Sporathe knowledge network
ES
Universidad Nacional de Colombia

Bogotá - Ciencias - Maestría en Ciencias - Física · 2021

Bayesian constraints on observable properties of exoplanetary systems using planet population synthesis simulations

Recent advances in exoplanet observations have led to the discovery of nearly 4400 planets in more than 3000 planetary systems. However, many physical properties of these systems remain largely unknown due to observational biases and stochasticity in their formation processes. We address current limitations of exoplanet classification by formulating a new, robust classification scheme based on the first moment of planet mass vs. semi-major axis distribution using Gaussian Mixture Models (GMM). We validate our method via Approximate Bayesian Computation and Information Criteria, which shows that it is robust with uncertainties measurement and adaptable to new observations. Our scheme yields four planetary system classes: Sub-Mercurial systems, Venusian systems, Solar-like systems, and Periphery systems. We also propose a new method for compensating observational biases while addressing stochasticity in star and planet formation. To this end, we develop a Bayesian probabilistic formalism in which we take priors from observed planetary systems and marginalize them over synthetic likelihood functions. It allows us to estimate probability distribution functions for variables of interest, such as the total number of planets, total planetary mass, rocky planetary mass, center of mass, among others. We generate our synthetic likelihood functions from a multivariate Kernel Density Estimation (KDE) model based on the results of a Monte Carlo simulation of 1200 planet population synthesis models, drawn from observational priors obtained from the literature. We assess the performance of the kernel parameter choice using cross-validation. Therefore, we got probability distributions of physical variables of interest for ten observed systems using data from public catalogues. For the selected systems, we infer that they had initial disks with masses around 0.1 M⊙ ± 0.01 M⊙, their centers of mass are located around 5 AU ± 2 AU, and they should have around seven more planets than are currently observed. We also conclude that the number of rocky planets significantly contributes to the total number of planets, so we expect to find more rocky planets in future observations. Our formalism allows getting the probability distributions of exoplanetary systems unobserved or with biased properties. It will help steering future astronomical surveys and motivating further questions of observed planetary systems.

Full text 100 pages with text out of 104

Read the full thesis Record in the repository

Its library's thesaurus