Gauge Symmetry, Relational Observables, and Locality

Diffeomorphism invariance changes what “an observable in a region” can mean. A coordinate-labelled field value is not gauge invariant, a finite boundary can carry charges that prevent a naive tensor-product split, and an excitation with nonzero energy necessarily has a long-range gravitational field. The chapter builds these statements from constraint generators through relational and dressed observables, then identifies what remains formulation- and boundary-dependent.

Helpful background. Flat-Spacetime Quantum Field Theory supplies operator algebras and microcausality; Einstein Equations supplies diffeomorphism-invariant dynamics; Quantum Information supplies subsystem language; Quantum Gravity supplies the volume’s scale and evidence conventions; and Variational and Symplectic Geometry supplies constraints and boundary symplectic terms.

Gauge invariance changes the question

For a metric and matter fields collectively denoted $\Phi$, an infinitesimal diffeomorphism generated by $\xi^\mu$ acts as $\delta_\xi\Phi=\mathcal L_\xi\Phi$. If $\xi^\mu$ vanishes suitably at the boundary, its canonical generator is a sum of constraints and the transformation relates redundant descriptions. If it approaches a nontrivial asymptotic symmetry, the generator acquires a surface term and can carry a physical charge. Regge and Teitelboim’s Hamiltonian analysis makes this boundary distinction explicit Regge and Teitelboim 1974, §§3–6.

That distinction propagates into quantum theory. A coordinate-local operator $\phi(x)$ generally fails to commute with the constraints because the coordinate label moves. One can instead specify $x$ through dynamical clocks and rods, solve the constraints, or attach a gravitational dressing to an asymptotic anchor. Each construction is invariant under its declared gauge transformations, but each also imports physical choices: reference fields fluctuate, dressings carry long-range fields, and asymptotic anchors depend on boundary conditions.

Six steps from constraints to asymptotic data

The sequence is cumulative.

1. Diffeomorphism Constraints and Gauge Orbits derives the bulk generator and the boundary charge, separating small gauge transformations from physical symmetries.
2. Dirac and Relational Observables constructs complete observables by evaluating one field where physical reference fields take specified values.
3. Gravitational Dressing shows perturbatively how a matter insertion is shifted by a metric-dependent functional so that its gauge variation cancels.
4. Subsystems and Algebraic Factorization explains why Gauss constraints obstruct a canonical regional tensor product and how edge data or algebraic choices restore an approximate split.
5. Locality, Microcausality, and Quantum Geometry compares ordinary QFT commutators with the dressing-dependent commutators of gravitational observables.
6. Soft Charges and Asymptotic Observables relates large asymptotic symmetries, Ward identities, soft gravitons, and memory while keeping their information content bounded.

The recurring thread is a three-part declaration: which transformations count as gauge, what boundary data are held fixed, and which operator is actually measured. Without all three, “local,” “inside,” and “commutes” are ambiguous.

Results and open questions as of 2026-08-20

The perturbative and asymptotic results are controlled in limited regimes; their nonperturbative generalization is unresolved.

Result or formulation	Established evidence	Scope and limitation
Canonical surface generators	Differentiability of the Hamiltonian requires boundary terms that become asymptotic charges	Depends on falloffs, topology, boundary conditions, and which transformations are quotiented as gauge
Perturbative gravitational dressing	Explicit weak-field operators can be invariant through a specified order and necessarily produce long-range fields Donnelly and Giddings 2016, §§II–V	Dressing is nonunique; calculations do not by themselves define exact nonperturbative local algebras
Soft-graviton Ward identities	The leading soft factor is universal for scattering with well-defined asymptotic states Weinberg 1965, pp. B516–B524	Infrared dressings and finite-time boundaries complicate the Hilbert-space interpretation; soft data need not encode all hard-state information
Regional factorization	Gauge constraints and boundary fluxes obstruct the most naive tensor-product split	Extended Hilbert spaces, centers, edge modes, and relational splits are useful but choice-dependent constructions

The conservative position treats these as precise restrictions on observables in perturbative or asymptotically controlled gravity, without inferring the disappearance of locality in every quantum-gravity theory. A permissive position sees their common structure as evidence that exact quantum gravity is fundamentally relational and only approximately local. These competing positions differ principally over extrapolation beyond the controlled weak-field, asymptotic, or semiclassical regime.

The conclusions could change if an exact construction establishes or refutes nonperturbative relational algebras, a regulator-independent continuum treatment changes subsystem factorization, a new boundary condition alters the charge–soft-theorem correspondence, or evidence after 2026-08-20 changes the comparison.

Boundaries of the chapter

The chapter does not claim that diffeomorphism invariance forbids all local experiments. Laboratory observables are relational—readings at a detector, proper times along worldlines, scattering data—and can be sharply localized relative to a semiclassical geometry. Nor does a nonzero commutator between two chosen dressings prove operational superluminal signalling. The commutator can describe shared long-range constraint fields rather than controllable causal influence.

String-theoretic and holographic realizations belong in their own volumes. Here the purpose is structural: establish the constraint, boundary, algebra, and perturbative facts that any microscopic construction must reproduce.

Review the chapter

Diagnostic. Consider two canonical pairs $(q_1,p_1)$ and $(q_2,p_2)$ with the constraint $C=p_1+p_2\approx0$. Show that $q_1-q_2$ is a Dirac observable, while $q_1$ is not. What changes if the common translation is declared to act nontrivially on a physical boundary reference?

Solution

The constraint generates $\delta_\epsilon q_i=\epsilon\{q_i,C\}=\epsilon$ and $\delta_\epsilon p_i=0$. Hence

$$\{q_1-q_2,C\}=1-1=0, \qquad \{q_1,C\}=1.$$

Only the separation is invariant under the common-translation gauge orbit. If a physical boundary reference is held fixed, a translation that moves both $q_i$ relative to that reference is no longer a redundancy; its generator must include a boundary degree of freedom or charge. Then an anchored position can be observable. The algebra alone does not decide the issue—the boundary conditions decide which transformations are gauge.

References

• Donnelly, William, and Steven B. Giddings. “Observables, Gravitational Dressing, and Obstructions to Locality and Subsystems.” Physical Review D 94 (2016): 104038. DOI. Open PDF.
• Regge, Tullio, and Claudio Teitelboim. “Role of Surface Integrals in the Hamiltonian Formulation of General Relativity.” Annals of Physics 88 (1974): 286–318. DOI.
• Weinberg, Steven. “Infrared Photons and Gravitons.” Physical Review 140 (1965): B516–B524. DOI.

Dated Research assessment. For open questions and recent results, follow the link below. Each assessment states the date through which it considers results and the limits of its conclusions.

• Operational Observables in Quantum Gravity