DESI Data Put Dark Sector Energy Transfer to a Precision Test

The dark side of the cosmos is refusing to give up its secrets. In a new theoretical analysis published in The European Physical Journal C, a team of cosmologists led by Anil Kandel, Phongpichit Channuie, Ertan Güdekli and Farruh Atamurotov has tested whether dark matter and dark energy secretly exchange energy through a non-gravitational channel. Using the latest baryon acoustic oscillation measurements from the Dark Energy Spectroscopic Instrument, together with cosmic chronometers, compressed cosmic microwave background likelihoods and three independent Type Ia supernova compilations, the researchers constrained a phenomenological interacting dark energy model known as ξIDE. Their verdict is nuanced: the data consistently hint at energy flowing from dark energy into dark matter, but the statistical significance is so weak that the standard cosmological constant model still wins the model-comparison contest outright.

The motivation for such a test comes from two directions at once. On the theoretical side, the cosmological constant that anchors the standard ΛCDM model faces the notorious discrepancy between its observed energy density and quantum field theory estimates of the vacuum energy, as well as the cosmic coincidence problem, which asks why the matter and dark energy densities happen to be comparable at the present epoch. On the observational side, the DESI 2024 baryon acoustic oscillation analysis, combined with supernova data, reported evidence for an evolving dark energy density at 2.5 to 3.9 sigma, and the inclusion of DESI Data Release 2 pushed the preference for dynamical dark energy beyond the 4 sigma level in joint analyses. Later recalibrations of the Dark Energy Survey five-year supernova sample, known as DES-Dovekie, softened that preference to around 3.2 sigma. These developments have energized the search for alternatives in which the dark sector is not simply two non-interacting fluids evolving side by side.

The ξIDE framework, originally proposed by Dalal and collaborators, offers a particularly clean way to probe the question. In this model, the ratio of dark energy density to dark matter density evolves as a power law of the scale factor, ρX/ρm proportional to a raised to the power ξ. The parameter ξ quantifies the severity of the coincidence problem: ξ equals 3 reproduces the standard ΛCDM cosmology, while ξ equals 0 corresponds to a self-similar solution in which the ratio of dark energy to matter stays constant over cosmic time, making the coincidence problem vanish. Intermediate values between 0 and 3 describe scenarios in which the coincidence problem is progressively alleviated. Crucially, the model also carries an equation-of-state parameter wX, the ratio of dark energy pressure to its density, and the combination ξ + 3wX directly determines the interaction term Q, the rate at which energy is exchanged between the two dark components. When ξ + 3wX equals zero, Q vanishes and the standard non-interacting cosmology is recovered. A negative value of ξ + 3wX implies Q greater than zero, meaning energy flows from dark energy into dark matter, a direction that could potentially mitigate the coincidence problem. A positive value implies the opposite flow, which would exacerbate it.

To constrain this model, the team performed a Bayesian statistical analysis with the cosmological inference code SimpleMC, exploring the parameter space through dynamic nested sampling implemented in the dynesty package, with the number of live points set at fifty times the number of free parameters. The observational arsenal was formidable. Fifteen cosmic chronometer measurements of the Hubble expansion rate, spanning redshifts from 0.1791 to 1.965, provided direct H(z) determinations through the differential age method using massive, passively evolving galaxies. The baryon acoustic oscillation constraints came separately from DESI Data Releases 1 and 2, which combine measurements from the bright galaxy sample, luminous red galaxies, emission line galaxies, quasars and Lyman-alpha forests, expressed as dimensionless ratios of the comoving angular diameter distance, the Hubble distance and the volume-averaged distance to the sound horizon at the drag epoch. The supernova component included the Pantheon+ compilation of 1701 light curves from 1550 supernovae, the recalibrated DES-Dovekie dataset with 1820 supernovae, and the Union3 compilation of 2087 supernovae drawn from 24 independent datasets. A compressed CamSpec likelihood for the cosmic microwave background, modeled as a three-dimensional Gaussian in the angular sound horizon scale and the physical baryon and matter densities, completed the picture. The two DESI releases were analyzed separately rather than combined, allowing the researchers to quantify how improved precision affects the inferred interaction.

The first headline result concerns the Hubble constant. Within ΛCDM, every dataset combination shows a significant discrepancy with the local distance-ladder measurement of the reduced Hubble parameter, 0.7304 plus or minus 0.0104, with tensions ranging from roughly 4.3 to 5.0 sigma. The ξIDE model shifts the inferred value of the Hubble parameter toward slightly larger values, reducing that tension considerably. The most dramatic improvement appears for the combination of the compressed CMB likelihood, DESI Data Release 1 and cosmic chronometers, where the tension collapses from 4.71 sigma in ΛCDM to only 1.91 sigma in the interacting model. However, the relief is incomplete and release-dependent. Replacing DESI Data Release 1 with the more precise Data Release 2 raises that same tension back to 2.87 sigma, indicating that the updated baryon acoustic oscillation data weaken the preference for a higher Hubble constant and thereby reduce the model’s ability to address the Hubble tension. For every other combination, the residual tension remains between 3.59 and 3.95 sigma. The dark-sector interaction, in other words, moves the inferred expansion rate closer to the local measurement but never eliminates the discrepancy.

The model parameters themselves stay remarkably close to their standard limits. The constraints on ξ fall in the range of roughly 2.7 to 2.9, mildly below the canonical value of 3 and suggesting a slight reduction of the coincidence problem, but fully consistent with the ΛCDM prediction within the statistical uncertainties. The equation-of-state parameter wX is confined between approximately minus 0.98 and minus 0.92, a regime that favors quintessence-like dark energy, in which the pressure is slightly less negative than that of a cosmological constant. Most dataset combinations prefer wX greater than minus 1, echoing the broader trend toward dynamical dark energy seen in the DESI analyses, yet the cosmological constant value of minus 1 remains consistent with all combinations at the one to two sigma level. The present matter density parameter Ωm is barely affected by the interaction at all, with deviations between the two models never exceeding 0.83 sigma, showing that current data pin down the matter content almost independently of the assumed dark-sector coupling.

The central quantity of the analysis, the interaction parameter combination ξ + 3wX, came out negative for every single dataset combination. This consistency across DESI Data Releases 1 and 2 and across all three supernova compilations is the most intriguing hint in the paper: if energy is being exchanged in the dark sector, the flow runs from dark energy into dark matter, a direction that would alleviate rather than worsen the coincidence problem. The strongest deviation from the non-interacting limit appears for the combination of the compressed CMB likelihood, DESI Data Release 2, cosmic chronometers and the Union3 supernova sample, where ξ + 3wX equals minus 0.035 plus or minus 0.023, corresponding to only 1.52 sigma. The DES-Dovekie and Pantheon+ combinations with the same CMB and DESI Data Release 2 data favor negative values at 1.35 and 1.32 sigma respectively. The comparison with earlier studies is striking. Previous analyses have reported values ranging from ones fully consistent with zero, such as 0.03 with uncertainties above one, to substantially negative values like minus 2.52 plus or minus 0.87, minus 2.98 plus or minus 0.08 and even minus 3.54, which would imply a substantial dark-sector interaction. The new constraints sit much closer to the ΛCDM limit, with far tighter error bars, and they suggest that some of the earlier indications of strong coupling were driven by larger uncertainties rather than by robust data preference.

Model comparison delivers the sobering counterweight. The team quantified the goodness-of-fit improvement of ξIDE over ΛCDM through the difference in the maximum-a-posteriori chi-squared statistic, converted into an equivalent Gaussian significance. For the DESI Data Release 1 combinations, the significance ranges from 1.10 to 1.90 sigma, and for the Data Release 2 combinations from 1.68 to 2.06 sigma, with the highest preference again found for the Union3 combination at 2.06 sigma. Under the adopted classification scheme, all of these values fall in the weak-evidence category, below the 2.5 sigma threshold required for moderate evidence. The Bayesian evidence tells an even clearer story. The logarithmic Bayes factors, computed with the revised Jeffreys scale, range from 4.18 to 5.38 in favor of ΛCDM for the Data Release 1 combinations, corresponding to moderate-to-strong support for the standard model, and from 2.60 to 4.16 for the Data Release 2 combinations, still moderate evidence for ΛCDM. Although the strength of that preference decreases with the more precise DESI Data Release 2 data, the simpler non-interacting cosmology wins every model-comparison contest in the analysis.

The overall picture that emerges is one of tantalizing consistency without detection. The interacting dark energy model remains a viable extension of the standard cosmology, offers a possible mechanism for softening the Hubble tension, and consistently points toward energy transfer from dark energy to dark matter, a direction that survives the transition from DESI Data Release 1 to the tighter Data Release 2 constraints. But the deviation from zero interaction never exceeds 1.52 sigma, the fit improvement stays in the weak-evidence band, and Bayesian evidence firmly favors the cosmological constant. As the DESI survey continues to accumulate data and future supernova and chronometer measurements sharpen the late-time expansion history, the sign of ξ + 3wX will be one of the numbers to watch. If a genuine dark-sector coupling exists, the current data have already revealed its direction; what remains is to establish whether the coupling itself is real or merely a mirage of statistical noise.

Subject of Research: Testing the direction and strength of dark matter–dark energy interaction with late-time cosmological datasets

Article Title: Testing the direction of dark sector interaction using late-time datasets

Article References: Kandel, A., Channuie, P., Güdekli, E., & Atamurotov, F. (2026). Testing the direction of dark sector interaction using late-time datasets. The European Physical Journal C, 86(10), Article 1218. https://doi.org/10.1140/epjc/s10052-026-16492-4

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-16492-4

Keywords: dark energy, dark matter, interacting dark energy, DESI, baryon acoustic oscillations, Hubble tension, Lambda-CDM, Type Ia supernovae, cosmic chronometers, cosmic microwave background, Bayesian evidence, cosmic coincidence problem