Worldline Quantum Energy Inequalities

Pointwise quantum energy density is unbounded below as the state varies, but a smooth timelike average obeys a state-independent lower bound in many free-field theories. In four-dimensional Minkowski space, shortening a sampling profile by a factor $\tau$ weakens the massless scalar bound as $\tau^{-4}$: deeper negative energy is permitted only when it is resolved for a shorter time.

Required background. Congruences and Optical Scalars supplies proper-time observers and their measured energy density, and Quantum Information supplies states and expectation values.

Helpful background. Quantum Fields in Curved Spacetime supplies Hadamard two-point functions, Einstein Equations supplies later focusing applications, and Black-Hole Geometry and Mechanics supplies the relevant timelike and horizon scales.

The four-dimensional inertial worldline bound

Theorem. Let $\phi$ be a free, real, minimally coupled Klein–Gordon field of mass $m\ge0$ in four-dimensional Minkowski spacetime. Let $\gamma(t)=(t,\boldsymbol x_0)$ be an inertial worldline parametrized by proper time, let $\omega$ be any Hadamard state, and normal order $T_{00}$ relative to the Minkowski vacuum. For every real $g\in C_0^\infty(\mathbb R)$,

$$\int_{-\infty}^{\infty}dt\, g(t)^2\, \langle{:}T_{00}{:}\rangle_\omega(\gamma(t)) \ge -\frac{1}{16\pi^2} \int_{-\infty}^{\infty}dt\,\lvert g''(t)\rvert^2.$$

The same inequality holds for $g$ in the Sobolev class $W^{2,2}(\mathbb R)$. It is sharp enough to expose the sampling law but is not generally the optimal lower bound. The exact statement and its massive spectral refinement are recorded in Fewster 2012, § 1.3, Eqs. (1)–(3), pp. 10–11, PDF; the original flat-space derivation is Fewster and Eveson 1998, Eqs. (1.7)–(1.8) and § III, PDF.

The hypotheses matter. The curve is timelike and inertial, the field is free and minimally coupled, the spacetime is flat and four-dimensional, the state is Hadamard, and the sampling weight is a square $g^2$. Normal ordering fixes the reference subtraction. Curved spacetime, acceleration, nonminimal coupling, boundaries, and interacting fields require different bounds and often additional state-independent geometric terms.

Why the bound scales as the inverse fourth power

Choose a fixed profile with $\int g^2dt=1$ and set

$$g_\tau(t)=\tau^{-1/2}g(t/\tau).$$

This preserves the normalization. Two derivatives give $g_\tau''(t)=\tau^{-5/2}g''(t/\tau)$, hence

$$\int dt\,\lvert g_\tau''(t)\rvert^2 =\tau^{-4}\int ds\,\lvert g''(s)\rvert^2.$$

Therefore

$$\int dt\,g_\tau(t)^2\langle{:}T_{00}{:}\rangle_\omega \ge -\frac{C_g}{\tau^4}, \qquad C_g=\frac{1}{16\pi^2}\int ds\,\lvert g''(s)\rvert^2.$$

Energy density has dimension $L^{-4}$, so the scaling is also forced dimensionally. As $\tau\to0$, the lower bound tends to $-\infty$, consistently recovering the absence of a pointwise lower bound. To recover a complete-worldline statement, multiply the normalized inequality by $\tau$ before taking the long-time limit:

$$\int dt\,g(t/\tau)^2\langle{:}T_{00}{:}\rangle_\omega \ge-\frac{C_g}{\tau^3}.$$

Its lower limit is nonnegative. If the energy density is absolutely integrable and $g(0)\ne0$, dominated convergence gives $\int dt\,\langle{:}T_{00}{:}\rangle_\omega\ge0$. Merely taking the limit of the normalized average would not prove this: a finite negative integral divided by $\tau$ also approaches zero.

The scaling also gives a duration–magnitude tradeoff. Suppose the measured density never exceeds $-\rho_0\lt 0$ throughout an interval of duration $\tau$. Choose a fixed normalized profile supported inside that interval. Its average is at most $-\rho_0$, while the QEI is at least $-C_g/\tau^4$, so

$$\rho_0\tau^4\le C_g.$$

The constant depends on the shape of the sampling profile, and optimizing it is a fourth-order eigenvalue problem; the inverse-fourth-power law does not. A nonzero mass strengthens the exact spectral bound when $\tau$ is long compared with $m^{-1}$, because modes below the mass threshold do not contribute. The mass-independent derivative estimate remains valid but then ceases to be sharp.

Proof architecture

The proof is not an application of positivity to the renormalized density itself. It uses positivity before the vacuum subtraction.

1. Point split. For the minimally coupled field, express the classical energy density as a finite sum of squares of differentiated fields. Pull the corresponding two-point functions back to $\gamma\times\gamma$.

2. Subtract a Hadamard reference. The difference between the state two-point function and the vacuum two-point function is smooth. Its diagonal is the normal-ordered density, so point splitting defines the left-hand side.

3. Unsplitting by Fourier modes. Insert the Fourier representation of the diagonal delta distribution. For a real sampling function, symmetry reduces the frequency integration to a positive half-line.

4. Use positive type. The unsubtracted state two-point function is of positive type. Dropping its nonnegative contribution leaves minus an integral involving only the reference two-point function.

5. Control the pullback and ultraviolet tail. The Hadamard wavefront set does not meet the conormal bundle of a timelike curve in the forbidden way, so the pullback exists. Microlocal decay makes the remaining positive-frequency integral finite.

6. Evaluate the vacuum spectrum. In Minkowski space the reference two-point function has an on-shell Fourier representation. With

   $$\widehat g(u)=\int_{-\infty}^{\infty}dt\,e^{iut}g(t),$$

   summing the differentiated squares yields the massive bound

   $$-\frac{1}{16\pi^3} \int_m^\infty du\, \lvert\widehat g(u)\rvert^2u^4Q_3(u/m),$$

   where, for $x\ge1$,

   $$Q_3(x)=\sqrt{1-x^{-2}}\left(1-\frac{1}{2x^2}\right) -\frac{\ln(x+\sqrt{x^2-1})}{2x^4}.$$

   This function satisfies $0\le Q_3\le1$ and tends to one as $x\to\infty$. For $m=0$, use that limit. Since $g$ is real, Parseval gives $\int_0^\infty du\,u^4|\widehat g(u)|^2=\pi\int dt\,|g''(t)|^2$, recovering the displayed coefficient $1/(16\pi^2)$.

Fewster states the general positive-type theorem, including the timelike curve, Hadamard reference, and smooth sampling hypotheses, as Fewster 2012, Theorem 2.4, pp. 24–26, PDF, and performs the four-dimensional inertial calculation in Fewster 2012, § 2.5, pp. 26–27, PDF.

What fails when a hypothesis is removed

Sharp switching is not covered: a discontinuous profile has a distributional second derivative whose squared norm is undefined, and increasingly sharp smooth approximations make the lower bound increasingly negative. This is a limitation of the proposed measurement, not a loophole in the theorem.

Nor may one replace the timelike curve by a null worldline in four dimensions. The required pullback fails its wavefront-set test, and state-independent null-worldline QEIs of this form do not exist. Fewster highlights this counterexample immediately after Fewster 2012, Theorem 2.4, p. 25, PDF. Spacetime-volume smearings and achronal complete-null averages are different statements with different hypotheses.

Finally, the inequality bounds matter energy on a fixed background. Applying it to a self-consistent geometry requires the sampling time to be short compared with curvature variation but long compared with the matter or material cutoff. If no such window exists, the flat-space bound does not justify a semiclassical focusing conclusion.

Exercise

Take the normalized Gaussian

$$g_\tau(t) =\pi^{-1/4}\tau^{-1/2} \exp\!\left(-\frac{t^2}{2\tau^2}\right),$$

which lies in $W^{2,2}(\mathbb R)$. Evaluate $\int\lvert g_\tau''\rvert^2dt$ and state the resulting QEI.

Solution

Differentiation gives

$$g_\tau''(t) =\left(\frac{t^2}{\tau^4}-\frac{1}{\tau^2}\right)g_\tau(t).$$

The probability density $g_\tau^2$ is a centered Gaussian with moments $\langle t^2\rangle=\tau^2/2$ and $\langle t^4\rangle=3\tau^4/4$. Therefore

$$\int dt\,\lvert g_\tau''(t)\rvert^2 =\frac{\langle t^4\rangle}{\tau^8} -\frac{2\langle t^2\rangle}{\tau^6} +\frac{1}{\tau^4} =\frac{3}{4\tau^4}.$$

Substitution into the theorem yields

$$\int dt\,g_\tau(t)^2 \langle{:}T_{00}{:}\rangle_\omega \ge-\frac{3}{64\pi^2\tau^4}.$$

This explicit profile checks the sign, dimension, and $\tau^{-4}$ scaling. Narrowing the detector permits a more negative lower bound, but it never produces a finite pointwise bound in the limit $\tau\to0$.

References

• Fewster, C. J. 2012. “Lectures on Quantum Energy Inequalities.” arXiv:1208.5399. Open PDF.
• Fewster, C. J., and Eveson, S. P. 1998. “Bounds on Negative Energy Densities in Flat Spacetime.” Physical Review D 58, 084010. DOI. Open PDF.