The Effective Field Theory of Inflation Models with Sharp Features

TL;DR

This paper develops a model-independent framework within the Effective Field Theory of Inflation to analyze sharp features in the inflaton potential, predicting enhanced non-Gaussianity and providing bounds on feature sharpness.

Contribution

It introduces a parametrization of sharp features in inflation models using EFT, linking features to observable signatures in the power-spectrum and bispectrum, and derives bounds on feature sharpness.

Findings

Enhanced non-Gaussianity with sharp features

Power-spectrum most sensitive to features

Upper bound on feature sharpness parameter eta"

Abstract

We describe models of single-field inflation with small and sharp step features in the potential (and sound speed) of the inflaton field, in the context of the Effective Field Theory of Inflation. This approach allows us to study the effects of features in the power-spectrum and in the bispectrum of curvature perturbations, from a model-independent point of view, by parametrizing the features directly with modified "slow-roll" parameters. We can obtain a self-consistent power-spectrum, together with enhanced non-Gaussianity, which grows with a quantity $\beta$ that parametrizes the sharpness of the step. With this treatment it is straightforward to generalize and include features in other coefficients of the effective action of the inflaton field fluctuations. Our conclusion in this case is that, excluding extrinsic curvature terms, the only interesting effects at the level of the bispectrum could arise from features in the first slow-roll parameter $\epsilon$ or in the speed of sound $c_s$. Finally, we derive an upper bound on the parameter $\beta$ from the consistency of the perturbative expansion of the action for inflaton perturbations. This constraint can be used for an estimation of the signal-to-noise ratio, to show that the observable which is most sensitive to features is the power-spectrum. This conclusion would change if we consider the contemporary presence of a feature and a speed of sound $c_s < 1$, as, in such a case, contributions from an oscillating folded configuration can potentially make the bispectrum the leading observable for feature models.