Accueil du site > Congrès-Séminaires > Archives > Soutenances Thèse > Soutenance de thèse de Javier Olivares



OSUG - Terre Univers Environnement OSUG

Soutenance de thèse de Javier Olivares

Bayesian Hierarchical Modelling of Young Stellar Clusters

Javier Olivares (IPAG)
Jeudi 19 octobre 2017 à 10h
Salle Forestini (bât. OSUG-A)

The origin and evolution of stellar populations is one of the greatest challenges in modern astrophysics. Understanding the origin and evolution of stars demands meticulous analyses of stellar clusters, where the majority of the stars are born, particularly at young and intermediate ages. The objective of my PhD thesis is to analyse the statistical distributions of the stellar populations in Nearby Young Clusters (NYC). In particular, the spatial, proper motions, and luminosity distributions.

The project Dynamical Analysis of Nearby Clusters (DANCe, an international collaboration), from which this work is part of, provides the scientific framework for the analysis of these NYCs. The DANCe carefully designed observations of the well known Pleiades cluster provide the perfect case study for the development and testing of statistical tools aiming at obtaining parametric representations of the cluster distributions. The statistical tool, developed in this work (called a probabilistic intelligent system in AI), is a Bayesian Hierarchical Model (BHM) which infers some of the prior distributions from the data, thus avoiding (but not eliminating) much of the subjectivity of prior choosing. In this BHM, the true values of the cluster distributions are specified by stochastic and deterministic relations representing the state of knowledge of the NYC, without the explicit use of stellar evolutionary models.

In performing the parametric inference, the BHM accounts for the properties of the data set, especially its heteroscedasticity and missing value objects (under very simplistic assumptions though). By accounting for these properties, the BHM : i) Increases the size of the data set (10^5 objects), with respect to previous studies working exclusively on fully observed data (i.e. with no-missing values) and ii) Reduces the biases associated to fully observed data sets, and restrictions to low-uncertainty objects (i.e. sigma-clipping procedures). As a by-product, the BHM also renders the cluster membership probability of each object in the data set. From the analysis of synthetic data sets, the expected value of the contamination rate in the cluster distributions is 5.8 ± 0.2%.

The results of the BHM applied to the Pleiades cluster are :

- It finds 1967 candidate members form which 205 are new. The agreement with previous lists of candidate members from the literature (Stauffer et al. 2007, Bouy et al. 2015 and Rebull et al. 2016) is larger than 90%.

- The derived Present-Day System Mass Distribution, (using the mass-luminosity relation from stellar evolutionary models) is in general agreement with previous results. However, it shows discrepancies with the IMFs (e.g. Chabrier et al.2005) in the low-mass end.

The BHM developed in this work could be directly applied to new photometric and astrometric data sets (e.g. Gaia+DANCe, Gaia+Pan-STARRS) accounting for their particular data properties (heteroscedastic and correlated uncertainties).

Sous la tutelle de:


Sous la tutelle de:

CNRS Université Grenoble Alpes