Locality, Microcausality, and Quantum Geometry

Microcausality in ordinary quantum field theory says that gauge-invariant observables with spacelike-separated supports commute. Gravity complicates every noun in that sentence: the causal relation is dynamical, coordinate support is gauge dependent, and an operator carrying energy needs a long-range dressing. In weak-field gravity, nominally spacelike dressed observables generally acquire commutators beginning at order $G$. Their coefficients depend on the dressing and state, so the result is an obstruction to exact background-local algebras—not by itself an operational signal outside a light cone.

Required background. Variational and Symplectic Geometry supplies constraints and brackets. Equivalence Principles and Gravitational Measurement supplies operational causal comparisons.

Helpful background. Flat-Spacetime Quantum Field Theory supplies microcausality, Einstein Equations supplies the dynamical metric, Quantum Information supplies signalling tests, and Quantum Gravity supplies the regime and evidence conventions.

Fixed-background benchmark

For a free linear scalar field on a fixed globally hyperbolic spacetime,

$$[\phi(x),\phi(y)]=i\Delta(x,y),$$

where the Pauli–Jordan function $\Delta$ is the causal propagator with sign fixed by this commutator and vanishes at spacelike separation. The causal relation and field supports use the same metric. Interacting local fields still commute at spacelike separation under the usual locality assumptions, but their general unequal-point commutator need not be a state-independent c-number function $i\Delta$.

In gravity, $x$ and $y$ are coordinate labels and do not specify gauge-invariant events. Write $g_{\mu\nu}=\eta_{\mu\nu}+\kappa h_{\mu\nu}$ and dress a scalar as

$$\Phi_V(x)=\phi(x)+V^\mu(x)\partial_\mu\phi(x)+O(\kappa^2), \qquad V^\mu=O(\kappa).$$

At background-spacelike separation, the bare matter commutator vanishes. The first generic dressing contribution contains

$$[V^\mu(x),V^\nu(y)]\, \partial_\mu\phi(x)\partial_\nu\phi(y),$$

which is $O(\kappa^2)=O(G)$. Donnelly and Giddings compute such nonlocal algebras for explicit gravitational line and Coulomb dressings Donnelly and Giddings 2016, “Diffeomorphism-Invariant Observables,” §§III–V.

Results and limitations

This discussion reflects results available through 2026-08-20. The order-$G$ statement is controlled perturbatively; its nonperturbative interpretation remains unresolved.

Formulation or observable	Result and evidence	Assumptions and regime	Scope or limitation
Bare local QFT field	Spacelike commutator vanishes by the Pauli–Jordan support property	Fixed globally hyperbolic background and local field algebra	Not diffeomorphism invariant when gravity is dynamical
Asymptotically anchored Coulomb dressing	Generic equal-time commutators begin at $O(G)$ and have a long-range kernel	Weak field about Minkowski space, asymptotic flatness, specified gauge and operator ordering	Coefficient and tensor structure depend on dressing; loop and nonlinear terms are omitted
Gravitational line dressing	Gauge-invariant through retained order; commutator support follows the dressing line and can differ from the Coulomb choice	Perturbative line to an asymptotic anchor	Singular idealized lines require smearing; changing the path changes the algebra by radiative data
Relational matter coordinates	Can define observables at physical clock-and-rod readings and recover approximate microcausality in sharply peaked semiclassical states	Invertible reference map, controlled clock fluctuations and backreaction	No universal state-independent causal support; global caustics or clock failure obstruct the construction
Charge-fixed gravitational split	Suitable dressings make exterior metric matrix elements depend only on total Poincaré charges at first order Donnelly and Giddings 2018, §IV, PDF	Perturbative localized state space, buffer region, and chosen common exterior dressing	A first-order matrix-element result is not an exact factorization theorem for the full exterior algebra

The conservative position says that locality remains an excellent emergent principle for relational observables in weak fields and at resolutions too coarse to detect $O(G)$ dressing effects. The permissive position says the absence of an exact compactly supported algebra is evidence for fundamentally nonlocal quantum geometry. These competing interpretations depend on the dressing, boundary condition, state, observable algebra, and parameter regime; the perturbative calculation alone does not choose between microscopic pictures.

What the first gravitational order means

The power counting is robust. Gauge invariance requires $V=O(\kappa)$, and a quantum commutator between two linear metric functionals is $O(\kappa^2)$. A special pair of dressings or an equal-time component can cancel, but that does not create a universal compact local algebra: other commutators, momenta, or dressings can remain nonzero. Changing $V$ by a gauge-invariant radiative functional changes the coefficient without changing the generic order.

The dimensionless size must be built from the state and separation. For localized energies $E_x,E_y$ separated by $r$, a typical weak-field scale is

$$\epsilon_{\mathrm{nl}}\sim\frac{G E}{r}$$

with the relevant energy combination fixed by the matrix element. It is tiny when the gravitational potential is weak. The expansion fails when the corresponding invariant becomes order unity, when accumulated long-time effects defeat fixed-order perturbation theory, or when no single semiclassical causal geometry describes the state.

Nonzero commutator is also not synonymous with controllable signalling. A signalling test compares outcome probabilities under an intervention in one laboratory. It requires normalized dressed states, a preparation operator, a detector observable, retarded evolution, and constraints satisfied throughout. The equal-time algebra can encode a common constraint field already present in both observables. A claim of superluminal communication would require an additional protocol-level calculation.

Causality with a quantum metric

When the metric fluctuates, “spacelike” may be state dependent. In a semiclassical state sharply peaked on $g^{(0)}_{\mu\nu}$, one can smear relational observables over regions separated by more than the metric uncertainty and demand that commutators be suppressed by both $G$ and the state’s geometric variance. That is an approximate, resolution-dependent microcausality condition.

In a superposition of causal structures, a single background separation may not exist. One may instead define causal relations using clock fields, boundary time, or conditional metric observables. Each choice specifies a different operator question. It is therefore premature to infer a universal noncommutative causal order from one truncation or symmetry-reduced model.

Adverse checks include changing Coulomb to line dressing, varying gauge and boundary falloffs, smearing the operator, switching the reference state, and computing a genuine intervention channel. A physical conclusion should survive these changes or state explicitly which one defines it.

An explicit order-counting check

Let two background-spacelike scalar insertions have dressings satisfying

$$[V^\mu(x),W^\nu(y)]=i\kappa^2 C^{\mu\nu}(x,y)+O(\kappa^3).$$

Metric and matter commute at zeroth order, so expansion of the two dressed operators gives

$$[\Phi_V(x),\Phi_W(y)] =i\kappa^2 C^{\mu\nu}(x,y) \partial_\mu\phi(x)\partial_\nu\phi(y)+O(\kappa^3),$$

up to same-order cross terms fixed by the complete dressing. Replacing $V$ with $V+R$, where $R=O(\kappa)$ is gauge invariant, shifts $C^{\mu\nu}$ and can cancel a chosen component, but generically leaves the $\kappa^2$ order. The existence of a nonlocal correction is robust; its detailed kernel is formulation dependent.

Exercises

Exercise. Suppose the dimensionless strength of a displayed commutator matrix element scales as $\epsilon_{\mathrm{nl}}=GE/r$. If $E$ is halved and $r$ is doubled, by what factor does the leading nonlocal strength change? Which conclusion is invalid if the result is then exactly zero for one special dressing?

Solution

The leading scale transforms as

$$\frac{G(E/2)}{2r}=\frac14\frac{GE}{r},$$

so it falls by a factor of four. If one special dressing gives an exact cancellation, it is invalid to conclude that all gauge-invariant gravitational observables obey exact background microcausality. One must check other components, conjugate momenta, and dressings. The cancellation can be a property of the chosen boundary attachment rather than a formulation-independent local algebra.

What could change these conclusions

The conclusions could change when a nonperturbative construction derives an exact causal algebra, an independent calculation changes the order-$G$ dressing commutator, an operational signalling protocol separates constraint correlations from causal influence, or evidence after 2026-08-20 changes the competing interpretations.

References

• Donnelly, William, and Steven B. Giddings. “Diffeomorphism-Invariant Observables and Their Nonlocal Algebra.” Physical Review D 93 (2016): 024030. DOI. Open PDF.
• Donnelly, William, and Steven B. Giddings. “Gravitational Splitting at First Order: Quantum Information Localization in Gravity.” Physical Review D 98 (2018): 086006. DOI. Open PDF.