Preprint Suggests Dark Matter–Dark Energy Interaction Could Explain DESI Anomalies

A revised arXiv preprint argues that the recent cracks in the standard model of cosmology highlighted by DESI data may not require evolving dark energy after all. Instead, the authors say the same anomaly can be explained by a non-gravitational interaction between dark matter and dark energy — a claim they describe as showing roughly 3- to 5-sigma statistical support. The work, posted on arXiv and not peer-reviewed, should be treated as an early research result rather than a discovery.

The paper, “Robust Preference for Dark Sector Interactions,” is by Tian-Nuo Li, William Giarè, Guo-Hong Du, Yun-He Li, Eleonora Di Valentino, Jing-Fei Zhang and Xin Zhang. It was first posted Jan. 12, 2026, and revised July 15. The authors compare the standard ΛCDM model, which includes a cosmological constant, against two interacting-dark-energy setups — coupled quintessence and a coupled fluid model — as well as the standard time-varying dark energy parameterization known as CPL. Their analysis combines a wide set of observations, including Planck 2018, ACT DR6, SPT-3G, joint cosmic microwave background lensing, DESI DR2 baryon acoustic oscillation measurements and several Type Ia supernova compilations, including PantheonPlus, DES Y5 and DES-Dovekie.

In the paper’s abstract, the authors write: “Recent DESI baryon acoustic oscillation data reveal deviations from ΛCDM 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.” They add that “both IDE scenarios show robust evidence for non-vanishing interactions at the 3–5σ level, with marginalized constraints significantly deviating from the ΛCDM limit.”

If that conclusion held up, it would matter because ΛCDM is the baseline model cosmologists use to describe the universe. A confirmed interaction between dark matter, the unseen matter inferred from gravity, and dark energy, the unknown component associated with cosmic acceleration, would point to physics beyond that framework.

The timing is important because DESI DR2, the latest major release from the Dark Energy Spectroscopic Instrument, has already fueled a wave of 2025 and 2026 studies reporting hints that dark energy may be changing over time. Depending on the dataset mix and statistical method, those claims have generally landed in the roughly 2.8- to 4-sigma range. This new paper enters that debate with a different argument: late-time data that seem to favor dynamical dark energy may instead be signaling a coupling within the dark sector. The authors say that preference remains even when using the DES-Dovekie supernova recalibration, which had weakened evidence for dynamical dark energy in some other recent analyses.

The paper includes a few representative parameter estimates. For one coupled quintessence case using cosmic microwave background data, DESI and DES-Dovekie, the authors report a coupling parameter of β = 0.0527 ± 0.0087. For the coupled fluid model using the same data combination, they report β = -2.59 with asymmetric uncertainties. They also report better overall fits than ΛCDM in some combinations. In the paper’s convention, more negative values indicate improvement; for example, with CMB plus DESI plus DES-Dovekie, the coupled fluid model gives Δχ²_MAP of about -15.64 and ΔDIC of about -12.89 relative to ΛCDM.

But the paper also contains an important warning sign. The coupled fluid model’s best-fit region implies what the authors call extreme late-time cosmology, including a present-day matter density around 0.6 and strongly suppressed growth of cosmic structure. The paper says that model worsens the fit to growth-rate data, often written as fσ8(z), even if it improves the overall combined fit to expansion-history measurements.

That distinction matters. Baryon acoustic oscillations and supernova data mainly trace the universe’s expansion history, and different theories can mimic one another there. Measurements that track how structure grows over time are more discriminating. The authors explicitly point to perturbation-level probes — especially weak lensing and galaxy clustering — as the next tests that could separate interacting dark-sector models from more conventional dynamical dark energy explanations.

For now, the main takeaway is narrower than the headlines that often accompany cosmology anomalies. DESI-era tensions with ΛCDM have not produced a settled answer. This preprint argues that one prominent interpretation — evolving dark energy — may not be unique, and that a dark matter-dark energy interaction can fit the same data as well as or better in some cases. Whether that claim survives closer scrutiny will depend on peer review, independent reanalysis and, above all, the next round of structure-growth measurements.