Considering these different properties, we suggest that they do not belong to the thin disk. Firstly, they are older (10-18Gyr) than other stars. Secondly, if they are thin disk stars, it is hard to understand why stars coming from both sides of the solar annulus have lower metallicity than the local region. Thirdly, these stars show a relatively small metallicity dispersion at such early Galactic time, i.e. smaller than stars at 8-10 Gyr. This is not in agreement with the effect of orbital diffusion working during the evolution of the thin disk, which suggests larger metallicity dispersion for older stars. Finally, the $\mbox{$W_{\rm LSR}$ }$ dispersion of these stars is about 40 $\mbox{\rm\,km\,s$^{-1}$ }$ , considerably larger than the typical value of about 20 $\mbox{\rm\,km\,s$^{-1}$ }$ for thin disk stars. Consistently, the kinematics, age, metallicity and abundance ratios of these stars follow the features of the thick disk: $\mbox{$V_{\rm LSR}$ }\lesssim -50 \mbox{\rm\,km\,s$^{-1}$ }$ , $\sigma(\mbox{$W_{\rm LSR}$ }) \simeq 40 \mbox{\rm\,km\,s$^{-1}$ }$ , $\tau > 10$ Gyr, $\mbox{\rm [Fe/H]}< -0.5$ and [ $\alpha$ /Fe] $\sim 0.2$ . We conclude that these oldest stars in both EAGLNT and this work are thick disk stars. Hence, they are probably not resulting from an inside-out formation of the Galactic disk, but have been formed in connection with a merger of satellite components with the Galaxy.

$\begin{figure}\resizebox{\hsize}{!}{\includegraphics{ds1800f10.eps}}\end{figure}$

Figure 10: The mean $\alpha$ (Mg, Si, Ca and Ti) abundance as a function of metallicity. The symbols are the same as in Fig. 7 and their size is proportional to stellar age

Concerning the abundance connection of the thick disk with the thin disk, our data for [ $\alpha$ /Fe], shown in Fig. 10, suggest a more smooth trend than those of EAGLNT, who found a correlation between [ $\alpha$ /Fe] and $\mbox{$R_{\rm m}$ }$ at $\mbox{\rm [Fe/H]} \sim -0.7$ . We leave the issue open considering the small number of these stars in our work. Two stars marked by their names in Fig. 10 may be particularly interesting because they show significantly higher [ $\alpha$ /Fe] than other stars. Fuhrmann & Bernkopf ([1999]) suggest that one of them, HD106516A, is a thick-disk field blue straggler. It is unclear if this can explain the higher [ $\alpha$ /Fe]. HD97916 is a nitrogen rich binary (Beveridge & Sneden [1994]) with $\mbox{$U_{\rm LSR}$ }=-117 \mbox{\rm\,km\,s$^{-1}$ }$ and $\mbox{$W_{\rm LSR}$ }=101 \mbox{\rm\,km\,s$^{-1}$ }$ (typical for halo stars), but with $\mbox{$V_{\rm LSR}$ }=22 \mbox{\rm\,km\,s$^{-1}$ }$ similar to the value for thin disk stars. Surprisingly, this star is also very young (5.5 Gyr) for it's metallicity.

It is interesting to re-inspect the observational results for thin disk stars excluding the thick disk stars. More direct information on the evolution of the Galactic thin disk will be then obtained. In summary, the thin disk is younger (not older than 12 Gyr), more metal-rich ( $\mbox{\rm [Fe/H]}>-0.8$ ) and has a smaller [ $\alpha$ /Fe] spread (0.1 dex) without the mixture of the thick disk stars. In particular, the AMR is more weak and there seems to exist a radial metallicity gradient. All these features agree better with the present evolutionary models for the Galactic disk.

We emphasize here that there is no obvious gradient in [ $\alpha$ /Fe] for the thin disk at a given metallicity. Such a gradient was suggested by EAGLNT based on higher [ $\alpha$ /Fe] of the oldest stars with $\mbox{$R_{\rm m}$ }< 7$ kpc than stars with $\mbox{$R_{\rm m}$ }> 7$ kpc (see their Fig. 21). After we have ascribed these oldest stars to the thick disk, the abundance gradient disappears.

Our study of relative abundance ratios as a function of $\mbox{\rm [Fe/H]}$ suggests that there are subtle differences of origin and enrichment history both within the group of $\alpha$ elements and the iron-peak elements. Nucleosynthesis theory predicts that Si and Ca are partly synthesized in SNe Ia, while O and Mg are only produced in SNe II (Tsujimoto et al. [1995]). Our data suggest, however, that SNe Ia may also be a significant synthesis site of Mg, because [Mg/Fe] shows a trend more similar to [Si/Fe] and [Ca/Fe] than to [O/Fe]. Ti may not lie in a smooth extension of Si and Ca, because there is a hint of a decrease of [Ti/Fe] for $\mbox{\rm [Fe/H]}> -0.4$ not seen in the case of Si and Ca. The situation for the odd-Z elements is more complicated. The available data for Na and Al show confusing disagreements; EAGLNT finds an overabundance of 0.1 to 0.2 dex for [Na/Fe] and [Al/Fe] among the metal-poor disk stars, whereas our study points at solar ratios. Two other odd-Z elements, K and Sc (Nissen et al. [2000]), behave like $\alpha$ elements, but the result for K is sensitive to the assumption of LTE. The iron-peak elements also show different behaviours: V, Cr and Ni follow Fe very well, while [Mn/Fe] (Nissen et al. [2000]) decreases with decreasing metallicity from [Mn/Fe] $\simeq 0.0$ at $\mbox{\rm [Fe/H]}=0.0$ to [Mn/Fe] $\simeq -0.4$ at $\mbox{\rm [Fe/H]}= -1.0$ . We conclude that the terms " $\alpha$ elements'' and "iron-peak elements'' do not indicate productions in single processes, and that each element seems to have a unique enrichment history.

8 Concluding remarks