Quantum Inequalities from the Second Law

Abstract: Quantum field theory allows local violations of the classical energy conditions, such as regions of negative energy density. The Ford-Roman quantum inequalities quantify how negative these regions can be and for how long. We present a derivation of these inequalities from the second law of thermodynamics, in an operational framework that couples the field to a small thermal detector. The detector acts as a physical probe of the field's energy density and, by the non-negativity of the entanglement entropy between the two systems, one can show that its heat loss is bounded from below by a state-independent quantity.

• The paper derives a state-independent lower bound on the heat lost by a thermal Unruh–DeWitt detector by combining entropy balance, mutual-information positivity, and perturbation theory.
• The bound depends only on detector frequency, temperature, coupling, and switching profile because the free field’s c-number commutator makes the relevant term independent of the field state.
• The result constrains an operational, thermometer-weighted two-point-function average rather than the canonical smeared energy density, leaving the recovery of Ford–Roman scaling and extensions to interacting or curved-spacetime fields open.

Motivation and context

Classical stress tensors obey pointwise positivity conditions, but quantum field theory permits locally negative renormalized energy densities, as exhibited by the Casimir effect, squeezed states, and moving mirrors. The Ford–Roman quantum inequalities quantify these violations: the time-averaged energy density along a worldline is bounded below by a state-independent negative quantity scaling as the inverse fourth power of the sampling time (2608.11817). Standard derivations rely on microlocal spectrum conditions or positivity of quadratic forms built from Wightman functions. Ford himself originally motivated such bounds thermodynamically—arbitrarily large negative energy fluxes would enable second-law violations—but a rigorous derivation of the inequalities from the second law has remained largely unexplored. This paper supplies one, in an operational framework where the "measurement" of negative energy is performed by a physical thermometer coupled to the field.

Setup

The system is a free massless scalar field $\phi$ in four-dimensional Minkowski spacetime, probed by a harmonic oscillator detector $A$ of frequency $\Omega$, initially thermal at temperature $T = 1/\beta$. The interaction is Unruh–DeWitt type,

$$H_I(t) = \lambda\, g(t)\, \phi(t,\mathbf{0})\, X_A,$$

with smooth switching function $g(t)$ of width $\tau$ and small coupling $\lambda$. The joint state evolves unitarily from a product initial state $\rho_\psi \otimes \rho_A^{(\beta)}$; the total von Neumann entropy is exactly conserved, while the reduced entropies change due to entanglement generation. The mean heat exchanged by the detector, $\langle Q\rangle = \Tr[H_A(\rho_A(t_f) - \rho_A)]$, serves as the operational proxy for the sampled energy density.

Entropy balance

Expanding the von Neumann entropy around the thermal state via the Fréchet derivative of the logarithm, the first-order term vanishes by trace preservation, and the second-order term is non-positive because the thermal state maximizes entropy at fixed Hamiltonian (a restatement of the Clausius inequality). Since $A$0, this yields

$A$1

Because $A$2 for physically relevant states, both $A$3 and $A$4 are $A$5, so to leading order $A$6.

For the field, assumed initially pure, the entropy is non-analytic at zero eigenvalues: writing the perturbed reduced state in block form relative to the projector onto $A$7, the leading contribution behaves as $A$8 with $A$9. The coefficient $\Omega$0 evaluates to a double integral involving the real part of the thermal two-point function $\Omega$1 times the field Wightman function $\Omega$2 smeared with $\Omega$3.

The mutual information bound

Since total entropy is conserved and the initial state is uncorrelated, $\Omega$4: the coarse-grained second law states that subsystem entropies increase by exactly the generated mutual information. Combining the two entropy changes gives, to leading order,

$\Omega$5

Evaluating the pieces explicitly—the detector susceptibility is $\Omega$6, so $\Omega$7, and the antisymmetric part of $\Omega$8 drops out of $\Omega$9—the heat reduces to $T = 1/\beta$0, where $T = 1/\beta$1 involves only the symmetric two-point function $T = 1/\beta$2. For $T = 1/\beta$3, the thermal averages give $T = 1/\beta$4 (consistent with the fluctuation-dissipation theorem), yielding $T = 1/\beta$5, where $T = 1/\beta$6 involves the commutator part $T = 1/\beta$7.

The crucial observation is that for a free field the commutator $T = 1/\beta$8 is a c-number (the Pauli–Jordan function), so $T = 1/\beta$9 is state-independent; all state dependence resides in $$H_I(t) = \lambda\, g(t)\, \phi(t,\mathbf{0})\, X_A,$$0. With $$H_I(t) = \lambda\, g(t)\, \phi(t,\mathbf{0})\, X_A,$$1 so that $$H_I(t) = \lambda\, g(t)\, \phi(t,\mathbf{0})\, X_A,$$2, the inequality becomes

$$H_I(t) = \lambda\, g(t)\, \phi(t,\mathbf{0})\, X_A,$$3

or equivalently a universal lower bound on the detector's heat:

$$H_I(t) = \lambda\, g(t)\, \phi(t,\mathbf{0})\, X_A,$$4

This bound depends only on detector parameters—frequency, temperature, coupling, switching profile—and holds for every field state. Physically, the thermometer cannot be cooled below a threshold set by its own characteristics, which translates into a restriction on how negative the field's energy density can appear to any physical probe.

Relation to the smeared energy density

An appendix establishes that both $$H_I(t) = \lambda\, g(t)\, \phi(t,\mathbf{0})\, X_A,$$5 and the conventionally smeared energy density $$H_I(t) = \lambda\, g(t)\, \phi(t,\mathbf{0})\, X_A,$$6 are linear functionals of the same symmetric two-point function $$H_I(t) = \lambda\, g(t)\, \phi(t,\mathbf{0})\, X_A,$$7, differing only in their kernels: point-splitting plus integration by parts gives $$H_I(t) = \lambda\, g(t)\, \phi(t,\mathbf{0})\, X_A,$$8, whereas $$H_I(t) = \lambda\, g(t)\, \phi(t,\mathbf{0})\, X_A,$$9 with $g(t)$0. The derived bound therefore constrains a nonlocal, operationally defined probe of the energy density rather than the standard Ford–Roman functional itself. This is the sense in which the result is "reminiscent of" rather than identical to the classical quantum inequalities: it bounds a thermometer-weighted average of $g(t)$1, not $g(t)$2 directly.

Limitations and open questions

Several restrictions should be noted plainly. The derivation is perturbative in $g(t)$3 and assumes $g(t)$4; the leading-order entropy balance neglects $g(t)$5 corrections. The field must be free and massless in flat spacetime, since the state-independence of $g(t)$6 hinges on the commutator being a c-number—curved backgrounds, interactions, or higher-spin fields would require separate analysis. The initial field state is taken pure; mixed states would alter the field-entropy expansion, which is invalid around rank-deficient density matrices. Most significantly, the bound applies to the detector's heat functional rather than to the canonical smeared $g(t)$7, so recovering the precise Ford–Roman scaling (e.g., the inverse fourth power of the sampling time) from this thermodynamic route remains open, as does the question of whether the bound tightens to match known quantum inequality constants in specific limits.

Conclusion

The paper demonstrates that the non-negativity of entanglement entropy, applied to a small thermal detector coupled to a quantum field, enforces a second-law constraint on coarse-grained entropies that translates into a state-independent lower bound on the heat the detector can lose. Because this heat probes the symmetric two-point function underlying the renormalized energy density, the result provides an operational, thermodynamic derivation of quantum-inequality-type bounds: negative energy cannot be made arbitrarily large or long-lived as measured by any physical thermometer, with the strength of the restriction fixed solely by the probe's frequency, temperature, and switching profile.