Backreaction of Cosmological Quantum Fields

Cosmological quantum fields appreciably backreact when their renormalized stress changes the expansion by an order larger than the calculation’s error while curvature, response, state regularity, fluctuations, and secular corrections remain controlled. A large infrared logarithm or particle number is not enough by itself. The relevant quantity is a conserved, gauge-invariant source inserted into a self-consistent equation before the approximation that produced it fails.

Helpful background. Quantum Fields in Curved Spacetime supplies states and particle production; Einstein Equations supplies constraints; Black-Hole Geometry and Mechanics supplies invariant horizon comparisons; Semiclassical Gravity supplies response and fluctuations; and Causality and Global Structure supplies the domain of dependence.

A quantitative backreaction criterion

For a spatially flat background satisfying $3H_0^2=8\pi G\rho_0+\Lambda$, compare homogeneous solutions at the same cosmic proper time with $H_0\ne0$. At fixed renormalized couplings and neglecting retained higher-curvature corrections at this step, the small total source change obeys

$$6H_0\,\delta H=8\pi G\,\delta\rho, \qquad \frac{\delta H}{H_0}=\frac{4\pi G\,\delta\rho}{3H_0^2}.$$

Define $\epsilon_{\rm br}=8\pi G|\delta\rho|/(3H_0^2)$. Detectable mean backreaction requires $\epsilon_{\rm br}$ to exceed numerical, subtraction, and truncation errors. Perturbative control simultaneously requires $\epsilon_{\rm br}\ll1$ unless the geometry and state are solved self-consistently rather than expanded about the original background.

“Total” matters: $\delta\rho$ includes any induced response of the background matter, not only the stress of an added field held on the old geometry. In an inhomogeneous perturbation, a change of time slicing also changes $\delta\rho$ and $\delta H$; specify a physical clock or construct a gauge-invariant combination before calling either shift observable.

The causal linear-response equation is more discriminating:

$$\delta G_{ab}[h]+\Lambda h_{ab}+\delta H^{\rm local}_{ab}[h] =8\pi G\left( \delta\langle T_{ab}\rangle_{\rm state} +\int d^4x'\sqrt{-g'}\, \Pi^{\rm ret}_{ab}{}^{cd}(x,x')h_{cd}(x') \right).$$

Here $\Pi^{\rm ret}$ is the complete retarded functional response of the renormalized stress to the covariant metric perturbation $h_{cd}$, including its local operator-contact terms. The finite curvature couplings assigned to $H^{\rm local}_{ab}$ stay on the left and are not counted again in this kernel; the explicit state term describes an independent change of the initial preparation. The resulting equation tests whether a gauge-invariant perturbation grows. Anderson, Molina-París, and Mottola formulate this criterion and show infrared stability for a scalar vacuum around Minkowski space, not for arbitrary cosmologies Anderson, Molina-París, and Mottola 2003, §§ II–V, PDF.

How an effect can accumulate

Nonadiabatic expansion can create particles, giving a state-dependent $\rho$ and $p$. Light, minimally coupled fields can accumulate infrared correlations across many e-folds. Conformal fields contribute local anomaly terms, and interacting fields can generate secular logarithms. These mechanisms are distinct: a logarithm in a coordinate-dependent correlator need not appear in a gauge-invariant expansion observable, while a local anomaly contribution need not represent particles.

Two clocks must be compared. Let $\epsilon_{\rm loop}(N)$ be the largest loop or secular correction after $N$ e-folds, and let $\epsilon_{\rm br}(N)$ measure its gravitational source. For a smeared gauge-invariant metric or expansion observable $H[F]$, let $H_{\rm scale}[F]$ be the background scale resolved by the same profile. A controlled backreaction window exists only if

$$\epsilon_{\rm error}\ll\epsilon_{\rm br}(N)\ll1, \qquad \epsilon_{\rm loop}(N)\ll1, \qquad \frac{\sqrt{\operatorname{Var}H[F]}}{H_{\rm scale}[F]}\ll1$$

for the declared smearing and retarded response. The stress noise kernel is an input to this last ratio, not the ratio itself: a stress mean may vanish, and the response can suppress or amplify different spectral components. If perturbation theory reaches unity before $\epsilon_{\rm br}$ becomes distinguishable, the calculation has found a breakdown time rather than a reliable altered cosmology.

Results and open questions through 2026-08-19

The evidence through 2026-08-19 establishes controlled effects in particular fields, states, and regimes. Whether infrared accumulation generically relaxes or destabilizes a cosmological background remains unresolved; no universal verdict follows.

Evidence and exact locator	Result under stated assumptions	Scope or limitation
Parker and Simon 1993, §§ II–IV, PDF	Order reduction retains perturbative curvature corrections while removing nonanalytic high-frequency roots	Local one-loop EFT analysis; it does not resum long-time state-dependent infrared effects
Anderson, Molina-París, and Mottola 2003, §§ III–V, PDF	Gauge-invariant linear response makes boundedness a validity test; scalar-vacuum Minkowski space is stable for wavelengths larger than the Planck scale after the stated treatment of cutoff-scale roots	The worked result is flat space, and the paper leaves de Sitter response as a further problem
Mottola and Vaulin 2006, §§ II–VI, PDF	Anomaly auxiliary fields expose state-dependent homogeneous solutions and possible macroscopic stress	Boundary data select the state; proposed cosmological implications are not a unique anomaly prediction
Garriga and Tanaka 2008, §§ II–IV, PDF	A proposed gauge-invariant local expansion observable does not show the claimed infrared screening in its controlled setup	A negative result for that observable and perturbative setting, not a theorem excluding every nonlocal or matter-driven effect

The strongest conservative position requires a renormalized gauge-invariant observable, causal response, and self-consistent constraint evolution before calling a secular term backreaction. The competing permissive position treats sufficiently large resummed infrared or anomaly contributions as evidence for a new branch even before a fully state-selected solution is available. The latter can motivate calculations, but branch, gauge, state, regulator, and secular-time axes must be held fixed before results are compared.

Adversarial checks and failure modes

First, change gauge or observable: a physical shift in the scalar expansion or curvature invariant should survive. Second, change the initial Hadamard state within the claimed basin: a universal attractor should not require finely tuned auxiliary data. Third, vary the subtraction scale and finite $R^2$ term while translating renormalized couplings; the same observable should agree. Fourth, compute the noise kernel with a physical smearing scale, propagate it through the retarded metric response, and compare the induced covariance with the resolved background and the claimed accuracy. A local stress fluctuation-to-mean ratio cannot replace that calculation.

Flat space supplies a null check: in a Poincaré-invariant vacuum, a purported secular FLRW effect must vanish as $H\to0$. A conformally coupled massless field provides another discriminator because state-dependent particle production in conformally flat geometry is absent for the conformal vacuum, although the trace anomaly remains. Failure of either check identifies which mechanism, rather than “quantum fields” generically, generated the claim.

The calculation fails when a response pole lies at the cutoff, the state stops being Hadamard, conservation residuals exceed the effect, gradients invalidate FLRW symmetry, or $\epsilon_{\rm loop}$ reaches unity before resummation has independent justification. A runaway of an unreduced fourth-order equation is not automatically physical.

A self-consistent workflow

Specify the field theory, state, initial hypersurface, renormalization scale and finite couplings. Evolve modes or correlators and verify canonical normalization. Construct $\rho$ and $p$ with the same subtraction, check $\dot\rho+3H(\rho+p)=0$, solve the Friedmann constraint and evolution together, and refine both time resolution and state initialization. Report $\epsilon_{\rm br}$ beside stress fluctuations, secular parameters, and the first omitted EFT term.

The conclusions could change after an independent state-selected nonlinear backreaction calculation changes a stability result, a theorem establishes or excludes a gauge-invariant secular mechanism, a corrected response kernel changes the spectrum, or new observations and a portable discriminator jointly identify a particular quantum source; any revised counterterm analysis or reproduced null result could also change the conclusion.

Exercise

In a flat FLRW model with $\Lambda=0$, a calculation finds $\delta\rho/\rho_0=0.04$, a smeared stress standard deviation divided by its mean equal to $0.6$, a loop parameter $0.08$, and combined numerical plus subtraction error $0.003$. Using the linearized constraint, find $\delta H/H_0$. Do the supplied data establish or exclude a controlled deterministic backreaction claim?

Solution

Because $3H_0^2=8\pi G\rho_0$, the constraint gives

$$\frac{\delta H}{H_0}=\frac12\frac{\delta\rho}{\rho_0}=0.02.$$

The mean shift exceeds the quoted error and both the shift and loop parameter are perturbative. The value $0.6$ does not by itself establish or exclude mean-field validity: it divides one smeared stress statistic by another, while the relevant test propagates the noise kernel through the retarded response to obtain the variance of a gauge-invariant metric or expansion observable. The deterministic claim is therefore not yet established, but it cannot be rejected from this ratio alone. One must report $\operatorname{Var}H[F]$ relative to the background scale and to the precision claimed for the $0.02$ shift.

References

• Anderson, P. R., Molina-París, C., and Mottola, E. (2003). “Linear Response, Validity of Semiclassical Gravity, and the Stability of Flat Space.” Physical Review D 67, 024026. DOI. Open PDF.
• Garriga, J., and Tanaka, T. (2008). “Can Infrared Gravitons Screen $\Lambda$?” Physical Review D 77, 024021. DOI. Open PDF.
• Mottola, E., and Vaulin, R. (2006). “Macroscopic Effects of the Quantum Trace Anomaly.” Physical Review D 74, 064004. DOI. Open PDF.
• Parker, L., and Simon, J. Z. (1993). “Einstein Equation with Quantum Corrections Reduced to Second Order.” Physical Review D 47, 1339–1355. DOI. Open PDF.