Phantom-Crossing Cosmology

• Phantom-Crossing Behaviour is the evolution of dark energy's equation-of-state parameter through the phantom divide (w=-1), marking transitions between quintessence and phantom regimes.
• It is realized through various frameworks including modified gravity, braneworld cosmology, and interacting dark sectors, each offering distinct observational signatures.
• Observational constraints and theoretical diagnostics show that an effective phantom crossing does not necessarily imply a fundamental phantom field, raising open questions on dark energy microphysics.

Phantom-crossing behaviour in cosmology denotes evolution of an equation-of-state parameter through the phantom divide $w=-1$, so that a component or effective component passes between a non-phantom regime $w>-1$ and a phantom regime $w<-1$. In contemporary late-time cosmology, the term is used in several related but non-identical senses: for a fundamental dark-energy sector, for an effective dark-energy component reconstructed from the background expansion, and for geometric terms in modified gravity or braneworld cosmology rewritten as an effective fluid. This distinction is central, because the same background signature can arise from intrinsically non-phantom physics, dark-sector interactions, or modified gravity rather than from a fundamental phantom field (Hirano et al., 2010, Mishra, 26 May 2026).

1. Definition and kinematic meaning

The phantom divide is the line

$w=-1,$

where $w\equiv p/\rho$. In the standard late-time classification used throughout the literature summarized here, $w=-1$ corresponds to a cosmological constant, $w>-1$ to quintessence-like behaviour, and $w<-1$ to phantom-like behaviour. Phantom crossing means that $w(z)$ changes sign relative to $w+1$ during cosmic evolution, either from $w>-1$0 to $w>-1$1 or in the reverse direction (Hirano et al., 2010).

In effective background reconstruction, the relevant quantity is often not a fundamental fluid but a residual component defined after subtracting a conserved matter sector from the observed expansion history. Under the assumptions of a spatially flat FLRW universe, the standard Friedmann equation of General Relativity, and separately conserved non-relativistic matter,

$w>-1$2

so that

$w>-1$3

The associated effective equation of state is

$w>-1$4

Within this framework, effective phantom behaviour is equivalent to a dark-energy density that grows with cosmic time: $w>-1$5 Phantom crossing is then the change in sign of this derivative across $w>-1$6 (Mishra, 26 May 2026).

A closely related viewpoint is to parameterize the dark-energy density directly around an extremum. In the phenomenological model

$w>-1$7

the first-order term is omitted because $w>-1$8 is taken to be an extremum. Since the continuity equation gives

$w>-1$9

crossing $w<-1$0 is tied to $w<-1$1, i.e. to the extremum at $w<-1$2 (Valentino et al., 2020).

For the CPL form

$w<-1$3

the crossing redshift is

$w<-1$4

provided the denominator is nonzero (Mishra, 26 May 2026).

2. Structural obstructions to crossing

A recurring result in the literature is that many simple models cannot cross the phantom divide for structural reasons. In the original self-accelerating DGP braneworld, the modified Friedmann equation is

$w<-1$5

and on the self-accelerating branch $w<-1$6 the effective equation of state of the DGP term can be written as

$w<-1$7

For realistic $w<-1$8 and $w<-1$9, this remains above $w=-1,$0, so late-time acceleration occurs without phantom crossing (Hirano et al., 2010).

The Dvali–Turner extension,

$w=-1,$1

adds a constant exponent $w=-1,$2, but does not resolve the obstruction. If $w=-1,$3, then $w=-1,$4 always; if $w=-1,$5, then $w=-1,$6 always; and $w=-1,$7 gives $w=-1,$8 identically. The sign of $w=-1,$9 fixes which side of the divide the model occupies, so a constant-$w\equiv p/\rho$0 DGP-type model cannot cross (Hirano et al., 2010).

An analogous obstruction appears in ordinary single-field dark-energy models. For a canonical quintessence field, the equation of state satisfies $w\equiv p/\rho$1, while for a phantom field with reversed kinetic sign one has $w\equiv p/\rho$2. The quintom literature summarized here further notes that in generic single-field $w\equiv p/\rho$3-essence, stable crossing is obstructed by the usual no-go logic: without extra degrees of freedom, stable phantom crossing is not possible in the usual single-field setup (Goh et al., 15 Sep 2025).

Horndeski theory provides a controlled route around this obstruction, but only under nontrivial conditions. One result emphasized in the Horndeski literature is that stable phantom behaviour is impossible if

$w\equiv p/\rho$4

because the stability and phantom inequalities reduce to contradictory requirements. Stable crossing requires at least one of two ingredients: an $w\equiv p/\rho$5-dependence in $w\equiv p/\rho$6, or a $w\equiv p/\rho$7-dependence in $w\equiv p/\rho$8 (Matsumoto, 2017).

3. Realizations in modified gravity and braneworld cosmology

The prototype braneworld realization is the Phantom Crossing DGP model. Its central modification is to promote the Dvali–Turner exponent to a time-dependent quantity,

$w\equiv p/\rho$9

with $w=-1$0. The modified Friedmann-like equation becomes

$w=-1$1

When $w=-1$2, one has $w=-1$3 and the effective DGP component behaves non-phantom; when $w=-1$4, $w=-1$5 and the effective equation of state reaches $w=-1$6; when $w=-1$7, $w=-1$8 and the effective equation of state becomes phantom-like. The paper defines

$w=-1$9

and, from the combined SNIa+CMB+BAO fit, obtains

$w>-1$0

at $w>-1$1, with the crossing occurring at $w>-1$2. Its best-fit $w>-1$3 is $w>-1$4, smaller than $w>-1$5 for the Dvali–Turner version and $w>-1$6 for the original DGP model (Hirano et al., 2010).

Modified teleparallel and non-metricity frameworks yield analogous effective-fluid realizations. In $w>-1$7 gravity, Model A,

$w>-1$8

crosses the phantom divide twice, whereas Model B,

$w>-1$9

can cross in either direction depending on the signs of $w<-1$0 and $w<-1$1. The best-fit values quoted are $w<-1$2, $w<-1$3, $w<-1$4 for Model A, and $w<-1$5, $w<-1$6, $w<-1$7, $w<-1$8 for Model B (Wu et al., 2010). In $w<-1$9 gravity, the combined exponential+logarithmic model produces a genuine transition from non-phantom to phantom, whereas the exponential and logarithmic models do not generically cross; the paper further argues that future phantom-divide crossing is a generic feature of feasible $w(z)$0 models (Arora et al., 2022).

In $w(z)$1 gravity, reconstruction can generate an explicit non-phantom to phantom transition through the effective equation of state

$w(z)$2

One reconstructed model evolves from $w(z)$3 at early times to $w(z)$4 near a future Big Rip singularity, with the large-curvature asymptotics $w(z)$5. The same work notes that adding an $w(z)$6 term or a non-singular viable $w(z)$7 deformation can make the phantom phase transient and avoid the singularity (Bamba et al., 2010). A separate thermodynamic analysis shows that a model of $w(z)$8 gravity realizing phantom crossing can still satisfy the generalized second law of thermodynamics on the apparent horizon in both the non-phantom and effective phantom phases (0901.1509).

Horndeski and Gauss–Bonnet constructions extend the range of possible behaviours. The Asymptotic Cubic Galileon subclass of luminal Horndeski models generates phantom crossing by a growing braiding term or a decaying kinetic term while remaining minimally coupled to matter; it provides a fit of comparable quality to $w(z)$9CDM, though perturbative observables sharply restrict the viable region (Naidoo et al., 18 Jun 2026). Scalar–Einstein–Gauss–Bonnet gravity and ghost-free $w+1$0 gravity admit both true and apparent inverse phantom crossing around $w+1$1; in the apparent case, dark matter can dilute more slowly than $w+1$2, so the inferred effective dark-energy sector appears to cross even though no component need violate the energy conditions (Nojiri et al., 6 Dec 2025).

4. Effective, intrinsic, and apparent crossing

A major conceptual clarification in recent work is that effective phantom behaviour does not automatically identify the microphysics. In background reconstruction, $w+1$3 does not by itself imply a ghostly scalar field with wrong-sign kinetic term, microscopic vacuum instability, violation of the null energy condition by the fundamental stress-energy tensor, or an unavoidable future Big Rip. The reconstructed quantity is a residual component inferred after imposing FLRW, the standard Friedmann equation, and conserved matter. Even if $w+1$4, the total equation of state

$w+1$5

can still satisfy the null energy condition, and a transient effective phantom phase does not imply the finite-time future singularity associated with eternal phantom domination (Mishra, 26 May 2026).

Interacting dark-sector models make this distinction explicit. With

$w+1$6

the effective equations of state become

$w+1$7

In the DESI DR2 BAO + Planck 2018 CMB + Pantheon+ analysis, the effective dark-energy equation of state shows phantom behaviour at higher redshift and non-phantom behaviour at lower redshift, but the phantom behaviour at higher redshift is well within $w+1$8. The intrinsic equation of state remains non-phantom even without imposing a non-phantom prior, while a nonzero interaction is favored at more than $w+1$9 at $w>-1$00 and changes sign near $w>-1$01 (Guedezounme et al., 24 Jul 2025).

Related mechanisms realize the same observational signature without fundamental phantom dynamics. In the dissipative dark-energy scenario, a canonical quintessence field transfers energy weakly to dark matter through a dissipation term $w>-1$02, altering the dark-matter redshift scaling; an observer assuming non-interacting dark matter and dark energy infers an effective $w>-1$03 that can cross $w>-1$04, even though the intrinsic scalar equation of state satisfies $w>-1$05 throughout (Chanda et al., 3 Jun 2026). In an interacting field-theoretic model with a tachyonic Born–Infeld scalar and Yukawa-coupled fermionic dark matter, the scalar itself remains nonphantom,

$w>-1$06

while the effective equation of state

$w>-1$07

can exhibit a recent double crossing of the phantom divide (Abdalla et al., 19 May 2026).

5. Observational constraints and statistical status

Observational studies of phantom-crossing behaviour have used broadly similar distance datasets but arrived at different levels of significance, depending on the parametrization and physical interpretation. In the 2010 DGP analysis, the combined likelihood used a 288-object Type Ia supernova sample from Kessler et al., the 5-year WMAP CMB shift parameter with

$w>-1$08

and the SDSS luminous red galaxy BAO constraint at

$w>-1$09

The total fit was obtained from

$w>-1$10

This analysis found the Phantom Crossing DGP model more compatible with the observations than the original DGP model or the Dvali–Turner model, although $w>-1$11CDM retained smaller AIC/BIC because it has one fewer free parameter (Hirano et al., 2010).

More recent DESI-era analyses are notably more cautious. A frequentist Monte Carlo study based on a fiducial non-phantom algebraic quintessence model generated 1,000 mock realizations using DESI DR2 BAO, Union3 supernovae, and compressed Planck CMB constraints. The real data favored CPL by approximately

$w>-1$12

but this improvement was reproduced or exceeded in $w>-1$13 of the mocks, while in over $w>-1$14 of the simulations CPL simply fit better than the non-phantom comparator. The conclusion was that evolving dark energy appears robust, but the more specific claim of phantom crossing remains suggestive rather than definitive (Keeley et al., 18 Jun 2025).

A complementary explanation attributes the preference for crossing to an $w>-1$15 “tug-of-war” among datasets. In $w>-1$16CDM fits, the ordering

$w>-1$17

was emphasized, with

$w>-1$18

Phantom-crossing models can shift these inferred matter densities toward mutual alignment, whereas quintessence can alleviate the tensions with SN data but only at the cost of exacerbating the BAO–CMB discrepancy. In that analysis, the preference for crossing was therefore driven primarily by CMB and BAO rather than by supernovae, and it was stronger in CMB+BAO than in CMB+BAO+SN fits (Shlivko et al., 23 Mar 2026).

The same theme appears in analyses linking crossing to the $w>-1$19 tension. In the model where $w>-1$20 is expanded around an extremum, CMB+BAO data yielded

$w>-1$21

at $w>-1$22 confidence and were described as confirming phantom crossing at $w>-1$23 confidence level. For the full combination CMB+lensing+BAO+R19+Pantheon, the quoted constraints were

$w>-1$24

with the transition still indicated at more than $w>-1$25 (Valentino et al., 2020).

6. Physical implications, diagnostics, and open problems

One persistent misconception is that phantom-crossing behaviour necessarily signals a fundamental phantom field. The literature summarized here repeatedly rejects that inference. Modified gravity, braneworld leakage into an extra dimension, interacting dark sectors, dissipative quintessence, and apparent effects from non-standard dark-matter dilution can all generate the observed background signature without introducing a fundamental wrong-sign kinetic term (Mishra, 26 May 2026).

A second theme is that crossing should not be studied through the background expansion alone. In shift-symmetric cubic Horndeski gravity, the same unusual kinetic and braiding structure that permits $w>-1$26 crossing also modifies growth and lensing. That work identifies three broad late-time paths after crossing: the effective equation of state can continue away from $w>-1$27 and even become positive, asymptote to a nontrivial constant with $w>-1$28, or return to de Sitter with $w>-1$29. The paper’s central lesson is that post-crossing behaviour is highly diagnostic of the underlying theory and must be tested with structure growth and lensing, not only with background distances (Linder, 2 Dec 2025).

Related analyses sharpen this point with concrete observables. In the ACG Horndeski subclass, galaxy–ISW cross-correlation and void force profiles substantially reduce the viable region of parameter space, even when the background fit is competitive with $w>-1$30CDM (Naidoo et al., 18 Jun 2026). In $w>-1$31 gravity, oscillating dark-energy eras can produce repeated standard and inverse crossings, so phantom-crossing behaviour need not be a single transition. The same work stresses that $w>-1$32 gravity can realize a transition from a phantom era to a quintessential era without resorting to phantom scalar fields (Nojiri et al., 26 Jun 2025).

The future evolution implied by crossing is likewise model-dependent. Some explicit reconstructions in $w>-1$33 gravity end in a Big Rip singularity (Bamba et al., 2010), whereas thermodynamic analyses and effective-reconstruction studies emphasize that transient phantom behaviour need not imply any catastrophic cosmic future (0901.1509, Mishra, 26 May 2026). This suggests that the most robust content of the term “phantom-crossing behaviour” is not a unique microphysical diagnosis, but a statement about the inferred redshift dependence of the cosmic acceleration sector. Whether that statement ultimately points to modified gravity, dark-sector interaction, or a genuinely dynamical dark-energy fluid remains an open problem tied to increasingly precise BAO, CMB, supernova, growth, and lensing data.