Millions of synthetic global populations, filtered through Kepler and then compared with actual detections, refine the occurrence rates according to the stars’ mass and metallicity.
Chloé Padois, Friedrich Anders and Daniel del Ser present a simulator capable of generating exoplanet populations for millions of stars in a few seconds, combining observed detections and recent models of planetary formation. It takes into account the links between the properties of the host star and the frequency of the planets, the distribution of orbital periods or the mass-radius relationship.
Learn to report observations to parameters
Because many of these parameters remain uncertain, the team couples them to simulation-based inference: 44 free parameters are drawn in wide intervals, millions of populations are simulated and then "observed" as Kepler would, and a machine learning algorithm learns to relate the input parameters to the resulting catalogues. The catalogue of planets confirmed by Kepler, the largest homogeneous sample from a single instrument, serves as a reference.
The first results focus on the frequency of the planets according to the mass and metallicity of the star, as well as on the distribution of periods and the dispersion of inclinations within the systems. They should be used to better predict the yields of future PLATO, Roman and HAYDN missions.
Read the full article on the original source
Go to the source →