Gauge Degeneracies and Boundary Symmetries

A diffeomorphism is pure gauge only if its tangent vector is a degeneracy of the presymplectic form on the chosen phase space. For a hypersurface with a boundary, the contraction of a diffeomorphism direction with that form reduces to a surface variation. A generator that vanishes sufficiently strongly at the boundary is usually degenerate; one that preserves the boundary conditions but acts nontrivially on boundary data can carry a Hamiltonian charge and represents a boundary symmetry.

Required background. The Gibbons–Hawking–York Boundary Term supplies the boundary variational principle; Presymplectic Potentials and Currents supplies degeneracies and reduction; and Killing Fields and Isometry Groups supplies diffeomorphism and symmetry generators.

Helpful background. The Einstein Equation supplies the constraints; Hypersurface Normals and Tangential Projectors supplies boundary projections; and Formulations of General Relativity orients the chapter.

Degeneracy test on solution space

Let $\Gamma$ be a space of solutions satisfying declared boundary conditions, and let $\delta g$ be tangent to $\Gamma$. Fix a field-independent vector field $\xi^a$, so $\delta\xi^a=0$. It induces $\delta_\xi g=\mathcal L_\xi g$. It is a presymplectic degeneracy when

$$\Omega_\Sigma(\delta g,\mathcal L_\xi g)=0$$

for every allowed $\delta g$. With the site’s Iyer–Wald ordering, the on-shell identity gives

$$\Omega_\Sigma(\delta g,\mathcal L_\xi g) =-\int_{\partial\Sigma} \left(\delta Q_\xi -\xi\mathbin{\lrcorner}\Theta\right).$$

The overall sign depends on the ordering convention; the condition that the surface integral vanish does not. The surface expression and its Hamiltonian interpretation follow from Iyer and Wald 1994, §6, eqs. (73)–(82), pp. 14–15, PDF.

If $\xi^a$ and the derivatives entering $Q_\xi$ vanish at $\partial\Sigma$, the integral normally vanishes, and quotienting by this direction removes redundancy. If the integral is a nonzero exact field-space variation, the reverse pairing $\Omega_\Sigma(\mathcal L_\xi g,\delta g)$ defines $\delta H_\xi$ and the direction is not a gauge degeneracy. A Hamiltonian symmetry can act nontrivially while preserving its own generator $H_\xi$: antisymmetry gives $\delta_\xi H_\xi=\Omega_\Sigma(\delta_\xi g,\delta_\xi g)=0$. Nondegeneracy is not a claim that a symmetry changes its own charge.

Proper gauge versus boundary-active transformations

Three tests must be kept distinct:

1. Tangency: $\mathcal L_\xi g$ must preserve the boundary and falloff conditions, possibly together with an allowed transformation of prescribed sources.
2. Degeneracy: its presymplectic pairing with every allowed perturbation must vanish.
3. Integrability: if the pairing does not vanish, the resulting one-form must integrate to a Hamiltonian charge for $\xi^a$ to generate a canonical symmetry.

A vector field can preserve the boundary conditions yet fail the degeneracy test. Conversely, a formal charge expression can vanish on a particular solution while its variation is nonzero nearby, so zero background charge does not prove that the direction is gauge throughout phase space.

The quotient between boundary-condition-preserving vectors and proper-gauge vectors defines the boundary symmetry algebra. Its concrete content depends on the boundary: spatial infinity, null infinity, an isolated horizon, and a finite timelike wall impose different falloff and flux conditions. This page develops the test, not a classification of those algebras.

Boundaries may require additional variables

Holding a geometric boundary fixed can make a bulk diffeomorphism appear boundary-active because it moves the embedding or frame. One consistent approach restricts the generator at the boundary. Another enlarges phase space with embedding, corner, or edge variables whose transformation restores gauge invariance. These are different phase spaces, so their degeneracies and charges need not coincide.

Likewise, adding an allowed boundary counterterm can change the symplectic potential by a corner contribution. It may improve differentiability or shift the representative of a charge, but it cannot be assessed independently of the boundary conditions. Harlow and Wu develop a boundary-consistent covariant phase space in Harlow and Wu 2020, §§2.1–2.4, pp. 6–23, PDF.

A compact Maxwell analogy

For source-free Maxwell theory on a fixed background, take $L=-\tfrac12F\wedge *F$ and $\Theta=-\delta A\wedge *F$, with $F=dA$ and coupling set to one. A field-independent gauge parameter gives $\delta_\lambda A=d\lambda$. The declared ordering gives $\omega(\delta A,d\lambda)=d\lambda\wedge\delta(*F)$. On the linearized Gauss-law surface, $d\delta(*F)=0$, so

$$\Omega_\Sigma(\delta A,d\lambda) =\int_{\partial\Sigma}\lambda\,\delta({*F}).$$

If $\lambda|_{\partial\Sigma}=0$, the direction is degenerate. If $\lambda$ takes a nonzero constant on one boundary component and vanishes on the others, the surface term measures the electric flux variation on that component. Gauss’ law still fixes the oriented sum of all boundary fluxes to zero. In particular, a globally constant $\lambda$ has $d\lambda=0$ and cannot acquire a nonzero pairing by ignoring another boundary component. Gravity follows the same logic, although its charge form depends on both $\xi^a$ and its derivative and its constraints generate spacetime deformations.

Common pitfalls

Diffeomorphism covariance does not make every diffeomorphism disposable. Boundary-active transformations can change measurable charges.

Preserving a boundary location is not the degeneracy test. The test is the pairing with every allowed perturbation, including corner contributions.

A symmetry algebra needs a phase space first. Falloff, flux, counterterms, and field dependence determine both the allowed generators and their bracket.

Exercise: Maxwell gauge parameter on an annulus

Let $\Sigma$ be a source-free annulus with inner and outer boundary spheres. Allow the outer electric flux $Q$ to vary, with inner oriented flux $-Q$. Let $\lambda$ equal $\lambda_0$ on the outer sphere and zero on the inner sphere, interpolating smoothly in between. Decide whether $d\lambda$ is a proper gauge direction. Compare with a globally constant parameter.

Solution

The surface pairing becomes

$$\Omega_\Sigma(\delta A,d\lambda) =\lambda_0\,\delta Q.$$

For $\lambda_0\ne0$, it is nonzero for allowed perturbations that change $Q$, so the direction is not a degeneracy. In our Hamiltonian ordering, its generator has variation $\delta H_\lambda=\Omega_\Sigma(d\lambda,\delta A)=-\lambda_0\delta Q$. If $\lambda$ is instead constant throughout the annulus, then $d\lambda=0$ and the two surface contributions cancel: $\lambda_0\delta(Q-Q)=0$. This is why a charge on one component must not be confused with the total Gauss-law flux.

References

• Harlow, Daniel, and Jie-qiang Wu. “Covariant Phase Space with Boundaries.” Journal of High Energy Physics 2020, 146 (2020). DOI; Open PDF.
• Iyer, Vivek, and Robert M. Wald. “Some Properties of Noether Charge and a Proposal for Dynamical Black Hole Entropy.” Physical Review D 50, no. 2 (1994): 846–864. DOI; Open PDF.