Imaging the Epoch of Reionization: limitations from foreground confusion and imaging algorithms

TL;DR

This paper analyzes the limitations of foreground confusion in 21 cm EoR imaging, introduces a formalism to estimate contamination, and proposes a new imaging algorithm to mitigate these effects, demonstrated with MWA data.

Contribution

It develops a formalism linking foreground contamination to PSF and visibility weights, and introduces a CZT-based imaging algorithm to reduce gridding contamination.

Findings

Identification of LOS foreground contamination sources

Existence of an EoR window with minimal contamination

CZT imaging algorithm reduces gridding contamination

Abstract

Tomography of redshifted 21 cm transition from neutral Hydrogen using Fourier synthesis telescopes is a promising tool to study the Epoch of Reionization (EoR). Limiting the confusion from Galactic and Extragalactic foregrounds is critical to the success of these telescopes. Instrumental response or the Point Spread Function (PSF) of such telescopes is inherently 3 dimensional with frequency mapping to the Line of Sight (LOS) distance. EoR signals will necessarily have to be detected in data where continuum confusion persists; therefore, it is important that the PSF has acceptable frequency structure so that the residual foreground does not confuse the EoR signature. This paper aims to understand the 3 dimensional PSF and foreground contamination in the same framework. We develop a formalism to estimate the foreground contamination along frequency, or equivalently LOS dimension, and establish a relationship between foreground contamination in the image plane and visibility weights on the Fourier plane. We identify two dominant sources of LOS foreground contamination-'PSF contamination' and 'gridding contamination'. We show that 'PSF contamination' is localized in LOS wavenumber space, beyond which there potentially exists an 'EoR window' with negligible foreground contamination where we may focus our efforts to detect EoR. 'PSF contamination' in this window may be substantially reduced by judicious choice of a frequency window function. Gridding and imaging algorithms create additional 'gridding contamination' and we propose a new imaging algorithm using the Chirp Z Transform (CZT) that significantly reduces this contamination. Finally, we demonstrate the analytical relationships and the merit of the new imaging algorithm for the case of imaging with the Murchison Widefield Array (MWA).