Precise Calibration of the One-Loop Bispectrum in the Effective Field Theory of Large Scale Structure

Abstract

The bispectrum is the leading non-Gaussian statistic in Large-Scale Structure (LSS) clustering and encodes the interactions in the underlying field. It is thus an important diagnostic for primordial non-Gaussianity and higher order galaxy biasing. In this paper we present a detailed test and calibration of the matter bispectrum counterterms in the Effective Field Theory of LSS against a suite of $N$-body simulations. We are going beyond previous studies in employing realisation based perturbation theory that allows for a significant reduction in cosmic variance error bars. This enables the measurement of the low-energy constants on large scales before two-loop corrections become relevant, around $k<0.09 h\mathrm{Mpc}^{-1}$ at $z=0$. We also go beyond previous work in using bispectrum propagator terms, i.e. correlators with linear and second order fields, to quantify the two new counterterms in isolation and to establish consistency with the power spectrum counterterm. By investigating the fully non-linear bispectrum, $B_{\mathrm{nnn}}$, as well as the terms $B_{\mathrm{n}11}$ and $B_{\mathrm{n}21}$, we find evidence for the new counterterms deviating from the shape suggested by the UV-limit of the relevant bispectrum contributions. We also show that the commonly used Einstein-de Sitter approximation for the time dependence of the tree-level bispectrum is insufficient for precise studies of the one-loop bispectrum and that it is necessary to use $\Lambda$CDM growth factors in order to obtain meaningful one-loop counterterm constraints. Finally, we also find evidence for small deviations in the growth factors that arise from time integration inaccuracies in the $N$-body simulations.