Leveraging Machine Learning to Constrain Spica’s Apsidal Constant through MESA Simulation

The eccentric orbit of the massive binary Spica (Period = 4.01 days) shows an advance of its line of apsides; one rotation of the line of apsides takes ∼100 years. This advance of longitude of periastron depends on the distribution of mass within the stellar components. The original measurement of the apsidal constant, which comes from intensity interferometry (Herbison-Evans et al. 1971), sparked several theoretical investigations (e.g. Odell 1974) into the apsidal constant for the primary star which predicted values different from the measured value. To constrain a modern theoretical value for the apsidal constant for Spica, we are using the MESA (Modules for Experiments in Stellar Astrophysics) code to predict the constant through the post main-sequence phase for Spica A to hydrogen-core exhaustion. We are exploring how the apsidal constant changes over a range of parameters including the mass, metallicity, rotation, convection and equations of state. This will be computationally intensive. By augmenting MESA with machine learning, we aim to enhance computational efficiency and cost-effectiveness, thereby advancing our understanding of this binary star system.

There are two primary machine learning objectives. The first objective is to apply regression methods to solve for the apsidal constant at grid points in parameter space not calculated by MESA. A regression model will be employed first to establish a baseline understanding of the relationship between the variables. The primary technique chosen for this project is Recursive Feature Elimination (RFE) due to its ability to pinpoint the subset of features that are most important. The second objective, focusing on feature engineering techniques, aims to uncover the critical features that significantly influence the determination of the apsidal constant.