Crosschecking Cosmic Distances from DESI BAO and DES SNe Points to Systematics

TL;DR

The paper investigates whether DESI DR2 BAO and DES SNe distances are consistent by binning SNe to the same effective redshifts z_eff as DESI BAO and reconstructing DM/r_d and DH/r_d under flat LCDM or wCDM with Planck rd. It then forms RD_M and RD_H ratios to test horizontality and the distance duality relation, assessing robustness to variations in r_d and late/early-Universe physics. The main finding is that RD_M is consistent with unity across models, confirming distance duality, while RD_H declines with z_eff at about 2.5σ, which cannot be explained by physics alone and points to observational systematics in either DESI BAO or DES SNe. Even with relaxed r_d and exotic pre-recombination physics, the trend persists at ~2.3σ, underscoring the need for more data and careful crosschecks before claiming new physics from combined datasets.

Abstract

We perform a consistency check of DESI DR2 BAO constraints ($D_M/r_d, D_H/r_d)$ by reconstructing the same quantities from DES supernovae (SNe) in bins with the same effective redshift $z_{\textrm{eff}}$. We find that the ratio of $D_M/r_d$ values are consistent with a horizontal, thus confirming that the distance duality relation holds up to calibration. However, the $D_H/r_d$ ratio shows a decreasing trend with $z_{\textrm{eff}}$ at $2.3 σ$ to $2.5 σ$ that cannot be explained by physics. We demonstrate that the result does not depend on the choice of cosmological model, but the radius of the sound horizon $r_d$ has a much greater influence. Studying ratios of $D_H/r_d$ is a stronger test than the distance duality relation, and the rejection of a horizontal confirms systematics in either DESI BAO or DES SNe. Claims of new physics based on combined data still have rudimentary hurdles to clear.

Figures (3)

• Figure 1: Redshift distributions for Pantheon+ (orange) and DES (blue) SNe Ia samples. Pantheon+ contains the majority of its supernovae at low redshift ($z \lesssim 0.2$), while DES provides a more uniform coverage up to $z \sim 1.0$.
• Figure 2: The ratio $R_{D_M}$ as a function of effective redshift $z_{\textrm{eff}}$ assuming $\Lambda$CDM and $r_d = 147.09 \pm 0.26$ Mpc from Planck. The dashed horizontal line indicates the null hypothesis of consistency, $R_{D_M} = 1$.
• Figure 3: Ratio $R_{D_H}$ as a function of effective redshift assuming $\Lambda$CDM and $r_d = 147.09 \pm 0.26$ Mpc from Planck. The solid red line and shaded band show the best-fit linear trend $R_{D_H}(z) = m \, z_{\textrm{eff}} + c$ inferred from MCMC sampling and its 68% confidence region, while the dashed horizontal line represents the null hypothesis of consistency, $R_{D_H} = 1$. The line is $2.5\sigma$ removed from horizontal (slope $m=0$).