Diffeomorphism Constraints and Gauge Orbits

In canonical gravity, a spacetime diffeomorphism is represented by constraints only after its behavior at the boundary has been fixed. Deformations that die off at the boundary generate gauge orbits; deformations that act on admissible boundary data can instead carry energy, momentum, or angular momentum. The same local formula can therefore represent either redundancy or physical symmetry, and the differentiability of its generator tells the two cases apart.

Required background. Variational and Symplectic Geometry supplies Hamiltonian generators and surface terms. Equivalence Principles and Gravitational Measurement supplies the operational meaning of coordinates and reference frames.

Helpful background. Flat-Spacetime Quantum Field Theory supplies constrained quantization, Einstein Equations supplies the covariant dynamics, Quantum Information supplies subsystem language, and Quantum Gravity sets this volume’s regimes and evidence standards.

Canonical data and constraints

Take a globally hyperbolic four-dimensional region foliated by spacelike hypersurfaces $\Sigma_t$, with induced metric $h_{ij}$, canonical momentum $\pi^{ij}$, lapse $N$, and shift $N^i$. Matter terms may be added without changing the logic. Up to the boundary action, the Hamiltonian form is

$$S=\int dt\left[\int_{\Sigma_t}d^3x \left(\pi^{ij}\dot h_{ij}-N\mathcal H_\perp-N^i\mathcal H_i\right)-H_{\partial\Sigma}\right].$$

The lapse and shift are Lagrange multipliers, giving the Hamiltonian and momentum constraints $\mathcal H_\perp\approx0$ and $\mathcal H_i\approx0$. Choose $K_{ij}=(\dot h_{ij}-D_iN_j-D_jN_i)/(2N)$ and the momentum density $\pi^{ij}=\sqrt h\,(K^{ij}-h^{ij}K)/(16\pi G)$, where $h=\det h_{ij}$. Then

$$\mathcal H_i=-2D_j\pi^j{}_i+\mathcal H_i^{\mathrm{matter}},$$

while $\mathcal H_\perp$ combines the spatial scalar curvature and a quadratic expression in $\pi^{ij}$. The complete canonical decomposition and its relation to the Einstein equation are derived in Arnowitt, Deser, and Misner 1962/2008, §§3–4.

On a closed hypersurface, the smeared momentum constraint

$$D[\xi]=\int_\Sigma d^3x\,\xi^i\mathcal H_i$$

generates spatial diffeomorphisms:

$$\{h_{ij},D[\xi]\}=\mathcal L_\xi h_{ij}, \qquad \{\pi^{ij},D[\xi]\}=\mathcal L_\xi\pi^{ij},$$

where $\pi^{ij}$ is a tensor density. The constraints are first class; their brackets close with structure functions. This closure expresses hypersurface-deformation consistency, not an ordinary Lie algebra with constant structure coefficients.

The surface term decides the transformation

For clarity, omit matter and write the ordinary tensor $p^{ij}=\pi^{ij}/\sqrt h$. Integration by parts gives

$$\int_\Sigma d^3x\,\pi^{ij}\mathcal L_\xi h_{ij} =\int_\Sigma d^3x\,\xi^i\mathcal H_i +2\int_{\partial\Sigma}d^2x\,\sqrt\sigma\,n_ip^{ij}\xi_j.$$

Here $n_i$ points outward within $\Sigma$ and $\sigma$ is the determinant of the induced two-metric. The surface measure already carries its density, so the integrand uses $p^{ij}$, not the weight-one bulk density $\pi^{ij}$. This is the covariant divergence theorem applied to $\sqrt h\,p^{ij}$. If $\xi^i$ vanishes at the boundary, the surface term vanishes and $D[\xi]$ is a sum of constraints. On the constraint surface its Hamiltonian flow moves along a gauge orbit.

If $\xi^i$ approaches an allowed nonzero boundary transformation, the bare bulk constraint is generally not functionally differentiable. One must add a surface functional $Q[\xi]$ so that

$$G[\xi]=\int_\Sigma d^3x\,\xi^i\mathcal H_i+Q[\xi]$$

has a well-defined variation. On shell, $G[\xi]=Q[\xi]$ need not vanish. In asymptotically flat gravity, the corresponding lapse and shift limits yield ADM energy, momentum, and angular momentum. Regge and Teitelboim give the systematic differentiability analysis and asymptotic generators Regge and Teitelboim 1974, §§3–6.

Thus “diffeomorphisms are gauge” is too coarse. Proper diffeomorphisms preserve the boundary data and have vanishing generators; allowed boundary symmetries can act nontrivially on the physical boundary sector. A charge vanishing on one particular state does not make its generator gauge on all states. Nor is “large” synonymous with “charged”: large diffeomorphisms can be disconnected from the identity, a separate global issue. The classification depends on falloff conditions, topology, corners, and the chosen ensemble.

Reduced phase space and observables

Let $\Gamma_C$ be the constraint surface in phase space. The reduced phase space is schematically

$$\Gamma_{\mathrm{red}}=\Gamma_C/\mathcal G_0,$$

where $\mathcal G_0$ is the group generated by proper constraints, not necessarily every transformation preserving the form of the boundary conditions. A Dirac observable $O$ satisfies $\{O,C_A\}\approx0$ for all proper first-class constraints $C_A$. A boundary charge may act nontrivially on $O$ and organize it into a physical symmetry representation.

There are two useful checks. On compact $\Sigma$ without boundary, $Q[\xi]$ is absent and the spatial generator reduces to the constraint. In the weak-field asymptotically flat limit, a constant asymptotic time translation instead gives the nonzero ADM Hamiltonian. Treating both transformations as gauge would incorrectly set the total energy to zero.

The quotient can also be singular where gauge orbits change dimension, and global gauge fixing can fail. The local first-class analysis therefore does not guarantee a globally smooth reduced phase space. Quantum constructions may use Dirac quantization instead, imposing constraints on states while representing boundary charges as operators.

Independent sign and boundary check

Starting from $\mathcal H_i=-2D_j\pi^j{}_i$, integration by parts gives

$$\int_\Sigma \xi^i\mathcal H_i =2\int_\Sigma \pi^{ij}D_i\xi_j -2\int_{\partial\Sigma}\sqrt\sigma\,n_ip^{ij}\xi_j.$$

Because $\pi^{ij}$ is symmetric, the volume term is $\int_\Sigma\pi^{ij}\mathcal L_\xi h_{ij}$. Moving the surface term to the other side reproduces the earlier identity and checks the sign of $Q[\xi]$. At a fixed finite boundary, take $\xi$ field independent, tangent to the boundary, and preserving the allowed boundary data; these conditions eliminate the residual normal boundary variation and make the spatial canonical generator differentiable. A normal deformation requires a separately specified boundary variational problem; integration by parts alone does not establish integrability of its charge. Reversing the boundary orientation reverses the charge integral. Field-dependent $\xi$ also requires additional terms in the charge algebra.

Common pitfalls

Constraint does not mean zero Hamiltonian in every setting. The bulk Hamiltonian is constrained, but a boundary Hamiltonian can generate physical time translations and be nonzero.

Coordinate covariance is not the full gauge test. An expression may be a scalar under passive coordinate changes yet still depend on where a dynamical point is placed along a gauge orbit. The relevant test is its Poisson bracket with proper constraints.

Exercises

Exercise. Let $\Sigma$ have a fixed boundary, ignore matter, and take $\xi$ field independent and tangent to the boundary with boundary-preserving data. Starting from $D_0[\xi]=\int_\Sigma\xi^i\mathcal H_i$, derive the boundary term needed so that the generator equals $\int_\Sigma\pi^{ij}\mathcal L_\xi h_{ij}$. State when the resulting generator is pure gauge.

Solution

Using $\mathcal H_i=-2D_j\pi^j{}_i$ and integrating by parts gives

$$D_0[\xi]=\int_\Sigma d^3x\,\pi^{ij}\mathcal L_\xi h_{ij} -2\int_{\partial\Sigma}d^2x\,\sqrt\sigma\,n_ip^{ij}\xi_j.$$

Therefore the differentiable generator with the desired canonical action is

$$D[\xi]=D_0[\xi]+2\int_{\partial\Sigma}d^2x\,\sqrt\sigma\,n_ip^{ij}\xi_j.$$

If $\xi^i$ obeys the proper-gauge boundary conditions so the surface functional vanishes throughout the allowed phase space, $D[\xi]\approx0$ and generates a gauge orbit. If the allowed boundary symmetry instead has a nontrivial charge functional, it must remain in the physical symmetry algebra. Its numerical value may vanish on a special state without changing that classification.

References

• Arnowitt, Richard, Stanley Deser, and Charles W. Misner. “The Dynamics of General Relativity.” In Gravitation: An Introduction to Current Research, edited by Louis Witten, 227–265. New York: Wiley, 1962. Republication, arXiv:gr-qc/0405109.
• Regge, Tullio, and Claudio Teitelboim. “Role of Surface Integrals in the Hamiltonian Formulation of General Relativity.” Annals of Physics 88 (1974): 286–318. DOI.