Continuum generalized entropy beyond the regulated model

Fix a fully specified four-dimensional semiclassical Einstein-scalar collapse model: its scalar field equation and couplings with SI units, fixed-coupling semiclassical equation and causal prescription, initial and boundary data, a globally hyperbolic domain, and a covariant renormalization prescription. Here causal curves are piecewise $C^1$ curves with consistently future- or past-directed causal tangent, and $J^{\pm }(K)$ is the set of points reachable from or reaching a set $K$ along such curves. “Globally hyperbolic” uses the supplied Hadamard-contract convention: strong causality (each point has arbitrarily small neighborhoods met in a connected parameter interval by every causal curve) and compact causal diamonds $J^{+}(p)\cap J^{-}(q)$. “Semiclassical” means the metric is treated as a classical field while the quantum scalar state contributes through the adopted effective equation. A covariant prescription here has the supplied local-covariance property: its assigned local data commute with pullback under smooth time-orientation-preserving isometric embeddings $\psi :(M,g)\to (M^{\mathrm{\prime }},g^{\mathrm{\prime }})$ with causally convex image that preserve the prescribed scalar potential, $V={\psi }^{\ast }{V}^{\mathrm{\prime }}$, comparing the same scale $\mu$ and finite normalization conditions $\mathbf{c}$. Potential preservation is equivalently $P({\psi }^{\ast }f)={\psi }^{\ast }(P^{\mathrm{\prime }}f)$ for every $f\in C^{\mathrm{\infty }}(M^{\mathrm{\prime }})$. This assigns no preferred state; covariance of stress expectations is required only for supplied states whose two-point distributions satisfy $W_{\omega }=(\psi \times \psi {)}^{\ast }{W}_{{\omega }^{\mathrm{\prime }}}$. The prescription fixes one scale $\mu$, the local curvature counterterms, and their finite normalization conditions; the same prescription and its prescribed regulator dependence are used at both cuts and for every regulator sequence compared below. The spacetime of the selected solution is a smooth time-oriented Lorentzian four-manifold $(M,g,N)$ of signature $(-,+,+,+)$, with Levi-Civita connection $\mathrm{\nabla }$ (the unique torsion-free, metric-compatible connection); $N$ is the supplied smooth future-timelike field fixing the time orientation. Select an actually existing smooth Lorentzian metric $g$ (a smooth symmetric covariant two-tensor field on $M$) and a positive normalized Hadamard state $\omega$ on the specified scalar-field algebra, such that $g$ and $\omega$ satisfy the adopted semiclassical equation and the model's supplied initial and boundary data. Here $\omega$ supplies the quantum scalar field in that equation; no additional classical scalar field $\varphi$ is part of the data unless the selected model separately specifies its equation and coupling. Its two-point distribution $W_{\omega }$ is a bisolution with $\mathrm{WF}(W_{\omega })={\mathcal{C}}_g^{+}$, where ${\mathcal{C}}_g^{+}$ consists of pairs $(x,\xi ;y,-\zeta )$ of nonzero future-null covectors that agree at a coincident point or are the metric-dual tangent covectors, with a common positive scale, along one nonconstant future-directed null geodesic. The supplied Hadamard contract also fixes a state-independent local singular bidistribution $H_{\mu }$ with $W_{\omega }-H_{\mu }$ smooth in each specified convex normal neighborhood; such a neighborhood is an open set on which the exponential map at every point is a diffeomorphism from a star-shaped neighborhood of zero in the tangent space. No existence, uniqueness, or stability of such a solution is assumed or claimed. The model remains the adopted effective theory below its stated cutoff; “nonperturbative in the scalar field” means that the proposed sufficient conditions and argument use no truncation in the scalar-field amplitude or state correlations, not that they construct exact quantum gravity.

For the relation ${\mathcal{C}}_g^{+}$ above, the two base-point parameters may occur in either order along the future-directed null geodesic; the positive-frequency designation applies to the covectors.

The separate fixed-background free-collapse scattering postulate is a Schwarzschild model for a massless minimally coupled free scalar with an incoming ground state, an exponential late-time ray map, and a supplied scattering channel. Its Schwarzschild exterior, state, ray-map, late-time convergence, and transmission assumptions are not imposed on the coupled model in this question unless the selected model independently supplies them.

The fixed-coupling equation is $G_{ab}+{\mathrm{\Lambda }}_{\mathrm{c}\mathrm{c}}{g}_{ab}+{\alpha }_{\mathrm{c}\mathrm{t}}{H}_{ab}^{(R^2)}+{\beta }_{\mathrm{c}\mathrm{t}}{H}_{ab}^{({\mathrm{Ric}}^2)}=\frac{8\pi G}{c^4}{T}_{ab}^{\mathrm{r}\mathrm{e}\mathrm{n}}[g,\omega ]$ below the supplied effective cutoff. Its scalar-field causal prescription supplies $P=-g^{ab}{\mathrm{\nabla }}_a{\mathrm{\nabla }}_b+V$, with smooth real $V$ of units ${\mathrm{m}}^{-2}$ in length coordinates, and a retarded map $E_R:C_0^{\mathrm{\infty }}(M)\to C^{\mathrm{\infty }}(M)$ satisfying $P{E}_{R}f=f=E_{R}Pf$ and $\mathrm{supp}(E_{R}f)\subseteq J^{+}(\mathrm{supp}f)$ for every $f\in C_0^{\mathrm{\infty }}(M)$; the initial state data are separate supplied inputs. These are adopted model premises, not a solution-existence theorem.

Select a smooth portion $\mathcal{H}\subset M$ of the future null horizon in this solution. Here $\mathcal{H}$ means a smooth embedded three-dimensional submanifold for which the restriction of $g_p$ to $T_p\mathcal{H}$ has one-dimensional radical $L_p:=\{v\in T_p\mathcal{H}:g_p(v,w)=0\text{ for every }w\in T_p\mathcal{H}\},$ and choose a smooth nowhere-zero future-directed field $k$ along $\mathcal{H}$ with $k_p\in L_p$ and ${\mathrm{\nabla }}_{k}k=0$. These are the exact smooth-null-hypersurface and affine-generator conditions of the supplied smooth null-screen definition. Its screen spaces ${\mathcal{S}}_p:=k_p^{\perp }\mathrm{/}\mathbb{R}{k}_p$ form a smooth rank-two bundle with the positive-definite quotient metric induced by $g$; for local screen sections represented by tangent fields $X,Y$, the screen deformation form $B([X],[Y])=g({\mathrm{\nabla }}_{X}k,Y)$ is well defined on this quotient. The expansion and shear are respectively $\theta :={\mathrm{tr}}_{\bar{g}}B$ and $\sigma :=B-\frac{1}{2}\theta \bar{g}$; this problem imposes no stationarity, completeness, or restriction on either quantity. If $\mathcal{H}$ is intended to be an event horizon, the model must additionally supply a conformal completion: $M$ is the interior of a manifold with boundary $\bar{M}$, a smooth defining function $\mathrm{\Omega }$ is positive on $M$ and vanishes with nonzero differential at the boundary, and $\bar{g}={\mathrm{\Omega }}^{2}g$ extends to a smooth nondegenerate metric there. Future null infinity ${\mathcal{I}}^{+}$ is the future null boundary component, where ${\bar{g}}^{-1}(d\mathrm{\Omega },d\mathrm{\Omega })=0$; the portion under study is $\mathrm{\partial }{J}^{-}({\mathcal{I}}^{+})\cap M$, where $J^{-}({\mathcal{I}}^{+})$ is the set of points from which a future-directed causal curve reaches ${\mathcal{I}}^{+}$. The local semiclassical equation alone does not supply this global structure.

Use length coordinates $x^0=ct$ and an affine parameter $\lambda$ with units of metres; the model fixes its origin and positive normalization. On the specified common generator patch, supply a nonempty compact boundaryless smooth two-manifold $C$, an open interval $J\subset \mathbb{R}$, and a smooth embedding $\mathrm{\Psi }:J\times C\to \mathcal{H}$ such that ${\mathrm{\partial }}_{\lambda }\mathrm{\Psi }=k\circ \mathrm{\Psi }$ and every map ${\mathrm{\Psi }}_{\lambda }:C\to \mathcal{H}$ is an embedding with spacelike image. A cut ${\mathrm{\Sigma }}_i$ is the image of the graph of a smooth function ${\lambda }_i:C\to J$ under $\mathrm{\Psi }$, with positive-definite induced metric; thus it is a nonempty compact smooth spacelike two-surface and meets each generator of this patch exactly once. The cuts ${\mathrm{\Sigma }}_1$ and ${\mathrm{\Sigma }}_2$ are ordered when ${\lambda }_2(x)>{\lambda }_1(x)$ for every $x\in C$. This order is preserved by every positive affine reparametrization of the generators.

At each finite ultraviolet regulator ${\mathrm{\Lambda }}_n$, whose type, units, and ultraviolet-limit convention are specified by the model, use the finite generalized-entropy data specified by the selected finite-regulator definition on each cut, with the same model, state, finite local counterterms, and fixed scheme. Each $S_{\mathrm{g}\mathrm{e}\mathrm{n},n}[{\mathrm{\Sigma }}_i]$ is a finite real entropy with units $\mathrm{J}{\mathrm{K}}^{-1}$. For a regulator sequence $\mathcal{R}=({\mathrm{\Lambda }}_n,\text{finite-regulator data})$, define

${\mathrm{\Delta }}_{\mathcal{R},n}({\mathrm{\Sigma }}_1,{\mathrm{\Sigma }}_2):=S_{\mathrm{g}\mathrm{e}\mathrm{n},n}[{\mathrm{\Sigma }}_2]-S_{\mathrm{g}\mathrm{e}\mathrm{n},n}[{\mathrm{\Sigma }}_1]\mathrm{.}$

An admissible sequence has ${\mathrm{\Lambda }}_n\to +\mathrm{\infty }$, uses finite-regulator data from this same selected solution and state, and applies the one fixed covariant prescription at both cuts. A sufficient criterion must specify a nonempty admissible class containing at least two regulator constructions; regulator independence means that the finite limit of ${\mathrm{\Delta }}_{\mathcal{R},n}$ is identical for every sequence in that class. The target does not require the two absolute sequences $S_{\mathrm{g}\mathrm{e}\mathrm{n},n}[{\mathrm{\Sigma }}_i]$ to converge separately.

Determine checkable sufficient conditions on the specified model, solution, state, ordered cuts, and admissible regulator class under which, for every such pair of later cuts,

${\mathrm{\Delta }}_{\mathcal{R},n}({\mathrm{\Sigma }}_1,{\mathrm{\Sigma }}_2)⟶\mathrm{\Delta }({\mathrm{\Sigma }}_1,{\mathrm{\Sigma }}_2)\in \mathbb{R},\mathrm{\Delta }({\mathrm{\Sigma }}_1,{\mathrm{\Sigma }}_2)\ge 0,$

The finite real value $\mathrm{\Delta }$ is expressed in SI entropy units $\mathrm{J}{\mathrm{K}}^{-1}$.

The same finite $\mathrm{\Delta }$ must be obtained for every admissible regulator sequence. The deduction must not assume a finite modular-geometric balance postulate or any other identity equating a modular-energy change to a gravitational-entropy or area change.

The finite comparison theorem assumes nonzero finite-dimensional Hilbert spaces $H_1=H_2\otimes H_E$, density operators ${\rho }_i,{\sigma }_i$ with faithful ${\sigma }_i$, and ${\rho }_2={\mathrm{Tr}}_E{\rho }_1$, ${\sigma }_2={\mathrm{Tr}}_E{\sigma }_1$. For such finite density operators, $D(\rho \mathrm{\parallel }\sigma ):=\mathrm{tr}(\rho (\log \rho -\log \sigma ))$ uses spectral logarithms and the convention $0\log 0=0$; $\mathrm{tr}$ is the finite-dimensional trace. Set ${\delta }_i:=\mathrm{tr}(({\rho }_i-{\sigma }_i)(-\log {\sigma }_i))$. Its adopted balance premise is $a_2-a_1={\delta }_1-{\delta }_2$, where $a_i$ are model-supplied geometric entropy data in nats, and it proves $S_{\mathrm{g}\mathrm{e}\mathrm{n},2}-S_{\mathrm{g}\mathrm{e}\mathrm{n},1}=k_B[D({\rho }_1\mathrm{\parallel }{\sigma }_1)-D({\rho }_2\mathrm{\parallel }{\sigma }_2)]\ge 0$. Its limit clause assumes both endpoint entropy sequences already converge finitely in one fixed scheme. This is a comparison point only; it establishes neither continuum convergence, regulator independence, semiclassical backreaction, nor the sufficient conditions requested here.

Semiclassical equation with fixed counterterms and state

Hadamard and renormalized stress contract

Specified late-time free-field collapse scattering model

Smooth null screen and subextremal Kerr–Newman parameter family

Specified regulated generalized entropy

Nested finite regulator and gravitational balance premises

Regulated generalized entropy monotonicity under explicit balance

Aron C. Wall, A proof of the generalized second law for rapidly changing fields and arbitrary horizon slices

H. S. Reall, Part III Black Holes lecture notes

Ted Jacobson, Introductory Lectures on Black Hole Thermodynamics

Thomas Faulkner and Antony J. Speranza, Gravitational algebras and the generalized second law

Josh Kirklin, Generalised second law beyond the semiclassical regime

Jonah Kudler-Flam, Samuel Leutheusser, and Gautam Satishchandran, Generalized black hole entropy is von Neumann entropy

Mohd Ali and Vardarajan Suneeta, A local Generalized second law in crossed product constructions

Weizhen Jia, Qiongyu Qi, and Christina Gao, Entanglement entropy and thermodynamics of dynamical black holes

Ahmed Almheiri et al., The entropy of Hawking radiation

Phil Saad, Stephen H. Shenker, and Douglas Stanford, JT gravity as a matrix integral

Adopted JT disk spectral model

Status checked on 22 October 2026. This is an open problem, not a theorem or an empirical claim. The review is targeted and non-exhaustive. Here the generalized second law (GSL) means nondecrease of the cited source's cut entropy $S_{\mathrm{g}\mathrm{e}\mathrm{n}}=A\mathrm{/}(4G_N)+S_{\mathrm{o}\mathrm{u}\mathrm{t}}$ in its natural units, where $A$ is cut area and $S_{\mathrm{o}\mathrm{u}\mathrm{t}}$ is the entropy of exterior quantum fields. A horizon algebra here is a unital complex *-subalgebra $\mathcal{A}\subseteq \mathcal{B}(\mathcal{H})$ of bounded operators on a Hilbert space $\mathcal{H}$, closed in the weak operator topology and assigned to the exterior of a specified cut; the cited proof requires the stated nested inclusions. “Semiclassical” means quantum matter on a classical metric with backreaction governed at the stated order by the adopted effective equation. Wall proves a conditional GSL under these horizon-algebra and semiclassical hypotheses, together with compatible renormalized-entropy prescriptions and decomposition/limit conditions; those hypotheses are not silently transferred to the fully specified coupled collapse problem above.

For the cited operator-algebra results, a von Neumann algebra is a unital weak-operator-closed, adjoint-closed algebra of bounded operators on a Hilbert space. A state is a positive normalized linear functional; it is faithful if it vanishes on no nonzero positive operator and normal if it is ultraweakly continuous. For a density operator $\rho$ (a positive trace-class operator of trace one on a Hilbert space), its von Neumann entropy is $-\mathrm{Tr}(\rho \log \rho )$, with $0\log 0:=0$ when the trace is defined. A Type II factor is a von Neumann algebra with scalar center, no minimal projections, and a faithful normal semifinite trace; Type II${}_{\mathrm{\infty }}$ has infinite trace of its identity.

A closer result is Kirklin's exact modified GSL in perturbative gravity, meaning his specified gravitational degrees of freedom are quantized around a classical background. An ideal dynamical quantum-reference-frame (QRF) cut in Kirklin's model has an auxiliary Hilbert factor $H_a=L^2(\mathbb{R},d{v}_a)$ for each cut. Its position operator ${\widehat{v}}_a$ is multiplication, $({\widehat{v}}_a\psi )(v_a)=v_a\psi (v_a)$, on domain $\{\psi \in L^2(\mathbb{R}):v_a\psi \in L^2(\mathbb{R})\}$; $v_a$ is the affine horizon coordinate. The conjugate translation generator is ${\widehat{k}}_a=-i{\mathrm{\partial }}_{v_a}$ on the domain $\{\psi \in L^2(\mathbb{R}):{\mathrm{\partial }}_{v_a}\psi \in L^2(\mathbb{R})\text{ as a weak derivative}\}$, with $e^{-is{\widehat{k}}_a}\psi (v_a)=\psi (v_a-s)$. The source's self-adjoint boost/dilation generator is ${\widehat{d}}_a:=\pi ({\widehat{v}}_a{\widehat{k}}_a+{\widehat{k}}_a{\widehat{v}}_a)$, defined as the closure from the dense space $\{\psi \in C^{\mathrm{\infty }}(\mathbb{R}):{\sup }_{v\in \mathbb{R}}\mathrm{\mid }{v}^m{\psi }^{(n)}(v)\mathrm{\mid }<\mathrm{\infty }\text{ for all }m,n\in {\mathbb{N}}_0\}$. For an earlier cut $S_b$ and a later cut $S_a$, define $H_{S_a\mathrm{\mid }{S}_b}:={\widehat{d}}_a-2\pi {\widehat{v}}_b{\widehat{k}}_a$, the later-cut boost generator. The exterior region is relationally defined by this quantum location, not by a fixed classical surface. Perturbative gravity here means that the source's gravitational degrees of freedom are quantized order by order about a classical background. For the outside-region von Neumann entropies $S(N)$, where $S_{\mathrm{v}\mathrm{N}}(\rho )=-\mathrm{Tr}(\rho \log \rho )$ when the state has density operator $\rho$, his exact inequality is $S(N_a)\ge S(N_b)+F(S_a\in N_b),F(S_a\in N_b)=⟨H_{S_a\mathrm{\mid }{S}_b}{⟩}_{\mathrm{\Phi }}-S_{S_a\in N_b}[\mathrm{\Phi }]\mathrm{.}$ Here $N_a\subset N_b$ are the exterior regions of the later and earlier cuts; ${\mathcal{A}}_{N_a}^G$ and ${\mathcal{A}}_{N_b}^G$ are the source's gauge-invariant von Neumann algebras assigned to them. The later-cut subalgebra ${\mathcal{A}}_{S_a\in N_b}^G$ is Type I; for $\mathrm{\Phi }$, let ${\rho }_a$ be its density operator relative to the source trace ${\mathrm{Tr}}_{S_a\in N_b}$. Thus $S_{S_a\in N_b}[\mathrm{\Phi }]:=-{\mathrm{Tr}}_{S_a\in N_b}({\rho }_a\log {\rho }_a)$. The source's $H_{S_a\mathrm{\mid }{S}_b}$ is the generator defined above, and $F(S_a\in N_b):=⟨H_{S_a\mathrm{\mid }{S}_b}{⟩}_{\mathrm{\Phi }}-S_{S_a\in N_b}[\mathrm{\Phi }]$. In the UV-cutoff semiclassical regime, Kirklin obtains only an approximate leading-order reduction to generalized entropy, up to a state-independent constant; the ordinary GSL follows only with further cut conditions giving $F\ge 0$, for example, a later-cut density operator ${\rho }_a=\mathrm{\mid }\psi ⟩⟨\psi \mathrm{\mid }$ with $\mathrm{\parallel }\psi \mathrm{\parallel }=1$ has zero cut entropy, and the additional hypothesis $⟨H_{S_a\mathrm{\mid }{S}_b}{⟩}_{\mathrm{\Phi }}\ge 0$ then gives $F\ge 0$. A product state with a mixed cut factor does not suffice. The source obtains the sign only at retained leading order in a classical-cut approximation using sharply peaked dressed relative position and a nonnegative mean dressed null-translation generator for the cut; these conditions are additional to the UV-cutoff semiclassical reduction and do not bound discarded corrections. This does not establish regulator independence for the finite-regulator definition here or identify its local fixed-scheme limit with Kirklin's Type II algebra entropy. His ideal cuts are auxiliary, and he leaves other symmetries and more realistic cut models for further work.

Reall and Jacobson give broad GSL discussions while recording backreaction, renormalization, and proof-status limits. Let $(M,g)$ be a smooth Lorentzian manifold with signature $(-,+,\dots ,+)$ and Levi-Civita connection $\mathrm{\nabla }$, the unique torsion-free, metric-compatible connection. A smooth vector field $\xi$ is Killing when ${\mathrm{\nabla }}_a{\xi }_b+{\mathrm{\nabla }}_b{\xi }_a=0$, equivalently ${\mathcal{L}}_{\xi }g=0$, where ${\xi }_b:=g_{bc}{\xi }^c$. A Killing horizon is a smooth embedded codimension-one null hypersurface $H\subset M$ whose induced metric has a one-dimensional radical, on which $\xi$ is null, tangent, and normal, with $\xi$ nonzero except possibly on a bifurcation surface $B$. A bifurcate Killing horizon is a pair of such hypersurfaces meeting at a spacelike codimension-two surface $B$ where $\xi =0$ and is nonzero on each horizon away from $B$. For a faithful normal state $\omega$ on $\mathcal{A}$, let $({\pi }_{\omega },{\mathcal{H}}_{\omega },{\mathrm{\Omega }}_{\omega })$ be its GNS representation, obtained by completing $\mathcal{A}\mathrm{/}{\mathcal{N}}_{\omega }$ for ${\mathcal{N}}_{\omega }:=\{a\in \mathcal{A}:\omega (a^{\ast }a)=0\}$ with inner product $⟨[a],[b]⟩:=\omega (a^{\ast }b)$, representation ${\pi }_{\omega }(c)[b]:=[cb]$, and cyclic vector ${\mathrm{\Omega }}_{\omega }:=[1]$, let $S_{\omega }$ be the closure of the antilinear map ${\pi }_{\omega }(a){\mathrm{\Omega }}_{\omega }\mapsto {\pi }_{\omega }(a^{\ast }){\mathrm{\Omega }}_{\omega }$, and put ${\mathrm{\Delta }}_{\omega }=S_{\omega }^{\ast }{S}_{\omega }$. Its modular automorphism group is defined by ${\pi }_{\omega }({\sigma }_t^{\omega }(a))={\mathrm{\Delta }}_{\omega }^{it}{\pi }_{\omega }(a){\mathrm{\Delta }}_{\omega }^{-it}$. For this modular action, on the Hilbert space $L^2(\mathbb{R},ds;{\mathcal{H}}_{\omega })$ of square-integrable ${\mathcal{H}}_{\omega }$-valued functions, define $(\mathrm{\Pi }(a)f)(s):={\pi }_{\omega }({\sigma }_{-s}^{\omega }(a))f(s)$ and $(U(t)f)(s):=f(s-t)$. The crossed product $\mathcal{A}{⋊}_{{\sigma }^{\omega }}\mathbb{R}$ is the weak-operator-closed unital $\ast$-algebra generated by $\mathrm{\Pi }(\mathcal{A})$ and these $U(t)$ for $t\in \mathbb{R}$; this regular representation satisfies $U(t)\mathrm{\Pi }(a)U(t{)}^{\ast }=\mathrm{\Pi }({\sigma }_t^{\omega }(a))$. The cited gravitational constructions also include global gravitational charges and their constraints. “Modular flow is geometric on the horizon” means that this automorphism group acts on the horizon-cut observables as the specified geometric flow there. Faulkner–Speranza derive a GSL for arbitrary cuts of a Killing horizon when this condition and their semiclassical state hypotheses hold; they do not establish monotonicity after their second-order entropy corrections. Kudler-Flam–Leutheusser–Satishchandran prove their Type II algebra structure result for a linear Klein–Gordon field, a bifurcate Killing horizon, a stationary state (invariant under the Killing flow, not necessarily thermal), and suitable asymptotic decay. Ali–Suneeta treat maximally extended static and Kerr settings and assume gravitational constraints on each partial Cauchy slice, here an achronal spacelike hypersurface segment that every causal curve meets at most once. Jia–Qi–Gao work to first order around stationary backgrounds. In their $f(R)$ model, where $f$ is the gravitational Lagrangian as a function of scalar curvature $R$, on a region where $f^{\mathrm{\prime }}(R)>0$, the generalized apparent horizon is foliated by cuts on which ${\tilde{\mathrm{\Theta }}}_{\tilde{k}}:={\tilde{\theta }}_{\tilde{k}}+{\tilde{k}}^a{\mathrm{\nabla }}_a\log f^{\mathrm{\prime }}(R)=0$, with $\tilde{k}$ the chosen outward null normal and ${\tilde{\theta }}_{\tilde{k}}$ its fractional area-density rate. Their entropy identification is first order and only the leading local area-law term is identified with renormalized generalized entropy.

In the cited island formula, an island is a disconnected bulk region included with radiation in the semiclassical field entropy; its boundary $X$ is chosen by extremizing and then minimizing $\mathrm{Area}(X)\mathrm{/}(4G_N)+S_{\mathrm{s}\mathrm{e}\mathrm{m}\mathrm{i}}({\mathrm{\Sigma }}_{\mathrm{R}\mathrm{a}\mathrm{d}}\cup {\mathrm{\Sigma }}_{\mathrm{I}\mathrm{s}\mathrm{l}\mathrm{a}\mathrm{n}\mathrm{d}})$ in the source's natural units. The replica trick computes $S=-{\mathrm{\partial }}_n\log \mathrm{tr}({\rho }^n){\mathrm{\mid }}_{n=1}$ from the path integral on $n$ copies. A replica saddle is a stationary (first variation of the action equal to zero), possibly complex, geometry contributing to the chosen replicated gravitational path integral; a replica-wormhole saddle connects interiors of different copies. The cited review leaves the integration contour, included saddles, cutoff treatment, and relation to the exact state unresolved. JT means two-dimensional Jackiw–Teitelboim gravity. The cited matrix integral matches its formal genus expansion, a sum of contributions indexed by the number of handles, through an ensemble (a probability law on matrices whose observables are averaged). A nonperturbative completion means additional exact model data beyond that formal expansion; the completion is not unique, and the disk supplier here specifies only a genus-zero continuous density, not an individual Hamiltonian. The sources reviewed do not establish the exact combined claim posed here; this bounded review does not establish that no other proof exists.