Limits of Semiclassical Gravity

Semiclassical gravity can mean a controlled approximation in which a classical geometry responds to quantum expectation values, or a proposed fundamental coupling of classical and quantum systems. These claims have different standards of validity. A useful research question specifies the dynamics and asks which fluctuation, coherence, or consistency test separates it from a matched alternative. Literature checked 11 October 2026: selected evidence includes the 2023 hybrid-dynamics results, a January 2026 interferometry proposal, and a stochastic-gravity preprint revised in September 2026; none establishes that gravity is fundamentally classical.

Helpful background. Quantum foundations and gravity distinguishes external fields and dynamical sources; gravity-mediated entanglement addresses a separate channel witness.

Mean-field equations and their domain

A schematic semiclassical equation is

$$G_{\mu\nu}[g]+\Lambda g_{\mu\nu} =8\pi G\,\langle T_{\mu\nu}\rangle_{\mathrm{ren}}$$

in units $c=1$. This suppresses the local curvature terms required in a renormalized gravitational effective action. State choice, boundary conditions, the derivative expansion, and stress-energy correlations matter. The fluctuation diagnostic is not simply a pointwise ratio of stress-tensor variance to squared mean: that ratio can be singular when the mean vanishes, and composite operators require a specified renormalization and smearing prescription. The response of suitably averaged metric observables is the relevant question.

Established effective-theory framework: the Hu–Verdaguer review develops stochastic semiclassical gravity using a stress-tensor noise kernel and an Einstein–Langevin response. The stochastic source represents quantum matter fluctuations in this approximation. It must not be identified automatically with the fundamental classical spacetime noise of a postquantum theory. Failure of a mean-field approximation does not imply that every classical–quantum model fails.

The nonrelativistic Schrödinger–Newton model uses a state-dependent gravitational potential. Its deterministic nonlinear evolution is not identical to a linear, completely positive stochastic hybrid theory. Applying a constraint derived for one class to another without checking the assumptions is invalid.

Evidence and competing theoretical routes

Oppenheim constructs a stochastic classical–quantum framework. Oppenheim, Sparaciari, Šoda, and Weller-Davies derive a tradeoff between quantum decoherence and classical diffusion under specified positivity and dynamical assumptions. Established within that framework: the joint classical–quantum dynamics is linear, completely positive, normalization preserving, and Markovian, with the generator and short-time expansion assumptions stated in their derivation. Its quantum marginal need not be Markovian after classical degrees of freedom are eliminated. A displayed coefficient inequality is not automatically a universal bound on every non-Markovian model.

The Oppenheim–Sajjad preprint, revised 17 September 2026, studies stochastic modes about Minkowski space and constraints on coupling combinations within that program. Positive dynamical-mode action is a necessary consistency check; it is not a proof of every nonlinear or ultraviolet consistency property. Liu and colleagues’ January 2026 article targets Schrödinger–Newton phenomenology through correlations associated with common/differential interferometer motion. These are model analyses and proposed tests, not observations selecting either framework.

Active: well-posed dynamics, conservation and constraint consistency, controlled approximation schemes, and observable discriminators that survive ordinary noise. Experimentally unresolved: whether any of these particular departures occurs in nature. An exclusion of one specified parameterization does not establish quantization of gravity without an argument covering the remaining alternatives.

A matched-force calculation isolates what noise adds

As a calibration model, let a probe experience the same mean force in two descriptions, with one adding a zero-mean Gaussian classical force $\xi(t)$ satisfying

$$\langle\xi(t)\xi(t')\rangle=2D\,\delta(t-t'), \qquad D\ge0.$$

The interaction is $-x\xi(t)$, and $D$ has units momentum squared per time. Averaging unitary realizations gives the noise contribution

$$\left.\dot\rho\right|_{\mathrm{noise}} =-\frac{D}{\hbar^2}[x,[x,\rho]].$$

For two narrow prescribed branches held at separation $\ell$, the random phase has variance $2D\ell^2T/\hbar^2$. Thus

$$\frac{V(T)}{V_0}=e^{-D\ell^2T/\hbar^2}, \qquad \left.\frac{d\langle p^2\rangle}{dt}\right|_{\mathrm{noise}}=2D.$$

The second identity follows from $[x,[x,p^2]]=-2\hbar^2$. For a free probe after subtracting its deterministic drift, $\Delta\operatorname{Var}(p)=2DT$. The two measurements therefore test a single assumed noise coefficient. Trapping, friction, colored noise, or path disturbance changes this relation and must enter the forward model.

This is an exactly defined random-unitary null model with an external force; it includes no self-consistent gravitational backreaction. Its dephasing–heating relation is consequently not the hybrid decoherence–diffusion theorem. It is a useful control because a purely classical environmental force can reproduce both signatures. A gravitational interpretation needs a specified spatial correlation, source dependence, or other prediction that distinguishes the proposed mechanism from calibrated ordinary forces.

An entry project: two models with the same mean force

Baseline. Reproduce the white-force calculation by Gaussian phase averaging and by density-matrix evolution. Compare with the deterministic matched-force case. Start with one separation and a free probe; add a trap only after accounting for its momentum response.

Resources. A laptop, stochastic differential equations, and density matrices. A two-branch model suffices for coherence; momentum heating needs a converged wavepacket basis. A backreacting hybrid extension additionally requires supervised study of complete positivity.

Deliverable. An analytic comparison and notebook reporting the mean phase, visibility, and momentum variance with sampling errors. Check $D\to0$, timestep convergence, and basis truncation separately. Then replace white noise with one specified measured or synthetic spectrum and determine whether the two observables identify its amplitude. Report when unknown environmental noise leaves only an upper bound rather than a mechanism identification.

Failure criterion. Stop if ensemble positivity or normalization fails, heating reaches the basis boundary, or the discrepancy comes from inconsistent initial states. A Monte Carlo discrepancy smaller than its sampling uncertainty is not evidence for a different master equation. A noise excess attributed to gravity must survive conventional sources with independently constrained parameters.

Research extension. Select one explicit backreacting generator from the cited hybrid literature, write its coefficient positivity conditions, and match its mean force to the baseline. Demonstrate positivity on the joint classical–quantum state, not only on the reduced density matrix; sampling a few initial states does not prove complete positivity. A new result could bound identifiable parameter combinations using measured noise or expose a controlled approximation’s failure. Merely adding arbitrary Gaussian noise and naming it gravity is not a new hybrid theory.

Choosing a defensible claim

For an effective-theory project, quantify the error relative to a better-controlled quantum model. For a fundamental-model project, state the altered dynamical assumptions and their empirical consequences. For a phenomenology project, specify which parameters and noise kernels are constrained.

An assessment should change when an independently reproduced constraint removes a meaningful model region, a consistency obstruction is proved under explicit hypotheses, or a controlled approximation is extended. A calculation titled “classical gravity” is not evidence for all classical-gravity alternatives.

References

• B. L. Hu and E. Verdaguer, “Stochastic Gravity: Theory and Applications,” Living Reviews in Relativity 11, 3 (2008). DOI — stress-tensor noise, metric response, and semiclassical validity.
• Y. Liu, Y. Chen, K. Somiya, and Y. Ma, “Testing the quantum nature of gravity through interferometry,” Physical Review D 113, 022002 (2026). Primary proposal.
• J. Oppenheim, “A postquantum theory of classical gravity?” Physical Review X 13, 041040 (2023). arXiv:1811.03116.
• J. Oppenheim and M. Sajjad, “Stochastic modes in postquantum classical gravity,” arXiv:2605.05375v2 (17 September 2026), preprint. Model-specific analysis.
• J. Oppenheim, C. Sparaciari, B. Šoda, and Z. Weller-Davies, “Gravitationally induced decoherence vs space-time diffusion: testing the quantum nature of gravity,” Nature Communications 14, 7910 (2023). Article.

Entry points

Read the dynamical assumptions and positivity conditions before the proposed exclusion plot. Reproduce a finite toy model before attempting gravitational field constraints.