Projected two- and three-point statistics: Forecasts and mitigation of non-linear RSDs

TL;DR

This paper explores how modeling non-linear redshift-space distortions affects the accuracy and precision of galaxy clustering measurements, demonstrating that improved models can recover most 3D power spectrum information and reduce biases.

Contribution

It provides an analysis of the bias-error trade-off in projected galaxy clustering statistics and shows that better modeling of non-linear distortions enhances information recovery.

Findings

Inflating error bars by 20% without proper modeling.

Reducing tomographic bin depth recovers over 99% of 3D power spectrum info.

Improved modeling mitigates biases in parameter estimation.

Abstract

The combination of two- and three-point clustering statistics of galaxies and the underlying matter distribution has the potential to break degeneracies between cosmological parameters and nuisance parameters and can lead to significantly tighter constraints on parameters describing the composition of the Universe and the dynamics of inflation. Here we investigate the relation between biases in the estimated parameters and inaccurate modelling of non-linear redshift-space distortions for the power spectrum and bispectrum of projected galaxy density fields and lensing convergence. Non-linear redshift-space distortions are one of the leading systematic uncertainties in galaxy clustering. Projections along the line of sight suppress radial modes and are thus allowing a trade-off between biases due to non-linear redshift-space distortions and statistical uncertainties. We investigate this bias-error trade-off for a CMASS-like survey with a varying number of redshift bins. Improved modelling of the non-linear redshift-space distortions allows the recovery of more radial information when controlling for biases. Not modelling non-linear redshift space distortions inflates error bars for almost all parameters by 20%. The information loss for the amplitude of local non-Gaussianities is smaller, since it is best constrained from large scales. In addition, we show empirically that one can recover more than 99% of the 3D power spectrum information if the depth of the tomographic bins is reduced to 10 $h^{-1}$Mpc.