Departmental Doctoral Oral Examination (Thesis Title: “The effects of calibration errors and foreground filters on the CHIME power spectrum measurement. A study with simulations and real data.”)
Physics and Astronomy
The Canadian Hydrogen Intensity Mapping Experiment (CHIME) is a drift-scan radio telescope designed to map large scale structure in the universe using the redshifted 21 cm line emitted by neutral hydrogen. By observing the 400 to 800 MHz frequency band, CHIME will measure the expansion rate of the universe in the redshift range z = 0.8 - 2.5 to constrain the nature of dark energy.
In this frequency range, astrophysical foregrounds from the Galaxy and extragalactic point sources are much brighter than the 21 cm emission. This requires aggressive foreground filtering. We therefore developed a new implementation of the Karhunen-Loeve (KL) transform that correctly tracks the signal and noise power in our data to enable us to filter bright foregrounds. As part of the development of the KL transform, we significantly improved the point source component of the foreground model.
The data volumes for CHIME are extremely large, therefore we developed an upgraded parallelized power spectrum estimation pipeline which is able to forecast the Fisher information matrix and estimate power spectra for a telescope almost as large as CHIME.
Due to the bright astrophysical foregrounds CHIME has very stringent calibration requirements. We wrote an end-to-end simulation pipeline and studied various realistic sources of calibration uncertainties with it. The calibration requirements are very stringent and we found that CHIME currently doesn't quite meet the requirements when using the Karhunen-Loeve (KL) foreground filter. We must therefore continue to develop innovative foreground filters that are more robust against systematics and we must explore further whether our requirements are too conservative or if a larger telescope will allow us to relax these requirements.
We then investigated different processing choices on power spectrum estimation with CHIME data in the frequency band 610 to 680 MHz with a selection of short baselines to ensure quick computation times. We found the even with our currently best calibration procedures our power spectrum estimates are several orders of magnitude higher than expectation. Using a delay filter as an intermediate processing step seemed to reduce the bias but we leave further development of a hybrid filter to future work.