BAO scale inference from biased tracers using the EFT likelihood

TL;DR

This paper demonstrates that forward modeling with EFT likelihood can infer the BAO scale from biased tracers with minimal bias and smaller errors compared to standard methods, enhancing high-precision cosmology.

Contribution

First application of EFT likelihood with forward modeling to BAO inference from halo catalogs, reducing systematic bias and improving precision over traditional power spectrum methods.

Findings

Systematic bias less than 2% for cutoff Λ ≤ 0.25 h/Mpc.

Field-level inference yields 1.1 to 3.3 times smaller errors than standard power spectrum analysis.

Bias remains below 1% for most biased halo samples.

Abstract

The physical scale corresponding to baryon acoustic oscillations (BAO), the size of the sound horizon at recombination, is precisely determined by CMB experiments. Measuring the apparent size of the BAO scale imprinted in the clustering of galaxies gives us a direct estimate of the angular-diameter distance and the Hubble parameter as a function of redshift. The BAO feature is damped by non-linear structure formation, which reduces the precision with which we can infer the BAO scale from standard galaxy clustering analysis methods. Many methods to undo this damping via the so-called BAO reconstruction have so far been proposed; however, they all rely on backward modeling. In this paper, we present the first results of BAO inference from rest-frame halo catalogs using forward modeling combined with the EFT likelihood, in the case where the initial phases of the density field are fixed. We show that the remaining systematic bias is less than 2% when we consider cutoff values of $\Lambda \leq 0.25 \,h\,{\rm Mpc}^{-1}$ for all halo samples considered, and below 1% and consistent with zero for all but the most highly biased samples. We also demonstrate that, when compared to the standard power spectrum likelihood approach under the same assumption of fixed phases, the 1$\sigma$ errors associated to the field level inference of the BAO scale are 1.1 to 3.3 times smaller, depending on the value of the cutoff and the halo sample. Our analysis therefore unveils another promising feature of using field-level inference for high-precision cosmology.