Monodromic k-essence: Oscillatory Dark Energy

• Monodromic k-essence is a dark-energy model that combines a noncanonical k-essence sector with a monodromic modulation in the scalar potential to produce rapid oscillations around the phantom divide (w = -1).
• The model leverages a quadratic k-essence Lagrangian with a power-law envelope and sinusoidal modulation, enabling a tracking solution that mimics ΛCDM on average while allowing dynamic dark energy behavior.
• Observational constraints from DESI BAO, CMB, and supernovae show consistency with ΛCDM on average, yet specific datasets reveal mild to moderate preferences for oscillatory signatures.

Monodromic k-essence is a scalar-field dark-energy scenario in which a noncanonical k-essence sector is combined with a monodromic potential so that the dark-energy density and equation-of-state parameter can undergo rapid oscillations about the phantom divide $w_{\rm DE}=-1$ while remaining close to $\Lambda$CDM on average. In the explicit realization that has been confronted with DESI-era data, the action is

$$S=\int d^4x\sqrt{-g}\left[\frac{1}{2}M_{\rm Pl}^2R+p(\phi,X)+\mathcal{L}_m \right], \qquad X\equiv -\frac{1}{2}g^{\mu\nu}\partial_\mu \phi\partial_\nu\phi,$$

with

$$p(\phi,X)=V(\phi)[-X+X^2], \qquad V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right].$$

This formulation was introduced as the first observationally constrained example of monodromic k-essence at the background level, motivated by axion-monodromy-type constructions and by renewed interest in dynamical dark energy after DESI baryon acoustic oscillation measurements (Goldstein et al., 22 Jul 2025).

1. Definition and model architecture

In its strict contemporary usage, monodromic k-essence refers to a k-essence dark-energy model with a monodromic modulation in the scalar-sector “potential,” rather than to noncanonical scalar dynamics in general. The defining background-level realization employs the quadratic k-essence form $p(\phi,X)=V(\phi)[-X+X^2]$, together with a power-law envelope multiplied by a sinusoidal modulation,

$$V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right],$$

where $A$ is the modulation amplitude, $\nu$ the dimensionless frequency, $\delta$ the phase, and $\alpha$ the power-law index of the smooth envelope (Goldstein et al., 22 Jul 2025).

The construction is physically motivated by a weakly broken shift symmetry, which allows a smooth, nearly flat dark-energy sector together with periodic modulation. In the fiducial analysis, the field is initialized on the tracking solution that exists when $\Lambda$0; accordingly, $\Lambda$1 and $\Lambda$2 are fixed by the present-day dark-energy density rather than sampled independently. The smooth power-law envelope governs the average evolution, while the sinusoidal factor generates the oscillatory component.

A central distinction from canonical quintessence is explicit. Canonical quintessence cannot cross the phantom divide in the relevant way, whereas k-essence can exhibit much richer oscillatory behavior. This is the core reason that monodromic k-essence was proposed as a scalar-field realization of oscillatory dark energy rather than as a purely phenomenological $\Lambda$3 parameterization (Goldstein et al., 22 Jul 2025).

2. Oscillatory dark-energy dynamics

At the background level, the monodromic modulation drives the dark-energy equation of state above and below $\Lambda$4, so that the model can cross or oscillate around the phantom divide. The reconstructed $\Lambda$5 obtained from the observational analysis “roughly oscillates around the cosmological constant $\Lambda$6” for all dataset combinations considered, and that behavior is the primary phenomenological target of the scenario (Goldstein et al., 22 Jul 2025).

The parameter roles are sharply separated. Increasing $\Lambda$7 increases the oscillation amplitude. Changing $\Lambda$8 shifts the average evolution and therefore the time-averaged dark-energy density tilt. The frequency $\Lambda$9 controls oscillation spacing, and $$S=\int d^4x\sqrt{-g}\left[\frac{1}{2}M_{\rm Pl}^2R+p(\phi,X)+\mathcal{L}_m \right], \qquad X\equiv -\frac{1}{2}g^{\mu\nu}\partial_\mu \phi\partial_\nu\phi,$$0 shifts the oscillation relative to the observed redshift window. The fiducial prior $$S=\int d^4x\sqrt{-g}\left[\frac{1}{2}M_{\rm Pl}^2R+p(\phi,X)+\mathcal{L}_m \right], \qquad X\equiv -\frac{1}{2}g^{\mu\nu}\partial_\mu \phi\partial_\nu\phi,$$1 was chosen so that the oscillations are neither too slow nor too rapid to resolve with the DESI redshift bins.

The same structure also explains why the analysis was deliberately restricted to the background level. Ordinary k-essence models that cross $$S=\int d^4x\sqrt{-g}\left[\frac{1}{2}M_{\rm Pl}^2R+p(\phi,X)+\mathcal{L}_m \right], \qquad X\equiv -\frac{1}{2}g^{\mu\nu}\partial_\mu \phi\partial_\nu\phi,$$2 can suffer gradient instabilities because the squared sound speed can become negative. For that reason, the observational study used only data that are insensitive to dark-energy perturbations, and a modified stabilized construction was deferred to an appendix (Goldstein et al., 22 Jul 2025).

3. Observational constraints from CMB, DESI BAO, and supernovae

The first dedicated observational study of monodromic k-essence combined a compressed late-time CMB likelihood $$S=\int d^4x\sqrt{-g}\left[\frac{1}{2}M_{\rm Pl}^2R+p(\phi,X)+\mathcal{L}_m \right], \qquad X\equiv -\frac{1}{2}g^{\mu\nu}\partial_\mu \phi\partial_\nu\phi,$$3, the full DESI DR2 combined BAO sample over $$S=\int d^4x\sqrt{-g}\left[\frac{1}{2}M_{\rm Pl}^2R+p(\phi,X)+\mathcal{L}_m \right], \qquad X\equiv -\frac{1}{2}g^{\mu\nu}\partial_\mu \phi\partial_\nu\phi,$$4, and two alternative supernova compilations: Pantheon-Plus, with 1550 spectroscopically confirmed SNe Ia over $$S=\int d^4x\sqrt{-g}\left[\frac{1}{2}M_{\rm Pl}^2R+p(\phi,X)+\mathcal{L}_m \right], \qquad X\equiv -\frac{1}{2}g^{\mu\nu}\partial_\mu \phi\partial_\nu\phi,$$5, and DESY5, with 1635 photometrically classified SNe Ia plus 194 low-redshift SNe. The sampled fiducial priors were

$$S=\int d^4x\sqrt{-g}\left[\frac{1}{2}M_{\rm Pl}^2R+p(\phi,X)+\mathcal{L}_m \right], \qquad X\equiv -\frac{1}{2}g^{\mu\nu}\partial_\mu \phi\partial_\nu\phi,$$6

with inference performed using hi_class, Cobaya, the Gelman-Rubin criterion, and MAP optimization (Goldstein et al., 22 Jul 2025).

The main background-level constraints are as follows.

Dataset combination	Main parameter result	Statistical note
$$S=\int d^4x\sqrt{-g}\left[\frac{1}{2}M_{\rm Pl}^2R+p(\phi,X)+\mathcal{L}_m \right], \qquad X\equiv -\frac{1}{2}g^{\mu\nu}\partial_\mu \phi\partial_\nu\phi,$$7 DESI BAO	$$S=\int d^4x\sqrt{-g}\left[\frac{1}{2}M_{\rm Pl}^2R+p(\phi,X)+\mathcal{L}_m \right], \qquad X\equiv -\frac{1}{2}g^{\mu\nu}\partial_\mu \phi\partial_\nu\phi,$$8 prior-dominated; $$S=\int d^4x\sqrt{-g}\left[\frac{1}{2}M_{\rm Pl}^2R+p(\phi,X)+\mathcal{L}_m \right], \qquad X\equiv -\frac{1}{2}g^{\mu\nu}\partial_\mu \phi\partial_\nu\phi,$$9	Consistent with $$p(\phi,X)=V(\phi)[-X+X^2], \qquad V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right].$$0CDM; $$p(\phi,X)=V(\phi)[-X+X^2], \qquad V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right].$$1
Baseline + Pantheon-Plus	$$p(\phi,X)=V(\phi)[-X+X^2], \qquad V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right].$$2 at $$p(\phi,X)=V(\phi)[-X+X^2], \qquad V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right].$$3 C.L.	Still consistent with zero; $$p(\phi,X)=V(\phi)[-X+X^2], \qquad V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right].$$4
Baseline + DESY5	$$p(\phi,X)=V(\phi)[-X+X^2], \qquad V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right].$$5, $$p(\phi,X)=V(\phi)[-X+X^2], \qquad V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right].$$6, $$p(\phi,X)=V(\phi)[-X+X^2], \qquad V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right].$$7, $$p(\phi,X)=V(\phi)[-X+X^2], \qquad V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right].$$8	Mild-to-moderate preference; $$p(\phi,X)=V(\phi)[-X+X^2], \qquad V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right].$$9

For the baseline $p(\phi,X)=V(\phi)[-X+X^2]$0DESI combination, the amplitude is not detected and remains prior-dominated, with $p(\phi,X)=V(\phi)[-X+X^2]$1 and $p(\phi,X)=V(\phi)[-X+X^2]$2 also unconstrained at $p(\phi,X)=V(\phi)[-X+X^2]$3. This means that CMB+BAO alone remain consistent with standard $p(\phi,X)=V(\phi)[-X+X^2]$4CDM. By contrast, the DESY5 supernova sample yields a non-zero oscillatory amplitude at the $p(\phi,X)=V(\phi)[-X+X^2]$5 level, while Pantheon-Plus yields no evidence for such a signal. In the DESY5 fit, the corresponding matter density and Hubble constant are roughly $p(\phi,X)=V(\phi)[-X+X^2]$6 and $p(\phi,X)=V(\phi)[-X+X^2]$7 km/s/Mpc (Goldstein et al., 22 Jul 2025).

The goodness-of-fit comparison was made relative to $p(\phi,X)=V(\phi)[-X+X^2]$8CDM and to the phenomenological CPL $p(\phi,X)=V(\phi)[-X+X^2]$9-$$V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right],$$0 model. Monodromic k-essence yields $$V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right],$$1, $$V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right],$$2, and $$V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right],$$3 for the baseline, baseline+Pantheon-Plus, and baseline+DESY5 combinations, respectively, while CPL gives $$V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right],$$4, $$V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right],$$5, and $$V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right],$$6. The monodromic model therefore matches the phenomenological alternative “with comparable $$V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right],$$7,” although it does so with more parameters (Goldstein et al., 22 Jul 2025).

4. Robustness, dataset dependence, and statistical interpretation

The observational status of monodromic k-essence is driven less by a uniform trend across all datasets than by a small number of influential measurements. A major theme of the analysis is sensitivity to the DESI DR2 LRG2 BAO distance, especially the parallel measurement at $$V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right],$$8. The best-fit monodromic model for the baseline data is largely driven by the LRG2 “dip” in the parallel BAO distance, and this same feature also helps drive the dynamical-dark-energy preference in the $$V(\phi)=C\left(\frac{\phi}{\phi_0}\right)^{-\alpha}\left[1-A\sin(\nu H_0 \phi+\delta)\right],$$9-$A$0 comparison (Goldstein et al., 22 Jul 2025).

When the LRG2 point is removed, the preference for monodromic k-essence drops markedly: from $A$1, $A$2, and $A$3 to $A$4, $A$5, and $A$6 for the baseline, baseline+Pantheon-Plus, and baseline+DESY5 combinations, respectively. The corresponding $A$7 values also decrease in magnitude, and the oscillation frequency becomes substantially less constrained. Mock $A$8CDM BAO data were used to test whether finite DESI redshift binning itself artificially selects an oscillation scale; the conclusion was negative, so the LRG2 point rather than the binning is the main source of the oscillatory preference.

The resulting interpretation is therefore cautious. The data do not provide substantial evidence for monodromic k-essence over $A$9CDM in general, because the significance is low for the baseline and Pantheon-Plus combinations and only moderate for DESY5. The model is data-compatible, and the DESY5 combination yields a mild to moderate preference for a nonzero oscillatory amplitude, but the evidence is not robust across supernova compilations and depends strongly on the LRG2 BAO measurement (Goldstein et al., 22 Jul 2025).

5. Perturbative viability and the broader k-essence stability problem

The restriction of the monodromic analysis to background observables is part of a broader issue in k-essence cosmology: perturbative dynamics depend on the scalar sound speed and cannot, in general, be inferred from the background equation of state alone. For a general k-essence action $\nu$0, the effective sound speed is

$\nu$1

and the full linear system can produce oscillatory matter-density modes of the form

$\nu$2

It was shown that the usual subhorizon approximation can be too strong in k-essence because small scale does not automatically imply quasi-static scalar perturbations (Bamba et al., 2011).

This perturbative lesson is directly relevant to monodromic k-essence. A plausible implication is that any perturbatively complete monodromic model must control not only $\nu$3 but also the sign and evolution of $\nu$4, especially near phantom-divide crossings. That inference is consistent with the explicit caution that ordinary k-essence crossing $\nu$5 can develop gradient instabilities and that the first monodromic analysis was therefore kept background-only (Goldstein et al., 22 Jul 2025).

A related stability perspective comes from a purely kinetic quadratic model,

$\nu$6

which arises from a vectorial-nonmetricity geometry and satisfies

$\nu$7

There it was found that, if physical viability conditions such as $\nu$8 and $\nu$9 are not enforced, the phase space generically exhibits instabilities and divergent behaviour; after imposing those conditions, the model becomes statistically indistinguishable from $\delta$0CDM at late times (Csillag et al., 21 May 2025). Although this construction is not monodromic, it sharpens the general point that viable k-essence parameter inference depends sensitively on stability priors.

6. Related non-monodromic k-essence constructions and scope of the term

A common source of ambiguity is the tendency to use “monodromic k-essence” loosely for any noncanonical scalar theory with unusual kinetic structure. The literature summarized here does not support that broad usage. Several papers study models that are relevant only indirectly, because they illuminate noncanonical kinetic sectors, dust-like limits, higher-dimensional origins, or branch-like scalar behavior, but they are not presented as monodromic k-essence in the axion-monodromy-inspired dark-energy sense.

In minisuperspace quantum cosmology, one class of models adopts

$\delta$1

with effective perfect-fluid equation of state

$\delta$2

In the distinguished pressureless limit $\delta$3, the Hamiltonian becomes

$\delta$4

so that the scalar field acts as an internal time variable and the quantum model yields a nonsingular bounce with

$\delta$5

This is a study of power-law k-essence and relational time, not of monodromy (Almeida et al., 2016).

In the Stephani-universe context, the exotic source can be written as a k-essence model linear in the scalar “velocity,”

$\delta$6

with homogeneous energy density $\delta$7 and inhomogeneous pressure. That model is then interpreted as a dimensional reduction of a five-dimensional nonlinear electrodynamics. The construction is explicitly said not to involve monodromy, branched potentials, axion monodromy, or winding sectors (Marais et al., 2015).

Within two-field measure theory, the k-essence sector is modified by the constraint $\delta$8, leading universally to

$\delta$9

This produces a dust term and a cosmological-constant term for any k-essence model, thereby furnishing a unified dark matter/dark energy description. Again, the result concerns a measure-theoretic reformulation of k-essence rather than a monodromic potential (Cordero et al., 2019).

Black-bounce solutions provide another indirect connection. In a bumblebee-gravity setting with Lorentz-symmetry violation, a power-law k-essence model

$\alpha$0

supports regular black-bounce geometries. The reconstructed scalar and potential contain $\alpha$1 and logarithmic structures and display nontrivial finite asymptotics, which is suggestive of branch-like behavior, but the model is not identified as monodromic k-essence (Pereira et al., 13 Mar 2025).

These neighboring constructions show that noncanonical kinetics, unusual fluid correspondences, higher-dimensional reductions, and branch-like scalar profiles are widespread in k-essence. This suggests that the term “monodromic k-essence” is best reserved for models that combine a k-essence kinetic sector with an explicit monodromic modulation of the scalar potential or background evolution, as in the DESI-era oscillatory dark-energy scenario (Goldstein et al., 22 Jul 2025).