Dynamics, Errors, and Operational Tests

A proposal becomes testable when it names a calculation that can fail inside its own claimed regime. Reproducing an entropy endpoint is not enough: the dynamics, approximation parameter, error norm, timescale, decoder, measurement access, and gravitational backreaction must all be specified. This page gives a method for turning statements such as “information is encoded in the radiation” or “the horizon remains smooth” into discriminating calculations whose outcomes differ among recovery frameworks.

Required background. Quantum Information for Black-Hole Physics supplies channel distance, recovery fidelity, sampling, and circuit resources.

Helpful background. Black-Hole Thermodynamics and Semiclassical Gravity supplies backreaction and evaporation scales; Quantum Fields in Curved Spacetime supplies detector and horizon states; Black-Hole Information supplies recovery and complexity results; Quantum Gravity supplies the domain of microscopic proposals.

Begin with a falsifiable claim graph

Write the target as a chain

$$\text{microscopic input} \longrightarrow \text{controlled dynamics} \longrightarrow \text{observable} \longrightarrow \text{decision rule}.$$

Each arrow needs its own assumptions. The input specifies state, charges, geometry, bath, and boundary conditions. The dynamics names a Hamiltonian, channel, path integral, or effective equation and its domain. The observable is a density-matrix distance, correlator, detector probability, stress tensor, or decoded reference entanglement—not simply “information.” The decision rule states which numerical or analytic outcome would contradict the proposal.

For example, Hayden–Preskill recovery predicts decoupling only after declaring an old hole, diary and radiation dimensions, random-unitary or design assumptions, access to early radiation, and an error criterion; see Hayden and Preskill 2007, §3, Eqs. (3)–(7). A calculation that omits the early-radiation register cannot test that statement.

Build the error budget before interpreting a result

Let $Q$ be the claimed observable and $Q_N$ its computed approximation. After inserting named intermediate quantities so that a telescoping triangle inequality applies, a conservative bound can separate distinct sources,

$$\lvert Q-Q_N\rvert \le \varepsilon_{\mathrm{EFT}} +\varepsilon_{\mathrm{replica}} +\varepsilon_{\mathrm{num}} +\varepsilon_{\mathrm{state}} +\varepsilon_{\mathrm{op}}.$$

The terms represent effective-field-theory truncation, saddle or continuation error, numerical error, state-preparation mismatch, and operational measurement error. Each term must bound its corresponding intermediate difference in the same observable norm; the displayed inequality does not follow merely by naming five uncertainties. The terms should not be collapsed into one adjustable tolerance. An exponentially small nonperturbative term can dominate an exact factorization question even when every perturbative coefficient is known; finite sampling can dominate a mathematically exact entropy formula.

At least one independent check is required: a solvable limit, conservation law, second algorithm, theorem bound, or adversarial input. Varying only the same solver tolerance tests implementation stability, not the underlying approximation. Production computation requires a maintained, reviewed executable artifact; a page here must still expose enough equations, inputs, and diagnostics for the scientific claim to be assessed without assuming that infrastructure.

Follow correlations through time

Endpoint purity and a Page curve underdetermine dynamics. A stronger test entangles the input $M$ with a reference $Q$ and tracks

$$\delta(t)= \frac12\left\lVert \rho_{Q B(t)}-\rho_Q\otimes\rho_{B(t)} \right\rVert_1,$$

where $B(t)$ is the remaining hole. Small $\delta$ is a decoupling statement; it implies the information can in principle be recovered from the complement under the theorem’s hypotheses. It does not give the circuit, collection efficiency, or physical time needed to implement recovery.

Useful time-resolved alternatives include mutual information with selected radiation bands, OTOCs in their valid growth window, response of a dressed operator, and the difference between competing generalized-entropy saddles. The time coordinate, clock, state, smoothing, and causal access must be stated. A late-time average cannot establish a pointwise recovery process.

Close the operational resource budget

A protocol is a tuple

$$\mathcal P=(n_{\mathrm{copy}},N_{\mathrm{gate}},M_{\mathrm{q}}, t_{\mathrm{run}},\Delta E,\eta_{\mathrm{col}},p_{\mathrm{err}}),$$

containing independent copies, gate count, quantum memory, runtime, injected energy, collection efficiency, and error probability. Entropy is a nonlinear state functional, so many protocols require multiple identical preparations. An astronomical black-hole evaporation does not automatically provide them. Copy-complexity bounds for entropy estimation make the access model explicit in Acharya et al. 2020, §§1.3, 4, 6–7.

The deadline belongs to the claimed experiment. A decoder followed by an interior comparison must fit within the available evaporation and causal window. A radiation-only measurement can in principle occur after complete evaporation if coherent storage and the observer’s lifetime permit it; the evaporation time is not its automatic cutoff. In either case, apparatus stress-energy must preserve the background used to derive the prediction. Harlow–Hayden complexity gives a possible separation between information-theoretic recovery and feasible decoding, but its asymptotic hardness assumptions must not be promoted from a small numerical circuit trend; see Harlow and Hayden 2013, §§2–4.

Design a discriminating failure test

Claim	Calculation	Failure criterion
Exact unitary recovery	Reference-system channel test over the declared input family	Nonzero uniform recovery error after all output sectors are included
Smooth horizon	Dressed infalling detector response and renormalized stress tensor	Order-one deviation inside the claimed low-curvature, low-frequency domain
Island reconstruction	Recovery of an island operator from the radiation code subspace	Error exceeds the declared nonperturbative bound or grows uncontrollably with code size
Single-theory replica interpretation	Fixed-theory multipoint observable including connected saddles	Failure to reconcile exact boundary factorization with the claimed bulk term
Feasible decoding	Explicit algorithm and full resource estimate	Runtime, memory, precision, collection, or backreaction exceeds the declared budget

The calculation should also include a positive control that the method can pass and a negative control that it can reject. Otherwise a null result may be an insensitive observable rather than evidence for the proposal. Generic tomography, detector engineering, and instrument-specific inference are treated in their dedicated subjects; this method retains only what is needed to test a black-hole recovery claim.

Results and open questions through 2026-08-19

Evidence or method	Assumptions	Controlled result	Limitation or unresolved step
Hayden–Preskill decoupling, Hayden and Preskill 2007, §3, Eqs. (3)–(7)	Old hole, random-unitary or design model, declared diary and radiation dimensions, early-radiation access, and trace-distance criterion	Gives a dimension- and norm-dependent recovery-existence statement	Does not supply a generic efficient decoder or gravitational implementation
Replica and island calculations, Almheiri et al. 2020, §§2–6	Specified low-dimensional bath model, state, replica saddles, contour, and continuation	Gives controlled entropy candidates and Page-like transitions in the stated approximation	Individual amplitudes, real-time transfer, contour uncertainty, and one-theory interpretation remain additional tests
Harlow–Hayden analysis, Harlow and Hayden 2013, §§2–4 and appendices	Circuit model, generic scrambling assumptions, access model, and asymptotic complexity conjectures	Motivates a large separation between information-theoretic recovery and generic decoding cost	Does not prove a universal physical lower bound for every structured black-hole decoder
Quantum-entropy estimation, Acharya et al. 2020, §§1.3, 4, and 6–7	Independent identical copies of a finite-dimensional unknown state, declared entropy order, accuracy, and success probability	Gives copy-complexity algorithms and bounds for von Neumann and Rényi entropies	Independent preparations and coherent control may be unavailable for one evaporation; copy cost is not gate or backreaction cost
Gravitational dressing, Donnelly and Giddings 2016, §§2–5	Perturbative gravity about flat spacetime, asymptotic anchor, gauge and dressing choice, and fixed order in $G$	Constructs gauge-invariant dressed operators and leading nonlocal commutators	A complete nonperturbative measurement protocol remains open

Competing positions may agree that an observable exists while disagreeing on whether it can be calculated or implemented. The conclusions could change if a structured decoder beats the assumed hardness with a closed resource budget, if a fixed-theory real-time calculation resolves replica ambiguity, or if a new lower bound joins copy, gate, time, energy, and backreaction costs in one operational theorem.

Check your understanding

A simulation’s Page time changes by less than $10^{-8}$ under three grid refinements. What has and has not been validated?

Solution

The test supports numerical convergence of that reported Page time for the implemented equations and branch, provided the refinements resolve all roots and the residual also decreases. It does not validate the effective action, state preparation, replica continuation, saddle set, or interpretation as a one-theory recovery mechanism. Those errors occupy different terms in the budget and need analytic limits, alternative saddles, or independent physical checks. A stable wrong model can converge perfectly.

References

• Acharya, Jayadev, Ibrahim Issa, Nirmal V. Shende, and Aaron B. Wagner, “Estimating Quantum Entropy,” IEEE Journal on Selected Areas in Information Theory 1 (2020) 454–468. arXiv
• Almheiri, Ahmed, Thomas Hartman, Juan Maldacena, Edgar Shaghoulian, and Amirhossein Tajdini, “Replica Wormholes and the Entropy of Hawking Radiation,” Journal of High Energy Physics 05 (2020) 013. DOI. Open manuscript
• Donnelly, William and Steven B. Giddings, “Observables, Gravitational Dressing, and Obstructions to Locality and Subsystems,” Physical Review D 94 (2016) 104038. arXiv
• Harlow, Daniel and Patrick Hayden, “Quantum Computation versus Firewalls,” Journal of High Energy Physics 06 (2013) 085. arXiv
• Hayden, Patrick and John Preskill, “Black Holes as Mirrors: Quantum Information in Random Subsystems,” Journal of High Energy Physics 09 (2007) 120. arXiv