Christopher M. Orban
Published: 2011
Total Pages: 85
Get eBook
Abstract: With the many questions raised by the discovery, about a decade ago, that the universe is in an accelerating phase of expansion, it is clear that cosmological N-body simulations will continue to play a key role in unravelling this mysterious phenomenon. Historically, the first generation of cosmological N-body codes were tested and verified using self-similar models for dark matter clustering that have special "scale-free" properties. These models, in addition to allowing novel tests of numerical accuracy, are also interesting as fundamental problems and have been insightful in illuminating the non-linear physics of cosmological structure formation. In this thesis I return to this theme, in the first part investigating in considerable detail a new class of self-similar dark matter clustering models with a large-scale clustering feature that closely resembles baryonic acoustic oscillations (BAO) - a key distance indicator for dark energy studies. The non-linear physics of this simplified model was investigated using cosmological N-body simulations and the results compared both to perturbation theory and a phenomenological model. In these comparisons the phenomenological model and one of the two perturbation theory models discussed generally matched the simulation data quite well - more specifically the "SimpleRG" perturbation theory scheme of McDonald (2007) was remarkably accurate. I also carried out (for the first time) a suite of numerical tests with this new self-similar model, concluding that with modest numerical requirements, current N-body simulations will accurately model the non-linear evolution of a BAO-like feature even for a wide range of broadband spectral power. Importantly, this statement is true of the shift of the BAO clustering feature - a crucial systematic for dark energy studies. In the second half of this thesis, again making use of self-similar numerical tests, I evaluate an alternative method for setting up and running ensembles of cosmological N-body simulations developed by Sirko (2005) based on the ideas of Pen (1997). This method maintains correspondence between the actual and simulated real-space (rather than fourier space) properties of the cosmological model and accordingly allows the average density in each box to vary slightly from one realization to another. Extensive tests show that this approach gives indistinguishable results, compared to the standard method, for the mean dark matter and halo clustering properties but that the box-to-box variance of these statistics in the new method is much higher than expected, making the scheme substantially sub-optimal for most uses. I discuss the assumptions in the method which cause this and comment on a regime where the Sirko (2005) approach may be very useful.