DESI’s Cosmic Ruler Hints Dark Energy May Change—But the Case Isn’t Closed

An echo from the infant universe has become one of cosmology’s most closely watched signals. Baryon acoustic oscillations, or BAO, are an imprint from sound waves in the early universe: they leave a preferred distance scale in the clustering of galaxies. Mapping that scale at different cosmic times lets astronomers trace how the universe’s expansion rate has changed.

In April 2024, the Dark Energy Spectroscopic Instrument (DESI) released its first BAO data, and later data releases sharpened the picture. When those measurements are combined with the cosmic microwave background (CMB) and type Ia supernovae, the data show an apparent departure from the standard cosmological model, ΛCDM. That model treats dark energy as a cosmological constant: a fixed energy density with equation of state w = −1, where w roughly describes how pressure and density are related. According to a review posted as a preprint on arXiv and not yet peer-reviewed, the new data prefer something more dynamic.

The Hint: Dark Energy That Doesn’t Stand Still

In the standard picture, dark energy is constant. But when researchers fit the DESI BAO data together with CMB and supernova measurements using the common two-parameter CPL description of an evolving equation of state, they find a preference for w greater than −1 today and a negative change parameter. In plain terms, dark energy would have been more phantom-like in the past—with w less than −1—and more quintessence-like now, crossing the phantom divide at w = −1. Phantom-like means w < −1; quintessence-like means w > −1. That crossing is called quintom behavior.

The review reports that the statistical preference varies. Depending on the supernova compilation and data combination, the quoted range is about 2.8σ to 3.8σ; here σ is a measure of statistical significance, with 5σ often taken as the threshold for discovery. DESI’s second data release, combined with CMB and supernova data, pushed the preference as high as 4.2σ. A recalibrated supernova sample called DES-Dovekie, when combined with DESI DR2 and CMB data, reduced the DESY5-based preference to about 3.2σ. In discovery terms, that is intriguing but not decisive.

The review stresses that the hint is not a single measurement but a delicate combination. BAO supplies geometric anchors across cosmic time. Supernovae provide the low-redshift luminosity-distance relation. The CMB constrains the early universe. Change one ingredient, and the preference shifts.

Different supernova compilations matter. DESY5 tends to strengthen the case for evolving dark energy, while PantheonPlus weakens it. A small calibration offset between low- and high-redshift supernovae, even a few hundredths of a magnitude, can mimic an evolving equation of state. That does not prove the signal is an artifact, but it shows why the supernova treatment is crucial.

The choice of mathematical parametrization also matters. A study of six forms of w(a)—CPL, JBP, BA, exponential, logarithmic and sinusoidal—finds that the preference for dynamics does not vanish, and some alternatives strengthen it. Yet a few DESI BAO bins—particularly luminous red galaxy measurements near redshifts 0.51 and 0.71—carry disproportionate influence. Removing the LRG2 distance point can bring the result back within 2σ of a cosmological constant. Non-parametric reconstructions also see deviations, but they, too, depend on data and assumptions.

Other Physics Can Wear the Same Disguise

Perhaps the review’s most important point is that an evolving equation of state is not the only way to explain the data. BAO and supernovae mostly measure the expansion history. Different physical models can produce the same background expansion.

One possibility is interacting dark energy. Here dark matter and dark energy exchange energy and momentum, so they are not separately conserved. Even if dark energy’s own equation of state is fixed at −1, the interaction can make the inferred equation of state appear to evolve or cross the phantom divide. Some analyses of DESI data find evidence for such interactions at roughly 3σ to 5σ, though the strength depends on the assumed interaction form.

A second possibility is non-minimally coupled gravity. Here a scalar field is coupled to spacetime curvature. That extra coupling can stabilize a phantom crossing that would be unstable in simple scalar-field models. One “hilltop thawing gravity” model fits the DESI-preferred expansion history with a log Bayes factor of about 7.3 relative to ΛCDM—a strong statistical preference. But such theories must also pass local tests of gravity, which require screening mechanisms on small scales.

A third possibility is non-standard dark matter. If dark matter is not perfectly cold and pressureless, its density evolution changes, and that can shift the inferred dark-energy parameters. Current data show a mild preference for a non-zero dark-matter equation of state at about 2.4σ to 3σ. But when dark energy is allowed to evolve as well, the significance drops to about 0.8σ to 1.1σ. The two dark sectors are strongly entangled.

Tensions That Refuse to Vanish

The review also examines how these ideas affect cosmology’s biggest disagreements. The H0 tension—the mismatch between early- and late-universe measurements of the expansion rate—is not solved by simple late-time dynamical dark energy. In fact, fitting the DESI preference with CMB data tends to lower the inferred H0. Interacting dark energy models can raise it, in one case to about 71.5 km/s/Mpc, reducing the tension with local measurements to roughly 1.5σ.

The S8 tension concerns how clumpy matter is on large scales. The review discusses interacting dark sectors, modified gravity, and neutrino–dark matter scattering as ways to suppress structure growth and ease the discrepancy.

The neutrino-mass problem is a striking place where allowing dark energy to evolve changes the bound. Within ΛCDM, combining Planck CMB data with DESI DR2 BAO tightens the upper limit on the sum of neutrino masses to below 0.064 eV—close to the minimum mass expected from neutrino oscillations. That creates tension with terrestrial experiments. Allowing dark energy to evolve relaxes the bound to about 0.163 eV. Adding weak lensing can even yield a positive neutrino mass near 0.098 eV, consistent with the normal hierarchy.

The review is careful: these are hints, not a discovery. The DESI measurements are real, but the interpretation depends on models, datasets, and calibration choices. The same expansion history can come from evolving dark energy, interacting dark sectors, modified gravity, or non-standard dark matter. Background measurements alone cannot tell them apart.

Breaking the degeneracy will require more than distance measurements. It will need precise probes of how cosmic structure grows, including weak gravitational lensing and redshift-space distortions. Future DESI releases, improved BAO consistency tests, independent supernova calibration, and better control of systematics will all matter.

The deeper question is not simply whether dark energy changes with time. It is which assumption of the standard model is being tested: that dark energy is constant, that dark matter and dark energy are separately conserved, that general relativity holds on cosmic scales, or that dark matter is cold and pressureless. For now, this early-universe sound echo has given cosmologists a tantalizing clue—and a difficult puzzle about how to read it.