Stress–Energy Fluctuation Diagnostics

A small expectation value of the stress tensor does not imply a sharply defined gravitational source. The omitted information is the connected stress–energy correlation, which is a bi-distribution and must be smeared or propagated before it becomes a finite diagnostic. The most useful question is not whether a formal pointwise variance is “large,” but whether fluctuations in a declared state, resolution scale, and tensor channel induce metric correlations comparable to the mean geometry or to the accuracy claimed for it.

Required background. Einstein Equations supplies the conserved source equation and geometric scale against which induced perturbations are compared.

Helpful background. Quantum Fields in Curved Spacetime supplies Hadamard states and operator-valued distributions, while Black-Hole Geometry and Mechanics supplies invariant frames and horizon observables for later diagnostics.

The connected stress bi-distribution

For a renormalized stress operator in a state $\rho$, define

$$t_{\mu\nu}(x)=T_{\mu\nu}(x)-\langle T_{\mu\nu}(x)\rangle_\rho,$$

and the symmetrized noise kernel

$$N_{\mu\nu\rho\sigma}(x,y) =\frac12\left\langle \left\{t_{\mu\nu}(x),t_{\rho\sigma}(y)\right\} \right\rangle_\rho.$$

At separated points, the local identity-valued counterterms cancel in $t_{\mu\nu}$. The coincidence limit is nevertheless singular: $N$ is a distribution, not an ordinary tensor field. Hu and Verdaguer state the definition, positivity, conservation, and distributional qualification in Hu and Verdaguer 2008, § 3.2, Eqs. (3.11)–(3.13).

Choose a smooth compactly supported test tensor $f^{\mu\nu}(x)$ and define

$$T[f]=\int d^4x\sqrt{-g}\,f^{\mu\nu}T_{\mu\nu}, \qquad \operatorname{Var}_\rho T[f] =\iint dV_xdV_y\,f^{\mu\nu}(x)N_{\mu\nu\rho\sigma}(x,y)f^{\rho\sigma}(y).$$

This variance is nonnegative for real $f^{\mu\nu}$. Its value depends on the sampling width, observer contraction, state, and field content. Those are physical parts of the question, not nuisances to be removed after the calculation.

Dimensionless diagnostics

A commonly used relative measure is

$$\Delta_f= \frac{\sqrt{\operatorname{Var}T[f]}} {\lvert\langle T[f]\rangle\rvert}.$$

It is informative when the denominator is a stable, nonzero reference source. It becomes meaningless when the mean vanishes by symmetry or subtraction. An absolute alternative starts from a physical observable $Q$ with background value $Q_0$ and linear perturbation $\delta Q[h]$:

$$\mathcal R_Q =\frac{\sqrt{\langle(\delta Q[h])^2\rangle_c}} {\lvert Q_0\rvert+Q_{\mathrm{tol}}}.$$

Here $\delta Q$ is gauge invariant and smeared, and $Q_{\mathrm{tol}}\gt 0$ is a tolerance with the same units as $Q$. Examples include a detector-defined distance variation or an averaged curvature observable. This avoids dividing one coordinate metric component by another. It asks what the detector or geometric diagnostic can actually resolve.

For linearized backreaction, the induced correlator follows by propagation:

$$\langle h_{ab}(x)h_{cd}(y)\rangle_{\mathrm{ind}} =(8\pi G)^2\!\iint dV_{x'}dV_{y'} \,G^{\mathrm{ret}}_{ab}{}^{ef}(x,x') N_{efgh}(x',y')G^{\mathrm{ret}}_{cd}{}^{gh}(y,y').$$

This weights stress fluctuations by the causal susceptibility of the geometry. A large ultraviolet variance can be harmless for a coarse observable, while a modest source near a long-lived response pole can produce a large metric fluctuation.

Conservation, scale, and state checks

Covariant stress conservation gives

$$\nabla_x^\mu N_{\mu\nu\rho\sigma}(x,y)=0, \qquad \nabla_y^\rho N_{\mu\nu\rho\sigma}(x,y)=0,$$

in the distributional sense. This is an independent Ward-identity check on both analytic and numerical kernels. Positivity of $\operatorname{Var}T[f]$ is another. For $N$ independent identical fields, the kernel grows as $N$ but the metric variance with $G=\bar G/N$ falls as $1/N$, recovering the large-$N$ limit.

For a four-dimensional massless scale-invariant theory in Minkowski vacuum, dilate every spacetime dimension of a normalized profile as $f_\tau(x)=\tau^{-4}f(x/\tau)$. The stress two-point function has dimension $L^{-8}$, so $\sqrt{\operatorname{Var}T[f_\tau]}\propto\tau^{-4}$, while the normal-ordered mean is exactly zero. A state with a thermal or occupation scale generally gives a different mean dependence; dimensional analysis alone cannot assign it a nonzero $\tau^{-4}$ value. Shrinking the sampling region probes increasingly singular short-distance structure. Hu and Verdaguer 2008, §§ 5.1–5.2 develops the point-separated kernel and explains why smearing is essential.

Interpretation and limits

Stress fluctuations are a warning about deterministic mean-field gravity, not by themselves proof of quantum spacetime. The Gaussian noise kernel captures the second cumulant. Non-Gaussian states may require higher stress correlators, and graviton self-fluctuations are not encoded in a matter noise kernel. Linear propagation can also fail once the induced perturbation is no longer small.

Fixed-background QFT determines the state and its stress correlators; this volume asks how they affect the metric. Fine-grained information questions and microscopic gravitational fluctuations remain outside this diagnostic. Anderson, Molina-París, and Mottola provide the complementary response-based validity test in Anderson, Molina-París, and Mottola 2003, §§ III–V.

Exercise

Suppose a smeared stress observable has mean zero and variance $\sigma^2$, while its induced gauge-invariant metric observable is $Q=\chi T[f]$ with real susceptibility $\chi$. Which fluctuation diagnostic remains meaningful, and how does it change when the sampling function is doubled?

Solution

The relative stress ratio $\Delta_f$ is undefined because its denominator vanishes. The induced rms metric signal $\sqrt{\langle Q^2\rangle}=\lvert\chi\rvert\sigma$ remains meaningful and should be compared with a geometric tolerance or background scale. If the test tensor is multiplied by two without changing its shape, both $T[f]$ and its rms value double, so the induced rms signal doubles. This linearity is a normalization check; it says nothing about changing the sampling width, which alters the kernel’s scale dependence.

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.