DES-Y5 Supernova Cosmology Overview

• DES-Y5 is the Dark Energy Survey five-year Type Ia supernova sample offering homogeneous photometry and controlled systematics across redshifts 0.025–1.13 for late-time cosmology studies.
• The sample employs SALT3 light curve modeling and the BEAMS with Bias Corrections framework to standardize data and mitigate selection biases, enhancing precision in cosmological measurements.
• Analyses of DES-Y5 have highlighted calibration anomalies and tension with DESI BAO data, which spurred a recalibration into DES-Dovekie and informed dark energy model comparisons.

Searching arXiv for recent DES-Y5 papers to support the article. DES-Y5 usually denotes the Dark Energy Survey five-year Type Ia supernova sample, DES-SN5YR, used in late-time cosmology as a supernova Hubble-diagram dataset. In the recent literature it is also the original DES supernova cosmology release that was later reassessed and reprocessed into DES-Dovekie, and it has become central to model-independent geometry tests, peculiar-velocity analyses, and dark-energy model comparison because joint fits with DESI BAO and Planck CMB respond to it differently than to Pantheon+ or Union3 (Dinda et al., 20 Jan 2026, Ong et al., 13 Nov 2025).

1. Definition and nomenclature

In the late-time cosmology literature summarized here, DES-Y5 refers to the Dark Energy Survey five-year Type Ia supernova sample built from the DES 5-year supernova program. Different analyses describe it in slightly different catalog terms: one gives “of order 1800” SNe Ia over $0.025 < z < 1.13$, another gives 1635 light curves from 1550 SNIa with $0.10 < z < 1.13$ plus 194 low-$z$ SNIa with $0.025 < z < 0.10$, and a model-independent FLRW analysis quotes 1829 spectroscopically confirmed SNe Ia in $0.10 < z < 1.13$ (Crisman et al., 7 Apr 2026, Xu et al., 12 Oct 2025, Millard et al., 28 Jan 2026). These descriptions all refer to the same DES five-year supernova cosmology program as used in different pipelines or summary conventions.

DES-Y5 is repeatedly contrasted with Pantheon+, a heterogeneous multi-survey compilation, and with Union3. In the peculiar-velocity literature it is singled out as having homogeneous photometry and well-controlled systematics, while Pantheon+ is described as more heterogeneous (Crisman et al., 7 Apr 2026). In the calibration-update literature, DES-Y5 is also the original release later revised into DES-Dovekie (Dinda et al., 20 Jan 2026).

The label is not unique across all DES work. In cosmic-shear forecasting, “DES Y5” denotes the final fifth-year weak-lensing dataset rather than the supernova sample; that forecast assumed 5000 deg$^2$, source density $\bar n=8$ galaxies arcmin$^{-2}$, 5 tomographic bins, and intrinsic shape noise $\sigma_\epsilon=0.27$ (Foreman et al., 2016). In the arXiv papers discussed below, however, DES-Y5 is overwhelmingly the supernova dataset.

2. Catalog construction and observable content

DES-Y5 is a standardizable-candle sample. The supernova light curves are modeled with SALT3, and both Pantheon+ and DES-Y5 use the BEAMS with Bias Corrections framework to determine nuisance parameters for color and stretch and to correct selection biases (Crisman et al., 7 Apr 2026). The published distance moduli are used directly in several analyses, rather than refitting the SALT nuisance parameters $\alpha$, $0.10 < z < 1.13$0, and $0.10 < z < 1.13$1, because those are already absorbed into the catalogued $0.10 < z < 1.13$2 (Crisman et al., 7 Apr 2026).

The DES-Y5 catalog covariance, denoted $0.10 < z < 1.13$3 in the peculiar-velocity analysis, contains photometric uncertainties, intrinsic dispersion, calibration systematics, and selection effects. In that framework the full supernova covariance is written

$0.10 < z < 1.13$4

with a non-linear velocity-dispersion term $0.10 < z < 1.13$5 (Crisman et al., 7 Apr 2026). The DES-Y5 posterior for $0.10 < z < 1.13$6 is reported as broad and platykurtic, allowing values in the entire range $0.10 < z < 1.13$7–$0.10 < z < 1.13$8 km/s (Crisman et al., 7 Apr 2026).

For geometry analyses, DES-Y5 enters through apparent peak magnitudes $0.10 < z < 1.13$9 or through distance moduli $z$0. One calibration-independent formulation uses

$z$1

while another writes the comoving distance directly as

$z$2

In both cases the unknown absolute magnitude $z$3 appears only as an overall normalization and therefore cancels in the geometric ratios used for consistency tests (Dinda et al., 24 Sep 2025, Dinda et al., 20 Jan 2026).

3. Calibration-independent geometry and the DESI tension

A major reason DES-Y5 attracted attention is its behavior in calibration-independent comparisons with DESI BAO. The common variable is the Alcock–Paczynski ratio

$z$4

which is independent of the sound horizon $z$5. In flat FLRW one also has $z$6, so for supernovae

$z$7

Because the overall $z$8 normalization cancels between numerator and denominator, the supernova version is likewise independent of absolute SN calibration (Dinda et al., 24 Sep 2025, Dinda et al., 20 Jan 2026).

The reconstruction is non-parametric. The DESI DR2 BAO $z$9 data and the SNIa-derived $0.025 < z < 0.10$0 data are each fitted with Gaussian Processes using a zero mean function and a squared-exponential kernel,

$0.025 < z < 0.10$1

with hyperparameters obtained by maximizing the marginal likelihood (Dinda et al., 20 Jan 2026). The redshift-dependent tension statistic is then

$0.025 < z < 0.10$2

Using this framework, Pantheon+ and Union3 were found to have tension $0.025 < z < 0.10$3 across their redshift ranges, whereas DES-Y5 showed a growing discrepancy that exceeded $0.025 < z < 0.10$4 near $0.025 < z < 0.10$5 (Dinda et al., 24 Sep 2025). The same analysis reported up to $0.025 < z < 0.10$6 evidence for redshift evolution of $0.025 < z < 0.10$7 around $0.025 < z < 0.10$8 in DES-Y5, while Pantheon+ showed only mild evidence slightly above $0.025 < z < 0.10$9 for $0.10 < z < 1.13$0 (Dinda et al., 24 Sep 2025). This localized the anomaly to a geometrical quantity that is already independent of both $0.10 < z < 1.13$1 and $0.10 < z < 1.13$2.

The update based on DES-Dovekie did not reanalyze DES-Y5 directly; instead it replaced DES-Y5 with the recalibrated DES release and repeated the same test. The result was that all uncalibrated data from DESI DR2 BAO and the three SNIa datasets Union3, Pantheon+, and DES-Dovekie became mutually consistent within $0.10 < z < 1.13$3, and DES-Dovekie stayed below $0.10 < z < 1.13$4 tension across the full range up to $0.10 < z < 1.13$5 (Dinda et al., 20 Jan 2026). This suggests that the original DES-Y5 anomaly was tied to the original release rather than to an irreducible BAO–SN mismatch.

4. Peculiar velocities, growth, and FLRW curvature tests

DES-Y5 has also been used as a peculiar-velocity catalog. In that setting the low-redshift subset is restricted to

$0.10 < z < 1.13$6

which yields 243 low-$0.10 < z < 1.13$7 DES-Y5 SNe; the full sample is used simultaneously for the background magnitude–redshift relation (Crisman et al., 7 Apr 2026). The supernova peculiar-velocity likelihood is

$0.10 < z < 1.13$8

with $0.10 < z < 1.13$9 (Crisman et al., 7 Apr 2026).

In flat $^2$0CDM with GR growth fixed, DES-Y5 alone gives

$^2$1

a broad but CMB-consistent amplitude constraint (Crisman et al., 7 Apr 2026). Combined with Planck PR4 while allowing both curvature and the growth index $^2$2 to vary, DES-Y5 yields

$^2$3

and

$^2$4

with $^2$5 km/s/Mpc, $^2$6, $^2$7, and $^2$8 in the same extended model (Crisman et al., 7 Apr 2026). When SH0ES $^2$9 information is added, the fit shifts to

$\bar n=8$0

and the point $\bar n=8$1 is excluded at around $\bar n=8$2 for DES-Y5 (Crisman et al., 7 Apr 2026).

A separate model-independent FLRW test reconstructs $\bar n=8$3, the dimensionless comoving distance $\bar n=8$4, and $\bar n=8$5 with an iterative smoothing algorithm and combines them with DESI DR2 BAO in the $\bar n=8$6 diagnostic (Millard et al., 28 Jan 2026): $\bar n=8$7 For DES-Y5 the overlap with DESI DR2 BAO is $\bar n=8$8, corresponding to three BAO bins (Millard et al., 28 Jan 2026). Over the subset of reconstructions that both improve the fit relative to flat $\bar n=8$9CDM and remain consistent with FLRW, the reported median curvature is

$^{-2}$0

for DES-Y5 + DESI DR2 (Millard et al., 28 Jan 2026). The same paper reports that all selected DES-Y5 reconstructions pass the constancy test for $^{-2}$1, so the issue is not a breakdown of FLRW within the tested redshift range (Millard et al., 28 Jan 2026).

5. DES-Y5 in dark-energy model selection

DES-Y5 is the supernova compilation that most strongly shifts joint DESI–CMB–SN fits away from $^{-2}$2CDM in several recent analyses. A Bayesian model-comparison study found that DESI DR2 BAO + Planck CMB alone gives

$^{-2}$3

which modestly favors $^{-2}$4CDM over $^{-2}$5CDM, but adding DES-Y5 changes this to

$^{-2}$6

equivalent to $^{-2}$7 in favor of $^{-2}$8CDM (Ong et al., 13 Nov 2025). For DESI DR2 + DES-Y5 alone the same paper reported $^{-2}$9, corresponding to $\sigma_\epsilon=0.27$0 (Ong et al., 13 Nov 2025).

The tension analysis in that work located the driver of the preference in a low-dimensional DESI–DES-Y5 inconsistency within $\sigma_\epsilon=0.27$1CDM. For DESI DR2 versus DES-Y5 the parameter-difference tension is

$\sigma_\epsilon=0.27$2

with $\sigma_\epsilon=0.27$3, $\sigma_\epsilon=0.27$4, and $\sigma_\epsilon=0.27$5 (Ong et al., 13 Nov 2025). In $\sigma_\epsilon=0.27$6CDM the tension drops to

$\sigma_\epsilon=0.27$7

$\sigma_\epsilon=0.27$8, and $\sigma_\epsilon=0.27$9, which the authors interpret as the extra $\alpha$0 freedom acting to absorb a specific low-dimensional dataset conflict rather than as a purely early-time demand for evolving dark energy (Ong et al., 13 Nov 2025).

A broader parameterization study using CMB + BAO + DES-Y5 reported the following representative values. In $\alpha$1CDM,

$\alpha$2

whereas among seven dynamical parameterizations the Barboza–Alcaniz model gave

$\alpha$3

$\alpha$4

with $\alpha$5 and $\alpha$6, the strongest improvement among the tested forms (Xu et al., 12 Oct 2025). The same paper states that all dynamical dark-energy models fitted to CMB+BAO+DES-Y5 lower $\alpha$7 relative to $\alpha$8CDM, so they do not alleviate the $\alpha$9 tension, but they reduce $0.10 < z < 1.13$00 relative to the $0.10 < z < 1.13$01CDM fit (Xu et al., 12 Oct 2025).

A separate multi-model study using CMB + DES-Y5 + DESI found that CPL reaches

$0.10 < z < 1.13$02

with $0.10 < z < 1.13$03, $0.10 < z < 1.13$04, and $0.10 < z < 1.13$05 (Pérez et al., 23 Dec 2025). In the same analysis the flipped RVM reaches $0.10 < z < 1.13$06, and the threshold RVM reaches $0.10 < z < 1.13$07 with DES-Y5 (Pérez et al., 23 Dec 2025). Across those model spaces, DES-Y5 is consistently the more dynamical-DE-friendly supernova dataset relative to Pantheon+ (Pérez et al., 23 Dec 2025).

Analysis context	DES-Y5 role	Representative result
Calibration-independent AP test (Dinda et al., 24 Sep 2025)	SN geometry vs DESI DR2 BAO	$0.10 < z < 1.13$08 tension near $0.10 < z < 1.13$09
Peculiar velocities + CMB (Crisman et al., 7 Apr 2026)	Low-$0.10 < z < 1.13$10 PV subset + full Hubble diagram	$0.10 < z < 1.13$11, $0.10 < z < 1.13$12
Bayesian evidence (Ong et al., 13 Nov 2025)	Joint DESI DR2 + CMB + SN model selection	$0.10 < z < 1.13$13 for $0.10 < z < 1.13$14CDM
Multi-model DE comparison (Pérez et al., 23 Dec 2025)	DESI DR2 + Planck PR4 + SN	CPL gives $0.10 < z < 1.13$15

DES-Y5 also enters joint DES SN + DES BAO analyses. A DESI-independent angular BAO measurement from DES Y6 reported that adding DES BAO-noDESI to DES Y5 SN + Planck + DESI DR1 increases the $0.10 < z < 1.13$16CDM preference from $0.10 < z < 1.13$17 to $0.10 < z < 1.13$18, and updating to DESI DR2 raises it from $0.10 < z < 1.13$19 to $0.10 < z < 1.13$20 (Mena-Fernández et al., 21 Jan 2026). The corresponding significances drop to $0.10 < z < 1.13$21 when DES-Y5 is replaced by SN-Dovekie (Mena-Fernández et al., 21 Jan 2026).

6. Reprocessing into DES-Dovekie and present interpretation

The revision from DES-Y5 to DES-Dovekie is central to the current status of the dataset. The DES-Dovekie release is described as a reanalysis of the DES-SN5YR program with revised calibration, involving a reassessment of systematics in SN photometry and light-curve standardization (Dinda et al., 20 Jan 2026). In the DESI DR2 cosmology context, replacing DES-Y5 by DES-Dovekie lowers the quoted deviation from $0.10 < z < 1.13$22CDM in DESI DR2 BAO + CMB + DES SN from $0.10 < z < 1.13$23 to $0.10 < z < 1.13$24 (Dinda et al., 20 Jan 2026).

A complementary “cosmological intercept tension” analysis isolates an internal DES-Y5 issue at lower redshift. Splitting the sample into the homogeneous DES-SN component and the external low-$0.10 < z < 1.13$25 subset, it reports a weighted-average intercept offset

$0.10 < z < 1.13$26

around $0.10 < z < 1.13$27 (Wang et al., 30 Apr 2026). In that paper, removing or phenomenologically debiasing the low-$0.10 < z < 1.13$28 intercept anomaly reduces the apparent Planck + DESI + DES-Y5 preference for CPL dynamical dark energy from $0.10 < z < 1.13$29 to $0.10 < z < 1.13$30, while removing SNe below $0.10 < z < 1.13$31 reduces it to $0.10 < z < 1.13$32 (Wang et al., 30 Apr 2026). The same paper further notes that DES-Dovekie still shows a late-time $0.10 < z < 1.13$33 tension of similar form, although somewhat softened (Wang et al., 30 Apr 2026).

Taken together, the recalibration-independent AP update, the intercept analysis, and the DESI-independent BAO combination indicate that DES-Y5 was both scientifically powerful and unusually sensitive to calibration structure in the original release. The factual pattern is that the original DES-Y5 sample produced stronger apparent tension with DESI BAO, stronger apparent preference for evolving dark energy, and a localized intercept anomaly, whereas DES-Dovekie reduces those effects and restores $0.10 < z < 1.13$34-level agreement in the calibration-independent BAO–SN geometry test (Dinda et al., 20 Jan 2026, Wang et al., 30 Apr 2026, Mena-Fernández et al., 21 Jan 2026). A plausible implication is that DES-Y5 remains indispensable as a case study in late-universe inference precisely because it sits at the interface between statistical preference for new dark-energy phenomenology and sensitivity to supernova calibration and standardization.