Quantifying and mitigating the effect of snapshot interval in light-cone Epoch of Reionization 21-cm simulations

TL;DR

This paper investigates the errors caused by finite snapshot intervals in light-cone 21-cm simulations of the Epoch of Reionization and proposes an efficient interpolation method to reduce these errors without high computational costs.

Contribution

It introduces a novel interpolation-based approach to mitigate snapshot interval errors in 21-cm light-cone simulations, reducing the need for numerous costly reionization snapshots.

Findings

Errors decrease as the inverse of snapshot number, N_RS^{-1}

Interpolation method effectively reduces discontinuities at stitching boundaries

Achieves accurate simulations with fewer snapshots, saving computational resources

Abstract

The Epoch of Reionization (EoR) neutral Hydrogen (HI) 21-cm signal evolves significantly along the line-of-sight (LoS) due to the light-cone (LC) effect. It is important to accurately incorporate this in simulations in order to correctly interpret the signal. The 21-cm LC simulations are typically produced by stitching together slices from a finite number $(N_{\rm RS})$ of "reionization snapshot'', each corresponding to a different stage of reionization. In this paper, we have quantified the errors in the 21-cm LC simulation due to the finite value of $N_{\rm RS}$. We show that this can introduce large discontinuities $(> 200 \%)$ at the stitching boundaries when $N_{\rm RS}$ is small $(=2,4)$ and the mean neutral fraction jumps by $\delta \bar{x}_{\rm HI} =0.2,0.1$ respectively at the stitching boundaries. This drops to $17 \%$ for $N_{\rm RS}=13$ where $\delta \bar{x}_{\rm HI}=0.02$. We find that we can achieve $\delta \bar{x}_{\rm HI} \le 0.01$ with $N_{\rm RS} =26$, and we use this as the reference for comparing the other simulations. We present and also validate a method for mitigating this error by increasing $N_{\rm RS}$ without a proportional increase in the computational costs which are mainly incurred in generating the dark matter and halo density fields. Our method generates these fields only at a few redshifts, and interpolates them to generate reionization snapshots at closely spaced redshifts. We use this to generate 21-cm LC simulations with $N_{\rm RS}=51,101$ and $201$, and show that the errors go down as $N_{\rm RS}^{-1}$.