---

WHY A SELF-ENRICHMENT SCENARIO IS NECESSARY FOR THE CHEMISTRY OF GC STARS?

---

Bimodal distribution of CN band strength in the giants of several GCs (refs. in Kraft 1994, Cannon et al. 1998). The CN vs CH anticorrelation is found also at the main sequence or on the subgiant branch stars where it can not be attributed to mixing. [Already for NGC 6752 -Suntzeff 1989 and 47 Tuc (Bell et al. 1983, Briley et al. 1991), M92 (Carbon et al. 1982)].

The O - Na anticorrelation found for the RGB stars (Kraft et al. 1993) has now been found also among the turnoff and subgiant stars in NGC 6752 (Gratton et al. 2000).

Bimodality can be due to the merging of two clouds of gas , whose collision triggers the GC stars formation, but why there is a great uniformity of `metal' abundance , and this is true at very different Z?

Some clusters present CNO abundance spreads and not a bimodality (e.g. the subgiants of M5 -Cohen et al. 2002), whose complexity is difficult to be attributed to primordial variations.

The HB morphology is correlated to other chemical anomalies (e.g. Oxygen is lower for Blue HB clusters; compare M13 vs. M3 (Catelan and de Freitas Pacheco 1995). What is the `structural meaning' if it is not related to deep mixing on the RGB?

next page figures