Logarithmic divergences in the $k$-inflationary power spectra computed through the uniform approximation

TL;DR

This paper examines the uniform approximation method for calculating power spectra in $k$-inflation, revealing logarithmic divergences at higher orders that limit the method's applicability and cautioning against over-reliance on existing results.

Contribution

It identifies the order limitations of the uniform approximation in $k$-inflation power spectra calculations and highlights the divergence issues affecting higher-order predictions.

Findings

Power spectra are reliable up to second order.

Tensor-to-scalar ratio is reliable up to third order.

Spectral indices and running converge to all orders.

Abstract

We investigate a calculation method for solving the Mukhanov-Sasaki equation in slow-roll $k$-inflation based on the uniform approximation (UA) in conjunction with an expansion scheme for slow-roll parameters with respect to the number of $e$-folds about the so-called \textit{turning point}. Earlier works on this method has so far gained some promising results derived from the approximating expressions for the power spectra among others, up to second order with respect to the Hubble and sound flow parameters, when compared to other semi-analytical approaches (e.g., Green's function and WKB methods). However, a closer inspection is suggestive that there is a problem when higher-order parts of the power spectra are considered; residual logarithmic divergences may come out that can render the prediction physically inconsistent. Looking at this possibility, we map out up to what order with respect to the mentioned parameters several physical quantities can be calculated before hitting a logarithmically divergent result. It turns out that the power spectra are limited up to second order, the tensor-to-scalar ratio up to third order, and the spectral indices and running converge to all orders. This indicates that the expansion scheme is incompatible with the working equations derived from UA for the power spectra but compatible with that of the spectral indices. For those quantities that involve logarithmically divergent terms in the higher-order parts, existing results in the literature for the convergent lower-order parts calculated in the equivalent fashion should be viewed with some caution; they do not rest on solid mathematical ground.