[Paper Review] Interpreting DESI 2024 BAO: late-time dynamical dark energy or a local effect?

Ioannis D. Gialamas, Gert Hütsi+5 more|arXiv (Cornell University)|Jun 11, 2024

Geophysics and Gravity Measurements9 citations

TL;DR

The paper analyzes DESI BAO, CMB, and SN data with linear, quintessence, and step-function dark energy models, finding that phantom indications from a linear w0-wa parametrization are artificial and that deviations from w = -1 arise mainly at very low redshift, potentially due to local effects or systematics in the nearby universe.

ABSTRACT

We perform fits to DESI, CMB and supernova data to understand the physical origin of the DESI hint for dynamical dark energy. We find that the linear parametrization of the equation of state $w$ may guide to misleading interpretations, such as the hint for a phantom Universe, which are not preferred by the data. Instead, physical quintessence models fit the data well. Model-independently, present observations prefer deviations from the constant dark energy, $w=-1$, only at very low redshifts, $z < \mathcal{O}(0.1)$. We find that this result is driven by low-$z$ supernova data. Therefore, either the fundamental properties of our Universe, characterised by the equation of state $w$ and the Hubble parameter $H$, underwent dramatic changes very recently or, alternatively, we do not fully understand the systematics of our local Universe in a radius of about $300\,h^{-1} m Mpc$.

Motivation & Objective

• Motivate understanding whether DESI hints of dynamical dark energy are physical or artefacts of parametrization.
• Test physically motivated dark energy models against DESI BAO, Planck CMB, and DES SN data.
• Identify redshift range where deviations from w = -1 occur and assess potential local/universal explanations.

Proposed method

• Fit DESI BAO, Planck CMB, and DES SN data to multiple dark energy parametrizations.
• Compare linear w(a) parametrization with quintessence and step-function w(a) models.
• Use a conserved DE energy-momentum framework to derive ρ_DE(a) for each model.
• Marginalize nuisance scaling parameters analytically (e.g., r_d h) to simplify the likelihood analysis.

Figure 1: Parameter constraints ( $1\sigma$ and $2\sigma$ contours) for the linear DE model assuming DESI data only, DESI+CMB, and DESI+CMB+SN with and without local SN, as shown in the legend. The locus of the $\Lambda$ CDM model is shown by the crossing of dashed lines in the $w_{0}-w_{a}$ panel.

Experimental results

Research questions

• RQ1 Does the DESI BAO data require a dynamical dark energy component beyond ΛCDM?
• RQ2 Is the apparent phantom behavior a robust feature or an artefact of the w0-wa linear parametrization?
• RQ3 At what redshift do deviations from w = -1 become significant according to current data?
• RQ4 Do quintessence or step-function w(a) models provide comparable or better fits than the linear model without phantom behavior?
• RQ5 Is the observed low-redshift signal driven primarily by local supernova data, and could local systematics explain it?

Key findings

• Linear w0-wa parametrization suggests phantom dark energy, but this is likely an artefact of the parametrization rather than a physical requirement.
• Quintessence and step-function w(a) parametrizations fit data comparably well or better and do not favor phantom behavior.
• Deviations from w = -1 are favored only at very low redshifts (z < O(0.1)) when all data are combined, driven mainly by low-z SNe.
• Removing low-z SN data removes the tension with ΛCDM, indicating sensitivity to nearby supernova systematics or local effects.
• All three models yield similar χ² values, with the step model performing slightly better than linear or quintessence in this analysis.
• The robust feature across models is a very recent/local (z ≲ 0.1) sharp change in DE density, potentially indicating either a local underdensity or very recent dark energy dynamics.

Figure 2: The $1\sigma$ and $2\sigma$ contours for the quintessence model parameter $K$ (upper panel) and the transition redshift $z_{t}$ of the step function parametrization (lower panel) against the current $w_{0}$ parameter, assuming DESI+CMB+SN data.