Null-Averaged Energy Conditions

Null averaging can restore a nonnegative statement even when $\langle T_{ab}k^ak^b\rangle$ is negative on finite segments, but only under geometric and state hypotheses. “ANEC” is not one universal theorem: completeness, achronality, convergence prescription, field theory, transverse smearing, and whether the geometry satisfies the semiclassical equation all change what has actually been proved.

Required background. Congruences and Optical Scalars supplies affine null tangents and focusing, and Quantum Information supplies states and the later entropy-based arguments.

Helpful background. Quantum Fields in Curved Spacetime supplies Hadamard states, Einstein Equations supplies the null-contracted field equation, and Black-Hole Geometry and Mechanics supplies complete generators and horizon applications.

ANEC and its achronal restriction

Let $\gamma:\mathbb R\to M$ be a complete affinely parametrized null geodesic with tangent $k^a=dx^a/d\lambda$. When the integral is absolutely convergent, the averaged null energy condition is

$$\int_{-\infty}^{\infty}d\lambda\, \langle T_{ab}\rangle_{\mathrm{ren}}k^ak^b\ge0.$$

ANEC requires this on every null geodesic in the stated class. Achronal ANEC, or AANEC, requires it only when no two points of $\gamma$ can be connected by a timelike curve. Achronality removes winding and delayed null curves that need not behave like fastest causal paths. Kontou and Sanders give the definitions and several inequivalent limiting prescriptions for non-absolutely-convergent integrals in Kontou and Sanders 2020, § 4.1, Eqs. (83)–(85), pp. 28–29, PDF.

Affine normalization is part of the numerical value. Under $\lambda'=a\lambda+b$ with $a\gt 0$, $k'^a=k^a/a$, so the integral changes by $1/a$. Its sign is invariant, but comparisons of magnitudes require a normalization fixed by asymptotic time, horizon generators, or another declared convention.

A theorem with exact hypotheses

One rigorous curved-spacetime result is deliberately narrow.

Theorem. Let $(M,g)$ be a two-dimensional globally hyperbolic spacetime, let $\phi$ be the free, real, massless, minimally coupled Klein–Gordon field, let $\omega$ be any Hadamard state in the algebraic sense, and let $\gamma$ be a complete achronal null geodesic with affine parameter $\lambda$. Define $\widehat c(\nu)=\int dx\,e^{i\nu x}c(x)$. For every real $c\in C_0^1(\mathbb R)$ such that

$$(1+\nu^2)^{1+\delta}\lvert\widehat c(\nu)\rvert$$

is bounded for some $\delta\gt 0$, and for every choice of affine origin,

$$\liminf_{\lambda_0\to\infty} \int_{-\infty}^{\infty}d\lambda\, \langle T_{ab}\rangle_{\omega,\mathrm{ren}}k^ak^b \,c(\lambda/\lambda_0)^2 \ge0.$$

If the unsmeared integral converges absolutely, choose $c(0)=1$; dominated convergence then reduces this statement to ordinary AANEC. A profile with $c(0)=0$ would instead give a trivial zero limit and could not establish its sign. The theorem does not state a four-dimensional result, does not cover incomplete or chronal null curves, and does not assume absolute convergence. Wald and Yurtsever define the limiting prescription in Wald and Yurtsever 1991, § II, Eq. (3), printed pp. 404–406 and prove the curved two-dimensional result in Wald and Yurtsever 1991, § V, printed pp. 411–412; its scope is summarized in Kontou and Sanders 2020, § 4.3, pp. 35–36, PDF.

The proof architecture is:

1. choose two-dimensional null coordinates along $\gamma$ and subtract a locally constructed Hadamard bidistribution, so the relevant null stress is the coincidence limit of derivatives of the smooth remainder;
2. apply positivity of the algebraic state to two global Klein–Gordon solutions, obtaining a quadratic inequality for the real two-point form and its fixed commutator part;
3. choose compactly supported oscillatory data proportional to $c(\lambda/\lambda_0)\cos(\kappa\lambda)$ and $c(\lambda/\lambda_0)\sin(\kappa\lambda)$ on the generator;
4. use global hyperbolicity and achronality to extend every such compactly supported datum on $\gamma$ to a global smooth solution, which is the precise point at which the curve hypothesis enters;
5. integrate the resulting spectral inequality over $\kappa$. The stated Fourier decay of $c$ controls the error terms, and the $\lambda_0\to\infty$ lower limit gives zero as the state-independent lower bound.

Each step uses a stated hypothesis. Four dimensions add transverse ultraviolet structure. On a chronal curve, arbitrary compact data need not extend to global solutions, so the positivity argument no longer tests the required functions. An incomplete curve is not the theorem’s integration domain, and a different limiting prescription is a different averaged condition.

Other proved routes and their domains

The comparison axes are curve class, dimension and background, matter theory and state, averaging prescription, and whether the geometry is fixed or self-consistent.

In a four-dimensional flat tubular region, one can obtain null averages from a timelike quantum null-energy inequality. Fewster, Olum, and Pfenning assume strong causality in the spacetime interior and require the complete null geodesic to lie in a simply connected flat open neighborhood $N'$ that misses the boundary, is causally isolated in the sense that $J^+(p,N)\cap N'=J^+(p,N')$ for every $p\in N'$, and contains a tube of fixed positive radius along the entire geodesic. Causal isolation here implies that the generator is achronal. The quantum theory must reduce on suitable globally hyperbolic open subsets to the usual minimally coupled scalar theory of mass $m\ge0$, and the state must be Hadamard on those subsets.

Under those hypotheses, highly boosted timelike segments rule out an ANEC integral that converges uniformly to negative values on a timelike family. Stationarity plus convergence gives ordinary ANEC as a corollary; alternatively, uniform continuity of $\langle T_{kk}\rangle$ gives the long-smearing lower-limit statement. These distinct conclusions are Fewster, Olum, and Pfenning 2007, § II A–C, Eqs. (2)–(3), Theorem II.1 and Corollary II.2, manuscript pp. 4–8; § III, Theorem III.1, manuscript p. 9, PDF.

Modern Minkowski-space arguments extend ANEC to broad interacting-QFT classes using causality, modular Hamiltonians, or conformal methods, but their assumptions are algebraic or field-theoretic rather than statements about every curved semiclassical spacetime. Conversely, curved-space test fields can violate AANEC after conformal transformations.

Backreaction changes the question again. Flanagan and Wald perturb a self-consistent semiclassical system about Minkowski vacuum, assume long wavelengths compared with the Planck length and restrict incoming gravitational radiation, and obtain nonnegativity after averaging along a complete achronal generator and transversely over several Planck areas. That result rules out a class of macroscopic violations; it is not an unsmeared universal AANEC theorem. The hypotheses and conclusion are reviewed in Kontou and Sanders 2020, § 4.3, pp. 35–36, PDF.

Statement	Geometry and curve	Matter/state	Status
Two-dimensional AANEC	Globally hyperbolic; complete achronal null geodesic	Free massless scalar; every Hadamard state	Theorem with a specified long-smearing limit
Flat-tube null averages from timelike QNEI	Simply connected flat, causally isolated fixed-radius tube around a complete null geodesic	Minimally coupled scalar; locally Hadamard state; convergence, stationarity, or uniform continuity as required by the chosen conclusion	No uniform ANEC violation; ordinary or long-smearing ANEC under the additional stated regularity
Perturbative self-consistent AANEC	Near Minkowski; complete achronal generator; transverse averaging	Long-wavelength scalar perturbations; restricted incoming radiation	Controlled second-order semiclassical result
General self-consistent AANEC	Generic curved solution; complete achronal generator	General renormalized QFT state	Conjectural without a universal proof

Counterexamples and discriminating hypotheses

The vacuum-plus-two-particle state from the previous page is Hadamard and can make $T_{ab}k^ak^b$ negative on an open segment. In Minkowski space, the applicable complete-geodesic ANEC nevertheless requires compensating positive contribution elsewhere. Local NEC violation therefore does not contradict an averaged theorem.

Drop achronality and explicit counterexamples appear. In a flat spacetime with a compact spatial direction, a null geodesic that winds around the compact direction has pairs of points connected timelike; the Casimir-type ground-state stress can give a negative null average along such a chronal curve. Kontou and Sanders review these compactification counterexamples in Kontou and Sanders 2020, § 4.3, p. 35, PDF. They do not refute AANEC because the curve fails its defining hypothesis.

Drop completeness and a segment can isolate only the negative lobe of an interference pattern. Drop self-consistency and conformally transformed test-field states can violate curved-space AANEC while their stress is not allowed to backreact on the geometry. These examples discriminate the formulations rather than delivering a single verdict about “quantum matter.”

Consequences for focusing

For a hypersurface-orthogonal null congruence in four dimensions,

$$\frac{d\theta}{d\lambda} =-\frac12\theta^2-\sigma_{ab}\sigma^{ab}-R_{ab}k^ak^b.$$

For the four-derivative renormalized equation used in this volume, null contraction instead gives

$$R_{ab}k^ak^b +16\pi G\left( \alpha H^{(R^2)}_{ab}k^ak^b +\beta H^{(C^2)}_{ab}k^ak^b \right) =8\pi G\langle T_{ab}\rangle_{\mathrm{ren}}k^ak^b.$$

Only in the Einstein truncation, or after a controlled order reduction that makes the displayed local curvature terms negligible at the retained order, does this reduce to $R_{kk}=8\pi G\langle T_{kk}\rangle$. Moving the curvature terms to an “effective stress tensor” does not make a matter-field ANEC theorem apply to them. In that qualified Einstein regime, a pointwise NEC makes the curvature term in Raychaudhuri nonpositive. ANEC is weaker: an integrated nonnegative matter stress does not prevent temporary defocusing, and focusing theorems require additional genericity, completeness, convergence, and boundary hypotheses.

Thus one must cite the actual averaged focusing theorem, not insert ANEC into a pointwise proof line by line. QNEC and quantum focusing, developed next, replace the matter-only average with local entropy or generalized-entropy variations.

Exercise

Let

$$I[\gamma,\lambda] =\int_{-\infty}^{\infty}d\lambda\, T_{ab}k^ak^b.$$

Show how $I$ transforms under $\lambda'=a\lambda+b$ with $a\gt 0$. Then explain why a negative integral over a finite subsegment neither violates ANEC nor supplies a counterexample to AANEC.

Solution

Because $d\lambda'=a\,d\lambda$ and $k'^a=dx^a/d\lambda'=k^a/a$,

$$I[\gamma,\lambda'] =\int a\,d\lambda\, T_{ab}\frac{k^a}{a}\frac{k^b}{a} =\frac{1}{a}I[\gamma,\lambda].$$

The magnitude depends on affine normalization, but multiplication by the positive number $1/a$ preserves the sign. ANEC concerns the entire complete generator, not a selected interval, so positive energy outside a negative segment can make the full integral nonnegative. AANEC adds the requirement that the complete generator be achronal. A finite segment is not the theorem’s integration domain, and a complete but chronal winding geodesic is not in the AANEC curve class.

References

• Fewster, C. J., Olum, K. D., and Pfenning, M. J. 2007. “Averaged Null Energy Condition in Spacetimes with Boundaries.” Physical Review D 75, 025007. DOI. Open PDF.
• Kontou, E.-A., and Sanders, K. 2020. “Energy Conditions in General Relativity and Quantum Field Theory.” Classical and Quantum Gravity 37, 193001. DOI. Open PDF.
• Wald, R. M., and Yurtsever, U. 1991. “General Proof of the Averaged Null Energy Condition for a Massless Scalar Field in Two-Dimensional Curved Spacetime.” Physical Review D 44, 403–416. DOI.