Phantom Scalar: Theory & Applications

• Phantom scalar is defined by a reversed kinetic term that violates the null energy condition, enabling an equation of state with w < -1.
• Cosmological models with phantom scalars can yield super-accelerating expansion and alleviate Hubble tension through interacting dark-sector dynamics.
• In gravitational settings, phantom scalars support exotic geometries like traversable wormholes and modified black holes, though they introduce potential ghost instabilities.

A phantom scalar is a scalar field whose kinetic term carries the opposite sign from the canonical one, so that its stress tensor can violate the null energy condition and its homogeneous cosmological equation of state can satisfy $w<-1$. In contemporary literature, the term spans several distinct but connected uses: late-time dark-energy models, effective ghost fluids produced by vacuum decay, exact wormhole and black-hole spacetimes, AdS solutions, braneworld constructions, and dynamical-systems analyses of interacting dark sectors. The common structural feature is the wrong-sign kinetic response; the main conceptual fault line is that a classical phantom description is straightforward, whereas a quantum interpretation generically introduces negative-energy ghost excitations and severe stability constraints (Cline et al., 2023, Martinez et al., 2020).

1. Definition, sign conventions, and stress-energy structure

A convention-independent way to encode the phantom sign is

$$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$

so that $\epsilon=-1$ denotes a phantom field and $\epsilon=+1$ a canonical one (Karakasis et al., 2021). In this form, the stress tensor is

$$T^{\phi}_{\mu\nu} = -\partial_\mu\phi\,\partial_\nu\phi + \frac{1}{2}\,g_{\mu\nu}\,\partial^\alpha\phi\,\partial_\alpha\phi - g_{\mu\nu}\,V(\phi),$$

and for any null vector $k^\mu$,

$$T_{\mu\nu}k^\mu k^\nu = -(k^\mu\partial_\mu\phi)^2 \le 0,$$

which directly exhibits null-energy-condition violation (Lora-Clavijo et al., 2012).

The literature does not use a single sign convention. In full numerical relativity, one finds

$$\mathcal{L}=R+\frac{1}{2}g^{\mu\nu}\partial_\mu\phi\partial_\nu\phi - V(\phi),$$

with the kinetic term written with the same sign as the Ricci scalar (Lora-Clavijo et al., 2012). In rotating wormhole constructions with a complex field, the phantom sector is written as

$$\mathcal{L}_{\rm ph} = \frac{1}{2} g^{\mu\nu}\bigl(\partial_\mu\Phi^*\,\partial_\nu\Phi+\partial_\nu\Phi^*\,\partial_\mu\Phi\bigr)-V(|\Phi|^2),$$

again with the plus sign marking the phantom character (Chew et al., 2019). In cosmology, authors often emphasize the sign flip through the resulting fluid variables rather than through a universal Lagrangian notation. For a homogeneous field in flat FRW, the phantom density and pressure are commonly written as

$$\rho_\phi=-\frac{1}{2}\dot\phi^2+V(\phi),\qquad p_\phi=-\frac{1}{2}\dot\phi^2-V(\phi),$$

so that

$$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$0

and $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$1 arises when $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$2 dominates the negative kinetic contribution (Cline et al., 2023).

A recurring misconception is that $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$3 is the universal local signature of a phantom scalar. That is only true in homogeneous isotropic cosmology. In strong-field, inhomogeneous settings the local ratio $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$4 need not remain below $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$5, even though NEC violation persists (Lora-Clavijo et al., 2012).

2. Cosmological realizations and dark-sector dynamics

In homogeneous cosmology, phantom scalars are used as dynamical dark-energy components whose late-time behavior differs from $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$6CDM by allowing super-acceleration, $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$7, whenever $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$8. One recent implementation adopts hyperbolic polar variables,

$$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$9

with

$\epsilon=-1$0

and parametrizes the potential sector by $\epsilon=-1$1. Using Planck 2018, SPT-3G+WMAP9, and ACTPol DR-4+WMAP9 baselines together with BAO, Pantheon, and cosmic clocks, this tracker phantom model yields $\epsilon=-1$2 for SPT-3G+WMAP9+Late and $\epsilon=-1$3 for ACTPol DR-4+WMAP9+Late, with the corresponding Hubble-tension metric reduced below $\epsilon=-1$4 in both cases (Nájera et al., 2024).

Interacting dark-sector models change the phase-space structure more radically. For a phantom field coupled to pressureless dark matter by

$\epsilon=-1$5

with $\epsilon=-1$6 or $\epsilon=-1$7, compactified variables

$\epsilon=-1$8

put the dynamics on a bounded phase space satisfying $\epsilon=-1$9. In that framework, Big-Rip-type configurations occur on $\epsilon=+1$0 boundaries, but the late-time attractors can instead be scalar-dominated, interacting scaling, or de Sitter points depending on the coupling model (Leon et al., 15 Jan 2025). A related arbitrary-potential analysis reaches a sharper conclusion: uncoupled phantom cosmology does not admit accelerated scaling solutions, whereas coupled models with $\epsilon=+1$1 can produce accelerating scaling attractors, though the resulting accelerated stage is not preceded by a long enough matter-dominated era (Halder et al., 20 Oct 2025).

A distinct route to crossing the phantom divide is the quintom construction, where a canonical scalar $\epsilon=+1$2 and a phantom scalar $\epsilon=+1$3 coexist: $\epsilon=+1$4 Then

$\epsilon=+1$5

and $\epsilon=+1$6 can be crossed smoothly because no single field crosses it alone. Gaussian hill-top and hyperbolic-tangent plateau potentials were shown to realize phantom-to-quintessence crossing consistent, at the background level, with DESI DR2 BAO, CMB, and supernova trends, although the viable parameter region is narrow and fine-tuning is substantial in the hyperbolic-tangent case (Goh et al., 15 Sep 2025).

3. Quantum interpretation, ghost pathology, and effective-field-theory bounds

The sharpest distinction in the subject is between classical phantom dynamics and quantum phantom quanta. If the wrong-sign field is quantized in the usual way, its excitations carry negative energy, so gravity-mediated vacuum decay becomes unavoidable. In the explicit analysis of a massless phantom scalar with vanishing potential and no classical condensate, vacuum decay into ghosts plus photons or gravitons diverges unless one imposes a three-momentum cutoff $\epsilon=+1$7, which is necessarily Lorentz-violating because the cutoff is placed on $\epsilon=+1$8 rather than on a Lorentz-invariant combination (Cline et al., 2023).

Using the diffuse $\epsilon=+1$9-ray background measured by COMPTEL, the updated bound is

$$T^{\phi}_{\mu\nu} = -\partial_\mu\phi\,\partial_\nu\phi + \frac{1}{2}\,g_{\mu\nu}\,\partial^\alpha\phi\,\partial_\alpha\phi - g_{\mu\nu}\,V(\phi),$$0

and the phenomenological summary is $$T^{\phi}_{\mu\nu} = -\partial_\mu\phi\,\partial_\nu\phi + \frac{1}{2}\,g_{\mu\nu}\,\partial^\alpha\phi\,\partial_\alpha\phi - g_{\mu\nu}\,V(\phi),$$1 (Cline et al., 2023). A weaker bound, $$T^{\phi}_{\mu\nu} = -\partial_\mu\phi\,\partial_\nu\phi + \frac{1}{2}\,g_{\mu\nu}\,\partial^\alpha\phi\,\partial_\alpha\phi - g_{\mu\nu}\,V(\phi),$$2, follows from $$T^{\phi}_{\mu\nu} = -\partial_\mu\phi\,\partial_\nu\phi + \frac{1}{2}\,g_{\mu\nu}\,\partial^\alpha\phi\,\partial_\alpha\phi - g_{\mu\nu}\,V(\phi),$$3-induced CMB spectral distortions, but COMPTEL dominates. Coupling the ghost sector to a light hidden fermion through higher-dimension operators,

$$T^{\phi}_{\mu\nu} = -\partial_\mu\phi\,\partial_\nu\phi + \frac{1}{2}\,g_{\mu\nu}\,\partial^\alpha\phi\,\partial_\alpha\phi - g_{\mu\nu}\,V(\phi),$$4

turns vacuum decay into a late-time source for a combined ghost-plus-hidden-sector “phantom fluid” obeying

$$T^{\phi}_{\mu\nu} = -\partial_\mu\phi\,\partial_\nu\phi + \frac{1}{2}\,g_{\mu\nu}\,\partial^\alpha\phi\,\partial_\alpha\phi - g_{\mu\nu}\,V(\phi),$$5

Its effective equation of state approaches $$T^{\phi}_{\mu\nu} = -\partial_\mu\phi\,\partial_\nu\phi + \frac{1}{2}\,g_{\mu\nu}\,\partial^\alpha\phi\,\partial_\alpha\phi - g_{\mu\nu}\,V(\phi),$$6 at early times and $$T^{\phi}_{\mu\nu} = -\partial_\mu\phi\,\partial_\nu\phi + \frac{1}{2}\,g_{\mu\nu}\,\partial^\alpha\phi\,\partial_\alpha\phi - g_{\mu\nu}\,V(\phi),$$7 at late times, and a global cosmological fit finds a best-fit phantom-fluid model with $$T^{\phi}_{\mu\nu} = -\partial_\mu\phi\,\partial_\nu\phi + \frac{1}{2}\,g_{\mu\nu}\,\partial^\alpha\phi\,\partial_\alpha\phi - g_{\mu\nu}\,V(\phi),$$8, $$T^{\phi}_{\mu\nu} = -\partial_\mu\phi\,\partial_\nu\phi + \frac{1}{2}\,g_{\mu\nu}\,\partial^\alpha\phi\,\partial_\alpha\phi - g_{\mu\nu}\,V(\phi),$$9 km/s/Mpc, and $k^\mu$0, corresponding to mild preference over $k^\mu$1CDM and modest amelioration of the $k^\mu$2 and $k^\mu$3 tensions (Cline et al., 2023).

At the same time, a broad no-go result has been proved for single-field effective cosmologies with Hubble-modulated kinetic response. Writing the background sector as $k^\mu$4 and $k^\mu$5, ghost freedom requires

$k^\mu$6

while gradient stability requires

$k^\mu$7

Then

$k^\mu$8

so $k^\mu$9 cannot be reached continuously without violating ghost freedom or gradient stability. In this class, $$T_{\mu\nu}k^\mu k^\nu = -(k^\mu\partial_\mu\phi)^2 \le 0,$$0 is a stability-protected boundary and a de Sitter-like attractor rather than a crossing surface (Sahoo, 25 Jan 2026).

A separate phenomenological line of work proposes a unified Lagrangian

$$T_{\mu\nu}k^\mu k^\nu = -(k^\mu\partial_\mu\phi)^2 \le 0,$$1

with $$T_{\mu\nu}k^\mu k^\nu = -(k^\mu\partial_\mu\phi)^2 \le 0,$$2 selecting the phantom branch, but this construction is presented at the level of classical background dynamics and numerical stability regions rather than as a ghost-free quantum completion (Joshi, 2023).

4. Wormholes, black holes, geons, and exact phantom geometries

Because the wrong-sign kinetic term violates the NEC, phantom scalars naturally support geometries unavailable to canonical matter. The most systematic static classification for a massless phantom scalar in $$T_{\mu\nu}k^\mu k^\nu = -(k^\mu\partial_\mu\phi)^2 \le 0,$$3 dimensions shows that the generalized Fisher and Ellis–Gibbons families generically develop parallelly propagated curvature singularities, even when scalar invariants remain finite, whereas the Ellis–Bronnikov family is the only regular two-ended asymptotically flat traversable wormhole in that class (Martinez et al., 2020).

Geometry	Phantom role	Representative result
Rotating wormhole	Complex phantom scalar induces rotation	Regular asymptotically flat solutions with $$T_{\mu\nu}k^\mu k^\nu = -(k^\mu\partial_\mu\phi)^2 \le 0,$$4; no ergoregions in the studied domain (Chew et al., 2019)
Static $$T_{\mu\nu}k^\mu k^\nu = -(k^\mu\partial_\mu\phi)^2 \le 0,$$5 wormhole	Phantom field sources exact Morris–Thorne-type throat	$$T_{\mu\nu}k^\mu k^\nu = -(k^\mu\partial_\mu\phi)^2 \le 0,$$6, $$T_{\mu\nu}k^\mu k^\nu = -(k^\mu\partial_\mu\phi)^2 \le 0,$$7, $$T_{\mu\nu}k^\mu k^\nu = -(k^\mu\partial_\mu\phi)^2 \le 0,$$8, with $$T_{\mu\nu}k^\mu k^\nu = -(k^\mu\partial_\mu\phi)^2 \le 0,$$9 and $$\mathcal{L}=R+\frac{1}{2}g^{\mu\nu}\partial_\mu\phi\partial_\nu\phi - V(\phi),$$0 (Karakasis et al., 2021)
Topological geon	$$\mathcal{L}=R+\frac{1}{2}g^{\mu\nu}\partial_\mu\phi\partial_\nu\phi - V(\phi),$$1 quotient of a phantom-supported wormhole	Three classes distinguished by the behavior of $$\mathcal{L}=R+\frac{1}{2}g^{\mu\nu}\partial_\mu\phi\partial_\nu\phi - V(\phi),$$2 near the throat (Kratovitch et al., 2018)
Phantom Curzon–Chazy counterpart	Phantom scalar replaces the vacuum Curzon–Chazy source	Wormhole throat at $$\mathcal{L}=R+\frac{1}{2}g^{\mu\nu}\partial_\mu\phi\partial_\nu\phi - V(\phi),$$3; interior ends on a non-scalar singularity and behaves as a one-directional time machine (Polcar et al., 2021)
Planar AdS black hole	Phantom scalars linear in transverse coordinates modify $$\mathcal{L}=R+\frac{1}{2}g^{\mu\nu}\partial_\mu\phi\partial_\nu\phi - V(\phi),$$4	Minimal temperature and Hawking–Page transition despite Ricci-flat horizon (Zhang et al., 2017)

The rotating wormhole sector supported by a complex phantom scalar uses a Mexican-hat potential,

$$\mathcal{L}=R+\frac{1}{2}g^{\mu\nu}\partial_\mu\phi\partial_\nu\phi - V(\phi),$$5

and the ansatz

$$\mathcal{L}=R+\frac{1}{2}g^{\mu\nu}\partial_\mu\phi\partial_\nu\phi - V(\phi),$$6

A distinctive result is the exact relation $$\mathcal{L}=R+\frac{1}{2}g^{\mu\nu}\partial_\mu\phi\partial_\nu\phi - V(\phi),$$7, together with a static but nonspherical subset at $$\mathcal{L}=R+\frac{1}{2}g^{\mu\nu}\partial_\mu\phi\partial_\nu\phi - V(\phi),$$8 and $$\mathcal{L}=R+\frac{1}{2}g^{\mu\nu}\partial_\mu\phi\partial_\nu\phi - V(\phi),$$9 (Chew et al., 2019).

In metric $$\mathcal{L}_{\rm ph} = \frac{1}{2} g^{\mu\nu}\bigl(\partial_\mu\Phi^*\,\partial_\nu\Phi+\partial_\nu\Phi^*\,\partial_\mu\Phi\bigr)-V(|\Phi|^2),$$0 gravity, the phantom profile $$\mathcal{L}_{\rm ph} = \frac{1}{2} g^{\mu\nu}\bigl(\partial_\mu\Phi^*\,\partial_\nu\Phi+\partial_\nu\Phi^*\,\partial_\mu\Phi\bigr)-V(|\Phi|^2),$$1 yields an exact wormhole with

$$\mathcal{L}_{\rm ph} = \frac{1}{2} g^{\mu\nu}\bigl(\partial_\mu\Phi^*\,\partial_\nu\Phi+\partial_\nu\Phi^*\,\partial_\mu\Phi\bigr)-V(|\Phi|^2),$$2

throat radius $$\mathcal{L}_{\rm ph} = \frac{1}{2} g^{\mu\nu}\bigl(\partial_\mu\Phi^*\,\partial_\nu\Phi+\partial_\nu\Phi^*\,\partial_\mu\Phi\bigr)-V(|\Phi|^2),$$3, and total NEC violation $$\mathcal{L}_{\rm ph} = \frac{1}{2} g^{\mu\nu}\bigl(\partial_\mu\Phi^*\,\partial_\nu\Phi+\partial_\nu\Phi^*\,\partial_\mu\Phi\bigr)-V(|\Phi|^2),$$4. Notably, the modified-gravity sector remains ghost-free and Dolgov–Kawasaki stable because $$\mathcal{L}_{\rm ph} = \frac{1}{2} g^{\mu\nu}\bigl(\partial_\mu\Phi^*\,\partial_\nu\Phi+\partial_\nu\Phi^*\,\partial_\mu\Phi\bigr)-V(|\Phi|^2),$$5 and $$\mathcal{L}_{\rm ph} = \frac{1}{2} g^{\mu\nu}\bigl(\partial_\mu\Phi^*\,\partial_\nu\Phi+\partial_\nu\Phi^*\,\partial_\mu\Phi\bigr)-V(|\Phi|^2),$$6 (Karakasis et al., 2021).

The phantom analogue of Curzon–Chazy is conformastatic,

$$\mathcal{L}_{\rm ph} = \frac{1}{2} g^{\mu\nu}\bigl(\partial_\mu\Phi^*\,\partial_\nu\Phi+\partial_\nu\Phi^*\,\partial_\mu\Phi\bigr)-V(|\Phi|^2),$$7

with areal radius $$\mathcal{L}_{\rm ph} = \frac{1}{2} g^{\mu\nu}\bigl(\partial_\mu\Phi^*\,\partial_\nu\Phi+\partial_\nu\Phi^*\,\partial_\mu\Phi\bigr)-V(|\Phi|^2),$$8. It has a throat at $$\mathcal{L}_{\rm ph} = \frac{1}{2} g^{\mu\nu}\bigl(\partial_\mu\Phi^*\,\partial_\nu\Phi+\partial_\nu\Phi^*\,\partial_\mu\Phi\bigr)-V(|\Phi|^2),$$9, finite curvature scalars, and a p.p. curvature singularity at $$\rho_\phi=-\frac{1}{2}\dot\phi^2+V(\phi),\qquad p_\phi=-\frac{1}{2}\dot\phi^2-V(\phi),$$0; ADM, Komar, and Brown–York masses all equal $$\rho_\phi=-\frac{1}{2}\dot\phi^2+V(\phi),\qquad p_\phi=-\frac{1}{2}\dot\phi^2-V(\phi),$$1, while the Misner–Sharp energy becomes negative and diverges as $$\rho_\phi=-\frac{1}{2}\dot\phi^2+V(\phi),\qquad p_\phi=-\frac{1}{2}\dot\phi^2-V(\phi),$$2 (Polcar et al., 2021).

A useful corrective to overly broad wormhole claims is that phantom support alone does not guarantee regularity. The broad static classification shows that only Ellis–Bronnikov remains free of both scalar curvature and p.p. singularities throughout the extended manifold (Martinez et al., 2020).

5. Accretion, perturbations, and observational signatures near compact objects

Once a phantom scalar is dynamical rather than merely a background source, its negative-energy flux alters standard black-hole intuition. Fully nonlinear numerical-relativity simulations in spherical symmetry show that accretion of a phantom scalar pulse shrinks the apparent horizon for every potential studied. For positive ADM-mass initial data the apparent-horizon mass can decrease by up to approximately $$\rho_\phi=-\frac{1}{2}\dot\phi^2+V(\phi),\qquad p_\phi=-\frac{1}{2}\dot\phi^2-V(\phi),$$3, while for negative ADM-mass initial data the reduction can reach approximately $$\rho_\phi=-\frac{1}{2}\dot\phi^2+V(\phi),\qquad p_\phi=-\frac{1}{2}\dot\phi^2-V(\phi),$$4 (Lora-Clavijo et al., 2012).

The local equation of state near the hole behaves differently from the cosmological one. Defining

$$\rho_\phi=-\frac{1}{2}\dot\phi^2+V(\phi),\qquad p_\phi=-\frac{1}{2}\dot\phi^2-V(\phi),$$5

the simulations find $$\rho_\phi=-\frac{1}{2}\dot\phi^2+V(\phi),\qquad p_\phi=-\frac{1}{2}\dot\phi^2-V(\phi),$$6, typically ranging between $$\rho_\phi=-\frac{1}{2}\dot\phi^2+V(\phi),\qquad p_\phi=-\frac{1}{2}\dot\phi^2-V(\phi),$$7 and $$\rho_\phi=-\frac{1}{2}\dot\phi^2+V(\phi),\qquad p_\phi=-\frac{1}{2}\dot\phi^2-V(\phi),$$8, even though the same phantom sign violates NEC and WEC and drives horizon shrinkage (Lora-Clavijo et al., 2012). This directly shows that cosmological $$\rho_\phi=-\frac{1}{2}\dot\phi^2+V(\phi),\qquad p_\phi=-\frac{1}{2}\dot\phi^2-V(\phi),$$9 does not translate into a universal local equation of state in strong gravity.

Accretion is also not always efficient. For a massless phantom wave packet sent toward a black hole, thick low-wavenumber packets leave residual scalar material outside the horizon: for $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$00, about $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$01 remains outside for $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$02, and about $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$03 remains outside for $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$04, whereas thin packets with $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$05 are accreted almost completely (González et al., 2016). This suggests that phantom accretion can produce a residual exterior halo rather than total absorption.

Perturbatively, massive static Ellis–Bronnikov wormholes supported by a massless phantom scalar have scalar and axial quasinormal spectra that approach the Schwarzschild values as $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$06 for fixed multipole number $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$07, while retaining a single unstable radial mode with

$$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$08

for $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$09 (Blázquez-Salcedo et al., 2018). In the eikonal limit, the wormhole frequencies differ from Schwarzschild by coefficients $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$10 instead of $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$11, which makes the ringdown nearly black-hole-like at large mass but not identical (Blázquez-Salcedo et al., 2018).

Strong-deflection lensing by black holes in Einstein–(anti-)Maxwell–(anti-)dilaton theory also depends sensitively on whether the scalar and electromagnetic sectors are canonical or phantom. Phantom electromagnetic fields tend to move the photon sphere outward and increase the critical impact parameter, while canonical electromagnetic fields draw the images inward. The resulting relativistic-image observables $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$12, $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$13, and $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$14 therefore distinguish canonical from phantom sectors at the level of image position, separation, and brightness (Gyulchev et al., 2012).

6. Conceptual tensions, misconceptions, and unresolved issues

The subject is unified by a sign flip but fragmented by interpretation. First, “phantom scalar” is not a single model class. It may denote a homogeneous dark-energy field, a real or complex wormhole source, a dilaton-like field in exact black-hole solutions, a braneworld support field, or an effective ghost fluid created by vacuum decay. The mathematical consequences depend strongly on the setting, the potential, and the convention for the action.

Second, the classical and quantum statements are not interchangeable. Classical NEC violation can support wormholes, shrink black-hole horizons, or yield $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$15, but quantized ghost excitations destabilize the vacuum unless the theory is treated as a low-cutoff EFT, typically with Lorentz violation in the ghost sector (Cline et al., 2023). A plausible implication is that many phenomenological phantom models should be interpreted as effective descriptions rather than fundamental quantum field theories.

Third, crossing the phantom divide is model-dependent. Single-field ghost-free Hubble-modulated effective theories admit approach to $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$16 but not continuous crossing (Sahoo, 25 Jan 2026), while two-field quintom systems can cross it smoothly because the total dark-energy fluid crosses even though each constituent remains on its own side of the divide (Goh et al., 15 Sep 2025).

Fourth, classical stability is highly nonuniform across applications. In a five-dimensional thick-brane model supported by two phantom scalar fields, the background can be linearly stable for the chosen parameters, with perturbation frequency $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$17, yet the warp factor grows exponentially and the effective four-dimensional Planck-mass integral diverges, so gravity is not localized in the usual sense (0804.0151). In cosmological doublet models, even an essentially weak phantom field significantly changes the phase-space structure, converting the origin into a saddle and leaving only two attractive fixed points out of nine singular points (Ignat'ev, 2017).

Finally, future singularities remain contingent rather than universal. Classical phantom cosmology is often associated with a Big Rip, but interacting dark-sector models can shift the late-time behavior to accelerating scaling or de Sitter attractors (Leon et al., 15 Jan 2025), and the quantum phantom-fluid EFT with a finite cutoff avoids continuation to arbitrarily high-energy phantom domination because the effective description itself breaks down above $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$18 (Cline et al., 2023).

Taken together, these results establish phantom scalars as a technically versatile but conceptually constrained sector: they are exceptionally effective at generating NEC violation and $$\mathcal{L}_\phi = -\frac{\epsilon}{2}\,g^{\mu\nu}\partial_\mu\phi\,\partial_\nu\phi - V(\phi),\qquad \epsilon=-1,$$19 behavior, yet every successful use is shadowed by questions of ghost freedom, gradient stability, UV completion, or global regularity.