Microcanonical Black-Hole State Spaces

A microcanonical black-hole state space is not defined by energy alone. One must specify the complete quantum theory, asymptotic boundary conditions, superselection sector, conserved charges, angular momentum labels, and finite energy windows, then choose which exact states are included. Only the logarithm of that regulated dimension can be compared with a semiclassical entropy. This definition prevents a density of states, an exact degeneracy, and a thermodynamic entropy from being silently identified.

Required background. Quantum Information supplies density operators, projectors, and entropy; Gauge Theory and String-Theory Preparation supplies constrained Hilbert spaces, charge sectors, and microscopic state-counting language.

Helpful background. Semiclassical Gravity supplies the Bekenstein–Hawking entropy and thermodynamic variables used for comparison; Quantum Fields in Curved Spacetime supplies the exterior algebra and boundary conditions held fixed.

The constrained energy shell

Let $H$ be the exact Hamiltonian in a regulated theory with fixed asymptotic boundary data. Let $Q_a$ denote commuting conserved charges whose sector is held fixed. For a centered energy window of width $\Delta E$, define

$$P_{E,\Delta E,\mathbf q} =\mathbf 1_{[E-\Delta E/2,E+\Delta E/2]}(H) \prod_a P_{Q_a=q_a}.$$

The microcanonical dimension and normalized state are

$$d(E,\Delta E,\mathbf q)=\operatorname{Tr}P_{E,\Delta E,\mathbf q}, \qquad \Omega_{E,\Delta E,\mathbf q} =\frac{P_{E,\Delta E,\mathbf q}}{d(E,\Delta E,\mathbf q)}.$$

The trace is over physical, constraint-satisfying states, not an unconstrained kinematical tensor product. Additional discrete quantum numbers, topology, or asymptotic sources must be included if the Hamiltonian does not mix them. A window spanning two superselection sectors defines a mixture of different questions rather than a single ensemble.

For a continuous approximation with density of states $\rho(E,\mathbf q)$, the count is the integral of $\rho$ over the shell. If that density is nearly constant across the window, it simplifies to

$$d(E,\Delta E,\mathbf q) \simeq \rho(E,\mathbf q)\Delta E, \qquad S_{\rm micro}=\log d.$$

Here $\rho$ has units of inverse energy, so the product is dimensionless. The window is part of the entropy convention. A sufficient local condition for this simplification is $\Delta E\,|\partial_E\log\rho|\ll1$; thermodynamically this is $\beta\Delta E\ll1$. A narrow relative energy window alone need not satisfy it, because a black-hole density of states grows very rapidly with energy.

The allowed hierarchy of scales

A useful shell contains many levels but changes no macroscopic control variable appreciably. If $\delta E$ is a typical exact level spacing and $E_{\rm macro}$ is the scale on which temperature or geometry changes, require

$$\delta E\ll\Delta E\ll E_{\rm macro}.$$

Locally, $\delta E\simeq1/\rho(E,\mathbf q)$; for $d=e^S$ approximately uniformly spaced levels in the declared shell, this is $\delta E\simeq\Delta E\,e^{-S}$. The lower inequality can therefore coexist with an extremely narrow relative window. Doubling $\Delta E$ changes the entropy by approximately $\log2$ only when the density remains nearly constant across both windows. Otherwise the density integral is required, even if both windows leave the geometry almost unchanged.

The upper inequality can fail near a phase transition or a sharp edge of the spectrum. It also requires care for asymptotically flat black holes with negative heat capacity, where a canonical ensemble is unstable unless a box, AdS boundary condition, or other stabilizing environment is specified. The microcanonical definition remains meaningful, but equivalence with a canonical state is then not automatic.

Comparing the count with black-hole entropy

For a stationary four-dimensional Einstein black hole, the leading semiclassical quantity is

$$S_{\rm BH}=\frac{A_H}{4G},$$

with $c=\hbar=k_{\rm B}=1$ and $G$ retained. The comparison

$$\log d(E,\Delta E,\mathbf q) =S_{\rm BH}+o(S_{\rm BH})$$

is a microscopic claim about a specified sector. It is stronger than obtaining the same canonical free energy at one saddle and weaker than identifying every individual state with a smooth geometry. Harlow discusses the finite-system entropy and black-hole state-space assumptions in Harlow 2016, § 4.6. The relation between fine-grained and coarse-grained black-hole entropy is organized in Almheiri et al. 2021, §§ 2–3.

Three limiting cases check the definition. A one-dimensional shell has $S_{\rm micro}=0$. An exactly $g$-fold degenerate eigenvalue gives $\log g$ even as $\Delta E$ shrinks around it. A nondegenerate dense spectrum gives a large entropy only because the declared finite window includes many nearby levels. These cases distinguish exact degeneracy from energy resolution.

What the definition does not provide

The maximally mixed state $\Omega$ is an ensemble description. A typical pure state in its support may reproduce restricted observables, but that requires a separate typicality theorem. Thermal behavior of individual eigenstates requires ETH or another dynamical statement. A common geometry requires an explicit coarse-graining map. None follows from the dimension $d$ alone.

This page does not redo generic quantum-information theory, stable black-hole thermodynamics, or explicit string and holographic state construction. Those topics remain separate. The result here is the precise list of constraints that defines the microcanonical ensemble whose logarithmic dimension is compared with black-hole entropy.

Exercises

A shell contains $d=10^{20}$ states. Doubling its energy width doubles the count without changing its charges or macroscopic geometry. By how much does the microcanonical entropy change, using natural logarithms, and what does that say about the leading entropy?

Solution

The original entropy is $S=\log d$. The doubled shell has $d'=2d$, so $S'-S=\log(2d)-\log d=\log2\simeq0.693$. This order-one shift is negligible relative to $\log(10^{20})\simeq46.05$ at leading order, provided the larger window satisfies $\Delta E\ll E_{\rm macro}$ and, as stipulated, doubles the count. The latter is an extra density-of-states assumption. The shift remains a real convention-dependent subleading term and must be kept when comparing precise counts.

References

• Almheiri, Ahmed, Thomas Hartman, Juan Maldacena, Edgar Shaghoulian, and Amirhossein Tajdini. 2021. “The Entropy of Hawking Radiation.” Reviews of Modern Physics 93, 035002. arXiv:2006.06872.
• Harlow, Daniel. 2016. “Jerusalem Lectures on Black Holes and Quantum Information.” Reviews of Modern Physics 88, 015002. arXiv:1409.1231.