Fast, large volume, GPU enabled simulations for the Ly-alpha forest: power spectrum forecasts for baryon acoustic oscillation experiments

TL;DR

This paper introduces a fast, GPU-enabled semi-analytic simulation method for large-volume Ly-alpha forest data, enabling accurate BAO scale forecasts crucial for dark energy research in upcoming surveys.

Contribution

The authors develop a rapid, large-volume simulation approach for the Ly-alpha forest that accurately reproduces key observational features and BAO signals, facilitating survey planning and systematic studies.

Findings

Simulations match hydrodynamical references and observations.

BAO scale can be recovered with ~1.4% accuracy in survey-like conditions.

Power-law estimates for BAO measurement errors are provided.

Abstract

High redshift measurements of the baryonic acoustic oscillation scale (BAO) from large Ly-alpha forest surveys represent the next frontier of dark energy studies. As part of this effort, efficient simulations of the BAO signature from the Ly-alpha forest will be required. We construct a model for producing fast, large volume simulations of the Ly-alpha forest for this purpose. Utilising a calibrated semi-analytic approach, we are able to run very large simulations in 1 Gpc^3 volumes which fully resolve the Jeans scale in less than a day on a desktop PC using a GPU enabled version of our code. The Ly-alpha forest spectra extracted from our semi-analytical simulations are in excellent agreement with those obtained from a fully hydrodynamical reference simulation. Furthermore, we find our simulated data are in broad agreement with observational measurements of the flux probability distribution and 1D flux power spectrum. We are able to correctly recover the input BAO scale from the 3D Ly-alpha flux power spectrum measured from our simulated data, and estimate that a BOSS-like 10^4 deg^2 survey with ~15 background sources per square degree and a signal-to-noise of ~5 per pixel should achieve a measurement of the BAO scale to within ~1.4 per cent. We also use our simulations to provide simple power-law expressions for estimating the fractional error on the BAO scale on varying the signal-to-noise and the number density of background sources. The speed and flexibility of our approach is well suited for exploring parameter space and the impact of observational and astrophysical systematics on the recovery of the BAO signature from forthcoming large scale spectroscopic surveys.