Cosmological implications of DESI DR2 BAO measurements in light of the latest ACT DR6 CMB data

TL;DR

This study reanalyzes DESI DR2 BAO in light of the latest ACT DR6 CMB data, using conservative multipole cuts to combine ACT with Planck and assess robustness of DESI in the ΛCDM framework and its extensions. It systematically explores how different Planck+ACT data combinations affect consistency with DESI, and revisits evolving dark energy and neutrino-mass inferences when including ACT data and SNe. The main findings show persistent evidence for a dynamical dark energy component (2.4–4.0σ depending on SN data), and neutrino-mass constraints tighten modestly under optimized ACT+Planck combinations, with results generally robust to the specific CMB dataset. Overall, DESI DR2 BAO conclusions remain broadly robust when ACT DR6 data are incorporated, and future consolidated CMB datasets may yield even tighter cosmological constraints.

Abstract

We report cosmological results from the Dark Energy Spectroscopic Instrument (DESI) measurements of baryon acoustic oscillations (BAO) when combined with recent data from the Atacama Cosmology Telescope (ACT). By jointly analyzing ACT and Planck data and applying conservative cuts to overlapping multipole ranges, we assess how different Planck+ACT dataset combinations affect consistency with DESI. While ACT alone exhibits a tension with DESI exceeding 3$σ$ within the $Λ$CDM model, this discrepancy is reduced when ACT is analyzed in combination with Planck. For our baseline DESI DR2 BAO+Planck PR4+ACT likelihood combination, the preference for evolving dark energy over a cosmological constant is about 3$σ$, increasing to over 4$σ$ with the inclusion of Type Ia supernova data. While the dark energy results remain quite consistent across various combinations of Planck and ACT likelihoods with those obtained by the DESI collaboration, the constraints on neutrino mass are more sensitive, ranging from $\sum m_ν< 0.061$ eV in our baseline analysis, to $\sum m_ν< 0.077$ eV (95\% confidence level) in the CMB likelihood combination chosen by ACT when imposing the physical prior $\sum m_ν>0$ eV.

Figures (7)

• Figure 1: Left: The 68% and 95% constraints on the $H_0r_\mathrm{d}$-$\Omega_\mathrm{m}$ 2D parameter space for DESI DR2 BAO, ACT, PR4+ACT, and DESY5 SNe Ia, under the $\Lambda$CDM model. The combination based on ACT alone with SRoll2 is shown in green, while our baseline combination PR4+ACT is shown in purple. The ACT dataset shows a 2.7$\sigma$ tension with DESI (3.2$\sigma$ if CMB lensing is excluded), while the combination of PR4+ACT shows a 2.0$\sigma$ tension with DESI, once CMB lensing and low-$\ell$ TT data have been also included. Right: Tension between DESI DR2 BAO and different CMB variations (expressed in n$\sigma$ units), in the $\Lambda$CDM model. The tension is calculated given the 2D posterior distributions of $\Omega_\mathrm{m}$ and $H_0r_\mathrm{d}$. The first bars shown are PR4 (pink), ACT (green), P-ACT (orange) and PR4+ACT (purple). The hatched bars of the corresponding color represent variations of the CMB dataset.
• Figure 2: Left: The 68% and 95% confidence contours in the $w_0$-$w_a$ plane using DESI DR2 BAO data in combination with CMB data. The blue contour describes the results from the combination DESI+ACT, while the unfilled orange contour shows DESI in combination with the P-ACT dataset described in ACT:2025fju. The combination between DESI and our baseline CMB dataset with mixed multipole cuts in both Planck PR4 and ACT is shown in the unfilled green contour. The intersection of the two straight dashed gray lines represents the $\Lambda$CDM model. Right: Similar to the left panel but now including DESY5 data but using a different ranges for $w_0$ and $w_a$.
• Figure 3: 1D constraints on $w_0$ (left panel) and $w_a$ (middle panel), highlighting the robustness of the results against variations in the CMB dataset chosen for the analysis. In the left and central panels, the thick bars represent the 68% errors while the thin bars correspond to the 95% errors. The vertical black dashed lines represent the $\Lambda$CDM value for $w_0$ and $w_a$. The right panel shows the corresponding significance in favor of the $w_0w_a$CDM model.
• Figure 4: Whisker plots showing the 95% confidence constraints on $\sum m_\nu$ from the combination of DESI DR2 BAO with various CMB likelihoods under the $\Lambda$CDM model. Dashed lines correspond to constraints obtained using the low-$\ell$ EE SRoll2 likelihood, while solid lines use low-$\ell$ EE SimAll likelihood in each corresponding combination. The vertical dashed line and shaded region indicate the minimal sum of neutrino masses for the normal ($\sum m_\nu > 0.06$ eV) mass ordering.
• Figure 5: 1D marginalized posterior constraints on $\sum m_\nu$ from various dataset combinations within the $w_0w_a$CDM model. The black curves show results from combining DESI DR2 BAO with PR4+ACT. We also present posteriors obtained using DESI DR2 BAO with the main CMB combinations, all combined with DESY5. Dashed curves correspond to datasets that include the low-$\ell$ EE SRoll2 likelihood, while solid curves instead use the low-$\ell$ EE SimAll likelihood. The vertical dashed lines and shaded regions indicate the minimum sum of neutrino masses allowed for the normal ($\sum m_\nu > 0.06$ eV) and inverted ($\sum m_\nu > 0.10$ eV) mass orderings.