Entanglement-Orthogonal Gauge

• Entanglement-Orthogonal Gauge is a framework that separates entanglement-relevant degrees of freedom from gauge redundancy by employing specialized gauge-fixing and operator algebra selections.
• It utilizes methods such as diagonalizing target-space coordinates in matrix models and decomposing lattice gauge theories into U(1) structures to clarify entanglement calculations.
• This approach refines entanglement entropy studies by aligning mathematical cuts with physical sectors, ensuring that superselection sectors and edge modes are explicitly isolated.

“Entanglement-orthogonal gauge” denotes a family of closely related constructions in which gauge fixing, subsystem algebra selection, or projector orthogonality is chosen so that the degrees of freedom relevant to entanglement are isolated from gauge-redundant, hidden-sector, or superselection data. The most concrete usage is in adjoint matrix quantum mechanics, where the matrix associated with the target-space coordinate normal to a planar entangling surface is diagonalized, reducing the problem to a $U(1)$ lattice gauge theory on the complete graph of D-brane sites (Hampapura et al., 2020). Related formulations appear in algebraic quantum theory, where a gauge group acts only on the commutant and is therefore “orthogonal” to subsystem entanglement (Balachandran et al., 2019); in semiclassical emergence, where reduced density matrices must become approximately proportional to mutually orthogonal projectors (Kirklin, 2022); and in lattice gauge theory, where one diagonalizes boundary-flux sectors and fixes pure-gauge modes without mixing those sectors (Spalvieri et al., 22 Dec 2025).

1. Terminological scope and structural idea

The phrase is not attached to a single universal gauge prescription. In current usage, it refers to a structural requirement: entanglement should be organized in variables adapted to an entangling cut, while gauge redundancy is either fixed away, shifted into edge modes, or confined to a sector that does not modify the relevant reduced state. In that sense, “orthogonal” does not primarily mean orthogonality in the Hilbert-space inner product, but separation of entanglement-carrying data from gauge data, often through a block decomposition into superselection sectors or through a hidden-sector action that leaves subsystem expectations invariant (Hampapura et al., 2020, Balachandran et al., 2019, Kirklin, 2022, Spalvieri et al., 22 Dec 2025).

Context	Orthogonalized object	Representative construction
Matrix models	Target-space direction normal to the cut	$X_{(1)} \to \mathrm{diag}$
Lattice gauge theory	Boundary-flux sectors vs pure-gauge modes	Center diagonalization; tree gauge
Algebraic/GNS setting	Observable subsystem vs commutant	Gauge group commuting with modular data
Classical limit	Classical labels vs entangled multiplicities	$\rho_A \simeq \pi_A/N_A$, $\pi_A(i)\pi_A(j)\simeq 0$

A common thread is that entanglement is not defined solely by a formal partial trace over arbitrary variables. It depends on the choice of local operator algebra, the treatment of centers, and the way Gauss-law constraints or modular constraints partition the physical state space. This suggests that entanglement-orthogonal constructions are best understood as cut-adapted reorganizations of the physical algebra rather than as ordinary gauge choices in isolation.

2. Matrix-model realization: diagonalizing the normal coordinate

In gauged multi-matrix models describing groups of D-branes separated by a planar entangling surface, the entanglement-orthogonal gauge is implemented by choosing the entangling surface to be the hyperplane $Y^1=0$ and using the adjoint $U(N)$ gauge symmetry to diagonalize the matrix $X_{(1)}$ corresponding to the target-space coordinate $Y^1$ (Hampapura et al., 2020). The gauge condition is

$$F_{ab}[X]= (X_{(1)})_{ab}\,(1-\delta_{ab})=0,\qquad a,b=1\ldots N,$$

or, covariantly, one sets the off-diagonal part of $n^iX_{(i)}$ to zero for $X_{(1)} \to \mathrm{diag}$0.

Under the change of variables

$X_{(1)} \to \mathrm{diag}$1

the Jacobian produces the square of the Vandermonde determinant,

$X_{(1)} \to \mathrm{diag}$2

This determinant can be represented by adjoint Grassmann ghosts $X_{(1)} \to \mathrm{diag}$3, with gauge-fixing and ghost terms

$X_{(1)} \to \mathrm{diag}$4

Integrating out $X_{(1)} \to \mathrm{diag}$5 enforces $X_{(1)} \to \mathrm{diag}$6, while integrating out $X_{(1)} \to \mathrm{diag}$7 reproduces $X_{(1)} \to \mathrm{diag}$8. An equivalent formulation absorbs one power of $X_{(1)} \to \mathrm{diag}$9 into the path-integral measure and discards the integral over $\rho_A \simeq \pi_A/N_A$0, leaving a flat measure in the $\rho_A \simeq \pi_A/N_A$1.

The physical interpretation is direct. Starting from axial gauge $\rho_A \simeq \pi_A/N_A$2, states $\rho_A \simeq \pi_A/N_A$3 are invariant under time-independent $\rho_A \simeq \pi_A/N_A$4 rotations. Diagonalizing $\rho_A \simeq \pi_A/N_A$5 identifies the eigenvalues $\rho_A \simeq \pi_A/N_A$6 as D-brane positions in the $\rho_A \simeq \pi_A/N_A$7 direction, while the off-diagonal modes $\rho_A \simeq \pi_A/N_A$8 become massive “fast” modes with mass $\rho_A \simeq \pi_A/N_A$9. The planar cut at $\pi_A(i)\pi_A(j)\simeq 0$0 splits the eigenvalues into visible branes with $\pi_A(i)\pi_A(j)\simeq 0$1 and hidden branes with $\pi_A(i)\pi_A(j)\simeq 0$2. The reduced density matrix is then obtained by tracing over diagonal eigenvalues with $\pi_A(i)\pi_A(j)\simeq 0$3, off-diagonal hidden–hidden strings, and, in the electric-center prescription, the strings crossing the cut.

After gauge fixing, the off-diagonal fields that connect the two sides of the cut carry $\pi_A(i)\pi_A(j)\simeq 0$4 charges. If $\pi_A(i)\pi_A(j)\simeq 0$5 is the total number of string quanta on the link $\pi_A(i)\pi_A(j)\simeq 0$6, the visible Hilbert space and reduced density matrix decompose as

$\pi_A(i)\pi_A(j)\simeq 0$7

The von Neumann entropy then splits into a classical Shannon term and a quantum term,

$\pi_A(i)\pi_A(j)\simeq 0$8

In a two-brane toy model, the second Rényi entropy behaves as

$\pi_A(i)\pi_A(j)\simeq 0$9

showing that the dependence of the off-diagonal-string ground state on the brane separation $Y^1=0$0 already induces $Y^1=0$1 entanglement.

The residual gauge symmetry after diagonalization is $Y^1=0$2. The theory can then be reinterpreted as a $Y^1=0$3 gauge theory on the complete $Y^1=0$4-site lattice: vertices are the eigenvalues $Y^1=0$5, link variables are the off-diagonal modes $Y^1=0$6, and Gauss’ law imposes net zero charge at each node. In this formulation, choosing the cut between two subsets of vertices turns target-space entanglement into the standard problem of entanglement in lattice gauge theory with electric-center boundary conditions. This is the clearest and most explicit realization of an entanglement-orthogonal gauge in the literature.

3. Local operator algebras, centers, and sectorwise entanglement

A second line of development emphasizes that entanglement in gauge theory is controlled by the local operator algebra attached to a region, not by naïvely gauge-fixed canonical variables. In $Y^1=0$7-dimensional Maxwell theory in Coulomb gauge, the equal-time commutator

$Y^1=0$8

contains the longitudinal projector $Y^1=0$9, which is non-local in position space; correspondingly, the fields $U(N)$0 and $U(N)$1 are not local operators after imposing Dirac brackets (Yang et al., 2017). To recover locality one defines

$U(N)$2

for which

$U(N)$3

a strictly local derivative-of-delta commutator. The same local algebra can be expressed in terms of gauge-invariant fields $U(N)$4 and $U(N)$5.

Once a local canonical algebra is chosen, the entropy of a Gaussian ground state with trivial center is computed from the commutator matrix $U(N)$6 and correlators $U(N)$7 and $U(N)$8 through

$U(N)$9

On the lattice, one finds an expansion

$X_{(1)}$0

For the trivial-center Coulomb-gauge algebra $X_{(1)}$1, $X_{(1)}$2, in agreement with a single scalar degree of freedom. If one instead includes a non-trivial $X_{(1)}$3-center or $X_{(1)}$4-center, $X_{(1)}$5, matching the standard gauge-field logarithmic coefficient. In the Higgsed phase, a deformed local algebra built from tilded fields yields a logarithmic coefficient that is independent of the choice of center and approaches zero at large $X_{(1)}$6.

The same structural point appears in lattice gauge theory without a local tensor-product decomposition. For a region $X_{(1)}$7, the gauge-invariant local algebra $X_{(1)}$8 is the subalgebra of the raw link algebra commuting with all Gauss operators. Its center is generated by the Gauss operators in $X_{(1)}$9 and the boundary-normal electric fields $Y^1$0, so simultaneous diagonalization produces sectors labeled by boundary fluxes $Y^1$1 (Spalvieri et al., 22 Dec 2025). The physical Hilbert space decomposes as

$Y^1$2

For a density matrix $Y^1$3, the sectorwise reduced state is $Y^1$4, the operational entanglement is

$Y^1$5

and the full entropy including classical sector mixing is

$Y^1$6

A generalized LOCC theorem then states that no sequence of local gauge-invariant operations in $Y^1$7 and classical communication can increase $Y^1$8.

Within this framework, an entanglement-orthogonal gauge is a gauge fixing that first diagonalizes the center and then fixes the remaining pure-gauge variables without mixing different $Y^1$9-sectors. A convenient example is the radial tree gauge inside $$F_{ab}[X]= (X_{(1)})_{ab}\,(1-\delta_{ab})=0,\qquad a,b=1\ldots N,$$0: one picks a spanning tree and sets $$F_{ab}[X]= (X_{(1)})_{ab}\,(1-\delta_{ab})=0,\qquad a,b=1\ldots N,$$1 on its links. The remaining non-trivial holonomies are then around plaquettes and along the boundary. In that gauge, the boundary fluxes act as sector labels, the magnetic loop operators and local electric fields carry the physical entanglement within each block, and the pure-gauge tree variables are orthogonal in the sense that they commute with the sector-label operators and carry no entanglement accessible to $$F_{ab}[X]= (X_{(1)})_{ab}\,(1-\delta_{ab})=0,\qquad a,b=1\ldots N,$$2.

4. Modular and algebraic entanglement-orthogonality

In the algebraic approach, entanglement-orthogonality is formulated without assuming any underlying spatial tensor-product factorization. Let $$F_{ab}[X]= (X_{(1)})_{ab}\,(1-\delta_{ab})=0,\qquad a,b=1\ldots N,$$3 be a finite-dimensional $$F_{ab}[X]= (X_{(1)})_{ab}\,(1-\delta_{ab})=0,\qquad a,b=1\ldots N,$$4-algebra of observables and $$F_{ab}[X]= (X_{(1)})_{ab}\,(1-\delta_{ab})=0,\qquad a,b=1\ldots N,$$5 a faithful mixed state. The Gelfand–Naimark–Segal construction yields a Hilbert space $$F_{ab}[X]= (X_{(1)})_{ab}\,(1-\delta_{ab})=0,\qquad a,b=1\ldots N,$$6, a representation $$F_{ab}[X]= (X_{(1)})_{ab}\,(1-\delta_{ab})=0,\qquad a,b=1\ldots N,$$7, and a cyclic and separating vector $$F_{ab}[X]= (X_{(1)})_{ab}\,(1-\delta_{ab})=0,\qquad a,b=1\ldots N,$$8 such that $$F_{ab}[X]= (X_{(1)})_{ab}\,(1-\delta_{ab})=0,\qquad a,b=1\ldots N,$$9 (Balachandran et al., 2019). The full system is $n^iX_{(i)}$0, the observable subsystem is $n^iX_{(i)}$1, and the complementary subsystem is the commutant $n^iX_{(i)}$2.

Tomita–Takesaki theory then supplies the antilinear Tomita operator $n^iX_{(i)}$3, its polar decomposition $n^iX_{(i)}$4, the modular operator $n^iX_{(i)}$5, and the modular conjugation $n^iX_{(i)}$6, satisfying

$n^iX_{(i)}$7

If $n^iX_{(i)}$8 is the positive operator with $n^iX_{(i)}$9, the emergent gauge group is

$X_{(1)} \to \mathrm{diag}$00

equivalently the unitaries commuting with $X_{(1)} \to \mathrm{diag}$01 or stabilizing $X_{(1)} \to \mathrm{diag}$02. On $X_{(1)} \to \mathrm{diag}$03, $X_{(1)} \to \mathrm{diag}$04 acts by inner automorphisms leaving $X_{(1)} \to \mathrm{diag}$05 invariant. On the full algebra $X_{(1)} \to \mathrm{diag}$06, it induces quantum channels

$X_{(1)} \to \mathrm{diag}$07

with projectors $X_{(1)} \to \mathrm{diag}$08. Each $X_{(1)} \to \mathrm{diag}$09 is completely positive, trace preserving, and entropy increasing.

The entanglement-orthogonal content is that these operations do not affect subsystem $X_{(1)} \to \mathrm{diag}$10. For every $X_{(1)} \to \mathrm{diag}$11,

$X_{(1)} \to \mathrm{diag}$12

so the states $X_{(1)} \to \mathrm{diag}$13 all restrict to the same $X_{(1)} \to \mathrm{diag}$14 on $X_{(1)} \to \mathrm{diag}$15. Theorem 2 gives

$X_{(1)} \to \mathrm{diag}$16

and the entropy increase is a relative entropy on the commutant side,

$X_{(1)} \to \mathrm{diag}$17

Because $X_{(1)} \to \mathrm{diag}$18, the modular flow and modularly defined entropies are $X_{(1)} \to \mathrm{diag}$19-invariant. In this precise sense, the gauge group acts orthogonally to entanglement: it changes only the hidden $X_{(1)} \to \mathrm{diag}$20-sector while preserving the restriction to $X_{(1)} \to \mathrm{diag}$21.

This formulation shifts the emphasis from geometric cuts to subalgebra inclusion. The relevant “orthogonality” is not spatial but modular: gauge transformations commute with the modular data that govern subsystem state reconstruction. A plausible implication is that entanglement-orthogonal gauge ideas extend naturally to settings where no preferred spatial factorization is available.

5. Classical limits, orthogonal projectors, and emergent gauge symmetry

A third usage arises in the emergence of classical gauge symmetry from quantum entanglement. Consider a bipartite system $X_{(1)} \to \mathrm{diag}$22 with a family of pure states whose classical limit is $X_{(1)} \to \mathrm{diag}$23. For any state corresponding to a definite classical configuration $X_{(1)} \to \mathrm{diag}$24, the reduced density matrix of subsystem $X_{(1)} \to \mathrm{diag}$25 must be approximately proportional to a projector,

$X_{(1)} \to \mathrm{diag}$26

where $X_{(1)} \to \mathrm{diag}$27 is the induced classical label on $X_{(1)} \to \mathrm{diag}$28 (Kirklin, 2022). The projectors for different classical labels must satisfy the approximate orthogonality condition

$X_{(1)} \to \mathrm{diag}$29

or equivalently

$X_{(1)} \to \mathrm{diag}$30

Together with the approximate resolution of identity

$X_{(1)} \to \mathrm{diag}$31

this defines the entanglement-orthogonal condition.

When $X_{(1)} \to \mathrm{diag}$32, the subsystem is classically resolvable but still carries shared non-local classical degrees of freedom. To obtain a local kinematical description one introduces an auxiliary edge-mode or purification space $X_{(1)} \to \mathrm{diag}$33 and isometric maps

$X_{(1)} \to \mathrm{diag}$34

The kinematical state space is then enlarged, and the physical space is recovered as a quotient by a gauge group. For a bipartite state $X_{(1)} \to \mathrm{diag}$35, that gauge group is

$X_{(1)} \to \mathrm{diag}$36

acting on the purifications by $X_{(1)} \to \mathrm{diag}$37 and $X_{(1)} \to \mathrm{diag}$38. The physical state space is $X_{(1)} \to \mathrm{diag}$39. In this setting, Wilson-line-type observables constructed from the purifications reproduce the non-local classical observables.

The three-spin large-$X_{(1)} \to \mathrm{diag}$40 toy model makes the mechanism explicit. Starting from the unique total-spin-zero intertwiner in $X_{(1)} \to \mathrm{diag}$41, one constructs coherent states by acting with independent $X_{(1)} \to \mathrm{diag}$42 rotations on spins 1 and 2. In the classical limit, these states become approximately orthogonal and define a global classical state space $X_{(1)} \to \mathrm{diag}$43. Each single-spin reduced state is maximally mixed, $X_{(1)} \to \mathrm{diag}$44, but each pair has a reduced state proportional to a projector $X_{(1)} \to \mathrm{diag}$45, and these projectors obey

$X_{(1)} \to \mathrm{diag}$46

The corresponding pair labels satisfy the closure constraint

$X_{(1)} \to \mathrm{diag}$47

in $X_{(1)} \to \mathrm{diag}$48. Introducing kinematical variables $X_{(1)} \to \mathrm{diag}$49 with gauge action $X_{(1)} \to \mathrm{diag}$50 then yields the physical quotient $X_{(1)} \to \mathrm{diag}$51.

This framework generalizes to compact Lie groups and has been proposed as evidence for emergent bulk diffeomorphism invariance: semiclassical reduced density matrices becoming approximately projectors onto minimal-area sectors, modular symmetries acting on edge modes, and tensor-network reduced states labeled by approximately orthogonal projectors all fit the same pattern. The key point is that orthogonality of reduced-state projectors can itself force the introduction of a gauge redundancy in any local classical description.

6. Gauge dependence, invariance of observables, and open formulations

Entanglement-orthogonal language also intersects a more familiar issue: gauge dependence of state descriptions versus gauge invariance of observables. In covariantly quantized electromagnetism, the Lorentz gauge contains scalar and longitudinal photons, and their exchange can entangle two distant atoms or harmonic oscillators (Franson, 2011). The resulting entangled state differs in form from the Coulomb-gauge state, where those modes are absent and their effect is encoded in the instantaneous Coulomb interaction. Nevertheless, an explicit generalized gauge transformation

$X_{(1)} \to \mathrm{diag}$52

maps the Lorentz-gauge state to the Coulomb-gauge state, and measurable quantities such as two-atom correlations or concurrence are identical once the transformation is taken into account. This establishes an important boundary condition on the concept: an entanglement-orthogonal gauge cannot mean that entanglement as a physical observable becomes gauge dependent, only that the decomposition of degrees of freedom used to represent it is gauge adapted.

A related but more tentative formulation appears in a sketch built around Coulomb-gauge entanglement methods. There, one asks which gauge choice or operator basis minimizes the mutual commutator between operators inside a region $X_{(1)} \to \mathrm{diag}$53 and outside $X_{(1)} \to \mathrm{diag}$54 (Yang et al., 2017). For a one-parameter family of gauge conditions $X_{(1)} \to \mathrm{diag}$55, one defines a boundary cross-commutator norm

$X_{(1)} \to \mathrm{diag}$56

and chooses $X_{(1)} \to \mathrm{diag}$57 minimizing $X_{(1)} \to \mathrm{diag}$58. In that sketch, generalized Coulomb gauges interpolate between more localized and more delocalized commutators, and the entanglement-orthogonal choice is the parameter for which the Dirac projector has minimal support on the entangling surface; spectral analysis of $X_{(1)} \to \mathrm{diag}$59 then places the entropy at a local minimum. Because this is presented as a possible formulation rather than a universally adopted construction, it is best understood as a programmatic extension of the operator-algebra viewpoint.

Several misconceptions follow from conflating these usages. First, entanglement-orthogonal gauge is not a single standard gauge comparable to Lorenz, Coulomb, or axial gauge. Second, it is not synonymous with “choose a center”: the center is one part of the construction, but the operational content comes from how Gauss-law constraints, local operators, and hidden sectors are organized. Third, it does not remove gauge-invariant entanglement; rather, it reorganizes the description so that entanglement-carrying variables, edge modes, or superselection labels are explicit.

Taken together, the existing formulations indicate a coherent theme. In matrix models, the gauge is adapted to the target-space normal direction and turns the problem into a complete-graph $X_{(1)} \to \mathrm{diag}$60 gauge theory. In lattice gauge theory, it diagonalizes boundary-flux sectors before fixing the pure gauge. In modular algebraic settings, it identifies gauge transformations that act only on the commutant. In semiclassical emergence, it appears as orthogonality of reduced-state projectors that forces a quotient by a gauge group. The shared content is a cut-adapted, algebra-adapted, or modular-adapted separation between entanglement data and gauge redundancy.