Suggestions of decreasing dark energy from supernova and BAO data: an update

TL;DR

The paper tests the possibility that dark energy is dynamical, driven by a scalar field with a linear potential, by constraining a two-parameter model $V_0$ and $V_1$ using the DES-SN5YR SN dataset and DESI DR2 BAO data. Using a Bayesian likelihood with MCMC and analytic marginalization over nuisance parameters, they find $V_1 \approx 1.49 \pm 0.25$, with ΛCDM ($V_1=0$) strongly disfavored under the combined data; the significance is sensitive to the inclusion of very low-redshift SNe. Mock data analyses reveal a positive $V_1$ bias, yet the real data lie several standard deviations away from pure ΛCDM expectations, supporting decreasing dark energy in this linear-potential framework. The quadratic extension provides no meaningful improvement, and the strength of the result depends on the low-z SNe, highlighting the need for complementary probes (e.g., ISW) to test late-time scalar dynamics and the holographic quantum gravity motivation. These findings elevate the case for dynamical dark energy and offer a concrete EFT-inspired scenario beyond ΛCDM, with implications for late-time cosmology and fundamental theory.

Abstract

In a previous work 2305.04946, we found that supernova and baryon acoustic oscillation data support the hypothesis that late time cosmic acceleration is caused by the potential energy of a scalar field descending its potential, as suggested by holographically defined models of quantum gravity. In this note, we update our analysis using the Dark Energy Survey 5 year supernova data set (DES-SN5YR) and the baryon acoustic oscillation data from the Dark Energy Spectroscopic Instrument Data Release 2 (DESI DR2). Approximating the scalar potential via a first order Taylor series $V \approx V_0 + V_1 φ$ about the present value, and making use of only recent-time data from DES-SN5YR and DESI DR2, we find that the slope parameter is constrained as $V_1 = 1.49 \pm 0.25$ in a standard likelihood analysis. This is naively a $>5 σ$ discrepancy with $Λ$CDM (which has $V_1 =0$), though a more detailed analysis not assuming a Gaussian likelihood distribution suggests $4 σ$ significance. Based only on the $Δχ^2 = -13.7$ improvement of fit while ignoring parameter space volumes disfavours $Λ$CDM at a $3 σ$ significance level. These significance measures are substantially improved from our previous analysis using older data sets. We also reproduce the DESI DR2 parameter constraints based on the same combination of data and find that the $Λ$CDM is more strongly disfavoured in the context of the linear potential extension (dubbed $V_0V_1$) as compared with the $w_0 w_a$ extension of $Λ$CDM. A caveat is that for both $w_0 w_a$ and $V_0 V_1$, much of the significance relies on the historical $z < 0.1$ supernova samples included in the DES-SN5YR data set.

Figures (9)

• Figure 1: Constraints on parameters $\Omega_M$, $V_0$, and $V_1$ for a linear potential scalar field model, using $e^{-\chi^2 / 2}$ likelihood calculated from combined DES-SN5YR supernova and DESI DR2 BAO data. The $\Lambda$CDM model corresponds to $V_1 = 0$, $\Omega_M + V_0 = 1$.
• Figure 2: Distribution of parameter values $[\Omega_M + V_0,V_1]$ for a linear potential scalar field model, using $e^{-\chi^2 / 2}$ likelihood calculated from DES-SN5YR supernova data (blue), DESI DR2 BAO data (orange), and combined data (purple). Each distribution includes 9,000 points. The $\Lambda$CDM model corresponds to $V_0 + \Omega_M = 1$, $V_1 = 0$. The two-pronged structure of the distribution is explained in the text.
• Figure 3: Distribution of parameter values $[\Omega_M + V_0,V_1]$ for a linear potential scalar field model with a prior that $\phi(t)$ is monotonic. The points follow an $e^{-\chi^2 / 2}$ distribution calculated from DES-SN5YR supernova data (blue), DESI DR2 BAO data (orange), and combined data (purple). Each distribution includes 1,100 points. The $\Lambda$CDM model corresponds to $V_0 + \Omega_M = 1$, $V_1 = 0$.
• Figure 4: Distribution of potential parameter values $V_0$ and $V_1$ for a linear potential scalar field model constrained by DES-SN5YR supernova data (green), DESI DR2 BAO data (pink), and combined data (blue). For each dataset, we show equal $\chi^2$ contours containing 68% and 95% of the $\exp(-\chi^2/2)$ distribution.
• Figure 5: Value of $\chi^2$ for the best fit linear potential model with various values of $V_1$. Here, we still allow for scalar kinetic energy, but the best fit $\Lambda$CDM model gives a similar $\chi^2$ to the best fit $V_1=0$ model overall.