DESI Supports Modified Expansion History but Leaves Gravitational vs Fluid Dark Energy Underdetermined
DESI data establish a preference for evolving dark-energy equation-of-state parameters inside a GR-plus-fluid framework but cannot yet rule out equivalent modified-gravity explanations. The analysis highlights an observational degeneracy that current BAO-only fits leave unresolved. Stronger tests will require joint expansion-plus-growth measurements from upcoming surveys.
The preprint by Bamonti and collaborators dissects three distinct claims bundled in recent DESI announcements. Claim D1 (disfavoring LambdaCDM expansion) and D2 (preference for CPL parameters inside GR plus smooth fluid) receive support from the BAO measurements, while claim D3 (existence of a new matter-sector degree of freedom) does not. Designer f(R) constructions reproduce the identical background history without exotic matter, exposing an observational degeneracy that fluid-only fits ignore. Phantom-crossing behavior preferred by the data requires fine-tuned negative kinetic terms under the fluid reading but emerges without pathology in non-minimally coupled scalar-tensor theories.
Related work from the DESI collaboration itself (arXiv:2404.03002) and the 2025 Euclid early-release cosmology paper both rely on the same CPL template, inheriting the same ambiguity. Neither survey has yet reported scale-dependent growth measurements capable of breaking the degeneracy. The resulting underdetermination means current evidence updates the expansion history without yet selecting a physical mechanism.
Future data releases from DESI Year-3 and full Euclid will tighten w0-wa contours but will still require separate growth probes, such as redshift-space distortions or weak-lensing ratios, to test whether the deviation lives in the matter sector or in gravity itself. Until those observables are combined at the required precision, headline claims of dynamical dark energy remain premature.
Bamonti et al.: DESI Year-3 plus Euclid DR1 joint analysis will still allow f(R) models within 1 sigma of the best-fit CPL point through 2028.
Sources (2)
- [1]Primary Source(https://arxiv.org/abs/2608.21405)
- [2]Supporting Source(https://arxiv.org/abs/2404.03002)