We present yields from stars of mass in the range M _ { \odot } \leq M \leq 8 M _ { \odot } of metallicities Z = 3 \times 10 ^ { -4 } and Z = 8 \times 10 ^ { -3 } , thus encompassing the chemistry of low– and high–Z Globular Clusters .
The yields are based on full evolutionary computations , following the evolution of the stars from the pre-Main Sequence through the Asymptotic Giant Branch phase , until the external envelope is lost .

Independently of metallicity , stars with M < 3 M _ { \odot } are dominated by Third Dredge–Up , thus ejecting into their surroundings gas enriched in carbon and nitrogen .
Conversely , Hot Bottom Burning is the main responsible for the modification of the surface chemistry of more massive stars , whose mass exceeds 3 M _ { \odot } : their gas shows traces of proton–capture nucleosynthesis .

The extent of Hot Bottom Burning turns out to be strongly dependent on metallicity .
Models with Z = 8 \times 10 ^ { -3 } achieve a modest depletion of oxygen , barely reaching -0.3 dex , and do not activate the Mg–Al chain .
Low–Z models with Z = 3 \times 10 ^ { -4 } achieve a strong nucleosynthesis at the bottom of the envelope , with a strong destruction of the surface oxygen and magnesium ; the most extreme chemistry is reached for models of mass \sim 6 M _ { \odot } , where \delta [ O/Fe ] \sim - 1.2 and \delta [ Mg/Fe ] \sim - 0.6 .
Sodium is found to be produced in modest quantities at these low Z ’ s , because the initial increase due to the combined effect of the second dredge–up and of ^ { 22 } Ne burning is compensated by the later destruction via proton capture .
A great increase by a factor \sim 10 in the aluminium content of the envelope is also expected .
These results can be used to understand the role played by intermediate mass stars in the self–enrichment scenario of globular clusters : the results from spectroscopic investigations of stars belonging to the second generation of clusters with different metallicity will be used as an indirect test of the reliability of the present yields .

The treatment of mass loss and convection are confirmed as the main uncertainties affecting the results obtained in the context of the modeling of the thermal pulses phase .
An indirect proof of this comes from the comparison with other investigations in the literature , based on a different prescription for the efficiency of convection in transporting energy and using a different recipe to determine the mass loss rate .