Fast computation of the non-Gaussian covariance of redshift-space galaxy power spectrum multipoles

TL;DR

This paper presents a rapid, analytical method to compute the non-Gaussian covariance matrix of galaxy power spectrum multipoles in redshift space, significantly speeding up calculations for cosmological analyses.

Contribution

It introduces an FFTLog-based analytical approach to efficiently compute the non-Gaussian covariance at tree-level, enabling fast evaluations across various parameters without parallelization.

Findings

Computation time is reduced to about 10 seconds for multiple parameters.

Achieves 0.1-1% accuracy in covariance calculations across relevant scales.

Applicable to future galaxy surveys and multi-tracer analyses.

Abstract

The non-Gaussian part of the covariance matrix of the galaxy power spectrum involves the connected four-point correlation in Fourier space, i.e. trispectrum. This paper introduces a fast method to compute the non-Gaussian part of the covariance matrix of the galaxy power spectrum multipoles in redshift space at tree-level standard perturbation theory. For the tree-level galaxy trispectrum, the angular integral between two wavevectors can be evaluated analytically by employing an FFTLog. The new implementation computes the non-Gaussian covariance of the power spectrum monopole, quadrupole, hexadecapole and their cross-covariance in $O(10)$ seconds, for an effectively arbitrary number of instances of cosmological and galaxy bias parameters and redshift, without any parallelization or acceleration. It is a large advantage over conventional numerical integration. We demonstrate that the computation of the covariance at $k = 0.005 - 0.4\,h\,\mathrm{Mpc}^{-1}$ gives results with $0.1 - 1\%$ accuracy. The efficient computation of the analytic covariance can be useful for future galaxy surveys, especially utilizing multi-tracer analysis.