Gravitational Observers in Modern Theories

• Gravitational observers are physical reference systems—such as clocks, tetrads, or detector arrays—that enable gauge-invariant localization in gravitational theories.
• They implement relational observables by dressing local operators, overcoming diffeomorphism invariance and Hamiltonian constraints in both classical and quantum contexts.
• Observer-induced algebras yield Type II von Neumann structures that clarify gravitational entropy, holography, and operational measurements across diverse spacetimes.

The term gravitational observer does not denote a single universally fixed construction. In the literature it can mean a timelike worldline carrying a standard clock, a tetrad-adapted local frame, a uniformly accelerated family tied to a causal horizon, an extended array of freely falling probes that reads tidal curvature, or a dynamical quantum reference system used to define gauge-invariant observables and entropy in gravity (Philipp et al., 2023, Spaniol et al., 2011, Kolekar et al., 2017, Ferreira et al., 2019, Vuyst et al., 2024). What unifies these uses is that gravity obstructs an observer-independent notion of localization: diffeomorphism invariance, Hamiltonian constraints, horizon dependence, and modular flow all force observables to be specified relative to some physical frame, congruence, or subsystem.

1. Relational localization and the need for observers

In gravitational theories, ordinary local subregion factorization fails because diffeomorphisms are gauge redundancies and local field insertions are not gauge invariant. In asymptotically AdS, the linearized Hamiltonian constraint implies

$$[\hat H_{\partial},\hat O^{\rm Phys}(x)] = -i\sqrt{16\pi G_N}\,\partial_t \hat O^{\rm Phys}(x),$$

so a genuinely time-localized bulk operator cannot commute with the asymptotic boundary Hamiltonian unless it is appropriately dressed. For open subregions reaching the asymptotic boundary, one may use gravitational Wilson lines ending where gravity decouples; for a closed subregion such as an entanglement island, this is obstructed. The observer was introduced precisely to supply a relational endpoint for such dressing, and in the island model that observer is identified with the Goldstone/Stückelberg field of bath-induced broken diffeomorphism symmetry, with dressed operators

$$\hat O^{\rm Phys}(x)=\hat O\!\left(x+\sqrt{16\pi G_N}\,V(x)\right).$$

In this construction the observer is emergent rather than fundamental, and the resulting island algebra is identified holographically with an emergent Type $\mathrm{II}_\infty$ von Neumann algebra, assuming the geometric modular flow conjecture (Geng et al., 13 Jun 2025).

A related relational construction appears in the JT black-hole throat. There, local fiducial observers are defined for each off-shell boundary reparametrization $F(t)$ by the null anchoring prescription

$U=F(t+z),\qquad V=F(t-z),$

which is uniquely fixed semiclassically by requiring that asymptotic boundary time translations extend into the bulk as the flow of a conformal Killing vector. These observers are accelerated, remain outside the horizon, and provide a notion of local bulk frame in a regime where the gravitational field itself is highly quantum (Mertens et al., 28 Jul 2025).

Near-boundary AdS observers furnish a complementary operational example. Observers confined to a short boundary time band can act only with simple low-energy unitaries and measure the boundary Hamiltonian through the asymptotic metric,

$$H = \frac{d}{16\pi G}\lim_{r\to\infty} r^{d-2}\int h_{00}(r,t,\Omega)\,d^{d-1}\Omega.$$

For low-energy bulk states, this lets them extract overlaps with states of the form $X|0\rangle$ from vacuum-projection probabilities of the form

$$\langle g_{\rm mod}|P_0|g_{\rm mod}\rangle = J^2 |\langle X|g\rangle|^2 + O(J^3),$$

and thereby reconstruct the bulk state. The same protocol fails in theories without gravity, including nongravitational gauge theories, because the crucial one-dimensional vacuum projector is not boundary-accessible there (Chowdhury et al., 2020).

2. Observer-induced algebras, Hilbert spaces, and entropy

A major recent use of gravitational observers is algebraic: they convert otherwise problematic local algebras into algebras that admit traces and entropies. In the island construction, the relevant extended algebra is the crossed product of the Type III$_1$ island QFT algebra by its modular automorphism group,

$$\mathcal A = \mathcal A_{I,\rm QFT}\rtimes \mathcal A_{\Pi[\xi_\Psi]+H_{\rm QFT}[\xi_\Psi]},$$

and by Takesaki’s theorem this crossed product is Type $$\hat O^{\rm Phys}(x)=\hat O\!\left(x+\sqrt{16\pi G_N}\,V(x)\right).$$0. The observer degree of freedom appears as the canonical mode $$\hat O^{\rm Phys}(x)=\hat O\!\left(x+\sqrt{16\pi G_N}\,V(x)\right).$$1, which turns modular flow into an internal algebraic direction and yields a semifinite trace, but not a normalized one (Geng et al., 13 Jun 2025).

The same issue appears in closed-universe quantum gravity. One recent analysis distinguishes two separate problems: the problem of time, attributed to summing over metrics, and the problem of dimension, attributed to summing over topologies. In that framework, an observer is defined as “a subsystem of the universe whose specification results in a nontrivial Hilbert space,” while a timekeeper is “an observer with a specified history that can be used as a reference for the time of the environment and experiences a nontrivial time evolution.” The corresponding gravitational path integral is then modified by restrictions such as existence, connectibility, replica distinguishability, and no replica collision, and fixing the observer history produces a timekeeper-dependent generalization of holography (Wei, 26 Jun 2025).

In holographic-map models of closed universes, observer inclusion can also be formulated by excising the part of the code that acts on the observer. In the construction implementing the Abdalla–Antonini–Iliesiu–Levine prescription, the observer/environment interface becomes an effective holographic boundary, and the toy model yields

$$\hat O^{\rm Phys}(x)=\hat O\!\left(x+\sqrt{16\pi G_N}\,V(x)\right).$$2

rather than a one-dimensional closed-universe Hilbert space. Related abstract work reformulates the gravitational path integral as a map from sources to complex numbers and introduces partial sources that glue into full sources; universes with spatial boundaries, prescribed observer worldlines, or observers entangled with external systems then admit nontrivial $$\hat O^{\rm Phys}(x)=\hat O\!\left(x+\sqrt{16\pi G_N}\,V(x)\right).$$3-sector Hilbert spaces with noncommuting operators (Akers et al., 12 Mar 2025, Chen, 21 May 2025).

This algebraic program is explicitly observer-relative. In the quantum-reference-frame treatment of gravitational entropy, the low-energy system consists of clocks $$\hat O^{\rm Phys}(x)=\hat O\!\left(x+\sqrt{16\pi G_N}\,V(x)\right).$$4 and a QFT sector $$\hat O^{\rm Phys}(x)=\hat O\!\left(x+\sqrt{16\pi G_N}\,V(x)\right).$$5 on

$$\hat O^{\rm Phys}(x)=\hat O\!\left(x+\sqrt{16\pi G_N}\,V(x)\right).$$6

and physical states are obtained by group averaging. The Page–Wootters reduction into the perspective of clock $$\hat O^{\rm Phys}(x)=\hat O\!\left(x+\sqrt{16\pi G_N}\,V(x)\right).$$7,

$$\hat O^{\rm Phys}(x)=\hat O\!\left(x+\sqrt{16\pi G_N}\,V(x)\right).$$8

is shown to reproduce the CLPW observer formalism exactly, hence the slogan $$\hat O^{\rm Phys}(x)=\hat O\!\left(x+\sqrt{16\pi G_N}\,V(x)\right).$$9. Different observers, or different sets of observers, generally define different physical subregion algebras and therefore different density operators and entropies; the paper’s central claim is that gravitational entropy is observer-dependent in precisely this sense (Vuyst et al., 2024).

3. Worldlines, clocks, and operational observation in general relativity

In the most traditional sense, a gravitational observer is a timelike worldline equipped with a standard clock. In the chronometric formulation, an observer is a future-pointing timelike curve $\mathrm{II}_\infty$0 with four-velocity $\mathrm{II}_\infty$1 satisfying $\mathrm{II}_\infty$2. The basic observable is the relativistic redshift between emitter and receiver,

$\mathrm{II}_\infty$3

where $\mathrm{II}_\infty$4 is the photon wave covector. In stationary spacetimes with

$\mathrm{II}_\infty$5

the function $\mathrm{II}_\infty$6 is the redshift potential, and for two stationary observers one has the exact relation

$\mathrm{II}_\infty$7

This underlies chronometric geodesy: clock comparisons between ground and space observers can recover the relativistic gravity potential and, in principle, the mass and spin multipole moments of the underlying spacetime (Philipp et al., 2023).

A different operational lesson comes from analyses of uniform gravitational fields. If observers, mirrors, detectors, and comparison clocks are treated as physical systems rather than idealized fixed backgrounds, then a strictly uniform gravitational field is locally unobservable. For clocks in free fall, the gravitational redshift

$\mathrm{II}_\infty$8

is exactly canceled by the gravitational Doppler shift

$\mathrm{II}_\infty$9

For a freely falling light-pulse atom interferometer, atoms and mirror fall together and the interferometer phase is

$F(t)$0

Observable effects arise only from non-gravitational acceleration of the apparatus or from non-uniformity, tidal gravity, and curvature (Asenbaum et al., 2024).

These two lines of work delimit an important distinction. In one case, observers are precision clocks whose mutual redshifts probe geometry; in the other, the same operational discipline is used to show that a uniform field has no local observable content when observer and apparatus are treated consistently. Taken together, they make the observer a physical subsystem rather than a coordinate convention.

4. Tetrads, local frames, and measured fields

In tetrad-based formulations, a gravitational observer is a local orthonormal frame. In TEGR, the tetrad $F(t)$1 determines both the metric

$F(t)$2

and the observer’s four-velocity $F(t)$3. Gravitoelectric and gravitomagnetic fields are defined from the teleparallel superpotential by

$F(t)$4

For a radially freely falling observer in Schwarzschild spacetime, the paper finds

$F(t)$5

while some components with internal indices $F(t)$6 are nonzero. Nevertheless the TEGR gravitational Lorentz force vanishes, as does the gravitational field energy density $F(t)$7; the conclusion is that the freely falling observer is locally dynamically indistinguishable from an observer at rest in Minkowski space, in agreement with the equivalence principle (Spaniol et al., 2011).

A different TEGR analysis reaches a less restrictive conclusion. There the tetrad is again interpreted as an observer congruence, and a freely falling, nonrotating frame is defined by vanishing acceleration tensor $F(t)$8. In exact pp-wave and Wyman spacetimes, however, the TEGR gravitational energy-momentum tensor can be nonzero even in such freely falling frames. For the exact pp-wave, the energy density is

$F(t)$9

which need not be sign-definite, while in Wyman spacetime the free-fall energy density is generally nonzero and differs from the static-frame result. The paper argues that this does not contradict the equivalence principle because a freely falling frame need not be a fully local inertial reference frame in the strong sense required to remove all gravitational observables (Formiga, 2018).

Electromagnetic measurements in Schwarzschild spacetime exhibit the same observer dependence. In a tetrad formulation of Maxwell theory, static observers and radially free-falling observers decompose the same covariant field equations differently. For static observers, Schwarzschild geometry modifies temporal and radial sectors through the lapse function $U=F(t+z),\qquad V=F(t-z),$0, with $U=F(t+z),\qquad V=F(t-z),$1 representing proper-time differentiation and $U=F(t+z),\qquad V=F(t-z),$2 proper radial differentiation. For radially free-falling observers, an additional local radial Lorentz boost with

$U=F(t+z),\qquad V=F(t-z),$3

mixes charge density with radial current and intertwines electric and magnetic sectors in the temporal-radial projections, while the angular Ampère–Maxwell and Faraday equations retain the static-observer structure (Carneiro et al., 10 Jun 2026).

A closely related invariant notion is that of the principal observer. In electromagnetism the Poynting vector $U=F(t+z),\qquad V=F(t-z),$4 measured by an observer may vanish; in gravity the analogous object is the Weyl super-Poynting vector

$U=F(t+z),\qquad V=F(t-z),$5

For Petrov type I or D spacetimes, observers with $U=F(t+z),\qquad V=F(t-z),$6 exist. In type D, the observer canonically associated with an arbitrary observer $U=F(t+z),\qquad V=F(t-z),$7 is reached by a boost in the direction of $U=F(t+z),\qquad V=F(t-z),$8,

$U=F(t+z),\qquad V=F(t-z),$9

This defines an observer class singled out by vanishing super-energy flux of the Weyl tensor (Wylleman et al., 2020).

5. Accelerated, horizon, and extended observers

Acceleration introduces a distinct family of gravitational observers: those for whom horizons and thermal effects are intrinsic. In Rindler space, a matter shock wave without planar symmetry can implant supertranslational hair on the horizon. The observers considered there are required to remain uniformly linearly accelerated through the wave, in the sense of the curved-spacetime Letaw–Frenet equations. Before the wave, they are orbits of a single boost Killing vector; after the wave, each still follows the orbit of a boost Killing vector, but the boost is generally different from trajectory to trajectory. The post-wave trajectory carries a transverse drift proportional to $$H = \frac{d}{16\pi G}\lim_{r\to\infty} r^{d-2}\int h_{00}(r,t,\Omega)\,d^{d-1}\Omega.$$0, and the corresponding observer-dependent Lorentz transformation is the classical memory effect (Kolekar et al., 2017).

In the near-extremal black-hole throat, the JT construction of fiducial observers provides the horizon-exterior analogue. These observers are accelerated and boundary-anchored; for each off-shell $$H = \frac{d}{16\pi G}\lim_{r\to\infty} r^{d-2}\int h_{00}(r,t,\Omega)\,d^{d-1}\Omega.$$1 they foliate the exterior and define the entangling surface used to compute thermal-atmosphere entropy. At fixed topology the matter entropy diverges near the horizon, but after including quantum-gravitational wormhole contributions the entropy saturates at a finite plateau. The corresponding proper distance

$$H = \frac{d}{16\pi G}\lim_{r\to\infty} r^{d-2}\int h_{00}(r,t,\Omega)\,d^{d-1}\Omega.$$2

defines a quantum version of the stretched horizon in the model (Mertens et al., 28 Jul 2025).

Not all gravitational observers are pointlike. In the laser gravitational compass proposal, the observable content of Einsteinian gravity is identified with curvature, and a detector must access the local tidal matrix $$H = \frac{d}{16\pi G}\lim_{r\to\infty} r^{d-2}\int h_{00}(r,t,\Omega)\,d^{d-1}\Omega.$$3. A standard three-mass Michelson geometry is only planar and accesses at most

$$H = \frac{d}{16\pi G}\lim_{r\to\infty} r^{d-2}\int h_{00}(r,t,\Omega)\,d^{d-1}\Omega.$$4

To probe all three spatial directions one needs at least four non-coplanar mass probes, yielding the full symmetric $$H = \frac{d}{16\pi G}\lim_{r\to\infty} r^{d-2}\int h_{00}(r,t,\Omega)\,d^{d-1}\Omega.$$5 tidal matrix

$$H = \frac{d}{16\pi G}\lim_{r\to\infty} r^{d-2}\int h_{00}(r,t,\Omega)\,d^{d-1}\Omega.$$6

The observer is therefore an extended relational configuration of freely falling masses and light signals rather than a single worldline (Ferreira et al., 2019).

Acceleration also changes quantum information seen by observers. For the entangled state

$$H = \frac{d}{16\pi G}\lim_{r\to\infty} r^{d-2}\int h_{00}(r,t,\Omega)\,d^{d-1}\Omega.$$7

one uniformly accelerated observer sees degradation of mutual information, classical correlation, and quantum discord, but they approach finite nonzero limits at large acceleration. If both observers accelerate simultaneously, the total, classical, and quantum correlations all rapidly go to zero. The paper interprets this through the equivalence principle as evidence that when two observers both remain in a gravitational field, quantum and classical correlations are much harder to preserve (Li et al., 2021).

6. Nonclassical geometry, observer relativity, and unresolved tensions

Observers become especially nontrivial when the gravitational state itself is nonclassical. One proposal models a medium of observers by scalar fields

$$H = \frac{d}{16\pi G}\lim_{r\to\infty} r^{d-2}\int h_{00}(r,t,\Omega)\,d^{d-1}\Omega.$$8

together with a clock field $$H = \frac{d}{16\pi G}\lim_{r\to\infty} r^{d-2}\int h_{00}(r,t,\Omega)\,d^{d-1}\Omega.$$9, so that the full wavefunctional is

$X|0\rangle$0

A relational distance operator $X|0\rangle$1 is then defined to be diagonal on classical geometry-plus-observer configurations. In a superposition of distinct classical geometries, the averaged distance

$X|0\rangle$2

can be locally metric-like but globally non-additive; the paper’s explicit example has

$X|0\rangle$3

The resulting relational geometry therefore fails to arise from any ordinary Riemannian metric, even though each classical branch separately does (Piazza, 2022).

Several current debates concern how far observer-based constructions can be pushed. In the island model, the observer mode naturally yields a Type $X|0\rangle$4 algebra because the added generator has spectrum $X|0\rangle$5 and

$X|0\rangle$6

The paper argues that earlier Type $X|0\rangle$7 constructions, which project the observer Hamiltonian to be bounded below so that $X|0\rangle$8 becomes finite, are physically questionable in that setting because the observer mode is not the Hamiltonian of a standalone system and mediates bidirectional energy transfer between AdS and bath (Geng et al., 13 Jun 2025).

A similar indeterminacy appears in closed-universe constructions. Different observer prescriptions in holographic maps and gravitational path integrals produce different Hilbert-space dimensions and different ontologies for the observer. In one line of work the observer/environment interface acts as an effective holographic boundary; in another, the observer conditions imposed on the path integral are proposed rather than derived. This suggests that the status of the observer is still partly formal rather than uniquely fixed microscopically (Akers et al., 12 Mar 2025, Wei, 26 Jun 2025).

Taken together, these works support a common conclusion. In gravitational theory, an observer is generally a physical reference system—a clock, tetrad, congruence, detector array, Goldstone mode, or quantum reference frame—whose inclusion changes what counts as an observable, what algebra is assigned to a region, and what entropy or geometry is seen. The notion is therefore not peripheral. It is part of the structure by which gravity becomes operationally and algebraically definable.