Cross-Correlations in the New Era of Intensity Mapping

Student thesis: Doctoral Thesis


Neutral Hydrogen (HI) in the Universe can be detected via the electron spin flip producing 21 centimetre wavelength electromagnetic radiation, which is equivalent to a frequency of 1420 MHz. This wavelength can already be observed with great sensitivity by several radio telescopes such as MeerKAT, the Canadian Hydrogen Intensity Mapping Experiment (CHIME), and the Green Bank Telescope. The abundance of the element is estimated by the observed surface brightness temperature.
A very useful new technique to measure this kind of signal is intensity mapping (IM). Instead of focusing on the individual sources of HI (e.g. particular galaxies), IM instead measures the density of the intensity on larger scales.
In this thesis, after introducing the overall cosmological picture in Chapter 1, I discuss intensity mapping in detail in Chapter 2. The main challenge for this intensity mapping technique is the low surface brightness temperature of the signal compared to the very bright foreground generated by our own Galaxy. Hence, foreground removal is an essential process in 21cm mapping observations.
The properties of the large-scale structure of the Universe and its evolution can be identified by studying the statistics of the IM. Since the cosmological field is near- Gaussian on the relevant scales, 2-point statistics are preferred when dealing with large sky survey data. We can cross-correlate the IM with other fields to gain further cosmological information.
In this thesis, I focus on the cross-correlation with optical weak gravitational lensing. This lensing is caused by the deflection of light due to the gravitational field. The effect can be seen in the distortion of galaxy images. The cross-correlation between IM and lensing therefore probes the geometry of the Universe, the growth of large scale structure, and the relationship between dark matter and radio-emitting matter.
IM foreground removal unfortunately also removes density fluctuations with long wavelengths along the line of sight. An unexplored issue is whether this process drastically removes the correlation of the remaining IM signal with the lensing field; this would spoil our ability to obtain good cosmological constraints.
Therefore, in Chapter 3 I assess the damage done to the cross-correlation by fore- ground removal, using realistic simulations. 35 realisations of N-body simulations are used to measure the lensing convergence (κ) and HI 2-point statistics (κ HI)(l). The result shows that foreground removal reduces cross-correlation signals by a factor of 2-5, but the shape of the correlation function is not substantially changed. Fisher information matrices are calculated to assess the likelihood of cosmological parameters. These show the viability of kappa-HI as a tool for cosmological constraints.
I proceed (Chapter 4) to examine the effects of instrument noise for Kilo-Degree Survey (KiDs)-like instruments for lensing and MeerKAT-like instruments for radio emission, to assess the feasibility of realistic experiments. I find similar levels of instrument noise and cosmic variance in the optimistic case when observation time is over 1000 hours. Even when the cross-correlations are dominated by noise, it is still possible to detect the signal in some cases with deep IM mapping and large redshift and angular bins. Both HI and lensing measurements are found to require at least a few hundred square degree common patch.
Finally I attempt to measure the cross-correlation with real data from 10 hours of MeerKAT integration together with the KiDS1000 survey (Chapter 5). I present the measured correlation functions for different redshift slices, and make tests of systematic effects. I find that due to the small shared sky patch between these surveys, the outcome is dominated by noise. However, the foreground removal shows a promising result for getting to cosmologically interesting noise levels on large scales. On small scales the 2-point statistics are still 2 orders of magnitude higher than the theoretical model. This indicates that we require additional processes to reduce noise levels for small scales. I present our final conclusions and topics for further work in Chapter 6.
Date of Award31 Jan 2024
Original languageEnglish
SupervisorDavid Bacon (Supervisor), Robert Crittenden (Supervisor) & Florian Beutler (Supervisor)

Cite this