\journalname
Astronomy&AstrophysicsReview

Introduction

\label{intro}
Stars are the main baryonic building blocks of galaxies and the central engines in their evolution. Accurate knowledge of the masses, radii, luminosities, chemical compositions, and ages of stars is fundamental to our understanding of their structure and evolution. This, in turn, underlies our models of the nucleosynthesis in stars and their interaction with their environment – the driving forces in the evolution of galaxies. It is important, therefore, to establish the basic properties of stars using a minimum of theoretical assumptions – preferably only geometry and Newtonian mechanics.
The required accuracy of these fundamental data depends on the intended application. When estimating the bulk properties of large groups of stars, an uncertainty of 10% may be quite acceptable, while for other purposes such data are useless. Testing models of stellar evolution is perhaps the most demanding application: The sensitivity of all properties of a stellar model to the initial mass is such that virtually any set of models will fit any observed star, if the mass is uncertain by just \(\pm\)5%. Only data with errors below \(\sim\)1–3% provide sufficiently strong constraints that models with inadequate physics can be rejected.
The aim of this paper is to present a critical assessment and summary of the currently available fundamental determinations of stellar masses and radii of sufficient accuracy for even the most demanding applications. Thus, this paper supersedes the earlier review by \cite{A91} (A91). As before, the stars included in our sample are all members of detached, non-interacting binary systems: Only for stars with detectable companions can the mass be determined directly and with errors of \(\sim\)1%, and only in eclipsing binary systems can both the stellar masses and radii be determined to this accuracy (with the sole exception of \(\alpha\) Centauri). We identify 95 binary systems in the literature that satisfy these selection criteria.
Significant progress since the earlier review includes the large body of data published since 1991, reflecting the improvements in observing and analysis techniques. In this paper, we also systematically recompute the stellar masses and radii from the original data with the same analysis techniques and modern, consistent values for the associated physical constants. Similarly, we critically review the published effective temperatures of the stars, which are used to compute their radiative properties. The stellar parameters listed in Table \ref{tableMR} are therefore not necessarily exactly identical to those given in the original papers.
To facilitate further discussion of the data presented here, we also provide individual rotational velocities, reddenings, metal abundances and distances whenever possible, and we compute approximate ages for all the systems. Finally, we list relevant additional data for the 29 systems that have well-studied apsidal motion.
Following a brief overview of the data, we discuss their use in testing state-of-the-art stellar evolution models. We also discuss the effects of tidal evolution in these systems, now from much-improved observational data, and explore the predictions of tidal theory for the axial rotation of binary components and the Newtonian and general relativistic contributions to the apsidal motion of the eccentric systems.
The data also allow us to devise calibrations that provide good estimates of mass and radius for single stars with reliable determinations of \(T_{\rm eff}\), \(\log g\) and [Fe/H].
Progress in long-baseline optical interferometry offers much promise for the future determination of stellar masses and radii, although the radii have not yet reached the level of accuracy aimed at here. We therefore list a number of (non-eclipsing) interferometric binaries in which the mass errors meet our selection criterion, but where more work is needed for the radii to also do so.
The paper ends with a number of considerations for the future.
We emphasise here that the criterion for inclusion in this review is quality of the data only, not any attempt to make the sample complete or unbiased in any sense. We warn the reader, therefore, that it is unsuitable for any kind of statistical analysis, its other qualities notwithstanding.