Our analysis is standard and essentially based on LTE model atmospheres. For each star we estimated effective temperature from through the colour-T calibration for giants of Alonso et al. (1999). We adopted log g = 2.5 for both stars, this value is compatible with the position of the stars in the colour-magnitude diagram, whichever the isochrone set considered. Furthermore we verified a posteriori that this gravity nearly satisfied the ionization equilibrium of both Fe I/Fe II and Ti I/Ti II. The atmospheric parameters are summarized in Table 1. The mean abundances for all the elements are given in Table 2.
Table 2. Mean abundances
With these parameters we computed model atmospheres using the ATLAS9 code (Kurucz 1993) using opacity distribution functions with microturbulence of 2 km s-1 and suitable metallicity.
We began by deriving the Fe abundance for both stars. We picked a set of lines which our preliminary synthetic spectra predicted to be substantially unblended, which had accurate laboratory or theoretical gf values and which spanned a range of line strengths. We excluded from the analysis lines stronger than to avoid an excessive dependence on microturbulence and lines weaker than to avoid lines which are too noisy.
For these lines we measured equivalent widths by fitting a gaussian with the iraf task splot and used these and the model atmosphere as input to the WIDTH9 code (Kurucz 1993). Some of the lines were removed from the analysis because they provided highly discrepant abundances. The microturbulence was determined by imposing that strong lines and weak lines give the same abundance. The results are given in Table 3. The excitation equilibrium for Fe I is very nearly satisfied for both stars (slopes are -0.05 dex/eV for star 139 0.13 dex/eV for star 143), we did not adopt an excitation temperature, so that the equilibria are better satisfied, since our lines cover a range of only about 2.7 eV, this means that in the worst case (star 143) the slope predicts a difference of slightly over 0.3 dex between the highest and lowest excitation lines, this is .
Table 3. Line data and abundances for Fe
As a check of our metallicity we derived Fe abundances also using MARCS models computed by Plez et al. (1992). The difference is not significative: for both stars the MARCS models provide an Fe I abundance which is 0.02 dex larger than that provided by the ATLAS models, while for Fe II it is 0.07 dex larger for star 139 and 0.05 dex larger for star 143.
Having fixed the metallicity and the microturbulence we proceeded to determine all the other abundances with the same method, again we disregarded the too strong and too weak lines, with a few exceptions, such as Sc, Cu and Ba, for which only one or two lines were available, in order to get information on as many elements as possible. The line data and abundances are given in Table 4. In addition for some blended lines we resorted to spectrum synthesis using the same model-atmosphere and the SYNTHE code (Kurucz 1993). We did not take into account hyperfine splitting (HFS) for Sc, V, Mn, Co and Cu, however given that the abundances of these elements are coherent with those of other elements we do not expect corrections due to HFS to be very large. For Eu we determined the abundance from the Eu II 664.5 nm line, taking into account HFS splitting, the relevant data is given in Table 5. The error on the Eu abundances estimated from the quality of the fit is 0.15 dex.
Table 4. Line data and abundances
Table 4. (continued)
Table 5. HFS data for Eu II
The main result of this analysis confirms the impression gathered by a direct comparison of the spectra of the two stars: the stars are very nearly identical, the few differences in their spectra are quite likely determined by slightly different T and log g, but not by chemical composition.
From the measure of the line centers of the unblended lines used for abundances we determined the radial velocity for the two stars. We obtain the following heliocentric radial velocities km s-1 from 60 lines for star 139 and km s-1 from 57 lines for star 143, the quoted error is just the rms. The measurement of the position of the atmospheric Na I D emission lines allowed to estimate the zero point shift to be less than 0.1 km s-1, this, coupled with the excellent reproducibility of the wavelength scale from night to night induced us to assume a null zero-point shift. These heliocentric radial velocities support membership to Sagittarius, Ibata et al. (1995) give a mean heliocentric radial velocity for Sgr of km s-1 and Ibata et al. (1997) find the intrinsic velocity dispersion to be km s-1 and constant across the face of the galaxy. The N-body model of Sgr computed by Helmi & White (2000) displays a similar velocity dispersion, if only the stars with 100 km s km s-1 are included, as done by Ibata et al. However if this condition is relaxed the velocity dispersion turns out to be much larger, due to the contribution of stars in the debris streams. Given the above considerations it is not surprising that we find a difference of 10 km s-1 between our stars. Our measured radial velocities compare quite well with those measured from our EMMI low resolution spectra (147 km s-1 for star 139 and 154 km s-1 for star 143, both are accurate to km s-1).
We cannot rule out the possibility that the stars studied here belong to the Bulge, rather than Sgr. If they were at a distance of 8.5 Kpc, rather than 25 Kpc their log g should be dex higher than what we assumed, but such a difference is within the errors of the analysis. However the radial velocity ought to be a very good discriminant. By looking at Fig. 1 of Ibata et al. (1995) we see that the distribution of radial velocity of Bulge stars in directions which do not intercept the Sgr dSph, shows a vanishingly small number of stars at the radial velocity of Sgr. Also the chemical composition suggests that the two stars do not belong to the Bulge: in fact Bulge stars are expected, theoretically, to have [/Fe], even at solar metallicities. Observationally the situation is not so clear, however our distinctly solar [/Fe] is a clue against Bulge membership.
© European Southern Observatory (ESO) 2000
Online publication: July 7, 2000