The metallicity of the two stars examined here is higher than all previous photometric estimates. Although it is possible that we happened to select two members of the high-metallicity tail of Sgr, this position is hardly tenable, the event of finding two such stars in a 9 square arcmin field must be quite rare. It is more likely that Sgr actually possesses a population, perhaps the main population, this metal-rich. For our two stars the Schlegel et al. (1998) maps provide . By comparison Marconi et al. (1998) used E(B-V)=0.18. The fact that the actual reddening could be 0.04 less than this could explain why the metallicity we find is 0.3 dex higher than the highest metallicity estimated by Marconi et al.
Quite obviously our results do not rule out the existence of a more metal-poor population. Preliminary results of abundance analysis in Sgr are given by Smecker-Hane & McWilliam (1999), who find two metal-poor Sgr member stars, with [Fe/H] and [Fe/H]. It is interesting that out of 11 stars analyzed by them 7 have metallicities in the range , two are metal-poor and two are metal rich ([Fe/H]). The 7 stars of intermediate metallicity, which should be analogous to the two under study here, show solar abundance ratios and no enhancement of elements, in agreement with our findings. Also the Na abundance displays a similar pattern: for all their stars, except the two metal-poor ones Na is over-deficient with respect to iron by 0.3-0.5 dex. Unfortunately, these results have not been published in a more detailed form and we lack information on the temperatures and luminosities of the stars considered by Smecker-Hane & McWilliam so we do not know if we are comparing similar giants.
It is also interesting to compare the present results with the abundances of the two Sgr planetary nebulae He 2-436 and Wray 16-423, studied by Walsh et al. (1998). The only element in common in the two analyses is O, for which Walsh et al. find [O/H] and [O/H] for He 2-436 and Wray 16-423, respectively. Our result for Sgr 143 is about 0.2 dex higher, but it is also more uncertain, because it is based on a single weak line, which is also very sensitive to gravity. An increase of gravity of 0.5 dex results in an increase of O abundance of 0.25 dex. O should be only marginally affected during AGB evolution, so that the O abundance in the PN ought to be quite close to that in the progenitor star. Walsh et al. (1998) argued that their abundances suggested a mild enhancement of O over Fe, because they assumed -0.8 to be the mean [Fe/H] of Sgr. Another scenario appears more likely, in view of our results: a solar O/Fe ratio, which suggests that the PNe have [Fe/H].
Having established that the two stars are quite similar in atmospheric parameters and abundances we must explain why their photometry is different and why the metallicity estimated from the low resolution spectra for star 143 is far lower than the one derived here. We consider 5 possibilities: 1) errors in V; 2) errors in ; 3) different reddening; 4) different age; 5) different distance. Let us examine all of these cases.
That a difference of 0.18 mag in V may be due to the photometric error may be discarded since this is a factor of ten larger than the photometric error of Marconi et al. (1998). An error in is more likely; a 0.03-0.04 mag error in would allow to slide sideways one of the two stars in the colour-magnitude diagram in such a way that both stars lie on the same isochrone, since the RGB, in this range of is very steep. The implied difference in is of K, the errors of our analysis.
Differential reddening seems unlikely for three reasons. The dust maps of Schlegel et al. (1998) give for both stars, suggesting that the reddening of the two stars is the same within 0.01 mag. The absence of detectable amounts of HI in Sgr (Burton & Lockman, 2000) also argues against a differential reddening. If the 0.18 mag difference in V were due to reddening it would imply a difference of almost 0.08 mag in , i.e. K in T. Although such a difference is within the errors of the present analysis and cannot be ruled out, it does seem somewhat unlikely, given the similarity of the two spectra.
An age difference of Gyr could be enough to explain the difference in the photometry of the two stars. A larger age spread would be necessary to explain the width of the RGB, like in the scenario proposed by Bellazzini et al. (1999). Although such a possibility is attractive, it appears somewhat contrived and it is not so clear that star formation may continue for several Gyrs without resulting in a spread in metallicity, as well as ages.
A distance difference of about 2 Kpc would be enough to explain the difference in V. This value is not unreasonable, Ibata et al. (1997), estimate the half-brightness depth of Sgr to be about 1.2 kpc. It is interesting that recent N-body simulations by Helmi & White (2000) support a considerable depth of Sgr: inspection of their Fig. 2 shows that the bulk of their model for Sgr has a depth of about 2 Kpc, however considering the debris shed during previous orbits, one has a sizeable population over a depth of 10 Kpc. Further support to the possibility that the two stars have a different distance comes from inspection of the Na I D lines (Fig. 2), three interstellar components belonging to our Galaxy are evident in both the spectra of Sgr 143 and of Sgr 139 at radial velocity +16.5 km s-1, km s-1 and km s-1; while the Na I D lines of star 143 appear symmetric and there is no hint of an interstellar component at the radial velocity of Sgr, the lines of star 139 show a weak but definite asymmetry, which we interpret as a weak interstellar line associated with Sgr. Star 139 is in fact the fainter of the two and hence the most distant, according to this interpretation, this would explain why the interstellar Na I D lines appear in its spectrum but not in the spectrum of star 143, which would then be in the side of Sgr nearer to us.
So of the five possibilities considered only the photometric error in V and the differential reddening are discarded. We may not decide which is the correct one with the present data, new accurate photometric measurements will allow to settle at least the issue of errors in . However we consider that the distance difference is the most likely explanation, because it is the simplest and is supported by several arguments. This suggests that the non-negligible line of sight depth of Sgr could explain at least a part of the width of the RGB of Sgr. Up to now all investigators have adopted a unique distance modulus for Sgr, in order to compare their photometry to fiducial ridge lines of Galactic clusters or to theoretical isochrones. This assumption may prove to be bit too naive.
A full discussion of the metallicity estimates from low resolution spectra shall be given elsewhere. Suffice to say here that the method of estimating abundances from low resolution needs a relatively high S/N ratio. In the case of star 139 the metallicity derived from high resolution analysis coincides with that estimated from low resolution to within the errors of the latter. We verified that the degraded UVES spectrum is very similar to the low resolution EMMI spectrum. The indices measured on this degraded spectrum yield in fact almost the same abundance provided by those measured on the low resolution spectrum. In the case of star 143 instead the method has been applied to a spectrum of too low signal to noise ratio, in this case the degraded UVES spectrum bears almost no resemblance to the low resolution spectrum, except for the strongest feature of the Mg I b triplet, which was enough to provide the correct radial velocity for this star.
The ratios of all elements are essentially solar, noticeable
exceptions are: Na which is overdeficient with respect to iron, and
the heaviest elements Ba, La, Ce, Nd, Eu, which appear over-abundant
while Y appears underabundant. Such anomalies are not readily
interpretable, deep mixing would produce an enhanced Na and low O and
Mg, at variance to what is observed. While it would be tempting to
interpret the overabundance of heavy elements as due to s-process
enrichment, the stars do not appear luminous enough to be on the
thermally pulsating asymptotic giant branch, where this mechanism is
operative. Moreover, s-process enrichement would produce also a high Y
abundance and no Eu (which is thought to be a "pure" r-process
element), at variance to the low Y and high Eu abundances observed
here. On the other hand, these stars could have been born in r-process
enhanced material (suggested by the Eu enhancement). Howevever this
seems also quite implausible since the r-process is thought to take
place in Type II supernovae which also produce large amounts of O and
other -elements, which are not
observed to be enhanced in our stars. This surprising pattern is
reminiscent to what is observed in the young supergiants in both
Magellanic Clouds, where the ratios of the moderate-mass s-process
elements Y and Zr to iron are essentially solar, whereas the heavier
species Ba to Eu are overabundant by ratios [X/Fe] of the order of 0.3
and 0.5 dex respectively in the LMC and SMC (Hill et al. 1995,
Hill 1997, Luck et al. 1998). In the Magellanic Clouds also, we are at
loss of an explanation for this behaviour (see discussion in Hill
1997). Note that our two Sgr giants have the same overall metallicity
as the LMC young population, and that the heavy elements
overabundances are also of the same order as in the LMC.
© European Southern Observatory (ESO) 2000
Online publication: July 7, 2000