Mapping the integrated emission of the 21-cm line of neutral hydrogen (HI) from all the galaxies at a given redshift, a technique known as HI intensity mapping is the new frontier to probe the cosmic large-scale structure. Indeed, it consists of treating the 21 cm signal as a diffuse background instead of measuring the 21 cm emission of each galaxy separately. Thus, it is less costly, less time consuming and it allows to treat a large cosmological volume. It has a great potential for cosmology, as thanks to its exquisite redshift resolution it allows us to track the evolution of cosmic structures across time in a way unparalleled by traditional galaxy redshift surveys. Furthermore, the study of individual galaxies is not required for the study of the large-scale structures of the Universe.
Objectives:
The project consists of forecasting the capability of such advanced higher-order statistics for HI intensity mapping, in the view of the upcoming SKA Observatory. By applying these new statistical methods to simulated HI intensity mapping data for the first time. To this purpose, it will also be necessary to modify the existing techniques to account for the peculiarities of the HI intensity mapping signal, like its superb redshift resolution, opposite to its rather poor angular resolution (compared to optical surveys). Moreover, the telescope beam plays an important role, and it will have to be modeled accurately to extract most of the information. All of this will be done with the most advanced techniques of machine learning and data analysis of cosmological maps. The goal is to forecast constraints on cosmological parameters as per the planned SKA Observatory HI intensity mapping survey, which will cover 5000 to 15,000 square degrees.
Methodology:
Standard methods for deriving cosmological information from HI intensity mapping is via measurements of its power spectrum, analogously to what done with galaxy redshift surveys. However, it is well known that the power spectrum does not contain all the cosmological information, as the nonlinear evolution of cosmic structures makes the density field non-Gaussian. Therefore it is necessary to use beyond second order statistics which are sensitive to non-gaussianities that need to be considered when studying the Universes after Early times. Higher-order statistics are routinely used in galaxy surveys and the most standard ones, like the bispectrum, have been proposed for HI intensity mapping too. However, this is not the case for the most advanced techniques, like wavelet l1-norm statistics or machine learning based summary statistics. The method is to simulate the 21 cm signal map and power spectrum in order using existing and publicly available codes like 21CM FAST a code for generating fast simulations of the cosmological 21cm signal. Then the foregrounds, the noise and the SKA characteristics will be added to the simulation in order to be closer to real measurements. The ultimate goal is to be able to constrain cosmological parameters, e.g. baryon density Ωb, matter density Ωm, dark energy density ΩΛ, σ8, primordial non-gausianities fNL, neutrino masses ∑mν, Hubble constant H0.