Sigma-Eight at the Percent Level: The EFT Likelihood in Real Space

TL;DR

This paper develops a real-space effective field theory likelihood for galaxy clustering, enabling unbiased cosmological parameter inference with improved accuracy and the ability to incorporate survey effects.

Contribution

It introduces a real-space formulation of the EFT likelihood with a Lagrangian forward model, allowing for unbiased $\sigma_8$ inference and better handling of survey window functions.

Findings

Achieves 2% accuracy in $\sigma_8$ inference for certain cutoff values.

Demonstrates convergence to ground truth in the low-cutoff limit.

Improves upon previous EFT likelihood results by including survey effects.

Abstract

The effective field theory likelihood for the density field of biased tracers allows for cosmology inference from the clustering of galaxies that consistently uses all available information at a given order in perturbation theory. This paper presents results and implementation details on the real-space (as opposed to Fourier-space) formulation of the likelihood, which allows for the incorporation of survey window functions. The implementation further uses a Lagrangian forward model for biased tracers which automatically accounts for all relevant contributions up to any desired order. Unbiased inference of $\sigma_8$ is demonstrated at the 2% level for cutoff values $\Lambda \leq 0.14h\,{\rm Mpc}^{-1}$ for halo samples over a range of masses and redshifts. The inferred value shows the expected convergence to the ground truth in the low-cutoff limit. Apart from the possibility of including observational effects, this represents further substantial improvement over previous results based on the EFT likelihood.