AdS Causality, Trapping, and Stability Interfaces

AdS global structure creates repeated causal contact between the bulk and its timelike boundary. Reflecting conditions can return radiation indefinitely; confining geometry can produce resonances; black-hole exteriors can add stable trapping and slow decay. These facts shape stability problems but do not by themselves prove turbulent collapse, universal instability, or nonlinear stability. Each verdict depends on the equations, dimension, boundary conditions, data class, symmetry, and norm.

Helpful background. Lorentzian Metrics and Causal Cones supplies boundary causality; Geodesics and Jacobi Fields supplies trapping; Causality and Global Structure supplies the global framework; Exact Solutions of Einstein’s Equations supplies AdS and black holes; Formulations of General Relativity supplies evolution systems; Gravitational Actions and Boundary Terms supplies boundary conditions.

Repeated reflection and resonant spectra

In global AdS a radial null ray reaches the boundary in time $\pi/2$ and, under reflecting conditions, returns to the center after total time $\pi$ in units where the AdS radius is one. Radiation cannot simply disperse to infinity as it does in an asymptotically flat spacetime. The same wave packet can revisit the central region many times and accumulate weak nonlinear interactions.

Linear normal modes on exact global AdS have discrete, highly commensurate frequencies in many field sectors. For a free scalar of effective mass $m$ on AdS$_{d+1}$, regular at the center and with the standard fast-falloff reflecting boundary condition, set $\Delta=\Delta_+=d/2+\sqrt{d^2/4+m^2\ell_{\rm AdS}^2}$. In dimensionless global time its positive frequencies are

$$\omega_{n\ell}=\Delta+2n+\ell$$

where $n=0,1,\ldots$ is the radial overtone and $\ell=0,1,\ldots$ the spherical-harmonic order; $\ell_{\rm AdS}$ denotes the AdS radius to avoid confusing it with angular momentum. Physical frequencies are $\omega_{n\ell}/\ell_{\rm AdS}$. Mixed boundary conditions can change this spectrum. Integer relations among the standard frequencies allow secular terms in naive perturbation theory. Multiple-scale or resonant-system methods then track slow energy transfer between modes.

This kinematic setup motivates an unresolved instability conjecture; it does not answer it. Resonant couplings can vanish by selection rules, energy can cycle rather than cascade, and nonlinear frequency shifts can detune resonances. Perturbatively constructed and numerically continued time-periodic scalar solutions, together with perturbative vacuum geons, give special noncollapsing families Maliborski and Rostworowski 2013; Horowitz and Santos 2015, §§2–4. Their existence does not prove that an open neighborhood is stable. Conversely, numerical evidence in selected symmetry classes shows cascades and black-hole formation for many data families; the original spherical scalar evidence and its scope are set out in Bizoń and Rostworowski 2011, pp. 1–4.

The formulation axis includes the size norm. Data small in total energy can be large in high Sobolev norms if concentrated at short scales. A theorem about an open set in one topology need not imply stability in another.

Trapping and decay

A bulk null geodesic is trapped when it remains in a spatially bounded region instead of escaping or crossing a horizon. Stable trapping means nearby rays share that behavior. For reflecting AdS evolution one must also consider rays reflected at the conformal boundary: repeated motion between the boundary and an angular-momentum barrier can prevent horizon absorption even without a stable circular null orbit in the interior. The relevant trapped set then belongs to the reflected ray flow. High-frequency waves can concentrate there, producing long-lived quasimodes and obstructing uniform decay estimates.

Schwarzschild–AdS makes the distinction concrete. With physical mass $M\gt 0$, lapse $f(r)=1-2GM/r+r^2/\ell_{\rm AdS}^2$, affine parameter $\lambda$, and null-geodesic constants $E$ and $L\ne0$, the radial equation is

$$\left(\frac{dr}{d\lambda}\right)^2+V_{\rm geo}(r)=E^2, \qquad \frac{V_{\rm geo}(r)}{L^2} =\frac1{\ell_{\rm AdS}^2}+\frac1{r^2}-\frac{2GM}{r^3}.$$

The maximum at $r=3GM$ is an unstable photon sphere. But rays arriving from the boundary with

$$\frac1{\ell_{\rm AdS}^2} \lt\frac{E^2}{L^2} \lt\frac1{\ell_{\rm AdS}^2}+\frac1{27G^2M^2}$$

turn outside that barrier, return to the boundary, and reflect back without reaching the horizon. The strict inequalities persist under small changes of the ray data. This is the stable trapping mechanism relevant to reflecting Schwarzschild–AdS. Waves can tunnel through the barrier, so the geometric-optics picture predicts long lifetimes rather than an exactly nondecaying wave Holzegel and Smulevici 2013, Introduction, pp. 2–4 of the Open PDF.

Stable trapping appears in several asymptotically AdS black-hole or horizonless geometries, depending on parameters and boundary conditions. It can lead to logarithmic or derivative-losing decay bounds. Slow linear decay weakens the estimates needed to control nonlinear terms, but it is not itself a nonlinear instability theorem. One must show that the persistent energy couples in a way that grows an appropriate norm or changes the geometry. A rigorous field-decay analysis with explicit Kerr–AdS hypotheses is given in Holzegel and Smulevici 2013, §§1–3.

Unstable trapping, such as a normally hyperbolic photon sphere, also slows decay but often supports stronger estimates than stable trapping. A Penrose diagram does not display either stability type: trapping is a phase-space property of the geodesic flow, including angular momentum, which radial conformal diagrams suppress.

Boundary recurrence and interior trapping are therefore distinguishable mechanisms, but they can cooperate. The reflected Schwarzschild–AdS example uses both the boundary and the photon-sphere barrier. A radial Penrose diagram displays boundary return while suppressing the angular momentum that makes those rays avoid the horizon.

Boundary causality

The timelike boundary has its own conformal light cones. A bulk causal curve connecting two boundary events should not arrive earlier than allowed by the boundary causal relation. The precise Gao–Wald theorem assumes a smooth conformal embedding with timelike boundary, null energy and null generic conditions in the physical spacetime, strong causality of the completed spacetime, and compact completed causal diamonds $J^+(p)\cap J^-(q)$. Under those hypotheses, any fastest causal curve joining a boundary point to the frontier of its bulk-reachable boundary set lies entirely in the boundary Gao and Wald 2000, Theorem 2, pp. 5005–5007. This boundary causality condition differs from bulk global hyperbolicity and from the choice of reflective field boundary conditions.

Time delay can be tested by comparing a bulk null geodesic with the fastest boundary null curve between its endpoints. Pure AdS saturates the relation for suitable antipodal paths but fails the theorem’s null generic condition. Perturbations that preserve the boundary geometry and satisfy all the hypotheses produce delay; the null energy condition may be replaced by the specific averaged condition cited by Gao and Wald, but averaged energy alone is not the full theorem. Higher-derivative corrections, exotic matter, or inappropriate effective couplings can threaten the inequality and thereby constrain a proposed theory.

Boundary causality is global and end dependent. It does not imply that every boundary correlator has ordinary local support without additional quantum assumptions, nor does a coordinate travel time prove violation. One must use the fixed boundary conformal metric, complete the bulk curve, and compare invariant causal relations.

Stability questions by formulation

As of the evidence cutoff 2026-08-12, no single proposition called “AdS instability” covers all relevant systems. At least these formulations must be separated:

Question	Necessary specification	Possible outcome
Exact AdS nonlinear stability	vacuum or matter equations, dimension, reflecting condition, data topology	global regularity, weak turbulence, or collapse for an open/generic set
Island stability	time-periodic/background solution, symmetry, perturbation norm	orbital stability, metastability, or instability
Black-hole exterior stability	horizon geometry, rotation, field sector, boundary condition	decay with rate/loss, mode stability, or growing mode
Stable-trapping effect	trapped-set geometry and frequency norm	quasimodes, slow decay, derivative loss; nonlinear consequence separate
Transparent-boundary evolution	coupled exterior/boundary system	dispersion can differ qualitatively from closed AdS

The evidence combines rigorous linear results, perturbative resonant analyses, numerical simulations, exact special solutions, and one theorem-level collapse result in a specific matter model. Moschidis constructs smooth, spherically symmetric Einstein–massless-Vlasov data that are arbitrarily small in the theorem’s stated norm and, with reflecting boundary conditions, evolve to contain a trapped sphere Moschidis 2023, Theorem 1 and §4. This proves the existence of arbitrarily small collapsing data for that well-posed matter model; it proves neither generic collapse nor the vacuum or scalar instability conjecture. Likewise, a numerical cascade in spherical symmetry is not a theorem for nonsymmetric vacuum data; a time-periodic solution is not proof that an open neighborhood is stable; and linear mode stability is not nonlinear stability.

Nonlinear global-existence or collapse theorems, stable counterexamples in specified unrestricted data topologies, corrected numerical results, or different boundary conditions can change the conclusions for a particular model. Dated research discussions compare those evolving results; the causal definitions developed here remain the starting point.

Conserved energy is not enough

Reflective boundary conditions can conserve a positive total energy. Conservation prevents the total norm represented by that energy from growing, but collapse can arise by concentration: energy moves to smaller spatial scales while the total remains fixed. Higher derivatives or curvature can grow even when the energy is constant.

Conversely, a turbulent spectral cascade need not form a horizon. It may reverse, saturate, or remain globally regular. A useful diagnostic must connect modal transfer to geometric concentration and verify convergence with resolution and boundary treatment.

Charges constrain endpoints but do not select them. Angular momentum, gauge charge, or additional integrals can suppress or redirect collapse in some sectors. Boundary sources can inject energy, invalidating closed-system conservation. Every stability argument must use the same boundary phase space as its charge statement.

Exercise

Classify the strength of the following inference: “A linear scalar wave equation on a reflecting AdS background has a discrete resonant spectrum; therefore every sufficiently small nonlinear gravitational perturbation forms a black hole.” Identify at least four missing steps.

Solution

The inference is not justified. A resonant linear spectrum establishes a possible channel for secular nonlinear transfer, not universal collapse. Missing steps include: deriving the actual nonlinear coupling coefficients and selection rules; controlling detuning and long-time perturbation errors; specifying the vacuum or matter equations, dimension, boundary condition, symmetry class, and small-data topology; and proving that energy transfer produces geometric concentration or a trapped surface rather than recurrence or a regular time-periodic state.

One must also address islands of stability, nonsymmetric modes, numerical convergence if simulations are used, and the distinction between generic/open-set behavior and every datum. The correct conclusion is that resonance makes long-time dispersion-based stability difficult and motivates a model-specific investigation.

The next chapter turns from timelike boundary dynamics to topology, beginning with Topology of Cauchy Surfaces.

References

• Bizoń, Piotr, and Andrzej Rostworowski, “On Weakly Turbulent Instability of Anti-de Sitter Space”, Physical Review Letters 107, article 031102, pp. 1–4, 2011.
• Gao, Sijie, and Robert M. Wald, “Theorems on Gravitational Time Delay and Related Issues”, Classical and Quantum Gravity 17, pp. 4999–5008, 2000. Open PDF.
• Holzegel, Gustav, and Jacques Smulevici, “Decay Properties of Klein–Gordon Fields on Kerr–AdS Spacetimes”, Communications on Pure and Applied Mathematics 66, §§1–3, pp. 1751–1802, 2013.
• Horowitz, Gary T., and Jorge E. Santos. “Geons and the Instability of Anti-de Sitter Spacetime.” Surveys in Differential Geometry 20 (2015): 321–335. DOI; arXiv:1408.5906.
• Maliborski, Maciej, and Andrzej Rostworowski. “Time-Periodic Solutions in an Einstein AdS–Massless-Scalar-Field System.” Physical Review Letters 111 (2013): 051102. DOI.
• Moschidis, Georgios. “A Proof of the Instability of AdS for the Einstein–Massless Vlasov System.” Inventiones Mathematicae 231 (2023): 467–672. DOI.