Evolving dark energy in DESI DR2: supernova-calibration systematics versus thawing quintessence
The DES-Dovekie and Union3.1 recalibrations lowered the preference to about 3σ and the deciding data (DESI DR3, Euclid DR1, LSST) are scheduled, so difficulty is 40, level with the ladder problem and far below the cosmological constant (90). Leverage 55, just below the cosmological constant problem (56): a confirmed w ≠ −1 would redirect dark-energy and quantum-gravity model building, but a systematic resolves it with little downstream effect. Application value 9 sits at the high end of cosmology because supernova calibration methods are reused across surveys.
The problem
Is dark energy changing over time? Sorting out whether a 3-sigma hint is real or a measurement artefact.
- 1Standard model: dark energy is constant
- 2DESI + CMB + supernovae hint at change
- 3Hint is about 3 sigma, not decisive
- 4Each input has its own possible errors
- 5Find which input drives the hint
The model. Dark energy is described by an equation of state w, the ratio of its pressure to its energy density. A cosmological constant has w = −1 at all times. A common parametrization allows change: w(a) = w₀ + w_a(1 − a), where a is the scale factor (a = 1 today). The standard model is (w₀, w_a) = (−1, 0), called ΛCDM; the wider model is w₀w_aCDM.
Built from
earlier problems → this oneThe problems this one is built from: progress or tools it needs (builds on), special cases that come first (generalizes), and weaker results on the way (implies). On the right, what builds on it.
- Start with
- Step 2
- Evolving dark energy in DESI DR2: supernova-calibration systematics versus thawing quintessenceThis problem
- Then
Path to understanding
foundations → problemThe research entries this problem rests on, in the order to read them. Each links to its explainer.
- Start with
- Step 2
- Step 3
- ΛCDM, the standard model of cosmologyTheory · host
- Then
- Evolving dark energy in DESI DR2: supernova-calibration systematics versus thawing quintessenceThe problem
State of the art
DESI DR2 BAO (arXiv:2503.14738, March 2025; PRD 112, 083515; extended analysis arXiv:2503.14743) prefers w₀w_aCDM over ΛCDM at 3.1σ with the CMB alone and at 2.8–4.2σ when supernovae are added, highest with DES-SN5YR (arXiv:2401.02929). The supernova leg has since moved. DES recalibrated its photometry ('Dovekie', arXiv:2506.05471) and reanalysed its 5-year sample: with Planck/ACT/SPT and DESI DR2 it finds w₀ = −0.803 ± 0.054, w_a = −0.72 ± 0.21, a preference that falls from 4.2σ to 3.2σ, which the authors describe as only weak (about 5:1 odds) (Popovic et al., arXiv:2511.07517, MNRAS 2026). Union3.1, with self-consistent host-galaxy measurements for about 2,000 SNe, gives 3.4σ (from 3.8σ) with DESI DR2 and the CMB (Hoyt et al., arXiv:2601.19424), and the combined Pantheon+ and DES-SN5YR 'Unite' compilation gives 3.1–3.3σ frequentist but only weak Bayesian evidence (Camilleri et al., arXiv:2609.05053, September 2026). The full-survey DES multiprobe analysis finds w₀ = −0.82 ± 0.05, w_a ≈ −0.63 at 3.0σ with the CMB, and 2.2σ from DES alone (arXiv:2605.27221, May 2026). On the statistics, a nested-sampling reanalysis (Ong et al., arXiv:2511.10631) finds that DESI DR2 + Planck alone modestly favour ΛCDM (ln B = −0.57 ± 0.26), that the 3σ evidence with DES-SN5YR comes from a 2.95σ DESI–DES tension inside ΛCDM, and that it vanishes with DES-Dovekie (ln B = −0.30 ± 0.19); a calibration-independent Alcock–Paczynski test finds DESI DR2, Union3, Pantheon+ and DES-Dovekie mutually consistent within about 1σ (Dinda & Maartens, arXiv:2601.16229; JCAP 05 (2026) 015). A 2026 invited review (Li et al., arXiv:2606.21826; RAA 26, 084002) stresses that the strength of the preference depends on parametrisation and dataset choice. Several of these are preprints. The next tests are not yet in: DESI's DR2 Lyman-α full-shape analysis was being validated in July 2026 (arXiv:2607.27411), DESI DR3 has not reported, Euclid DR1 is split into a data-only 'Foundation' release (scheduled for November 2026) and the full cosmology release (mid-2027), and Rubin/LSST began its 10-year survey in mid-2026 with no cosmology results. As of October 2026 the preference sits at about 3σ in most combinations and is unresolved.
Realistic applications
If real, the preference would be the first evidence of dynamics in the dominant component of the universe, redirecting dark-energy and string-landscape model building. If a systematic, the identification improves SN cosmology. Inference tools (model-independent w(z) reconstructions, calibration-marginalised likelihoods) are reusable. There is no practical application.
What counts as solved
Resolved if DESI DR3 BAO and full-shape, Euclid DR1 and LSST supernova and lensing data give a model-independent w(z) departing from −1 at ≥5σ (confirmation), or if an identified, quantified systematic (for example in supernova host-galaxy or photometric calibration) removes the preference to below 2σ in independent samples (refutation).
How it gets verified
ObservationSettled by new measurements or observations.
Needs new data: DESI DR3 BAO and full-shape, Euclid DR1 and LSST supernova and lensing data must give a model-independent w(z) that departs from -1 at 5 sigma or more, or a quantified systematic must remove the preference to below 2 sigma in independent samples.
Claims under verification
+ add a claimWhere the work is
Mostly new BAO and supernova data. Compute is likelihood and covariance work. Theory is the construction of consistent evolving models.
Bottlenecks
ranked by votes- #1EmpiricalSupernova photometric calibration across samples
Different calibration of low-z and high-z SN samples shifts the preference by 1σ or more. A common cross-calibration at 0.01 mag is needed.
Community grades
Sign in to add your own grade.
Connected problems
Progress
- May 26, 2026Full-survey DES multiprobesource
DES alone 2.2σ; with the CMB w₀ = −0.82 ± 0.05 at 3.0σ.
- Jan 20, 2026BAO–SN consistency testsource
DESI DR2 BAO and Union3, Pantheon+ and DES-Dovekie SNe are mutually consistent within about 1σ.
- Nov 10, 2025DES-Dovekie recalibrationsource
DES reanalysis with updated SN calibration lowers the w₀w_a preference from 4.2σ to 3.2σ.
- Mar 18, 2025DESI DR2 BAOsource
2.8–4.2σ preference for evolving dark energy with CMB and SNe.