Fast Computation of First-Order Feature-Bispectrum Corrections

TL;DR

This paper introduces an efficient method for computing first-order feature-bispectrum corrections in inflation models, accurately capturing large non-Gaussian effects caused by rapid potential features.

Contribution

It develops a simple, fast computational approach for first-order corrections in generalized slow-roll inflation, improving accuracy over previous methods.

Findings

Matches numerical computations within 1% for small features.

Achieves better than 10% accuracy for large features.

Provides a practical tool for analyzing non-Gaussianity in inflation models.

Abstract

Features in the inflaton potential that are traversed in much less than an e-fold of the expansion can produce observably large non-Gaussianity. In these models first order corrections to the curvature mode function evolution induce effects at second order in the slow roll parameters that are generically greater than ~ 10% and can reach order unity for order unity power spectrum features. From a complete first order expression in generalized slow-roll, we devise a computationally efficient method that is as simple to evaluate as the leading order one and implements consistency relations in a controlled fashion. This expression matches direct numerical computation for step potential models of the dominant bispectrum configurations to better than 1% when features are small and 10% when features are order unity.