Many controlled realizations of chaotic inflation employ pseudo-scalar axions. Pseudo-scalars φ are naturally coupled to gauge fields through cφFF . In the presence of this coupling, gauge field quanta are copiously produced by the rolling inflaton. The produced gauge quanta, in turn, source inflaton fluctuations via inverse decay. These new cosmological perturbations add incoherently with the "vacuum" perturbations, and are highly nongaussian. This provides a natural mechanism to generate large nongaussianity in single or multi field slow-roll inflation. The resulting phenomenological signatures are highly distinctive: large nongaussianity of (nearly) equilateral shape, in addition to detectably large values of both the scalar spectral tilt and tensor-to-scalar ratio (both being typical of large field inflation). The WMAP bound on nongaussianity implies that the coupling c of the pseudo-scalar inflaton to any gauge field must be smaller than about 10 2 M −1 p .

Abstract. We consider a model in which a pseudo-scalar field σ rolls for some e-folds during inflation, sourcing one helicity of a gauge field. These fields are only gravitationally coupled to the inflaton, and therefore produce scalar and tensor primordial perturbations only through gravitational interactions. These sourced signals are localized on modes that exit the horizon while the roll of σ is significant. We focus our study on cases in which the model can simultaneously produce (i) a large gravitational wave signal, resulting in observable B-modes of the CMB polarizations, and (ii) sufficiently small scalar perturbations, so to be in agreement with the current limits from temperature anisotropies. Different choice of parameters can instead lead to a localized and visible departure from gaussianity in the scalar sector, either at CMB or LSS scales.arXiv:1509.07521v1 [astro-ph.CO]

We study the possibility that particle production during inflation could source observable gravity waves on scales relevant for Cosmic Microwave Background experiments. A crucial constraint on such scenarios arises because particle production can also source inflaton perturbations, and might ruin the usual predictions for a nearly scale invariant spectrum of nearly Gaussian curvature fluctuations. To minimize this effect, we consider two models of particle production in a sector that is only gravitationally coupled to the inflaton. For a single instantaneous burst of massive particle production, we find that localized features in the scalar spectrum and bispectrum might be observable, but gravitational wave signatures are unlikely to be detectable (due to the suppressed quadrupole moment of non-relativistic quanta) without invoking some additional effects. We also consider a model with a rolling pseudoscalar that leads to a continuous production of relativistic gauge field fluctuations during inflation. Here we find that gravitational waves from particle production can actually exceed the usual inflationary vacuum fluctuations in a regime where non-Gaussianity is consistent with observational limits. In this model observable B-mode polarization can be obtained for any choice of inflaton potential, and the amplitude of the signal is not necessarily correlated with the scale of inflation.

In any realistic particle physics model of inflation, the inflaton can be expected to couple to other fields. We consider a model with a dilaton-like coupling between a U(1) gauge field and a scalar inflaton. We show that this coupling can result in observable non-gaussianity, even in the conventional regime where inflation is supported by a single scalar slowly rolling on a smooth potential: the time dependent inflaton condensate leads to amplification of the large-scale gauge field fluctuations, which can feed-back into the scalar/tensor cosmological perturbations. In the squeezed limit, the resulting bispectrum is close to the local one, but it shows a sizable and characteristic quadrupolar dependence on the angle between the shorter and the larger modes in the correlation. Observable non-gaussianity is obtained in a regime where perturbation theory is under control. If the gauge field is identified with the electromagnetic field, the model that we study is a realization of the magnetogenesis idea originally proposed by Ratra, and widely studied. This identification (which is not necessary for the non-gaussianity production) is however problematic in light of a strong coupling problem already noted in the literature.