Testing pre-main-sequence models: the power of a Bayesian approach

Abstract

Pre-main sequence (PMS) models provide invaluable tools for the study of star forming regions as they allow to assign masses and ages to young stars. Thus it is of primary importance to test the models against observations of PMS stars with dynamically determined mass. We developed a Bayesian method for testing the present generation of PMS models which allows for a quantitative comparison with observations, largely superseding the widely used isochrones and tracks qualitative superposition. Using the available PMS data we tested the newest PISA PMS models establishing their good agreement with the observations. The data cover a mass range from ~0.3 to ~3.1 Msun, temperatures from ~3x10^3 to ~1.2x10^4 K and luminosities from ~3x10^-2 to ~60 Lsun. Masses are correctly predicted within 20% of the observed values in most of the cases and for some of them the difference is as small as 5%. Nevertheless some discrepancies are also observed and critically discussed. By means of simulations, using typical observational errors, we evaluated the spread of log \tau_sim - log \tau_rec, i.e. simulated minus recovered ages distribution of the single objects. We also found that stars in binary systems simulated as coeval might be recovered as non coeval, due to observational errors. The actual fraction of fake non coevality is a complex function of the simulated ages, masses and mass ratios. We demonstrated that it is possible to recover the systems' ages with better precision than for single stars using the composite age-probability distribution, i.e. the product of the components' age distributions. Using this valuable tool we estimated the ages of the presently observed PMS binary systems.