Non-Gaussian tails without stochastic inflation

Abstract

We show, both analytically and numerically, that non-Gaussian tails in the probability density function of curvature perturbations arise in ultra-slow-roll inflation from the δN formalism, without invoking stochastic inflation. Previously reported discrepancies between both approaches are a consequence of not correctly accounting for momentum perturbations. Once they are taken into account, both approaches agree to an excellent degree. The shape of the tail depends strongly on the phase space of inflation.