Lesson 11.3 · 11. Research Frontiers

The Island Formula

What is Hawking radiation?

Stephen Hawking showed in 1975 that black holes are not completely black: they slowly emit thermal radiation, like a warm object cooling down. This radiation causes the black hole to lose mass until it eventually evaporates entirely. The problem is that this radiation appears to carry no information about what fell into the black hole.

What is unitarity?

Unitarity is a fundamental principle of quantum mechanics stating that information is never destroyed. If you know the state of a system perfectly at one moment, you can in principle reconstruct its entire past and predict its entire future. The apparent loss of information in black holes violates this principle, hence the paradox.

The island formula is a major advance in the black hole information problem. In controlled semiclassical models, it reproduces the Page curve when one includes “islands,” gravitational regions that belong to the radiation's entanglement wedge. It therefore explains how entropy compatible with unitarity can emerge in those models. It does not yet provide the complete microscopic mechanism for a four-dimensional astrophysical black hole.

The Central Idea

When computing the entropy of Hawking radiation, we must include not just the radiation itself, but also an "island", a region inside the black hole that is part of the radiation's quantum description. After the Page time, the island contribution dominates and the entropy decreases, restoring unitarity. Information is not lost; it is encoded in the radiation through the island.

The island formula: after the Page time, a region inside the black hole (the island) is included when computing the entropy of the Hawking radiation, restoring unitarity.

The Formula

S(R) = \min_{\mathcal{I}}\, \text{ext}_{\mathcal{I}} \left[ \frac{\text{Area}(\partial \mathcal{I})}{4 G_N} + S_{\text{bulk}}(R \cup \mathcal{I}) \right]

where:

$R$ is the radiation region (outside the black hole)
$\mathcal{I}$ is the island (a region inside or near the black hole)
$\partial \mathcal{I}$ is the boundary of the island, the quantum extremal surface (QES)
$S_{\text{bulk}}$ is the von Neumann entropy of bulk quantum fields
The minimization/extremization is over all possible islands, including the trivial island $\mathcal{I} = \emptyset$

The formula is a quantum generalization of the Ryu-Takayanagi prescription. The boundary $\partial \mathcal{I}$ is a quantum extremal surface (QES): a surface that extremizes the generalized entropy $S_{\text{gen}} = A/(4G_N) + S_{\text{bulk}}$.

Quantum Extremal Surfaces

A quantum extremal surface (QES) is the key technical concept underlying the island formula. It is a codimension-2 surface $\chi$ that extremizes the generalized entropy:

S_{\text{gen}}(\chi) = \frac{A(\chi)}{4G_N} + S_{\text{bulk}}(\chi)

The first term is the Bekenstein-Hawking area entropy. The second is the von Neumann entropy of quantum fields in the bulk region bounded by $\chi$ and the boundary. The QES prescription, due to Engelhardt and Wall (2014), generalizes the classical RT surface, which extremizes area alone.

Historical Development of the QES

2006: Ryu-Takayanagi: $S = A_{\min}/(4G_N)$ for static spacetimes (classical, leading order)

2007: HRT: Hubeny-Rangamani-Takayanagi generalize to covariant settings, replacing "minimal" with "extremal"

2013: FLM: Faulkner-Lewkowycz-Maldacena add the quantum correction: $S = A/(4G_N) + S_{\text{bulk}} + O(G_N)$

2014: Engelhardt-Wall: Define the QES and prove that $S_{\text{gen}}$ obeys a generalized second law

2019: Penington; AEMM: The QES undergoes a phase transition at the Page time, yielding the island formula and resolving the Page curve

Phase Transitions

The QES can undergo phase transitions: discontinuous jumps in location when a previously subdominant saddle becomes dominant. This is the mechanism behind the Page curve. For an evaporating black hole:

Before Page time: The trivial QES (empty surface, $\mathcal{I} = \emptyset$) dominates, giving $S = S_{\text{bulk}}(R)$
After Page time: A non-trivial QES appears just inside the horizon, giving $S \approx A_{\text{hor}}/(4G_N)$, which decreases as the black hole evaporates

The Large Boost Mechanism

In the explicit JT gravity calculation by Almheiri, Engelhardt, Marolf, and Maxfield (AEMM), the QES moves far from the classical extremal surface due to large boosts near the horizon. Even though the absolute bulk entropy is $O(1)$, its gradient is $O(1/G_N)$ because of the exponential blueshift factor $\exp(2\pi v/\beta)$ near the horizon. This gradient competes with the dilaton (area) gradient, placing the QES at a physically significant location.

The Page Curve

The Page curve describes how the entropy of Hawking radiation should behave if black hole evaporation is unitary:

Before vs. After the Page Time

Before Page time: No island ($\mathcal{I} = \emptyset$) dominates. $S(R) = S_{\text{bulk}}(R)$ grows as more Hawking quanta are emitted. This agrees with Hawking's original calculation.

After Page time: The island $\mathcal{I}$ inside the black hole appears. $S(R) = \frac{\text{Area}(\partial\mathcal{I})}{4G_N} + S_{\text{bulk}}(R \cup \mathcal{I})$. The area term decreases as the black hole shrinks. The net entropy decreases, information comes out.

The overall behavior is: entropy rises, peaks at the Page time, then falls back to zero, exactly the Page curve expected from unitary evolution.

In Penington's calculation, the post-Page-time QES sits at infalling time:

v_{\text{QES}} = -\frac{\beta}{2\pi}\log\left(\frac{S_{\text{BH}}}{c_{\text{evap}}}\right) + O(\beta)

at radial distance $O(G_N)$ from the horizon. The result for the entropy is elegantly simple:

S(R) = \min\left(S_{\text{rad}},\, \frac{A_{\text{hor}}}{4G_N}\right)

Three Foundational Derivations

The fact that three independent approaches all yield the same answer provides strong evidence for the formula.

1. QES Prescription (Penington, May 2019)

Geoffrey Penington applied the quantum RT formula to an evaporating AdS-Schwarzschild black hole with absorbing boundary conditions. He showed that the QES undergoes a phase transition at the Page time. The key insight: after the Page time, the entanglement wedge of the radiation jumps to include an island inside the black hole interior. By entanglement wedge reconstruction (JLMS theorem), operators in the island can be reconstructed from the radiation. This is how information gets out, it is encoded in the radiation via state-dependent reconstruction.

2. Explicit JT Gravity Calculation (AEMM, May 2019)

Almheiri, Engelhardt, Marolf, and Maxfield provided the first explicit computation of the Page curve in JT gravity (2D dilaton gravity) coupled to a CFT. The generalized entropy is:

S_{\text{gen}} = \frac{\phi_0 + \phi(x^+, x^-)}{4G_N} + S_{\text{bulk}}(x^+, x^-)

where the dilaton $\phi$ plays the role of area in 2D. Their calculation precisely reproduces the Hayden-Preskill scrambling time:

t_{\text{HP}} = \frac{\beta}{2\pi}\log(\Delta S) + O(1)

confirming the Hayden-Preskill prediction exactly: information thrown into the black hole enters the radiation's entanglement wedge after exactly one scrambling time.

3. Replica Wormholes (AHMST, November 2019)

Almheiri, Hartman, Maldacena, Shaghoulian, and Tajdini provided a gravitational path integral derivation. They computed $\text{Tr}(\rho_R^n)$ using the replica trick in gravity and discovered new replica wormhole saddle points that connect different copies. Taking the $n \to 1$ limit yields the island rule. The island formula follows from including all saddle points in the gravitational path integral, not from any modification of Hawking's original calculation.

What the Island Formula Resolves

Resolved Puzzles

Information paradox: Hawking radiation entropy follows the unitary Page curve
No firewall needed: The island provides a smooth interior consistent with unitarity, via state-dependent reconstruction
Entanglement wedge of radiation: After the Page time, the radiation encodes interior information
Hayden-Preskill consistency: Information thrown into an old black hole is recoverable after the scrambling time $t_{\text{scr}} \sim (\beta/2\pi)\log S_{\text{BH}}$

Islands in Cosmology

The island formula was originally developed for black holes in AdS spacetime. Extending it to cosmological settings introduces new features and challenges that are absent in the black hole case.

Islands in Closed Universes (Bousso-Wildenhain 2022)

The most systematic study of cosmological islands considers homogeneous, isotropic FRW universes purified by a reference system via the thermofield double construction. The key results depend on spatial curvature and cosmological constant:

Results by Geometry

Closed universe ($k=+1$), $\Lambda \geq 0$: Arbitrarily small positive curvature guarantees the entire universe is an island. For the full spatial slice, $\partial\Sigma_M = \emptyset$, so the area term vanishes and the entropy condition is automatically satisfied. Proper subset islands do not exist.
Closed universe ($k=+1$), $\Lambda < 0$: Both the entire universe and proper subsets can be islands, if $\Lambda$ is sufficiently negative.
Open universe ($k=-1$), $\Lambda \geq 0$: No islands exist. The volume-to-area ratio $V_c/A_c \leq 1/2$ prevents the entropy condition from being satisfied.
Flat universe ($k=0$), $\Lambda < 0$: Islands exist for proper radii $r \gtrsim t_\Lambda^{3/2}/\ell_P^{1/2}$.

Islands in de Sitter Space

In pure 2D de Sitter gravity with a bath (Sybesma 2020), islands exist but the island moves backward in time: a physically puzzling feature whose interpretation remains unclear. The Page curve is formally recovered, but the meaning is debated.

In quasi-de Sitter space (Seo 2022), modeling slow-roll inflation, islands may not exist when the deviation from exact de Sitter is too large. This connects to the swampland program: if the de Sitter conjecture is correct and exact de Sitter is impossible, the island mechanism may fail in realistic cosmologies.

Conceptual Challenges

Challenges for Cosmological Islands

No natural radiation reservoir: In black hole physics, Hawking radiation escapes to infinity. In cosmology, there is no natural "outside" to collect radiation. The thermofield double purification is a mathematical device, not a physical process.
No Page time analog: The black hole Page time marks when the island saddle overtakes the no-island saddle. In cosmology, there is no clear analog driven by evaporation dynamics.
Reference system dependence: Island existence depends on how the purification is constructed, a feature absent in the black hole case where the radiation is physical.

The Gauss Law Objection

A fundamental challenge to the island program was raised by Geng, Karch, Perez-Pardavila, Raju, Randall, Riojas, and Shashi (2021). Their argument identifies a potential inconsistency between islands and long-range gravity.

The Gauss Law Argument

Setup: Suppose an island $\mathcal{I}$ appears as part of the entanglement wedge of radiation $R$. Entanglement wedge reconstruction implies operators in $\mathcal{I}$ can be reconstructed on $R$ and should commute with operators in the complement $\bar{R}$.

Gravitational Gauss law: In a theory with a massless graviton, the ADM energy is a boundary observable. The total energy within any region can be measured by a surface integral at the boundary of that region, at arbitrary distance.

The contradiction: An excitation in the island changes the energy in $\mathcal{I}$. By the Gauss law, this energy change is detectable at $\partial\mathcal{I}$, which lies in $\bar{R}$. Formally: $[\hat{O}_{\mathcal{I}}, \hat{Q}_{\bar{R}}] \neq 0$, violating the requirement that entanglement wedge operators commute with complement operators.

Resolution: All known consistent island computations involve massive gravity. In Karch-Randall braneworlds, the graviton has mass $m_g \sim 1/\ell_{\text{AdS}}$, screening the Gauss law. In JT gravity with a bath, gravity is effectively massive. Standard 4D GR with a massless graviton may not support islands.

This objection raises the disturbing possibility that islands may not exist in standard (massless) gravity, including astrophysical black holes in 4D general relativity. However, possible loopholes exist, including compact spatial topologies where there is no asymptotic boundary from which to perform the Gauss law measurement.

Open Frontiers

Key Open Questions

Real black holes: The formula is proven in JT gravity and simple models. Does it work for 4D Schwarzschild and Kerr black holes?
De Sitter: Can islands resolve cosmological information puzzles?
Mechanism: What is the physical process by which information gets from the island into the radiation? The formula is bookkeeping, not dynamics.
Infalling observer: What does the interior look like after the Page time?
Beyond semiclassical: The formula relies on semiclassical gravity. What happens at the Planck scale?
Closed universes: How does the island formula interact with the gravitational Gauss law that makes the Hilbert space trivial in closed universes?
Compact topology: How does nontrivial spatial topology affect island existence?

Criticisms and Limitations

The formula has been rigorously derived only in 2D JT gravity and simple AdS$_2$ setups. Extension to 4D requires assumptions about the gravitational path integral that are not established.
The replica wormhole derivation relies on Euclidean quantum gravity and analytic continuation. The physical interpretation of replica wormhole saddles is debated.
The formula does not explain how information physically gets from the island into the radiation, it is a bookkeeping device, not a dynamical description.
Cosmological islands may be a mathematical artifact of the thermofield double purification, with no physical content.
The semiclassical approximation may break down precisely where the QES becomes physically interesting.

Key Takeaways

The island formula computes the entropy of Hawking radiation by including contributions from "islands", disconnected spacetime regions inside the black hole
The quantum extremal surface (QES) generalizes the RT surface by extremizing the full generalized entropy: area plus bulk entanglement
The QES undergoes a phase transition at the Page time, producing the Page curve and resolving the information paradox
Three independent derivations (Penington, AEMM, AHMST) all yield the same result
Extension to cosmology reveals that positive curvature guarantees the entire closed universe is an island, while open universes with $\Lambda \geq 0$ have no islands
The Gauss law objection suggests islands may require massive gravity, potentially limiting their applicability to standard 4D GR
The formula resolves the entropy puzzle but does not explain the dynamical mechanism of information transfer

← 11.2 Entanglement & Spacetime 11.4 Holographic Error Correction →