Gravitationally Generated Entanglement

• Gravitationally generated entanglement is the emergence of nonseparable quantum correlations mediated solely by gravity, indicating that the gravitational mediator must exhibit nonclassical properties.
• Various experimental schemes, from interferometric branch-phase setups to many-body optomechanical models, enable quantitative probes of entanglement through controlled phase shifts and Gaussian correlations.
• Theoretical models show that factors like spacetime curvature, retardation effects, and gravitomagnetic interactions critically influence entanglement, offering insights into the quantum nature of gravity.

Searching arXiv for recent and foundational papers on gravitationally generated entanglement. Gravitationally generated entanglement is the production of nonseparable quantum correlations between subsystems whose only relevant coupling is gravitational. In contemporary usage, the term most often refers to branch-dependent phase accumulation in interferometric superposition experiments, but the literature also includes entanglement induced by quantized spacetime vacuum fluctuations, retarded metric perturbations of light pulses, gravitomagnetic frame dragging, optomechanical and phononic many-body realizations, and curved-spacetime generalizations. Its foundational significance is tied to the claim that, under local mediation and absence of direct interactions, a mediator capable of entangling two quantum systems cannot be purely classical in the sense of possessing only a single observable (Marletto et al., 2017).

1. Conceptual foundations

The modern subject was sharply formulated by Marletto and Vedral, who argued that if two initially separable quantum systems become entangled solely through a mediator that interacts locally with each of them, then that mediator must possess at least two non-commuting observables; in that operational sense, it cannot be purely classical (Marletto et al., 2017). Their proposal distinguished gravitationally generated entanglement from earlier gravitational phase experiments such as single-particle interferometric phase shifts, which can be produced by a classical background field and therefore do not by themselves witness nonclassicality of gravity.

This foundational argument is narrower than a full derivation of quantum gravity. It concerns the mediator properties needed to generate entanglement, not the full Hilbert-space structure of gravity, the existence of gravitons as asymptotic observables, or the validity of any particular quantization scheme. Several later papers retain this distinction explicitly: observation of gravity-induced entanglement is presented as evidence for nonclassical features of the mediator, not as a complete proof of a fundamental theory of quantum gravity (He et al., 2023).

A complementary line of work studies gravity not as a static potential but as a fluctuating quantum environment. In linearized quantum gravity, two gravitationally polarizable two-level subsystems can become entangled through the vacuum fluctuations of the gravito-electric tensor $E_{ij}=C_{i0j0}$. In that open-system setting, entanglement generation near the initial time depends crucially on the quadrupolar polarizations of the subsystems, their geometry, and boundary conditions: the subsystems may fail to entangle when their polarizations differ in certain ways, while a boundary can enable entanglement for parallel alignment that is absent in free space (Cheng et al., 2018). This places gravitationally generated entanglement within the broader theory of environment-induced quantum correlations.

2. Standard branch-phase mechanism

The canonical mechanism is the two-mass interferometric scenario. Each mass is prepared in a spatial superposition of two localized states, often denoted $|L\rangle$ and $|R\rangle$, so that the initial two-body state is a product superposition over the four branches $|LL\rangle$, $|LR\rangle$, $|RL\rangle$, and $|RR\rangle$. Because the Newtonian interaction energy depends on branch-dependent separation, each branch accumulates a different phase. Entanglement appears when the phase pattern cannot be factorized into single-particle contributions; a standard nonlocal combination is

$$\Delta\phi=\phi_{LL}+\phi_{RR}-\phi_{LR}-\phi_{RL},$$

with $\Delta\phi\neq 0$ generating entanglement (Brahma et al., 2023).

A resource-theoretic reformulation makes the same mechanism more explicit. In the Bose-type orbital-qubit model, gravity acts as a diagonal branch-dependent unitary,

$$U_G=\sum_{i,j\in\{L,R\}} e^{i\phi_{ij}} |ij\rangle\langle ij|,$$

which preserves total coherence of the closed bipartite pure state while converting initially local spatial coherence into bipartite entanglement. For equal initial superpositions, the reduced local $|L\rangle$0-coherence and the negativity obey the exact complementarity relation

$|L\rangle$1

and maximal entanglement requires maximal initial local coherence (Dolatkhah et al., 10 Feb 2026). This interpretation does not change the dynamics; it identifies the pre-existing local superposition coherence as the resource that gravity redistributes into nonlocal correlations.

The same branch-phase mechanism can also be recast in explicitly interferometric language. Rostom argues that the decisive operational step is destructive interference: by tuning controllable phases to the Pancharatnam-relative-phase shifts, one can eliminate selected output amplitudes and expose a pure entangled state of the form

$|L\rangle$2

with the entanglement tied to a sign change associated with destructive interference (Rostom, 2024). This does not replace the branch-phase picture; it isolates how those phases become experimentally manifest at the interferometer output.

Continuous-variable models show that large cat-like superpositions are not the only route. Two trapped or released masses interacting through the quadratic term of the Newtonian potential can become entangled through Gaussian correlations, and in the released-mass case detectable entanglement can accumulate on timescales shorter than coherence times, even though no macroscopic quantum superposition develops during the evolution (Krisnanda et al., 2019). This broadens the subject beyond qubit-like path encoding.

3. Mediator-explicit weak-field formulations

A major development has been the replacement of effectively instantaneous Newtonian descriptions by explicitly local weak-field mediator models. In linearized gravity around Minkowski spacetime,

$|L\rangle$3

two laser pulses can source weak retarded metric perturbations and become entangled through the interaction Hamiltonian

$|L\rangle$4

In that proposal the sources are massless radiation pulses rather than massive particles, the mediator is explicitly propagating and retarded, and the conditional phases are engineered with paired Holometer-like interferometers. The resulting entanglement witness depends on the retardation delay $|L\rangle$5, and increasing $|L\rangle$6 decreases the induced entanglement (He et al., 2023). The proposal is presented as a locality-aware proof of principle rather than a near-term experiment.

A related relativistic extension moves beyond the Newtonian scalar sector to gravitomagnetism. In a frame-dragging proposal, a rotating source mass is placed in a superposition of opposite angular momenta and a probe traverses a circular interferometer in a superposition of opposite orbital directions. In the large-angular-momentum limit, the Lense–Thirring interaction generates the entangling phase

$|L\rangle$7

and the same phase is recovered from the on-shell action of local linearized quantum gravity in a stationary-phase treatment. The path-integral derivation makes retardation explicit through causal step functions, so only causally connected portions of the histories contribute (Wakakuwa et al., 30 Jun 2026). This extends gravitationally generated entanglement into a genuinely post-Newtonian regime.

Microscopic propagator-based analyses further sharpen the role of dynamical gravity. For fermionic qubits treated as spin-$|L\rangle$8 particles in spatially separated wave packets, a quantum Boltzmann equation with a graviton propagator shows that entanglement arises from forward-scattering graviton exchange. In that framework, only the dynamical limit of the propagator generates entanglement; the purely static limit leaves an initially factorized state unentangled. In one of the two explicit microscopic models, the entanglement depends on the Larmor frequency of the qubits rather than their masses, and in both models the effect decreases as the wave-packet size increases (Zarei et al., 23 Oct 2025). This suggests that some entanglement claims tied only to static Newtonian sectors may miss the role of genuinely dynamical mediator degrees of freedom.

4. Platforms and implementations

The subject has diversified into a wide range of physical platforms. In optomechanics, Matsumura and Yamamoto analyze a four-cavity, two-oscillator “quantum Cavendish” system in which gravity couples the mechanical modes and the optomechanical interaction maps the resulting joint mechanical energy shift onto a photon-photon conditional phase. The exact reduced photonic state is obtained nonperturbatively under photon-number conservation, and the gravity-induced entangling rate is

$|L\rangle$9

Large photonic entanglement develops on timescales of order $|R\rangle$0, while photon leakage suppresses the negativity as $|R\rangle$1 (Matsumura et al., 2020). The model is entirely Newtonian on the gravity side, but it provides an exact solvable many-body realization of gravity-induced conditional phases.

A different strategy is to infer an interaction channel’s entangling capability without measuring both subsystems. In an atom-interferometer proposal, the atom is a spatial qubit and the second subsystem is a mechanical oscillator subject to a gravitationally induced conditional displacement,

$|R\rangle$2

The key observable is collapse and revival of single-subsystem visibility: under the authors’ semigroup and population-preserving assumptions, such non-monotonic behavior cannot arise from a separable, non-entangling channel. The same framework also shows that thermal occupation of the oscillator can enhance the visibility dip, and that a pre-entangling non-gravitational stage can make the gravitational sensitivity linear in the weak coupling (Carney et al., 2021). This shifts attention from final-state entanglement to entangling power.

Many-body phononic extensions use Bose–Einstein condensates rather than individual masses. In a quantized linearized-gravity model for two separated BECs, the second-order graviton-induced energy shift becomes an operator-valued coupling between Bogoliubov modes, producing a term proportional to $|R\rangle$3 and hence entanglement between phonon sectors. In the truncated two-mode description, the concurrence scales at leading order as

$|R\rangle$4

so short-distance entanglement can be significantly larger than in two-test-mass QGEM, but it also decays much faster with separation because of the Gaussian overlap factor (Sen et al., 22 Apr 2026). This is a many-body enhancement, not merely a reparametrization of the two-particle case.

Not every gravity-related entanglement proposal is mediator-based in the modern sense. An early neutron proposal uses Earth’s gravitational field together with a rectangular cavity to filter ultracold neutrons into a common spatial ground state, after which fermionic antisymmetry forces the spin singlet

$|R\rangle$5

Here gravity is essential in shaping the one-particle level structure, but the entanglement is most accurately described as gravity-assisted state preparation rather than entanglement generated by mutual gravitational interaction (Li et al., 2012). Likewise, discrete-time quantum walks with Newtonian-inspired branch-dependent phases serve as analog models of gravity-mediated entanglement rather than direct realizations of gravitational dynamics (Badhani et al., 2019).

5. Curved spacetime and relativistic generalizations

Once gravity is treated as a field on a nontrivial background, the entangling interaction becomes background dependent. In a conceptual extension of the standard two-mass scenario to FLRW and de Sitter spacetimes, the effective potential depends on the graviton vacuum and mode functions rather than only on a flat-space Newtonian kernel. In de Sitter space the potential is

$|R\rangle$6

where the first term reproduces Newtonian gravity in terms of physical distance and the second is a curvature correction arising from the Bunch–Davies squeezing term. The notable result is that this correction does not alter the entanglement profile at leading nonrelativistic order, even though it changes the effective potential. The authors then argue that this is special to de Sitter and that generic FLRW backgrounds should modify the entanglement profile (Brahma et al., 2023). This makes gravitationally generated entanglement a possible probe of background curvature rather than merely a weak-field flat-space phenomenon.

A more astrophysical generalization places microscopic particles in curved static backgrounds and replaces the constant branch separation of laboratory proposals by a history-dependent separation $|R\rangle$7 determined by geodesic deviation. In the Schwarzschild-background study, the entangling phase becomes

$|R\rangle$8

with $|R\rangle$9 computed from the geodesic deviation equation in cluster or Schwarzschild geometries. The resulting entanglement witness is predicted to oscillate as a function of observed kinetic energy across families of geodesics, producing what the paper calls “characteristic spectra of QGEM” (Zhang et al., 2023). This shifts the control variables from laboratory arm length and hold time to redshift, impact parameter, travel proper time, and mass profile.

These curved-spacetime constructions remain conceptual. They generally assume a fixed classical background, weak-field or nonrelativistic matter dynamics, and highly idealized coherence over long propagation times. Their significance lies less in immediate feasibility than in showing that entanglement generation, if formulated covariantly, depends on propagators, vacuum structure, and geodesic history.

6. Classical reconstructions, limitations, and interpretational disputes

A central controversy concerns what entanglement generation does and does not prove. One line of criticism shows that the standard two-mass branch-phase signal can be reproduced by evolving the initial Wigner function with classical Liouville dynamics under a suitable quadratic approximation to the Newtonian potential. In that framework the reduced-state purity obtained from classical phase-space flow closely matches the usual quantum result on the relevant timescale, leading to the claim that entanglement generation in these Newtonian-potential experiments does not by itself isolate genuinely quantum gravitational dynamics; to do that, one would need sensitivity to higher-order $|LL\rangle$0 terms in the Moyal bracket (Marchese et al., 2024). This does not deny entanglement generation; it disputes the inference from entanglement to uniquely quantum gravitational dynamics.

A different criticism targets claims that a prescribed classical gravitational potential can itself entangle quantum systems. In a fixed-particle-number, nonrelativistic scenario responding to Aziz and Howl, Gundhi, Infantino, and Bassi argue that the apparent entanglement comes from discarding off-diagonal transition amplitudes. When those amplitudes are retained, the evolved coefficients factorize and the state remains separable, so classical gravity in that specific external-potential model does not generate entanglement (Gundhi et al., 21 Apr 2026). The paper is not a universal no-go theorem for all classical models; it is a rebuttal to a particular truncation.

Classical gravitational driving can nevertheless generate entanglement in other senses. In a LIGO-motivated oscillator model driven by a classical gravitational wave, the Hamiltonian contains a two-mode squeezing term

$|LL\rangle$1

and the reduced subsystem purity and Rényi entropy show gravity-induced entanglement-like dynamics enhanced by finite temperature. The authors explicitly note, however, that this is entanglement generated by a classical external gravitational interaction and therefore not evidence that gravity itself is quantum (Dutta et al., 25 Mar 2025). The distinction between “gravity-generated entanglement” and “entanglement witnessing a quantum mediator” is therefore essential.

Across the literature, several common limitations recur. Most analyses work in linearized gravity, weak coupling, static or near-static regimes, idealized two-path or Gaussian-state truncations, and simplified decoherence models. Retardation, losses, realistic mode matching, thermal environments, and competing nongravitational forces are often neglected or treated only schematically. Even the most explicit mediator models state that a positive result would support the nonclassicality of the weak gravitational interaction under locality assumptions, not settle the structure of full quantum gravity (Wakakuwa et al., 30 Jun 2026). Conversely, a null result would not prove that gravity is classical, since the gravitational channel could be entanglement-breaking or experimentally masked (Marletto et al., 2017).

In that sense, gravitationally generated entanglement is best regarded as a family of operational probes of gravity as an interaction channel. The most robust conclusion shared across the field is conditional: if two initially separable systems become entangled solely through local gravitational mediation, then the mediator must carry nonclassical structure. What remains actively contested is which experimental signatures establish that premise most cleanly, whether Newtonian branch phases are sufficient, and how much of the conclusion survives once locality, retardation, curvature, and realistic noise are treated in full detail.