Entropy Bounds, Species, and Regulator Dependence

An entropy–energy bound is meaningful only after specifying which entropy, subsystem algebra, regulator, species content, and gravitational input it uses. Absolute subregion entropy in continuum QFT is ultraviolet divergent and grows with the number of fields. Relative entropy and mutual information evade particular divergences, but they answer different questions.

Required background. Relative-entropy Bekenstein bounds provide the finite state-comparison inequality tested here.

Helpful background. Entropy counterterms and renormalization explain the local surface terms that cancel only in controlled combinations.

Species scaling

For $N$ decoupled fields in a product state, leading regulated entanglement contributions add:

$$S_A^{(\epsilon)} \sim N\,c_{d-2}\frac{\operatorname{Area}(\partial A)}{\epsilon^{d-2}} +\cdots .$$

At fixed cutoff $\epsilon$, increasing $N$ can overwhelm a proposed bound involving only a fixed total excitation energy. This “species problem” shows that a cutoff entropy cannot be inserted into a regulator-independent inequality without additional physics. The lattice calculation of Srednicki 1993, Eqs. (5)–(12), pp. 667–669 is a canonical demonstration of the cutoff-dependent area term.

For a state $\rho$ compared with a reference $\sigma$ that shares the same ultraviolet structure, the leading local divergences cancel in

$$S(\rho_A\Vert\sigma_A) =\Delta\langle K_\sigma\rangle-\Delta S_A.$$

The cancellation is species-by-species. The finite result still depends on the state pair, region, and algebra; it does not turn absolute $S_A$ into an observable. Casini 2008, §§ 2–3 gives the relative-entropy formulation appropriate to the nongravitational bound used in this chapter.

Matched one- and many-species test

Regulate $N$ free scalar fields on the same lattice and prepare excitations with total energy $\Delta E$ distributed among species. Compare:

1. the raw entropy $S_A^{(\epsilon)}$;
2. the vacuum-subtracted entropy $\Delta S_A$;
3. the relative entropy $S(\rho_A\Vert\sigma_A)$;
4. the modular energy $\Delta\langle K_\sigma\rangle$.

Increase $N$ at fixed lattice, then repeat at smaller $\epsilon$ while keeping physical region size and total energy fixed. Raw entropy grows with $N$ and the area cutoff. Relative entropy remains nonnegative, while the distribution of energy among species changes its finite value. Reversing the $N\to\infty$ and $\epsilon\to0$ limits can produce a different sequence and must be reported.

The finite regional bound belongs to the relative-entropy branch. Absolute regulated entropy, gravitational area terms, and species-dependent cutoff contributions are additional inputs, not consequences of positivity alone. The diagram is schematic.

Regulator and gravitational input

A lattice tensor factorization, a continuum local algebra, and a gauge-theory extended Hilbert space define different subsystem data. A change among them can move surface contributions without changing long-distance correlators. If Newton’s constant is invoked to absorb species-dependent area divergences, the claim has acquired gravitational renormalization input and lies outside a purely nongravitational entropy–energy theorem.

Changing species number, cutoff, factorization convention, or limit order can change an absolute entropy while leaving the long-distance state similar. A valid bound declares which of these data are held fixed. The map is schematic.

Common pitfalls

Comparing unmatched regulators. Subtract states on the same lattice or within the same algebra. Two separately fitted area laws do not define a relative entropy.

Calling every finite entropy universal. Finite parts may depend on shape, scheme, and edge-mode convention. Universality requires an identified invariant combination.

References

• Casini, Horacio. “Relative Entropy and the Bekenstein Bound.” Classical and Quantum Gravity 25 (2008): 205021. DOI.
• Srednicki, Mark. “Entropy and Area.” Physical Review Letters 71 (1993): 666–669. DOI.