Bayesian data analysis for sky-averaged 21-cm experiments with contamination from linearly polarised foreground

TL;DR

This paper demonstrates that using a physically informed data analysis pipeline, the global 21-cm signal can be reliably recovered from contaminated data, including effects from polarised foregrounds and Faraday rotation, with low error.

Contribution

It introduces a method to accurately extract the sky-averaged 21-cm signal despite complex polarised foreground contamination using the REACH pipeline.

Findings

Successful recovery of a 0.16 K 21-cm signal between 80-120 MHz.

Effective separation of the signal from polarised foregrounds with less than 30% error.

Linear mixing of multiple Faraday-rotated patches aids in distinguishing the signal.

Abstract

The precise measurement of the sky-averaged HI absorption signal between 50 and 200 MHz is the primary goal of global 21-cm cosmology. This measurement has the potential to unravel the underlying physics of cosmic structure formation and evolution during the Cosmic Dawn. It is, however, hindered by various non-smooth, frequency-dependent effects, whose structures resemble those of the signal. One such effect is the leakage of polarised foregrounds into the measured intensity signal: polarised foreground emission undergoes Faraday rotation as it passes through the magnetic fields of the interstellar medium, imprinting a chromatic structure in the relevant frequency range which complicates the extraction of the cosmological HI absorption feature. We investigate the effect of polarised Galactic foregrounds on extracting the global 21-cm signal from simulated data using REACH's data analysis pipeline; the Radio Experiment for the Analysis of Cosmic Hydrogen (REACH) is an experiment designed to detect the sky-averaged 21-cm HI signal from the early Universe using physically informed models. Using the REACH pipeline, we successfully recover an injected global 21-cm signal with an amplitude of approximately 0.16 K, centred between 80 and 120 MHz, achieving a low root-mean-square error (less than 30\% of the injected signal strength) in all the tested cases. This includes scenarios with simulated polarised Galactic diffuse emissions and polarised point source emissions, provided the overall polarisation fraction is below $\sim 3\%$. The linear mixing of contamination, caused by the superposition of multiple patches with varying strengths of Faraday rotation, produces patterns that are more distinct from the global signal. This distinction makes global signal recovery easier compared to contamination resulting from a single, slow oscillation pattern.