2005A&A...442..961C

Thanks to the accurate determination of the baryon density of the universe by the recent cosmic microwave background experiments, updated predictions of the standard model of Big Bang nucleosynthesis now yield the initial abundance of the primordial light elements with unprecedented precision. In the case of ⁷Li, the CMB+SBBN value is significantly higher than the generally reported abundances for Pop II stars along the so-called Spite plateau. In view of the crucial importance of this disagreement, which has cosmological, galactic and stellar implications, we decided to tackle the most critical issues of the problem by revisiting a large sample of literature Li data in halo stars that we assembled following some strict selection criteria on the quality of the original analyses. In the first part of the paper we focus on the systematic uncertainties affecting the determination of the Li abundances, one of our main goal being to look for the ``highest observational accuracy achievable" for one of the largest sets of Li abundances ever assembled. We explore in great detail the temperature scale issue with a special emphasis on reddening. We derive four sets of effective temperatures by applying the same colour-T<sub>eff</sub>calibration but making four different assumptions about reddening and determine the LTE lithium values for each of them. We compute the NLTE corrections and apply them to the LTE lithium abundances. We then focus on our ``best'' (i.e. most consistent) set of temperatures in order to discuss the inferred mean Li value and dispersion in several T_effand metallicity intervals. The resulting mean Li values along the plateau for [Fe/H]≤-1.5 are A(Li)_NLTE=2.214±0.093 and 2.224±0.075 when the lowest effective temperature considered is taken equal to 5700K and 6000K respectively. This is a factor of ∼2.48 to 2.81 (depending on the adopted SBBN model and on the effective temperature range chosen to delimit the plateau) lower than the CMB+SBBN determination. We find no evidence of intrinsic dispersion. Assuming the correctness of the CMB+SBBN prediction, we are then left with the conclusion that the Li abundance along the plateau is not the pristine one, but that halo stars have undergone surface depletion during their evolution. In the second part of the paper we further dissect our sample in search of new constraints on Li depletion in halo stars. By means of the Hipparcos parallaxes, we derive the evolutionary status of each of our sample stars, and re-discuss our derived Li abundances. A very surprising result emerges for the first time from this examination. Namely, the mean Li value as well as the dispersion appear to be lower (although fully compatible within the errors) for the dwarfs than for the turnoff and subgiant stars. For our most homogeneous dwarfs-only sample with [Fe/H]≤-1.5, the mean Li abundances are A(L)_NLTE = 2.177±0.071 and 2.215±0.074 when the lowest effective temperature considered is taken equal to 5700 K and 6000 K respectively. This is a factor of 2.52 to 3.06 (depending on the selected range in T_efffor the plateau and on the SBBN predictions we compare to) lower than the CMB+SBBN primordial value. Instead, for the post-main sequence stars the corresponding values are 2.260±0.1 and 2.235±0.077, which correspond to a depletion factor of 2.28 to 2.52. These results, together with the finding that all the stars with Li abnormalities (strong deficiency or high content) lie on or originate from the hot side of the plateau, lead us to suggest that the most massive of the halo stars have had a slightly different Li history than their less massive contemporaries. In turn, this puts strong new constraints on the possible depletion mechanisms and reinforces Li as a stellar tomographer.