Massive Islands in Gravitational Theories

• Massive islands are entanglement regions where bath couplings induce an effective graviton mass, altering the boundary stress tensor scaling.
• They facilitate tractable Page-curve calculations in higher-dimensional AdS-plus-bath setups by breaking exact energy conservation.
• Recent advances, including massless wedge holography, reveal that a healthy additional entropy term can substitute graviton mass as the balancing mechanism.

“Massive islands” is a term from the entanglement-island literature denoting the observation that, in more than $2+1$ spacetime dimensions, the controlled Page-curve constructions were initially realized in setups where the effective graviton is massive rather than exactly massless (Geng et al., 2020). In the canonical AdS-plus-bath constructions, transparent boundary conditions or equivalent bath couplings spoil exact conservation of the gravitating stress tensor, shift its scaling dimension away from $\Delta=d$, and thereby produce a massive or effectively massive graviton (Antonini et al., 4 Jun 2025). This association was sharpened into a consistency thesis based on gravitational Gauss law and dressing, then challenged by papers arguing that islands and Page curves also arise in massless settings, and more recently refined by massless wedge-holographic constructions in which a healthy additional entropy term replaces graviton mass as the balancing mechanism (Geng et al., 2021, Antonini et al., 4 Jun 2025, Kumar, 30 May 2026).

1. Origin of the term and basic prescription

The immediate background of the term is the island prescription

$$S(R)= \underset{I}{\rm min\,\, ext}\left[\frac{\mathcal{A}(\partial I)}{4G}+S_{\rm eff}(R\cup I)\right],$$

or equivalently the generalized-entropy extremization of a candidate island $I$ for a radiation subsystem $R$ (Antonini et al., 4 Jun 2025). In the original higher-dimensional AdS constructions, the gravitating region was coupled to an external nongravitating bath. The crucial structural point was that energy could flow across the AdS boundary, so the boundary gravitational stress tensor was not conserved. In the AdS representation-theoretic language reviewed in later work, a “massless graviton” is precisely the case in which the boundary gravitational stress tensor is conserved, $\nabla_a T^{ab}_{\rm grav}=0$, and therefore has scaling dimension $\Delta=d$; when the bath coupling makes $\nabla_a T^{ab}_{\rm grav}\neq 0$, the stress tensor acquires anomalous dimension and the graviton is called “massive” (Antonini et al., 4 Jun 2025).

This terminology was made explicit in “Massive Islands,” which stated that all reliable calculations of the Page curve in more than $2+1$ spacetime dimensions had been performed in systems with massive gravitons (Geng et al., 2020). That paper emphasized a tension that would dominate subsequent debate: the graviton mass did not seem to enter the standard island calculations explicitly, yet varying the graviton mass within the Randall–Sundrum / AdS-BCFT family produced an analytically tractable model in which the island contribution disappeared in the zero-mass limit (Geng et al., 2020).

A standard misconception in the literature is that “massive islands” means an explicit Fierz–Pauli deformation was inserted by hand. The papers under discussion typically do not make that claim. Rather, the operative mechanism is a bath-induced or boundary-condition-induced mass, often described as a Higgs-like or Stückelberg-completed effect in the effective gravitational description (Geng, 2023).

2. Gauss law, dressing, and the necessity thesis

The strongest early necessity claims were formulated in terms of long-range gravity and gravitational dressing. “Inconsistency of Islands in Theories with Long-Range Gravity” argued that islands are potentially inconsistent as entanglement wedges in higher-dimensional massless gravity because the gravitational Gauss law makes excitations inside the island detectable from outside the island (Geng et al., 2021). In linearized massless gravity on a spatial slice, the constraint

$$-\partial_j\partial_j h_{ii}+\partial_j\partial_i h_{ij} = 16\pi G\,\rho$$

integrates to a boundary Gauss law,

$\Delta=d$0

The Hamiltonian is then determined by asymptotic data, and a gauge-invariant bulk operator must be dressed to asymptotia (Geng et al., 2021).

The same paper stressed that this is precisely what islands lack by definition: an island is an entanglement wedge in the gravitating spacetime that does not extend to the asymptotic boundary of the gravitating spacetime (Geng et al., 2021). In ordinary AdS/CFT wedges this is harmless, because the wedge reaches asymptotia and the dressing can remain inside the wedge. For a compact island, however, any gravitationally gauge-invariant dressing must involve the complement. In that sense, the problem is not merely geometric but algebraic.

“Graviton Mass and Entanglement Islands in Low Spacetime Dimensions” reformulated the same issue in terms of the standard gravitational Gauss law rather than the existence of a propagating helicity-2 mode (Geng, 2023). In JT gravity, the matter Hamiltonian becomes a pure boundary term,

$\Delta=d$1

which is the AdS$\Delta=d$2 version of Gauss law. The paper then showed that bath coupling induces a Stückelberg-completed effective mass term. In unitary gauge this takes the form

$\Delta=d$3

and the Hamiltonian acquires a genuine bulk term, so bulk energy is no longer completely visible from asymptotic data (Geng, 2023). The necessity thesis was therefore sharpened as follows: islands in the known constructions are compatible with subregion physics because the exact long-range Gauss-law relation has failed.

A related line of argument went further and tied operator localization itself to massive gravity. The leading-order dressed scalar operator was written as

$\Delta=d$4

with the Stückelberg field $\Delta=d$5 providing a dressing medium that does not refer to asymptotic infinity (Geng, 2023). This was presented as the mechanism by which information localization differs between massive and massless gravity.

3. Canonical constructions and quantitative control

The higher-dimensional examples most often cited in this connection are doubly holographic or braneworld constructions. In the subcritical Randall–Sundrum / AdS-BCFT family, the brane angle $\Delta=d$6 controls both the induced AdS radius and the graviton sector: $\Delta=d$7 Near $\Delta=d$8, the effective Newton constant scales as

$\Delta=d$9

while the almost-zero-mode graviton mass scales as

$$S(R)= \underset{I}{\rm min\,\, ext}\left[\frac{\mathcal{A}(\partial I)}{4G}+S_{\rm eff}(R\cup I)\right],$$0

(Geng et al., 2020).

In “Massive Islands,” the $$S(R)= \underset{I}{\rm min\,\, ext}\left[\frac{\mathcal{A}(\partial I)}{4G}+S_{\rm eff}(R\cup I)\right],$$1 limit produced an analytically tractable higher-dimensional island model, while the $$S(R)= \underset{I}{\rm min\,\, ext}\left[\frac{\mathcal{A}(\partial I)}{4G}+S_{\rm eff}(R\cup I)\right],$$2 limit showed that the nontrivial island surface becomes parametrically too expensive: the crossing time scales like

$$S(R)= \underset{I}{\rm min\,\, ext}\left[\frac{\mathcal{A}(\partial I)}{4G}+S_{\rm eff}(R\cup I)\right],$$3

so the island never dominates in the massless limit of that model family (Geng et al., 2020).

Setup	Graviton status	Island consequence
Subcritical RS / AdS-BCFT	Bath-induced massive graviton	Controlled higher-dimensional Page-curve constructions (Geng et al., 2020)
Karch–Randall braneworld	No normalizable zero mode on the brane	Island operators admit bath-directed dressing (Geng, 12 Feb 2025)
Type IIB AdS$$S(R)= \underset{I}{\rm min\,\, ext}\left[\frac{\mathcal{A}(\partial I)}{4G}+S_{\rm eff}(R\cup I)\right],$$4/bath embedding	Lowest graviton small but nonzero	Islands exist only in a restricted mass window (Demulder et al., 2022)

The string-theoretic realization in type IIB made the correlation quantitative. In the “bag + thin throat” regime, the lightest graviton mass is

$$S(R)= \underset{I}{\rm min\,\, ext}\left[\frac{\mathcal{A}(\partial I)}{4G}+S_{\rm eff}(R\cup I)\right],$$5

so making the graviton lighter requires increasing $$S(R)= \underset{I}{\rm min\,\, ext}\left[\frac{\mathcal{A}(\partial I)}{4G}+S_{\rm eff}(R\cup I)\right],$$6 and/or the dilaton variation $$S(R)= \underset{I}{\rm min\,\, ext}\left[\frac{\mathcal{A}(\partial I)}{4G}+S_{\rm eff}(R\cup I)\right],$$7 (Demulder et al., 2022). The same paper found numerically that zero-temperature islands disappear above a critical $$S(R)= \underset{I}{\rm min\,\, ext}\left[\frac{\mathcal{A}(\partial I)}{4G}+S_{\rm eff}(R\cup I)\right],$$8, while finite-temperature islands persist but change behavior. Its central conclusion was that the graviton cannot be made arbitrarily light while retaining islands in that explicit string-theoretic class (Demulder et al., 2022).

4. Massive gravity as the mechanism of encoding

Later work shifted from the necessity question to the positive mechanism of encoding. “The Mechanism behind the Information Encoding for Islands” argued that the graviton mass is not merely a consistency condition but the actual local mechanism by which a disconnected bath encodes operators in the gravitational region (Geng, 12 Feb 2025). In the weakly coupled AdS-plus-bath model, the total Hamiltonian

$$S(R)= \underset{I}{\rm min\,\, ext}\left[\frac{\mathcal{A}(\partial I)}{4G}+S_{\rm eff}(R\cup I)\right],$$9

implies

$I$0

The resulting Higgs phase of gravity yields the gauge-invariant combination

$I$1

and the effective mass term

$I$2

(Geng, 12 Feb 2025).

The same paper then proposed the physical island operator

$I$3

and argued that the Goldstone/Stückelberg field $I$4 is dual to a mixed AdS-bath operator. This is why the dressed operator evolves under the bath Hamiltonian but not under the AdS boundary Hamiltonian. In Karch–Randall, the dressing becomes a literal gravitational Wilson line through the extra dimension, so the information-transfer mechanism is geometrized (Geng, 12 Feb 2025).

“Making the Case for Massive Islands” generalized this line of thought and emphasized that the induced graviton mass should not be identified with the dissipation timescale (Geng, 26 Sep 2025). In JT gravity with transparent conformal matter, the paper derived

$I$5

for the induced graviton mass, while the boundary-energy leakage obeys

$I$6

The distinction was presented as conceptual as well as quantitative: $I$7 is an equilibrium property of the effective gravitational theory, whereas $I$8 is a nonequilibrium relaxation rate (Geng, 26 Sep 2025).

5. Dispute over massless islands

The claim that islands are necessarily “massive” was explicitly challenged. “An apologia for islands” argued that the phrase is misleading because islands and Page-curve behavior also occur in setups with massless gravitons and no external nongravitating reservoir (Antonini et al., 4 Jun 2025). The paper gave three classes of examples: boundary subregion entanglement wedges in AdS/CFT, radiation algebras at future null infinity in asymptotically flat spacetimes, and radiation inside a semiclassical but gravitating spacetime. Its broader claim was that islands are a generic consequence of the gravitational replica trick and connected replica wormholes whenever a semiclassical radiation entropy would otherwise exceed the black hole’s capacity to purify it (Antonini et al., 4 Jun 2025).

The same paper also tried to dissolve the operator-localization objection by arguing that compactly supported gauge-invariant operators exist to all orders in perturbation theory whenever no isometries of the background spacetime exist. Its criterion was that the linearized gravitational constraint operator $I$9 be surjective, or equivalently that its adjoint be injective. On that basis, it treated the obstruction as one tied to exact symmetries rather than to the asymptotic graviton notion of masslessness (Antonini et al., 4 Jun 2025).

A direct response came from “Seeing Page Curves and Islands with Blinders On,” which argued that in ordinary closed gravity the asymptotic algebra of observables, including the Hamiltonian, is complete (Geng et al., 6 Feb 2026). On that view, the bulk Hilbert space does not factorize along the radial direction, so the full black-hole interior, not just an island, is reconstructible from exterior data. Page curves and islands then arise only after one deliberately restricts the observable algebra by “putting on blinders,” for example by removing the Hamiltonian from the exterior algebra or restricting access to part of the asymptotic region (Geng et al., 6 Feb 2026). The disagreement is therefore not only about examples but about what counts as the relevant algebra for entropy and entanglement wedge reconstruction.

This controversy has produced two competing uses of “massive islands.” In one, the phrase names a necessity condition rooted in Gauss law and dressing; in the other, it names only a historically convenient subclass of models and should not be elevated into a universal principle (Geng, 2023, Antonini et al., 4 Jun 2025).

6. Massless wedge holography and the current synthesis

The most explicit recent refinement of the debate appears in “Massless Islands in Wedge Holography” (Kumar, 30 May 2026). Wedge holography provides a sharp arena because the minimal Neumann-on-both-branes branch supports a normalizable massless graviton. In that minimal model, the candidate island surface $R$0 has area functional

$R$1

and outside the horizon this obeys

$R$2

Free endpoint variation further imposes the pure orthogonality condition

$R$3

The result is that the naive island saddle collapses to the horizon (Kumar, 30 May 2026).

Earlier wedge-holographic work had shown that negative Dvali–Gabadadze–Porrati terms can change the endpoint condition and restore nontrivial islands, but the required branch contains a massive ghost (Kumar, 30 May 2026). The new proposal keeps the wedge gravitational action free of DGP terms and instead adds a unitary defect CFT localized at the codimension-two corner,

$R$4

The generalized entropy becomes

$R$5

and the endpoint equation becomes the quantum extremality condition

$R$6

For a two-dimensional defect CFT,

$R$7

so with $R$8 and $R$9 one has

$\nabla_a T^{ab}_{\rm grav}=0$0

while

$\nabla_a T^{ab}_{\rm grav}=0$1

The paper’s central conclusion is that the obstruction in minimal wedge holography is not masslessness itself, but the absence of an additional healthy entropy term capable of balancing the horizon-minimizing area variation (Kumar, 30 May 2026).

The local endpoint model then yields an isolated stable non-horizon saddle when

$\nabla_a T^{ab}_{\rm grav}=0$2

and this time-independent island eventually dominates over the Hartman–Maldacena surface after

$\nabla_a T^{ab}_{\rm grav}=0$3

A plausible implication is that the field has moved from a simple necessity slogan—“islands are massive islands”—to a more discriminating statement: in long-range, massless, ghost-free gravity, island formation depends on whether the full generalized entropy contains an additional healthy contribution whose endpoint variation can compete with the geometric pull toward the horizon (Kumar, 30 May 2026).

A separate but related generalization appears in cosmology, where “islands in cosmology” refers to large cosmological quantum extremal regions controlled by regulated area-bound violation and quantum-normal conditions rather than by bath-induced graviton mass (Hartman et al., 2020). That usage broadens the island framework, but it does not alter the specific technical meaning of “massive islands” in the black-hole and braneworld literature.