Observational Tests and Predictive Stellar Evolution. II. Nonstandard Models

We examine contributions of second-order physical processes to the results of stellar evolution calculations that are amenable to direct observational testing. In the first paper in the series, we established baseline results using only physics that were common to modern stellar evolution codes. In this paper we establish how much of the discrepancy between observations and baseline models is due to particular elements of new physics in the areas of mixing, diffusion, equations of state, and opacities. We then consider the impact of the observational uncertainties on the maximum predictive accuracy achievable by a stellar evolution code. The Sun is an optimal case because of the precise and abundant observations and the relative simplicity of the underlying stellar physics. The standard model is capable of matching the structure of the Sun as determined by helioseismology and gross surface observables to better than a percent. Given an initial mass and surface composition within the observational errors, and no current observables as additional constraints for which the models can be optimized, it is not possible to predict the Sun's current state to better than ~7%. Convectively induced mixing in radiative regions, terrestrially calibrated by multidimensional numerical hydrodynamic simulations, dramatically improves the predictions for radii, luminosity, and apsidal motions of eclipsing binaries while simultaneously maintaining consistency with observed light element depletion and turnoff ages in young clusters. Systematic errors in core size for models of massive binaries disappear with more complete mixing physics, and acceptable fits are achieved for all of the binaries without calibration of free parameters. The lack of accurate abundance determinations for binaries is now the main obstacle to improving stellar models using this type of test.