Removable Matter-Power-Spectrum Covariance from Bias Fluctuations

TL;DR

This paper presents a simple, accurate model for the covariance matrix of the real-space matter power spectrum on slightly nonlinear scales, highlighting a method to remove non-Gaussian covariance by normalizing overdensities.

Contribution

The authors introduce a one-parameter model for the covariance matrix and demonstrate a normalization technique to nearly diagonalize it, improving analysis of nonlinear scales.

Findings

The model accurately describes covariance on scales k~0.1-0.8 h/Mpc at z=0.

Normalizing overdensity by its cell variance reduces non-Gaussian covariance.

The covariance model depends mainly on the variance of the nonlinear density field.

Abstract

We find a simple, accurate model for the covariance matrix of the real-space cosmological matter power spectrum on slightly nonlinear scales (k~0.1-0.8 h/Mpc at z=0), where off-diagonal matrix elements become substantial. The model includes a multiplicative, scale-independent modulation of the power spectrum. It has only one parameter, the variance (among realizations) of the variance of the nonlinear density field in cells, with little dependence on the cell size between 2-8 Mpc/h. Furthermore, we find that this extra covariance can be modeled out by instead measuring the power spectrum of (delta/sigma_cell), i.e. the ratio of the overdensity to its dispersion in cells a few Mpc in size. Dividing delta by sigma_cell essentially removes the non-Gaussian part of the covariance matrix, nearly diagonalizing it.