Primordial non-Gaussianity from the DBI Galileons
Shuntaro Mizuno, Kazuya Koyama
TL;DR
The paper develops a generalized DBI Galileon framework via a P(X,φ)−G(X,φ) action to study primordial fluctuations during inflation. It derives the power spectrum and a bispectrum that includes a novel Galileon-induced shape, showing the equilateral f_NL estimator remains a robust probe thanks to strong shape overlap. For DBI Galileons, f_NL^equil scales like 1/c_s^2 with bounds -0.32/c_s^2 < f_NL^equil < -0.16/c_s^2, while a near-b_D–independent relation links f_NL^equil to (1−n_s) and r; in G-inflation, f_NL^equil = 4.62 r^−2/3, imposing tight constraints from tensor mode measurements. Overall, the work highlights how Galileon terms modify inflationary predictions, including a broken consistency relation and enhanced non-Gaussianity for small sound speeds, with implications for string-theory embeddings and future observations.
Abstract
We study primordial fluctuations generated during inflation in a class of models motivated by the DBI Galileons, which are extensions of the DBI action that yield second order field equations. This class of models generalises the DBI Galileons in a similar way with K-inflation. We calculate the primordial non-Gaussianity from the bispectrum of the curvature perturbations at leading order in the slow-varying approximations. We show that the estimator for the equilateral-type non-Gaussianity, $f_{\rm NL} ^{equil}$, can be applied to measure the amplitude of the primordial bispectrum even in the presence of the Galileon-like term although it gives a slightly different momentum dependence from K-inflation models. For the DBI Galileons, we find $-0.32 /c_s^2 < f_{\rm NL} ^{equil} < -0.16/c_s^2$ and large primordial non-Gaussianities can be obtained when $c_s$ is much smaller than 1 as in the usual DBI inflation. In G-inflation models, where a de Sitter solution is obtained without any potentials, the non-linear parameter is given by $f_{\rm NL}^{equil} = 4.62 r^{-2/3}$ where $r$ is the tensor to scalar ratio, giving a stringent constraint on the model.
