Bayesian evidence of non-standard inflation: isocurvature perturbations and running spectral index

Bayesian model comparison penalizes models with more free parameters that are allowed to vary over a wide range, and thus offers the most robust method to decide whether some given data require new parameters. In this paper, we ask a simple question: do current cosmological data require extensions of the simplest single-field inflation models? Specifically, we calculate the Bayesian evidence of a totally anti-correlated isocurvature perturbation and a running spectral index of the scalar curvature perturbation. These parameters are motivated by recent claims that the observed temperature anisotropy of the cosmic microwave background on large angular scales is too low to be compatible with the simplest inflation models. Both a subdominant, anti-correlated cold dark matter isocurvature component and a negative running index succeed in lowering the large-scale temperature power spectrum. We show that the introduction of isocurvature perturbations is disfavored, whereas that of the running spectral index is only moderately favored, even when the BICEP2 data are included in the analysis without any foreground subtraction.