Firewall Paradox
The 2012 AMPS argument that black hole complementarity is self-contradictory for old black holes. The preferred resolution: an infalling observer hits a wall of high-energy quanta at the horizon and burns up, breaking the equivalence principle.
Placeholder for a 3D visualisation of Black Hole Information Paradox. The interactive scene will land in Phase 3. Hawking showed in 1974 that black holes radiate. If the radiation is purely thermal (random) and the black hole eventually evaporates, the information about what fell in is destroyed. That breaks one of the foundational rules of quantum mechanics, that information is preserved by any allowed physical process. Hawking himself argued for decades that information really is lost and quantum mechanics has to be modified at black hole horizons. Most of the field disagreed and treated the apparent contradiction as a paradox to be solved within standard quantum theory. Hawking conceded the bet in 2004. Today nearly everyone agrees information must be preserved; the contested question is by what mechanism. Black Hole Complementarity, the Firewall paradox, ER=EPR, Soft Hair, and Fuzzballs are five proposals for that mechanism. The 2019 Page curve calculations using replica wormholes (Penington; Almheiri, Engelhardt, Marolf, Maxfield) showed that unitarity can be recovered within semiclassical gravity, which is the strongest concrete progress the field has made in two decades.
In one sentence
Almheiri, Marolf, Polchinski, and Sully (AMPS) proved in 2012 that the standard dual-picture resolution quietly assumes three things that cannot all be true at once for a heavily-evaporated black hole: information is preserved, the late Hawking radiation is correlated with the early radiation in a specific way, and the infalling observer sees a smooth horizon. Their preferred fix is a high-energy firewall at the horizon, breaking the equivalence principle to save quantum mechanics.
The claim
The AMPS paradox starts from a counting argument due to Don Page. For an evaporating black hole, the entropy of the Hawking radiation initially grows, but once the black hole has evaporated past its halfway point, the radiation entropy has to start decreasing again so the final state can be pure (information preserved). For the entropy to come back down, each piece of late-emitted radiation has to be highly entangled with the radiation that was emitted earlier. This is what the Page curve says, and it is what unitarity demands. So far, no contradiction.
The contradiction shows up when AMPS combine this with the equivalence principle. Each Hawking particle is emitted in an entangled pair with a partner falling into the black hole. For the infalling observer to see a smooth horizon, that partner has to be entangled with degrees of freedom just inside the horizon. But the late-radiation particle is already entangled with the early radiation by the Page argument. Quantum mechanics does not allow the same degree of freedom to be maximally entangled with two different systems at once (the monogamy of entanglement). One of the three assumptions has to break: information is not really preserved (Hawking's original position, now widely rejected), the late radiation is not really correlated with the early radiation, or the horizon is not really smooth. AMPS preferred dropping smoothness, hence the firewall.
The 2019 Page curve calculations using replica wormholes showed unitarity can be preserved in semiclassical gravity without requiring a literal firewall at the horizon. Most of the field now interprets this as making firewalls unnecessary, though AMPS's original argument has not been formally refuted within its own assumptions. The argument was the single most important paper in the modern paradox debate: it broke the consensus that complementarity was the answer and forced every subsequent proposal (ER=EPR, soft hair, fuzzballs) to take a position on what the firewall paradox demonstrated.
The family stance
Most physicists now accept that information is preserved when matter falls into a black hole. Hawking conceded this point publicly in 2004, paying off a 1997 bet with John Preskill. The contested question is the mechanism. Black hole complementarity, ER=EPR, soft hair, and fuzzballs each propose different machinery for how information escapes; the firewall paradox is the argument that exposed why a naive resolution cannot work. The 2019 Page curve calculations using replica wormholes have shown that unitarity can be recovered within semiclassical gravity, but the question of what an infalling observer experiences at the horizon locally remains open.
Predictions
- An observer falling into an old black hole (more than half evaporated by Hawking radiation) encounters a high-energy 'firewall' at the horizon and burns up; the equivalence principle is violated locally there
- Young black holes (less than half evaporated) still have smooth horizons; the firewall only forms once the radiation has accumulated enough entanglement to force the contradiction
- If the firewall is real, it is a curtain of Planck-scale quanta at the horizon whose temperature is set by the black hole's age and mass; in principle detectable by an infalling probe if the equivalence principle locally fails
Evidence
- The AMPS argument's logical structure has been independently re-derived by multiple groups (Mathur 2009 had a closely related earlier version; Bousso et al replicated the AMPS argument shortly after the original paper)
- Page's 1993 calculations showing that the von Neumann entropy of the Hawking radiation must follow the Page curve for unitarity are not contested; this is the input AMPS relies on
- The monogamy of entanglement is a theorem of quantum mechanics, not an extra assumption; AMPS's combinatorial argument follows from standard QM applied to the Page-curve setup
- No explicit construction has shown how complementarity, in its 1993 form, avoids the AMPS contradiction; subsequent proposals (ER=EPR, soft hair, replica wormholes) accept that something in the original framing had to give
Counterpoints
- Violating the equivalence principle at the horizon is a heavy theoretical price; general relativity predicts no special local physics at the horizon of a sufficiently large black hole, since spacetime curvature there can be arbitrarily small
- No astrophysical observation supports the existence of firewalls; ringdown spectra from LIGO mergers and EHT shadow images are consistent with classical Kerr horizons within current sensitivities
- Post-2019, the replica wormhole calculations recover the Page curve without invoking a literal firewall, providing an explicit semiclassical gravity mechanism that AMPS treated as impossible; this is widely read as evidence that AMPS missed a contribution to the gravitational path integral
- The AMPS argument is constructed using effective field theory near the horizon, which is exactly where one expects effective field theory to break down for old black holes; some authors argue the paradox is an artifact of pushing semiclassical methods past their domain of validity
Variants in this family
▸Go deeperTechnical detail with proper terminology
Page time: the moment in a black hole's evaporation at which the radiation entropy equals the remaining black hole entropy. For a Schwarzschild black hole this is roughly when half the initial mass has evaporated. Before the Page time, AMPS's argument does not bind; after it, the monogamy contradiction triggers.
Monogamy of entanglement: if system A is maximally entangled with system B, then A cannot also be maximally entangled with any other system C. This is a strict consequence of the structure of pure states in a Hilbert space and is the technical lever AMPS use to force the three-way contradiction.
The 'A = R_B' identification: ER=EPR's proposed escape is to identify the black hole interior modes (A) with a coarse-grained version of the early Hawking radiation (R_B). If this identification is valid then the infalling-partner entanglement and the early-late radiation entanglement are the same entanglement, monogamy is not violated, and the firewall is not forced. Whether this identification is consistent is the central technical question that ER=EPR raises.
Replica-wormhole resolution: the 2019 entanglement-wedge calculations show that for the radiation past the Page time, the dominant gravitational path integral saddle point includes a wormhole connecting copies of the spacetime, and this saddle re-includes a chunk of the black hole interior in the radiation's entanglement wedge. The Hawking-radiation-to-interior identification AMPS treated as illegal turns out to be enforced by the gravitational path integral itself.
References
Last reviewed May 18, 2026
Spotted an error? Have a source to add?
Prefer email?
You can also send a prefilled email with the variant URL already filled in.