Non-Gaussian likelihoods for Type Ia supernova cosmology: implications for dark energy and H0

ABSTRACT The latest improvements in the scale and calibration of Type Ia supernova (SN Ia) catalogues allow us to constrain the specific nature and evolution of dark energy through its effect on the expansion history of the Universe. We present the results of Bayesian model comparison on the SNe Ia catalogue Pantheon+, where flat Lambda cold dark matter ($\Lambda$CDM) is preferred by the data over all other models and we find moderate evidence ($\Delta \log \mathcal {Z} \sim 2.5$) to reject a number of alternate dark energy models. The effect of peculiar velocity corrections on model comparison is analysed, where we find that removing peculiar velocity corrections lessens the $\Delta \log \mathcal {Z}$ of each cosmology against $\Lambda$CDM and overall reduces the strength of $\Lambda$CDM in the model comparison. As well as comparing cosmological models, the Bayesian methodology is extended to testing for non-gaussianity in the distribution of the Pantheon+ Hubble residuals. We find that adding a scale parameter to the Pantheon+ covariances, or alternately using a multivariate Student’s t-distribution fits the data better than the fiducial analysis, producing a marginalized evidence increase of $\Delta \log \mathcal {Z} = 1.70$ and 2.70, respectively. This improved treatment of the scatter produces a 5 per cent decrease in the uncertainty of the Hubble constant as compared to SH0ES, finding $H_0 = 73.67 \pm 0.99$${\rm km}\,{\rm s}^{-1}\,{\rm Mpc}^{-1}$ in $5.6\sigma$ tension with Planck, guided by the evidence when marginalizing over both cosmological and scatter models. We also explore $M_B$ transition models as a potential solution for the Hubble tension, finding no evidence to support these models among the SN data.