Master's Thesis Defense by Rasmus Damgaard Nielsen

Title: Massively parallel simulation of dusty protostellar systems

Abstract: The formation of protoplanetary and protostellar systems is a complex and heterogeneous astrophysical phenomenon driven by the interaction between magnetohydrodynamical dynamics, self-gravitational forces, radiation, dust etc., all occurring at temporal and spatial scales spanning many orders of magnitude. These complexities and more make studying the system very difficult, and especially when it relates to the dust behaviour outside of the mid-plane of a homogeneous disk, many open questions are yet to be answered. One such unknown is the degree to which the outflows of the disk can transport away heated dust particles, enriching the surrounding GMC.

In this study, we introduce a high-performance dust simulation incorporated into the state-of-the-art three-dimensional code, \textsc{dispatch}\cite{nordlund_dispatch_2018}, integrating dust simulation into a highly detailed model of the magnetohydrodynamics and self-gravity in the system, modelling full dust-gas feedback. This allows realistic physical modelling of these complex environments all the way from initial collapse to disk formation. This approach provides a level of detail in studying the process of planetary formation that would not otherwise be possible, and its usefulness will continue to increase with the rapid development of supercomputer hardware.

With this implementation, we were able to simulate a protoplanetary system discretized into 1.5 million cells of gas and magnetism, as well as 60 million dust particles through more than 10,000 years of early evolution, at a cost of 150ns/particle update. Our findings validate the implementation's accuracy. Furthermore from this simulation, we can conclude that the rate of particles passing near the star and being ejected is highly particle-size dependent but may be up to on order 10\% of the total dust budget. This result is validated by analysing gas tracers simulated by the simulation framework \textsc{ramses}, although further study at higher resolution is needed.

With this implementation, we have thus taken a step towards enabling further computational studies of the behaviour of gas in the highly complex environment that is planet formation.

Supervisor: Troels Haugbølle

Censor: Steen Hannestad, Aarhus University