An efficient method for removing point sources from full-sky radio interferometric maps

TL;DR

This paper introduces an efficient linear method using Wiener filtering and advanced matrix inversion techniques to remove point sources from full-sky radio interferometric maps, improving data analysis for 21-cm cosmology surveys.

Contribution

It extends the $m$-mode formalism to incorporate realistic sky models with bright point sources and develops an efficient algorithm for their removal.

Findings

Algorithm successfully removes point sources in simulations

Efficient matrix inversion reduces computational cost

Method enables optimal data filtering for large datasets

Abstract

A new generation of wide-field radio interferometers designed for 21-cm surveys is being built as drift scan instruments allowing them to observe large fractions of the sky. With large numbers of antennas and frequency channels the enormous instantaneous data rates of these telescopes require novel, efficient, data management and analysis techniques. The $m$-mode formalism exploits the periodicity of such data with the sidereal day, combined with the assumption of statistical isotropy of the sky, to achieve large computational savings and render optimal analysis methods computationally tractable. We present an extension to that work that allows us to adopt a more realistic sky model and treat objects such as bright point sources. We develop a linear procedure for deconvolving maps, using a Wiener filter reconstruction technique, which simultaneously allows filtering of these unwanted components. We construct an algorithm, based on the Sherman-Morrison-Woodbury formula, to efficiently invert the data covariance matrix, as required for any optimal signal-to-noise weighting. The performance of our algorithm is demonstrated using simulations of a cylindrical transit telescope.