We have analysed high resolution and high signal-to-noise spectra of 21 mildly metal-poor stars ([Fe/H] ~ -1). The correlations between the relative abundances of 16 elements have been studied, with a special emphasis on the neutron-capture ones. This analysis reveals the existence of two sub-populations of field metal-poor stars which differ by the behaviour of the s-process elements versus the and r-process elements.

     We suggest a scenario for the formation of metal-poor stars, which closely relates the origin of these stars to the evolution of globular clusters. According to this scenario, thick disk and field halo stars were born in proto-globular clusters from which they escaped, either during an early disruption of the cluster or through a later disruption or an evaporation process.

1. Introduction
2. Observational data and abundance analysis
   1. Atmospheric parameters
   2. A strictly differential analysis
   3. A genuine cosmic scatter
3. Abundance correlations
4. The two-branches diagram and the EASE scenario
5. Figures

1. Introduction

     Traditional abundance analyses of metal-poor stars aim at determining abundance ratios of some chemical elements as a function of the overall metallicity, usually measured by the iron abundance [Fe/H]¹ . These trends are then compared to predictions from models of nucleosynthesis and chemical evolution of the Galaxy, in order to provide constraints on the sites and mechanisms for element synthesis.
However, many of these abundance ratios show rather considerable star-to-star scatter, so that they provide only weak constraints on the models.
With the improvement of observing and spectroscopic analysis techniques, it is now possible to reduce considerably the observational uncertainties in the abundance determinations. In a first step, this allows to decrease the scatter in the abundance ratios, but only down to a certain point, since there is a genuine cosmic scatter which can now be measured and analysed, provided the data are of sufficient quality.
With such high quality data, we can therefore investigate the cosmic scatter in the relative abundances at a given metallicity, and look for correlations between different elements. These correlations should give better constraints on the sites of formation of these elements, and on the nucleosynthetic mechanisms responsible for their formation.

     Here, we present the results of such an analysis (Jehin et al. 1999) for a sample of moderately metal-poor stars and we suggest a scenario explaining the observed trends.


¹ We adopt the usual spectroscopic notation: for elements A and B.

2. Observational data and abundance analysis

     We have selected a sample of 21 unevolved metal-poor stars with roughly solar temperatures and one tenth of the solar metallicity. This metallicity range is very interesting because it corresponds more or less to the transition between the halo and the disk.
The observations were carried out with the Coudé Echelle Spectrometer (CES) fed by the 1.4 m Coudé Auxilliary Telescope (CAT) at the European Southern Observatory (La Silla, Chile). The spectral resolution is of the order of 65 000 and the signal-to-noise ratio in the continuum of the reduced and coadded spectrum is at least 200 in the 4 spectral regions chosen for each star. About 100 lines belonging to 16 elements have been measured and a comparison of our EW's with those of Zhao and Magain (1991) indicates that our precision is better than 1 mÅ.

2.1. Atmospheric parameters

     The effective temperatures T_eff were determined from the Strömgren b-y and Johnson V-K colour indices, using the calibration of Magain (1987) which is based on the infrared flux method (Blackwell and Shallis, 1977). The agreement between the temperatures deduced from the two colour indices is very good (T_eff = 45 K ± 40 K) and indicates that the internal precision is around 20 K. The model metallicities were taken from previously published analyses. For the surface gravities we have used the Strömgren c₁ index, with the calibrations of VandenBerg and Bell (1985) for the adopted temperatures and metallicities. A comparison of this method for stars having Hipparcos parallaxes based gravities (Nissen et al. 1997) indicates that these photometric gravities are excellent (log g = -0.07 ± 0.11). Microturbulence velocities were obtained by forcing the FeI lines with different EW's to indicate the same abundance. The precision is around 0.1 km s^-1. The adopted model parameters for the 21 stars are listed in Table 2 of Jehin et al. (1999).

2.2. A strictly differential analysis

     In order to reduce the analysis uncertainties, the lines were chosen, whenever possible, to have similar dependences on the stellar atmospheric parameters. Moreover, as the stars have similar atmospheric parameters, the analysis was carried out differentially inside the sample (each star was compared to all other stars in the sample). The zero point was then fixed by analyzing one of the stars using oscillator strengths available in the literature (the precise value of this zero point is relatively unimportant as it does not affect the abundance correlations).

2.3. A genuine cosmic scatter

     Following the typical abundance analysis, we show in Fig. 1, the abundance of Ti relative to that of Fe, [Ti/Fe], plotted as a function of [Fe/H]. In the following figures, the subscripts I and II stand for neutral and ionized species and no specifications means that we have used a mean of the two species. We note a roughly constant overabundance of Ti relative to iron, with a 1  scatter amounting to 0.080 dex (20 %).
Now we address the following point: is this scatter real or is it due to observational and/or analysis uncertainties ? To answer this crucial question, we compare the values of [Ti/Fe] deduced from neutral lines with those deduced from lines of singly ionized species, as shown in Fig. 2. We can see a very nice correlation, with a scatter of 0.026 dex (6 %) only. As the neutral and ionized lines have different dependences on the stellar atmospheric parameters, the analysis and observational uncertainties should not exceed 6 %, and most of the scatter in the abundance of Ti relative to Fe represents therefore real cosmic scatter.
We can thus conclude that our data are of sufficient quality to investigate the cosmic scatter in the relative abundances of the chemical elements, and proceed in the analysis of these correlations.

3. Abundance correlations

     In Fig. 3 we show the abundance of Ca relative to Fe, as a function of [Ti/Fe]. We see immediately that the two elements Ca and Ti are closely correlated. The same is true for Mg. We conclude, and this is certainly no surprise, that the so-called -elements were synthesized by the same process in the same objects. This is in agreement with the accepted view that the -elements are mainly produced during supernova explosions of massive stars.
The abundance of Cr and Fe relative to Ti also show a remarkable correlation (Fig. 4), indicating a common origin for these two iron-peak elements. The same holds true for Ni (Fig. 5), with two exceptions: the stars HD193901 and HD194598 appear to be somewhat depleted in Ni. These two stars, which present other abundance pecularities, will be identified by open symbols in all subsequent figures.

     Recently we have also determined Na abundances for some of our stars and we confirm (circles in Fig. 6) the clear correlation between Ni and Na, already pointed out by Nissen and Schuster (1997) (open squares in Fig. 6). Moreover these abundances seem to be also correlated with the kinematics (Jehin and Bancken, these proceedings).

     The main purpose of this work was to study the behaviour of the neutron-capture elements, in order to identify the sites and mechanisms for the synthesis of these elements, in a relatively early phase of the galactic evolution.
In Fig. 7 the abundance of the prototypical r-process element Eu is compared to the Ti abundance. The correlation is almost perfect, except for the same two stars which show a Ni depletion. Now, they stand up as relatively enriched in Eu. The nice correlation allows us to conclude that, in general, the r-process elements are synthesized in the same objects as the -elements, i.e. most probably in the supernova explosion of massive stars, which confirms the generally accepted scenario.
A more complex situation appears when one examines the s-process elements (Fig. 8). In a first group of stars (Pop IIa) the relative abundance of Y increases slightly with the -elements abundance, until a maximum value for the -elements abundance is reached. In the second group of stars (Pop IIb), the -elements to iron ratio is constant and maximum while the [Y/Fe] shows a large range of enhancement. We find similar results when any of the light s-elements, Sr, Y and Zr is compared to any of the -elements. We called this behaviour the "two-branches diagram".

4. The two-branches diagram and the EASE scenario

     This peculiar and well defined behaviour, which should be related to nucleosynthesis processes, has led us to distinguish between two sub-populations of metal-poor stars, namely Pop IIa and Pop IIb. The first interpretation which comes to mind is to relate one of these populations to the most metal-rich stars of the halo and the other to the most metal-poor stars of the disk.
In Fig. 9 we have added to our data (black squares), and after zero-point corrections, the stars from Zhao and Magain (1991) (full pentagons), Nissen and Schuster (1997) (full triangles), and Edvardsson et al. (1993) (full circles for stars with [Fe/H] < -0.6 and open circles for the more metal rich ones). All the stars with [Fe/H] < -0.6 fall on the two-branches diagram, but it is not true for the more metal rich stars, which have disk kinematics. On the other hand, no obvious distinction in term of metallicity or kinematics was found between the two branches.
This leads us to propose an alternative interpretation linking field metal-poor stars to globular clusters (GCs). We propose that thick disk and halo stars were born in GCs or proto-GCs, from which they escaped, either during an early disruption of the cluster (Pop IIa) or later through an evaporation process (Pop IIb). We distinguish between two distinct phases of chemical enrichment in GCs : a first one consisting in supernovae explosions of massive stars which could disrupt the cluster and account for the Pop IIa (G. Parmentier et al., these proceedings) and, later, a second phase, to account for Pop IIb, where the stars atmospheres are enriched in s-process elements by the accretion of the products expelled by AGB stars onto the intra-cluster medium (A. Thoul et al., these proceedings).

5. Figures

Figure 1	Figure 2		Figure 3
Figure 4	Figure 5	Figure 6	Figure 7
Figure 8	Figure 9		Figure 10

Acknowledgements

     This work has been supported by contracts ARC 94/99-178 "Action de Recherche Concertée de la Communauté Française de Belgique", the Pôle d'Attraction Interuniversitaire P4/05 (SSTC, Belgium) and by FRFC F6/15-OL-F63 (FNRS, Belgium).

References

Blackwell D.E., Shallis M.J. 1977, MNRAS 180, 177
Edvardsson B., Andersen J., Gustafsson B., Lambert D.L., Nissen P.E., Tomkin J. 1993, A&A, 275, 101
Jehin E., Magain P., Neuforge C., Noels A., Parmentier G., Thoul A. 1999, A\&A 341, 241
Magain P. 1987, A&A 181, 323
Nissen P.E., Høg E., Schuster W.J. 1997, in Hipparcos Venice '97, ESA SP-402, 225
Nissen P.E., Schuster W.J. 1997, A&A 326, 751
VandenBerg D.A., Bell R.A. 1985, ApJSS 58, 561
Zhao G., Magain P. 1991, A&A 244,425

---