1 Introduction

Globular clusters (GCs) are (with a few significant exceptions, most notably the case of $\omega$ Cen) generally very homogenous as far as Fe-peak elements are considered; on the other hand, abundances of the lighter elements (from C to Al) show a complex, not yet adequately explained pattern. Rather wide variations in the strength of the CH and CN bands among stars on the red giant branch (RGB) were already noticed in the seventies (Osborn 1971, and many other references; for reviews see Smith 1987; Kraft 1994), and found to be correlated with other cluster properties (Norris 1987). More recently, a CN-CH anticorrelation was found in Main Sequence (MS) stars of NGC 6752 (Suntzeff & Smith 1991), 47 Tuc (Cannon et al. 1998), and M 71 (Cohen 1999a) from low dispersion spectra. Even more striking is the Na-O anticorrelation among GC red giants discovered by the Lick-Texas group (see e.g. Kraft et al. 1997).

Figure 1: Positions of program stars (large symbols) on the Strömgren colour-magnitude diagram of NGC 6397 (panel a)) and NGC 6752 (panel b)). Photometry is from Grundahl et al. (1999). Note that photometry for NGC 6397 has not been calibrated yet, so that only instrumental magnitudes are given

These features are indications of the presence of elements processed through the complete CNO-cycle in the atmospheres of some globular cluster stars; at temperatures where complete CNO cycle occurs, quite large amounts of Na may be produced by proton capture on ²²Ne nuclei (Denissenkov & Denissenkova 1990; Langer & Hoffman 1995; Cavallo et al. 1996). Whatever the cause of these abundance anomalies, it appears to be related to the dense environment of GCs. In fact, field halo stars present surface abundances of Li, C, N, O, Na that are well explained by standard models, once a second mixing episode (only affecting Li, C and N abundances) after the RGB-bump is allowed (Gratton et al. 2000). Note that the RGB bump plays an important rôle, because there the molecular weight barrier created by the deepest inward penetration of the outer convective envelope (at the base of the RGB) is canceled by the outward shift of the H-burning shell. The absence of this barrier in the subsequent evolution allows deep mixing caused by e.g. meridional circulations (Sweigart & Mengel 1979; Charbonnel 1994; Denissenkov & Tout 2000). Hence, both the O-Na anticorrelation observed among giants, and the C and N anomalies observed in stars in different evolutionary phases, are peculiar to GCs.

The proposed explanations of the abundance anomalies in globular cluster stars have been oscillating for years between an "in situ'' mechanism (very deep mixing in RGB stars, possibly activated by a core rotation that might be larger in cluster than in field stars) and an external source of material (either primordial proto-cluster gas or processed material from a polluting companion). Since MS stars do not have a convective envelope able to enhance N while depleting C, these anomalies should be attributed mainly to a primordial origin and/or to pollution, while evolutionary mixing could possibly be responsible only for likely minor effects in RGB stars. However, rather surprisingly, an analysis of stars below the Turn-off (TO) of M 13 (Cohen 1999b) shows no variations in the strength of CN and CH molecular features, likely excluding primordial variation of C and N abundance in this cluster, which instead presents the clearest Na-O anticorrelation on the RGB (Kraft et al. 1997), and a range in the strength of CN and CH bands among giants (Suntzeff 1981). While the result of Cohen is very surprising, the published analysis is only qualitative. Very recently, Dr. J. Cohen kindly provided us with a copy of her new quantitative analysis of these spectra in advance of publication: this analysis shows that large variations in the abundances of C and N may indeed be accomodated within the observed upper limit for star-to-star variations in the CN and CH bands in M 13 main-sequence stars. However, in general, the relation between the Na-O and C-N anticorrelation is not yet well established. Observation of a Na-O anticorrelation among main sequence stars of globular clusters would be crucial, clearly ruling out the deep mixing scenario.

Earlier explorative studies of stars close to the TO and at the base of the giant branch of a few clusters have been presented by King et al. (1998), Deliyannis et al. (1995) and Boesgaard et al. (1998) who used HIRES at Keck to gather spectra of three stars in M 92; and Pasquini & Molaro (1996,1997) who exploited the presence of closer clusters in the southern emisphere to obtain spectra of a few stars in NGC 6397 and 47 Tuc using EMMI at NTT (these stars are about a magnitude brighter than those of M 92). However, very low S/N observations of a handful of stars could be obtained. Only UVES (D'Odorico et al. 2000) at the Kueyen (VLT UT2) Telescope is able to provide abundances of O, Na, Mg, Al and other elements for significant samples of well studied stars at the TO and at the base of RGB (well below the RGB bump), crucial to clarify most of these issues. Here we present a short description of the first results of the ESO Large Program 165.L-0263 devoted to the acquisition and analysis of high dispersion spectra ( $R\;\lower.6ex\hbox{$\sim$ }\kern-7.75pt\raise.65ex\hbox{$>$ }\;40\,000$) of a fairly large sample of stars at the TO and at the base of the RGB in NGC 6752, NGC 6397, and 47 Tuc. A more detailed presentation of the analysis methods, and of related results for Li, will be given in forthcoming papers, still in preparation.