🔭 astrophysics

Cosmological constraints from a joint DESI DR1 Full-Shape and DR2 BAO

This paper presents a robust, prior-robust Bayesian analysis combining DESI DR1 full-shape and DR2 BAO data using the ShapeFit compression method to deliver the first reliable DESI-only constraints on extensions to the $\Lambda$ CDM model, including dynamical dark energy and neutrino masses.

Original authors: D. Forero-Sánchez, H. Gil-Marín, L. Verde, Z. Ding, A. J. Ross, A. Carnero Rosell, J. Aguilar, S. Ahlen, S. Bailey, D. Bianchi, C. Blake, A. Brodzeller, D. Brooks, R. Canning, F. J. Castander, T. Clay

Published 2026-02-24

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: D. Forero-Sánchez, H. Gil-Marín, L. Verde, Z. Ding, A. J. Ross, A. Carnero Rosell, J. Aguilar, S. Ahlen, S. Bailey, D. Bianchi, C. Blake, A. Brodzeller, D. Brooks, R. Canning, F. J. Castander, T. Claybaugh, S. Cole, A. Cuceu, A. de la Macorra, Arjun Dey, P. Doel, S. Ferraro, A. Font-Ribera, J. E. Forero-Romero, E. Gaztañaga, G. Gutierrez, J. Guy, C. Hahn, H. K. Herrera-Alcantar, K. Honscheid, D. Huterer, M. Ishak, R. Joyce, S. Juneau, R. Kehoe, D. Kirkby, T. Kisner, J. Kneib, A. Kremin, O. Lahav, C. Lamman, M. Landriau, L. Le Guillou, M. Manera, A. Meisner, R. Miquel, J. Moustakas, G. Niz, N. Palanque-Delabrouille, W. J. Percival, F. Prada, I. Pérez-Ràfols, G. Rossi, E. Sanchez, E. F. Schlafly, D. Schlegel, H. Seo, J. Silber, D. Sprayberry, G. Tarlé, B. A. Weaver, C. Zhao, R. Zhou, H. Zou

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, expanding balloon. For decades, cosmologists have been trying to figure out exactly how fast this balloon is inflating, what it's made of, and whether the inflation is speeding up or slowing down. To do this, they need a "ruler" to measure the distances between galaxies.

This paper is about a team of astronomers using a massive new tool called DESI (Dark Energy Spectroscopic Instrument) to take the most precise measurements of this cosmic ruler yet. They are combining two different "snapshots" of the universe taken at slightly different times (Data Release 1 and Data Release 2) to get a clearer picture than ever before.

Here is the breakdown of what they did and what they found, explained with some everyday analogies.

1. The Tool: A Cosmic Ruler and a "Full-Shape" Scan

DESI is like a super-powered camera that can take pictures of 5,000 galaxies at once. It looks at the light from these galaxies to measure two things:

The "Echo" (BAO): Long ago, sound waves rippled through the hot soup of the early universe. When the universe cooled, these waves froze, leaving a specific "fingerprint" or echo in how galaxies are spaced out. This is the Baryon Acoustic Oscillation (BAO). Think of it like finding a perfectly spaced row of trees in a forest. If you know the trees were planted 10 meters apart, and you see them looking 12 meters apart, you know the forest has stretched. This is a very reliable, rigid ruler.
The "Full Shape" (FS): But the forest isn't just a row of trees; it's a complex landscape with hills, valleys, and clumps. The Full-Shape analysis looks at the entire pattern of the forest, not just the spacing of the trees. It tells us about how gravity is pulling things together and how the universe is growing.

2. The Problem: The "Prior Volume" Trap

In the past, when scientists tried to combine these two measurements (the rigid ruler and the complex landscape) to test new theories (like "Is dark energy changing over time?"), they ran into a mathematical problem called Prior Volume Effects.

The Analogy: Imagine you are trying to find a lost coin in a giant warehouse.

The Rigid Ruler (BAO) tells you the coin is definitely in the warehouse.
The Complex Landscape (Full Shape) gives you clues about where in the warehouse it might be.
The Trap: If you don't have enough clues, your computer simulation might get confused. It might say, "Well, the coin could be in the corner, or the middle, or the ceiling!" Because the "ceiling" area is so huge, the math accidentally makes it seem like the coin is most likely on the ceiling, even if the clues point to the floor. This is the "Prior Volume Effect"—the math gets biased by the sheer size of the possibilities, not the actual data.

3. The Solution: "ShapeFit" Compression

The authors of this paper used a clever new method called ShapeFit.

The Analogy: Instead of trying to map every single tree in the forest (which is computationally heavy and prone to the "ceiling" trap), ShapeFit compresses the data into a few key "summary stats." It's like taking a high-resolution photo of the forest and shrinking it down to a small, clear thumbnail that still keeps all the important details.

By using this "thumbnail" approach, they avoided the mathematical trap. They could now look at the data without the computer getting confused by the size of the possibilities. This allowed them to make reliable conclusions using only the DESI data, without needing to lean heavily on outside information (like the Cosmic Microwave Background) to fix the math.

4. The Results: What Did They Find?

By combining the new, larger dataset (DR2) with the detailed Full-Shape data (DR1), they found:

The Universe's Ingredients: They confirmed that the universe is made of about 30% matter (stuff like galaxies and dark matter) and the rest is dark energy. They measured the expansion rate (Hubble constant, or h) to be about 0.69.
Is Dark Energy Changing? The standard theory says dark energy is a constant "cosmological constant" (like a steady pressure pushing the universe apart). However, their data hints that dark energy might be wobbly—changing strength over time.
- The Tension: If you only look at the "rigid ruler" (BAO), the data disagrees with the standard theory by about 1.7 standard deviations (a bit of a disagreement).
- The Fix: When they added the "Full Shape" data, the disagreement dropped to 1.4 standard deviations. It's still a hint, but it's less of a scream and more of a whisper. It suggests the universe might be more dynamic than we thought, but it's not a full-blown crisis yet.
Neutrinos (Ghost Particles): They put a new limit on the total mass of neutrinos (tiny, ghostly particles that pass through everything). They found the total mass must be very small, less than 0.54 eV (electron volts) without using outside data. This is a big step forward in understanding these elusive particles.
Curvature: They checked if the universe is flat (like a sheet of paper) or curved (like a sphere). The data says it's flat, consistent with our current best theories.

5. Why This Matters

This paper is a milestone because it proves that DESI can stand on its own. Previously, to get reliable answers about the universe's expansion, scientists needed to mix DESI data with data from the Big Bang (CMB).

Now, thanks to the ShapeFit method, DESI can answer complex questions about the universe's evolution using only its own data. It's like a detective who used to need a second detective to confirm a suspect, but now has a new forensic tool that makes them 100% confident on their own.

In a nutshell: The universe is expanding, it's mostly flat, and dark energy might be doing something interesting, but we need more data to be sure. The new "ShapeFit" tool is the key that unlocked these answers without getting lost in the math.

1. Problem Statement

The Dark Energy Spectroscopic Instrument (DESI) has produced high-precision cosmological data through its Data Releases (DR1 and DR2). However, combining these datasets presents two significant challenges:

Data Correlation: DR1 and DR2 are highly correlated because DR2 includes the DR1 data plus additional observations. A naive combination of their likelihoods without accounting for cross-correlations leads to incorrect error estimates.
Prior Volume Effects: When analyzing DESI data alone (without Cosmic Microwave Background, CMB, priors) using Bayesian frameworks, constraints on models beyond the standard $\Lambda$ CDM (such as dynamical dark energy or neutrino masses) are often unreliable. This is due to "prior volume effects," where the choice of priors on nuisance parameters dominates the posterior distribution, leading to shifts between the Maximum A Posteriori (MAP) and the true profile likelihood. Previous Full-Modeling (FM) approaches suffered significantly from this issue.

2. Methodology

The authors propose a novel framework to combine DESI DR1 Full-Shape (FS) clustering measurements with DESI DR2 Baryon Acoustic Oscillation (BAO) measurements, termed "DESI1.5".

ShapeFit Compression: Instead of using the computationally expensive Full-Modeling (FM) approach, the authors utilize the ShapeFit method. This compresses the power spectrum information into a small set of parameters:
- $q_{iso}, q_{ap}$ : Isotropic and Alcock-Paczynski dilation parameters.
- $f$ : Logarithmic growth of structure.
- $m, n$ : Shape parameters describing the slope of the power spectrum.
- This compression is cosmology-agnostic, separating data compression from model inference.
Joint Covariance Estimation via EZmocks:
- To handle the correlation between DR1 and DR2, the authors generated a suite of EZmocks (mock catalogs based on the Zel'dovich approximation).
- They reprojected 100 existing DR1 EZmocks onto the DR2 survey footprint to create a consistent mock suite for both releases.
- They computed the cross-correlation matrix between the compressed DR1 ShapeFit parameters and DR2 BAO parameters. This allowed them to construct a full $6 \times 6$ covariance matrix (combining 4 SF parameters and 2 BAO parameters) for each tracer.
- Three covariance strategies were tested: (A) full mock covariance, (B) ignoring cross-correlations, and (C) the baseline choice using data-derived diagonal blocks and mock-derived off-diagonal cross-terms.
Priors:
- BBN Prior: A tight prior on the physical baryon density ( $\omega_b$ ) derived from Big Bang Nucleosynthesis.
- Spectral Index ( $n_s$ ): A wide prior ( $n_s10$ ) with uncertainty inflated by a factor of 10 relative to Planck, to minimize external bias.
- External Data: Analyses were performed both with DESI-only data and combined with CMB (Planck) and Supernovae (PantheonPlus) data.

3. Key Contributions

First Consistent DR1+DR2 Combination: The paper provides the first optimal combination of DESI DR1 Full-Shape and DR2 BAO data, properly accounting for their statistical correlation using mock-based cross-covariances.
Mitigation of Prior Volume Effects: The study demonstrates that the ShapeFit compression approach effectively regularizes the parameter space. Unlike Full-Modeling, ShapeFit yields Bayesian posteriors where the MAP aligns with the high-probability region, eliminating the misleading shifts caused by prior volume effects in models with evolving dark energy.
DESI-Only Constraints: For the first time, robust Bayesian constraints on extensions to $\Lambda$ CDM (dynamical dark energy, neutrino masses) are derived using only DESI data (plus minimal BBN/ $n_s$ priors), without relying on CMB to break degeneracies.
Sound-Horizon-Free Analysis: The authors perform an analysis agnostic to the sound horizon scale ( $r_d$ ), using the matter-radiation equality scale as a standard ruler to constrain $H_0$ independently of recombination physics.

4. Key Results

$\Lambda$ CDM Constraints (Flat Universe)

Using the combined DESI1.5 dataset with BBN and wide $n_s$ priors:

$\Omega_m = 0.3035 \pm 0.0085$
$h = 0.6876 \pm 0.0059$
$\sigma_8 = 0.822 \pm 0.034$
These results are consistent with previous DR1 and DR2-only analyses but with improved precision.

Dynamical Dark Energy ( $w_0w_a$ CDM)

DESI-Only: $w_0 = -0.49 \pm 0.25$ $w_{0} = - 0.49 \pm 0.25$ , $w_a = -1.52 \pm 0.77$ $w_{a} = - 1.52 \pm 0.77$ .
- This represents a ~30% improvement in constraints compared to DR1-only ShapeFit results.
- The tension with $\Lambda$ CDM is reduced to 1.4 $\sigma$ (compared to 1.7 $\sigma$ for BAO-only), showing that the Full-Shape data helps stabilize the constraints without introducing prior bias.
DESI + CMB: The preference for evolving dark energy increases to approximately 3 $\sigma$ .

Neutrino Mass Constraints

DESI-Only ( $\Lambda$ CDM): Upper limit $\sum m_\nu < 0.54$ eV (95% CI). The posterior peaks at a positive mass ( $\approx 0.03$ eV), contrasting with previous DR1 FM results which peaked in negative mass regions.
DESI-Only ( $w_0w_a$ CDM): Upper limit $\sum m_\nu < 1.00$ eV (95% CI).
DESI + CMB: Constraints tighten significantly to $\sum m_\nu < 0.062$ eV ( $\Lambda$ CDM) and $< 0.144$ eV ( $w_0w_a$ CDM).

Spatial Curvature

Adding Full-Shape information to DR2 BAO tightens the curvature constraint ( $\Omega_k$ ) by a factor of two.
DESI1.5 alone is consistent with a flat universe.
Combined with CMB, the data shows a preference for an open universe ( $\Omega_k > 0$ ) at ~2.4 $\sigma$ , though this sensitivity depends on the specific CMB likelihood used.

Sound-Horizon-Free ( $r_d$ -free)

Without assuming the sound horizon scale, the analysis yields $h = 0.693 \pm 0.026$ .
This value is in excellent agreement with CMB determinations, demonstrating that Large Scale Structure (LSS) can constrain $H_0$ independently of early-universe physics.

5. Significance

This work establishes ShapeFit as a robust, computationally efficient, and prior-robust alternative to Full-Modeling for analyzing Large Scale Structure data. By successfully combining DR1 and DR2 data while mitigating prior volume effects, the authors demonstrate that DESI can provide standalone, reliable constraints on complex cosmological models (dynamical dark energy, neutrino masses) without needing external CMB data to "fix" the priors. This marks a significant step toward using LSS as an independent probe of fundamental physics and dark energy evolution. The methodology sets a precedent for future data releases where combining different analysis flavors (BAO vs. Full-Shape) across different data releases will be essential.