Modelling redshift-space distortion in the post-reionization ${\rm HI}$ 21-cm power spectrum

TL;DR

This paper models the redshift-space distortion effects on the post-reionization HI 21-cm power spectrum, incorporating bias and velocity dispersion to improve predictions for cosmological studies.

Contribution

It extends previous real-space HI 21-cm simulations by including redshift space distortions with new prescriptions for bias and velocity damping profiles.

Findings

Models with scale-dependent bias and Lorentzian damping fit simulations well.

Best-fit velocity dispersion scales as (1+z)^{-m} with m=2 and 1.2.

Predictions match simulations for k<0.3 Mpc^{-1} at all z, and at lower z for larger k.

Abstract

The post-reionization ${\rm HI}$ 21-cm signal is an excellent candidate for precision cosmology, this however requires accurate modelling of the expected signal. Sarkar et al. (2016) have simulated the real space ${\rm HI}$ 21-cm signal, and have modelled the ${\rm HI}$ power spectrum as $P_{{\rm HI}}(k)=b^2 P(k)$ where $P(k)$ is the dark matter power spectrum and $b(k)$ is a (possibly complex) scale dependent bias for which fitting formulas have been provided. This paper extends these simulations to incorporate redshift space distortion and predict the expected redshift space ${\rm HI}$ 21-cm power spectrum $P^s_{{\rm HI}}(k_{\perp},k_{\parallel})$ using two different prescriptions for the ${\rm HI}$ distributions and peculiar velocities. We model $P^s_{{\rm HI}}(k_{\perp},k_{\parallel})$ assuming that it is the product of $P_{{\rm HI}}(k)=b^2 P(k)$ with a Kaiser enhancement term and a Finger of God (FoG) damping which has $\sigma_p$ the pair velocity dispersion as a free parameter. Considering several possibilities for the bias and the damping profile, we find that the models with a scale dependent bias and a Lorentzian damping profile best fit the simulated $P^s_{{\rm HI}}(k_{\perp},k_{\parallel})$ over the entire range $1 \le z \le 6$. The best fit value of $\sigma_p$ falls approximately as $(1+z)^{-m}$ with $m=2$ and $1.2$ respectively for the two different prescriptions. The model predictions are consistent with the simulations for $k < 0.3 \, {\rm Mpc}^{-1}$ over the entire $z$ range for the monopole $P^s_0(k)$, and at $z \le 3$ for the quadrupole $P^s_2(k)$. At $z \ge 4$ the models underpredict $P^s_2(k)$ at large $k$, and the fit is restricted to $k < 0.15 \, {\rm Mpc}^{-1}$.