The Bekenstein Bound and Generalized Entropy

Generalized entropy combines a horizon’s geometric entropy with the von Neumann entropy accessible outside it. A quasistatic lowering experiment shows why the two terms belong together: the redshift can make the delivered Killing energy small, but the body’s finite size limits how close it can be released. Neglecting the atmosphere and buoyancy, the leading increase of area entropy is at least approximately $2\pi ER$ for a weakly gravitating body of energy $E$ and radius $R$. This motivates an entropy bound; it is not a universal theorem that an independent bound on ordinary matter entropy is required for the generalized second law.

Required background. Quantum Information supplies density matrices, von Neumann entropy, and reduced exterior states.

Helpful background. Quantum Fields in Curved Spacetime supplies the horizon state and renormalized exterior entropy; Einstein Equations supplies the backreacting geometry; Black-Hole Geometry and Mechanics supplies the first law; and Semiclassical Gravity supplies the mean-field regime in which the two contributions are combined.

Generalized entropy across a horizon cut

For a cut $C$ of a causal horizon, the leading semiclassical definition is

$$S_{\mathrm{gen}}[C] =\frac{A[C]}{4G_{\mathrm{ren}}} +S_{\mathrm{out,ren}}[C] +S_{\mathrm{local,grav}}[C].$$

The last term denotes higher-curvature geometric contributions when the renormalized gravitational action requires them. Neither the area term nor the exterior entropy is generally a separately cutoff-independent observable; the renormalized sum is the object entering the generalized second law.

In the elementary lowering argument, higher-curvature corrections are neglected, the black hole is much larger than the body, and backreaction is treated to first order. If a body of entropy $S_{\mathrm m}$, initially uncorrelated with the remaining exterior, crosses the horizon, the accessible exterior entropy decreases approximately by $S_{\mathrm m}$. The generalized second law requires a compensating geometric increase.

Redshifted lowering calculation

Let $\chi^\mu$ be the horizon Killing generator normalized by the asymptotic or cavity clock. Close to a nonextremal horizon, its lapse at proper distance $\ell$ is

$$N=\sqrt{-\chi^2}\simeq\kappa\ell.$$

A slowly held body with local energy $E$ therefore delivers Killing energy $\delta M\simeq EN\simeq E\kappa\ell$ when released. Its center cannot be lowered closer than roughly its radius $R$ without the body already crossing the horizon, so $\ell\gtrsim R$. The first law and Hawking normalization give

$$\delta S_{\mathrm{BH}} =\frac{\delta M}{T_{\mathrm H}} \simeq\frac{E\kappa\ell}{\kappa/(2\pi)} =2\pi E\ell \gtrsim2\pi ER.$$

The surface gravity cancels. This is the crucial normalization check: rescaling the Killing field rescales both delivered energy and temperature, leaving the entropy change invariant. Combining the geometric gain and exterior loss,

$$\Delta S_{\mathrm{gen}} \gtrsim2\pi ER-S_{\mathrm m}.$$

Thus the sufficient condition

$$S_{\mathrm m}\le 2\pi ER$$

protects the generalized second law in this idealized experiment. Bekenstein proposed the bound for complete bounded systems Bekenstein 1981, §I, pp. 287–289. Wald reviews both the lowering argument and its entropy-bound role Wald 2001, §§4.1–4.2, PDF.

Assumptions behind the bound

The derivation assumes an approximately isolated, weakly self-gravitating body, a well-defined positive total energy, size small compared with the curvature radius, quasistatic lowering, and a nonextremal stationary horizon. The body’s stress tensor may polarize the thermal atmosphere; buoyancy and finite-size effects modify the optimal drop point. The compact inequality is therefore not a universal local entropy-density condition.

Several extrapolations fail. A strongly gravitating object cannot be assigned an independent flat-space radius and energy without including its geometry. A negative-energy quantum excitation need not obey the same one-body reasoning. An arbitrary region in an interacting QFT has ultraviolet-divergent von Neumann entropy until a regulator and subtraction are specified. Finally, a cosmological horizon lacks the same asymptotic energy normalization. Each case needs its own generalized-entropy statement rather than a direct substitution into $S\le2\pi ER$.

The thermal atmosphere also explains why an apparent violation from lowering the box arbitrarily close is spurious. The force needed to hold the box and the displaced atmosphere contribute to the conserved Killing energy and entropy balance. Treating the box but not the state around it is not a self-consistent semiclassical comparison. Indeed, Unruh and Wald show how buoyancy protects the GSL in their box-lowering analysis without imposing the proposed matter bound as a separate axiom Unruh and Wald 1982, p. 942.

A precise quantum-field-theory version

For a region with vacuum reference state $\sigma$, let $K_{\mathrm{mod}}=-\log\sigma$ in a regulated description. Positivity of relative entropy gives

$$D(\rho\Vert\sigma) =\Delta\langle K_{\mathrm{mod}}\rangle-\Delta S\geq0, \qquad \Delta S\leq\Delta\langle K_{\mathrm{mod}}\rangle.$$

Here both differences compare the same region in $\rho$ and $\sigma$; $\Delta S$ is not its divergent, unsubtracted entropy. In the Minkowski half-space $x\gt0$ at $t=0$,

$$\Delta\langle K_{\mathrm{mod}}\rangle =2\pi\int_{x\gt0}d^3x\,x\,\Delta\langle T_{00}\rangle.$$

A localized positive-energy packet centered a distance $R$ from the cut has weighted energy approximately $2\pi ER$. This explains the scale while replacing an ambiguous size and absolute entropy with specified operators and subtractions. The exact inequality concerns modular energy, not $2\pi ER$ for every region and state Casini 2008, §§3–4, Eqs. (7), (16)–(18), PDF.

Interpretation and checks

The bound is dimensionally consistent in the inherited units: $ER$ is dimensionless because energy has inverse-length dimension. It is independent of $G$ even though its derivation uses the gravitational first law; $G$ cancels between area response and black-hole entropy. A body saturating the bound gives zero leading generalized-entropy change, while any smaller entropy produces a positive margin.

This argument motivates generalized entropy but does not prove the generalized second law for arbitrary rapid processes. Modern proofs use relative-entropy monotonicity and precise horizon-algebra hypotheses. Nor does the calculation answer whether Hawking radiation recovers fine-grained information: that question concerns a different region, algebra, and time dependence.

Exercise

A box has local energy $E$, entropy $S_{\mathrm m}=1.5\pi ER$, and radius $R$. It is released with its center at proper distance $\ell=R$. Compute the leading change in generalized entropy and explain which approximations make the answer an estimate within the idealized no-buoyancy model.

Solution

The area term increases by $\delta S_{\mathrm{BH}}\simeq2\pi ER$, while the exterior loses $S_{\mathrm m}=1.5\pi ER$. Hence $\Delta S_{\mathrm{gen}}\simeq0.5\pi ER\gt0$. It is a leading estimate because the body is treated as weakly gravitating and small, the horizon is stationary and nonextremal, the lowering is quasistatic, and buoyancy, higher-curvature entropy, finite-size structure, and higher-order backreaction have been neglected. Those omissions have not been bounded here, so the number is not a rigorous lower bound for the full physical lowering experiment.

References

• Bekenstein, J. D. (1981), “Universal Upper Bound on the Entropy-to-Energy Ratio for Bounded Systems,” Physical Review D 23, 287–298. DOI.
• Casini, H. (2008), “Relative Entropy and the Bekenstein Bound,” Classical and Quantum Gravity 25, 205021. DOI. Open PDF.
• Unruh, W. G., and R. M. Wald (1982), “Acceleration Radiation and the Generalized Second Law of Thermodynamics,” Physical Review D 25, 942. DOI.
• Wald, R. M. (2001), “The Thermodynamics of Black Holes,” Living Reviews in Relativity 4, 6. Open PDF.