Modelling the Galactic Foreground and Beam Chromaticity for Global 21-cm Cosmology

TL;DR

This paper develops a method to model and generate eigenvectors for the galactic foreground in 21-cm cosmology experiments, accounting for beam chromaticity and sky variations to improve foreground subtraction accuracy.

Contribution

It introduces a combined analytical and observational approach to generate basis eigenvectors that account for beam chromaticity effects in 21-cm foreground modeling.

Findings

Eigenvectors for achromatic, isotropic beams are similar regardless of foreground model.

Beam chromaticity causes eigenvector distortion and dependence on beam and foreground features.

Using SVD with detailed beam and sky models reduces uncertainties to milli-Kelvin levels.

Abstract

In order to characterize and model the beam-weighted foreground for global 21-cm signal experiments, we present a methodology for generating basis eigenvectors that combines analytical and observational models of both the galactic spectral index and sky brightness temperature with simulations of beams having various angular and spectral dependencies and pointings. Each combination creates a unique beam-weighted foreground. By generating eigenvectors to fit each foreground model using Singular Value Decomposition (SVD), we examine the effects of varying the components of the beam-weighted foreground. We find that the eigenvectors for modelling an achromatic, isotropic beam -- the ideal case -- are nearly identical regardless of the unweighted foreground model used, and are practicably indistinguishable from polynomial-based models. When anisotropic, chromatic beams weight the foreground, however, a coupling is introduced between the spatial and spectral structure of the foreground which distorts the eigenvectors away from the polynomial models and induces a dependence of the basis upon the exact features of the beam (chromaticity, pattern, pointing) and foreground (spectral index, sky brightness temperature map). We find that the beam has a greater impact upon the eigenvectors than foreground models. Any model which does not account for its distortion may produce RMS uncertainties on the order of $\sim 10$ - $10^3$ Kelvin for six-parameter, single spectrum fits. If the beam is incorporated directly using SVD and training sets, however, the resultant eigenvectors yield milli-Kelvin level uncertainties. Given a sufficiently detailed description of the sky, our methodology can be applied to any particular experiment with a suitably characterized beam for the purpose of generating accurate beam-weighted foreground models.