Bayesian evidence of the post-Planck curvaton

We perform a Bayesian model comparison for scenarios within the quadratic curvaton model, determining the degree to which both are disfavoured with respect to the $\Lambda$CDM concordance model and single-field quadratic inflation, using the recent \emph{Planck} data release. Despite having three additional model parameters, the simplest curvaton scenario is not disfavoured relative to single-field quadratic inflation, and it becomes favoured against this single-field model when we include the joint BICEP/Keck/\emph{Planck} analysis. In all cases we assume an instantaneous inflaton decay and no surviving isocurvature perturbations. Despite the success of \emph{Planck} reaching its forecast measurement accuracy, we show that the current constraints on local non-Gaussianity are insufficiently precise to have any significant impact on the evidence ratios so far. We also determine the precision $\sigma(f_{\mathrm{NL}})$ required by future measurements assuming a fiducial value of $f_{\mathrm{NL}}=-5/4$ or $10.8$ to no longer disfavour the curvaton against the $\Lambda$CDM parametrisation, and we discuss the effect that the predicted increase in precision from future measurements on $f_{\mathrm{NL}}$ may have. We show that our results are not very sensitive to our choice of priors.