Unitarity and Entropy Requirements

A recovery proposal passes the unitarity test only after it identifies a complete quantum system and a single evolution law on that system. A decreasing radiation entropy is important evidence, but it is not by itself a unitary S-matrix, a density operator, or a recovery channel. This page turns the broad question into checks of Hilbert-space completeness, linearity, final purity, energy conservation, entropy type, approximation error, and whether the result belongs to one theory or an ensemble.

Required background. Quantum Information for Black-Hole Physics supplies density operators, purifications, quantum channels, trace distance, and von Neumann entropy.

Helpful background. Black-Hole Thermodynamics and Semiclassical Gravity supplies generalized entropy and evaporation; Quantum Fields in Curved Spacetime supplies the Hawking calculation; Black-Hole Information supplies Page, island, and reconstruction results; Quantum Gravity supplies the microscopic-program context.

The complete map comes first

Let $X$ denote every input degree of freedom allowed by the proposal and $Y$ every final degree of freedom. Exact unitarity between complete in- and out-spaces means that there is a single linear, surjective map

$$U:\mathcal H_X\longrightarrow\mathcal H_Y, \qquad U^\dagger U=\mathbf 1_X, \qquad UU^\dagger=\mathbf 1_Y.$$

The first identity alone makes $U$ an isometry into $\mathcal H_Y$; it does not establish surjectivity. If remnants, baby universes, soft sectors, an interior at late time, or superselection labels remain, they belong in $Y$ or must be shown not to carry correlations. When the regulated output admits the stated tensor factor, tracing its complement gives a completely positive trace-preserving channel on the radiation,

$$\mathcal N(\rho_X) =\operatorname{Tr}_{Y\setminus\mathcal R} \left(U\rho_XU^\dagger\right),$$

which can be noninvertible even though the enlarged evolution is unitary. This is ordinary subsystem mixing, not fundamental loss. For a nonfactorizing gravitational algebra, the analogous operation is restriction to the declared radiation subalgebra rather than an unqualified partial trace.

The original superscattering proposal instead allowed a map from density operators to density operators that need not factor as $S(\cdot)S^\dagger$; see Hawking 1976, §§II–IV, pp. 2462–2471. A modern proposal must say whether it rejects that conclusion, enlarges the output, conditions on a sector, or only computes an averaged observable.

Purity and entropy requirements

For a pure input $\rho_X=\lvert\psi\rangle\langle\psi\rvert$, a complete unitary process has

$$\operatorname{Tr}\rho_Y^2=1, \qquad S_{\mathrm{vN}}(Y)=0.$$

The radiation can be mixed at intermediate times because it is entangled with the remaining hole. If evaporation ends with no other correlated output, its fine-grained entropy must end at zero. If it grew earlier, it must eventually decrease. Unitarity alone does not require a Haar-typical Page shape or even an initial rise: a specially structured unitary can emit unentangled pure registers throughout. The Page curve adds typicality and dimension assumptions to the endpoint constraint. Page conjectured the exact Haar-average expression and Sen proved it. The resulting finite-dimensional benchmark is not a dynamical law; see Page 1993, Eqs. (1)–(11) and Sen 1996, complete proof.

Every entropy claim must name the state, subsystem or algebra, regulator, subtraction, logarithm base, and type. A thermodynamic $S_{\mathrm{BH}}$, a coarse-grained radiation entropy, a swap or ensemble-averaged entropy, and $S_{\mathrm{vN}}(\rho_{\mathcal R})$ answer different questions. Agreement of one scalar curve does not fix the spectrum of $\rho_{\mathcal R}$, its phases, its correlations with a reference, or a decoder.

Exact, approximate, conditional, or averaged

The conclusion should carry one of four qualifiers.

Claim	Required mathematical object	Discriminating check
Exact recovery	One-theory isometry or unitary and a complete output algebra	Verify inner products, trace, charges, and final purity for every allowed input.
Approximate recovery	Channel and recovery map with a norm and error $\varepsilon$	Bound reference-system trace distance or entanglement fidelity uniformly on the code.
Conditional recovery	A sector, postselection outcome, or final boundary condition with probability and normalization	Check linearity, signaling, interaction sensitivity, and the unconditional channel.
Ensemble-averaged result	Measure over Hamiltonians or theories and the averaged moment being computed	Show whether a corresponding observable exists in one member and how factorization is restored.

Energy conservation is an independent requirement. If information transfer is carried by extra quanta, long-lived remnants, or nonlocal couplings, the proposal must close the asymptotic charge and stress-energy balance. An isometric map that violates the stated Hamiltonian constraint is not a black-hole evolution, while an energy-conserving thermal channel can still lose quantum coherence.

Adverse checks that separate proposals

Three tests are especially useful. First, entangle the input with an external reference $Q$ and require the final channel to preserve distinguishability and entanglement with $Q$. This exposes maps that reproduce expectation values for one state but fail on superpositions. Second, include every proposed endpoint sector and ask whether information is merely displaced into an inaccessible factor. Third, compare two channels with the same radiation entropy curve but different off-diagonal elements or mutual information with $Q$.

The final comparison can be written as

$$\varepsilon_{\mathrm{rec}} =\sup_{\rho_{QX}} \frac12 \left\lVert (\operatorname{id}_Q\otimes\mathcal R\mathcal N)(\rho_{QX}) -\rho_{QX} \right\rVert_1.$$

A proposal claiming exact recovery requires $\varepsilon_{\mathrm{rec}}=0$ on its declared domain; an approximate proposal must show how the error scales. Generic quantum-channel theory is not redeveloped here, and explicit string or holographic constructions remain outside this page’s scope. The task here is to apply those tools to the complete black-hole system without claiming more than the stated norm establishes.

Results and open questions through 2026-08-19

Result or formulation	Assumptions	Controlled result	Limitation
Hawking superscattering, Hawking 1976, §§II–IV, pp. 2462–2471	Semiclassical collapse, complete evaporation, and no additional correlated output	Gives a precise density-operator formulation of a nonunitary complete-system map	Does not supply a microscopic quantum-gravity output space
Page benchmark, Page 1993, Eqs. (1)–(11) and Sen 1996, complete proof	Haar-random pure state on a finite bipartite Hilbert space with declared dimensions and logarithm base	Gives the exact average subsystem entropy and its dimension crossover	Haar typicality does not specify evaporation dynamics, locality, or energy conservation
Island Page curves, Almheiri et al. 2020, §§2–6	Specified low-dimensional or holographic bath model, state, QES candidates, replica saddles, contour, and continuation	Gives controlled generalized-entropy turnovers in the stated approximation	Does not alone give one-theory amplitudes, a Lorentzian transfer channel, or an efficient decoder
Holographic boundary evolution, Harlow 2016, §§6.2 and 6.8–6.9	A fixed AdS/CFT boundary theory with its complete boundary Hilbert space and Hamiltonian	Supplies a unitary microscopic boundary description in that setting	The map to generic asymptotically flat evaporation and its accessible radiation algebra remains unresolved
Remnant or hidden-sector purification, Harlow 2016, §4.6	Enlarged final Hilbert space retaining correlations after the semiclassical hole disappears	Is kinematically compatible with purity of the complete output	Capacity, production, lifetime, stability, and observable coupling are additional dynamical constraints

Competing interpretations therefore agree that a Page turnover is meaningful but differ on what it establishes: one treats it as strong evidence of unitary microscopic recovery, another restricts it to the entropy observable computed by a gravitational path integral. The modern derivations and explicit limits are compared in Almheiri et al. 2021, §§4–11.

The conclusions could change if a calculation supplies reference-system amplitudes and uniform recovery errors in one fixed theory, if a proposed endpoint closes both information and energy budgets, or if a counterexample produces the same claimed Page curve with demonstrably nonunitary complete evolution.

Check your understanding

A model gives $S_{\mathrm{vN}}(\mathcal R)=0$ at the end of evaporation for one chosen initial state. Does that establish a unitary evaporation map?

Solution

No. A pure output for one input does not show linear, inner-product-preserving evolution on a complete family of inputs. One must specify all output sectors, entangle the input with a reference, and test whether the channel preserves coherences and distinguishability uniformly. A reset channel, for example, sends every input to the same pure state and has zero output entropy while destroying all input information. Energy and charge conservation must also be checked independently.

References

• Almheiri, Ahmed, Thomas Hartman, Juan Maldacena, Edgar Shaghoulian, and Amirhossein Tajdini, “Replica Wormholes and the Entropy of Hawking Radiation,” Journal of High Energy Physics 05 (2020) 013. DOI. Open manuscript
• Almheiri, Ahmed, Thomas Hartman, Juan Maldacena, Edgar Shaghoulian, and Amirhossein Tajdini, “The Entropy of Hawking Radiation,” Reviews of Modern Physics 93 (2021) 035002. arXiv
• Harlow, Daniel, “Jerusalem Lectures on Black Holes and Quantum Information,” Reviews of Modern Physics 88 (2016) 015002. DOI. Open manuscript
• Hawking, Stephen W., “Breakdown of Predictability in Gravitational Collapse,” Physical Review D 14 (1976) 2460–2473. DOI
• Page, Don N., “Average Entropy of a Subsystem,” Physical Review Letters 71 (1993) 1291–1294. DOI. Open manuscript
• Sen, Siddhartha, “Average Entropy of a Quantum Subsystem,” Physical Review Letters 77 (1996) 1–3. DOI. Open manuscript