Cosmic Acceleration in Question: Why DESI Sees Deceleration but SDSS Doesn't? ✛ ✦
Context
The question of whether the universe is accelerating now is one of the key issues in cosmology. The foundations of modern cosmology were laid by Edwin Hubble, who discovered the expansion, and Georges Lemaître, who predicted it theoretically. The acceleration was discovered through observations of distant type Ia supernovae by Saul Perlmutter and colleagues. After that discovery, the model with a cosmological constant (ΛCDM) became the standard for dark energy. However, new data from spectroscopic surveys DESI and SDSS, measuring baryon acoustic oscillations (BAO) in the distribution of galaxies and quasars, combined with cosmic microwave background (CMB) and gravitational lensing of the CMB, allow precise constraints on the properties of dark energy. Recent DESI DR2 results hint at a possible evolution of the equation of state over time, and the most intriguing aspect is that the median deceleration parameter q0 turns out to be positive, contradicting the established picture of an accelerating universe. This discrepancy with SDSS conclusions requires an explanation: either there is hidden systematics in the data, or we are misinterpreting the parameterization of dark energy. The authors hypothesize that it all comes down to the different redshift depths of the surveys.
Methods
The analysis used public MCMC chains from the Planck, SDSS DR16, and DESI DR2 collaborations. Within a unified CPL (w0-wa) parameterization, where w(a)=w0+wa(1-a), data combinations were compared: Planck 2018 CMB (temperature and polarization spectra plus lensing), BAO measurements from DESI (lowest effective zeff≈0.295 from the Bright Galaxy Sample BGS) and SDSS (including the MGS sample with zeff≈0.15), as well as the Pantheon+ supernova compilation. Models with constant w (wCDM) and massive neutrinos were also considered. The key tool was the deceleration parameter q0, which in a flat universe is expressed through Friedmann equations and Hubble's law as q0 = 1/2 Ωm + 1/2 (1+3w0) ΩDE. A robustness test was performed: the low-redshift anchor was removed from the SDSS analysis to mimic the DESI situation.
Results
In the CPL model, the Planck+SDSS combination yields an accelerating universe: w0 = -0.71 (+0.19/-0.18), q0 = -0.22 (+0.20/-0.21), and the transition from deceleration to acceleration occurred at zcrit ≈ 0.76. A completely different picture emerges from Planck+DESI: the median q0 = +0.10 (+0.21/-0.23) formally lies in the deceleration regime (though compatible with zero within 1σ), w0 = -0.41 (+0.21/-0.22), and the q(z) function crosses zero twice — at z∼0.08 and ∼0.86. This indicates that without additional low-redshift data it is impossible to claim that present-day expansion is accelerating. Adding Pantheon+ to Planck+DESI radically changes the situation: q0 = -0.37±0.06, w0 = -0.84±0.06, and acceleration is robustly recovered. The key experiment — removing the nearby MGS sample from SDSS — raises the effective redshift to ~0.38 and shifts q0 to -0.10 (+0.33/-0.35), bringing it closer to the DESI value. This proves that the discrepancy is indeed due to the difference in the minimum z eff.
Implications
The main conclusion: the apparent preference for a non-accelerating universe in the DESI BAO+CMB combination is not necessarily a hint of new physics. It naturally arises from the need to extrapolate the CPL parameterization into the z < 0.3 range, where DESI lacks direct distance measurements. This does not negate the stronger evidence for dynamical dark energy (non-zero wa) obtained when adding supernovae, but it emphasizes that interpreting current data requires caution. To reliably answer the question “is the universe accelerating now”, a robust anchor at z → 0 is necessary.
Future development
The most direct development is the implementation of finer “tomographic” binning in the Bright Galaxy Sample (BGS) of the DESI survey. Currently, BGS covers the range 0.1 < z < 0.4 as a single bin; splitting into two (e.g., 0.1–0.25 and 0.25–0.4) would yield a point with z eff ≈ 0.18, practically equivalent to the best SDSS measurement. This would allow testing whether the conclusions of the two surveys converge at the same redshift “depth”. Further prospects are tied to future spectroscopic surveys (Euclid, Nancy Grace Roman), which will provide an even more detailed map of expansion in the local universe.
Impact
The results will impact the interpretation of DESI data and the planning of observational strategies, as well as cross-validation between different surveys. They are important for the whole of cosmology, testing the expansion of the universe ΛCDM model, and searching for deviations from standard physics.
Next steps
The next step is a detailed forecast of what minimum z eff and what accuracy in determining q0 can be achieved with fine BGS binning, taking into account cosmic variance. A cross-analysis with fully unified versions of CMB data is also needed to isolate the BAO contribution.
Key open problems
The study is directly connected to the fundamental problem of the nature of dark energy. Is it a cosmological constant, or a dynamical field whose equation of state changed over time? The uncertainty about present-day acceleration is part of a broader crisis of the standard model, including the “Hubble tension”. Additionally, the work touches upon systematic errors in BAO measurements and the reliability of parameterizations used to describe dark energy.
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 compared constraints on dark energy parameters from Planck+SDSS and Planck+DESI data combinations within the CPL parametrization, showing that the median value of the deceleration parameter q0 is positive for DESI and negative for SDSS. They hypothesized that the discrepancy is due to different redshift coverage of the surveys and tested this by removing the low-redshift anchor from the SDSS analysis.
Where its strength is
The work clearly demonstrates how data samples with different redshift coverage can affect conclusions about the current acceleration of the Universe, and offers a concrete test to check for systematics.
Where it could go next
- It might be interesting to see how the proposed test for robustness to low-redshift data relates to recent attempts to reconcile discrepancies between different cosmological datasets, for example, through models with evolving dark energy.based on: Why Cosmic Data Refuse to Converge arXiv:2606.12822
- Nearby there is work considering dark energy with a crossing of w=-1; it might be worth checking how such models change the constraints on q0 in combinations with DESI and SDSS.based on: Dark Energy That Changes Sign arXiv:2606.01360
- In the archive there is an analysis of a scalar field model solving cosmological puzzles; one could compare how its predictions for the deceleration parameter agree with the results of this work.based on: Dark-Sector Seesaw: How a Scalar Spring Solves Cosmological Puzzles arXiv:2607.27515v1
Worth trying
- Check whether the positive q0 for DESI persists when using alternative dark energy parametrizations, for example, with a crossing of w=-1.
- Apply the method of excluding low-redshift data to other datasets, such as future Euclid data, to assess expected systematics.
- Compare the constraints on q0 from DESI with predictions of specific dark sector models, such as a scalar field with a potential.
Why this may matter
Understanding the causes of the discrepancy between DESI and SDSS is critical for interpreting future data, especially in the context of possible dark energy evolution, which could lead to a revision of the standard cosmological model.
Nearby works in our archive
- Dark Matter Decays, the Universe Smooths Outsimilarity 0.73
- A Dark Deal: How the Universe Reconciles Its Clockssimilarity 0.70
- Why Has Cosmology Cracked?similarity 0.69
- Dark-Sector Seesaw: How a Scalar Spring Solves Cosmological Puzzlessimilarity 0.69
- Dark Energy That Changes Signsimilarity 0.68
- Cosmic Friction: How Disorder Fixes the Universesimilarity 0.67
- Why Cosmic Data Refuse to Convergesimilarity 0.67
The archive is crowded around this work: 9 reviews nearby, closest at 0.73. This ground is well walked — and there is plenty to compare against.
Original abstract
as written by the authors arXiv:2607.07348
The DESI collaboration's Data Release~2 (DR2) provides baryon acoustic oscillation (BAO) measurements from over 14 million galaxies and quasars, and a joint analysis of DESI BAO, CMB, and Type~Ia Supernovae reveals a preference for time-evolving dark energy. We quantify this preference relative to SDSS BAO and report three key results. First, DESI+Planck favors a higher $$w_0 = -0.41^{+0.21}_{-0.22}$$ than SDSS+Planck ($$w_0 = -0.71^{+0.19}_{-0.18}$$). Second, DESI+Planck prefers a deceleration parameter whose median lies on the decelerating side ($$q_0 = 0.10^{+0.21}_{-0.23}$$, consistent with $$q_0 = 0$$ at $$1\sigma$$), while SDSS+Planck prefers a negative value ($$q_0 =-0.22^{+0.20}_{-0.21}$$) indicating accelerated expansion. Third, we argue that this discrepancy arises from the difference in the lowest effective redshift probed by each survey: $$z_{\rm eff} \approx 0.295$$ for DESI versus $$z_{\rm eff} \approx 0.15$$ for SDSS. As present-day quantities, $$w_0$$ and $$q_0$$ are sensitive to the lowest probed redshift: data near $$z = 0$$ constrain them directly, whereas higher-redshift data rely on extrapolating the dark energy parametrization (here CPL). Reaching $$z_{\rm eff} \approx 0.15$$, SDSS constrains $$w_0$$ and $$q_0$$ in a data-driven way, finding consistency with $$w_0 = -1$$ and acceleration. Limited to $$z_{\rm eff} \gtrsim 0.295$$, DESI relies more on extrapolation, driving $$q_0$$ positive and $$w_0$$ well above $$-1$$. Adding the Pantheon+ supernova sample restores low-redshift information, returning $$q_0$$ to negative values and reducing tension with $$\Lambda\text{CDM}$$. We therefore propose that the apparent DESI preference for a non-accelerating present epoch in the BAO+CMB combination reflects redshift sampling rather than new physics, and suggest future DESI analyses adopt finer tomographic binning of the Bright Galaxy Survey sample to access lower mean redshifts and test this conclusion.
If the era of acceleration really ended 1.5 billion years ago (as formally allowed by the median DESI reconstruction), we would now be witnessing the very beginning of a new deceleration phase — and in a few billion years galaxies would start moving closer together again.
Key numbers
- w0 (Planck+DESI): -0.41 ± 0.21
- w0 (Planck+SDSS): -0.71 ± 0.19
- q0 (Planck+DESI): +0.10 ± 0.23
- q0 (Planck+SDSS): -0.22 ± 0.20
- z_transition (SDSS): 0.76