Maximin Constructions and Extremal-Surface Existence

Maximin converts the Lorentzian HRT saddle problem into a nested optimization: minimize area on each complete achronal bulk slice, then maximize that minimum over slices. Under AdS global-hyperbolicity, compactness or suitable limiting assumptions, stability, and the null-curvature focusing condition, a stable maximin surface is an HRT surface and inherits useful causal and nesting properties. The construction is not a hypothesis-free definition of HRT; remove its existence or focusing inputs and its conclusions need not follow. We work in a classical Lorentzian asymptotically AdS Einstein spacetime with regulated boundary anchors.

Required background. HRT defines the covariant extremal target, and causal wedges supplies the null geometry that maximin constrains.

Helpful background. Raychaudhuri focusing supplies the null-curvature step, and causal propagators clarifies the global-hyperbolicity domain.

Minimize on a slice, then maximize over slices

Let $C_\partial$ be a boundary Cauchy slice containing $A$. For every complete achronal bulk slice $\Sigma$ with $\partial\Sigma=C_\partial$, define $m(A,\Sigma)$ to be the least-area codimension-two surface in $\Sigma$ that is anchored on $\partial A$ and homologous to $A$ within $\Sigma$. Then choose

$$M(A)=m(A,\Sigma_*) , \qquad \operatorname{Area}[m(A,\Sigma_*)] =\max_\Sigma\operatorname{Area}[m(A,\Sigma)].$$

The order matters. Minimization controls spatial variations tangent to $\Sigma_*$. Maximization controls the independent timelike normal variation of the slice. At a stable interior optimum, both first variations vanish, so the two null expansions of $M(A)$ vanish: it is spacetime extremal.

To identify $M(A)$ with the least-area admissible HRT surface, the proof compares any extremal surface with its representative obtained by following orthogonal null congruences to $\Sigma_*$. If

$$R_{ab}k^ak^b\geq0$$

for every null $k^a$, Raychaudhuri’s equation implies that the cross-sectional area cannot increase along the relevant focused null sheet. This bounds the representative’s area and closes the global comparison. Wall gives the precise stable-maximin construction and its relation to HRT under these hypotheses (Wall 2014, §§2–3).

Two-sided AdS application

Consider the time-reflection-symmetric two-sided AdS black hole and take $A$ to be the complete right boundary on the $t_R=0$ boundary slice. The homology constraint requires a bulk cut separating right from left. On the reflection-symmetric slice $\Sigma_0$, the least-area cut is the bifurcation surface $B$, so

$$m(A,\Sigma_0)=B, \qquad S_A^{(0)}=\frac{\operatorname{Area}(B)}{4G_N}.$$

To check the maximization, launch future- and past-directed null sheets from $B$. At $B$, $\theta_{(k)}=\theta_{(\ell)}=0$. Raychaudhuri gives

$$\frac{d\theta}{d\lambda} =-\frac{\theta^2}{d-1}-\sigma_{ab}\sigma^{ab} -R_{ab}k^ak^b\leq0.$$

Hence cross sections of the relevant light sheets away from $B$ have area no larger than $\operatorname{Area}(B)$, until caustics where the representative is treated piecewise. For any admissible $\Sigma$, intersect an appropriate light sheet with $\Sigma$ to obtain a surface $\widetilde B(\Sigma)$ homologous to $A$. Since $m(A,\Sigma)$ is the minimum on that slice,

$$\operatorname{Area}[m(A,\Sigma)] \leq\operatorname{Area}[\widetilde B(\Sigma)] \leq\operatorname{Area}(B).$$

The symmetric slice saturates the bound, so $M(A)=B$. This calculation shows what each assumption does: reflection symmetry supplies the candidate, homology fixes the separating class, compactness supplies the slice minimum, and null focusing supplies the upper bound on every other slice.

For a general two-sided time-dependent geometry, one repeats the construction without assuming the bifurcation surface. The maximizing slice and its minimal cut must be found together; a wormhole throat on a convenient coordinate slice is not enough.

Causal consequences under the same hypotheses

Focusing also places the maximin/HRT surface outside the causal future and past of $D[A]$ under the standard asymptotically AdS causal assumptions. Otherwise a null deformation tied to the boundary domain would contradict extremality and the area comparison. This supports causal consistency and, for nested boundary regions with jointly chosen slices, entanglement-wedge nesting.

These are conditional geometric results. The null-curvature condition is classical. At order $G_N^0$, matter entropy changes the functional and quantum focusing or generalized-entropy statements replace the classical area argument; one cannot simply append $S_{\rm bulk}$ to the last line of a classical proof.

Three ways the theorem can fail

Loss of global hyperbolicity. If the spacetime has naked timelike singularities, unprescribed internal boundaries, or causal pathology, complete achronal slices may not exist or may not cover the target region. Then the maximization domain is undefined until boundary conditions and a causal completion are supplied.

Loss of compactness or stability. In a noncompact variational family, minimizing surfaces can run to infinity or a supremum over slices need not be attained. A regulator may produce a sequence $\Sigma_n$ with increasing minima but no limiting slice. One may report a regulated supremum, not an existing HRT surface.

Loss of null focusing. If $R_{ab}k^ak^b<0$ sufficiently strongly, the Raychaudhuri inequality reverses its useful area monotonicity. A null representative may have larger area than the extremal surface, breaking the comparison used for HRT equivalence and nesting. The variational surface may still exist in a particular geometry, but the primary theorem no longer licenses it.

This is the adversarial boundary: verify every hypothesis before exporting existence, causal placement, strong subadditivity, or nesting. A numerical extremum found beyond the theorem’s domain is evidence about that example, not a general result.

Controlled extensions and handoff

Classical maximin is appropriate at leading order in Einstein gravity. Higher-curvature actions alter the functional and can alter focusing arguments. Quantum maximin and QES constructions require renormalized generalized entropy and quantum focusing assumptions. Degenerate maxima require a stability prescription, and multiple homology classes still require global competition.

Higher-derivative functionals changes the variational problem; FLM and QES supplies the semiclassical quantum replacement; and wedge nesting applies the classical result without extending it past its assumptions.

The chapter overview contains the structure diagram and validity and failure diagram. They are embedded there once so that their shared chapter-level context is not repeated on every article.

For the chapter-wide comparison of assumptions, counterevidence, falsifiers, and claim ceilings, see the claim-domain table.

References

• Wall, A. C. (2014). “Maximin surfaces, and the strong subadditivity of the covariant holographic entanglement entropy.” Classical and Quantum Gravity 31, 225007. DOI.