Exploring the interplay of late-time dynamical dark energy and new physics before recombination

Abstract Cosmological models in which the effective dark energy (DE) equation-of-state parameter crosses the phantom divide at z c ∼ 0.3-0.6, while exhibiting quintessence-like behavior at low redshift, provide a significantly improved fit to current datasets, yielding indications of late-time DE dynamics at the ∼ 3 σ CL. However, such scenarios continue to favor low values of the Hubble constant, remaining in tension with the SH0ES distance-ladder determination. In the absence of unaccounted-for systematic effects, this tension may point to the presence of new physics operating prior to the decoupling of CMB photons. In this work, we reconstruct the relevant background DE functions using the Weighted Function Regression (WFR) method, introducing three main improvements with respect to our previous analysis in JCAP 12 (2025) 049. First, we adopt the Frequentist-Bayesian approach for determining the WFR weights, and demonstrate full consistency with results obtained using information criteria. Second, we combine state-of-the-art CMB and BAO measurements with the newly recalibrated DES supernova sample (DES-Dovekie), and compare our findings with those derived from Pantheon+, still assuming standard pre-recombination physics. Third, we study in a model-independent manner the viability of the early-time “solutions” to the Hubble tension and how they affect the evidence for dynamical DE at late times, under the influence of the SH0ES and the more conservative CCHP calibration of the cosmic ladders, separately. We find that, if the physics prior to decoupling is unmodified, the probability of phantom crossing is ∼ 96.7–98.5%, with ΛCDM excluded at ∼ 2.5 σ and ∼ 3 σ CL for the models with the largest weights when using Pantheon+ and DES-Dovekie, respectively. New physics before recombination can alleviate the Hubble tension, but requires extremely large values of the reduced matter density parameter ω m when the SH0ES calibration is employed, in strong tension with those inferred from full CMB analyses. This raises serious concerns about the actual viability of these models to explain the SH0ES measurement. We find that phantom crossing, while not excluded, is no longer required, with only a very mild preference for quintessence. Nevertheless, given the aforementioned tension in ω m , it would be rash to draw firm conclusions about how the dynamical DE signal is affected in these scenarios.