Edge Modes and Extended Observables at Gauge Boundaries

Edge variables are needed when the chosen regional phase space and gauge group otherwise fail to fit together. They can restore gauge covariance of the symplectic potential and dress open observables, but they also carry a genuine boundary symmetry distinct from the gauge redundancy. Adding them for transformations already fixed to vanish, or quotienting their surface symmetry away, changes the observable count.

Required background. Boundary phase spaces, constraints, and the BFV charge distinguishes gauge directions from charged boundary transformations. Gluing, reduction, and composition theorems supplies fusion across a cut. Proper and improper gauge transformations fixes the physical terminology.

Helpful background. Gauge constraints, centers, and edge data explains the algebraic center choice. Edge modes, subregions, and factorization gives the physical construction. Gauge-field entanglement and edge contributions treats the entropy question.

Extended Maxwell phase space

Let $\Sigma$ be a spatial region with boundary $S$. In Abelian gauge theory, the bulk symplectic potential may be written

$$\Theta_\Sigma=\int_\Sigma E\wedge\delta A,$$

where $E$ is the form dual to the electric field. Under $A\mapsto A+d\lambda$, integration by parts leaves a surface term unless $\lambda|_S=0$ or the normal flux is fixed. Introduce a boundary scalar $\varphi$ with

$$\varphi\longmapsto\varphi+\lambda|_S$$

and extend

$$\Theta_{\mathrm{ext}} =\int_\Sigma E\wedge\delta A -\int_S E_\perp\,\delta\varphi.$$

The two gauge variations cancel with the indicated orientation convention. The Gauss generator then vanishes on shell for all allowed $\lambda$, including those supported at $S$. A separate surface symmetry shifts $\varphi$ while leaving $A$ fixed; its Hamiltonian is the normal electric flux. This separation between right gauge action and left boundary symmetry is explicit in the non-Abelian construction of Donnelly and Freidel 2016, §§2.1–2.4, pp. 11–15.

The construction is conditional. If the phase space already restricts $\lambda|_S=0$, the surface variation vanishes and no compensating scalar is forced. If instead boundary gauge transformations carry physical charge, declaring them gauge requires the extended field and its constraint; otherwise one has erased a Hamiltonian action rather than resolved it.

Dressed open observables

For an oriented curve $\gamma$ from $x_1\in S$ to $x_2\in S$, define

$$W_\gamma^{\mathrm{dress}} =e^{-i\varphi(x_2)} \exp\left(i\int_\gamma A\right) e^{i\varphi(x_1)}.$$

Under $A\mapsto A+d\lambda$, the Wilson factor gains $e^{i\lambda(x_2)-i\lambda(x_1)}$, while the endpoint factors contribute the inverse. The dressed line is gauge invariant. In a non-Abelian theory, a group-valued boundary frame replaces $e^{i\varphi}$ and the path-ordered holonomy is conjugated by its endpoint values Donnelly and Freidel 2016, §2.3, pp. 13–14.

This dressing does not make the endpoint unobservable. The independent boundary symmetry acts on the frame and on the dressed electric flux, producing a surface current algebra. Whether the regional observable algebra includes the flux center, a conjugate edge algebra, or only neutral combinations is a physical choice developed at Gauge Constraints, Centers, and Edge Data.

Gluing complementary regions

Let $\Sigma_L$ and $\Sigma_R$ meet at $S$, with outward fluxes $E_{\perp L}$ and $E_{\perp R}$. Their extended phase spaces glue by the constraints

$$E_{\perp L}+E_{\perp R}=0, \qquad \varphi_L-\varphi_R=0$$

up to the diagonal surface action. Symplectic reduction pairs the two boundary frames and removes the duplicated edge degree of freedom. An open dressed Wilson line ending on $S$ from the left composes with one from the right to give an ordinary bulk Wilson line in $\Sigma_L\cup\Sigma_R$. Donnelly and Freidel formulate this fusion as a symplectic quotient in 2016, §2.5, pp. 15–20.

The BV–BFV version adds ghosts and implements the same matching through the boundary charge and pushforward. It does not by itself choose a quantum representation or an entanglement prescription.

Failure tests: two inconsistent counts

First impose $\lambda|_S=0$ and still add an unconstrained $\varphi$. The scalar then has no gauge transformation to compensate and becomes a spurious physical boundary mode. Second, allow arbitrary $\lambda|_S$ but quotient the original bulk phase space by those transformations without an edge field or fixed flux. The charged flux directions disappear. The first construction overcounts and the second undercounts.

The strongest safe statement is conditional: edge modes are required relative to a declared regional gauge group, boundary condition, symplectic potential, and observable algebra. They are neither universally physical nor universally removable.

The same warning applies after quantization. A classical surface current algebra does not choose its representation, and a direct-sum decomposition into flux sectors does not by itself choose an algebraic center or entropy. Those steps require a state, a completion, and a declared regional observable algebra. BV–BFV consistency supplies the constrained complex and gluing map; it does not settle every information-theoretic interpretation of the edge sector.

Exercises

Verify the gauge invariance of $W_\gamma^{\mathrm{dress}}$.

Solution

The line integral changes by $\lambda(x_2)-\lambda(x_1)$. The endpoint factors change by $e^{-i\lambda(x_2)}$ and $e^{i\lambda(x_1)}$, cancelling that phase exactly.

Why does gluing impose a sum rather than equality of outward fluxes?

Solution

The two regions induce opposite orientations on their common boundary. A single smooth electric field therefore has outward normal component $E_{\perp L}=-E_{\perp R}$. The moment-map constraint is their sum.

References

• Cattaneo, Alberto S., Pavel Mnev, and Nicolai Reshetikhin. “Classical BV Theories on Manifolds with Boundary.” Communications in Mathematical Physics 332 (2014): 535–603. DOI; Open PDF.
• Donnelly, William, and Laurent Freidel. “Local Subsystems in Gauge Theory and Gravity.” Journal of High Energy Physics 2016, no. 9 (2016): 102. DOI; Open PDF.