A new scenario --early disc accretion-- has been recently proposed to explain the discovery of multiple stellar populations in Galactic globular clusters. According to this model, the existence of well defined (anti)-correlations amongst light elemen
t abundances (i.e. C, N, O, Na) in the photospheres of stars belonging to the same cluster (and the associated helium enrichment), is caused by accretion of the ejecta of short lived interacting massive binary systems (and single fast rotating massive stars) on fully convective pre-main sequence low- and very low-mass stars, during the early stages of the cluster evolution. We investigated the constraints provided by considering simultaneously the observed spread of lithium and oxygen (and when possible also sodium) abundances for samples of turn-off stars in NGC6752, NGC6121 (M4), and NGC104 (47Tuc), and the helium abundance of their multiple main sequences. These observations provide a very powerful test for the accretion scenario, because the observed O, Li and He abundance distributions at the turn off can be used to constrain the composition (and mass) of the accreted matter, and the timescales of the polluting stars. In case of NGC6752 we could not find a physically consistent solution. In case of M4, spectroscopic errors are too large compared to the intrinsic spread, to constrain the properties of the accreted matter. As for 47Tuc, we could find a physically consistent solution for the abundances of He and O (and Na) in the accreted gas, and predict the abundances of these elements in the accreted matter only if pollution happens with timescales of ~1 Myr, hence polluters are objects with masses of the order of several tens of solar masses (abridged).
