Influence of Reheating on the Trispectrum and its Scale Dependence

TL;DR

This paper investigates how reheating affects the evolution of the trispectrum parameters and scale dependence in multifield inflation models, highlighting the importance of reheating details for observational predictions.

Contribution

It provides a detailed analysis of the evolution of trispectrum parameters during reheating and establishes relations between observables that remain valid regardless of reheating specifics.

Findings

Large trispectrum at end of inflation generally remains large after reheating.

Generating a very large $ aunl$ compared to $nl^2$ is difficult in studied models.

The relation $ aunl eq (6/5 nl)^2$ persists through reheating.

Abstract

We study the evolution of the non-linear curvature perturbation during perturbative reheating, and hence how observables evolve to their final values which we may compare against observations. Our study includes the evolution of the two trispectrum parameters, $\gnl$ and $\taunl$, as well as the scale dependence of both $\fnl$ and $\taunl$. In general the evolution is significant and must be taken into account, which means that models of multifield inflation cannot be compared to observations without specifying how the subsequent reheating takes place. If the trispectrum is large at the end of inflation, it normally remains large at the end of reheating. In the classes of models we study, it is very hard to generate $\taunl\gg\fnl^2$, regardless of the decay rates of the fields. Similarly, for the classes of models in which $\gnl\simeq\taunl$ during slow--roll inflation, we find the relation typically remains valid during reheating. Therefore it is possible to observationally test such classes of models without specifying the parameters of reheating, even though the individual observables are sensitive to the details of reheating. It is hard to generate an observably large $\gnl$ however. The runnings, $\nfnl$ and $\ntaunl$, tend to satisfy a consistency relation $\ntaunl=(3/2)\nfnl$, but are in general too small to be observed for the class of models considered regardless of reheating timescale.