Generalised Visibility ILC: a new foreground mitigation strategy for interferometric HI intensity mapping in the low-redshift Universe

Abstract

HI intensity mapping is a powerful probe of the large-scale matter distribution and of the underlying astrophysics and cosmology. However, its application is limited by the challenge of separating the faint neutral hydrogen (HI) signal from much brighter foreground emission. We present a new non-parametric foreground-cleaning method in visibility space for low-redshift HI intensity mapping and apply it to interferometric MeerKAT and SKA-Mid-like simulations. Our newly-developed Generalised Visibility Internal Linear Combination (GVILC) method combines Principal Component Analysis (PCA) for foreground subspace identification, informed by a prior HI signal-plus-noise covariance matrix, with a multi-dimensional ILC in visibility space to suppress foregrounds and recover the HI signal with reduced contamination. The number of modes to remove in the PCA is calculated using a data-driven statistical approach, leading to significant foreground residual suppression with limited HI signal loss. Moreover, GVILC in combination with foreground avoidance can prevent most of the signal loss, while simultaneously recovering modes that were initially contaminated by foregrounds. The mean squared error statistic shows that for both MeerKAT and SKA-Mid-like simulations, GVILC outperforms the foreground avoidance only strategy. We find that including a more realistic HI model with line broadening and spatial correlations between HI and continuum sources leads to increased signal loss after foreground removal, without affecting the number of modes removed by GVILC. This paper demonstrates that GVILC is an effective method to clean foreground contamination in HI intensity mapping with radio interferometric data, enabling the recovery of cosmological information from previously inaccessible regions of Fourier space.