The equation of state is a fundamental ingredient in stellar evolution codes. It is implemented in the form of pre-calculated tables for a number of chemical mixtures. Using the FreeEOS program new tables for a larger variety of chemical mixtures, in particulas for so-called alpha-enhanced composition, plus an update of EoS for higher densities, should be prepared.
Detached eclipsing binaries are perfectly suited systems to a) determine ages of the system, or b) to check physical input of stellar modes. Several surveys or dedicated programs find an increasing number of suitable system, such that the conventional “by hand” analysis is no longer efficient enough. Based on a previous master thesis this project aims at developing a random forest machine learning tool that quickly returns first estimates of stellar physics parameters and the system age for a set of user-provided system parameters (mass, radius, composition, Teff, L, distance, ...). An extension of the training set for the random forest (semi-)automatically, and a consequent re-training would be additional aims of the project, as well as a user-friendly web-tool.
While stellar evolution programs are able to follow intermediate-mass stars through many Thermal Pulses of the helium-burning shell in models of Asymptotic Giant Branch (AGB) stars, all modern codes encounter severe convergence problems for the more massive models, once they have evolved up the AGB. The physical reasons are so far unclear, and may be related to locally exceeding the Eddington limit within the convective envelope, related to similar problems for massive stars past the main-sequence. The project is to identify the physical reasons and possible, physically well motivated solutions, which are not just numerical tricks to overcome the problem. A connection to multi-dimensional models may be possible.