THE INFORMATION PARADOX

In 1997 three physicists signed a wager and notarized it. and bet that information falling into a black hole is destroyed forever; John Preskill bet that it survives. The stakes were an encyclopedia — "from which information can be recovered at will" — to be handed to the winner. The bet was the public face of a disagreement that had been running, mostly in private, for twenty-one years.

It started in 1976. Two years after he showed that black holes radiate, published a paper with a deliberately provocative title — Breakdown of predictability in gravitational collapse — arguing that the radiation carried a fatal implication. The radiation is thermal: featureless, characterized by nothing but a temperature, like the glow of a hot coal. If a hole forms from a collapsing star and then radiates itself away entirely into thermal photons, the detailed quantum state of the star — every bit of structure it ever had — is simply gone. Two different stars that collapsed into the same-mass hole would produce identical radiation. The map from "what went in" to "what came out" is many-to-one. That is forbidden in quantum mechanics.

heard the claim at a small 1981 gathering in 's house and, in his own telling, was alarmed. To accept Hawking's conclusion was to give up the most basic rule of quantum theory: that information is never created or destroyed. Susskind called the ensuing decades the "black hole war" — not a war between Hawking and reality, but between Hawking and a small group, Susskind and Gerard 't Hooft chief among them, who refused to let the rule go. In 2004, at a conference in Dublin, conceded the bet. He handed Preskill a baseball encyclopedia. He did not, most of his colleagues felt, fully explain why he had lost — and the problem he conceded is, two decades later, still not closed.

The paradox is the head-on collision of two statements that are each, on their own home turf, beyond reproach.

The first is Hawking's 1974 result: a black hole of mass $M$ emits Hawking radiation at the Hawking temperature, and that radiation is thermal to an extraordinary approximation. Each emitted quantum is entangled with a partner that falls behind the Event Horizon; the quantum you detect outside is, on its own, in a mixed state with no hidden message in it. Add up the whole emission and you appear to get pure noise.

The second is Unitarity, the spine of quantum mechanics. Time evolution is generated by a Hamiltonian through a unitary operator $U(t)$, and unitary operators preserve the inner product. A consequence is that a pure state stays pure forever — the information defining it is reshuffled but never lost. The von Neumann entropy of a closed system never changes.

$|\psi(t)\rangle = U(t)\,|\psi(0)\rangle, \qquad U^\dagger U = \mathbb{1}, \qquad S = -\operatorname{Tr}(\rho \ln \rho) = \text{const}$

In plain terms: quantum evolution is a perfect, reversible bookkeeper. Whatever state the universe is in, you could in principle run the film backward and recover exactly where it started. A black hole that turns a pure star into a cloud of thermal static has torn a page out of that ledger.

The contradiction is now sharp. Start with a pure state — a collapsing star. Let it form a hole and evaporate completely. Hawking's calculation says the end product is thermal radiation, a mixed state with positive entropy. Unitarity says the end product must still be pure, with zero entropy. Both derivations look airtight. One of them has a hidden flaw, and for forty years nobody could point to it with confidence.

For its first seventeen years the paradox was argued in words. In 1993 turned it into a single graph that made the disagreement quantitative and falsifiable in principle. His move was to stop asking "is information lost?" and start asking "what does the entanglement entropy of the radiation do over time?"

Track $S_{\text{rad}}$, the von Neumann entropy of everything that has been emitted so far, as the hole evaporates. Hawking's bookkeeping is simple: each new thermal quantum adds entanglement, so $S_{\text{rad}}$ climbs steadily and reaches its maximum exactly when the hole vanishes. The final radiation is maximally mixed; the endpoint entropy is the lost information.

Page's bookkeeping uses a theorem about entanglement. The entropy of a subsystem can never exceed the entropy of the smaller of the two complementary pieces. Early on, the radiation is the small piece, so $S_{\text{rad}}$ may rise. But the hole's own information capacity is its Bekenstein–Hawking entropy, $S_{\text{BH}} \propto M^2$, and that is shrinking. Once the hole is the smaller piece, $S_{\text{rad}}$ must track the falling $S_{\text{BH}}$ down to zero. The result is the Page curve: a rise, a turnover at the Page time, and a fall back to zero — leaving a pure final state.

$S_{\text{rad}}(f) \;=\; \min\!\big(S_{\text{thermal}}(f),\; S_{\text{BH}}(f)\big) \;\approx\; \min\!\big(f,\; 1-f\big)$

Here $f$ is the fraction of the hole's entropy already radiated: $f=0$ is the fresh hole, $f=1$ is gone. The radiation's entropy is bounded by whichever description holds less information at that moment — the growing pile of quanta, or the dwindling hole. The turnover happens at the Page curve time, near $f=\tfrac12$.

The power of Page's reframing is that the two answers now disagree about something concrete and local in time — the slope of a curve after the halfway mark — rather than about a philosophical endpoint. Any correct theory of evaporation must produce the Page curve, not Hawking's monotone ramp. For decades, no semiclassical calculation could.

Page's curve also explains why the paradox is so stubborn, through a tension in the ledger between two quantities that leave the hole at different times: energy and information.

Energy leaves early. The luminosity rises as the hole shrinks, but even a slowly-evaporating large hole is steadily losing mass-energy from the start. Track the radiated energy against the radiated entropy and it is convex: with $S_{\text{BH}} \propto M^2$, a hole that has shed half its entropy still has $\sqrt{1/2} \approx 71\%$ of its mass. Most of the energy pours out in the final stages.

Information, if it comes out at all, must come out late. Before the Page time the radiation is genuinely thermal — it carries essentially no information about the initial state, exactly as Hawking computed. The information can only begin to emerge after the turnover, encoded in subtle correlations between late quanta and the early radiation already collected. The amount available in the radiation is the gap between the thermal entropy and the true (Page) entropy:

$I_{\text{rad}}(f) \;=\; S_{\text{thermal}}(f) - S_{\text{Page}}(f) \;=\; \max\!\big(0,\; 2f - 1\big)$

This is exactly zero until $f=\tfrac12$ and then climbs — information is locked in the hole through the entire first half of its life and only leaks out once the hole has passed the Page time. That is a brutal constraint. By the time information can escape, most of the energy is already gone and the hole is small; the late, low-energy quanta have to carry the entire informational content of the original star. Whether thermal-looking radiation can secretly do that is the technical heart of the whole subject.

Faced with the contradiction, physicists generated a menu of resolutions, each paying a different price. Information is simply destroyed (Hawking's original position): clean, but it abandons unitarity and, worse, threatens energy conservation — 't Hooft and others argued the same machinery that loses information would let it leak from the vacuum. Information escapes in the radiation: keeps unitarity, but demands that the "thermal" radiation be subtly non-thermal in just the way Page's curve requires, with no agreed mechanism. A stable Planck-mass remnant locks the information away forever in a tiny object — but a tiny object forced to store unbounded entropy causes its own pathologies. A baby universe buds off carrying the information into a region we can never reach, which for our universe is loss by another name.

Then in 2012 the AMPS argument (Almheiri, Marolf, Polchinski, Sully) sharpened the screw: they showed you seemingly cannot have all of unitarity, a smooth horizon, and ordinary local physics at once. Saving information might require a firewall — a wall of high-energy quanta at the horizon that incinerates an infalling observer, in flat violation of the equivalence principle that built relativity on.

The breakthrough came in 2019. Using the gravitational path integral, Penington and, independently, Almheiri and collaborators found that the entropy of the radiation receives a contribution from disconnected regions of spacetime — islands — that sit behind the horizon yet count as part of the radiation's information. Including them, and the replica-wormhole geometries that justify them, the calculation reproduces the Page curve directly. The falling branch fell out of the math for the first time, rather than being assumed. It did not require a firewall, and it did not require abandoning the smooth horizon. It did leave a deep puzzle — by what concrete process the information physically transfers — but it converted a forty-year standoff into a computation.

The information paradox is not a footnote about an exotic object. It is the cleanest place where general relativity and quantum mechanics are forced to contradict each other, and so it functions as a proving ground for any theory that hopes to unify them.

The reason the conflict bites here and nowhere else is that a black hole is the one system in which gravity's geometry and quantum mechanics' bookkeeping cannot be kept in separate rooms. Hawking radiation is a quantum effect that depends on the global causal structure — the Event Horizon — of a curved spacetime. Resolve it and you have learned something true about Quantum gravity that no tabletop experiment could have told you. The 2019 results suggest the answer is unitary and that spacetime is, at the deepest level, built out of entanglement rather than the other way around — a claim that grew directly out of this fight and now organizes much of theoretical physics.

It is also unfinished, which is the honest note to end on. The island calculations reproduce the Page curve but do not yet say, in ordinary spacetime language, how a bit travels from inside the hole to a photon far away. That gap is where the next decade of work lives. The story runs straight into the wider question of what relativity doesn't say — the regimes where Einstein's theory hands the problem to a quantum theory of gravity it cannot itself provide — and it began, two sections of physics ago, with the discovery that a black hole has a temperature at all, in black-hole thermodynamics and the radiation that follows from it.