Detecting the HI Power Spectrum in the Post-Reionization Universe with SKA-Low (2302.11504v2)

Abstract: We present a survey strategy to detect the neutral hydrogen (HI) power spectrum at $5<z<6$ using the SKA-Low radio telescope in presence of foregrounds and instrumental effects. We simulate observations of the inherently weak HI signal post-reionization with varying levels of noise and contamination with foreground amplitudes equivalent to residuals after sky model subtraction. We find that blind signal separation methods on imaged data are required in order to recover the HI signal at large cosmological scales. Comparing different methods of foreground cleaning, we find that Gaussian Process Regression (GPR) performs better than Principle Component Analysis (PCA), with the key difference being that GPR uses smooth kernels for the total data covariance. The integration time of one field needs to be larger than $\sim 250$ h to provide large enough signal-to-noise ratio (SNR) to accurately model the data covariance for foreground cleaning. Images within the primary beam field-of-view give measurements of the HI power spectrum at scales $k\sim 0.02\,{\rm Mpc^{-1}}-0.3\,{\rm Mpc^{-1} }$ with SNR $\sim 2-5$ in $\Delta[{\rm log}( k/{\rm Mpc^{-1}})] = 0.25$ bins assuming an integration time of $600$ h. Systematic effects, which introduce small-scale fluctuations across frequency channels, need to be $\lesssim 5\times 10^{-5}$ to enable unbiased measurements outside the foreground wedge. Our results provide an important validation towards using the SKA-Low array for measuring the HI power spectrum in the post-reionization Universe.