A small fraction of the halo field is made up of stars that share the light element (Z<=13) anomalies characteristic of second generation globular cluster (GC) stars. The ejected stars shed light on the formation of the Galactic halo by tracing the d
ynamical history of the clusters, which are believed to have once been more massive. Some of these ejected stars are expected to show strong Al enhancement at the expense of shortage of Mg, but until now no such star has been found. We search for outliers in the Mg and Al abundances of the few hundreds of halo field stars observed in the first eighteen months of the Gaia-ESO public spectroscopic survey. One halo star at the base of the red giant branch, here referred to as 22593757-4648029 is found to have [Mg/Fe]=-0.36+-0.04 and [Al/Fe]=0.99+-0.08, which is compatible with the most extreme ratios detected in GCs so far. We compare the orbit of 22593757-4648029 to GCs of similar metallicity and find it unlikely that this star has been tidally stripped with low ejection velocity from any of the clusters. However, both chemical and kinematic arguments render it plausible that the star has been ejected at high velocity from the anomalous GC omega Centauri within the last few billion years. We cannot rule out other progenitor GCs, because some may have disrupted fully, and the abundance and orbital data are inadequate for many of those that are still intact.
The influence of the uncertainties in the rate coefficient data for electron-impact excitation and ionization on non-LTE Li line formation in cool stellar atmospheres is investigated. We examine the electron collision data used in previous non-LTE ca
lculations and compare them to recent calculations that use convergent close-coupling (CCC) techniques and to our own calculations using the R-matrix with pseudostates (RMPS) method. We find excellent agreement between rate coefficients from the CCC and RMPS calculations, and reasonable agreement between these data and the semi-empirical data used in non-LTE calculations up to now. The results of non-LTE calculations using the old and new data sets are compared and only small differences found: about 0.01 dex (~ 2%) or less in the abundance corrections. We therefore conclude that the influence on non-LTE calculations of uncertainties in the electron collision data is negligible. Indeed, together with the collision data for the charge exchange process Li(3s) + H <-> Li^+ + H^- now available, and barring the existence of an unknown important collisional process, the collisional data in general is not a source of significant uncertainty in non-LTE Li line formation calculations.
