6 Summary and conclusions

We have presented the first results of the ESO Large Program 165.L-0263: we have used the UVES spectrograph on VLT2 to obtain high resolution ( $R\;\lower.6ex\hbox{$\sim$ }\kern-7.75pt\raise.65ex\hbox{$>$ }\;40\,000$) spectra for a quite large number of stars at the turn-off (14 stars between the two GCs) and the base of the giant branch (12 stars) in the globular clusters NGC 6397 and NGC 6752. Thanks to the efficiency and wide spectral coverage of UVES we were able to obtain reliable EWs for a number of lines of Fe, Li, O, Na, and other elements. The main results of this first analysis are:

(i)

The [Fe/H] value for NGC 6397 is ${\rm [Fe/H]}=-2.03\pm 0.02\pm 0.04$ (internal and systematic errors within the abundance scale defined by observations of the field stars); less than that derived from giants by Carretta & Gratton (1997) and Zinn & West (1984), but in agreement with the recent analysis of giants and subgiants by Castilho et al. (2000). The [Fe/H] value for NGC 6752 ( ${\rm [Fe/H]}=-1.42\pm 0.02\pm 0.04$) coincides with the value derived from giants by Carretta & Gratton (1997), and is slightly higher than the value quoted by Zinn & West (1984);

(ii)

In both clusters, [Fe/H] obtained for TO stars agrees perfectly (within a few percent) with that obtained for stars at the base of the giant branch; this is a constraint on the impact of diffusion in stellar models. The star-to-star scatter is extremely small: the rms scatter is only 0.04 dex (i.e. 10%) for NGC 6397. This seems a very homogeneous cluster as far as Fe abundances are considered;

(iii)

For NGC 6397, the O abundance is ${\rm [O/Fe]}=+0.21\pm 0.05$. This is a quite low value in comparison with those usually found in metal-poor stars, but it agrees well with previous determinations for the red giants of this cluster. The scatter of individual values is small, and none of the program stars seem to be oxygen-poor;

(iv)

For NGC 6752, there is a clear anticorrelation between Na and O among TO stars and subgiants (similar to that seen among giants in this and other clusters). Also, an anticorrelation is observed between Mg and Al, most clearly among subgiants, but likely existing also among TO-stars. Na-poor stars (i.e. stars with ${\rm [Na/Fe]}<0.2$) in NGC 6752 have ${\rm [O/Fe]}=+0.30\pm 0.04$ (7 stars, ${\rm rms}=0.10$ dex). Extremes in Al abundances differ by over 1 dex, while for Mg the star-to-star scatter is smaller. The large variations in Mg and Al abundances suggests that nearly half of Mg has been converted into Al in the most Al-rich stars (1481 and 1460); note that these stars are also the most Na-rich ones. Note that given the adopted sample selection criteria, extreme cases of Na-poor and Na-rich stars may be overrepresented among observed stars.

We think the present results are very difficult to reconcile with deep mixing scenarios. To our knowledge, there are not appropriate calculations for main sequence stars. For giants, some calculations have been made by Langer et al. (1993) and Denissenkov & Tout (2000): they show that the temperature required for p-capture on ²⁴Mg to finally produce the Mg-Al anticorrelation is $\sim$6 10⁷ K; even if Al is produced starting from the far less abundant ²⁵Mg isotope, the temperature required is $\;\lower.6ex\hbox{$\sim$ }\kern-7.75pt\raise.65ex\hbox{$>$ }\;4~ 10^7$ K; finally, the temperature required for extensive O-burning and Na production by p-capture on ²²Ne is $\sim 3\ 10^7$. All these values are much higher than expected even at the center of a globular cluster TO-star ($\sim$2  10⁷ K). Admittedly, these computations assume densities about 20 times lower than expected at the center of TO-stars, and much shorter timescales than MS lifetimes. However complete mixing of MS stars is unacceptable for several other reasons (e.g. it would bring large amount of fresh H to the center); furthermore, Li would be completely destroyed (while we see some Li even in O-depleted stars; paper in preparation). Hence, we think deep mixing scenarios cannot explain our results.

We are then forced to some primordial mechanism, like those proposed years ago by Cottrell & Da Costa (1981) and D'Antona et al. (1983). In both scenarios, the inhomogeneities are due to the mass lost by intermediate mass stars ( $M = 4 - 5~M_\odot$) during the Asymptotic Giant Branch (AGB) evolution and the planetary nebula expulsion: the two scenarios differ because Cottrell & Da Costa think of a prolonged star formation, with most recently formed stars having a different chemical composition from the first ones; while D'Antona et al. consider pollution of the outer layers of already formed stars by other objects in the cluster. Anyhow, in both scenarios the intracluster gas is heavily nuclearly processed, due both to the occurrence of the third dredge-up from the helium buffer (Iben 1975) and by Hot Bottom Burning (HBB) at the basis of the convective envelopes of these massive AGB stars. Models for these evolutionary phases (Sackmann & Boothroyd 1992; Ventura et al. 2000) successfully explain the occurrence and evolution of the lithium rich, oxygen rich massive AGBs in the Magellanic Clouds (Smith et al. 1995). Very recent models by Ventura et al. (2001) show that, in full stellar models computed for these intermediate mass stars at the low metallicities of Globular Clusters, the HBB temperature can reach values as large as 10⁸ K. At this temperature, the complete CNO cycle operates, depleting oxygen. At the same time, p-captures on ²⁴Mg and ²⁰Ne produce Al (see also Denissenkov et al. 1998) and Na. This model can then explain the anticorrelations O-Na and Mg-Al. In this context it is very interesting to note that no O-Na anticorrelation is seen in NGC 6397. Also, Li abundances in this cluster follows the paradigma set by field metal-poor stars (Castilho et al. 2000). NGC 6397 is a quite small cluster (mass $\;\lower.6ex\hbox{$\sim$ }\kern-7.75pt\raise.65ex\hbox{$<$ }\;10^5~M_\odot$, from the integrated magnitude M_V=-6.58, Harris 1996, and a mass-to-light ratio of $\sim$2, typical for a globular cluster). On the other side, the O-Na anticorrelation is seen in the more massive cluster NGC 6752 (mass $\sim$ $2\,10^5~M_\odot$, from the integrated magnitude M_V=-7.68, Harris 1996, and the same mass-to-light ratio used for NGC 6397). A (cluster) mass threshold should be present in accretion scenarios (see e.g. Gratton 2001), and likely also in the prolonged star formation models. We plan to address thoroughly such problems in forthcoming papers.

Acknowledgements

This research has made use of the SIMBAD data base, operated at CDS, Strasbourg, France. We wish to thank V. Hill for help during the observations, and P. Bertelli for useful comments. We thank our referee (J. Cohen) for having provided very useful suggestions and data in advance of publication