Thèse Simulations Numériques de l'Évolution Dynamique d'Amas Stellaires Enfouis H/F - 38

Établissement : Université Grenoble Alpes
École doctorale : PHYS - Physique
Laboratoire de recherche : Institut de Planetologie et d'Astrophysique de Grenoble
Direction de la thèse : Estelle MORAUX ORCID 0000000341277295
Début de la thèse : 2026-10-01
Date limite de candidature : 2026-05-11T23:59:59

L'objectif de ce projet de thèse est de réaliser des simulations numériques de l'évolution dynamique des amas stellaires enfouis, depuis l'effondrement du nuage moléculaire jusqu'à plus de 10 millions d'années, afin de fournir un cadre théorique permettant d'interpréter les résultats d'observation récents d'amas stellaires. En confrontant les simulations aux observations, ce travail permettra de vérifier si les propriétés structurelles et cinématiques, les gradients d'âge et la dynamique des amas jeunes permettent de remonter aux conditions physiques des nuages moléculaires formant des étoiles (turbulence, masse, densité, état dynamique). Les résultats seront ensuite comparés à des modèles de formation stellaire, ce qui permettra de contraindre les processus dominants qui régissent l'assemblage et l'évolution des amas jeunes.

During the last decade, the new unprecedented quality astrometric measurements from Gaia (one of ESA's major missions, launched in 2013) have revolutionised our understanding of stellar clusters - which we take here as any grouping of stars from a few tens to thousands of members. Our capability to detect them and characterise them has dramatically improved, and led to important discoveries. In particular, the study of the spatial distribution and kinematics of young cluster populations have revealed that they often follow the filamentary structures of their parental molecular cloud. As a consequence, young stellar clusters typically exhibit complex morphological and dynamical substructures. They are composed of multiple substructures in the 5 or 6-D phase space, with an age gradient between them, suggesting several star-forming events propagating through the cloud, and show a certain degree of mass segregation in their pivotal early embedded phase of evolution.
These recent findings show that the traditional view of a stellar cluster as being the result of a single, isolated formation process has clearly reached its limits. In particular, a complete understanding on how the stars inherit their properties from their parental gas, the order in which they form and how star-forming events propagate is still lacking. This has called for a new paradigm of stellar formation, which appears much more dynamic than originally thought.
Star clusters are born from a large reservoir of diffuse gas and dust, the interstellar medium (ISM). The system is organised in a hierarchy of scales, thus leading to the formation of substructures. The formation process is governed by the intricate interplay of often competing physical agents such as gravity, turbulence, magnetic fields, and radiation. To add to this complexity, stellar feedback from young massive stars creates highly non-linear effects that strongly influence the dynamical evolution across the entire cascade of scales. Depending on this feedback, the star formation efficiency of the region, and the internal dynamics within each group, a star-forming region evolve within a few Myr into either a relatively dense and concentrated star cluster where clustering has been erased or a low-density sub-structured association that will disperse into the Galactic disk. Thus, even though 75%-80% of all stars in the solar neighbourhood form in embedded systems with N>100 members, most of them will end up in the Galactic field within ~10 Myr.
Deciphering the embedded cluster evolution is necessary to reveal the propagation of star-forming events and trace back the structuration of stellar systems at birth, i.e., the hierarchical filamentary structure and dynamics of the interstellar gas out of which stars form.
In addition, knowing the initial conditions of star formation and the early evolution of stellar systems is essential to understand the effect of environment on the properties of young stars. In particular, the way stars assemble their mass depends on the availability of a mass reservoir, which may be reduced in the presence of neighbours competing for accretion, or ripped out by UV flux of massive stars and dynamical interactions.

In order to be able to simulate the dynamical evolution of young star clusters from the cloud collapse to several tens of Myr, it is necessary to start from the cluster formation phase, follow the intricate gas and star dynamics during the embedded stage and then compute the stellar dynamical evolution after gas expulsion. However, while state-of-the-art MHD codes (such as Arepo or RAMSES) are now able to simulate the collapse and fragmentation of large turbulent molecular clouds taking into account complex physics and stellar feedback consistently, these simulations are still CPU expensive and stop after a few free fall times (<1Myr), when the proto-stellar cluster population is still fairly embedded in the parental gas and accreting, and cannot be directly compared to observations of stellar clusters. On the other hand, pure N-body calculations have become very fast thanks to the advent of GPUs. They can follow the dynamical evolution of large stellar clusters for several 100 Myr very precisely but cannot treat the early phases of cluster evolution when a lot of gas remains. Some simulations combining hydrodynamical and N-body codes have started to emerge,.
In our group, we have developed the required numerical scheme to perform such calculations, taking into account both the gas and stellar dynamics as well as stellar feedback. Methods originating from the N-body community, including regularisation and slow-down methods (SDAR), have been added to the SPH hydrodynamical code Phantom. This has been completed by a prescription for H II region and a subgrid model of star formation to initialise stars with a low numerical cost, but in a way that is consistent with the gas distribution during the cloud collapse. These new developments allow accurate hybrid simulations in minimal calculation time, which is the key to generating a larger set of simulations, and analysing the stochastic nature of dynamical interactions.
Using this new numerical framework, the PhD student will run simulations of the early dynamical evolution of stellar clusters, starting with various molecular cloud initial conditions in terms of turbulence, mass and density. For each set of initial conditions, ~10 calculations will be run in order to define average evolutionary paths when possible. We will test in particular if the evolution of substructures depends on specific initial conditions or on the dynamical history. We will also investigate the set-up of mass segregation and look at the dynamical decay of unstable multiple systems. Another aspect that these simulations will allow us to address is the fate of the cluster after gas dispersal. Whether or not a cluster will survive depends on its virial ratio as well as on the star formation efficiency. Moreover, if a cluster survives, the gas removal may affect its properties, such as the mass function, especially if mass segregation has already occurred. Indeed, in that case the lowest mass objects lie at the outskirt of the cluster and will be preferentially lost when the gas is removed.
Depending on the interest of the PhD student, he/she may also constrain further the subgrid prescription of star formation by analysing the sink formation in high resolution numerical simulations performed with RAMSES to derive their statistical properties in terms of clustering and masses.