Electronic Theses and Dissertations (PhDs)

Permanent URI for this collectionhttps://hdl.handle.net/10539/38021

Browse

Now showing 1 - 2 of 2

An unsupervised search of non-thermal diffuse emission in extended sources
(University of the Witwatersrand, Johannesburg, 2022-01) Thwala, Siphiwe Anthony; Beck, Geoff
Low-surface brightness diffuse radio emission is a useful probe for studying the connection between non-thermal processes in radio sources and their environment, as well as gaining insights on galaxy formation. The presence of this emission in different astrophysical and cosmological signals is not well understood in the literature due to challenges in detection caused by the low-surface brightness of diffuse objects and their origins, particularly in relation to galaxy feedback and formation. We explored the utility of publicly available data to detect, characterise, and study diffuse emission in extended radio sources at the scales of radio galaxies and galaxy clusters. We performed multiple analyses on extended radio continuum sources found in large publicly available radio surveys. The analyses range from in-depth multi-frequency examinations of individual sources to novel approaches for grouping and classifying radio sources with unsupervised machine learning. The analysis of individual sources brought together a range of datasets and techniques to provide insight about their nature. To group and classify continuum radio sources at scale, a novel unsupervised machine learning approach was designed to combine self-organising maps with convolutional neural networks for automatically detecting and clustering similar sources in radio surveys. To the best of our knowledge, this is the first implementation of an architecture that allows for the training of a machine learning model using multi-frequency and multi-scale radio continuum data cubes, as input, to automatically detect and cluster similar sources in radio surveys. This comes at a time when radio astronomy is undergoing a transformation and data mining methods are critical for optimum scientific utilisation of data from telescopes like MeerKAT, JVLA, MWA, and LOFAR (as well as the upcoming SKA). The different works showcase the most interesting radio sources with diffuse emission observed in these investigations.
Hunting dark matter with faint radio halos
(University of the Witwatersrand, Johannesburg, 2023-10) Sarkis, Michael David; Beck, Geoff
The nature of Dark Matter (DM), the elusive substance that constitutes a significant amount of the total matter in the universe, remains an unsolved problem in modern physics despite a decades-long search effort. One approach to this problem has been to search for faint emission signatures that are produced indirectly from the DM present in large astrophysical structures, and thus infer properties about theoretical DM models from observational data. In recent years, the results from studies that use this type of indirect search have produced stringent constraints on the most popular DM particle candidate parameter spaces, ruling out swathes of viable DM models. These compelling results have been enabled by the arrival of sophisticated interferometric radio telescopes, which are excellent DM hunters due to their high sensitivity and resolution. In this thesis, we focus on the use of the latest data from the MeerKAT radio interferometry telescope, through the first public release of the MeerKAT Galaxy Cluster Legacy Survey, to search for DM emissions in a set of nearby galaxy clusters. Each step of this process, from the creation of theoretical DM emission models to the statistical analysis of the observational data, has been described in detail in this thesis. With this data, we find an almost universal improvement to results found with corresponding modelling scenarios in the literature. Since this work is among the first to use MeerKAT data in astrophysical DM searches, these results present a strong argument for continued work in this field. Another central focus of this thesis is the accurate modelling of the physical processes involved in the production of the DM-induced radio emissions, as the quality of current radio data requires theoretical models that are sufficiently accurate to describe the emission at such high resolutions. One aspect of the modelling that has lacked this accuracy has been the solution to the diffusion-loss equation, which is a crucial factor in determining indirect DM emissions. A new algorithm for solving this equation, which provides higher accuracy and computational efficiency than previous public methods, has thus been developed and presented in this thesis. These aspects of DM indirect detection study will become ever more important as we approach the inauguration of the Square Kilometre Array (SKA), which will provide data with unprecedented potential with which to continue the hunt for DM.