Needlets and foreground removal for SKAO hydrogen intensity maps
Carmelita Carbone · 2025
Read Needlets and foreground removal for SKAO hydrogen intensity maps by Carmelita Carbone (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-02644.txt.
Introduction
Bianca De Caro, Isabella P. Carucci, Stefano Camera, Mathieu Remazeilles, and Carmelita Carbone present a 2025 cosmology paper on cleaning hydrogen maps from the Square Kilometre Array telescope. The 21-centimetre signal from neutral hydrogen traces cosmic structure, but foreground glow outshines it by orders of magnitude.
Their method adapts principal component analysis using needlets, wavelets localised on the sphere in both position and scale. Simulated maps for a planned telescope configuration test the pipeline against two rival separation methods. All three recovered the hydrogen signal within about ten percent across channels. Polarisation leakage did not hurt the cleaners. With a realistic beam, the needlet versions held up at large scales where standard methods failed, while one rival over-cleaned and lost signal.
Catalog record
| CREATOR | Carmelita Carbone |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 56 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-02644
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Needlets and foreground removal for SKAO hydrogen intensity maps Carmelita Carbone Abstract Intensity Mapping (IM) of the 21-cm line of the neutral hydrogen ( Hi ) has become a compelling new technique to map the large-scale structure of the Universe. One of the main challenges is the presence of strong foreground emissions of several orders of magnitude larger than the Hi signal. Here, we implement a version of the Principal Component Analysis, a blind component-separation technique, based on a kind of spherical wavelets called needlets. These functions exploit double localization both in real and in harmonic space. We test Need-PCA performances on a set of maps that simulates the SKA MID radio telescope in the AA4 configuration. We compare our results with other component separation methods such as Generalised Morphological Component Analysis (GMCA) and Generalized Needlet Internal Linear Combination (GNILC). All the methods have comparable results, recovering the Hi signal within 10% accuracy across the frequency channels, in the multipole range 30 ≲ ℓ ≲ \lesssim\ell\lesssim 136. We also test our pipeline in the presence of systematics such as polarization leakage. We find that the cleaning methods are insensitive to the presence of such systematic, yielding the same results as in the leakage-free case. 1 Introduction Detecting the 21-cm line emission from atomic neutral hydrogen ( Hi ) has emerged as a powerful observable in cosmology. Hi permeates the Universe from the Dark Ages following the recombination of the first atoms, through the epoch of reionization when the first stars formed, and into later times, where hydrogen is found in the gas clouds within galaxies [ 1 ] . At those late epochs, Hi is a valuable tracer of the galaxy distribution. However, the 21-cm - or, the frequency \nu_{21}= 1420 MHz - line, emitted by the hyperfine transition of Hi , is particularly weak: detecting a sufficiently large number of galaxies for cosmology is time- and resource-consuming, and becomes prohibitive at redshift larger than z\gtrsim 0.5 [ 2 ] . In the last two decades, Intensity Mapping (IM) has been developed as a new technique to efficiently map large portions of the sky at a relatively low cost. Instead of resolving individual galaxies, IM measures the sky’s intensity by integrating the 21-cm line emission over a large area [ 3 , 4 , 5 ] . This approach yields to a modest angular resolution in favour of high redshift resolution, made possible by the redshifting of the line. Similarly to the cosmic microwave background (CMB), the fluctuations in the observed temperature intensity of Hi emission act as a (biased) tracer of the underlying matter density fluctuations. Thus, IM of the Hi emission allows us to reconstruct matter density and constrain the matter power spectrum, which provides statistical information about the field on large scales and at different redshifts. The future of Hi cosmology is particularly promising, thanks to a growing number of large-scale 21-cm surveys that are are being conducted [ 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 ] or being proposed [ 14 , 15 , 16 , 17 ] . Among the latter, we stress how the SKA Observatory (SKAO) 1 1 1 https://www.skao.int/ . will become a game-changer in the field [ 18 ] . This work focuses on single-dish intensity mapping, which relies on the auto-correlation data of individual telescope dishes [ 19 , 20 ] , as opposed to standard radio interferometry that would require extremely short baselines to probe large-scale modes. To date, the Hi signal has only been detected via cross-correlation with galaxy surveys [ 21 , 22 , 9 , 13 , 23 ] , which helps mitigate systematics. IM observations are affected by strong astrophysical foregrounds, which are 3 – 5 orders of magnitude brighter than the Hi signal. Systematics and calibration errors further complicate foreground removal [ 24 , 25 , 26 ] . The main sources of foreground in the radio range are the Galactic synchrotron (cosmic-ray electrons accelerated by the magnetic field in the galactic medium), free-free emission (sourced by free electrons scattering off ions in our Galaxy), and extragalactic point sources (e.g. active galactic nuclei and radio galaxies in general). While these foregrounds are stronger in amplitude, their spectra are typically smooth in frequency, unlike the Hi signal, which has near-Gaussian oscillations [ 27 ] . Thus, in principle, it is possible to separate the foreground contributions from the cosmological signal using removal methods. In practice, instrumental effects complicate this separation. Of particular concern is the leakage of polarized emission into the intensity signal due to instrumental miscalibration [ 24 , 25 ] . If the instrument is partially miscalibrated, a fraction of the polarized synchrotron emission can leak into the observation and dominate the Hi signal. Moreover, polarization leakage has a non-smooth frequency dependence, making it hard to distinguish from the cosmological signal. Several foreground cleaning methods have been proposed, many exploiting the spectral smoothness of foregrounds. These methods differ primarily in whether they assume a model for the foregrounds. The so-called blind methods do not make any assumption on the foregrounds. Common blind methods include the the Principal Component Analysis ( [ 28 ] ), Generalized Morphological Component Analysis ( [ 25 ] ), Generalized Internal Linear Combination ( [ 29 , 30 ] ), Independent Component Analysis ( [ 31 ] ). For example, PCA has been applied in the analysis of MeerKAT intensity maps cross-correlated with optical galaxies from the WiggleZ Dark Energy Survey ( [ 9 , 13 ] ). In this study, we optimize a foreground cleaning pipeline to recover the Hi signal for SKAO-MID in the AA4 configuration. We use simulated maps of both the cosmological signal and foregrounds, including instrumental effects such as the thermal noise and the telescope beam smoothing. Our method is based on a version of PCA implemented with a kind of spherical wavelets called needlets [ 32 , 33 , 34 , 35 ] . Needlets enjoy some features which are shared by other wavelets, such as double localization in pixel and harmonic space, simple reconstruction formula after decomposition, uncorrelation properties which make them useful when dealing with incomplete data. Additionally, we assess the performance of the cleaning method in the presence of polarization leakage. We analyse its impact on signal recovery for SKAO-like simulations and evaluate the robustness of the pipeline under such systematics. The paper is structured as follows. In Sec. 2 we describe the formalism of the different foreground cleaning methods and the assumptions assumed by PCA, GMCA and GNILC. In Sec. 3 we describe the simulations and the full data-cube we use in our analysis. In Sec. 4 we summarize the properties of the needlets. In Sec. 5 we describe the application of the different cleaning methods to the simulated data and assess their performances. Sec. 6 shows the obtained results and the discussion. We conclude in Section 7 . 2 Foreground cleaning methods In this section we review some of the most used and studied methods for the 21-cm foreground cleaning. All methods rely on the assumptions that the foreground emissions are slowly varying with frequency, i.e. they are highly correlated,