing that the effective temperatures derived from detailed analyses are not tightly connected to the true effective temperatures. The effective temperatures derived from on the contrary are tightly bound to those obtained by the IRFM, as it can be directly checked in the full Alonso sample. This is not surprising, as effective temperatures from detailed analyses are affected by other parameters: gravity, departures from LTE in ionisation equilibria, and a variety of different techniques in fixing the triad (, log g and [Fe/H]).
© European Southern Observatory (ESO) 1999
Online publication: October 4, 1999