While limited to low spatial resolution, the next‐generation low‐frequency radio interferometers that target 21‐cm observations during the era of reionization and prior will have instantaneous fields of view that are many tens of deg2 on the sky. Predictions related to various statistical measurements of the 21‐cm brightness temperature must then be pursued with numerical simulations of reionization with correspondingly large volume box sizes, of the order of 1000 Mpc on one side. We pursue a semi‐numerical scheme to simulate the 21‐cm signal during and prior to reionization by extending a hybrid approach where simulations are performed by first laying down the linear dark matter density field, accounting for the non‐linear evolution of the density field based on second‐order linear perturbation theory as specified by the Zel'dovich approximation, and then specifying the location and mass of collapsed dark matter haloes using the excursion‐set formalism. The location of ionizing sources and the time evolving distribution of ionization field is also specified using an excursion‐set algorithm. We account for the brightness temperature evolution through the coupling between spin and gas temperature due to collisions, radiative coupling in the presence of Lyman α photons and heating of the intergalactic medium, such as due to a background of X‐ray photons. The hybrid simulation method we present is capable of producing the required large volume simulations with adequate resolution in a reasonable time, so a large number of realizations can be obtained with variations in assumptions related to astrophysics and background cosmology that govern the 21‐cm signal.

We explore the use of different radio galaxy populations as tracers of different mass halos and therefore, with different bias properties, to constrain primordial non-Gaussianity of the local type. We perform a Fisher matrix analysis based on the predicted auto and cross angular power spectra of these populations, using simulated redshift distributions as a function of detection flux and the evolution of the bias for the different galaxy types (Star forming galaxies, Starburst galaxies, Radio-Quiet Quasars, FRI and FRII AGN galaxies). We show that such a multi-tracer analysis greatly improves the information on non-Gaussianity by drastically reducing the cosmic variance contribution to the overall error budget. By using this method applied to future surveys, we predict a constraint of σ f nl = 3.6 on the local non-Gaussian parameter for a galaxy detection flux limit of 10µJy and σ f nl = 2.2 for 1 µJy. We show that this significantly improves on the constraints obtained when using the whole undifferentiated populations (σ f nl = 48 for 10 µJy and σ f nl = 12 for 1 µJy). We conclude that continuum radio surveys alone have the potential to constrain primordial non-Gaussianity to an accuracy at least a factor of two better than the present constraints obtained with Planck data on the CMB bispectrum, opening a window to obtain σ f nl ∼ 1 with the Square Kilometer Array.

The large-scale structure of the Universe supplies crucial information about the physical processes at play at early times. Unresolved maps of the intensity of 21 cm emission from neutral hydrogen HI at redshifts z=/~1-5 are the best hope of accessing the ultralarge-scale information, directly related to the early Universe. A purpose-built HI intensity experiment may be used to detect the large scale effects of primordial non-Gaussianity, placing stringent bounds on different models of inflation. We argue that it may be possible to place tight constraints on the non-Gaussianity parameter f(NL), with an error close to σ(f(NL))~1.