Dark Energy Is Changing: Signs of Evolution in DESI and Supernova Data ✛ ✎
Context
Cosmology in recent years has been living in tension: the standard ΛCDM model with a cosmological constant describes most data remarkably well, but precise distance measurements in the universe increasingly show small discrepancies. This work joins a line of research looking for signs of dark energy evolution in DESI DR2 data and modern supernova catalogs. Neighboring studies have already reported hints of dynamics, often with higher significance, up to 4–5 sigma. The present paper deliberately restricts itself to late epochs (supernovae and BAO) and uses compressed CMB data to avoid systematic biases associated with full spectra.
Methods
The authors used Bayesian inference with dynamic nested sampling in the SimpleMC code. They applied BAO data from DESI DR2, three Type Ia supernova compilations (Pantheon+, DES-Dovekie, Union3), and a compressed CMB likelihood folded into three parameters. To test dark energy evolution, they considered eight parameterizations of the equation of state: logarithmic, exponential, CPL, BA, JBP, Thawing, Mirage, and GEDE. They separately analyzed the influence of the LRG1 and LRG2 tracers, which are luminous red galaxies at redshifts around 0.5, by removing them from the sample one at a time.
Results
The key result is that the preference for evolving dark energy in DESI DR2 data is almost entirely driven by the LRG1 and LRG2 tracers. When they are included, the parameter w0 shifts to values greater than −1 and wa becomes negative; without them, the constraints return to the cosmological constant. For all models and data combinations, the preferred region lies in the quadrant w0 > −1, wa < 0, corresponding to the Quintom-B scenario. The statistical significance, estimated from the chi-square difference, does not exceed 2.3 sigma, and Bayesian factors show only inconclusive or weak evidence for dynamics. The Mirage model shows the highest values, up to 2.95 in the log Bayesian factor for Union3, which is still moderate. The evolution of w(z) in many models crosses the w = −1 boundary around redshift 0.5, indicating a phantom crossing, but this effect may be an artifact of the parameterization.
Implications
This work differs from neighboring studies in that it systematically checks the influence of individual BAO tracers and restricts itself to late-time data, avoiding the full CMB spectrum. This allows separating a real cosmological signal from possible systematic effects in low-redshift samples. The conclusion is moderate: current data do not require abandoning ΛCDM, but they leave room for evolving dark energy. Importantly, the apparent phantom crossing should not be interpreted as evidence for exotic physics; it may result from fitting a parameterization over a limited redshift range.
Future development
Future surveys such as Euclid and the Roman Space Telescope will provide much more precise BAO and supernova measurements over larger volumes. This will allow testing whether the preference for dynamics persists after improved statistics and control of systematics. A separate direction is the development of model-independent methods for reconstructing w(z) that do not rely on specific parameterizations. A more thorough analysis of DESI's low-redshift tracers will also be needed to understand the nature of their anomalous influence on the constraints.
Impact
The results will affect the interpretation of DESI data and the planning of future cosmological missions. They also stimulate theoretical work on dark energy models that can produce an effective phantom crossing without violating fundamental theorems.
Next steps
The immediate next step is to repeat the analysis with the addition of full CMB and lensing data to check the robustness of the conclusions. It is also necessary to study in detail the systematic errors in the LRG1 and LRG2 tracers, possibly using simulations.
Key open problems
For cosmology, the key unresolved problem is the nature of dark energy and the possible connection of its evolution to the Hubble tension. This work shows that late-time data alone do not provide convincing evidence for dynamics, which narrows the field for speculation and underscores the need to account for systematics.
The knowledge machine's view
This section is addressed to the author of the paper and their colleagues. Neighbouring works were found by meaning-based search across our archive, not by word overlap. We do not review: we show the field around the work.
What the work does
The authors analyzed DESI DR2 BAO data and three Type Ia supernova compilations, testing eight models with evolving dark energy. Using Bayesian inference with dynamic nested sampling, they showed that the preference for dynamics is mainly driven by the LRG1 and LRG2 tracers at redshifts around 0.5. The statistical significance does not exceed 2.3 sigma, indicating moderate evidence.
Where its strength is
A systematic check of the influence of individual BAO tracers and constraints from late-time data allow separating a possible cosmological signal from systematic effects in low-redshift samples. This approach gives a more conservative estimate of the significance of dark energy evolution compared to studies using full CMB spectra.
Where it could go next
- It might be interesting to compare the obtained constraints with an analysis where dark energy is treated as a dynamical field with phantom crossing, which would test the robustness of the preference for the quadrant w0 > -1, wa < 0.based on: The Cosmic Metronome Falls Out of Rhythm: Dark Energy Rewrites Hubble's Symphony arXiv:2606.05358v1
- A related work directly investigates the influence of a local inhomogeneity (bubble) on the interpretation of BAO data; comparing it with your LRG1 and LRG2 results could clarify whether the signal is linked to local structure.based on: Bubble in a Sea of Dark Energy arXiv:2607.18376v1
- There is an arXiv paper using DESI and supernova data to test models with evolving dark energy but with a different parametrization; it may be useful to compare how much the conclusions about significance depend on the chosen functional form of w(z).based on: Dark Energy: When the Constant Stops Being Constant arXiv:2607.01226v1
Worth trying
- Test whether the preference for dynamics persists when excluding only one of the LRG1 or LRG2 tracers, rather than both at once.
- Use alternative supernova compilations, for example DES-SN5YR, to assess the robustness of the w0 shift.
- Apply Bayesian model comparison that accounts for prior parameter probabilities to refine the strength of the evidence.
Why this may matter
If the dark energy evolution signal is confirmed, it would require a revision of the standard cosmological model and open the way to new physical theories. Even moderate hints of dynamics are important for planning future surveys such as Euclid and Roman.
Nearby works in our archive
- The Cosmic Metronome Falls Out of Rhythm: Dark Energy Rewrites Hubble's Symphonysimilarity 0.66
- Bubble in a Sea of Dark Energysimilarity 0.65
- Dark Energy: Constancy Questionedsimilarity 0.65
- The Universe Whispers: [tag:dark_energy]dark energy[/tag] is not alonesimilarity 0.64
- The Universe Is Not Standing Still: Dark Energy Is Changingsimilarity 0.63
- Dark Energy: When the Constant Stops Being Constantsimilarity 0.63
The closest work in the archive sits at 0.66 by meaning — the usual level for our corpus.
The Mirage model gets its name from its way of fooling the observer: it is specifically constructed to reproduce the distance to the last scattering surface of the CMB just like ΛCDM, while still allowing the equation of state to evolve at late times. This makes it almost indistinguishable from a cosmological constant with some data, but potentially distinguishable with others.
The idea of changing dark energy echoes Greg Egan's novel 'Schild's Ladder', where physical laws turn out to be non-constant in different regions of spacetime. In reality, the evolution of w(z) is not a change in laws, but a possible property of the dark energy field itself.
Key numbers
- number of dark energy models: 8
- maximum significance of deviation from ΛCDM: 2.3σ
- redshift of phantom crossing: z ≈ 0.5
- number of supernovae in Union3: 2087
- number of supernovae in Pantheon+: 1701