Hint towards inconsistency between BAO and Supernovae Dataset: The Evidence of Redshift Evolving Dark Energy from DESI DR2 is Absent

Abstract

The combination of independent cosmological datasets is a route towards precise and accurate inference of cosmological parameters if these observations are free from systematic effects. However, unknown systematics in different datasets can lead to biased inference of cosmological parameters. In this work, we test the consistency of two independent tracers of low-redshift cosmic expansion, namely the supernova dataset from Pantheon$+$ and the BAO dataset from DESI DR2, using the distance duality relation, a cornerstone of cosmology within General Relativity. We find that these datasets violate the distance duality relation and show a redshift-dependent signature, hinting at unaccounted physical effects or observational artifacts. This effect mimics a redshift-evolving dark energy scenario when Pantheon$+$ and DESI datasets are combined without accounting for this inconsistency. Accounting for this effect in the likelihood refutes the previous claim of evidence for non-cosmological constant dark energy from DESI DR2, yielding results consistent with a cosmological constant with $w_0= -0.92\pm 0.08$ and $w_a= -0.49^{+0.33}_{-0.36}$. This is further supported by an increased Bayes factor at ($w_0 = -1$, $w_a = 0$) when the inconsistency is accounted for. This indicates that current conclusions from DESI DR2 combined with Pantheon$+$ likely arise from combining inconsistent datasets, leading to precise but inaccurate inference of cosmological parameters. Future tests of consistency between cosmological datasets will be essential for robust inference and for identifying unaccounted physical effects or observational artifacts.[Abridged]

Figures (8)

• Figure 1: Illustration of how parameter inference from two datasets which are biased can can cause precise but an inaccurate inference. The dashed ellipses (green and blue) represent unbiased, independent constraints derived separately from Data Set 1 and Data Set 2, respectively. The solid ellipses (green and blue) indicate the same parameters measured in the presence of uncorrected systematic inconsistencies, causing shifts away from the true parameter values. The magenta filled ellipse shows the biased combined constraint resulting from naively merging these inconsistent measurements. After identifying and correcting the inconsistencies, the corrected combined inference is represented by the red filled ellipse, which realigns closely with the true parameter values. This schematic highlights the necessity of checking and correcting for inter-dataset inconsistencies prior to performing joint cosmological analyses. The boxes in different color explain different contours shown in the figure.
• Figure 2: Contours in the dark energy EoS (EoS) parameters denoted by $(w_0,w_a)$ plane for three analyses: the Baseline fit (without correcting from mismatch in CDDR) (filled orange) constraining $\{H_0,\Omega_m,w_0,w_a\}$; the 6-parameter fit (dashed orange) adding $(d_0,d_1)$; and the 7-parameter fit (dash-dot green) further including $d_2$ (More details are given in Sec. \ref{['sec:DataDrivenDDTest']}). Allowing for distance duality deviations broadens the parameter space and shifts the best‐fit region toward the $\Lambda$CDM reference point $(w_0=-1,\;w_a=0)$ in agreement with previous cosmological results, highlighting the importance of consistency checks between distance indicators before combining datasets for cosmological inference.
• Figure 3: Comparison of cosmological distance measurements as a function of redshift, displayed as distance modulus $\mu$ [mag]. Red points show the Pantheon+ supernova sample (1701 SNIa), while blue points show DESI BAO measurements converted to distance modulus using the transverse observable $D_M/r_s$ with the fiducial sound horizon $r_d = 147.09$ Mpc from Planck 2018 Planck:2018vyg. The conversion follows $\mu = 5\log_{10}[(1+z) (D_M / r_d) \times r_d] + 25$, where $D_L = (1+z)D_M = (1+z)^2 D_A$. Error bars represent 1$\sigma$ uncertainties; for DESI, these are derived from the diagonal elements of the covariance matrix. This visualization demonstrates both datasets used in our CDDR consistency analysis on a common distance scale, facilitating direct visual comparison between the two independent cosmological probes.
• Figure 4: Baseline constraints on the parameter set $\{H_0, \Omega_m, w_0, w_a\}$ in a flat $w_0\text{-}w_a$CDM model without including the distance duality parameters. The blue contours show the results of our analysis using DESI DR2 + Pantheon+ data, while the red shaded regions indicate the DESI DR2 (BAO + SNIa + Planck $\rm{\Omega_m}$ prior) $1\sigma$ constraints. We impose a Gaussian Planck prior on $\Omega_m$ (mean = 0.315, standard deviation = $0.007$), and adopt flat priors on $H_0 \in [60, 80]$, $w_0 \in [-2, 1]$, and $w_a \in [-3, 3]$, together with the condition $w_0 + w_a < 0$ to ensure viable past-light-cone histories. This plot shows that our results are in excellent agreement with the DESI DR2 constraints when no distance duality parameter is included.
• Figure 5: Joint six‐parameter posterior constraints on the parameter set $\{H_0,\Omega_m,w_0,w_a,d_0,d_1\}$ in a flat $w_0\text{-}w_a$CDM model including distance-duality coefficients. The blue contours represent the joint constraints from DESI DR2 + Pantheon+ for our extended framework, while the red shaded regions correspond to the DESI DR2 (BAO + SNIa + Planck $\rm{\Omega_m}$ prior) results. We impose a Gaussian Planck prior on $\Omega_m$ (mean = 0.315, $\sigma$ = 0.007), flat priors on $H_0\in[60,80]$, $w_0\in[-2,1]$ and $w_a\in[-3,3]$ with the requirement $w_0 + w_a < 0$, and flat priors on the distance-duality coefficients $d_0\in[-1,2]$ and $d_1\in[-1,1]$.