The Dark Energy Survey Supernova Program: Investigating Beyond-$\Lambda$CDM

TL;DR

This study uses five years of DES supernova data to test various non-standard cosmological models, finding no strong evidence against them and suggesting more flexible models may be needed beyond the cosmological constant.

Contribution

It provides a comprehensive analysis of non-standard cosmological models using DES supernova data and evaluates the impact of assumptions in the analysis process.

Findings

No strong evidence for or against non-standard models from DES data.

11 out of 15 models are moderately preferred over Flat-$\\Lambda$CDM when combined with external probes.

Bias from approximate cosmological models is sub-dominant to statistical uncertainties, with a proposed method to reduce it.

Abstract

We report constraints on a variety of non-standard cosmological models using the full 5-year photometrically-classified type Ia supernova sample from the Dark Energy Survey (DES-SN5YR). Both Akaike Information Criterion (AIC) and Suspiciousness calculations find no strong evidence for or against any of the non-standard models we explore. When combined with external probes, the AIC and Suspiciousness agree that 11 of the 15 models are moderately preferred over Flat-$\Lambda$CDM suggesting additional flexibility in our cosmological models may be required beyond the cosmological constant. We also provide a detailed discussion of all cosmological assumptions that appear in the DES supernova cosmology analyses, evaluate their impact, and provide guidance on using the DES Hubble diagram to test non-standard models. An approximate cosmological model, used to perform bias corrections to the data holds the biggest potential for harbouring cosmological assumptions. We show that even if the approximate cosmological model is constructed with a matter density shifted by $\Delta\Omega_m\sim0.2$ from the true matter density of a simulated data set the bias that arises is sub-dominant to statistical uncertainties. Nevertheless, we present and validate a methodology to reduce this bias.