Gauss Law, Higher Categorical Coherence, and the Island Resolution of the Black Hole Information Paradox

Gauss Law, Higher Categorical Coherence, and the Island Resolution of the Black Hole Information Paradox Andrei T. Patrascu 1 1 FAST Foundation, Destin FL, 32541, USA email: andrei.patr[email protected] We propose a unified perspective on gauge constraints, the measurement problem, and the black hole information paradox. In ordinary gauge theories, Gauss’s law enforces a rigid entanglement structure across spatial boundaries: the algebra of observables in a region A ( R )acquires a nontrivial center Z ( A ( R )), whose spectrum labels superselection sectors (boundary charges or fluxes). Locally, these labels behave as classical random variables, while globally they appear as entanglement indices sewing together complementary regions via a categorical pullback condition. This dual classical/quantum role of boundary data underlies both the operational emergence of spacetime causal propagation and the “island” prescription in black hole entropy computations, where interior and exterior algebras are glued along their common center. We extend this structure using the framework of higher categorical coherence breakdown. At the triangle level, central extensions reproduce the familiar Heisenberg commutation relations and guarantee unitarity. At higher levels, however, noncentral extensions appear, leading to operators that act nontrivially on the center itself. Physically, this promotes the Gauss–law superselection labels from immutable classical data to dynamical quantum channels, enabling “self–measurement” of the spacetime fabric. We show that such deformations naturally generalize the algebraic equalizer underlying the island rule to a weak or homotopy equalizer, and argue that this provides a microscopic mechanism for information recovery in black hole evaporation. In this view, the black hole information paradox, the measurement problem, and the emergence of spacetime structure are interwoven facets of a single categorical principle: gauge is quantum, and coherence breakdown governs the transition between classical superselection and dynamical self–measurement of spacetime itself. I. INTRODUCTION A. Overview and main claims This paper develops a unified, operator-algebraic and categorical account of (i) the Gauss-law constraints in gauge theories and their entanglement consequences, (ii) the appearance of classical superselection labels as spectra of local centers together with their quantum role as entanglement indices, (iii) the island prescription for black-hole entropy as a categorical gluing (equalizer/pullback) along those centers, and (iv) a new framework—higher categorical coherence breakdown—in which the rigid Gauss-law sewing becomes dynamical, thereby providing a microscopic mechanism for information flow and, potentially, for the resolution of the black-hole information paradox. At a technical level, the paper rests on the following pillars: 1. For a regular spatial region R in a gauge theory, the gauge-invariant local algebra A ( R )generally has a nontrivial center Z ( A ( R )) generated by boundary charges/fluxes (edge modes) [ 82 , 83 , 117 ]. The inclusion of a boundary flux operator Φ∂R =Z∂R E·dS into Z ( A ( R )) codifies Gauss’s law ∇· E = ρ as an operator identity. Interior, gauge-invariant observables commute with Φ∂R. 2. The commutative von Neumann algebra Z ( A ( R )) ∼ =L∞ ( XR, µR )has spectrum XR that labels superselection sectors (boundary data). The central (direct-integral) decomposition A(R)∼ =Z⊕ XR Ax(R)dµR(x), ωA(R)=ZXR ωxdµω R(x) (von Neumann–Dixmier–Mackey decomposition) yields an intrinsic split into classical randomness over XRand quantum states in the factor fibers Ax(R)[2, 3, 114]. 3. If a Cauchy slice Σis decomposed as Σ = R∪Rc with common boundary ∂R , then the global physical algebra is the pullback (equalizer) over the boundary center, A(Σ) ∼ =A(R)×Z(A(∂R)) A(Rc) = {(a, b)|φR(a) = φRc(b)},(1) 2 where φR and φRc “report” the boundary labels from inside/outside. Equation (1) is Gauss’s law as a categorical equalizer constraint. 4. The same pullback mechanism underlies the island formula for black-hole entropy: the relevant radiation algebra must be extended by gluing to an “island” algebra along a diffeomorphism/gauge center on the quantum extremal surface (QES) [9–11]. 5. We introduce higher categorical coherence breakdown as a controlled deformation of the operator-algebraic structure. At the triangle (2-simplex) level, only central extensions arise (projective/Heisenberg-type), preserving unitarity [ 13 , 86 ]. At higher levels (pentagon, hexagon), noncentral extensions appear; these permit interior operators to act on the center, thus mixing superselection labels. Algebraically this can be modeled by crossed-product (unitary) or completely-positive (CP/Lindbladian) deformations of the boundary/edge algebra [ 14 , 15 , 121 ]. In this regime, the Gauss equalizer (1) is replaced by a weak/homotopy equalizer in a noncommutative or 2-categorical sense. Notation. We fix a Cauchy slice Σ(smooth 3-manifold or lattice) and a regular bounded region Rb Σ with boundary ∂R . The gauge-invariant von Neumann algebra of strictly interior observables is A ( R ); its center is Z ( A ( R )). We write XR := Spec Z ( A ( R )) and identify Z ( A ( R )) ∼ =L∞ ( XR, µR ). The global algebra on Σis A (Σ). For finite group/lattice models we use direct sums; for continuum models, direct integrals. B. Gauss’s law as a central constraint: algebraic derivations We recall the canonical form of Gauss’s law and its representation in the local algebra. Definition I.1 (Gauss’s operator constraint) . Let E be the electric-field operator and ρ the charge density. For any real test function fsupported in an open set O⊂Σ, define G(f) := ZO f(x)∇· E(x)−ρ(x)d3x. The physical subspace Hphys is the joint kernel G ( f ) ψ = 0 for all f ; the physical algebra A ( O )is the commutant of the G(f)within the field algebra restricted to O. Integrating by parts with f≡1on Rand supported in a slightly larger region, one formally obtains Φ∂R := Z∂R E·dS=ZR ρ d3x=: QR.(2) Proposition I.2 (Centrality of boundary flux) . For any strictly interior, gauge-invariant operator O∈ A(R), we have [ Φ∂R, O ]=0. Hence Φ∂R ∈Z(A(R)). Proof. Let χ be a smooth cutoff equal to 1on R and supported in a slightly larger region R . Then G ( χ ) = RRχ ( ∇· E−ρ ) = Φ ∂R −QR by (2) . For any O∈ A ( R )one has [ G ( χ ) , O ] = 0 since O commutes with all constraints supported where O lives. Therefore [Φ ∂R −QR, O ]=0. But QR is the integral of ρ over the support of O and thus also commutes with O (they are both gauge-invariant and strictly supported in R). Hence [Φ∂R, O] = 0. Non-Abelian and lattice variants. For a finite group gauge theory on a lattice, boundary magnetic data are captured by the holonomy conjugacy class C⊂G around ∂R ;electric data are irreducible representations α of the centralizer of a representative g∈C . Gauge-invariant functions of these boundary data generate Z ( A ( R )) [ 82 ]. The proof of centrality follows from the commutation of interior gaugeinvariant operators with vertex/plaquette constraints and the fact that changing boundary data requires an operator whose support reaches ∂R (open Wilson line). 3 C. Central (direct-integral) decomposition and superselection Because Z ( A ( R )) is commutative, it is (isomorphic to) L∞ ( XR, µR ). The standard decomposition theorem is: Theorem I.3 (Direct-integral decomposition).Let πbe a normal representation of A(R)on H. There exist a measure space ( XR, µR )and a measurable field {Ax ( R ) ,Hx}x∈XR of factor von Neumann algebras on Hilbert spaces such that π(A(R))00 ∼ =Z⊕ XR Ax(R)dµR(x),H∼ =Z⊕ XR HxdµR(x), and Z ( π ( A ( R )) 00 ) ∼ =L∞ ( XR, µR ) ·1 acts by multiplication. Any normal state ω on π ( A ( R )) 00 decomposes as ω(O) = ZXR ωx(O)dµω R(x). Sketch. See [ 2 , 3 , 114 ]. One constructs the maximal abelian subalgebra generated by the center, diagonalizes it via the spectral theorem to obtain ( XR, µR ), and disintegrates the representation accordingly. Factoriality of the fibers follows from maximality of the center and the relative commutants. Corollary I.4 (Operational superselection) . If f∈L∞ ( XR ) ⊂Z ( A ( R )) and O∈ A ( R ), then [ f, O ] = 0. Thus no O∈ A ( R )connects different x∈XR sectors, and any reduced state on A ( R )is operationally indistinguishable from a classical mixture over XR. Proof. Immediate, since facts as multiplication on R⊕HxdµRand Oacts fiberwise. Entropy split (lattice). In a finite lattice region where each fiber Ax ( R )admits a density matrix ρx , the reduced density matrix is block-diagonal: ρR=Lxpxρx, with px=µω R({x}). Then S(ρR) = H({px}) + X x pxS(ρx), where His Shannon entropy (“edge” or “center” contribution). D. Categorical gluing: equalizers, pullbacks, and Gauss’s law Let R⊂Σbe complemented by Rcwith common boundary ∂R. There are canonical boundary maps φR:A(R)→Z(A(∂R)), φRc:A(Rc)→Z(A(∂R)), that compute the boundary label (e.g. flux) from either side. Definition I.5 (Equalizer/pullback gluing).The glued algebra is the pullback A(R)×Z(A(∂R)) A(Rc) := {(a, b)∈ A(R)⊕ A(Rc)|φR(a) = φRc(b)}. Proposition I.6 (Gauss equalizer).The global physical algebra on Σobeys A(Σ) ∼ =A(R)×Z(A(∂R)) A(Rc). Idea. The full field algebra on Σsubject to Gauss constraints G ( f ) = 0 is generated by interior observables from R and Rc modulo relations ensuring consistent boundary data. The universal property of the pullback identifies exactly the subalgebra where boundary reports agree. (See [ 117 ] for Haag duality/gluing principles and [82, 83] for edge-mode implementations.) Functoriality and naturality. The local net R7→ A ( R )is a covariant functor from the inclusion poset of regions to the category of von Neumann algebras ( vNAlg ). Heisenberg time evolution αt : A ( R ) → A ( R ) is a natural transformation: αS t◦ιR⊂S = ιR⊂S◦αR t . Since αt preserves constraints, it preserves the center and acts fiberwise on the direct integral decomposition. 4 E. Propagation and causal structure on the Gauss scaffold In Coulomb gauge, the field splits into longitudinal (constraint) and transverse (propagating) parts: A=AT,E=ET+EL,∇· EL=ρ. The Hamiltonian H=1 2Z(E2 T+B2)d3x+1 2ZZ ρ(x)ρ(y) 4π|x−y|d3x d3y+eZψ†α·ATψ d3x propagates only the transverse modes; the longitudinal part is slaved to Gauss’s law. Microcausality (vanishing of commutators at spacelike separation) holds for gauge-invariant local fields and, for potentials, for the transverse components with Pauli–Jordan support on the light cone. Conceptually: the global continuity enabling propagation is enforced kinematically by the Gauss equalizer; the dynamics then propagate disturbances within the fiber algebras. F. The measurement problem revisited: labels as pointers Corollary I.4 shows that the reduced state on A ( R )is operationally a classical mixture over XR . From the interior perspective, the spectrum XR behaves as a classical pointer basis: no interior observable can reveal phases between different labels. Globally, however, the state is pure and exhibits entanglement across the cut; the boundary label is the Schmidt index. This intrinsic classicalization, coming from constraints rather than an external environment, rephrases the measurement problem in strictly algebraic terms [5, 6, 117]. G. Higher categorical coherence breakdown: from central to noncentral extensions We outline a deformation framework in which the rigid classicalization is relaxed. Triangle level (2-cocycles) and unitarity. Projective representations of symmetry groups are classified by H2 ( G, U (1)); the associated central extensions lead to the Heisenberg group and canonical commutation relations (CCR). In categorical terms, the associator obeys Mac Lane’s coherence, and only central phases appear [13, 86]. Unitary dynamics are preserved; centers remain immutable. Higher levels (3-cocycles and beyond). Relaxing coherence (e.g. weakening the pentagon) allows nontrivial associators valued in noncentral elements (quasi-Hopf/weak structures) [ 14 ]. In an operatoralgebraic incarnation, we capture two minimal effects: •Unitary label mixing (crossed product). Let Z ( A ( R )) ∼ =L∞ ( XR )and let T : XR→XR be a measure-class-preserving map (e.g. flux shift q7→ q+ 1). Adjoin a unitary Uimplementing T: UfU =f◦T−1, U Ax(R)U=AT x(R). The deformed algebra Amix ( R ) := R⊕AxoTZ has Z ( Amix ( R )) = L∞ ( XR ) Z , which reduces to C if T is ergodic. Then [ Φ ∂R, U ] = U , so interior operators can raise flux: superselection collapses. •Nonunitary “self-measurement” (CP/Lindbladian). Introduce boundary-local CP maps that mix labels via a Markov kernel pt ( x0|x )and intertwiners between fibers Ax ( R ) → Ax0 ( R ). The Heisenberg evolution acts as a Lindbladian on the center [121]: d dt ft(x) = X x0pt(x|x0)ft(x0)−pt(x0|x)ft(x), turning the formerly static edge distribution into a dynamical variable; fiber coherences may be created/damped depending on dilation. Categorical consequence. The commutative base Z is replaced by a noncommutative (crossed-product) or 2-categorical (CP) boundary object, and the strict pullback (1) is replaced by a weak/homotopy equalizer compatible with the new boundary dynamics. Intuitively, the “glue” of spacetime—the Gauss sewing—becomes a quantum/dynamical channel. 5 H. Black holes, islands, and information flow In semiclassical gravity, Hawking radiation appears thermal; the paradox arises from assuming a naive tensor factorization between exterior and interior. Gauge/diffeomorphism constraints invalidate this factorization; dressed operators tie excitations to asymptotic charges [ 9 ]. The island formula corrects the algebra by gluing an interior (island) algebra along the QES center, reproducing the Page curve [ 10 , 11 ]. In our framework, this is precisely a pullback over the appropriate center. Under higher coherence breakdown, the boundary center can become dynamical: • With unitary mixing, the dressing sector can coherently transport information (center becomes a quantum bus). • With CP mixing, the edge distribution evolves stochastically (self-measurement), providing channels for information leakage consistent with locality when mixing is boundary-local. Either way, the “equalizer” underpinning the island rule generalizes to a weak/homotopy equalizer, offering a microscopic mechanism for information recovery consistent with gauge/diffeo constraints. I. Roadmap of the paper Section I (this section) set the narrative and core constructs. Section 2 reviews local nets and Gauss constraints, proving centrality of boundary fluxes in several models. Section 3 develops the direct-integral decomposition and operational superselection. Section 4 formulates Gauss’s law as an equalizer/pullback and introduces the base/total category and fibration viewpoint. Section 5 relates the Gauss scaffold to propagation and causality. Section 6 revisits the measurement problem algebraically. Section 7 presents higher categorical coherence breakdown and the two deformations (crossed-product and CP). Section 8 interprets these as “self-measurement of spacetime”. Section 9 develops the algebraic and categorical tools in preparation for gravity. Section 10 analyzes modular theory, KMS structure, and Connes invariants in the presence of nontrivial centers. Section 11 illustrates the framework in concrete models: lattice U (1), finite non-Abelian groups, and continuum Maxwell theory. Section 12 formulates the categorical semantics via the Grothendieck fibration and homotopy equalizers. Section 13 applies the full framework to black holes and islands, deriving the QES and JLMS formulae in direct-integral form. Section 14 extends the analysis to operational diagnostics, information-theoretic capacities, condensed-matter analogues, and gravitational predictions. Section 15 provides a broad discussion and outlook, synthesizing implications for quantum information, condensed matter, and gravity. Section 16 concludes with a summary of results and a set of open problems. The appendices collect operator-algebraic preliminaries, modular/KMS details, lattice gauge theory calculations, CP semigroup dilations, categorical background, and deferred proofs with a notation index. II. GAUGE THEORIES, GAUSS’S LAW, AND LOCAL ALGEBRAS In this section we place the Gauss-law constraint and its consequences on rigorous algebraic footing. We proceed in three steps. First, we recall the local net of von Neumann algebras and fix notation (§II A). Second, we review canonical quantization with constraints and exhibit Gauss’s law as an operator identity (§II B). Third, we prove that the boundary flux/holonomy operators belong to the center of the strictly interior algebra A ( R )and illustrate this both in the continuum and on the lattice, including non-Abelian generalizations (§II C). Throughout, we interleave the physics intuition with the mathematical statements. A. Local nets and basic structures Let Σbe a (fixed) Cauchy slice (e.g. R3 or a compact 3-manifold). We consider a directed set Reg of regular regions Rb Σwith smooth piecewise boundaries ∂R , partially ordered by inclusion. A (Haag–Kastler) local net is a covariant functor A: Reg −→ vNAlg, R 7−→ A(R)⊂B(H) into the category of von Neumann algebras on a common Hilbert space H, satisfying: 6 1. Isotony: R1⊂R2⇒ A(R1)⊂ A(R2). 2. Locality (Einstein causality): If R1 and R2 are spacelike separated (when embedded in spacetime), then [A(R1),A(R2)] = 0. 3. Additivity: If R=SiRiwith Ridirected, then A(R)is generated by SiA(Ri). In non-gauge relativistic QFTs, each local algebra A ( R )is (under mild hypotheses) a factor (often of type III1 ), so its center is trivial, Z ( A ( R )) = C1 . In gauge theories, by contrast, Gauss-law constraints render Z(A(R)) nontrivial once Rhas boundary, and this will be our main focus. Field vs. observable algebra. We distinguish a field algebra F ( R ), generated (say) by canonical fields and their exponentials smeared with test functions supported in R , from the gauge-invariant observable algebra A ( R ) = F ( R ) G(R) , the fixed-point subalgebra under the (local) gauge group G ( R ). The physical Hilbert space Hphys is obtained by imposing first-class constraints (Gauss laws) following Dirac/Henneaux– Teitelboim [17, 18]. B. Gauss constraints and the physical subalgebra We now display Gauss’s law both in continuum canonical form and on the lattice, derive the integrated (boundary) version, and state its operator content. 1. Continuum, Abelian case (QED) Work in temporal gauge or Coulomb gauge so that A0 plays the role of a Lagrange multiplier imposing Gauss’s law. Let E and B be the electric and magnetic field operators, and ρ the charge density (e.g. ρ=e ψ†ψin QED). For any real test function f∈C∞ c(Σ) we define the Gauss operator G(f) := ZΣ f(x)∇· E(x)−ρ(x)d3x. (3) Physical vectors satisfy G ( f ) Ψ = 0 for all f . The gauge-invariant (local) observable algebra is the commutant of {G ( f ) }f in the field algebra; concretely, A ( R )is generated by gauge-invariant polynomials and exponentials of the field strengths Fµν and neutral matter bilinears smeared inside R. Let Rb Σbe regular with outward normal ˆn . Choose χ∈C∞ c (Σ) with χ≡ 1on R and supp χ⊂R slightly larger. Using (3) and integrating by parts, G(χ) = ZR χ(∇· E−ρ)d3x=Z∂R E·dS | {z } Φ∂R −ZR ρ d3x | {z } QR .(4) On Hphys,G(χ)=0; hence the operator identity Φ∂R =QR(5) holds on all physical states. Here Φ ∂R is the boundary flux operator. Equation (5) is the familiar divergence theorem in operator form. Proposition II.1 (Centrality of boundary flux for strictly interior observables) . If O∈ A ( R )is supported strictly in R, then [ Φ∂R, O ]=0. Equivalently, Φ∂R ∈Z(A(R)). Proof. By construction [ G ( f ) , O ]=0whenever supp f contains supp O . In particular, with f = χ above, [ G ( χ ) , O ] = 0. Using (4) ,[ Φ ∂R −QR, O ] = 0. But QR = RRρ is a (gauge-invariant) integral of fields supported where Olives, so [QR, O]=0. Hence [Φ∂R, O]=0. Physics intuition. Any gauge-invariant operator entirely inside Rcan only create/annihilate neutral excitations (e.g. charge–anticharge pairs whose flux lines close inside R , or local field-strength fluctuations). It cannot change the net charge (equivalently, the net flux through ∂R ). Equation (5) enforces this: to change Φ ∂R (and thus QR ) one needs an operator whose support reaches ∂R (e.g. an open Wilson line ending on ∂R), which is not in A(R)by definition. 7 2. Non-Abelian case: conjugacy classes and class functions For a compact non-Abelian gauge group G , Gauss’s law arises from the non-Abelian generalization of (3) ; the local gauge-invariant algebra A ( R )is generated by Wilson loops contained in R , local curvature operators, and neutral matter bilinears. The relevant boundary data are: •Amagnetic label: the conjugacy class C⊂Gof the boundary holonomy around ∂R. • An electric label: an irreducible representation α of the centralizer Zg := {h∈G : hg = gh} of a representative g∈C. Gauge-invariant class functions f : C7→ C of the boundary holonomy and Casimirs built from boundary electric fields generate Z ( A ( R )). Strictly interior gauge-invariant operators commute with these boundary operators for the same reason as in Proposition II.1: changing C or α requires support reaching ∂R (open Wilson lines/surfaces). 3. Lattice formulation On a cubic lattice Λ ⊂ Σ, links ` carry compact variables U`∈U (1) (or G ) and conjugate electric fields E`(for U(1), with integer spectrum). The kinematics are [20, 80]: U`=eiθ`,[E`, θ`0] = −i δ``0,[E`, U`0] = δ``0U`. Vertices vcarry matter (charges) with density ρv. Gauss’s law at a vertex is Gv:= X `3v s(v, `)E`−ρv= 0,(6) where s ( v, ` ) = ± 1encodes the link orientation relative to v . Let R be a connected subgraph; define the boundary flux as the oriented sum over boundary-crossing links: Φ∂R := X `⊥∂R E`.(7) Summing (6) over all vertices in R telescopes the interior link contributions and yields the lattice version of (5): Φ∂R =X v∈R ρv=: QR.(8) Proposition II.2 (Centrality on the lattice) . Let A ( R )be generated by Wilson loops fully contained in R , plaquette operators in R , and neutral matter bilinears on vertices of R . Then [ Φ ∂R, O ] = 0 for all O∈ A(R); i.e. Φ∂R ∈Z(A(R)). Proof. Each generator of A ( R )is a product of link variables along a closed path in R (or functions of plaquette holonomies) and matter operators localized in R but neutral. Their commutators with E` vanish for links ` crossing ∂R , since these generators have no support on boundary-crossing links. Therefore they commute with the sum (7) . Neutral matter bilinears commute with Φ ∂R by neutrality. Linearity and weak operator closure extend the result to all of A(R). Flux-raising by open lines (diagnostic). If γ is a path in Λfrom a vertex in R to a vertex on ∂R , the open Wilson line Wγ = Q`∈γU` is not an element of A ( R )(its support touches ∂R ), and it satisfies a discrete version of [ Φ∂R, Wγ] = Wγ,(9) i.e. it raises the boundary flux by one unit in U (1). Equation (9) explicitly exhibits why only nonlocal (boundary-touching) operators can change Φ∂R. C. Central elements induced by Gauss law: physics and proofs We now gather the conclusions and give a detailed proof of centrality in the continuum, including the commutator with a dressed charged operator, and state the non-Abelian generalization. 8 1. Continuum proof and commutator with dressings Let Ψ † γ ( x )be a gauge-invariant dressed charged operator creating a unit charge at x∈R dressed by a path γthat runs from xto the boundary ∂R (or to infinity), e.g. Ψ† γ(x) := ψ†(x)PexpiZγ:x→∂R A.(10) The path-ordered exponential compensates the local gauge transformation of ψ† so that Ψ † γ is gauge invariant (see e.g. [21, 22]). Lemma II.3 (Flux commutator with a dressed charge).Let Ψ† γ(x)be as in (10). Then [ Φ∂R,Ψ† γ(x) ] = Ψ† γ(x).(11) Sketch. Smear the electric field flux with a thin pillbox around ∂R and write Φ ∂R = R∂R ˆn·E . The commutator with the exponential Pexp ( iRγA )arrives from the canonical commutation relations [ Ei ( y ) , Aj ( y0 )] = i δij δ3 ( y−y0 )and the fact that γ crosses the boundary once with positive orientation. The matter factor ψ† ( x )commutes with Φ ∂R (they have disjoint supports). Together these yield (11) . A careful derivation uses smooth test functions approximating the characteristic function of R and the Baker–Campbell–Hausdorff formula for line integrals of A. Interpretation. Equation (11) shows: any gauge-invariant operator that changes the net charge in R must be nonlocal (supported along γ up to ∂R ) and necessarily changes the central boundary flux by one unit. Conversely, any strictly interior gauge-invariant operator (closed Wilson loops, neutral bilinears) has vanishing commutator with Φ∂R by Proposition II.1. This is the algebraic content of Gauss’s law. 2. Conservation and locality Proposition II.4 (Conservation of boundary flux label) . Let H be the gauge-invariant Hamiltonian (e.g. the Coulomb-gauge QED Hamiltonian). Then [H, Φ∂R ] = 0,(12) so the boundary flux label is conserved in time. Time evolution preserves the decomposition into flux sectors (no mixing by interior dynamics). Idea. In Coulomb gauge, H = HT + HC + Hint , where HT depends only on transverse fields, Hint couples transverse ATto the conserved current, and HCis a functional of ρ(the instantaneous Coulomb term). The longitudinal electric field EL is a functional of ρ (constraint). The commutator with the flux Φ ∂R vanishes term-wise: (i) HT has no EL ; (ii) Hint is built from transverse fields which do not contribute to Φ ∂R ; (iii) HC depends on ρ and commutes with Φ ∂R by (5) . A rigorous statement uses the constraint surface and gauge-invariant algebraic dynamics. Locality bound. Any operator supported in a collar neighborhood of ∂R can alter Φ ∂R ; operators strictly supported in the interior cannot. This demarcates the edge (boundary) vs. bulk roles: the edge carries the central label; the bulk algebra acts within sectors. 3. Non-Abelian lattice: class sums and centrality For a finite group G , the gauge-invariant algebra A ( R )contains class sums on the boundary: for each conjugacy class C⊂G, CC(∂R) := 1 |C|X g∈C δ(g∂R, g),(13) where g∂R is the boundary holonomy. These CC generate a commutative subalgebra isomorphic to class functions on G . Together with electric (representation) data on the boundary edges/vertices (irreps of the centralizer), they generate Z ( A ( R )). Strictly interior gauge-invariant operators commute with CC ( ∂R ) and with boundary electric Casimirs, because changing the conjugacy class or electric label requires support intersecting ∂R (open ribbon operators). This is the non-Abelian analog of Propositions XI.1 and II.1. 9 D. Intuitive narrative and summary Why centers appear. Gauss’s law is a global constraint on every subregion. It ties the bulk charge to a boundary flux/holonomy. Because strictly interior gauge-invariant operators cannot affect this tie, the boundary flux/holonomy operators commute with the entire A ( R ): they are central. Algebraically, this forces A(R)to have a nontrivial center generated by edge (boundary) observables. What the center means operationally. The spectrum of the center Z ( A ( R )) labels superselection sectors: no interior operation can coherently mix them. The reduced state on A ( R )therefore looks like a classical mixture over boundary labels; globally, however, the same labels serve as entanglement indices sewing Rto Rc(see Sections I and III). Propagation on the scaffold. The constraint fixes the longitudinal sector and enforces a rigid sewing across ∂R . Propagation (causal waves) lives in the transverse sector within each superselection sector; the Gauss-law center provides the kinematic scaffold that makes a single global field possible. Non-Abelian refinements. For non-Abelian groups, the magnetic (conjugacy-class) and electric (centralizer irrep) data at the boundary form the center; interior loop/ribbon operators cannot change these labels; only boundary-touching operators can. The class sums (59) act as central projectors onto magnetic sectors. What comes next. In Section III we pass from the existence of central elements to the full central decomposition (direct-integral of factor fibers), which makes precise the “classical over labels + quantum inside sectors” dictum and prepares the categorical gluing (equalizer/pullback) of Section IV. III. CENTRAL DECOMPOSITION AND SUPERSELECTION STRUCTURE In Section II we established that Gauss’s law forces the existence of central boundary observables in the strictly interior algebra A ( R ), and that strictly interior gauge-invariant operators cannot change the boundary label. In this section we explain in full mathematical detail how this yields a central (direct–integral) decomposition of A ( R )and of states on A ( R ), why this makes the reduced state on R classical over boundary labels but quantum within each sector, and how all interior observables act fiberwise. We also present explicit examples (discrete and continuous label spaces), a toy two-sector calculation highlighting the disappearance of interference terms, and entropy/relative entropy decompositions. We keep the notation introduced previously: Σis a Cauchy slice, Rb Σis a regular region with boundary ∂R , A ( R )is the gauge-invariant von Neumann algebra of strictly interior observables, and Z ( A ( R )) denotes its center. A. Commutative centers and spectra The center Z ( A ( R )) is a commutative von Neumann algebra. By the spectral theorem for commutative von Neumann algebras (Gelfand–Naimark, see e.g. [ 23 , 24 ]), there exists a (localizable) standard Borel measure space (XR, µR)such that Z(A(R)) ∼ =L∞(XR, µR)·1,(14) where 1 is the unit of A ( R ). We call XR = Spec ( Z ( A ( R ))) the boundary-label space. Central projections correspond to indicator functions 1E of measurable sets E⊂XR ; more generally, f∈L∞ ( XR )acts by multiplication. Physically, XR collects all gauge-invariant boundary data that are fixed by strictly interior operations (fluxes/holonomies and electric data, cf. Section II). Spectral measure from a state. Fix a normal state ω on A ( R ). For any measurable E⊂XR , let PE∈Z ( A ( R )) be the corresponding central projection (i.e. PE is the image of 1E under the isomorphism (14)). Then µω R(E) := ω(PE)(15) defines a probability measure on ( XR, µR )that is absolutely continuous with respect to (and equivalent to) the underlying measure class. Intuitively, µω Ris the classical distribution over boundary labels in the state ω. 16 D. Dynamics as a natural transformation Let αtdenote the (Heisenberg) time evolution automorphisms on the net. Definition IV.6 (Natural dynamics) . The dynamics αt is a natural transformation αt : A⇒A if for any inclusion i:R ,→Swe have the commutative square αS t◦ιR⊂S=ιR⊂S◦αR t. Lemma IV.7 (Center preservation and fiberwise action).If αtpreserves Gauss constraints, then αtZ(A(R))= idZ(A(R)) and αtZ⊕ XR OxdµR(x)=Z⊕ XR αt,x(Ox)dµR(x) for some measurable field of ∗-automorphisms αt,x :Ax(R)→ Ax(R)(i.e. dynamics is fiberwise). Proof. The constraints are central relative to strictly interior observables and thus fixed by αt . Hence all central boundary operators are fixed: αt ( f ) = f for f∈L∞ ( XR ). The commutant characterization (Lemma III.1) then implies decomposability of αtand establishes fiberwise action. Modular automorphisms. Exactly the same conclusions hold for modular automorphisms σω t of a faithful normal state ω: central elements are fixed, and the action is fiberwise, cf. (21)–(22). Physics reading. Dynamics cannot change the boundary label; it evolves each sector independently. This is the kinematic rigidity arising from Gauss’s law: the entanglement scaffold (the equalizer) is preserved in time. Propagation (microcausality, light-cone support) is therefore a property of the factor fibers, sewn together along their common label space. E. Worked models of the pullback/equalizer Discrete U(1) flux labels. Let X=Zindex the integer flux Φ∂R. Then A(R) = M q∈Z Aq(R),A(Rc) = M q∈Z Aq(Rc), Z(A(∂R)) = `∞(Z). The boundary maps pick the diagonal component for each q. The pullback is A(R)×`∞(Z)A(Rc)∼ =M q∈ZAq(R)¯ ⊗ Aq(Rc), i.e. a direct sum over matched flux sectors with ordinary tensor factors inside each q. Finite non-Abelian group G .Let X be the finite set of allowed boundary pairs ( C, α )(magnetic conjugacy class Cand electric irrep α). Then A(R) = M (C,α)∈X AC,α(R),A(Rc) = M (C,α)∈X AC,α(Rc), Z(A(∂R)) = CX, and A(R)×CXA(Rc)∼ =M (C,α)∈XAC,α(R)¯ ⊗ AC,α(Rc). F. Remarks on existence and uniqueness of pullbacks in vNAlg While general pullbacks need care in vNAlg , the present situation is particularly benign because the legs of the pullback land in a commutative von Neumann algebra Z ( A ( ∂R )) ∼ =L∞ ( X )and the algebras to be glued admit central decompositions over X . In such cases the pullback exists and is canonically identified with the direct integral over X of the fiberwise tensor products, as in the proof of Proposition IV.2. This also ensures that the construction is independent (up to canonical isomorphism) of the choice of faithful normal representation implementing the constraints: any two such representations give the same fiberwise description almost everywhere. 17 Tensor product choice. We used the spatial tensor product ¯ ⊗ on the fiber Hilbert spaces; this is the natural choice for commuting representations on a common Hilbert space (the global representation). If the two sides act on independent Hilbert spaces, the spatial product implements the physical commuting product on the glued system. G. Intuitive narrative and synthesis Equalizer = agreement. Equalizers isolate precisely those pairs of objects for which two ways of mapping to a third agree. Gauss’s law is an agreement condition: the boundary label computed from inside must match that computed from outside. The global algebra is the equalizer. Pullback = sewing. Pullbacks encode gluing along a base. Here the base is the commutative boundary center L∞ ( X ): the global algebra is obtained by sewing A ( R )and A ( Rc )along their common report to the boundary center. Bundle picture. The net is a Grothendieck fibration: a bundle of quantum fibers Ax ( R )over the classical label space XR . The pullback becomes a fiber product over X : inside and outside are glued labelwise, yielding R⊕ XAx(R)¯ ⊗Ax(Rc)dµ(x). Dynamics. Time evolution and modular flow preserve the base and act fiberwise, so the equalizer (sewing) is stable in time. Causal propagation is therefore a fiberwise (factor) property; the Gauss-law equalizer is the kinematic scaffold that makes a single global field possible. V. PROPAGATION, CAUSALITY, AND SPACETIME In this section we connect the Gauss-law sewing developed in Sections II–IV with the dynamical and causal structure of gauge theories. We make three points: (i) in canonical quantization, Gauss’s law fixes the longitudinal sector while the transverse sector propagates as waves; (ii) the equalizer/pullback sewing is the kinematic scaffold that makes a single global field possible, while causality (light-cone propagation) is a fiberwise feature of the transverse sector; (iii) microcausality of gauge-invariant observables follows from the support properties of the Pauli–Jordan commutator distribution and extends fiberwise in the central decomposition. Throughout we keep the notation: Σis a Cauchy slice; Rb Σis a regular region with boundary ∂R ; A ( R )is the strictly interior, gauge-invariant von Neumann algebra; Z ( A ( R )) ∼ =L∞ ( XR, µR )is its center; and A(Σ) ∼ =A(R)×Z(A(∂R)) A(Rc) is the Gauss-law pullback (Proposition IV.2). We set c=~= 1. A. Canonical analysis: transverse vs. longitudinal, Hamiltonian split We briefly recall the Coulomb-gauge canonical quantization of QED on R3 and emphasize the decomposition into transverse/longitudinal modes and the resulting Hamiltonian structure; see e.g. [31, 32]. Helmholtz decomposition. Any sufficiently regular vector field V on R3 admits a unique decomposition V=VT+VL,∇· VT= 0,∇ × VL= 0, with projectors (ΠT)ij =δij −∂i∂j ∇2,(ΠL)ij =∂i∂j ∇2. Applied to the canonical fields, A=AT,E=ET+EL,∇· AT= 0,∇ × EL= 0. Gauss’s law, ∇· E=ρ, 18 determines the longitudinal part by a Poisson equation: ∇2Φ = −ρ, EL=− ∇Φ,(25) with solution Φ(x) = ZR3 ρ(y) 4π|x−y|d3y. (26) Thus ELis an instantaneous functional of ρ, fixed by the constraint. Canonical commutation relations (CCR). In Coulomb gauge, the nontrivial equal-time CCR are Ai T(t, x), Ej T(t, y)=iδij −∂i∂j ∇2δ(3)(x−y),Ai T, Aj T=Ei T, Ej T= 0.(27) The longitudinal field EL is a functional of ρ and commutes with strictly interior gauge-invariant observables. Hamiltonian split. The Hamiltonian decomposes as H=HT+HC+Hint,(28) with HT=1 2ZE2 T+B2d3x, HC=1 2ZZ ρ(x)ρ(y) 4π|x−y|d3x d3y, Hint =Zj(x)·AT(x)d3x, where B = ∇ × AT and j is the conserved current. The longitudinal sector appears only through the instantaneous Coulomb term HC; all propagating dynamics are carried by the transverse sector. Proposition V.1 (Wave equation for the transverse field).The Heisenberg equations imply ∂2 t− ∇2AT(t, x) = jT(t, x),(29) where jT = Π Tj . In particular, in the vacuum ( j = 0) each component of AT satisfies the free wave equation with unit propagation speed. Proof. Using (27) and HT , one finds ∂tAT = ET and ∂tET = ∇ × B− Π Tj = ∇ × ( ∇ × AT ) −jT = −∇2AT−jT. Combining yields (29). Finite propagation speed. Solutions of (29) propagate with finite speed 1(domain-of-dependence property) [ 35 , Ch. 7]: compactly supported initial data influence only the causal future/past of their support. This already hints that causality (light-cone propagation) is a property of the transverse sector. B. Microcausality and Pauli–Jordan commutators We now state and prove microcausality for gauge-invariant observables, first in the free theory and then in the interacting theory via the locality axiom. Pauli–Jordan distribution. Let ∆0(x)be the massless Pauli–Jordan commutator function, ∆0(x) := Dret(x)−Dadv(x) = 1 2πsgn(x0)δx2,(30) where x2 = ( x0 ) 2− kxk2 . In 3 + 1 dimensions, ∆ 0 has support on the light cone (Huygens principle). For the quantized free Maxwell field, the field-strength commutator is a c-number distribution built from derivatives of ∆0[33, 34]: Fµν(x), Fρσ(y)=iηµρ∂ν∂σ−ηµσ∂ν∂ρ−ηνρ∂µ∂σ+ηνσ∂µ∂ρ∆0(x−y),(31) with ηµν = diag(1,−1,−1,−1). 19 Theorem V.2 (Microcausality for gauge-invariant fields (free Maxwell)) . Let fµν, gρσ ∈C∞ c ( R1,3 )have spacelike-separated supports. Then F(f), F(g)= 0, F(f) := ZFµν(x)fµν(x)d4x. Proof. By (31), F(f), F(g)=iZd4x d4y fµν (x)Kµν,ρσ ∆0(x−y)gρσ(y), where K is the differential operator in (31) acting on ∆ 0 . Since supp ∆ 0⊂ { ( x, y ):( x−y ) 2 = 0 } and supp f, supp g are spacelike separated, the integral vanishes by support considerations (standard distribution theory: the convolution of compactly supported test functions with a distribution supported on the light cone vanishes if the supports are spacelike separated). Hence the commutator is zero. Interacting case. In interacting (renormalized) gauge theory, microcausality for gauge-invariant local fields is an axiom (locality) of the Wightman/Haag–Kastler framework [ 33 , 117 ]. The Pauli–Jordan distribution does not appear explicitly, but the same conclusion holds: smeared field strengths (and gauge-invariant composites) with spacelike-separated supports commute. C. Propagation on the Gauss scaffold We now make precise how the Gauss-law equalizer/pullback provides the kinematic scaffold on which the propagating transverse sector lives, and how causality is a fiberwise property of the central decomposition. Fiberwise dynamics. Let A ( R ) ∼ =R⊕ XRAx ( R ) dµR ( x )be the central decomposition (Section III). By Lemma IV.7, the Heisenberg dynamics and modular flows act fiberwise and leave the center fixed: αtZ⊕ OxdµR=Z⊕ αt,x(Ox)dµR, σω tZ⊕ OxdµR=Z⊕ σωx t(Ox)dµR. In each fiber, the transverse fields satisfy wave equations of the form (29) and obey microcausality (Theorem V.2 in the free approximation, and by locality in the interacting case). Gluing across the boundary. Let Σ = R∪Rcas in Section IV. The Gauss pullback gives A(Σ) ∼ =Z⊕ XAx(R)¯ ⊗ Ax(Rc)dµ(x), with X = Spec Z ( A ( ∂R )) and common label x (Proposition IV.2). Thus the global algebra is literally sewn from inside/outside fibers with the same boundary label. Physically, x enforces the boundary matching: • the normal component of the electric field (flux) through ∂R agrees between inside and outside (no spurious surface charge on the artificial cut); •the tangential components of Bsimilarly match (no spurious surface current at the cut); •the longitudinal sector is completely fixed by Gauss’s law and does not propagate. The transverse sector then propagates within each fiber and the equalizer ensures that the two patches assemble into a single global wave. Theorem V.3 (Fiberwise microcausality implies global microcausality) . Let O1∈ A ( O1 )and O2∈ A ( O2 ) be gauge-invariant observables supported in spacelike separated regions O1, O2⊂R1,3 . Then [ O1, O2 ]=0 in A(Σ). Moreover, in the central decomposition Oj=Z⊕ X Oj,x dµ(x), j = 1,2, one has [O1,x, O2,x] = 0 for a.e. x, and hence R⊕[O1,x, O2,x]dµ = 0. Proof. Microcausality for the net is an axiom (Haag–Kastler locality) and holds in any faithful representation [ 117 ]. Decomposing with respect to the center L∞ ( X )shows that [ O1, O2 ]is decomposable (commutes with the center); since it vanishes, the fiber components vanish a.e. Conversely, if the fiber commutators vanish a.e., their direct integral vanishes. 20 Lattice locality and Lieb–Robinson bounds. On a lattice, local Hamiltonians generate dynamics with a finite group velocity characterized by Lieb–Robinson bounds [ 36 ]. Gauge-invariant lattice observables supported in disjoint regions approximately commute for times shorter than their Lieb–Robinson light-cone separation, and exactly commute at spacelike separation in the continuum limit. The Gauss constraint acts locally (vertex/plaquette constraints); the boundary flux remains central in the strictly interior algebra (Proposition XI.1), and the fiberwise picture persists. D. Emergence of causal structure from the fibers We now articulate the conceptual punchline: the global causal behavior of the gauge theory is the fiberwise causal behavior of the propagating (transverse) sector, assembled by the Gauss equalizer. Kinematics vs. dynamics. Gauss’s law is kinematical: it fixes the longitudinal sector and enforces the boundary matching (equalizer). This produces a global object (the glued algebra) from local pieces. Dynamics (Heisenberg evolution) then act fiberwise and transport information at finite speed (unit speed for the free massless field). Causality is therefore an intrinsic property of the fiber dynamics. Huygens principle and light-cone support. In 3 + 1 dimensions, the Pauli–Jordan commutator ∆ 0 has support on the light cone. For Fµν , commutators are supported on the light cone as in (31) . This justifies the intuitive picture that electromagnetic disturbances propagate at the speed of light and their commutators vanish outside the light cone. The equalizer ensures that the inside/outside descriptions are the same disturbance (same boundary label), so the global field inherits the light-cone propagation. Operational summary. Given two spacelike separated regions O1, O2 , any gauge-invariant O1∈ A ( O1 ) and O2∈ A ( O2 )commute. In the central decomposition this is the statement that, for each boundary label x , the factor algebras Ax ( O1 )and Ax ( O2 )commute (fiberwise microcausality); integrating over x (sewing via the center) preserves commutativity. E. Intuitive narrative: the Gauss scaffold and waves Why does the boundary matter for propagation? Because fields are continuous objects: to propagate across a cut you must ensure the two sides agree on the cut. Gauss’s law enforces exactly that agreement: the normal component of E and the tangential component of B match when there is no physical surface charge/current at the cut. In the algebra, this is the equalizer condition φR ( a ) = φRc ( b )and the centrality of Φ∂R. Where does the light cone come from? From the wave equation for the transverse sector (Proposition V.1) and the support properties of the Pauli–Jordan commutator (Theorem V.2). The longitudinal sector is instantaneous but non-propagating; it is fixed by the constraint and lives in the center (strictly interior observables commute with it). Thus causal propagation is entirely a property of the transverse fiber algebras. How are global waves built? Fiber by fiber: for each boundary label x, the transverse wave evolves in Ax ( R )and Ax ( Rc )and is glued at ∂R by the equalizer. The direct integral over x assembles the unique global wave in A(Σ). VI. THE MEASUREMENT PROBLEM AND SUPERSELECTION This section makes precise the sense in which the Gauss-law–induced center Z ( A ( R )) ∼ =L∞ ( XR, µR ) renders reduced states on A ( R )classical over boundary labels yet quantum within sectors (Sections II–III), and shows how this reproduces the structural hallmarks of measurement: (i) pointer observables that commute with interior operations (quantum non-demolition), (ii) repeatability and Bayesian updating, (iii) the disappearance of interference between different outcomes for all interior observables, and (iv) the emergence of a Kolmogorov classical stochastic layer describing boundary labels. We also relate this intrinsic, constraint-driven superselection to environment-induced decoherence (einselection) [ 38 – 40 ] and to the general instrument/POVM formalism [41–43]. Throughout we retain the notation: Σis a Cauchy slice, Rb Σa regular region with boundary ∂R . The strictly interior, gauge-invariant von Neumann algebra is A ( R ), with center Z ( A ( R )) ∼ =L∞ ( XR, µR ) 21 (Section III). The direct-integral (central) decomposition reads A(R)∼ =Z⊕ XR Ax(R)dµR(x),H∼ =Z⊕ XR HxdµR(x),(32) with factor fibers Ax(R)⊂B(Hx)(Proposition III.2). A. Superselection as an algebraic non-interference principle We begin by recasting superselection as the statement that interior observables cannot induce or detect coherence between different boundary labels. Theorem VI.1 (No interior interference across the center) . Let E, F ⊂XR be disjoint measurable sets with µR ( E ) , µR ( F ) > 0. For any O∈ A ( R )and ξ∈ H ( E ) := R⊕ EHxdµR , η∈ H ( F ), one has hξ, O ηi = 0 . Equivalently, if PE, PF∈Z ( A ( R )) are the corresponding central projections then PEOPF = 0 for all O∈ A(R). Proof. Since PE, PF∈Z ( A ( R )),[ PE, O ]=[ PF, O ]=0. As PEPF = 0 for disjoint E, F , PEOPF = OPEPF= 0.For vectors, ξ=PEξand η=PFη, hence hξ, Oηi=hξ, PEOPFηi= 0. Physics. Any superposition that spans different boundary labels is operationally indistinguishable (with strictly interior probes) from the corresponding classical mixture. This is not a dynamical loss of phase information (no bath required), but a kinematic limitation enforced by Gauss’s law: interior observables commute with the center and hence cannot connect label sectors. B. States, classical layers, and a state-dependent conditional expectation Let ω be a normal state on A ( R ). As recalled in (32) , there exists a probability measure µω R on XR and a measurable field {ωx}of fiber states with ωZ⊕ XR OxdµR(x)=ZXR ωx(Ox)dµω R(x).(33) From ωwe can build a state-dependent conditional expectation onto the center. Proposition VI.2 (Center-valued expectation induced by a state) . Fix a normal state ω with disintegration (33). Define Eω:A(R)→L∞(XR, µω R)by EωZ⊕ OxdµR(x) := ωx(Ox)for a.e. x. (34) Then Eω is a normal, unital, completely positive (UCP), idempotent map with the bimodule property Eω(f O g) = f Eω(O)gfor all f, g ∈L∞(XR)∼ =Z(A(R)) and O∈ A(R). Proof. Normality and complete positivity follow fiberwise from positivity of ωx and measurability. Unitality: Eω ( 1 )( x ) = ωx ( 1 ) = 1. Idempotence: if h∈L∞ ( XR )then Eω ( h )( x ) = h ( x ), so Eω ( Eω ( O )) = Eω ( O ). Bimodule property is immediate: (fOg)x=f(x)Oxg(x)and ωxis linear. Interpretation. Eω collapses any interior observable to its classical shadow on the label space XR , by evaluating its fiber expectation values. Dually, for f∈L∞ ( XR )we have ω ( fO ) = Rf ( x ) Eω ( O )( x ) dµω R ( x ) , which is precisely the rule of classical conditioning on a random variable f. C. Measuring the center: spectral measures, instruments, and Lüders updates Let Z∈Z ( A ( R )) be a self-adjoint central observable. By the spectral theorem there is a projectionvalued measure (PVM) ∆ 7→ PZ (∆) ∈Z ( A ( R )) on Borel sets ∆ ⊂R such that Z = Rλ dPZ ( λ ). For a normal state ω , the Born rule probability of outcomes in ∆is ω ( PZ (∆)) = µω R ( {x : Z ( x ) ∈ ∆ } ), i.e. the pushforward of the boundary-label distribution under x7→ Z(x). 22 Definition VI.3 (Central instrument (Lüders form)) . For Borel ∆ ⊂R define the normal CP map (instrument component) I∆(O) := PZ(∆) O PZ(∆) = PZ(∆) O=O PZ(∆), O ∈ A(R).(35) The unconditioned post-measurement channel is I ( O ) = RIdλ ( O ) = PkPkOPk in the discrete case Z=PkzkPk. Proposition VI.4 (QND & repeatability for central measurements) . Central measurements are quantum non-demolition (QND) and repeatable: [I∆(O), Z ]=0,I∆◦ I∆=I∆,I∆1◦ I∆2= 0 if ∆1∩∆2=∅. Proof. Commutativity with Z is immediate since PZ (∆) ∈Z ( A ( R )). Idempotence and orthogonality follow from PZ(∆) being a projection and PZ(∆1)PZ(∆2)=0for disjoint ∆i. Posterior states and Bayesian update. Given ω, the posterior (selective) state conditioned on ∆is ω∆(O) := ωI∆(O) ωPZ(∆)=1 µω R(Z−1∆) Zx:Z(x)∈∆ ωx(Ox)dµω R(x). In particular, for a partition {Ek} of XR corresponding to the spectral decomposition, the update rule is Bayesian on the classical layer: p0 k = Pr ( Ek|obs ) = pk/Pjpj , and inside each Ek the fiber states remain unchanged (Lüders rule is trivial since PZ(∆) acts as identity on that fiber). Operational consequence. Because PZ (∆) ∈Z ( A ( R )), a central measurement cannot disturb any strictly interior observable O : I ( O ) = PkPkOPk = PkOPk = O . Thus measuring boundary labels is nondemolition for the interior quantum dynamics—precisely the measurement structure one expects for a pointer variable. D. Naimark dilation and POVMs for boundary labels Any POVM E : B ( XR ) →Z ( A ( R )) + (a positive-operator valued measure with values in the center) is classical in the sense that E (∆) are mutually commuting effects. By Naimark’s dilation theorem [ 41 , 43 ], there exists a Hilbert space K , a projection-valued measure e E on K , and an isometry V : H → K such that E(∆) = V∗e E(∆)V. Because E(∆) ∈Z(A(R)), one can choose the dilation so that e E(∆) acts only on an ancilla living in a thin collar around ∂R (or in Rc ), and V intertwines the Gauss-law pullback. Thus any boundary-label measurement can be implemented by coupling to an external pointer without disturbing the interior fibers—again a QND property. E. Constraint-induced superselection vs. environment-induced decoherence In environment-induced decoherence (EID) [ 38 – 40 ], a system S coupled to an environment E via Hint = PαAα⊗Bα loses coherence in the eigenbasis of “pointer” observables Aα that (approximately) commute with Hint . By contrast, in the present setting the constraint algebra enforces exact commutation of interior observables with the center Z(A(R)). Proposition VI.5 (Einselection selects the center in gauge-invariant couplings) . Let the interaction between R and Rc be described by a gauge-invariant Hamiltonian Hint ∈ A ( ∂R ) 00 supported in a thin collar of ∂R . Assume [ Hint, Z ] = 0 for all Z∈Z ( A ( R )) (true since Z are central boundary observables). Then any EID pointer observable for the interior algebra must be a (function of a) central element of A(R). Idea. In EID, pointer observables A are singled out by (approximate) commutation [ A, Hint ] ≈ 0, guaranteeing stability under monitoring by the environment. But Z ( A ( R )) is the maximal commutative algebra that already commutes with all strictly interior operations and with Hint (since the latter is boundary-local and gauge invariant). Therefore any stable pointer must lie in Z ( A ( R )) up to small corrections. 23 Moral. EID confirms the algebraic superselection structure rather than creating it: the same center that appears from constraints is the unique stable pointer algebra under gauge-invariant couplings to the outside. Unlike standard decoherence, however, here the classicalization over XR is exact at the algebraic level (Theorem VI.1); no dynamical or perturbative approximation is needed. F. A canonical QND measurement: flux sectors in lattice U(1) Consider the lattice U (1) setting of Section II B 3. Let XR = Z label the integer boundary flux, and let {Pq}q∈Z⊂Z(A(R)) be the spectral projections of Φ∂R (flux projectors). Define the instrument Iq(O) := PqO Pq, O ∈ A(R). Given a state ω = Pqpqωq (central decomposition), the probability of outcome q is pq = ω ( Pq ), and the posterior state on A(R)is ωpost q(O) = ω(Iq(O)) ω(Pq)=ωq(O). As expected for a QND measurement of a central observable, the posterior equals the prior within the selected sector. Physically, one may realize this measurement by coupling a boundary ancilla to a set of boundary links with Hamiltonian Hint = g Φ ∂R ⊗ Π(pointer momentum) and then reading out the ancilla position; Naimark dilation ensures the existence of such a construction. G. Entropies and information balance in central measurements In a finite (or finite-entropy) setting, the entropy of the reduced state obeys the split S ( ρR ) = H ( {px} ) + PxpxS ( ρx )(Section III E). A sharp measurement of the center (Section VIC) converts the a priori distribution {px} into a posterior {p0 x} concentrated on the observed x0 ; correspondingly, the classical piece H ( {px} )is converted into Shannon information gained, ∆ I = H ( {px} ), while the quantum piece PxpxS ( ρx )is unaffected (QND). Thus the center behaves as a classical random variable in the Shannon sense, while the interior fibers carry the quantum uncertainty. H. Summary and narrative •Superselection as kinematic classicalization. The Gauss-law center Z ( A ( R )) enforces an exact no-interference principle (Theorem VI.1): interior observables cannot detect relative phases between different boundary labels. Reduced states are classical mixtures over XR and quantum states within sectors. •Measurements of the center are QND and repeatable. Because central projections commute with all interior observables, measurements of Z∈Z ( A ( R )) neither disturb interior quantum dynamics nor change the fiber state in the selected sector (Proposition VI.4); updates are Bayesian on the classical layer. •POVMs and dilations stay at the boundary. Any POVM with effects in the center admits a Naimark dilation that couples only to a boundary ancilla, implementing a measurement that respects the Gauss-law sewing. •Environment-induced decoherence selects the same pointer. Under gauge-invariant couplings, EID stabilizes exactly the central algebra: the pointer equals the Gauss-law center (Proposition VI.5). In the next section we turn to higher categorical coherence breakdown, where operators that act on the center become admissible. There, the perfect classicalization over labels is relaxed: central measurements cease to be strictly QND, and the equalizer gluing is weakened to a homotopy notion—opening a path to “self-measurement” of spacetime. 24 VII. HIGHER CATEGORICAL COHERENCE BREAKDOWN This section formalizes and develops the key conceptual move of the paper: the passage from strict Gauss-law sewing—in which the boundary labels form a commutative center that is immutable under interior operations—to deformed sewings in which higher categorical coherence defects allow interior operators to act on the center. Technically, we structure the discussion in four parts. §7.1 Triangle-level coherence failure (2-cocycles): central extensions and projective representations yield the Heisenberg/Weyl commutation relations; the extension is central and preserves unitarity and superselection of the Gauss center. §7.2 Higher-level coherence failure (3-cocycles, quasi-Hopf, and noncentral associators): the pentagon constraint is relaxed by an associator Φthat is not a central scalar, permitting noncentral extensions. This makes it possible for interior operations to act on boundary labels. §7.3 Consequences for Gauss-law centers and sewing : we replace the strict equalizer by a weak/homotopy equalizer, and give precise algebraic models (unitary crossed products; nonunitary CP/Lindblad channels) in which the center is dynamically mixed. §7.4 Worked deformations : rigorous constructions with proofs for (U) unitary label mixing and (CP) nonunitary self-measurement, including the structure of the deformed center, flux-raising commutators, and entropy/Spohn inequalities. We maintain the notation of previous sections: Rb Σis a regular region with boundary ∂R , the strictly interior gauge-invariant algebra is A(R), and Z(A(R)) ∼ =L∞(XR, µR) is its center, with XRthe boundary-label space. Central (direct–integral) decomposition is A(R)∼ =Z⊕ XR Ax(R)dµR(x), with factor fibers Ax(R)(Section III). A. Triangle-level coherence failure: central extensions and unitarity At the “triangle” level (2-simplex), coherence failures are governed by 2-cocycles. The canonical manifestation is the passage from linear to projective representations, classified by H2 ( G, U (1)), and, in the continuous case, the Heisenberg/Weyl central extension of phase space. Projective representations and H2 ( G, U (1)).Aprojective unitary representation of a group G on H is a map U:G→ U(H)such that U(g)U(h) = ω(g, h)U(gh), ω :G×G→U(1).(36) Associativity of the product imposes the 2-cocycle identity ω(h, k)ω(g, hk) = ω(g, h)ω(gh, k),(37) and a change U ( g ) 7→ ξ ( g ) U ( g ), ξ : G→U (1), twists ω by a 2-coboundary, ω7→ ω δξ . Cohomology classes [ω]∈H2(G, U(1)) classify central extensions 1→U(1) →e G→G→1 and projective representations lift to honest representations of e G. Heisenberg–Weyl central extension. Let R2n be classical phase space with symplectic form σ((q, p),(q0, p0)) = p·q0−q·p0. The Weyl operators W(ξ),ξ∈R2n, obey W(ξ)W(η) = ei 2σ(ξ,η)W(ξ+η),(38) which is again a projective representation with multiplier ei 2σ . Equivalently, the Lie algebra has a central commutator [ Qj, Pk ] = i δjk 1 . The Stone–von Neumann theorem ensures uniqueness (up to unitary equivalence) of irreducible representations satisfying (38). 25 Unitarity and the Gauss center. In all these cases, the “coherence defect” (the multiplier ω or the central commutator) is central (scalar). As a result: • the dynamics remain unitary; associators on multi-particle tensor products are still scalar phases (coherence theorems), • the Gauss-law center Z ( A ( R )) is untouched: central defects commute with everything and cannot generate operators that mix boundary labels. This is the precise, technical sense in which “triangle-level coherence breakdown produces only central extensions and preserves unitarity.” B. Higher-level coherence failure: noncentral associators and quasi-Hopf structure We now step to higher simplices. In monoidal categories (C,⊗,1, α), the associator αX,Y,Z : (X⊗Y)⊗Z−→ X⊗(Y⊗Z) must satisfy the pentagon identity. “Strict” monoidal categories have α = id ; Mac Lane’s coherence shows this suffices. To deform coherence one permits nontrivial α. Scalar vs. noncentral associators. If α is a natural isomorphism by scalars (phases)—as in Vecω G for ω∈H3 ( G, U (1))—then α acts by central scalars on all morphism spaces. This is still “triangle-level” in spirit: it does not enable interior operations to act on centers. In contrast, in Drinfel’d’s quasi-bialgebras/quasi-Hopf algebras, there is an associator element Φ∈H⊗3 such that coassociativity holds only up to conjugation by Φ, and Φsatisfies a nontrivial pentagon relation. If Φis not proportional to 1⊗1⊗1 in all representations, then the induced associators on representation categories are operator-valued, hence noncentral. Physically, these associators can imprint nontrivial operators into the “parentheses flip” (X⊗Y)⊗Z⇒X⊗(Y⊗Z). Mechanism to affect the Gauss center. Let C be a monoidal category controlling the composition of edge-mode degrees of freedom on ∂R (e.g. charge/flux labels and their fusion). If the associator α on C is scalar, then rebracketing operations cannot change the values of central boundary observables; they only contribute phases. If, however, α carries noncentral operator content, then the very act of recombining boundary degrees of freedom can enact nontrivial transformations on the algebra generated by these labels, thereby allowing interior operators (built from compositions) to act on the center. Toy schematic. Let XR be discrete and let the boundary label algebra be L∞ ( XR ). Suppose the associator on three boundary “edge modes” induces a permutation T:XR→XRat the algebra level: α: (f⊗g)⊗h7→ f◦T−1⊗(g⊗h). If this associator is implemented by an interior unitary Uin the local algebra, then U f U∗=f◦T−1, f ∈L∞(XR), which is precisely the mechanism that mixes boundary labels. This U cannot exist in the strict theory (it would violate centrality), but appears when the associator is a noncentral operator originating in higher coherence failure. C. From strict equalizer to weak/homotopy equalizer We translate these deformations into the language of Section IV. Recall that strict Gauss sewing gives the pullback A(Σ) ∼ =A(R)×Z(A(∂R)) A(Rc), with boundary maps φR, φRcequalizing strictly. Higher coherence failure admits two algebraic avatars: 32 H. Summary of Section VIII • We defined the self-measurement regime as the existence of interior implementers that enact a nontrivial boundary action βon the label algebra. •In the strict theory, no such implementer exists (Theorem VIII.2); under deformations, the center shrinks to invariants or to the fixed-point algebra (Proposition VIII.3). • Edge channels transport information: relative entropy splits into a classical boundary term and a bulk conditional term obeying data-processing monotonicity (Proposition VIII.4); in the unitary case, the edge is a noiseless classical bus on XR/hTi. •Microcausality is preserved: self-measurement is boundary-local (Theorem VIII.5). • Central QND fails when the edge is dynamical: reading/writing the boundary label are now genuine operations with nontrivial disturbance (Proposition VIII.6). These results prepare the ground for Section IX, where the same mechanism is applied to the blackhole/island framework: the dynamical edge becomes the microscopic channel through which information flows across quantum extremal surfaces. IX. BLACK HOLES, ISLANDS, AND INFORMATION We now apply the algebraic and categorical framework developed in Sections II–VIII to the black–hole information problem. The two core messages of this section are: 1. Islands as pullbacks. The fine–grained algebra of the Hawking radiation is not the naive factor A(R∞)attached to a region R∞in the asymptotic domain; rather, gravity’s constraints and edge data on a candidate quantum extremal surface (QES) Γforce a gluing Arad(Γ) ∼ =A(R∞)×Z(A(Γ)) A(IΓ),(45) where IΓis the island bounded by Γ. This is the gravitational version of the Gauss–law equalizer. 2. QES/replica to center. The area (or Wald) term Area (Γ) / (4 G~ )and the replica–wormhole selection of Γare reinterpreted as the classical layer of a central decomposition over Z ( A (Γ)), while the bulk matter entropy in R∞∪IΓ is the quantum layer inside each sector. Optimizing over Γcorresponds to choosing the pullback over the dominant central sector. We proceed in stages. First we recall the paradox and the naive factorization failure; next we identify the gravitational edge algebra at a putative QES and show it is central relative to interior observables; then we construct the radiation algebra as a pullback and recover the JLMS/FLM/QES relations from the central decomposition. We then extract the Page curve and discuss how the higher–coherence deformations of Section VII modify the island mechanism. Throughout we allow either asymptotically AdS or asymptotically flat evaporating setups; R∞ denotes the distant radiation region on I+ (or the AdS boundary reservoir), B denotes the black–hole exterior region on a Cauchy slice, and I⊂Bcdenotes a bulk island. The candidate QES is denoted Γ = ∂I. A. The paradox and the failure of naive factorization Hawking’s semiclassical calculation suggests that the state of the outgoing radiation becomes thermal and more and more mixed as evaporation proceeds, yielding an ever–increasing fine–grained entropy for the radiation S ( R∞ ), in tension with unitarity. The naive tensor–factor argument takes Htot ? ∼ =Hrad ⊗ HBH and traces over the black hole. Gravity, however, does not admit such a factorization: diffeomorphism constraints and boundary charges obstruct the split. Algebraically, the local von Neumann algebras attached to complementary regions have a nontrivial center supported on their common boundary (Section II). In gravity the relevant “boundary” is the (a priori unknown) QES Γ. 33 B. Gravitational edge algebra on a QES and centrality Let Γbe a smooth spacelike codimension–two surface. The gravitational symplectic form acquires a boundary term on Γwhose integral generates the Noether–Wald charge for diffeomorphisms normal to Γ[ 56 , 57 ]. For Einstein gravity the associated observable reduces to the area operator A (Γ); more generally it is the Wald entropy density integrated on Γ. In addition, there are soft gravitational edge modes (analogous to gauge edge modes [ 58 ]) that encode large diffeomorphisms acting nontrivially at Γ. Proposition IX.1 (Centrality of gravitational edge observables at Γ) . Let A ( O )be the diffeomorphism–invariant local algebra of a region O whose closure does not meet Γ. Then any gravitational boundary observable ZΓ supported on Γ(e.g. the area/Wald operator, soft charges) commutes with A ( O ): [ZΓ,A(O) ] = 0. Equivalently, ZΓ∈ZA(N(Γ)), the center of the algebra in a thin collar N(Γ). Sketch. Hamiltonian and momentum constraints generate diffeomorphisms; the diffeo–invariant algebra A ( O )lies in the commutant of these constraints. Smearing the constraints with a test vector field supported in a collar of Γyields a boundary term which equals the variation of the Wald charge on Γ[ 56 ]. Since the local observables commute with the constraints and their support is disjoint from Γ, they commute with the induced boundary observables, establishing centrality. The argument parallels Proposition II.1 with Gauss law replaced by the diffeo constraints. Thus the edge algebra on Γ(generated by A (Γ) and soft charges) plays the role of the commutative center at the gluing surface. Denote this commutative von Neumann algebra by Z ( A (Γ)) ∼ =L∞ ( XΓ, µΓ ), with label space XΓparameterizing boundary data on Γ. C. Radiation algebra with an island as a pullback Let IΓ be the domain of dependence of a bulk region bounded by Γ, and write A ( IΓ )for the diffeo–invariant local algebra therein. Likewise let A ( R∞ )be the algebra of the far radiation. There are boundary maps (“boundary reports”) φR∞:A(R∞)→Z(A(Γ)), φI:A(IΓ)→Z(A(Γ)), encoding the edge labels seen from the outside and inside, respectively (Section IV). Gravity’s sewing condition is that these agree. Proposition IX.2 (Islands as pullbacks) . The algebra of the radiation including the island IΓ is the pullback Arad(Γ) ∼ =A(R∞)×Z(A(Γ)) A(IΓ) = {(a, b) : φR∞(a) = φI(b)}.(46) In the central decomposition over XΓthis reads Arad(Γ) ∼ =Z⊕ XΓAx(R∞)¯ ⊗ Ax(IΓ)dµΓ(x). Proof. Identical to Proposition IV.2, with L∞ ( X )replaced by Z ( A (Γ)) and using Proposition IX.1 for centrality. The fiberwise form follows from the central decomposition (Section III). Physics reading. Equation (46) is the algebraic heart of the island rule: the radiation algebra must be extended by gluing an island along the QES edge center. D. Generalized entropy and the QES formula from the center Let ρ be the global state on a code subspace of semiclassical configurations. The island prescription states (heuristically) S(ρArad ) = ext ΓhA(Γ)iρ 4G~+SbulkρA(R∞)∨A(IΓ),(47) with quantum corrections included [ 59 – 62 ] and replica–wormhole justification [ 64 , 65 ]. We now explain how (47) emerges structurally from the central decomposition. 34 Central decomposition at Γ.Decompose over the edge center: Z(A(Γ)) ∼ =L∞(XΓ, µΓ), ρArad(Γ) =ZXΓ ρxdµρ Γ(x), with ρx a state on Ax ( R∞ ) ¯ ⊗Ax ( IΓ ). In a semiclassical saddle–point regime the distribution on XΓ is sharply peaked by an effective action dµρ Γ ( x ) ∝e−Ieff (x)dν ( x ), where Ieff includes the area functional and local counterterms. The entropy splits as in (19): SρArad(Γ)=H(µρ Γ) + ZXΓ S(ρx)dµρ Γ(x).(48) Under the saddle approximation, H ( µρ Γ ) ≈ hA (Γ) iρ/ (4 G~ )(up to one–loop terms), while the fiber entropy equals the bulk matter entropy in R∞∪IΓ . Extremizing with respect to x yields the QES condition δSgen = 0 [61, 62]. JLMS from the pullback. The JLMS relation equates boundary relative entropy to bulk relative entropy plus the area term [63]. In our language, Srel(ρkσ;Arad(Γ)) = hA(Γ)iσ− hA(Γ)iρ 4G~+ZXΓ Srel(ρxkσx)dµρ Γ(x),(49) which is the direct–integral version of JLMS: the area term is the (center) modular–Hamiltonian piece, while the integral is the bulk relative entropy on the fiber algebras Ax ( R∞ ) ¯ ⊗Ax ( IΓ ). Equation (49) is rigorously the Araki relative–entropy split (Section III E) applied to the pullback algebra, together with the identification of the central generator with the Wald charge [56]. E. Replica wormholes as center selection In the replica trick, the gravitational path integral over n replicas admits disconnected “wormhole” saddles gluing the replicas through a common surface homologous to Γ[ 64 , 65 ]. In the present framework this gluing is reinterpreted as conditioning on a common central label x∈XΓ across replicas. The dominance of a given saddle is the statement that µρ Γ is concentrated near the corresponding x ; evaluating the Shannon term H ( µρ Γ )at this peak reproduces the area piece. Switching dominance at the Page time corresponds to a transfer of weight in µρ Γfrom the “no–island” sector to an “island” sector. F. The Page curve from a two–sector toy model Consider a minimal model with two central sectors x∈ { 0 , 1 } : x = 0 (no island) and x = 1 (island present). Let ρArad =p0ρ0⊕p1ρ1, p1=e−∆A(t)/(4G~) 1 + e−∆A(t)/(4G~), with ∆ A ( t ) = A1 ( t ) −A0 ( t )the generalized–area difference between the two candidates. The radiation entropy is Srad(t) = H(p0, p1) + p0S(ρ0(t)) + p1S(ρ1(t)). At early times ∆ A 1so p1≈ 0and Srad ≈S ( ρ0 )grows with the Hawking flux; at late times ∆ A− 1 so p1≈ 1and Srad ≈S ( ρ1 )decreases because ρ1 purifies the previously emitted quanta via the island attachment. The crossover reproduces the Page curve [ 66 ]. This is the central–decomposition avatar of the island transition. G. Quantum error correction viewpoint and code subspace In AdS/CFT, bulk reconstruction and entanglement wedge reconstruction are captured by quantum error correction (QEC) [ 67 , 68 ]. In our language, the code subspace restriction ensures that: (i) the central decomposition over XΓis sharply peaked (semiclassical geometry), and (ii) the fiberwise JLMS identity (49) holds. The pullback algebra (46) is precisely the algebra recovered by QEC from the boundary degrees of freedom on the radiation side, with the island IΓsitting in the entanglement wedge. 35 H. Higher–coherence deformations and modified islands Section VII introduced deformations in which interior operators can act on the center. Interpreted at a QES, this yields: •(U) Unitary label mixing. If there exist interior unitaries U supported near Γthat implement a nontrivial automorphism of Z ( A (Γ)) (e.g. shift a discrete flux/soft–hair label), the edge center collapses to the invariant subalgebra. Operationally, the island label becomes a quantum bus supporting coherent transport of information between R∞ and IΓ . The classical area–like term diminishes to the entropy of invariants (often trivial if the action is ergodic). •(CP) Boundary CP channels. A boundary–localized GKSL semigroup {EΓ t} induces a Markov semigroup on Z ( A (Γ)) and CP intertwiners between fibers. The island label undergoes stochastic evolution (“self–measurement of the QES”); the Shannon edge term obeys data–processing monotonicity (Section VIII C). The strict equalizer is replaced by a weak/homotopy equalizer. Proposition IX.3 (Bounds on the deformed radiation entropy) . Let ρt evolve under a boundary CP semigroup EΓ tacting on the edge. Then for any stationary faithful ϕ, Srel(ρtkϕ;Arad(Γ)) is nonincreasing in t, and the classical edge contribution D ( µρ tkµϕ )decreases under the induced Markov kernel on XΓ . In particular, edge dynamics cannot increase the generalized entropy beyond the QES extremum value. Idea. Apply Proposition XIV.2 to the pullback algebra and restrict to the center for the classical piece (data processing). The bulk fiber maps are CPTP; relative entropy is monotone under CPTP maps. Interpretation. Deforming the center modifies the mechanism by which information exits the black hole (through edge channels), but preserves causality and the overall monotonicity dictated by complete positivity. In the (U) case, the island label can coherently carry information (Hayden–Preskill–type mirroring across the QES [ 69 ]); in the (CP) case, the edge behaves like a noisy conduit that self–measures the gluing data. I. Consistency: causality and energy conditions All the deformations considered are localized in a thin collar of Γand respect microcausality (Theorem VIII.5). Moreover, the QES condition arises from the first law of entanglement [ 61 ]: small variations of the state deform Γto keep δSgen = 0. In the present language, this is the stationarity of the center distribution µρ Γ plus the fiber entropies under deformations that preserve the constraints, consistent with the null–energy condition that underlies focusing of extremal surfaces [62]. J. Summary of Section IX •Gravity induces a central edge algebra on candidate QES surfaces Γ(Proposition IX.1). • The correct radiation algebra is the pullback (46) , i.e. islands are equalizers over the edge center (Proposition IX.2). • The generalized–entropy/QES formula and JLMS follow from the central decomposition and its relative–entropy split (48)–(49). • Replica wormholes select a central sector; the Page transition is a redistribution of weight in the edge–label distribution µρ Γ(§IX E–IX F). • Higher–coherence deformations turn the edge into a dynamical information channel, coherent (crossed product) or noisy (CP), without violating causality; generalized entropy obeys monotonicity bounds (§IX H). 36 X. MODULAR THEORY, KMS STRUCTURE, AND TYPE III CLASSIFICATION This section develops the Tomita–Takesaki modular theory and KMS structure for local gauge-invariant algebras with nontrivial center and explains how Connes’ invariants classify the factor fibers that arise in the central decomposition (Section III). Our goals are: 1. Prove that the modular data (∆ω, Jω, σω t)of a faithful normal state ωon A(R)fix the center and decompose fiberwise on the direct integral A(R)∼ =R⊕ XRAx(R)dµR(x). 2. Show how KMS states on A ( R )factor into a classical component on the center and fiberwise KMS components. 3. Connect the modular spectra of fibers to Connes’ S -invariant and the type IIIλ classification, and recall why local algebras in relativistic QFT are typically of type III1. 4. Give explicit illustrations (Bisognano–Wichmann, Unruh/KMS; relative entropy and modular Hamiltonians) in the present direct-integral language. Throughout, Rb Σis a regular region, A ( R ) ⊂B ( H )is the strictly interior gauge-invariant von Neumann algebra, and Z(A(R)) ∼ =L∞(XR, µR),A(R)∼ =Z⊕ XR Ax(R)dµR(x), as in (17) . We use the standard Tomita–Takesaki notation: given a faithful normal state ω on A ( R ), its GNS triple is ( πω,Hω, Ω ω ); the Tomita operator is Sω with polar decomposition Sω = Jω ∆ 1/2 ω ; the modular automorphism group is σω t(A)=∆it ωA∆−it ω[114, 115]. A. Standard form, direct integrals, and fiberwise modular objects We first recall the standard form of a von Neumann algebra and its behavior under central decompositions. Theorem X.1 (Standard form and central disintegration) . Let M be a von Neumann algebra with center Z(M)∼ =L∞(X, µ)and let ωbe a faithful normal state on M. Then there exist: • a measurable field { ( Mx,Hx, Jx,Px ) }x∈X of standard forms (so Jx is the modular conjugation and Pxthe natural cone) for factor algebras Mx; •a measurable field of cyclic separating vectors {Ωx}with ωx(Ax) = hΩx, AxΩxi; such that, writing M=R⊕ XMxdµ(x)and H=R⊕ XHxdµ(x), Ω := Z⊕ X Ωxdµ(x)is cyclic separating for M,(M, J, P) = Z⊕ X (Mx, Jx,Px)dµ(x). Sketch. This is a standard consequence of the spectral theorem for the center together with Takesaki’s construction of the standard form [ 114 , III.2, III.4]. The measurability assertions are ensured by the existence of countable dense sets of sections and the fact that Z ( M )acts as the multiplication algebra L∞(X, µ). Theorem X.2 (Fiberwise Tomita–Takesaki) . With the hypotheses of Theorem X.1, the Tomita operator, modular conjugation, and modular operator decompose as Sω=Z⊕ X Sωxdµ(x), Jω=Z⊕ X Jωxdµ(x),∆ω=Z⊕ X ∆ωxdµ(x),(50) and the modular automorphism group is decomposable: σω tZ⊕ X Axdµ(x)=Z⊕ X σωx t(Ax)dµ(x)for all t∈R.(51) 37 Proof. By definition Sω is the closure of πω ( M )Ω ω3A Ω ω7→ A∗ Ω ω . Since M and Ω ω decompose fiberwise and πω is decomposable, so is the graph of Sω , hence Sω = R⊕Sωx . The polar decomposition is compatible with direct integrals of closed operators, yielding the decompositions for Jω and ∆ ω [ 114 , III.4.21]. Equation (51) follows from σω t= Ad ∆it ωand decomposability of ∆it ω. Proposition X.3 (The center is fixed by modular flow) . For any faithful normal state ω on M , σω t acts trivially on Z(M): σω t(f) = f, ∀f∈Z(M),∀t∈R. Proof. In standard form, Z ( M )acts as multiplication operators Mf on H and commutes with both M and M0 . Since Sω is defined by SωA Ω = A∗ Ω, one has SωMf = MfSω on πω ( M )Ω. Hence Mf commutes with both Jωand ∆ω, and σω t(Mf) = Mf. Physics. Proposition X.3 sharpens the statement used in Sections III and IV: modular flow cannot change the boundary label. Theorem X.2 then says all modular dynamics (and modular Hamiltonians) act within sectors. B. Relative modular operators and Araki relative entropy Given faithful normal states ω, ϕ on M , the relative Tomita operator Sω|ϕ is the closure of A Ω ϕ7→ A∗ Ω ω , with polar decomposition Sω|ϕ=Jω|ϕ∆1/2 ω|ϕ. The Araki relative entropy is S(ωkϕ) := −hΩω,log ∆ω|ϕΩωi ∈ [0,∞]. Theorem X.4 (Fiberwise relative modular objects and relative entropy) . Under the hypotheses of Theorem X.1, for faithful ω, ϕ one has ∆ω|ϕ=Z⊕ X ∆ωx|ϕxdµ(x), S(ωkϕ) = ZX S(ωxkϕx)dµω(x), where dµωis the measure induced by ωon X(cf. (15)). Idea. The proof parallels Theorem X.2: Sω|ϕ is defined fiberwise and is decomposable, so is ∆ ω|ϕ . The relative-entropy identity follows from the spectral calculus for direct integrals and the identification Ωω=R⊕Ωωxdµω(x); see [115, Ch. VIII] and [72]. Physics. Equation (20) in Section III is the special case of Theorem X.4 when ϕ is a reference state implementing a conditional expectation onto Z(M). C. KMS states with center: factorization and classical layer Let ( M, αt )be a W∗ -dynamical system (a normal one-parameter automorphism group). A normal state ω is KMS at inverse temperature β > 0if for any A, B in a σ -weakly dense subalgebra Man of entire analytic elements, FA,B(t) := ω(A αt(B)) extends holomorphically to {0<=z < β},with FA,B(t+iβ) = ω(αt(B)A). When αt=σω tand β= 1, this holds for all faithful ω(Tomita–Takesaki, KMS condition) [114, 115]. Theorem X.5 (KMS factorization when αt fixes the center) . Assume αtZ(M) = id for all t and M = R⊕ XMxdµ ( x ). Let ω be a normal state disintegrated as ω = RXωxdµω ( x ). Then ω is KMS β for (M, αt)iff, for µω-a.e. x,ωxis KMSβfor (Mx, αt,x), where αt,x is the fiber action in (51). Proof. “Only if”: Fix A = R⊕Ax and B = R⊕Bx with Ax, Bx analytic. Then FA,B ( z ) = RFAx,Bx ( z ) dµω ( x )and the strip analyticity and boundary values imply the KMS identity fiberwise for a.e. x . “If”: Reverse the argument using dominated convergence and analyticity of the integrand for a dense set of sections; extend by continuity to Man. 38 Classical layer. Theorem X.5 shows that, if dynamics leave the center fixed (as in gauge-invariant Hamiltonians; cf. Lemma IV.7), the KMS condition imposes no dynamics on the boundary label distribution besides stationarity: µω is arbitrary subject to normality and ωx being KMS fiberwise. If αt were to act nontrivially on Z ( M )(e.g. through a deformed, label-mixing dynamics as in Section VII), then a necessary condition for ω to be KMS is that µω be invariant under the induced classical flow on X . D. Connes’ S-invariant and type IIIλclassification Let M be a factor (center trivial). For a faithful normal state ω , write Sp (∆ ω )for the spectrum of its modular operator. Connes’ S-invariant is S(M) := \ ω Sp(∆ω)⊂[0,∞), the intersection taken over all faithful normal states. Connes proved that S ( M )depends only on M and characterizes the type III subclasses [73, 74]: Type S(M) III0{0,1} IIIλ(0 <λ<1) {0}∪{λn:n∈Z} III1[0,∞) Equivalently, one may use the flow of weights on the core MoσωR to define the T -invariant; for factors these descriptions agree up to natural identifications [114, 115]. Centers and direct integrals. If M=R⊕ XMxdµ(x)with factor fibers Mx, then by Theorem X.2, Sp(∆ω) = ess [ x∈X Sp(∆ωx), and hence S(M) = \ ω ess [ x Sp(∆ωx) = \ measurable fields {ωx} ess [ x Sp(∆ωx).(52) In particular, if Mx are a.e. of type III1 then S ( M ) = [0 ,∞ ); if Mx are a.e. of type IIIλ , then S ( M ) = {0}∪{λn}. QFT locality and type III1 .In relativistic QFT, under standard hypotheses (covariance, spectrum condition, existence of the vacuum vector, additivity), local algebras A ( O )are factors of type III1 [ 116 , 117 ]. Intuitively, the abundance of short-distance degrees of freedom and the Reeh–Schlieder property force absence of minimal projections and a continuous modular spectrum, yielding S ( A ( O )) = [0 ,∞ ). In our gauge-theory setting, Proposition III.2 asserts that the fiber algebras Ax ( R )are factors; in continuum limits they are expected to be of type III1generically, hence so is A(R)modulo its center. E. Bisognano–Wichmann property and Unruh/KMS Let W be a Rindler wedge in Minkowski space and A ( W )the associated algebra in the vacuum representation. The Bisognano–Wichmann (BW) theorem states that the modular group of ( A ( W ) , Ω) coincides with Lorentz boosts preserving W[118, 119]: ∆it Ω=U(ΛW(2πt)), σΩ t(A) = U(ΛW(2πt)) A U(ΛW(−2πt)). Therefore Ω A(W) is KMS at inverse temperature β = 2 π with respect to the boost generator, i.e. the Unruh temperature in suitable units [ 79 ]. In our language, if A ( W )has a nontrivial center generated by boundary data on ∂W (e.g. in gauge theory), Proposition X.3 implies the BW modular flow acts trivially on that center and fiberwise as boosts on the factor fibers. 39 F. Modular Hamiltonians, relative entropy, and JLMS splits Given a faithful normal state ω on M , the modular Hamiltonian (relative to ω ) is Kω := −log ∆ ω . For ω, ϕ faithful, the relative modular Hamiltonian is Kω|ϕ := −log ∆ ω|ϕ . By Theorem X.2 and Theorem X.4, Kω=Z⊕ X Kωxdµ(x), Kω|ϕ=Z⊕ X Kωx|ϕxdµ(x), and S(ωkϕ) = hΩω, Kω|ϕΩωi=ZX hΩωx, Kωx|ϕxΩωxidµω(x). In gravitational applications (Section IX), the modular Hamiltonian of the center includes the area/Wald charge on the gluing surface; the fiberwise piece is the bulk modular Hamiltonian. This is the direct-integral statement behind the JLMS identity (49). G. Remarks on the core and flow of weights For completeness we recall the Connes–Takesaki core construction: For a faithful normal state ω , the crossed product Cω:= MoσωR is a semifinite von Neumann algebra with a canonical faithful semifinite normal trace τω . The flow of weights θt is the dual action on Cω ; its orbit structure (up to conjugacy) is an invariant of M independent of ω and yields Connes’ T -invariant [ 74 , 114 ]. In the direct-integral situation M = R⊕ XMxdµ ( x ), one has Cω∼ =Z⊕ X (MxoσωxR)dµ(x), and the flow θt decomposes fiberwise. Thus the classification reduces to the factor fibers, in agreement with (52). H. Summary of Section X • Modular objects ( Sω, Jω, ∆ ω )and modular flow σω t fix the center and decompose fiberwise on the central direct integral (Theorems X.2 and Proposition X.3). • Relative modular operators and Araki relative entropy also decompose fiberwise (Theorem X.4), yielding the classical+quantum entropy split we used earlier. • If the dynamics αt leave the center fixed, KMS states factor into a classical distribution on labels and fiberwise KMS states (Theorem X.5). • Connes’ S -invariant and flow of weights reduce to those of the factor fibers; in QFT, local fibers are typically type III1, hence S= [0,∞)(Section X D). • The Bisognano–Wichmann property fits seamlessly: modular flow equals boosts on each fiber, while the center is untouched (Section X E). XI. CONCRETE MODELS AND COMPUTATIONS This section makes the abstract statements of Sections III–X fully explicit in representative models. We treat (i) lattice U (1) gauge theory (Hamiltonian/Kogut–Susskind) and identify the center, its spectral projectors, and a canonical QND measurement; (ii) finite non-Abelian gauge theories, where the center on a boundary loop is generated by class sums and the superselection labels are the Drinfel’d doubles ( C, α ); (iii) continuum Maxwell theory, where the normal electric flux through ∂R and suitable loop data form 40 the boundary center; and (iv) two deformations introduced in Section VII: a unitary crossed-product (flux-raising operator) and a CP/Lindblad birth–death edge channel. We close with compact worked examples. Throughout, Rb Σis a regular region with (piecewise smooth) boundary ∂R ; A ( R )denotes the strictly interior, gauge-invariant von Neumann algebra, and Z ( A ( R )) ∼ =L∞ ( XR, µR )its center. We keep the global notations and conventions of earlier sections. A. Lattice U(1): center, flux projectors, and QND measurement Hilbert space, CCR, and constraints. Place a cubic lattice Λ ⊂ Σwith oriented links ` and vertices v . The Hilbert space is H = N`L2 U (1) , dθ . On each link ` we write the unitary “link” U` = eiθ` and the self-adjoint electric field E`=−i ∂θ`, obeying [E`, U`0] = δ``0U`,[E`, U∗ `0] = −δ``0U∗ `0,[E`, E`0]=[U`, U`0]=0.(53) Gauge transformations gv=eiαvact by U`7→ gs(`)U`g−1 t(`)and E`7→ E`; the Gauss constraint at vis Gv:= X `→v E`−X `←v E`−ρv≡0,(54) with ρv the (integer) matter charge at v (if present). The gauge-invariant observables are generated by electric energies E2 `, magnetic plaquettes Bp= Arg Q`∈∂p U`, and Wilson loops. Strict interior algebra. Fix Rb Σand let Λ R be the subcomplex of links and plaquettes strictly contained in R (i.e. not touching ∂R ). The strictly interior gauge-invariant algebra A ( R )is generated by {E2 ` : `⊂ Λ R} and by Wilson loops/plaquettes supported in Λ R . Define the boundary electric flux through ∂R by Φ∂R := X `t∂R ε(`, ∂R)E`,(55) where the sum runs over links crossing the boundary once transversely and ε ( `, ∂R ) = ± 1is the outward normal sign. Φ∂R is integer-valued on the physical subspace (lattice Gauss law). Proposition XI.1 (Centrality of boundary flux (lattice U (1))) . Φ ∂R commutes with all strictly interior, gauge-invariant observables: [ Φ∂R,A(R) ] = 0. Hence the von Neumann algebra generated by Φ∂R lies in Z(A(R)). Proof. Every generator O∈ A ( R )is supported on links and plaquettes strictly inside R , hence disjoint from the boundary-crossing links in (55) . Using (53) , E` commutes with U`0 and E2 `0 unless ` = `0 , and by disjointness there is no overlap; thus [Φ ∂R, O ]=0. Gauge-invariant completions do not reintroduce boundary support, so centrality holds. Spectrum and projectors. Let XR = Z index eigenvalues of Φ ∂R . Spectral projectors {Pq}q∈Z resolve the identity on Hand belong to Z(A(R)). The central decomposition of Section III specializes to A(R)∼ =M q∈Z Aq(R), ωA(R)=X q∈Z pqωq, pq=ω(Pq).(56) Each fiber Aq(R)is a factor (Proposition III.2); in the continuum limit it is typically type III1. QND instrument and entropy split. Define the flux measurement instrument Iq(O) := PqO Pq, O ∈ A(R).(57) By Proposition VI.4, {Iq} is QND and repeatable for the central variable Φ ∂R , and the posterior state inside each sector is unchanged: ωpost q=ωq. For density matrices (finite-volume), ρR=M q pqρq, S(ρR) = H({pq}) + X q pqS(ρq), as in (19). 41 B. Unitary deformation: explicit flux-raising operator and crossed product We now realize the unitary label-mixing mechanism of Section VII D 1 explicitly on the lattice. Open Wilson line to the boundary. Let γ be a path from a vertex v∈R to a boundary vertex ˜v∈∂R , and define the dressed open Wilson line Wγ:= Y `∈γ U`,(58) with an implicit dressing at ˜v that renders Wγ invariant under gauge at v but charged under the boundary edge mode at ˜v (the edge dressing lives in the collar of ∂R ). Physically this creates a unit U (1) charge at vwhose electric string exits through the boundary. Proposition XI.2 (Flux-raising commutator on the lattice).Let U:= Wγas in (58). Then UΦ∂R U∗= Φ∂R + 1,i.e. [ Φ∂R, U ] = U. Proof. Wγ multiplies by eiθ` on each link along γ with orientation sign. Commuting E` through U` via (53) adds +1 on the unique boundary-crossing link where γ exits R , leaving all strictly interior contributions unchanged. Crossed-product algebra and center. Let α ( f ) = f◦T−1 on L∞ ( Z )with ( Tq ) = q + 1. Adjoining U to A(R)yields the crossed product Amix(R) := M q Aq(R)!oαZ, with UPqU∗ = Pq+1 and U OqU∗ = Oq+1 for Oq∈ Aq ( R ). By Proposition VII.1, the center collapses to the Z-invariants: ZAmix(R)=C1(ergodicity of T). C. Finite non-Abelian G: class sums, central projectors, and (C, α)labels Edge Hilbert space and operators. For a finite group G , the link Hilbert space is C [ G ]with orthonormal basis {|gi}g∈G . Left/right regular actions Lh|gi = |hgi , Rh|gi = |gh−1i implement gauge and electric operators. Plaquette (magnetic) operators project the holonomy around a face to the identity. Boundary loop algebra and class functions. Let Γ ⊂∂R be a simple loop (boundary-parallel). The holonomy operator UΓ acts by |gi 7→ |h−1ghi under gauge at Γ; hence class functions in the group algebra are gauge invariant on Γ. For a conjugacy class C⊂G, define the class sum zC:= X g∈C g∈C[G].(59) These generate the center ZC[G]. Proposition XI.3 (Action of class sums in irreps) . For any irreducible representation ρ : G→GL ( Vρ ) with character χρand dimension dρ, ρ(zC) = |C|χρ(C) dρ 1Vρ, where χρ(C)denotes the common value of χρ(g)for g∈C. Proof. By Schur’s lemma, ρ ( zC )is a scalar multiple of the identity because zC is central. Taking the trace gives Tr ρ(zC) = Pg∈Cχρ(g) = |C|χρ(C), hence the scalar equals |C|χρ(C)/dρ. Boundary center and sector labels. Let A ( R )be the strictly interior, gauge-invariant algebra. Then Z ( A ( R )) is generated by the class functions in the boundary loop algebra (magnetic data) together with electric Casimirs associated with the stabilizers of punctures on ∂R . In Kitaev’s quantum double D( G ) picture, superselection sectors are labeled by (C, α), C ⊂Gconjugacy class, α ∈c Zg(irrep of the centralizer of g∈C). The pair ( C, α )plays the role of x∈XR in the central decomposition A ( R ) = L(C,α)AC,α ( R ), with spectral projectors built from the minimal central idempotents of C [ G ](via character orthogonality; see e.g. [128]). 48 B. Boundary maps and the pullback radiation algebra Let R∞ denote the radiation region at infinity and IΓ the bulk region bounded by Γ(the candidate island). There are normal unital ∗-homomorphisms (boundary reports) φR∞:A(R∞)→Z(A(Γ)), φI:A(IΓ)→Z(A(Γ)), sending each strictly interior observable to its induced edge value on Γ(determined by the gravitational constraints in the collar N (Γ)). Precisely, they are obtained by restricting the action of O∈ A ( R )on states to the center via the conditional expectation onto Z(A(Γ)).[133] Theorem XIII.3 (Radiation algebra with island is a pullback) . Let Γbe a candidate QES. Then the von Neumann algebra of the radiation including the island is the equalizer Arad(Γ) := A(R∞)×Z(A(Γ)) A(IΓ) = {(a, b)∈ A(R∞)⊕ A(IΓ) : φR∞(a) = φI(b)}.(63) Moreover, under the central decomposition Z(A(Γ)) ∼ =L∞(XΓ, µΓ)one has the direct integral Arad(Γ) ∼ =Z⊕ XΓ (Ax(R∞)¯ ⊗ Ax(IΓ)) dµΓ(x).(64) Proof. The equalizer construction is identical to Proposition IX.2, now supplied with the explicit φ ’s. The direct–integral form follows from the central decomposition of Z ( A (Γ)) and Theorem XII.2: the pullback over an abelian algebra decomposes into the fiberwise tensor products over each label x∈XΓ. Uniqueness. Arad (Γ) is unique up to isomorphism as a limit in the 1–category vNAlg ; any other algebra satisfying the same universal property is canonically isomorphic to (63). C. Generalized entropy and the QES extremality equation Let ρ be a normal state on the global algebra, and consider its restriction to Arad (Γ). By the direct–integral form (64) and Theorem X.4, the von Neumann entropy splits as SρArad(Γ)=H(µρ Γ) + ZXΓ SρxAx(R∞)¯ ⊗Ax(IΓ)dµρ Γ(x),(65) where µρ Γ is the probability measure on XΓ induced by ρ (its central decomposition), and ρx are the fiber states. Theorem XIII.4 (QES extremality from the center) . Let Γvary smoothly in a family Γ s with s∈ ( −,  ), and suppose ρ varies within a semiclassical code subspace where the saddle point approximation holds. Then stationarity of the generalized entropy, d dss=0 Hµρ Γs+ZS(ρx,s)dµρ Γs(x)!= 0, is equivalent to the QES condition δhA(Γ)iρ 4G~+δSbulkρA(R∞)∨A(IΓ)= 0 for first variations δinduced by the deformation of Γ. Idea. In a semiclassical saddle, the central distribution has density dµρ Γ ( x ) ∝e−Ieff (Γ;x)dν ( x ), where Ieff contains the area/Wald functional plus local counterterms. Then H ( µρ Γ ) = hIeffi + const + . . . , and its first variation yields δhA (Γ) i/ (4 G~ )by the Iyer–Wald variational identity. The fiber entropy term varies by the entanglement first law (relative entropy linearization), producing the δSbulk piece. Stationarity gives the QES equation. A fully rigorous derivation can be organized by replacing S with relative entropy against a reference KMS state and invoking Theorem X.4 plus the first law (see also Section X). Physics. The “area term” is thus the classical, center–valued contribution (the Shannon layer), while the bulk matter entropy is the quantum contribution inside fibers. Extremizing the sum is the QES prescription. 49 D. JLMS as a direct–integral relative–entropy identity Let ρ, σ be faithful normal states on Arad(Γ) in a common code subspace. By Theorem X.4, Srel(ρkσ;Arad(Γ)) = ZXΓ Srel(ρxkσx;Ax(R∞)¯ ⊗Ax(IΓ)) dµρ Γ(x) + Srel(µρ Γkµσ Γ). Identifying the center relative entropy with the Wald/area modular contribution (the difference of expectation values of the center modular Hamiltonian) yields: Theorem XIII.5 (Direct–integral JLMS).For ρ, σ as above, Srel(ρkσ;Arad(Γ)) = hA(Γ)iσ− hA(Γ)iρ 4G~+ZXΓ Srel(ρxkσx)dµρ Γ(x).(66) Proof. The center piece equals the difference of the center modular Hamiltonians’ expectation values; by the covariant phase space analysis, this is the Wald charge evaluated on Γ(area term). The fiber piece is exactly the bulk relative entropy inside Ax(R∞)¯ ⊗Ax(IΓ). Summing gives (66). E. Replica wormholes as center conditioning Consider the n –replicated path integral for Tr ρn of the radiation state. The gravitational path integral admits saddles where the replicas are connected through a common QES Γ, glued along their copies of Γ. In the present formalism, this corresponds to imposing a common center label x∈XΓ across the replicas. Proposition XIII.6 (Renyi entropies as center averages) . Let ρ = Rρxdµ ( x )be the central decomposition on Arad(Γ). Then formally, Tr ρn=ZXΓ Tr ρn xdνn(x), for a sequence of measures dνn concentrating on the dominant saddle(s) in the semiclassical limit. The transition from “no–island” to “island” dominance at Page time is a transfer of weight in dνn from one region of XΓto another. Idea. In the saddle–point evaluation, contributions factorize across center labels; conditioning all replicas to share the same x enforces the wormhole gluing. The measure dνn includes the classical action (area) and one–loop determinants. Taking n→1recovers (65) and the QES extremality. F. Causality, monotonicity, and energy conditions Causality. The implementers of the pullback condition (boundary reports and conditional expectations) are supported in N (Γ), hence commute with algebras of spacelike–separated regions (Theorem VIII.5). Therefore, the island gluing does not allow superluminal signaling. Monotonicity. For inclusions of radiation regions R∞⊂R0 ∞ with compatible QES choices, the inclusion of pullback algebras induces monotonicity of relative entropy (data processing). This follows from the functoriality of the central decomposition and Theorem X.4. Energy conditions and extremality. The first variation in Theorem XIII.4 is consistent with the quantum focussing/energy conditions, which enforce the sign of the second variation and stability of the extremum. In the present language, this is the positivity of the center relative entropy plus the convexity properties of the bulk relative entropy under deformations normal to Γ. G. Deformations of the edge and modified island mechanics Section VII introduced deformations that allow interior operations to act on the center (unitary crossed products) or to mix it stochastically (boundary CP channels). Transposed to the QES: 50 •Unitary label mixing. If the collar algebra admits a covariance unitary implementing a nontrivial automorphism of Z ( A (Γ)), the center reduces to its invariants and the classical area–like term is correspondingly reduced (often to constants). The radiation algebra becomes a pullback over the noncommutative boundary algebra Z(A(Γ)) o Z (Theorem XII.4). •CP edge channels. A boundary–localized GKSL semigroup {EΓ t} induces a Markov semigroup on Z ( A (Γ)) and inter–fiber CP maps; the homotopy equalizer strictifies via Stinespring/Evans–Hudson (Theorem XII.5). Relative entropy is nonincreasing (Spohn monotonicity), so generalized entropy cannot be driven above the QES extremum by such dynamics. H. Example: JT gravity In JT gravity, the QES reduces to a point with dilaton value Φ(Γ). The center variable on Γis precisely Φ, and the “area term” is Φ(Γ) / (4 G~ ). The pullback algebra Arad (Γ) is built by equalizing over Z ( A (Γ)) ∼ =L∞ ( R+ )(dilaton values). The direct–integral formula (64) becomes an integral over Φ. The generalized entropy Srad(Γ) = Hµρ Γ+ZSbulk(Φ) dµρ Γ(Φ) reduces, in the semiclassical saddle, to Φ(Γ) / (4 G~ ) + Sbulk , reproducing the standard JT island formula. Replica–wormhole dominance corresponds to concentration of µρ Γnear the saddle value of Φ. I. Summary of Section XIII • The gravitational edge algebra at a candidate QES is central relative to strictly interior diffeo–invariant observables (Proposition XIII.2). •The radiation algebra with an island is the pullback (63) and decomposes fiberwise as in (64). • The QES extremality equation is the stationarity of the direct–integral entropy split, with the area/Wald term as the classical center piece and the bulk matter entropy as the fiber piece (Theorem XIII.4). •JLMS is the direct–integral identity (66) for relative entropy (Theorem XIII.5). • Replica wormholes implement center conditioning across replicas; the Page transition is a shift of weight in the center distribution (Proposition XIII.6). • Edge deformations (unitary/CP) translate to crossed–product or dilated pullbacks; causality and monotonicity persist. XIV. PREDICTIONS, DIAGNOSTICS, AND ANALOGUES This section turns the structural results of Sections III–XIII into concrete operational diagnostics and testable predictions in gauge theories, quantum simulators, condensed-matter analogues, and the gravitational (island) setting. The leitmotif is that the reduced state on a region R is classical over boundary labels and quantum within fibers: ρR=Z⊕ XR pR(x)ρR,x dµR(x), Z(A(R)) ∼ =L∞(XR, µR), so entropies and correlation measures split into classical (edge) and quantum (bulk) contributions. We systematize this into a set of invariants, derive inequalities, and propose protocols to measure them in lattice gauge-theory simulators and topological phases, and we extract sharpened predictions for islands. 51 A. Operational invariants at a cut: edge &bulk Let Σ = R∪Rcwith common boundary ∂R, and write ρR Rc=Z⊕ X p(x)ρxdµ(x), ρxa state on Ax(R)¯ ⊗Ax(Rc), where X = Spec Z ( A ( ∂R )). Denote by S ( · )the von Neumann entropy and by H ( · )the Shannon entropy. Proposition XIV.1 (Entropy and mutual information splits).With the above notation, one has: S(ρR) = H(p) + ZX S(ρR,x)dµp(x),(67) S(ρRc) = H(p) + ZX S(ρRc,x)dµp(x),(68) S(ρR Rc) = H(p) + ZX S(ρx)dµp(x),(69) and therefore the mutual information decomposes as I(R:Rc)ρ:= S(ρR) + S(ρRc)−S(ρR Rc) = H(p) + ZX IR:Rcρxdµp(x).(70) Proof. The entropy identities are those of Section III applied to R , Rc , and R∪Rc , using the direct-integral form of Aand ρ; (70) is the telescoping difference of the three splits. Invariants. We define three operational invariants across ∂R: •Edge entropy Hedge(R) := H(p)(classical label uncertainty). •Bulk entanglement Ebulk(R) := RXS(ρR,x)dµp(x). •Edge-assisted correlation Iedge(R) := Hedge(R), i.e. the purely classical piece of I. In the strict Gauss theory these are intrinsic; under deformations (Sections VII and VIII) they evolve according to the induced boundary action (unitary or CP). B. Tomography of the edge and of fibers We now present minimal protocols to extract Hedge,Ebulk, and to witness label mixing. Protocol A: QND edge tomography. Measure the central projections {PE}E⊂X associated with a finite measurable partition P= {Ek} of X by the QND instrument IEk ( O ) = PEkOPEk (cf. (57) ). Empirical frequencies ˆp ( Ek )converge (Hoeffding) to p ( Ek )with sample complexity O ( ε−2log (1 /δ )). Estimating over a refining sequence Pngives H(p)up to controlled bias. Protocol B: Fiber tomography (conditional). Post-select on Ek (or apply classical post-processing on the outcomes of Protocol A), and perform standard tomography on Ax ( R )within that sector (to the extent allowed by experimental access) to estimate S ( ρR,x )averaged over x∈Ek . Summing with weights ˆp(Ek)approximates Ebulk(R). Protocol C: Unitary label-mixing witness. Assume a covariance unitary U exists with UPxU∗ = PT x (e.g. the flux-raising operator of Section XI B). Prepare a Ramsey sequence (i) QND Px,(ii) U, (iii) bulk phase V(φ) (fiber unitary),(iv) U∗,(v) QND Px, and record the oscillatory dependence of the return probability on φ . Nontrivial fringes witness off-diagonal coherences across labels generated by U; in the strict theory the signal is flat. 52 Protocol D: CP edge-channel identification. With a stationary preparation ρ , repeat Protocol A at times t and t + ∆ t ; the empirical transition counts between label bins estimate the generator of the Markov semigroup on X , see (61) . The classical entropy production rate obeys the Spohn inequality (below), giving a consistency check. Proposition XIV.2 (Spohn inequality for boundary GKSL semigroups) . Let {Et}t≥0 be a normal UCP semigroup on A ( R )with GKSL generator and suppose ϕ is faithful and stationary ( ϕ◦ Et = ϕ ). Then for any normal state ω, d dt Sω◦ Etϕ≤0. If, in addition, Et acts on Z ( A ( R )) ∼ =L∞ ( X )as a Markov semigroup with stationary law π , then the classical divergence D ( ptkπ )is nonincreasing and the bulk conditional relative entropy RS ( ωt,xkϕx ) dµωt ( x ) is nonincreasing when the inter-fiber maps are CP-unital (bistochastic). Sketch. The first claim is Spohn’s theorem for quantum dynamical semigroups. The classical piece is the data-processing inequality for Markov kernels; the bulk piece uses monotonicity of relative entropy under CPTP maps on each fiber. C. Channel-theoretic view: edge capacities and bounds Consider using boundary operations to communicate between R and Rc . In the strict theory, only the classical label x∈X can carry information; in the unitary deformation, coherent control of U allows interference across labels; in the CP deformation, label evolution acts as a classical noisy channel. Strict theory (classical label channel). Let a sender choose x with distribution q ( x )by local operations on her side (e.g. prepare charge sectors) and a receiver perform Protocol A on the other side. The single-use classical capacity is bounded by Cedge ≤log NX, with NX the cardinality of effectively accessible labels (finite partitions in the continuum). The bound is saturated when sectors are prepared and measured sharply and the prior is uniform. Unitary mixing. If T decomposes X into orbits Oj of sizes Lj and the receiver can coherently undo/compose U ’s, the noiseless classical capacity is log Lmax per use on the quotient X/hTi (Section VIII C). This is realized by encoding in the phase accumulated around the T-cycle. CP mixing. Let P be the one-step Markov kernel and suppose the sender can choose initial labels. Then the classical capacity is upper bounded by the Shannon capacity of P: Cedge ≤C(P) := sup q IX0;X1, and the per-step achievable rate is limited further by dynamical constraints (e.g. detailed balance). The Spohn inequality implies that using the channel cannot increase the receiver’s knowledge relative to stationarity. D. Cold-atom and trapped-ion diagnostics (lattice U(1)) We now translate Protocols A–D to standard quantum-simulation platforms. Flux QND measurement (Protocol A). In U (1) Kogut–Susskind implementations [ 101 – 104 ], electric flux through ∂R equals the sum of link electric fields crossing the boundary. In spin or bosonic encodings, E` is (up to constants) a number operator or spin projection and can be read out nondestructively. Binning the outcomes estimates p(q)and Hedge. Fiber tomography (Protocol B). Condition on flux sectors using QND readout and perform tomography (e.g. via local rotations and parity measurements) on strictly interior degrees of freedom to estimate S(ρR,q). Averaging yields Ebulk. Unitary label mixing (Protocol C). Implement U as an open Wilson line string to the boundary (a product of link operators with suitable dressing). A Ramsey sequence (string on/off with a bulk phase) yields fringes whose contrast detects off-diagonal coherences across q7→ q+ 1.[134] 53 CP edge channels (Protocol D). Engineer stochastic hopping of flux between boundary-crossing links (e.g. weak coupling to a tailored ancilla bath at the edge) to realize (61) . Repeated edge readout reconstructs the rates γ± by maximum likelihood; the decay of D ( ptkπ )verifies the classical Spohn inequality (62) (Section XI E). E. Condensed-matter analogues: topological order and fractons Toric code and string-net models. In Z2 toric code (and more generally in Levin–Wen string-nets [ 105 , 106 ]), the algebra of strictly interior observables in a region R has a center generated by loop operators parallel to ∂R. The label xenumerates anyon flux threading R. Then S(ρR) = H(p) + X x pxS(ρR,x), and H ( p )reproduces the classical contribution behind the topological entanglement corrections when sectors are mixed. Protocol A coincides with measuring loop operators; Protocol C amounts to braiding/loop interference. Subsystem codes and fractons. In fracton phases, rigid sub-dimensional constraints generate an extensive center on cuts aligned with foliation planes. Our decomposition predicts a large edge entropy H ( p )reflecting immobile charge patterns, with limited unitary mixing (small orbits of T ). CP mixing is naturally slow, set by constrained kinematics. F. Gravitational predictions and island diagnostics We translate the invariants to the QES setting of Section XIII. Edge entropy as area term. At a QES Γ, the edge label x∈XΓ encodes boundary data; in the semiclassical saddle, Hedge(Γ) ≃hA(Γ)i 4G~+· · · , so measuring the classical piece of Srad is tantamount to the area term. The fiber entropy is the bulk matter term inside R∞∪IΓ. JLMS as an operational equality. Equation (66) identifies the change in the area with the center modular contribution to relative entropy. Any process that changes Hedge must be accompanied by a compensating change in bulk relative entropy; Protocols A/B provide a way to “separate variables” in toy models or holographic codes. Edge channel capacity bounds on information recovery. If the QES edge admits only CP mixing (no unitary label coherences), information leakage from the island to radiation through the edge is bounded by the classical capacity of the induced Markov process on XΓ ; in particular, Spohn monotonicity ensures no overshoot beyond the QES extremum. Unitary mixing and coherent edge transport. If crossed-product unitaries exist at the QES, coherent edge transport can, in principle, realize Hayden–Preskill–type mirroring across Γrestricted to XΓ/hTi (Section VIII C); the invariant subalgebra quantifies the residual classical piece. G. Worked quantitative illustrations Two-sector U (1) truncation. With X = { 0 , 1 } and a Ramsey-accessible unitary that swaps the labels, ρR=p ρ0⊕(1 −p)ρ1, I(R:Rc) = H(p) + p I(ρ0) + (1 −p)I(ρ1). Interference visibility V ≤ kρ01k1is zero in the strict theory and positive under unitary mixing (ρ01 the off-diagonal block in the mixed algebra). CP mixing with symmetric rates gives pt = 1 2 + ( p0−1 2 ) e−2γt and H(pt)%log 2. Non-Abelian G = S3 .Edge labels ( C, α )(Section XI C) yield Hedge = Hp ( C, α )  ; class-sum eigenvalues in Proposition XI.3 are directly measurable loop observables. Unitary mixing exists along conjugacy-class permutations; CP mixing is a random walk on the class graph. 54 H. Practical checklist for implementations 1. Choose a cut and its center: Determine the boundary algebra Z ( A ( ∂R )) and its label set X (flux, conjugacy class, loop charge). 2. Edge readout (QND): Implement central projectors PE (loop or flux measurements) to estimate p(E)and Hedge. 3. Fiber access: Condition on Eand perform interior tomography to estimate Ebulk. 4. Dynamics: If strings to the boundary are available, implement unitary mixing and run Ramsey sequences (Protocol C); otherwise, engineer CP edge noise and fit a generator (Protocol D). 5. Consistency: Check Spohn monotonicity and the MI split (70) within experimental error bars. I. Summary of Section XIV • We introduced operational invariants at a bipartition: edge entropy Hedge , bulk entanglement Ebulk , and the MI split (70). • We proposed four diagnostic protocols (QND edge tomography, fiber tomography, unitary labelmixing witness, and CP edge-channel identification), with quantitative bounds (Hoeffding, Spohn). • We translated these to cold-atom and trapped-ion platforms (flux QND, strings, engineered edge noise). • We identified condensed-matter analogues (string-nets, fractons), where the same central decomposition controls entanglement across cuts. • In gravity, the edge entropy is the area term; JLMS becomes an operational identity, and edge capacities bound island information flow. XV. PREDICTIONS, DIAGNOSTICS, AND ANALOGUES (EXTENDED) In this section we turn the structural picture of Sections III–XIII into operational tools with quantitative guarantees. We (i) define edge/bulk information functionals tied to the central decomposition, (ii) prove split identities and continuity/monotonicity inequalities, (iii) analyze communication capacities mediated by the edge (label) degrees of freedom under strict, unitary-mixing, and CP-mixing dynamics, (iv) give statistically sound tomography/estimation protocols with sample-complexity bounds, and (v) present concrete implementations in lattice gauge simulators and topological phases, and sharpened predictions for islands. Throughout we use the notations fixed earlier. In particular, a bipartition Σ = R∪Rc with common boundary ∂R has center ZA ( ∂R ) ∼ =L∞ ( X, µ ), and any normal state on the global algebra disintegrates as ρR Rc=Z⊕ X p(x)ρxdµ(x), ρxa state on Ax(R)¯ ⊗ Ax(Rc).(71) A. Edge &bulk functionals: identities and convexity Define the edge entropy,bulk entropy, and edge-assisted mutual information by Hedge(R) := H(p), Ebulk(R) := ZX S(ρR,x)dµp(x), Iedge(R) := Hedge(R),(72) and, similarly, Ebulk(Rc)with ρRc,x. We will also use the fiber mutual information Ix:= I(R:Rc)ρx. 55 Theorem XV.1 (Split identities and convexity) . With (71) , the von Neumann entropies and mutual information split as S(ρR) = Hedge(R) + Ebulk(R),(73) S(ρRc) = Hedge(R) + Ebulk(Rc),(74) S(ρR Rc) = Hedge(R) + ZX S(ρx)dµp(x),(75) I(R:Rc)ρ=Hedge(R) + ZX Ixdµp(x).(76) Moreover, Ebulk ( R )and RXIxdµp ( x )are affine in the state ρ (convex-linear on mixtures), while Hedge is strictly concave in p. Proof. The entropy identities are the direct-integral split of Section III applied to R , Rc , and R∪Rc . Subtracting yields (76) . Affinity follows because S ( · )is affine on each factor fiber while integration against pis linear; His strictly concave on the simplex. Continuity under perturbations. Let k·k1 denote the trace norm (in finite volume) or the Araki–Masuda L1 -norm more generally. For two reduced states ρ, σ with disintegrations {p, ρx} and {q, σx} , define the edge total variation kp−qk1:= R|p−q|dµ and the average fiber distance δ:= Rp(x)kρx−σxk1dµ. Proposition XV.2 (Fannes–Audenaert type bounds for the split) . Assume finite-dimensional fiber algebras (or suitable energy cutoffs). Let dR := supxdim Ax ( R )and dRRc := supxdim Ax ( R∪Rc ). Then for kp−qk1≤ε≤1/e and δ≤δ≤1/e, Hedge(ρ)−Hedge(σ)≤εlog |X| ε, Ebulk(R;ρ)−Ebulk(R;σ)≤δlog dR+h2(δ), I(ρ)−I(σ)≤εlog |X| ε+δlog dRRc+ 2 h2(δ), where h2is the binary entropy. Sketch. The classical bound is a standard continuity inequality for Shannon entropy. The quantum pieces follow from the Audenaert refinement of Fannes’ inequality applied fiberwise and then integrated, plus a triangle inequality for the MI split. Details are routine once finite-dimensionality (or cutoffs) are in place. B. Monotonicity and data processing for edge/bulk Consider an inclusion of regions R⊂R0 , with corresponding label maps r : XR0→XR (restriction of boundary data). Let ρbe a global state. Theorem XV.3 (Isotony/data processing) . The edge distribution of R0 coarse-grains that of R : pR = Coarser(pR0). Consequently, Hedge(R)≤Hedge(R0), Ebulk(R)≤Ebulk(R0), and the classical divergence to any fixed reference label law contracts: D(pRkπR)≤D(pR0kπR0) whenever πRis the pushforward of πR0under r. Proof. The center homomorphisms compose contravariantly (Section XII), so the induced probability laws obey the usual law-of-the-iterated-expectation structure; Shannon entropy is monotone under (classical) channels, and the quantum part increases by isotony of local algebras and strong subadditivity. The divergence contraction is the data-processing inequality for the classical Markov kernel induced by r . Now let {Et}t≥0 be a boundary-local normal UCP semigroup as in Section XI E, with stationary faithful ϕ. 56 Theorem XV.4 (Split Spohn inequality).For any normal state ωand all t≥0, d dt Sω◦ Etϕ≤0. Writing ωt=ω◦ Et, its edge and bulk contributions obey d dt Dptkπ≤0,d dt ZX S(ωt)xϕxdµωt(x)≤0 whenever the inter-fiber maps are CP-unital. Proof. Spohn’s inequality for quantum dynamical semigroups gives the global claim. Projecting to the center yields the classical data-processing inequality for the induced Markov semigroup on X . The bulk monotonicity follows from monotonicity of relative entropy under CPTP maps, applied to each fiber. C. Edge-mediated communication: capacities and bounds We quantify how much information can be transmitted across a cut by operations that are strictly interior on each side, thus constrained to act through the boundary center. Strict Gauss theory (no mixing). Only label preparation/measurement is available. A single-shot protocol induces a classical channel q ( x ) 7→ y (a coarse-graining of x by the receiver’s POVM). The Holevo bound implies the accessible information is at most H(q); the capacity per use satisfies Cstrict edge ≤log Neff, with Neff the number of distinguishable labels (finite partitions if X is continuous). Equality is achieved by sharp sector preparation and QND measurement (Section XIV B). Unitary label mixing. Let T : X→X generate orbits Oj of sizes Lj . If sender and receiver can implement U and U∗ (covariance unitary) and perform QND edge readout, then for noiseless control the classical capacity is Cunitary edge = log max jLj,(77) achieved by coding along the longest cycle of T (Ramsey interferometry across labels). If bulk phases can be inserted coherently between U–steps, phase modulation across the cycle realizes the code. CP label mixing. Let P be the one-step Markov kernel induced on X . The per-use capacity is upper bounded by the Shannon capacity C ( P ) := supqI ( X0 ; X1 )of P . For a stationary ergodic chain, the asymptotic rate is bounded by the entropy rate constraints and detailed-balance limits. Spohn monotonicity yields that repeated use of the edge channel cannot create information beyond stationarity if πis enforced by the environment. D. Statistical diagnostics: consistent estimators and sample complexity We now provide end-to-end, finite-sample guarantees for the protocols. Protocol A (edge tomography). Fix a finite measurable partition P= {Ek}m k=1 of X and let Pk := 1Ek be the central projectors. Perform n i.i.d. QND measurements of {Pk} , producing counts Nk and empirical frequencies ˆpk=Nk/n. Then: Pr[ kˆp−pk1≥]≤2e−n2/2=⇒H(ˆp)−H(p)≤log m (78) with the quoted probability (Hoeffding/DKW plus continuity of H ). Thus to estimate Hedge within ±η with confidence 1−δit suffices to take n=O1 η2log m η·log 1 δ. Protocol B (fiber tomography). Condition on an edge bin Ek and perform tomography on A ( R ) restricted to that sector. Let b Sk be a consistent entropy estimator with mean-square error σ2 k/nk after nk sectoral samples (e.g. via classical shadows in finite-dimensional encodings). Then the plug-in estimator b Ebulk = Pkˆpkb Sk is unbiased up to O (1 /n )and concentrates with variance Var ( b Ebulk ) ≤ Pkpk(1−pk) nS2 k+p2 kσ2 k nk. Allocating nk∝pkminimizes the second term. 57 Protocol C (unitary-mixing witness). In a Ramsey sequence U−V ( φ ) −U∗ followed by QND edge readout, the return probability as a function of φ is Pret ( φ ) = Pxp ( x ) Tr(ρR,x Wx(φ)) with Wx ( φ )a fiber unitary derived from V(φ). The visibility V= maxφPret(φ)−minφPret(φ)obeys the bound V ≤ X O kρO,offk1, where ρO,off is the block of ρ connecting labels inside each T -orbit O (zero in the strict Gauss theory). Standard Bernstein/Hoeffding bounds control the estimation of Vfrom Bernoulli samples at each φ. Protocol D (CP-channel identification). From two-time edge histograms ( ˆp (0) ,ˆp (∆ t )), the maximumlikelihood estimator of the generator rates in the birth–death chain (61) is asymptotically normal with covariance given by the Fisher information matrix of the multinomial model; the plug-in estimate of D(ˆptkˆπ)converges almost surely and its decrease per step is a consistent test of Spohn monotonicity. E. Concrete implementations Cold atoms (optical lattices). Kogut–Susskind U (1) and finite-group gauge theories can be emulated by Rydberg arrays or bosonic/spin mixtures. The edge flux operator is the sum of link electric fields crossing ∂R (number operators/spinz ), readable QND via dispersive shifts. Open strings U are realized by products of local unitaries along a path to the boundary; compiled Trotter sequences implement Ramsey protocols. CP edge channels arise by engineered loss/drive on boundary links; their rates are extracted by repeated QND edge readouts (as in Protocol D). Trapped ions. Global Mølmer–Sørensen entangling gates dressed with local phases synthesize string operators. Edge flux is measured by local Z -basis readout along the boundary crossing; coherent label mixing is probed with Ramsey sequences on collective modes dressed by local rotations. Topological order. In toric-code/string-net models, center projectors are loop operators parallel to ∂R . Measuring them yields Hedge ; postselection enables fiber tomography inside R . Braiding/loop interference realizes Uin non-Abelian models, giving access to unitary label mixing across conjugacy classes. F. Gravitational diagnostics and bounds At a QES Γ(Section XIII), the edge label distribution µρ Γ encodes the classical part of the radiation entropy. In the semiclassical regime, Hedge(Γ) ≈hA(Γ)iρ 4G~+O(log ~−1), so operational separation of Hedge from the fiber entropy reconstructs the area term in principle. Edge capacities bound information release. Suppose the edge admits only CP mixing with stationary π (no coherent unitaries). Let Rt denote the rate at which mutual information between early radiation and late-time radiation grows. Then Rt≤CPt, bounded by the instantaneous Shannon capacity of the induced Markov kernel Pt on XΓ . In particular, if Pt is rapidly mixing to π , the additional information that can be extracted through the edge channel after the Page time is limited by the classical contraction D(ptkπ)↓(by Theorem XV.4). Coherent edge transport and Hayden–Preskill. If covariance unitaries exist, the noiseless capacity (77) over XΓ/hTi bounds the coherent information that can be mirrored across Γ. This identifies the portion of the Page-curve turnover attributable to coherent edge transport vs. bulk-fiber entanglement. G. Worked illustrations Two-bin edge with unitary mixing. Take X = { 0 , 1 } and T the swap. With initial p (0) = p , p (1) = 1 −p and pure fibers, I ( ρ ) = H ( p ). Implement U and a bulk phase V ( φ )that acts as eiφ on label 1fibers and trivially on label 0fibers. Then Pret ( φ ) = 1 − 2 p (1 −p )  1 −cos φ , so V = 2 p (1 −p ), peaking at p = 1 2 . Thus the interference directly measures the classical edge uncertainty in this toy. 64 C.3 Finite non-Abelian groups: class sums and idempotents Let C [ G ]be the group algebra. For a conjugacy class C⊂G , set zC = Pg∈Cg . Then zC generates Z(C[G]). Irreducible representations ρsatisfy ρ(zC) = |C|χρ(C) dρ1(Schur). Proposition C.3 (Minimal central idempotents).The elements eρ:= dρ |G|X g∈G χρ(g)g are minimal central idempotents of C[G], with eρeρ0=δρρ0eρand Pρeρ=1. Proof. Standard character orthogonality: 1 |G|Pgχρ ( g ) χρ0(g) = δρρ0 , and e2 ρ = eρ follows from the convolution structure. Physical meaning. The spectral projections of zC and eρ are QND edge measurements: they label magnetic flux sectors (C) and, when electric data are included, the full (C, α)superselection labels. Appendix D. CP semigroups on centers: generators, dilations, Spohn Relation to main text. This appendix elaborates Section XI E and the split Spohn inequality in Section XIV: GKSL generators localized at edges, Stinespring/Evans–Hudson dilations, and entropy production. D.1 GKSL form and edge-locality A (norm–)continuous semigroup {Et}t≥0 of normal UCP maps on M has generator of the GKSL form L(A) = i[H, A] + X jL∗ jALj−1 2{L∗ jLj, A}, with H = H∗ , {Lj} bounded (Lindblad, GKSL). Edge locality means H and Lj are supported in a collar of ∂R and commute with strictly interior gauge constraints. D.2 Stinespring and Evans–Hudson dilations Theorem D.1 (Stinespring).For a normal UCP map Φ : M→M , there exists a Hilbert space K , a normal representation π:M→B(K), and an isometry V:H → K such that Φ(A) = V∗π(A)V. Theorem D.2 (Evans–Hudson).For a (sufficiently regular) GKSL generator L there exists a Fock space F , a dilation von Neumann algebra c M = M¯ ⊗B ( F ), and an E0 -semigroup { Θ t} of endomorphisms with a conditional expectation E:c M→Msuch that Et=E◦Θt|M. D.3 Spohn inequality and split version Let ϕbe faithful stationary: ϕ◦ Et=ϕ. Theorem D.3 (Spohn).d dt Sω◦ Etkϕ≤0for any normal ω. Sketch. Differentiate the Umegaki relative entropy along Et , use the generator in GKSL form and operator convexity of x7→ xlog x; see [120]. In our decomposed setting M = RMx with center L∞ ( X )and edge-local Et , the induced classical Markov semigroup Pt on X satisfies D ( ptkπ )monotone, while the fiber maps are CPTP; hence the split monotonicities of Theorem XV.4 hold. Physical meaning. Boundary noise cannot increase distinguishability from stationarity either in the classical edge register or in the (average) quantum fibers. 65 Appendix E. Categorical background and 2-limits Relation to main text. This appendix supplies the categorical underpinnings of Section XII: Grothendieck construction, 2-pullbacks (homotopy equalizers) in bicategories, and strictification via crossed products and dilations. E.1 Grothendieck construction for a presheaf of labels Let R be the category of regions and X : Rop →Meas the boundary-label pseudofunctor. Its category of elements B has objects ( R, x )and morphisms i : ( R, x ) → ( S, y )if i : R ,→S and x = rS→R ( y ). The projection π : B → R is a Grothendieck fibration: for each i and ( S, y )there exists a cartesian lift (R, x)→(S, y). E.2 The bicategory W∗-Corr and 2-pullbacks Objects are von Neumann algebras; 1-cells are normal correspondences; 2-cells are adjointable bimodule maps; composition is Connes fusion. A homotopy equalizer (2-pullback) of parallel 1-cells f, g : A→C is an object E , a 1-cell e : E→A , and an invertible 2-cell η : f◦e⇒g◦e with a universal property [126, 127]. E.3 Strictification: crossed products and dilations If α is a boundary automorphism of the center B , the pseudo-cospan A ( R ) φR −−→ BφRc ←−− A ( Rc ), together with a 2-cell η : φRc⇒α◦φR , is strictified to a cospan over BoαZ . The 2-pullback equates to the strict pullback over the crossed product (Theorem XII.4). For a CP edge channel with Stinespring triple ( K, π, V ), the pseudo-cospan strictifies to a cospan over e B = π ( B ); the 2-pullback is the image under the conditional expectation E ( · ) = V∗ ( · ) V of the strict pullback (Theorem XII.5). For semigroups, Evans–Hudson supplies an endomorphic dilation b B. Physical meaning. The “gluing up to an action” at the boundary is encoded as a 2-cell. Strictification builds an extended boundary algebra where the gluing becomes ordinary equality—precisely mirror to adding strings to the boundary (unitary) or adding an environment (CP). Appendix F. Deferred proofs and notation index Relation to main text. This appendix gathers proofs that would have interrupted the flow, and a compact notation index for quick reference. F.1 Pullbacks of von Neumann algebras along normal maps Let φi:Ai→Bbe normal unital ∗-monomorphisms (i= 1,2). Define A1×BA2:= {(a1, a2)∈A1⊕A2:φ1(a1) = φ2(a2)}. Then A1×BA2 is a von Neumann subalgebra of A1⊕A2 , and for any C with normal ∗ -homomorphisms ψi : C→Ai satisfying φ1ψ1 = φ2ψ2 there exists a unique normal ∗ -homomorphism Ψ : C→A1×BA2 with πiΨ = ψi. Proof. Closedness under ultraweak topology follows since the equality constraint is ultraweakly closed in B. Universal property is immediate from the definition. 66 F.2 Continuum Maxwell: centrality via smearing Let Ei be the (tempered) operator-valued distribution of the electric field and Φ ∂R = lim↓0R∂REinidS . For F(f)smeared strictly inside R, [Φ∂R, F(f)] = lim ↓0Z∂RZ[Ei(x), Fµν(y)] ni(x)fµν (y)dSxd3y= 0, since the equal-time commutator is supported on the diagonal and collars are disjoint from supp f for small enough . F.3 Notation index Symbol Meaning A(R)Strictly interior, gauge-invariant von Neumann algebra in region R Z(A(R)) Center of A(R),L∞(XR, µR) Ax(R)Factor fiber at label x∈XR H,K,FHilbert space, dilation space, Fock space σω tModular automorphism group for state ω ∆ω, JωModular operator and modular conjugation S(ωkϕ)Araki relative entropy B,EBase and total categories in Grothendieck fibration W∗-Corr Bicategory of W∗-correspondences UCovariance (label-mixing) unitary/string EtBoundary-local UCP semigroup (edge CP channel) Φ∂R Boundary electric flux operator zCClass sum in [G]for conjugacy class C [1] R. Haag, Local Quantum Physics, 2nd ed., Springer, 1996. [2] H. Araki, Mathematical Theory of Quantum Fields, Oxford Univ. Press, 1999. [3] J. Dixmier, Von Neumann Algebras, North-Holland, 1981. [4] M. Takesaki, Theory of Operator Algebras I–III, Springer, 2002–2003. [5] S. Doplicher, R. Haag, J. E. Roberts, Local Observables and Particle Statistics I, Commun. Math. Phys. 23 (1971) 199. [6] S. Doplicher, R. Haag, J. E. Roberts, Local Observables and Particle Statistics II, Commun. Math. Phys. 35 (1974) 49. [7] H. Casini, M. Huerta, J. A. Rosabal, Remarks on entanglement entropy for gauge fields, Phys. Rev. D 89 , 085012 (2014). [8] W. Donnelly, L. Freidel, Local subsystems in gauge theory and gravity, JHEP 2016, 102 (2016). [9] D. Harlow, Jerusalem Lectures on Black Holes and Quantum Information, Rev. Mod. Phys. 88 , 015002 (2016). [10] G. Penington, Entanglement Wedge Reconstruction and the Information Paradox, JHEP 09 (2020) 002. [11] A. Almheiri, N. Engelhardt, D. Marolf, H. Maxfield, The entropy of bulk quantum fields and the entanglement wedge of an evaporating black hole, JHEP 12 (2019) 063. [12] S. Mac Lane, Categories for the Working Mathematician, 2nd ed., Springer, 1998. [13] M. Stone, Linear transformations in Hilbert space (1932); J. von Neumann, Über die Eindeutigkeit der Schrödingerschen Operatoren, Math. Ann. 104 (1931) 570–578. [14] V. G. Drinfeld, Quasi-Hopf algebras, Leningrad Math. J. 1(1990) 1419–1457. [15] D. E. Evans, Y. Kawahigashi, Quantum Symmetries on Operator Algebras, Oxford Univ. Press, 1998. [16] G. Lindblad, On the generators of quantum dynamical semigroups, Commun. Math. Phys. 48 (1976) 119–130. [17] P. A. M. Dirac, Lectures on Quantum Mechanics, Belfer Graduate School of Science, Yeshiva University, 1964. [18] M. Henneaux, C. Teitelboim, Quantization of Gauge Systems, Princeton Univ. Press, 1992. [19] J. B. Kogut, L. Susskind, Hamiltonian formulation of Wilson’s lattice gauge theories, Phys. Rev. D 11 (1975) 395. [20] K. G. Wilson, Confinement of quarks, Phys. Rev. D 10 (1974) 2445. [21] P. P. Kulish, L. D. Faddeev, Asymptotic conditions and infrared divergences in quantum electrodynamics, Theor. Math. Phys. 4(1970) 745. [22] M. Lavelle, D. McMullan, Constituent quarks and gauge invariance, Phys. Rept. 279 (1997) 1. [23] S. Sakai, C∗-algebras and W∗-algebras, Springer, 1971. 67 [24] R. V. Kadison, J. R. Ringrose, Fundamentals of the Theory of Operator Algebras, Vol. I: Elementary Theory, Amer. Math. Soc., 1997. [25] H. Araki, Relative entropy of states of von Neumann algebras, Publ. RIMS Kyoto Univ. 11 (1976) 809–833. [26] M. Ohya, D. Petz, Quantum Entropy and Its Use, Springer, 1993. [27] F. Borceux, Handbook of Categorical Algebra I: Basic Category Theory, Cambridge Univ. Press, 1994. [28] T. Leinster, Basic Category Theory, Cambridge Univ. Press, 2014. [29] P. T. Johnstone, Sketches of an Elephant: A Topos Theory Compendium, Vols. 1–2, Oxford Univ. Press, 2002. [30] R. Brunetti, K. Fredenhagen, R. Verch, General covariance and algebraic quantum field theory, Commun. Math. Phys. 237 (2003) 31–68. [31] S. Weinberg, The Quantum Theory of Fields, Vol. I: Foundations, Cambridge Univ. Press, 1995. [32] M. E. Peskin, D. V. Schroeder, An Introduction to Quantum Field Theory, Addison–Wesley, 1995. [33] R. F. Streater, A. S. Wightman, PCT, Spin and Statistics, and All That, Princeton Univ. Press, 2000. [34] R. M. Wald, Quantum Field Theory in Curved Spacetime and Black Hole Thermodynamics, Univ. of Chicago Press, 1994. [35] L. C. Evans, Partial Differential Equations, 2nd ed., Amer. Math. Soc., 2010. [36] E. H. Lieb, D. W. Robinson, The finite group velocity of quantum spin systems, Commun. Math. Phys. 28 (1972) 251–257. [37] D. Buchholz, Gauss’ law and the infraparticle problem, Phys. Lett. B 174 (1986) 331–334. [38] W. H. Zurek, Decoherence, einselection, and the quantum origins of the classical, Rev. Mod. Phys. 75 (2003) 715–775. [39] M. Schlosshauer, Decoherence and the Quantum-To-Classical Transition, Springer, 2007. [40] E. Joos, H. D. Zeh, C. Kiefer, D. Giulini, J. Kupsch, I.-O. Stamatescu, Decoherence and the Appearance of a Classical World in Quantum Theory, 2nd ed., Springer, 2003. [41] E. B. Davies, J. T. Lewis, An operational approach to quantum probability, Commun. Math. Phys. 17 (1970) 239–260. [42] M. Ozawa, Quantum measuring processes of continuous observables, J. Math. Phys. 25 (1984) 79–87. [43] P. Busch, P. J. Lahti, P. Mittelstaedt, The Quantum Theory of Measurement, 2nd ed., Springer, 1996. [44] P. Etingof, S. Gelaki, D. Nikshych, V. Ostrik, Tensor Categories, Amer. Math. Soc., 2015. [45] A. Joyal, R. Street, Braided tensor categories, Adv. Math. 102 (1993) 20–78. [46] J. C. Baez, A. D. Lauda, A Prehistory of n -Categorical Physics, in Deep Beauty, ed. H. Halvorson, Cambridge Univ. Press, 2011. [47] W. F. Stinespring, Positive functions on C∗-algebras, Proc. Amer. Math. Soc. 6(1955) 211–216. [48] V. Gorini, A. Kossakowski, E. C. G. Sudarshan, Completely positive dynamical semigroups of N-level systems, J. Math. Phys. 17 (1976) 821–825. [49] W. Arveson, Continuous analogues of Fock space, Mem. Amer. Math. Soc. 80 (1989) no. 409. [50] H. Spohn, Entropy production for quantum dynamical semigroups, J. Math. Phys. 19 (1978) 1227–1230. [51] H. Umegaki, Conditional expectation in an operator algebra, Tôhoku Math. J. 14 (1962) 177–181. [52] E. H. Lieb, M. B. Ruskai, Proof of the strong subadditivity of quantum-mechanical entropy, J. Math. Phys. 14 (1973) 1938–1941. [53] A. Frigerio, Stationary states of quantum dynamical semigroups, Commun. Math. Phys. 63 (1978) 269–276. [54] B. Kümmerer, Markov dilations on W∗-algebras, J. Funct. Anal. 63 (1985) 139–177. [55] A. Strominger, Lectures on the Infrared Structure of Gravity and Gauge Theory, Princeton Univ. Press, 2018. [56] V. Iyer, R. M. Wald, Some properties of Noether charge and a proposal for dynamical black hole entropy, Phys. Rev. D 50 (1994) 846–864. [57] R. M. Wald, A. Zoupas, A general definition of “conserved quantities” in general relativity and other theories of gravity, Phys. Rev. D 61 (2000) 084027. [58] W. Donnelly, A. C. Wall, Entanglement entropy of electromagnetic edge modes, Phys. Rev. Lett. 114 (2015) 111603. [59] S. Ryu, T. Takayanagi, Holographic derivation of entanglement entropy from AdS/CFT, Phys. Rev. Lett. 96 (2006) 181602. [60] V. E. Hubeny, M. Rangamani, T. Takayanagi, A covariant holographic entanglement entropy proposal, JHEP 07 (2007) 062. [61] T. Faulkner, A. Lewkowycz, J. Maldacena, Quantum corrections to holographic entanglement entropy, JHEP 11 (2013) 074. [62] N. Engelhardt, A. C. Wall, Quantum extremal surfaces: holographic entanglement entropy beyond the classical regime, JHEP 01 (2015) 073. [63] D. L. Jafferis, A. Lewkowycz, J. Maldacena, S. J. Suh, Relative entropy equals bulk relative entropy, JHEP 06 (2016) 004. [64] G. Penington, S. H. Shenker, D. Stanford, Z. Yang, Replica wormholes and the black hole interior, JHEP 03 (2022) 205. [65] A. Almheiri, T. Hartman, J. Maldacena, E. Shaghoulian, A. Tajdini, The entropy of Hawking radiation, Rev. Mod. Phys. 93 (2021) 035002. [66] D. N. Page, Average entropy of a subsystem, Phys. Rev. Lett. 71 (1993) 1291–1294. [67] A. Almheiri, X. Dong, D. Harlow, Bulk locality and quantum error correction in AdS/CFT, JHEP 04 (2015) 68 163. [68] X. Dong, D. Harlow, A. C. Wall, Reconstruction of bulk operators within the entanglement wedge in gauge–gravity duality, Phys. Rev. Lett. 117 (2016) 021601. [69] P. Hayden, J. Preskill, Black holes as mirrors: quantum information in random subsystems, JHEP 09 (2007) 120. [70] M. Takesaki, Theory of Operator Algebras I, Springer, 2002. [71] O. Bratteli, D. W. Robinson, Operator Algebras and Quantum Statistical Mechanics II, 2nd ed., Springer, 1997. [72] H. Araki, Relative entropy of states of von Neumann algebras, Publ. RIMS Kyoto Univ. 11 (1976) 809–833. [73] A. Connes, Une classification des facteurs de type III, Ann. Sci. Éc. Norm. Sup. 6(1973) 133–252. [74] A. Connes, M. Takesaki, The flow of weights on factors of type III, Tôhoku Math. J. 29 (1977) 473–575. [75] W. Driessler, On the type of local algebras in quantum field theory, Commun. Math. Phys. 53 (1977) 295–297. [76] R. Haag, Local Quantum Physics, 2nd ed., Springer, 1996. [77] J. J. Bisognano, E. H. Wichmann, On the duality condition for a Hermitian scalar field, J. Math. Phys. 16 (1975) 985–1007. [78] J. J. Bisognano, E. H. Wichmann, On the duality condition for quantum fields, J. Math. Phys. 17 (1976) 303–321. [79] G. L. Sewell, Quantum fields on manifolds: PCT and gravitationally induced thermal states, Ann. Phys. 141 (1982) 201–224. [80] J. B. Kogut, L. Susskind, Hamiltonian formulation of Wilson’s lattice gauge theories, Phys. Rev. D 11 (1975) 395–408. [81] J. B. Kogut, An introduction to lattice gauge theory and spin systems, Rev. Mod. Phys. 51 (1979) 659–713. [82] H. Casini, M. Huerta, J. A. Rosabal, Remarks on entanglement entropy for gauge fields, Phys. Rev. D 89 (2014) 085012. [83] W. Donnelly, L. Freidel, Local subsystems in gauge theory and gravity, JHEP 09 (2016) 102. [84] J.-P. Serre, Linear Representations of Finite Groups, Springer, 1977. [85] A. Y. Kitaev, Fault-tolerant quantum computation by anyons, Ann. Phys. 303 (2003) 2–30. [86] S. MacLane, Categories for the Working Mathematician, 2nd ed., Springer, 1998. [87] G. M. Kelly, Basic Concepts of Enriched Category Theory, Cambridge Univ. Press, 1982. [88] R. Street, Limits indexed by category-valued 2-functors, J. Pure Appl. Algebra 8(1976) 149–181. [89] S. Lack, A 2-categories companion, in Towards Higher Categories, Springer, 2010, pp. 105–191. [90] W. L. Paschke, Inner product modules over B∗-algebras, Trans. Amer. Math. Soc. 182 (1973) 443–468. [91] J.-L. Sauvageot, Produit tensoriel de Z-modules et applications, in Operator Algebras and their Connections with Topology and Ergodic Theory, Springer LNM 1132 (1985) 468–485. [92] D. E. Evans, R. L. Hudson, Perturbations of quantum diffusions, Rep. Math. Phys. 20 (1984) 331–354. [93] P. S. Muhly, B. Solel, Tensor algebras over C∗ -correspondences: representations, dilations, and C∗ -envelopes, J. Funct. Anal. 158 (1998) 389–457. [94] M. Skeide, Hilbert modules and applications in quantum probability, in Quantum Probability and White Noise Analysis XIII, World Scientific (2001) 155–238. [95] V. Chandrasekaran, E. E. Flanagan, K. Prabhu, Symmetries and charges of general relativity at null boundaries, JHEP 11 (2018) 125. [96] D. Harlow, J.-Q. Wu, Covariant phase space with boundaries, JHEP 10 (2020) 146. [97] G. Penington, Entanglement wedge of an evaporating black hole, JHEP 09 (2020) 002. [98] D. Marolf, H. Maxfield, Transcending the ensemble: baby universes, spacetime wormholes, and the order and disorder of black hole information, JHEP 08 (2020) 044. [99] C. Akers, N. Engelhardt, Simple holographic proof of the quantum null energy condition, Phys. Rev. D 100 (2019) 126006. [100] H. Spohn, Entropy production for quantum dynamical semigroups, J. Math. Phys. 19 (1978) 1227–1230. [101] D. Banerjee, M. Dalmonte, M. Müller, E. Rico, P. Stebler, U.-J. Wiese, P. Zoller, Atomic quantum simulation of dynamical gauge fields coupled to fermionic matter, Phys. Rev. Lett. 109 (2012) 175302. [102] E. A. Martinez, C. A. Muschik, P. Schindler, D. Nigg, A. Erhard, M. Heyl, P. Hauke, M. Dalmonte, T. Monz, P. Zoller, R. Blatt, Real-time dynamics of lattice gauge theories with a few-qubit quantum computer, Nature 534 (2016) 516–519. [103] E. Zohar, J. I. Cirac, B. Reznik, Quantum simulations of lattice gauge theories using ultracold atoms in optical lattices, Rep. Prog. Phys. 79 (2016) 014401. [104] U.-J. Wiese, Ultracold quantum gases and lattice systems: quantum simulation of lattice gauge theories, Ann. Phys. 525 (2013) 777–796. [105] M. A. Levin, X.-G. Wen, String-net condensation: A physical mechanism for topological phases, Phys. Rev. B71 (2005) 045110. [106] X.-G. Wen, Quantum Field Theory of Many-Body Systems, Oxford Univ. Press, 2004. [107] W. Hoeffding, Probability inequalities for sums of bounded random variables, J. Amer. Stat. Assoc. 58 (1963) 13–30. [108] K. M. R. Audenaert, A sharp continuity estimate for the von Neumann entropy, J. Phys. A 40 (2007) 8127–8136. [109] A. S. Holevo, Bounds for the quantity of information transmitted by a quantum communication channel, Probl. Inf. Transm. 9(1973) 177–183. 69 [110] I. Csiszár, J. Körner, Information Theory: Coding Theorems for Discrete Memoryless Systems, Cambridge Univ. Press, 2011. [111] P. Massart, The tight constant in the Dvoretzky–Kiefer–Wolfowitz inequality, Ann. Probab. 18 (1990) 1269–1283. [112] M. Tomamichel, Quantum Information Processing with Finite Resources, Springer, 2016. [113] M. M. Wilde, Quantum Information Theory, 2nd ed., Cambridge Univ. Press, 2017. [114] M. Takesaki, Theory of Operator Algebras I, Springer, 2002. [115] O. Bratteli, D. W. Robinson, Operator Algebras and Quantum Statistical Mechanics II, 2nd ed., Springer, 1997. [116] W. Driessler, On the type of local algebras in quantum field theory, Commun. Math. Phys. 53 (1977) 295–297. [117] R. Haag, Local Quantum Physics, 2nd ed., Springer, 1996. [118] J. J. Bisognano, E. H. Wichmann, On the duality condition for a Hermitian scalar field, J. Math. Phys. 16 (1975) 985–1007. [119] J. J. Bisognano, E. H. Wichmann, On the duality condition for quantum fields, J. Math. Phys. 17 (1976) 303–321. [120] H. Spohn, Entropy production for quantum dynamical semigroups, J. Math. Phys. 19 (1978) 1227–1230. [121] G. Lindblad, On the generators of quantum dynamical semigroups, Commun. Math. Phys. 48 (1976) 119–130. [122] V. Gorini, A. Kossakowski, E. C. G. Sudarshan, Completely positive dynamical semigroups of N -level systems, J. Math. Phys. 17 (1976) 821–825. [123] D. E. Evans, R. L. Hudson, Perturbations of quantum diffusions, Rep. Math. Phys. 20 (1984) 331–354. [124] W. L. Paschke, Inner product modules over B∗-algebras, Trans. Amer. Math. Soc. 182 (1973) 443–468. [125] J.-L. Sauvageot, Produit tensoriel de Z-modules et applications, in Operator Algebras and their Connections with Topology and Ergodic Theory, Springer LNM 1132 (1985) 468–485. [126] R. Street, Limits indexed by category-valued 2-functors, J. Pure Appl. Algebra 8(1976) 149–181. [127] S. Lack, A 2-categories companion, in Towards Higher Categories, Springer, 2010, pp. 105–191. [128] J.-P. Serre, Linear Representations of Finite Groups, Springer, 1977. [129] We return to KMS/modular aspects in detail in Section X. [130] See [124, 125] and Section XIIE. [131] We use the Paschke–Sauvageot GNS correspondence BHβB with cyclic vector ξβ ; the Stinespring representation is a concrete realization. [132] We write φfor the dynamical fields (metric and matter). [133] Equivalently, by composing the inclusion A ( R ) ,→ A ( R ) ∨ A ( N (Γ)) with the center–valued expectation EΓ:A(N(Γ)) →Z(A(Γ)). [134] Finite-depth realizations can be compiled by Trotterization in cold atoms; analogous strings exist in trapped ions through global Mølmer–Sørensen couplings dressed by local phases.