Black-Hole Formation and Thermalization

In gauge/gravity duality, black-hole formation can encode equilibration of a boundary quantum state, but the dictionary is narrower than the slogan “a horizon means thermalization.” Horizon formation, entropy production, relaxation of one-point functions, isotropization, hydrodynamization, and approach to a thermal density matrix are different statements with different diagnostics. The connection is controlled only after the bulk theory, boundary theory, state, sources, and classical-gravity limit have been declared.

Helpful background. Black-Hole Geometry and Mechanics supplies event and apparent horizons; Black-Hole Perturbation Theory supplies relaxation modes; Formulations of General Relativity supplies characteristic evolution; Exact Solutions of Einstein’s Equations supplies AdS black branes; Causality and Global Structure supplies the conformal boundary; and Computational Relativity supplies numerical validation.

A precise holographic laboratory

A clean example uses strongly coupled $\mathcal N=4$ supersymmetric Yang–Mills theory at large color number $N_c$ and large ‘t Hooft coupling. For the neutral stress-tensor sector considered here, the classical bulk description lies in the five-dimensional Einstein sector of the consistent truncation of type-IIB supergravity on AdS$_5\times S^5$. This is not QCD, not the complete finite-$N_c$ duality, and not a claim about finite-$N_c$ unitary dynamics; it is a controlled limit of a particular conformal theory.

Chesler and Yaffe start with the field-theory vacuum and drive it through a homogeneous, anisotropic boundary metric,

$$ds^2_{\partial}=-dt^2+e^{B_0(t)}d\mathbf x_\perp^2 +e^{-2B_0(t)}dx_\parallel^2,$$

where the spatial volume element is constant. Setting the AdS radius to $L_{\rm AdS}=1$, they use infalling bulk coordinates,

$$ds^2=-A(v,r)dv^2+2\,dv\,dr +\Sigma(v,r)^2 \left(e^{B(v,r)}d\mathbf x_\perp^2 +e^{-2B(v,r)}dx_\parallel^2\right).$$

The boundary deformation sends gravitational radiation inward and performs work on the quantum system. The bulk evolution equations, radial constraints, boundary conditions, and residual radial gauge are specified in Chesler and Yaffe 2009, Eqs. (1)–(8), pp. 1–2. An apparent horizon forms in the driven geometry; the exterior later approaches a stationary AdS black brane.

The near-boundary coefficients determine the renormalized stress tensor,

$$\langle T^\mu{}_{\nu}\rangle =\operatorname{diag}(-\mathcal E, \mathcal P_\perp,\mathcal P_\perp,\mathcal P_\parallel).$$

After the source becomes static, conformal invariance gives the equilibrium relation $\mathcal P=\mathcal E/3$. The directly measured relaxation criterion in this model is pressure isotropization, for example

$$\frac{|\mathcal P_\parallel-\mathcal P_\perp|}{\mathcal E}\lt \epsilon_{\mathrm{iso}}$$

for a declared tolerance. It is a boundary one-point-function statement, not a proof that every correlation function has thermalized.

The injected energy can be checked without identifying a horizon. Stress-tensor conservation in the time-dependent boundary metric gives

$$\dot{\mathcal E} =\dot B_0\left(\mathcal P_\parallel-\mathcal P_\perp\right).$$

Indeed the two transverse scale factors have expansion rate $\dot B_0/2$ and the longitudinal one $-\dot B_0$, so their total volume expansion vanishes while anisotropic work remains. Once $\dot B_0=0$, the energy density is constant even though the pressures can continue to relax. Conservation and equilibration are therefore independent tests; the conformal trace anomaly during driving does not replace this conservation identity.

Five clocks that should not be conflated

Evidence cutoff: 12 August 2026. The current interpretation is best organized by observable.

Event or diagnostic	What it establishes	What it does not establish
Apparent-horizon appearance	A chosen foliation contains an outer marginally trapped surface; it supports excision and a quasi-local entropy proxy	Foliation-independent formation time or complete boundary equilibration
Event horizon	The global bulk solution separates rays that reach the boundary from those that do not	A causal local onset time; its early growth is teleological
Exterior ringdown	Perturbations of the final black object decay in the measured channels	Decay of every nonlinear observable or exact thermality
Hydrodynamization	A constitutive expansion describes the stress tensor within tolerance	Local isotropy or small nonhydrodynamic corrections in all channels
Isotropization	Principal pressures agree within a declared threshold	Thermalization of higher-point functions, entanglement, or finite-$N_c$ recurrences

In the homogeneous example no spatial hydrodynamic modes are excited, so isotropization is the useful boundary clock. Chesler and Yaffe compare the pressure history, apparent and event horizons, and final temperature in Chesler and Yaffe 2009, Figs. 1–3 and Table I, pp. 3–4. Their event horizon grows before the strongest boundary driving precisely because its location depends on the future. The apparent horizon responds more locally, yet its location still depends on the bulk slicing.

This separation corrects a common inference: bulk horizon formation is strong evidence for coarse-grained entropy in the classical limit, but it does not provide a gauge-invariant boundary timestamp by itself. A boundary statement should be based on boundary observables and their Ward identities.

Competing endpoints in global AdS

The global AdS scalar problem adds another boundary topology and another set of endpoints. A sufficiently concentrated pulse can form a black hole, whose exterior may relax. Other initial data lie near time-periodic solutions or stability islands and need not form a horizon on the observed timescale Martinon 2017, §§4.2–4.6, pp. 18–29. Thus even within classical reflecting AdS, “generic perturbation thermalizes” is an unresolved formulation-dependent claim.

The status can be separated into three positions.

1. Collapse-led equilibration. For data that form a sufficiently large stable AdS black hole, absorption and exterior decay make equilibration of accessible coarse observables plausible.
2. Nonthermal classical motion. Perturbatively and numerically constructed periodic solutions, geons, boson stars, and long-lived neighborhoods support alternatives to horizon formation. Their existence and stability must be judged within each field model; finite-time survival alone does not prove global regularity.
3. Fine-grained unitarity. At finite $N_c$, a boundary theory evolves unitarily. Poincaré recurrence additionally requires finite spatial volume or an appropriately discrete spectrum; it does not follow from finite $N_c$ alone, especially in the planar infinite-volume problem. Exact information loss and recurrence scales both lie outside classical bulk evolution.

These positions are compatible once their approximations are stated. The open question concerns the topology and measure of their basins for each boundary theory and energy range, not whether one carefully chosen driven state relaxes.

An isolated pure boundary state remains pure under unitary evolution, so its full density matrix cannot become an exact mixed Gibbs state. Thermalization in such a setting concerns a specified subsystem or set of coarse observables. The classical horizon describes that restricted information, not a change in the eigenvalues of the complete density matrix.

A decisive update would require a theorem or validated survey characterizing an open or generic set of boundary states, together with multiple boundary observables and controlled limits in $N_c$, coupling, volume, and time. One-point isotropization in a homogeneous classical calculation cannot by itself resolve that question.

Interpretation checks

Three tests protect the dictionary.

• Ward identities: verify the boundary stress tensor obeys the sourced conservation law and conformal trace relation, including anomaly terms when the boundary metric is curved.
• Horizon independence: state whether the reported surface is event, apparent, or trapping, and compare its timing only with observables defined on the same foliation or at the boundary.
• Observable hierarchy: report separate tolerances for pressure anisotropy, hydrodynamic residual, and any nonlocal observable. Do not rename the earliest one “thermalization.”

The numerical methods used for general asymptotically AdS characteristic evolution—including radial nesting, spectral domains, horizon excision, boundary expansions, and convergence tests—are developed in Chesler and Yaffe 2014, §§3–4, pp. 10–36. Those methods are essential evidence for a computation, but no new evolution was rerun for this page.

Exercise: diagnose three reported times

Consider a more general expanding plasma than the homogeneous, source-free final stage above. A simulation reports an apparent horizon at $tT=0.2$, a constitutive-relation residual below $10\%$ at $tT=0.6$, and pressure anisotropy below $10\%$ at $tT=1.1$, using one fixed reference temperature $T$. Which times may be called horizon formation, hydrodynamization, and isotropization? Has full thermalization been established?

Solution

With the foliation stated, $tT=0.2$ is the apparent-horizon time. The residual criterion defines hydrodynamization at $tT=0.6$. The pressure criterion defines isotropization at $tT=1.1$. Their order is physically possible in a flowing plasma because hydrodynamic viscous stresses can themselves be anisotropic. In a homogeneous static state with no gradients, that explanation is unavailable: hydrodynamic equilibrium is isotropic.

Full thermalization has not been established. That stronger claim would require a declared set of local and nonlocal observables, their tolerances, and control of the classical large-$N_c$, strong-coupling approximation. The apparent horizon alone is not a boundary observable, and pressure isotropy alone does not constrain higher-point functions.

References

• Chesler, Paul M., and Laurence G. Yaffe. “Horizon Formation and Far-from-Equilibrium Isotropization in Supersymmetric Yang–Mills Plasma.” Physical Review Letters 102 (2009): 211601. doi:10.1103/PhysRevLett.102.211601. Open PDF.
• Chesler, Paul M., and Laurence G. Yaffe. “Numerical Solution of Gravitational Dynamics in Asymptotically Anti-de Sitter Spacetimes.” Journal of High Energy Physics 07 (2014): 086. doi:10.1007/JHEP07(2014)086. Open PDF.
• Martinon, Grégoire. “The Instability of Anti-de Sitter Space-Time.” arXiv:1708.05600v3 (2017), 58 pp. Abstract. Open PDF.