Subtracting Foregrounds from Interferometric Measurements of the Redshifted 21 cm Emission

TL;DR

This paper presents a Fourier space foreground subtraction method for 21 cm cosmology, effectively removing bright point sources and diffuse emission in simulated data, enabling better detection of the underlying cosmological signal.

Contribution

Introduces a novel Fourier space foreground cleaning technique that handles all foreground types simultaneously, improving over traditional line-of-sight methods.

Findings

Effective removal of foreground contamination with residuals around 10 mK.

The method suppresses large-scale power spectrum systematic underestimates.

Combines uniform uv weighting and inverse-variance spectral fitting for optimal results.

Abstract

The ability to subtract foreground contamination from low-frequency observations is crucial to reveal the underlying 21 cm signal. The traditional line-of-sight methods can deal with the removal of diffuse emission and unresolved point sources, but not bright point sources. In this paper, we introduce a foreground cleaning technique in Fourier space, which allows us to handle all such foregrounds simultaneously and thus sidestep any special treatments to bright point sources. Its feasibility is tested with a simulated data cube for the 21 CentiMeter Array experiment. This data cube includes more realistic models for the 21 cm signal, continuum foregrounds, detector noise and frequency-dependent instrumental response. We find that a combination of two weighting schemes can be used to protect the frequency coherence of foregrounds: the uniform weighting in the uv plane and the inverse-variance weighting in the spectral fitting. The visibility spectrum is therefore well approximated by a quartic polynomial along the line of sight. With this method, we demonstrate that the huge foreground contamination can be cleaned out effectively with residuals on the order of \sim10 mK, while the spectrally smooth component of the cosmological signal is also removed, bringing about systematic underestimate in the extracted power spectrum primarily on large scales.