Foreground removal and 21 cm signal estimates: comparing different blind methods for the BINGO Telescope
Alessandro Ribeiro Marins, Filipe Batoni Abdalla, Elcio Abdalla A, Chang Feng, Luiza O. Ponte, Giancarlo de Gasperis, Luiz H.F. Assis, Mathieu Remazeilles, Carlos Alexandre Wuensche, Luciano Barosi, Edmar C. Gurjão, Thyrso Villela, Bin Wang, Jiajun Zhang, Ricardo G. Landim, Vincenzo Liccardo, Camila P. Novaes, Amilcar R. de Queiroz, Larissa C. O. Santos, Marcelo V. dos Santos · Journal of Cosmology and Astroparticle Physics · 2026
Abstract The BINGO radiotelescope will map the large-scale distribution of neutral hydrogen using Intensity Mapping (IM) to analyze the Dark Energy paradigm through Baryon Acoustic Oscillations. The target signal is contaminated by unwanted signals and instrumental noise, making accurate estimations essential for characterizing the 21 cm signal. In this work, we evaluated the performance of three blind foreground-removing algorithms — FastICA, GNILC, and GMCA — in the BINGO simulated sky. Using multifrequency simulations including foreground emission and thermal noise, we estimate the H i angular power spectrum through a debiasing procedure for amplitude correction, jackknife covariance estimates, and χ 2 and signal-to-noise diagnostics. Our findings indicate that, across the multipole range 2 ≤ ℓ ≤ 300, the three methods produce statistically consistent reconstructions of the H i signal, with no method statistically outperforming the others. Motivated by this result and by its lower computational cost, we apply FastICA to maps generated from BINGO time-ordered-data simulations using HIDE. For five years of simulated observations, we recover the H i angular power spectrum with overall statistics χ 2 overall = 0.65 and SNR overall = 203 over 2 ≤ ℓ ≤ 300, after excluding the first frequency channel, which is the most affected by edge effects. These results show that blind foreground removal, combined with a debiasing correction, can reliably recover the H i angular power spectrum from controlled BINGO-like observations on angular scales relevant for BAO studies.