Far-From-Equilibrium Dynamics and Turbulence

Far-from-equilibrium holography replaces a kinetic description by a nonlinear asymptotically AdS gravitational initial-value problem. It has established that hydrodynamics can become accurate before local isotropy in several strongly coupled large-N theories. Stronger claims—one universal attractor, one rapid-thermalization time, or one turbulent cascade—depend on dimension, field content, symmetry, initial data, and the diagnostic used. Evidence is assessed through 2026-08-20.

Required background. Gauge Theory and String-Theory Preparation supplies the large-N, strong-coupling regime and the stress-tensor dictionary. Helpful background. Quantum Gravity supplies semiclassical control, Quantum Information supplies state and entropy distinctions, Black-Hole Perturbations supplies nonhydrodynamic modes, and Gravitational Radiation supplies characteristic evolution and constraint monitoring.

Nonlinear data and three different endpoints

The cleanest setting is a boundary state with a classical Einstein-gravity dual. One specifies the bulk theory, boundary metric and sources, a regular initial slice or characteristic null data, and asymptotically AdS boundary conditions. Near-boundary coefficients determine a conserved stress tensor,

$$\nabla_\mu T^{\mu\nu}=0, \qquad T^{\mu\nu}=T_{\rm hyd}^{\mu\nu}[\varepsilon,u^\mu] +\Pi_{\rm nonhyd}^{\mu\nu} .$$

The three endpoints commonly conflated in “thermalization” are distinct. Isotropization means that principal pressures become nearly equal in a chosen local rest frame. Hydrodynamization means that a constitutive approximation predicts the stress tensor to a declared accuracy. Thermalization additionally invokes an equilibrium ensemble and decay of all monitored disturbances. Where $T^{\mu}{}_{\nu}$ has a future-timelike Landau eigenvector, choose the associated orthonormal Landau-frame tetrad. Where no such eigenvector exists—for example, for a nonzero null stress—choose and report a future-timelike observer tetrad instead. In either case define the positive component norm $\|X\|_{\rm comp}^2=\sum_{\hat a,\hat b}(X^{\hat a\hat b})^2$. A useful operational measure is

$${\cal R}_{\rm hyd}(x)= \frac{\|\Delta T\|_{\rm comp}} {\|T\|_{\rm comp}}, \qquad \Delta T^{\mu\nu}=T^{\mu\nu}-T_{\rm hyd}^{\mu\nu},$$

with the tetrad, component weights, hydrodynamic order, and threshold specified. This Euclidean component norm is deliberately frame-fixed and positive; outside a Landau-frame region its numerical value is observer-dependent and must not be compared across frames without a stated transport prescription. A Lorentzian contraction would not be a norm and could vanish for nonzero null stress. A small residual can coexist with substantial pressure anisotropy because viscous terms themselves are anisotropic. Homogeneous isotropization calculations demonstrate that nonhydrodynamic quasinormal contributions can decay rapidly in strongly coupled plasma Chesler and Yaffe 2009, complete Letter, pp. 1–4, but their symmetry removes shocks, vorticity, and spatial cascades.

For collisions, the formulation adds incoming energy sheets, their longitudinal profile and transverse homogeneity or structure. For turbulence, it adds a domain, forcing or initial spectrum, dissipation scale, and a gauge-invariant boundary observable. Numerical convergence requires the radial constraints and total boundary energy to agree across resolutions; agreement of a visually complicated horizon alone is not a boundary-field-theory diagnostic.

State of the evidence through 2026-08-20

The matrix records controlled results and their adverse scope conditions. “Supported” means reproduced within the listed gravitational model, not established for finite N, finite coupling, QCD, or arbitrary initial states.

setting and assumptions	controlled evidence or position	adverse result, limitation, or discriminator	primary source
Homogeneous, anisotropic states of strongly coupled large-N supersymmetric Yang–Mills theory with an AdS5 Einstein dual	Regular numerical evolutions show rapid isotropization toward equilibrium	After driving ends, homogeneous static flow has no viscous gradient stress; it does not independently demonstrate hydrodynamization with large persistent anisotropy. Initial radial profiles change transients	Chesler and Yaffe 2009, complete Letter, pp. 1–4
Planar gravitational shocks in AdS5, with varied longitudinal widths and fixed transverse symmetry	Collisions permit a direct constitutive-residual test; varying width reveals stopping and transparent regimes	The outcome depends on the dimensionless shock thickness, so no single collision narrative applies to all data	Chesler and Yaffe 2011, complete Letter, pp. 1–4; Casalderrey-Solana et al. 2013, pp. 1–4, Figs. 1–4
Boost-invariant strongly coupled plasma expanded at late proper time	Many gradient coefficients exist, but their factorial growth shows that the hydrodynamic series is asymptotic and encodes nonhydrodynamic scales	This is evidence against identifying hydrodynamization with convergence of the gradient expansion; resummation and attractor statements retain observable and model dependence	Heller, Janik, and Witaszczyk 2013, complete Letter, pp. 1–5
Long-wavelength flows dual to AdS4 black holes, effectively two-dimensional on the boundary	Boundary velocity spectra and bulk evolution support an inverse energy cascade in a controlled hydrodynamic regime	Dimensionality fixes the cascade direction, and the evidence does not establish the same cascade in three spatial dimensions or far outside hydrodynamics	Adams, Chesler, and Liu 2014, complete Letter, pp. 1–5

Two conclusions are robust within these examples. First, regular black-brane evolution plus holographic renormalization supplies a quantitative, conservation-constrained route from initial data to the stress tensor. Second, hydrodynamization need not wait for small gradients in a naive sense or for isotropy. The divergence of the gradient series is not a failure of hydrodynamics; it means the constitutive expansion requires asymptotic or resummed interpretation, with transient modes setting its large-order behavior.

Competing positions and discriminating tests

One position emphasizes the recurrence of early hydrodynamization across homogeneous relaxation, boost-invariant expansion, and planar shock collisions. On this view, a low-dimensional hydrodynamic attractor organizes many initial states once energy density fixes the local scale. Evidence for this position is substantial within particular Einstein-gravity models. A stronger universal-attractor claim is nevertheless model-dependent: higher-derivative corrections move nonhydrodynamic modes, conserved charges add slow sectors, and different observables need not define the same attractor.

An alternative position treats “hydrodynamization” as a practical accuracy statement rather than a universal dynamical object. It asks whether a specified truncation predicts a specified observable below a specified tolerance. This conservative definition travels cleanly between models but says less about why many trajectories collapse onto a common curve. The decisive comparison is not visual overlap; it is the residual after varying state class, field content, hydrodynamic frame, truncation order, and observable.

Turbulence exposes a sharper dimension dependence. For an incompressible boundary velocity field, define the shell-averaged kinetic spectrum $E(k)$ so that

$$\frac12\int d^{D-1}x\,|\mathbf{u}|^2=\int_0^\infty E(k)\,dk .$$

An apparent power law over less than a decade is not a cascade. One must show a scale-independent flux over an inertial interval, energy conservation, ultraviolet refinement, and insensitivity to gauge and extraction choices. In two boundary spatial dimensions, the inverse cascade supported by Adams, Chesler, and Liu 2014, pp. 2–5 is compatible with ordinary two-dimensional fluid mechanics. It is an adverse case for any claim that holographic black holes enforce one dimension-independent cascade direction.

Shock collisions provide another counterexample to overgeneralization. Varying the longitudinal thickness changes whether energy piles up near midrapidity or passes through before hydrodynamic flow emerges Casalderrey-Solana et al. 2013, pp. 1–4, Figs. 1–4. The gravitational equations are the same, but the data class differs. “Rapid” consequently means a dimensionless time such as $tT_{\rm eff}$ measured from a declared event, not an absolute universal constant.

Numerical controls and what remains open

A credible nonlinear calculation reports the bulk action and dimension; coordinates and gauge; initial and boundary data; spectral or finite-difference discretization; radial and boundary resolution; apparent-horizon handling; filtering; and the observables extracted. At every refinement it should monitor radial constraints, boundary conservation, total energy, and the change in hydrodynamic residual. A proposed turbulent cascade must also survive a gauge-invariant boundary spectrum and ultraviolet extension of its inertial range. A proposed attractor must survive changes in initial-state families rather than only denser sampling of one family.

It remains open how broadly attractor behavior persists at finite coupling and finite N, how conserved charges and phase structure reorganize far-from-equilibrium evolution, and which turbulent phenomena survive in fully three-dimensional, inhomogeneous boundary flows at adequate scale separation. It is also unresolved which coarse-grained observables offer an operational thermalization criterion shared by gravity and field theory. Present evidence does not establish a universal ordering of isotropization, hydrodynamization, entanglement growth, and microscopic equilibration.

What could change these conclusions

The conclusions could change when an independent high-resolution simulation reproduces or overturns a claimed cascade, when a counterexample initial-data class defeats a proposed attractor, or when a theorem or controlled finite-coupling calculation fixes its domain. A new gauge-invariant boundary diagnostic, a change of field content or boundary condition, or a nonperturbative error estimate could also change the interpretation. Merely increasing horizon-image resolution without demonstrating smaller constraints, conserved energy, and a longer physical inertial range does not change the assessment.

Exercises

Determine whether a reported evolution has hydrodynamized. At one event, suppose the normalized first-order constitutive residual is $0.06$, the second-order residual is $0.018$, the pressure anisotropy is $0.35$, and the declared accuracy threshold is $0.02$. State what follows and what does not.

Solution

The second-order residual lies below the stated threshold, whereas the first-order residual does not. Therefore the event has hydrodynamized only relative to the second-order constitutive model, chosen norm, observable, and two-percent criterion. The large pressure anisotropy does not contradict that conclusion because viscous stresses can be anisotropic. It does show that isotropization has not occurred. Nothing in these numbers establishes a thermal density matrix, decay of every nonhydrodynamic observable, or universality across initial data; those require separate checks.

References

• Adams, Allan, Paul M. Chesler, and Hong Liu. “Holographic Turbulence.” Physical Review Letters 112 (2014): 151602. DOI. Open PDF.
• Casalderrey-Solana, Jorge, Michal P. Heller, David Mateos, and Wilke van der Schee. “From Full Stopping to Transparency in a Holographic Model of Heavy Ion Collisions.” Physical Review Letters 111 (2013): 181601. DOI. Open PDF.
• Chesler, Paul M., and Laurence G. Yaffe. “Holography and Colliding Gravitational Shock Waves in Asymptotically AdS5 Spacetime.” Physical Review Letters 106 (2011): 021601. DOI. Open PDF.
• 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. Open PDF.
• Heller, Michal P., Romuald A. Janik, and Przemyslaw Witaszczyk. “Hydrodynamic Gradient Expansion in Gauge Theory Plasmas.” Physical Review Letters 110 (2013): 211602. DOI. Open PDF.