Dynamical Dark Energy: Models and Constraints

• Dynamical dark energy is defined as models with a time-evolving equation of state, contrasting the constant value of Λ and addressing issues like fine-tuning and data tensions.
• Observational analyses using BAO, CMB, and Type Ia supernovae data indicate improved fits and moderate-to-strong statistical preference for DDE, though systematic discrepancies persist.
• Future multi-probe surveys and advanced simulations are essential to validate DDE models and resolve current uncertainties between different cosmological datasets.

Dynamical dark energy (DDE) refers to models in which the cosmic acceleration is driven by a component with time-varying equation of state, in contrast to the constant vacuum energy density of the standard cosmological constant ($\Lambda$). DDE models are motivated both by attempts to resolve fine-tuning and coincidence problems intrinsic to $\Lambda$CDM, and by possible tensions or anomalies found in cosmological datasets. The last several years have seen increasingly strong constraints and, in some analyses, significant statistical preference for DDE over $\Lambda$CDM, largely due to improved baryon acoustic oscillation (BAO), cosmic microwave background (CMB), and Type Ia supernova (SN) data, notably from DESI and JWST. However, these claims are subject to crucial systematic concerns and ongoing debate regarding dataset consistency and model selection.

1. Theoretical Frameworks and Parameterizations

DDE is most commonly parameterized via the equation-of-state (EoS) parameter $w(a) = p_{\rm DE} / \rho_{\rm DE}$, generalizing $w(a) = -1$ for $\Lambda$. The Chevallier–Polarski–Linder (CPL) form is widely adopted: $w(a) = w_0 + w_a (1 - a)$ where $w_0$ is the present-day EoS, $w_a = -dw/da$ sets the first-order deviation, and $a = (1+z)^{-1}$ is the scale factor. $\Lambda$0CDM is recovered for $\Lambda$1. Several two-parameter forms with distinct redshift evolution are also used, including the Barboza-Alcaniz (BA), Jassal-Bagla-Padmanabhan (JBP), and exponential parameterizations. Extended multi-parameter forms have also been explored, such as transition models with asymmetric width and variable early-time limit, e.g.

$\Lambda$2

with $\Lambda$3 encoding the rapidity and timing of the transition from $\Lambda$4 (early time) to $\Lambda$5 (today) (Cheng et al., 10 Dec 2025). Physically, DDE may arise from scalar field models (quintessence/phantom/“quintom”), running vacuum, or phenomenological extensions.

The impact of DDE is encoded in the modified Friedmann equation: $\Lambda$6 where BAO, SN, CMB, and large-scale structure observables depend on the expansion and growth histories integrated over $\Lambda$7.

2. Observational Constraints and Statistical Evidence

Recent DESI BAO data, when combined with Planck CMB and high-precision SNe, have delivered stringent joint constraints on DDE. Multiple data combinations now show frequentist significances for DDE at $\Lambda$8–$\Lambda$9 and Bayesian evidence classified as “moderate” to “strong” in favor of DDE over $\Lambda$0CDM, provided SN data (especially DESY5) are included (Li et al., 27 Nov 2025, Giarè et al., 2024, Sabogal et al., 30 May 2025). Across all tested two-parameter forms, best-fit values remain robustly in the “quintessence today, phantom in past” region (i.e., $\Lambda$1, $\Lambda$2):

Model	$\Lambda$3	$\Lambda$4	Crossing redshift $\Lambda$5
CPL (DESI+DESY5)	$\Lambda$6	$\Lambda$7	$\Lambda$8
BA (DESI+DESY5)	$\Lambda$9	$w(a) = p_{\rm DE} / \rho_{\rm DE}$0	$w(a) = p_{\rm DE} / \rho_{\rm DE}$1
JBP	$w(a) = p_{\rm DE} / \rho_{\rm DE}$2	$w(a) = p_{\rm DE} / \rho_{\rm DE}$3	$w(a) = p_{\rm DE} / \rho_{\rm DE}$4

Combined analyses using galaxy–CMB lensing cross-correlations (DESI LRG × CMB lensing) and geometric probes (BAO, SN) independently yield $w(a) = p_{\rm DE} / \rho_{\rm DE}$5 evidence for DDE, with $w(a) = p_{\rm DE} / \rho_{\rm DE}$6, $w(a) = p_{\rm DE} / \rho_{\rm DE}$7 (Sabogal et al., 30 May 2025). Four-parameter forms show smooth transitions from phantom at early times to quintessence at present, with $w(a) = p_{\rm DE} / \rho_{\rm DE}$8 and $w(a) = p_{\rm DE} / \rho_{\rm DE}$9; these models are moderately favored over $w(a) = -1$0CDM when Planck, DESI BAO, and DESY5 SNe are combined (Cheng et al., 10 Dec 2025).

The improvement in fit quality is substantial; Planck + DESI + DESY5 typically achieves $w(a) = -1$1 to $w(a) = -1$2 over $w(a) = -1$3CDM, with moderate Bayes factors ($w(a) = -1$4–$w(a) = -1$5). The effective dark energy density $w(a) = -1$6 generically peaks at intermediate redshift and declines toward $w(a) = -1$7, and the deceleration parameter shows an earlier transition to acceleration than $w(a) = -1$8CDM (Li et al., 27 Nov 2025).

3. Robustness, Systematics, and Dataset Tension

Despite strong statistical claims, the robustness of DDE detections is actively debated due to significant internal tension among CMB, BAO, and SN datasets (Colgáin et al., 2024, Wang et al., 21 Apr 2025). When each probe is analyzed independently (CMB, BAO, SN), all prefer DDE over $w(a) = -1$9CDM by model-selection criteria (Bayesian Information Criterion or Bayes factor), but the allowed regions for $\Lambda$0 are broad and sometimes mutually incompatible. The most significant driver of the joint DDE signal is low-redshift SNe, especially from DESY5, which are in $\Lambda$1–$\Lambda$2 tension with DESI BAO/full-shape measurements at the same effective redshifts (Colgáin et al., 2024).

Inconsistencies also appear in reconstructed $\Lambda$3: SNe prefer higher values at low $\Lambda$4 compared to BAO/full-shape clustering, translating into geometry (distance modulus) tensions. Cross-checks reveal that neither BAO nor SN data alone robustly detect cosmic acceleration or negative pressure at $\Lambda$5; only combined datasets excluding clear outliers achieve $\Lambda$6 detections (Wang et al., 21 Apr 2025).

Several works argue that the apparent DDE signal may be an artifact of unaccounted systematics in SNe (calibration, selection, environmental effects) or BAO (modeling, sample variance at low $\Lambda$7). A bona fide DDE detection should be consistently confirmed by independent observables (BAO, SN, CMB, lensing, growth rate) tracing the same evolution at overlapping redshift (Colgáin et al., 2024).

4. Astrophysical and Structure Formation Constraints

DDE models imprint percent- to 10%-level effects on the large-scale structure, growth rate, and abundance of massive objects. High-resolution hydrodynamical and $\Lambda$8-body simulations tailored to CMB-consistent DDE cosmologies show that "thawing" models ($\Lambda$9) enhance matter clustering and massive halo abundances, while "freezing" models ($w(a) = w_0 + w_a (1 - a)$0) suppress them (Pfeifer et al., 2020, Ishiyama et al., 25 Mar 2025). For example, a DDE model with $w(a) = w_0 + w_a (1 - a)$1, $w(a) = w_0 + w_a (1 - a)$2 (DESI+Y1 best-fit) produces a 10% power excess at small scales, 15% suppression at large scales, and up to 70% more massive halos at $w(a) = w_0 + w_a (1 - a)$3 compared to Planck18 $w(a) = w_0 + w_a (1 - a)$4CDM (Ishiyama et al., 25 Mar 2025). Baryonic feedback and galaxy formation effects can be modeled independently from DDE, with their impact typically separable to $w(a) = w_0 + w_a (1 - a)$5 uncertainty (Pfeifer et al., 2020, Casarini et al., 2010).

Indirect constraints on DDE have also been derived from the abundance of high-redshift supermassive black holes and massive galaxies at $w(a) = w_0 + w_a (1 - a)$6, as these require extended cosmic time and rapid structure growth. The observed abundance of massive galaxies at $w(a) = w_0 + w_a (1 - a)$7 (CANDELS, submm galaxies at $w(a) = w_0 + w_a (1 - a)$8, SPT0311-58 at $w(a) = w_0 + w_a (1 - a)$9, and $w_0$0 JWST galaxies) excludes major fractions of the DDE parameter space favored by other cosmological probes (Santini et al., 2023). In particular, models with $w_0$1 or positive $w_0$2 (slowly evolving or “Quintom” scenarios) are strongly disfavored except in the phantom or rapidly evolving regime.

5. Signatures in Fluctuations and the Lyman-Alpha Forest

DDE influences not only the background but also the perturbation sector, including scale-dependent linear growth and halo bias. Two-fluid DDE models with sound speeds $w_0$3–$w_0$4 generate observable scale-dependent growth in galaxy power spectra and bispectra; bispectrum tomography is required to detect the DDE sound mode at significant signal-to-noise ratios in surveys with $w_0$5Gpc$w_0$6 (Die et al., 5 Mar 2026).

Hydrodynamical simulations reveal that DDE models matching DESI constraints predict percent-level changes in the Lyman-alpha forest transmitted flux power spectrum: a redshift- and scale-dependent “spectral tilt,” higher IGM temperatures, and reduced Ly$w_0$7 opacity compared to $w_0$8CDM (Garza et al., 2 Jan 2026). These effects are within reach of current and forthcoming surveys (DESI DR2+, WEAVE, ELT) but require percent-level control of systematics.

6. Model Selection and Alternative Explanations

While DDE models improve fit quality relative to $w_0$9CDM and, in some cases, outperform even four-parameter modified early recombination models, the parameter constraints and interpretation are highly dependent on the choice of $w_a = -dw/da$0 parameterization, redshift sensitivity, and adopted priors (Mirpoorian et al., 21 Apr 2025, Colgáin et al., 2021). The CPL form is conservative in sensitivity to $w_a = -dw/da$1; alternatives with more rapidly varying $w_a = -dw/da$2 can deliver tighter constraints or earlier apparent detections.

Alternative explanations for joint BAO–CMB tension, such as early-modified recombination or running vacuum models, can provide equally good or better fits to current data and help to alleviate the $w_a = -dw/da$3 discrepancy without invoking late-time DDE (Mirpoorian et al., 21 Apr 2025, Peracaula et al., 2016).

7. Future Prospects and Open Issues

Establishing DDE as a genuine physical component requires next-generation datasets with reduced systematics, greater cross-survey consistency, and multi-probe coherence. Full-shape galaxy clustering, weak lensing, growth rate ($w_a = -dw/da$4), CMB lensing, and Lyman-alpha forest measurements will provide complementary, independent constraints. Refinement of astrophysical systematics in SNe and galaxy samples is mandatory. Theoretical advances in model selection, parameterization priors, and field-theory consistency for non-parametric $w_a = -dw/da$5 reconstructions are also required (Colgáin et al., 2021).

At present, observational evidence—while increasingly suggestive—remains subject to cross-dataset consistency checks. Only persistent, high-significance departures from $w_a = -dw/da$6CDM found across all independent channels can establish dynamical dark energy as a physical reality. Ongoing and planned surveys (DESI, Euclid, LSST, CMB-S4, next-generation Lyman-alpha programs) will be decisive in resolving the dynamical vs. constant dark energy debate.