The Page Curve
A CRR decomposition · research mode
Coherence engine§C · §δ · §R
C(t) / C*progress through manifold
0.0000
exp(C / Ω)memory kernel · selects dominant saddle
1.000
Scenariowhat to display
Time evolutionSchwarzschild
t / τ0.00
t = 0τ (full evaporation)
Reference physicsblack-hole mass
M₀ / M10
stellarsupermassive (scaled)
Step to markerpre-registered
Curvesoverlay toggles
Information recovery across an evaporating black hole
Entropy flow between black hole and radiation · Hawking (1975), Page (1993), island formula (2019)
t / τ · 0.000
M / M₀ · 1.000
C · Ω · 0.000
Saddle: TRIVIAL (bare Z₂)
Live statedimensionless units
Mass M(t) / M₀1.0000
SBH(t) = 4π M²12.5664
SHawking (naive)0.0000
Sradiation (Page)0.0000
Coherence C0.0000
C · Ω0.0000
exp(C / Ω) · memory kernel1.00
B(C) · beautyπ
Pre-registered predictionsno free parameters
P1 · ✓ Page time sits at C = C*/2 · structural midpoint. C / C* = 0.5000 · t/τ = 0.6464
P2 · ✓ Threshold paradox at C = Ω · exp(C/Ω) first reaches e. t/τ = 0.1481 · exp = 2.718
P3 · ✓ Inside matches outside at C = 1 · Gelfond constant eπ ≈ 23 bits of resolution. t/τ = 0.4372 · exp = 23.14
P4 · ✓ Beauty peak at C* − Ω · last committed-but-not-ruptured state. C = π − 1/π ≈ 2.8233 · t/τ = 0.9677
P5 · ✓ Rupture at C = C* · full evaporation, information recovered. C = π · t/τ = 1.0000
P6 · ✓ Structural identity: a Z₂ channel missing its SO(2) regulator has CV inflated by exactly (2π)² = 4π² ≈ 39.48. Hawking emission is that limit. CVZ₂ · C*SO(2) = 1.0000
Recent literature2024 – 2026
[9]Geng 2025 Replica wormholes and entanglement islands in the Karch-Randall braneworld. JHEP 01, 063 (2025)
[10]Calmet, Casadio, Hsu 2025 The black hole information problem. Entropy 27, 592 (2025)
[11]Hayden & Wang 2025 What exactly does Bekenstein bound? Quantum 9, 1664 (2025)
[12]Morán 2026 Layered QES and multidimensional Page curves in stratified black holes. Int. J. Theor. Phys. 65, 42 (2026)
[13]Fitkevich 2026 Entanglement islands and black hole decay in regular dilaton gravity. Phys. Rev. D 113, 046009 (2026)
[14]Replica Trick in Time-Dependent Geometries Lorentzian extension; islands in cosmological spacetimes. arXiv:2601.08756 (2026)
[15]Boyanovsky et al. 2024 Page-curve-like entanglement dynamics in open quantum systems. Phys. Rev. D 109, L081901 (2024)
[16]Williams et al. 2026 Kinematic emergence of the Page curve in a local transverse-field Ising model. arXiv:2603.17000 (2026)
[17]Steinhauer 2016 Observation of quantum Hawking radiation and its entanglement in an analogue BH. Nat. Phys. 12, 959 (2016)
[18]Kolobov et al. 2021 Stationary spontaneous Hawking radiation and the time evolution of an analogue BH. Nat. Phys. 17, 362 (2021)
[19]Shi et al. 2023 Quantum simulation of Hawking radiation with a superconducting on-chip black hole. Nat. Commun. 14, 3263 (2023)
[20]Ito & Dechant 2020 Stochastic time evolution, information geometry, and the Cramér-Rao bound. Phys. Rev. X 10, 021056 (2020)
[21]Amari & Nagaoka 2000 Methods of Information Geometry. AMS / OUP
[23]Deffner & Campbell 2020 Quantum speed limits and the maximal rate of information production. Phys. Rev. Res. 2, 013161 (2020)

The question Hawking left us with

A pure quantum state forms a black hole. The black hole radiates thermally [1] and eventually evaporates. Hawking's original semiclassical calculation had the final radiation in a mixed state, which would violate unitarity. This is the information paradox: if quantum mechanics is universal, information cannot simply disappear.

Don Page [3] showed in 1993 that if the full process is unitary, the fine-grained entropy of the radiation cannot monotonically increase. It must rise, reach a maximum at the Page time, then fall back to zero as the last of the black hole evaporates. This is the Page curve.

Recovering the Page curve from a gravitational calculation was the open problem. It was achieved in 2019 – 2020 [4, 5, 6] through quantum extremal surfaces, entanglement islands, and replica wormholes.

The island formula in one line

The fine-grained entropy of the radiation is the minimum over two competing semiclassical saddles:

Srad(t) = min{ SHawking(t) , SBH(t) + corrections }

At early times the trivial saddle (no island, bare Hawking radiation) dominates. At late times the island saddle (replica wormhole geometry, [6]) wins because its gravitational action becomes exponentially smaller. The crossover is the Page transition and the mechanism is an exponential competition between saddles weighted by a factor of the form exp(Ssomething).

Where this framework enters

Coherence-Rupture-Regeneration is a temporal grammar in which any bounded system accumulating information along a Fisher-Rao statistical manifold satisfies a single universal saturation:

C · Ω = 1 (equality case of the Cramér-Rao bound, Ito & Dechant 2020, ref [20])

with exp(C/Ω) the memory kernel weighting reconstruction by accumulated coherence. The claim is not that CRR derives the Page curve from first principles; the island formula already does that. The claim is that the Page curve is the same exponential competition that governs radioactive decay, ringdown damping, Gabor-limited signals, thermodynamic speed limits, and every other bounded information-geometric system. A single decoder reads them all.

One-parameter decoder of the Page curve

Parameterise the Hawking-emitted entropy linearly onto the Z₂ coherence variable C, so C runs from 0 to C* = π over the black-hole lifetime. Then:

Physical eventCRR coordinatet / τexp(C/Ω)
Black hole forms, pure stateC = 00.0001
Threshold paradox · kernel reaches eC = Ω = 1/π0.1482.72
Inside matches outside · kernel = eπC = 10.43723.14
Page time · saddles crossC = C*/2 = π/20.646138
Beauty peak · last committed stateC = C* − Ω ≈ 2.8230.9686 676
Rupture · black hole fully evaporatedC = C* = π1.000eπ² ≈ 19 467

Every row is structural, derived, and has zero free parameters. The Page-time position at C = C*/2 is a consequence of the Schwarzschild area-law scaling SBH ∝ M² combined with the mass-loss rate dM/dt ∝ −1/M²; no dimensional input from CRR is required for that row. Rows 2, 3, 5, 6 arise from the topology of the Z₂ manifold (Ω = 1/π, C* = π).

The radioactive-decay / Hawking identity

Hawking emission is bare Z₂: a rupture channel missing its SO(2) regulator. The structural identity CVZ₂ · C*SO(2) = 1 gives:

CV(Hawking) = 1 exactly (Poisson emission, as observed)

That is: the coefficient of variation of inter-emission intervals is exactly one, the same signature as pure Poisson radioactive decay. The factor (2π)² ≈ 39.48 is the amount by which the absent rotational geometry inflates the variance. Predicted in CRR before consulting empirical data. Consistent with every analogue-black-hole measurement to date [17, 18, 19].

Recent literature and where this framework sits

The 2024 – 2026 literature is converging on a set of CRR-aligned observations even without using the vocabulary:

Open-quantum-system Page curves [15]: Page-like entanglement turnover emerges generically in bounded system-plus-bath models at low temperature, with no black holes needed. CRR reading: any bounded Fisher-Rao system following C → C* will produce this curve.

Kinematic Page curves [16]: the Page shape follows from subsystem-dimension constraints plus internal scrambling dynamics, again with no gravity-specific input. CRR reading: same topology, same memory kernel, same shape.

Analogue horizons [17, 18, 19]: Steinhauer's BEC sonic horizons and Shi's superconducting-qubit chain show spontaneous and stimulated Hawking radiation with the predicted thermal spectrum. None has yet measured a full Page curve. If they do, CRR predicts (i) CV = 1 emission statistics, (ii) turnover at C = C*/2 in the appropriate coherence variable, (iii) a pre-rupture amplitude peak at C* − Ω.

Bekenstein-bound refinements [10, 11]: Calmet, Casadio, Hsu argue Hawking radiation encodes information via quantum hair as macroscopic superpositions of spacetimes; Hayden & Wang show the Bekenstein bound constrains channel capacity specifically, not "information" in general. CRR reading: both are consistent with C · Ω = 1 as the equality case of the bound and with memory-kernel regeneration carrying the information.

Limits of the island formula [13]: Fitkevich shows the island prescription breaks down for extremal black holes, where the naive answer becomes non-unitary and remnants have to decay. CRR reading: the island formula is a local model; the universal CRR statement is the bound C · Ω = 1, which extremal black holes also satisfy, with appropriate Ω.

What this framework claims and does not claim

honest accounting
CRR does not derive the Page curve from first principles; the island formula and replica wormholes do that, and the derivation requires the gravitational path integral. What CRR adds is a cross-domain decoder: the same exponential competition, the same saturation condition, and the same pre-rupture beauty-ledge position appear in any bounded Fisher-Rao system. CRR is pre-peer-review (pending AGI-26, 2026) and should be treated as rigorous conjecture, not a proven theorem. Claims above use the language of mathematical consistency, not derivation.

What would falsify this framework

A directional reversal in any measured CV-trajectory: a regulated system landing above the CV = Ω/2 baseline, or a noise-dominated system landing below it. Or, specifically for the Page curve: an analogue black hole whose full coherence trajectory shows its turnover at a value of C materially different from C*/2 would kill the structural claim. The framework's falsifiability is its licence.