Gravity-Mediated Entanglement: Evidence and Open Questions

Gravity-mediated entanglement proposals ask whether two initially separable systems can acquire entanglement attributable to their gravitational interaction. The inference concerns a specified mediator and communication model; it is not an unconditional test of every theory called classical gravity. Reviewed 11 October 2026, this assessment includes the 2025 Aziz–Howl claim, the August 2026 Gundhi–Infantino–Bassi criticism, and an October 2026 experimental requirements preprint.

Helpful background. The stable gravity-mediated-entanglement treatment derives the branch-phase mechanism and its assumptions.

The interaction-dependent quantity

For four weak-field, nonrelativistic branches, define

$$\Phi=\phi_{00}+\phi_{11}-\phi_{01}-\phi_{10}.$$

Local one-body phases cancel from this combination. In the ideal equal-amplitude pure-state model, concurrence is $|\sin(\Phi/2)|$. Real experiments must infer an entanglement witness after imperfect splitting, evolution, recombination, and readout; a measured phase in one interferometer is insufficient.

The scale of the nonlocal phase matters. For pointlike masses with branch positions $x_A=as$ and $x_B=d+bt$, where $s,t\in\{0,1\}$ and $a,b\ll d$, expanding the four Newtonian phases gives

$$|\Phi|\simeq\frac{2Gm_A m_B T\,ab}{\hbar d^3}.$$

This curvature dependence can make the entangling phase much smaller than an individual branch phase of order $Gm_A m_BT/(\hbar d)$. The expansion requires nonoverlapping, narrow packets and nearly fixed separations during the hold. Actual trajectories require time integration; a large common phase does not compensate for a small $\Phi$.

The Bose and Marletto–Vedral proposals establish theoretical witness routes. Their inference requires identifying initially separable subsystems and excluding unintended quantum communication between them. A classical-channel model restricted to local operations and classical communication cannot generate entanglement from a separable state. An arbitrary operator-valued two-body potential already supplies an entangling interaction and does not identify its mediator by itself.

Marletto and Vedral’s 2025 review develops the information-theoretic interpretation and compares two-probe entanglement schemes with single-probe tests. It is an entry point to the witness assumptions; the subsequent dispute below requires examining the later primary models as well.

A witness margin, rather than an ideal entanglement curve

At $\Phi=\pi$, the ideal output is maximally entangled and can be mapped by known local rotations to $|\Phi^+\rangle=(|00\rangle+|11\rangle)/\sqrt2$. In these calibrated measurement axes, its fidelity is

$$F=\frac{1+\langle X\otimes X\rangle -\langle Y\otimes Y\rangle+\langle Z\otimes Z\rangle}{4}.$$

Every separable two-qubit state obeys $F\le1/2$, because its overlap with a maximally entangled state is at most $1/2$. Thus $W=I/2-|\Phi^+\rangle\langle\Phi^+|$ is a measurable witness. This algebraic statement is established; achieving a negative expectation with gravitational attribution is experimentally unresolved in the cited proposals.

For the diagnostic mixture $\rho=(1-p)|\Phi^+\rangle\langle\Phi^+|+pI/4$, $F=1-3p/4$, so the witness detects entanglement for $p<2/3$. White mixing is a transparent software test, not a realistic noise model. Local phase noise, leakage, loss, and correlated readout have different effects. Away from the maximally entangled target, the rotations and appropriate witness must be recalculated; applying this benchmark unchanged can miss real entanglement.

Finite counts must establish a lower confidence bound on $F$ above $1/2$, with calibration uncertainty included. Choose measurement axes using independent calibration data. Optimizing them on the test data without accounting for selection can overstate significance. A witness can certify the state and still leave its interaction mechanism unidentified.

A live dispute about mechanism and partition

Aziz and Howl’s 2025 paper argues that a field-theoretic matter description allows entanglement in a classical gravitational theory. Gundhi, Infantino, and Bassi’s revised August 2026 preprint challenges that interpretation: in their analysis, particle diffusion and the chosen subsystem partition generate the entanglement, including when the interaction potential is absent. This is a disputed mechanism, not an established general refutation of mediator arguments.

The disagreement identifies an actionable research requirement. Define the Hilbert-space partition, particle-transfer channels, initial correlations, and the interaction-off control before attributing generated entanglement to gravity. Spatial regions, particle labels, and modes are not interchangeable partitions.

The October 2026 QGEM white paper discusses required interferometry and low-frequency noise suppression. It contributes a feasibility agenda rather than a report of the target witness.

An entry project: isolate the entangling channel

Baseline. Implement a four-branch phase model and reproduce the ideal concurrence. Add calibrated local dephasing and calculate a measurable witness rather than using pure-state concurrence on a mixed state.

Resources. A laptop, two-qubit density matrices, partial transposition, and elementary open-system evolution. For the disputed field-theory mechanism, additional many-body and mode-entanglement background is essential.

Deliverable. First reproduce the $p=2/3$ diagnostic threshold and cross-check mixed-state entanglement with the partial-transpose criterion. Generate finite-count correlators and check interval coverage on known separable states. Then make a channel audit with interaction on/off and bounded nuisance couplings. A separate, more advanced lattice project can add transfer allowed/forbidden and fixed/changed partition controls to study the disputed diffusion mechanism; do not mistake that many-body task for the two-qubit baseline.

Failure criterion. If the witness remains when the claimed gravitational mechanism is removed, it cannot be assigned uniquely to that mechanism. If subsystem definitions change between initial and final calculations, justify the operational meaning before reporting entanglement growth.

Research extension. Derive a robust inequality distinguishing a bounded set of classical-channel models from the proposed dynamics with imperfect readout and nuisance interactions. Alternatively, compute how an independently constrained electromagnetic channel limits gravitational attribution. Specify the nuisance bound before examining the witness and test worst-case allowed channels. The basic witness and controls are training baselines; novelty requires an improved bound or an experimentally feasible discriminator absent from the cited analyses.

Update triggers and project selection

The decisive experimental advance would be an interaction-dependent witness with a complete channel and noise budget. The decisive theoretical advance would clarify which assumptions are sufficient, necessary, or violated by a concrete alternative model.

A student should first reproduce an existing witness or disputed toy model. Designing a full gravity experiment or claiming a universal theorem about all classical gravity theories is far beyond that initial project. Valuable progress can instead be a corrected partition, a tighter nuisance bound, or an experimentally accessible control that resolves a specific ambiguity.

References

• J. Aziz and R. Howl, “Classical theories of gravity produce entanglement,” Nature 646, 813–817 (2025). Primary claim.
• S. Bose et al., “Spin entanglement witness for quantum gravity,” Physical Review Letters 119, 240401 (2017). Proposal.
• S. Bose et al., “White paper: 1-10 Hz matter-wave interferometer to test the spin entanglement witness for quantum gravity,” arXiv:2610.07422v1 (2026), preprint. Requirements.
• A. Gundhi, G. Infantino, and A. Bassi, “Can classical theories of gravity produce entanglement?” arXiv:2604.19696v3 (7 August 2026), preprint. Critical analysis.
• C. Marletto and V. Vedral, “Gravitationally induced entanglement between two massive particles is sufficient evidence of quantum effects in gravity,” Physical Review Letters 119, 240402 (2017). Proposal.
• C. Marletto and V. Vedral, “Quantum-information methods for quantum gravity laboratory-based tests,” Reviews of Modern Physics 97, 015006 (2025). DOI — witness principles, variants, and interpretive assumptions.

Entry points

Read the proposal and the criticism with the same subsystem diagram beside them. Reproduce the interaction-off control before deciding what either calculation implies about a mediator.