From theoretical promise to observational reality : calibration, foreground subtraction, and signal extraction in hydrogen cosmology

Abstract

By using the hyperfine 21 cm transition to map out the distribution of neutral hydrogen at high redshifts, hydrogen cosmology has the potential to place exquisite constraints on fundamental cosmological parameters, as well as to provide direct observations of our Universe prior to the formation of the fist luminous objects. However, this theoretical promise has yet to become observational reality. Chief amongst the observational obstacles are a need for extremely well-calibrated instruments and methods for dealing with foreground contaminants such as Galactic synchrotron radiation. In this thesis we explore a number of these challenges by proposing and testing a variety of techniques for calibration, foreground subtraction, and signal extraction in hydrogen cosmology. For tomographic hydrogen cosmology experiments, we explore a calibration algorithm known as redundant baseline calibration, extending treatments found in the existing literature to include rigorous calculations of uncertainties and extensions to not-quite-redundant baselines. We use a principal component analysis to model foregrounds, and take advantage of the resulting sparseness of foreground spectra to propose various foreground subtraction algorithms. These include fitting low-order polynomials to spectra (either in image space or Fourier space) and inverse variance weighting. The latter method is described in a unified mathematical framework that includes power spectrum estimation. Foreground subtraction is also explored in the context of global signal experiments, and data analysis methods that incorporate angular information are presented. Finally, we apply many of the aforementioned methods to data from the Murchison Widefield Array, placing an upper limit on the Epoch of Reionization power spectrum at redshift z = 9:1.