Information Loss, Remnants, and Long-Lived Interiors

Resolution classes differ most sharply in where they place correlations that leading semiclassical radiation does not contain. Fundamental loss says no carrier survives. A remnant assigns them to a small exterior object; a long-lived interior assigns them to a large or persistent internal sector; a baby-universe description assigns them to another component; unitary release puts them in subtle radiation correlations. Dimension, energy, causal accessibility, and the meaning of the complete algebra must be compared separately.

Required background. Quantum Information supplies isometries, channel dilations, entropy, and dimension bounds used to track the correlations.

Helpful background. Semiclassical Gravity supplies evaporation and endpoint approximations; Quantum Fields in Curved Spacetime supplies the Hawking radiation state; Causality and Global Structure supplies horizons, singular boundaries, and disconnected future sectors.

Correlation carriers and their capacity

Let $Q$ purify an initial black-hole state, let $R$ be all exterior radiation, and let $H$ denote any hidden final sector. A unitary completion has an isometry

$$V:\mathcal H_{B_0}\longrightarrow\mathcal H_R\otimes\mathcal H_H, \qquad V^\dagger V=I.$$

The competing positions can be organized by five questions. Is $H$ physically present? How large is $\mathcal H_H$? Does an asymptotic observer have access to it? How long does it persist? Is the map $V$ state independent and compatible with energy and gravitational constraints? Merely naming a carrier does not answer its capacity or dynamics.

If $H$ is one-dimensional, the radiation alone must distinguish every initial state. If $R$ is approximately independent of the initial microstate, orthogonality must instead be carried by $H$. For $N$ mutually orthogonal inputs that yield the same pure radiation state, $\dim\mathcal H_H\geq N$. An initial black hole with thermodynamic entropy $S_{\mathrm{BH}}$ suggests $N\sim e^{S_{\mathrm{BH}}}$ distinguishable microstates, so a universal small remnant may require an exponentially large internal state space despite its small exterior mass.

Fundamental loss and unitary release

In fundamental information loss, the complete map is a noninvertible channel $\Phi:B_0\to R$, not the reduction of a physical hidden sector. Distinct pure inputs may reach the same mixed state. Hawking’s original proposal used this complete-system interpretation of the semiclassical endpoint Hawking 1976, §§ II–IV. It is logically consistent as a map on density operators, but it abandons closed-system unitarity and must independently address conservation laws, locality, and correlations with an external reference.

In unitary release, $H$ is trivial at the end and the correlations reside in high-order observables of $R$. Approximate thermality of each mode does not exclude this. The hard question is dynamical: leading pair creation makes each new exterior quantum entangled with an interior partner, so small independent corrections cannot generically reverse the accumulating entanglement. Mathur’s entropy inequality makes this obstruction precise under its pair-factorization assumptions Mathur 2009, §§ 4–6. A successful release mechanism must change a stated assumption rather than merely attach phases invisible to the entropy bound.

Remnants, long interiors, and other sectors

A remnant retains a bounded exterior mass after the semiclassical evaporation era. Its attraction is immediate: the enlarged state $RH$ may remain pure even when $R$ is mixed. Its burdens are capacity, production, stability, and eventual communication. A large internal degeneracy is not automatically inconsistent, but an effective theory must explain why it does not cause uncontrolled production or violate the regime in which species are counted.

A long-lived interior separates small exterior size from internal state capacity. Large interior volume, a long throat, or delayed decay can hold correlations without making them available at ordinary future null infinity. Geometry alone, however, is not a microscopic Hilbert-space count. One must specify boundary conditions at the endpoint and whether the interior is an independent factor, a gauge-redundant description, or data reconstructible from the exterior.

A baby-universe sector plays the algebraic role of $H$ but may be causally disconnected from the parent asymptotic region. The parent radiation channel is then mixed while the enlarged multiverse state can be pure. Whether this resolves “the” paradox depends on which algebra the question calls complete. A delayed-release model instead lets $H$ persist for a long time and later transfer its state to very soft quanta; its lifetime, available energy, and backreaction must be shown rather than assumed.

Page’s quantitative evaporation model takes a different route: it assumes complete unitary release and estimates the radiation entropy from the smaller of two coarse-grained entropy capacities Page 2013, Eqs. (6)–(12). That conditional curve illustrates the unitary-release class; it does not settle whether a remnant or hidden sector exists.

Results and open questions as of 2026-08-19

No general theorem selects one resolution class for realistic four-dimensional evaporation. Evidence depends on the semiclassical model, factorization, endpoint boundary condition, and whether an enlarged Hilbert space or only one asymptotic algebra is called complete.

Evidence or proposal	Correlation carrier and result	Scope and limitation
Hawking semiclassical endpoint Hawking 1976, §§ II–IV	No final carrier in the exterior theory; pure states undergo a mixed-state map	Leading semiclassical geometry does not define a microscopic endpoint Hilbert space
Small-correction entropy bound Mathur 2009, §§ 4–6	Perturbatively corrected independent pairs do not transfer enough correlation to radiation	Conditional on the named pair structure and smallness norm; it does not exclude order-one structural changes
Long-lived remnant	Correlations remain in a high-dimensional low-mass sector	Dimension counting is elementary, but production, stability, and coupling are model-dependent and unresolved
Long interior or baby universe	Correlations occupy an internal or disconnected algebra, so the enlarged state may be pure	Purity of the parent asymptotic algebra is not restored unless that sector is reconstructible or later returns
Unitary radiation release Page 2013, Eqs. (6)–(12)	Fine-grained correlations ultimately reside in radiation under the model’s unitary-release assumption	Entropy expectations and consistency arguments do not by themselves provide a complete microscopic four-dimensional amplitude

The strongest conservative interpretation is that semiclassical theory diagnoses its own insufficiency but does not choose the replacement. The stronger unitary interpretation treats exact loss as unacceptable and regards remnants, new interior identifications, or nonlocal release as competing completions. Evidence for one formulation is not automatically evidence against every other formulation.

The conclusions could change after a controlled microscopic endpoint calculation fixes the complete Hilbert space, or after a remnant model passes production and stability tests, or after a recovery protocol demonstrates radiation purification while respecting causal, energy, and backreaction limits.

A finite-dimensional adverse test

Take $N$ orthogonal inputs $\lvert i\rangle$ and define $V\lvert i\rangle=\lvert0\rangle_R\lvert i\rangle_H$. The enlarged evolution is an exact isometry, yet every radiation state equals $\lvert0\rangle\langle0\rvert$. Declaring $R$ complete produces maximal operational loss; declaring $RH$ complete produces none. If one now forces $\dim H=1$, the map cannot preserve inner products. This test exposes the formulation choice without deciding whether gravity supplies such an $H$.

The model is deliberately kinematic. It does not establish locality, a semiclassical geometry, a coupling that releases information, or finite energy cost. Generic quantum channels belong to Quantum Information; fixed-background radiation belongs to Quantum Fields in Curved Spacetime; explicit string and holographic carriers are treated separately.

Exercise

Suppose $N$ orthogonal initial states all give radiation supported in the same subspace of dimension at most $d_R$, while the complete final evolution is an isometry into $R\otimes H$. Derive a necessary lower bound on $\dim H$. Why would a separate rank bound on each output not suffice?

Solution

The image of the isometry contains $N$ mutually orthogonal vectors. A common radiation support of dimension $d_R$ gives total dimension at most $d_R\dim H$. Therefore $N\leq d_R\dim H$, or $\dim H\geq\lceil N/d_R\rceil$. If all radiation states are identical pure states, their common support has dimension one and the hidden sector needs at least $N$ states. By contrast, $V\lvert i\rangle=\lvert i\rangle_R\lvert0\rangle_H$ has rank-one radiation outputs with different supports and needs only $\dim H=1$. The bound tests shared capacity, not dynamical viability.

References

• Hawking, Stephen W. 1976. “Breakdown of Predictability in Gravitational Collapse.” Physical Review D 14: 2460–2473. DOI.
• Mathur, Samir D. 2009. “The Information Paradox: A Pedagogical Introduction.” Classical and Quantum Gravity 26: 224001. Open text.
• Page, Don N. 2013. “Time Dependence of Hawking Radiation Entropy.” Journal of Cosmology and Astroparticle Physics 2013 (09): 028. Open text.