Strong Evidence for Dark Sector Interactions

TL;DR

DESI BAO data indicate deviations from $\Lambda$CDM that could arise from dark-sector interactions rather than a time-varying dark energy EoS. The authors test two IDE realizations, coupled quintessence (CQ) and coupled fluid (CF), within a unified perturbation framework against Planck, ACT, SPT, DESI BAO, and SN data, employing $H_0$ as a sampling parameter and deriving $\theta_*$ self-consistently with $\theta_* = r_{\mathrm{d}}(z_*)/D_{\mathrm{A}}(z_*)$. They find robust evidence for a non-vanishing coupling parameter $\beta$ at the $3$–$5\sigma$ level in both models, with $\beta>0$ in CQ (energy transfer from DM to DE) and $\beta<0$ with a phantom $w$ in CF, often improving upon $\Lambda$CDM and competing with CPL in goodness-of-fit for the same number of parameters. The work demonstrates that IDE is a physically well-motivated alternative to dynamical DE, producing similar background histories and CMB signatures but potentially distinct perturbation evolution, underscoring the need for future weak-lensing and galaxy-clustering data to break degeneracies.

Abstract

Recent DESI baryon acoustic oscillation data reveal deviations from $Λ$CDM cosmology, conventionally attributed to dynamical dark energy (DE). We demonstrate that these deviations are equally, if not better, explained by interactions between dark matter and dark energy (IDE), without requiring a time-varying DE equation of state. Using a unified framework, we analyze two IDE models -- coupled quintessence and coupled fluid -- against the latest CMB (Planck, ACT, SPT), DESI BAO, and SN (including DES-Dovekie recalibrated) data. Both IDE scenarios show robust evidence for non-vanishing interactions at the 3-5$σ$ level, with marginalized constraints significantly deviating from the $Λ$CDM limit. This preference persists even under DES-Dovekie SN recalibration, which weakens dynamical DE evidence. Crucially, for the same number of free parameters, IDE models provide fits to low- and high-redshift data that match or exceed the performance of the CPL dynamical DE parametrization. Our results establish IDE as a physically motivated alternative to dynamical DE, highlighting the necessity of future cosmological perturbation measurements (e.g., weak lensing, galaxy clustering) to distinguish between these paradigms.

Figures (8)

• Figure 1: Two-dimensional marginalized contours (68.3% and 95.4% confidence levels) from DESI, CMB, and SN data. Upper panel: Constraints in the $\beta$--$\alpha$ plane within the CQ model. Lower panel: Constraints in the $\beta$--$w$ plane within the CF model.
• Figure 2: Using the CMB+DESI+DES-Dovekie data, the profile likelihood for the parameter $\beta$ is presented for the CQ (upper panel) and CF (lower panel) models.
• Figure 3: Comparison of the temperature power spectrum for the $\Lambda$CDM, CPL, CQ, and CF models, using the best-fit values inferred from the CMB+DESI+DES-Dovekie analysis, over the Planck, ACT, and SPT temperature power spectrum data points. We also present the fractional difference between the $\Lambda$CDM and CPL, CQ, or CF models.
• Figure 4: Best-fit predictions for (rescaled) distance-redshift relations for $\Lambda$CDM, CPL, CQ, and CF obtained from the analysis of CMB+DESI data.
• Figure 5: Observables from the best-fit $\Lambda$CDM, CPL, CQ, and CF models are compared with the distance modulus $\mu$, normalized to the Planck 2018 best-fit $\Lambda$CDM cosmology. The binned distance modulus residuals for DES-Dovekie and DESY5 are shown, represented by black squares and gray circles, respectively. The bin edges for the SN bins are indicated by vertical gray dashed lines, and the SN binning method is described in Section IVC of Ref. DESI:2025zgx.