A General Bayesian Framework to Account for Foreground Map Errors in Global 21-cm Experiments

TL;DR

This paper presents a Bayesian framework for global 21-cm signal detection that accounts for foreground map errors, improving the robustness of cosmic dawn and reionization measurements.

Contribution

It introduces an advanced foreground modeling approach that incorporates measurement errors, enabling unbiased 21-cm signal recovery in the presence of base map inaccuracies.

Findings

Unaccounted foreground map errors bias 21-cm signal recovery.

The new model effectively mitigates bias caused by amplitude errors.

Simulated data shows successful detection of the 21-cm signal with the improved model.

Abstract

Measurement of the global 21-cm signal during Cosmic Dawn (CD) and the Epoch of Reionization (EoR) is made difficult by bright foreground emission which is 2-5 orders of magnitude larger than the expected signal. Fitting for a physics-motivated parametric forward model of the data within a Bayesian framework provides a robust means to separate the signal from the foregrounds, given sufficient information about the instrument and sky. It has previously been demonstrated that, within such a modelling framework, a foreground model of sufficient fidelity can be generated by dividing the sky into $N$ regions and scaling a base map assuming a distinct uniform spectral index in each region. Using the Radio Experiment for the Analysis of Cosmic Hydrogen (REACH) as our fiducial instrument, we show that, if unaccounted-for, amplitude errors in low-frequency radio maps used for our base map model will prevent recovery of the 21-cm signal within this framework, and that the level of bias in the recovered 21-cm signal is proportional to the amplitude and the correlation length of the base-map errors in the region. We introduce an updated foreground model that is capable of accounting for these measurement errors by fitting for a monopole offset and a set of spatially-dependent scale factors describing the ratio of the true and model sky temperatures, with the size of the set determined by Bayesian evidence-based model comparison. We show that our model is flexible enough to account for multiple foreground error scenarios allowing the 21-cm sky-averaged signal to be detected without bias from simulated observations with a smooth conical log spiral antenna.