Revisiting the evidence for dynamical dark energy: from DES-SN5Y to DES-Dovekie

Abstract Recent analyses combining DESI DR2 BAO measurements with CMB and supernova data have suggested a preference for dynamical dark energy, particularly when using the DES-SN5Y and DES-SN5Y $$(z>0.1)$$ ( z > 0.1 ) supernova samples. With the revision of this dataset into the DES-Dovekie compilation, it is important to reassess the robustness of this result. In this work, we analyze a broad class of dark energy parametrizations, including w CDM, $$w_0w_a$$ w 0 w a CDM, BA, JBP, Efstathiou, and GEDE models, using DESI DR2 BAO, a compressed CMB likelihood, and supernova data. We find that the DES-SN5Y dataset consistently yields a stronger preference for deviations from $$\Lambda $$ Λ CDM compared to DES-Dovekie. In particular, redshift-dependent parametrizations favor the region $$w_0 > -1$$ w 0 > - 1 and $$w_a < 0$$ w a < 0 , corresponding to a Quintom-B–type evolution with phantom crossing. This behavior is observed across multiple models, indicating that it is not specific to a single parametrization. We find that the DES-SN5Y $$(z>0.1)$$ ( z > 0.1 ) sample also preserves the same qualitative behavior, although the deviations from the $$\Lambda $$ Λ CDM limit become significantly weaker compared to the full DES-SN5Y sample. However, when DES-Dovekie is used, the statistical preference for dynamical dark energy is significantly reduced. The improvement over $$\Lambda $$ Λ CDM drops to below the $$1\sigma $$ 1 σ level for most models, and the Bayesian evidence weakens from moderate to weak or inconclusive. Similarly, for the DES-SN5Y $$(z>0.1)$$ ( z > 0.1 ) sample, the preference for dynamical dark energy becomes intermediate between the DES-SN5Y and DES-Dovekie cases, with mostly weak or inconclusive evidence. Overall, the DES-Dovekie and DES-SN5Y $$(z>0.1)$$ ( z > 0.1 ) datasets weaken the preference for dynamical dark energy compared to the full DES-SN5Y dataset.