Speaker
Description
Intensity mapping for the 21 cm line is a promising route to map large-scale structure, but the cosmological HI signal is buried beneath Galactic and extragalactic foregrounds orders of magnitude brighter. We implement the Generalized Needlet Internal Linear Combination (GNILC) method to clean foregrounds from simulated MeerKAT single-dish observations in the UHF band (z ≈ 0.4–1.4). Simulations include a Gaussian-realization HI signal, spatially correlated Galactic synchrotron, and extragalactic point sources; the initial data cube was compressed from 252 to 20 channels and analysed at HEALPix nside = 256, probing multipoles ℓ ≈ 30–300. GNILC operates in needlet (wavelet) space, using a localized covariance comparison with a prior HI covariance to identify the HI subspace and apply a constrained ILC reconstruction. In tests excluding bright point sources, the recovered map shows high accuracy, matching the true signal (pixel-space correlation rp = 0.956) with rms residual ≈ 0.26 K, demonstrating GNILC’s effectiveness for diffuse foregrounds. Inclusion of extragalactic point sources degrades recovery, indicating that point-source needs masking or preprocessing and beam–frequency modelling to reach comparable performance on realistic data. We present validation in pixel and harmonic space, discuss limitations introduced by non-Gaussian small-scale contaminants, and outline steps to adapt GNILC for MeerKLASS single-dish analyses. These results support GNILC as a promising foreground-removal tool for upcoming HI intensity-mapping surveys.
| Stream | Science or Engineering |
|---|