Foreground removal and 21 cm signal estimates: comparing different blind methods for the BINGO Telescope

分类： 天文学 >> 天文学 提交时间： 2023-02-19

Alessandro Marins Filipe B. Abdalla Karin S. F. Fornazier Elcio Abdalla Luiz H. F. Assis Mathieu Remazeilles Carlos Alexandre Wuensche Luciano Barosi Amilcar R. Queiroz Thyrso Villela Bin Wang Chang Feng Ricardo Landim Vincenzo Liccardo Camila P. Novaes Larissa Santos Marcelo V. dos Santos Jiajun Zhang

摘要： BINGO will observe hydrogen distribution by means of the 21 cm line signal by drift-scan mapping through a tomographic analysis called \emph{Intensity Mapping} (IM) between 980 and 1260 MHz which aims at analyzing Dark Energy using \emph{Baryon Acoustic Oscillations}. In the same frequency range, there are several other unwanted signals as well as instrumental noise, contaminating the target signal. There are many component separation methods to reconstruct signals. Here, we used just three blind methods (FastICA, GNILC and GMCA), which explore different ways to estimate foregrounds' contribution from observed signals from the sky. Subsequently, we estimate 21 cm signal from its mixing with noise. We also analyzed how different number of simulations affect the quality of the estimation, as well as the effect of the binning on angular power spectrum to estimate 21 cm from the mixing with noise. For the BINGO sky range and sensitivity and the foreground model considered in the current simulation, we find that the effective dimension of the foreground subspace leading to best results is equal to three, composed of non-physical templates. At this moment of the pipeline configuration, using 50 or 400 simulations is statistically equivalent. It is also possible to reduce the number of multipoles by half to speed up the process and maintain the quality of results. All three algorithms used to perform foreground removal yielded statistically equivalent results for estimating the 21cm signal when we assume 400 realizations and GMCA and FastICA's mixing matrix dimensions equal to three. However, concerning computational cost in this stage of the BINGO pipeline, FastICA is faster than other algorithms. A new comparison will be necessary when the time-ordered-data and map-making are available.

同行评议状态:待评议

 点击量 16  下载量 13  评论