Matter bispectrum of large-scale structure with Gaussian and non-Gaussian initial conditions: Halo models, perturbation theory, and a three-shape model

TL;DR

This paper compares models of the matter bispectrum in large-scale structure with N-body simulations, introduces a new three-shape model that fits the data across scales and redshifts, and extends it to primordial non-Gaussianity.

Contribution

It develops a new three-shape bispectrum model calibrated to simulations and applicable to non-Gaussian initial conditions, improving over existing models.

Findings

Effective field theory extends validity on intermediate scales.

Halo model underestimates power on intermediate scales and in the squeezed component.

The three-shape model accurately fits simulations across scales and redshifts.

Abstract

We study the matter bispectrum of large-scale structure by comparing the predictions of different perturbative and phenomenological models with the full three-dimensional bispectrum from $N$-body simulations estimated using modal methods. We show that among the perturbative approaches, effective field theory succeeds in extending the range of validity furthest on intermediate scales, at the cost of free additional parameters. By studying the halo model, we show that although it is satisfactory in the deeply non-linear regime, it predicts a deficit of power on intermediate scales, worsening at redshifts $z>0$. By comparison with the $N$-body bispectrum on those scales, we show that there is a significant squeezed component underestimated in the halo model. On the basis of these results, we propose a new three-shape model, based on the tree-level, squeezed and constant bispectrum shapes we identified in the halo model; after calibration this fits the simulations on all scales and redshifts of interest. We extend this model further to primordial non-Gaussianity of the local and equilateral types by showing that the same shapes can be used to describe the additional non-Gaussian component in the matter bispectrum. This method provides a HALOFIT-like prototype of the bispectrum that could be used to describe and test parameter dependencies and should be relevant for the bispectrum of weak gravitational lensing and wider applications.