2015A&A...579A..59V

We investigate the performance of grid-based techniques in estimating the age of stars in detached eclipsing binary systems. We evaluate the precision of the estimates due to the uncertainty in the observational constraints - masses, radii, effective temperatures, and [Fe/H] - and the systematic bias caused by the uncertainty in convective core overshooting, element diffusion, mixing-length value, and initial helium content. We adopted the SCEPtER grid, which includes stars with mass in the range [0.8; 1.6]M_☉ and evolutionary stages from the zero-age main sequence to the central hydrogen depletion. Age estimates have been obtained by a generalisation of the maximum likelihood technique described in our previous work. We showed that the typical 1σ random error in age estimates - due only to the uncertainty affecting the observational constraints - is about ±7%, which is nearly independent of the masses of the two stars. However, such an error strongly depends on the evolutionary phase and becomes larger and asymmetric for stars near the zero-age main sequence where it ranges from about +90% to -25%. The systematic bias due to the including convective core overshooting - for mild and strong overshooting scenarios - is about 50% and 120% of the error due to observational uncertainties. A variation of ±1 in the helium-to-metal enrichment ratio ΔY/ΔZ accounts for about ±150% of the random error. The neglect of microscopic diffusion accounts for a bias of about 60% of the error due to observational uncertainties. We also introduced a statistical test of the expected difference in the recovered age of two coeval stars in a binary system. We find that random fluctuations within the current observational uncertainties can lead genuine coeval binary components to appear to be non-coeval with a difference in age as high as 60%.