Semiclassical Breakdown and Quantum Gravity

Semiclassical predictions cease to be parametrically controlled when an omitted effect changes a declared observable by order one relative to the retained result or its accuracy target. There is no universal scalar “breakdown meter”: curvature, frequency, stress fluctuations, response poles, secular growth, and state regularity test different failure channels. The conservative conclusion is operational—identify the first failed hierarchy—rather than the stronger claim that every Planckian-looking intermediate quantity destroys every low-energy observable.

Required background. Einstein Equations supplies the mean geometric equation whose corrections are being compared.

Helpful background. Quantum Fields in Curved Spacetime supplies admissible states and renormalized observables, while Black-Hole Geometry and Mechanics supplies invariant horizon and curvature diagnostics.

Observable criteria for loss of control

Let $Q$ be a gauge-invariant, smeared observable and write a controlled expansion as

$$Q=Q_0+\epsilon Q_1+R_2, \qquad \frac{\lvert R_2\rvert}{\lvert Q_0+\epsilon Q_1\rvert+Q_{\mathrm{tol}}}\ll1.$$

Here $Q_{\mathrm{tol}}\gt 0$ is the resolution relevant to the claim. An order-one estimated remainder demonstrates failure of this expansion for that observable. If no estimate is available, control is unestablished; that alone does not prove physical breakdown. Useful warning signs are:

• a measured curvature component or resolved proper frequency reaches the declared EFT cutoff; $\ell_{\mathrm P}^2\lvert R_{\hat a\hat b\hat c\hat d}\rvert\sim1$ is one example when the cutoff is Planckian;
• a response-weighted metric variance $\langle(\delta Q[h])^2\rangle_c$ becomes comparable to the squared background observable or tolerance;
• a causal response pole below the cutoff produces an unbounded gauge-invariant mode;
• secular corrections such as $\epsilon t/\tau$ become order one even though instantaneous curvature is small;
• the state ceases to be Hadamard in the region where a local renormalized stress tensor is required; or
• allowed finite higher-order terms change the answer as much as the retained quantum effect.

These are observable- and state-relative conditions. A coordinate component, an unsmeared stress variance, or a high-frequency root introduced by a truncated higher-derivative equation is not by itself decisive.

Results and open questions as of 2026-08-19

Evidence or result	Regime and assumptions	What it supports	Limitation or unresolved boundary
Hu and Verdaguer 2008, §§ 3.2–3.3.1, Eqs. (3.16)–(3.20)	$N$ independent free fields, $NG$ fixed, perturbations about a self-consistent background	mean backreaction survives while leading metric variance is $O(1/N)$	does not control graviton loops, strong cross-species correlations, or trans-cutoff modes
Anderson, Molina-París, and Mottola 2003, §§ III–V, especially Eqs. (3.4) and (4.6)	Minkowski vacuum, renormalized couplings, gauge-invariant perturbations, retarded prescription	no unacceptable growing low-energy mode under the stated prescriptions	stability of one background and state is not a state-independent theorem
Hu and Verdaguer 2008, § 3.2, Eqs. (3.11)–(3.16)	Gaussian second cumulant, smeared observables, linear response	detects induced metric fluctuations hidden by a small mean stress tensor	higher cumulants and nonlinear metric fluctuations can matter
Parker and Simon 1993, § 2, Eqs. (2.10)–(2.13)	corrections truncated at a stated order in $\hbar$ or curvature	removes high-frequency branches not analytic in the expansion parameter	cannot prove that every extra mode of a UV completion is absent
Anderson, Molina-París, and Mottola 2003, §§ II and V, pp. 7, 13–14	invariant local scale compared with a declared cutoff	marks loss of the derivative expansion for local observables	low-energy inclusive observables can sometimes remain insensitive to short-distance details

The large-$N$, noise, and response statements are developed together in Hu and Verdaguer 2008, §§ 3.2–3.3.1 and 6.5. The flat-space criterion requires a causal closed-time-path response and removal or interpretation of cutoff-scale roots, as shown in Anderson, Molina-París, and Mottola 2003, §§ III–V. These results establish controlled domains and necessary tests; they do not establish one background-independent boundary between semiclassical gravity and quantum gravity.

Conservative and permissive interpretations

The conservative position treats any failed necessary condition as a stop sign. If a Hadamard stress tensor is undefined, a below-cutoff physical perturbation grows without bound, or the response-weighted metric variance is order one, deterministic semiclassical geometry cannot support the claimed precision. This position minimizes extrapolation but may discard coarse observables that are protected from ultraviolet sensitivity.

The permissive position asks whether the particular observable remains insensitive after resummation, dressing, or coarse graining. A trans-cutoff intermediate frequency, for example, need not contaminate an infrared flux if the result is regulator independent and reproducible in several low-energy descriptions. This is plausible effective-field-theory reasoning, but it must be demonstrated rather than asserted. A formal cancellation in one subtraction scheme is not an independent check.

Both positions agree that small mean backreaction is insufficient. They differ on how much observable robustness is needed before a locally failed hierarchy can be bypassed. The answer depends on the quantum state, sampling norm, response prescription, solution branch, and cutoff—not simply on whether the geometry contains a horizon.

Adverse cases and independent checks

A useful counterexample to a mean-only criterion is a state with $\langle T[f]\rangle=0$ but $\operatorname{Var}T[f]\gt 0$. The mean metric can equal the classical solution even though induced metric correlations are nonzero. Conversely, an unsmeared coincidence variance diverges even in the Minkowski vacuum; that divergence does not prove macroscopic metric breakdown because it has not specified an observable resolution.

Higher-derivative runaways provide a second discrimination. Parker and Simon show that solutions nonanalytic in the perturbative parameter can be artifacts of treating a truncated equation as exact Parker and Simon 1993, § 2. Removing such a root by order reduction can restore low-energy control, but a genuine retarded pole that remains below the cutoff and appears in a gauge-invariant observable cannot be dismissed by the same argument.

The strongest practical check combines three methods: estimate the first omitted local operator, compute causal linear response, and propagate a smeared noise kernel. Agreement in a classical or large-$N$ limit is necessary; stability under reasonable changes of sampling scale and finite renormalization is also required.

When quantum gravity is needed

Once no controlled metric-and-matter expansion exists, microscopic dynamics requires Quantum Gravity rather than a more elaborate semiclassical ansatz. Questions about unitary fine-grained recovery are studied separately in Black-Hole Information. The breakdown of an approximation does not imply that quantum gravity begins at a unique surface or proper time. It means that semiclassical equations no longer determine the requested observable with a quantified remainder.

Topology change, singularity resolution, and final endpoint degrees of freedom therefore remain unresolved here. Fixed-background Hawking calculations also remain in Quantum Fields in Curved Spacetime; their backreaction is treated only while the state, renormalized source, and scale hierarchy stay controlled.

This page should be reassessed after any of the following events: a state-independent theorem relates stress fluctuations to bounded metric observables; an independent nonlinear calculation overturns a linear-response stability result; a reproduced backreaction solution crosses a previously uncontrolled regime; or a microscopic theory derives a low-energy matching condition that changes a stated cutoff criterion. A correction or withdrawal of a central source also triggers reassessment.

Exercise

A calculation has $\ell_{\mathrm P}^2\lvert R\rvert=10^{-6}$ and $G\lvert\langle T\rangle\rvert L^2=10^{-4}$, but a smeared gauge-invariant metric observable has rms fluctuation $0.8$ times its mean and grows secularly as $t^{1/2}$. Is the mean-field prediction controlled for arbitrarily long times?

Solution

No. The curvature and mean-backreaction ratios show that two local expansion parameters are small, but the response-weighted fluctuation is already order one for precision claims and grows further with time. The deterministic mean metric may remain a useful coarse central value over a bounded interval, yet its claimed small uncertainty is not controlled. One must specify a tolerance and solve for the time at which the rms-to-mean ratio reaches it; low curvature alone cannot extend the result indefinitely.

References

• Anderson, P. R., Molina-París, C., and Mottola, E. 2003. “Linear Response, Validity of Semi-Classical Gravity, and the Stability of Flat Space.” Physical Review D 67, 024026. Open PDF.
• Hu, B. L., and Verdaguer, E. 2008. “Stochastic Gravity: Theory and Applications.” Living Reviews in Relativity 11, 3. DOI.
• Parker, L., and Simon, J. Z. 1993. “Einstein Equation with Quantum Corrections Reduced to Second Order.” Physical Review D 47, 1339–1355. Open PDF.