Towards AMR Simulations of Galaxy Formation

Abstract

Numerical simulations present a fundamental building block of our modern theoretical understanding of the Universe. As such the work in this thesis is primarily concerned with understanding fundamental differences that lie between the different hydrodynamic schemes. In chapter 3 I outline the optimisations I make to the FLASH code to enable larger simulations to be run. These include developing and testing a new FFT gravity solver. With these complete, in chapter 4 I present results from a collaborative code comparison project in which we test a series of different hydrodynamics codes against a suite of demanding test problems with astrophysical relevance. As the problems have
known solutions, we can quantify their performance and are able to develop a resolution criteria which allows for the two different types to be reliably compared.

In chapter 5 we develop an analytic model for ram pressure stripping of the hot gaseous haloes of galaxies in groups and clusters. We test the model against a suite of hydrodynamic simulations in the SPH GADGET-2 code. To ensure that the spurious suppression of hydrodynamic instabilities by SPH codes does not bias our results, I compare our findings to those obtained with the FLASH AMR code and find excellent agreement.

Chapter 6 presents work in which we unambiguously determine the origin of the difference between the entropy cores formed in AMR and SPH codes. By running mergers of model clusters we are able to systematically explore the various proposed mechanisms and determine that turbulent mixing generates the higher entropy cores within AMR codes but is suppressed in SPH codes.

The startling differences between the two hydrodynamic schemes presented in chapter 6 leads me to investigate their affect upon different sub-grid physical recipes. In chapter 7 I present the implementation of a sub-grid star formation recipe in FLASH and find strong differences in the way the two codes model pressure laws. I extend the work in chapter 8 by implementing a kinetic supernova feedback mechanism in FLASH and contrasting it with the results from the GADGET-2 code. I find that AMR codes dissipate energy much more efficiently than in SPH codes.