Statistical errors and systematic biases in the calibration of the convective core overshooting with eclipsing binaries

Abstract

We attempt to constrain the initial helium abundance, the age and the efficiency of the convective core overshooting of the binary system TZ Fornacis. Our main aim is in pointing out the biases in the results due to not accounting for some sources of uncertainty. We adopt the SCEPtER pipeline, relying on stellar models computed with two stellar evolutionary codes (FRANEC and MESA). We found multiple independent groups of solutions. The best one suggests a system of age 1.10 $\pm$ 0.07 Gyr (primary star in the central helium burning stage, secondary in the sub-giant branch), with a helium-to-metal enrichment ratio of $\Delta Y/\Delta Z = 1$ and core overshooting parameter $\beta = 0.15 \pm 0.01$ (FRANEC) and $f_{\rm ov}= 0.013 \pm 0.001 $ (MESA). The second class of solutions, characterised by a worse goodness-of-fit, still suggest a primary star in the central helium-burning stage but a secondary in the overall contraction phase, at the end of the main sequence. The FRANEC grid provides an age of $1.16_{-0.02}^{+0.03}$ Gyr and $\beta = 0.25_{-0.01}^{+0.005}$, while the MESA grid gives $1.23 \pm 0.03$ Gyr and $f_{\rm ov} = 0.025 \pm 0.003$. We also show that very precise mass determinations with uncertainty of a few thousandths of solar mass are required to obtain reliable determinations of stellar parameters, as mass errors larger than approximately 1% lead to estimates that are not only less precise but also biased. Moreover, we show that a fit obtained with a grid of models computed at a fixed $\Delta Y/\Delta Z$ can provide biased age and overshooting estimates. The possibility of independent overshooting efficiencies for the two stars of the system is also explored. This work confirms that to constrain the core overshooting parameter by means of binary systems is a very difficult task that requires an observational precision still rarely achieved and robust statistical methods.