DESI Dark Secrets

TL;DR

DESI DR1/DR2 analyses probe whether dark energy deviates from a cosmological constant by testing the standard $w_0w_a$ parameterization against physics-motivated scalar-field models. The study finds that while scalar-field realizations with a single parameter $\beta$ can mimic the DESI expansion history at the percent level, they do not significantly improve fits over ΛCDM, and a sharply peaked dark-energy density around $z\simeq 0.5$ is most favored by DESI data. Joint constraints from DR2, CMB, and Pantheon+SH0ES mildly prefer a scalar-field interpretation with $\beta$ in the approximate range $0.23$–$0.95$, though $w_0w_a$ remains preferred in DESI-only analyses; Bayesian evidence still favors ΛCDM when model complexity is penalized. The work highlights the limitations of $w_0w_a$ as a universal descriptor for dark energy, the potential value of scalar-field dynamics, and the possibility that the DESI-inferred evolution could reflect nonstandard matter evolution or other physics beyond a simple two-component model.

Abstract

The first year results of DESI (DR1) provide evidence that dark energy may not be quantum vacuum energy ($Λ$). If true, this would be an extraordinary development in the 25-year quest to understand cosmic acceleration. The best-fit DESI $w_0w_a$ models for dark energy, which underpin the claim, have strange behavior. They achieve a maximum energy density around $z\simeq 0.5 $ and rapidly decrease before and after. We explore physics-based models where the dark energy is a rolling scalar-field. Our four scalar-field models are characterized by one dimensionless parameter $β$, which in the limit of $β\rightarrow 0$ reduces to $Λ$CDM. While none of our models fit the DESI data significantly better than $Λ$CDM, for values of $β$ of order unity, they fit about as well as $Λ$CDM. We also consider the second data release from DESI (DR2), CMB data and supernovae data. The DR2 results are consistent with the DR1, and the combination of DESI, CMB and SNe favor $β= 0.23 - 0.95$, providing some evidence for a scalar-field explanation for dark energy. While the DESI data prefer $w_0w_a$ to a scalar field, the SNe data prefer a scalar field to $w_0w_a$, and together they favor a $w_0w_a$ model. We study the limits of $w_0w_a$ in describing dark energy, especially scalar field models, and also point out that the strange behavior of the best-fit DESI models could arise due to the matter density not varying as expected or an unaccounted for component of energy density in the Universe.

Figures (19)

• Figure 1: Dark energy EOS as a function of redshift for $w_0w_a$ models: $\Lambda$ ($w_0 = -1$, $w_a =0$); DESI+ best fit ($w_0 = -0.7$, $w_a = -1.0$) and the behavior expected for a rolling scalar field ($w_0 = -0.7$ and $w_a = -0.3$). Unless $\alpha$ and $w_a$ have opposite signs, $w(z)$ will cross the phantom line ($w=-1$), which is what the DESI+ best fit model does.
• Figure 2: Evolution of the energy density of dark energy (in units of the critical density today) for the extreme DESI best-fit (blue), the DESI+ best-fit (red), and $\Lambda$. The extreme DESI best fit has $\Omega_M = 0.4$, $w_0=0.016$, $w_a = -3.69$, $\chi^2 = 8.53$ (as discussed in Sec. \ref{['sec3.1']}). The DESI+ best-fit includes the other data sets and has $\Omega_M = 0.33$, $w_0=-0.7$, $w_a = -1$, $\chi^2 = 9.00$ (for the DESI data only).
• Figure 3: The expansion rate for the DESI only and DESI+ best-fit $w_0w_a$ models divided by that for $\Lambda$CDM, with the same value of $\Omega_M$.
• Figure 4: The ratio of the expansion rate squared of the DESI+ best-fit $w_0w_a$ (red) and several of our scalar field model to $\Lambda$CDM. In particular, $\beta = 1.8$ (massive, green), $\beta = 1.9$ (quartic, orange), $\beta = 6$ (exponential, blue) and $\beta = 1.8$ (tachyonic, violet). We have chosen the values of $\beta$ to achieve good "visual" agreement with the expansion history for the best-fit DESI+ model.
• Figure 5: EOS $w(z)$ for the DESI+ best-fit $w_0w_a$ model and several of our scalar field models. In particular, $\beta = 2$ (massive), $\beta = 3$ (quartic), $\beta = 6.9$ (exponential) and $\beta = 1.8$ (tachyonic). While scalar field models can reproduce the expansion history of the DESI+ model, cf. Fig. \ref{['H2ALL']}, they do so with very different EOS histories. Further, we have chosen $\beta$ for the non-tachyonic models such that the dimensionless initial slope is the same, cf. Sec. \ref{['universal_sf_behavior']}, illustrating that the evolution of the EOS is similar but not identical when $\beta$ is of order unity.