Final-State Boundary Conditions

A final-state proposal replaces unconstrained evolution at the spacelike singularity by a selected entangled boundary condition. Contracting infalling matter with that boundary can teleport its state into the outgoing Hawking radiation. In an exactly tuned finite-dimensional model the resulting exterior map is unitary. Generic interior interactions spoil the tuning, however, and postselection can make probabilities state dependent, evolution nonlinear, or causal ordering operationally problematic. The proposal is therefore a sharp algebraic possibility, not established black-hole dynamics.

Helpful background. Semiclassical Gravity supplies the horizon and singularity regime; Quantum Information supplies teleportation, channels, and no-signaling; Quantum Fields in Curved Spacetime supplies the Hawking-pair idealization; Black-Hole Information supplies the unitarity target; Quantum Gravity supplies possible boundary-law origins; and Causality and Global Structure supplies the meaning of a spacelike final boundary.

Postselected teleportation in one equation

Use three $d$-dimensional Hilbert spaces: collapsing matter $M$, infalling Hawking partners $I$, and outgoing radiation $O$. The idealized near-horizon state is

$$\lvert\Phi\rangle_{IO} =\frac{1}{\sqrt d}\sum_{j=1}^{d} \lvert j\rangle_I\lvert j\rangle_O.$$

Impose at the singular boundary the maximally entangled bra $\langle\Phi\rvert_{MI}$. Direct contraction gives

$${}_{MI}\!\langle\Phi\rvert \bigl(\lvert\psi\rangle_M\otimes\lvert\Phi\rangle_{IO}\bigr) =\frac{1}{d}\lvert\psi\rangle_O.$$

The unnormalized success probability is $1/d^2$, independent of the input; conditioning therefore recovers the same normalized state in $O$. A unitary $S$ inserted in the final bra makes the output $S\lvert\psi\rangle$, up to a basis-dependent transpose convention. This is the postselected-teleportation core of Horowitz and Maldacena’s construction Horowitz and Maldacena 2004, §§ 2–3.

Nothing in this algebra explains why one final outcome occurs with certainty. Ordinary quantum mechanics would assign it probability $1/d^2$ and retain the other outcomes. The proposal instead elevates the selected bra to a boundary law. Its physical burden is to derive that law, including its entanglement, normalization, and relation to interior interactions.

The exact condition for a linear exterior map

Let $U$ describe all interactions among $M$ and $I$ before the final projection, and let the chosen bra be $\langle F\rvert$. Their contraction defines a linear transfer operator $T:M\to O$ before normalization. On density operators, the conditioned rule is

$$\rho\longmapsto \frac{T\rho T^\dagger} {\operatorname{Tr}(T\rho T^\dagger)}.$$

This rule is a normalized linear quantum channel for every input precisely when

$$T^\dagger T=c\,\mathbf 1, \qquad c\gt0.$$

Then the denominator equals the constant $c$ and $T/\sqrt c$ is an isometry, or a unitary when input and output dimensions agree. If $T^\dagger T$ has unequal eigenvalues, different input states have different postselection weights. Dividing by that weight makes the conditioned state map nonlinear. This singular-value test is more informative than asking whether $U$ itself is unitary: a perfectly unitary interior interaction can still produce a nonisometric partial contraction.

Interior interactions are the decisive adverse check

The Hawking partners and collapsing matter occupy the same interior and should interact gravitationally and through any other permitted couplings. Gottesman and Preskill show that generic entangling interactions modify the effective transfer operator and can degrade fidelity or destroy exact unitarity unless the final state compensates for them Gottesman and Preskill 2004, complete analysis, pp. 1–8. Compensation is algebraically possible: choose the final bra with detailed knowledge of $U$ so that the contraction remains proportional to a unitary. But that is a correlation between ultraviolet boundary data and the entire prior interior dynamics, not an automatic consequence of maximal entanglement.

Lloyd and Preskill examine a smaller input sector $M_1$ describing rapid collapse, while the complementary input $M_2$ is held fixed. For Haar-generic final data, their average deviation of the reference marginal from maximal mixing is bounded by $\sqrt{d_{M_1}/d_I}$, which is small when $d_{M_1}\ll d_I$ Lloyd and Preskill 2014, § 4.2, equation (3). The dimension hierarchy is essential: for $d_{M_1}=d_I$ the bound is order one. It is not a statement that a generic transfer matrix on the full equal-dimensional space becomes unitary merely because the black-hole entropy is large. Small average error is not the same as a completely positive trace-preserving map on every input, nor does it bound an adversarial code subspace without a specified norm.

Linearity, signaling, and causal interpretation

Postselection changes more than a singular endpoint. If probabilities depend on a later boundary condition, an intervention behind the horizon can alter the conditioned statistics assigned to radiation that is already outside. Lloyd and Preskill show how measurements in old-black-hole settings can create such chronology concerns, while decoding complexity may make some effects inaccessible Lloyd and Preskill 2014, §§ 5–6. Operational inaccessibility is not restoration of ordinary linear quantum mechanics; it is a separate resource claim.

Three questions must therefore be answered independently. Does the asymptotic S-matrix preserve inner products? Are intermediate probabilities governed by a linear normalized rule for all allowed interventions? Can one assign commuting algebras to causally separated operations without contradiction? A model may pass the first and fail the latter two. Declaring only asymptotic unitarity hides that distinction.

Results and open questions through 2026-08-19

This discussion reflects results available through 2026-08-19. The permissive position regards the teleportation identity as a consistent route to a unitary exterior S-matrix if quantum gravity supplies the special boundary law. The conservative position rejects the proposal until the law and its interaction compensation arise from a microscopic theory without nonlinear signaling. An intermediate position accepts approximate unitarity on a restricted state family while treating postselected measurements as outside the effective domain. These competing positions make different claims about formulation and regime.

Evidence or test	Result	Scope or limitation
Maximally entangled contraction in Horowitz and Maldacena 2004, §§ 2–3	The ideal transfer operator is proportional to a unitary, giving postselected teleportation to the radiation.	Finite-dimensional idealization with a stipulated final bra and maximally entangled Hawking state.
Interaction analysis in Gottesman and Preskill 2004, complete analysis, pp. 1–8	Generic entangling dynamics makes the transfer operator nonunitary unless the final data compensate.	Diagnoses sensitivity; it does not prove that a microscopic theory cannot enforce compensation.
Generic-final-state estimates in Lloyd and Preskill 2014, § 4.2, equation (3)	Reference-marginal deviation is bounded on average by $\sqrt{d_{M_1}/d_I}$ for a restricted rapid-collapse input.	Requires $d_{M_1}\ll d_I$ and fixed complementary input; does not establish full-domain channel linearity or a worst-case bound.
Measurement analysis in Lloyd and Preskill 2014, §§ 5–6	Interior interventions can entail cloning, polygamous entanglement, or chronology-sensitive statistics.	Detectability may require decoding resources; computational protection is conjectural and operational.

The finite-dimensional identity is controlled. Its elevation to a final law, its robustness under realistic interactions, and its gravitational realization remain unresolved. A special choice can always pass the algebraic test; the adverse check is whether a small physically allowed change in $U$ makes the singular values of $T$ unequal.

Requirements for a gravitational realization

A complete proposal should identify the physical Hilbert spaces and constraints rather than assume a tensor factorization at the singularity. It should derive the final functional from an action, contour, or microscopic state; show that the induced map is linear, normalized, and stable over a declared code domain; and account for fields with unequal effective dimensions. It must also specify how local interior interactions are represented and why their effects are cancelled without future-input fine-tuning.

Finally, it should calculate observable consequences of intermediate postselection and show whether causal signaling is forbidden, exponentially suppressed in a named norm, or merely hard to decode. The conclusions could change after a microscopic gravitational construction derives the final-state functional, or a theorem proves stability of $T^\dagger T=c\mathbf 1$ under a physical interaction class, or an explicit low-resource protocol exposes state-dependent probabilities or backward signaling in the proposed regime.

Exercises

Let $T$ act on a qubit as $T\lvert0\rangle=\tfrac12\lvert0\rangle$ and $T\lvert1\rangle=\lvert1\rangle$. Compare the postselection probabilities for the basis states and explain why the normalized conditioned rule is not a linear quantum channel.

Solution

The probabilities for this postselected outcome are $1/4$ for $\lvert0\rangle$ and $1$ for $\lvert1\rangle$, because $T^\dagger T$ has eigenvalues $1/4$ and $1$ rather than being proportional to the identity. Normalization therefore depends on the input. For example, the output of a classical mixture is not obtained by mixing the separately normalized outputs with the original weights. The equal mixture $I/2$ is sent to $\operatorname{diag}(1/5,4/5)$, whereas the equal mixture of the two normalized basis outputs remains $I/2$. The underlying contraction is linear, but conditioning on a state-dependent success probability produces a nonlinear normalized rule.

References

• Gottesman, Daniel, and John Preskill. 2004. “Comment on ‘The Black Hole Final State.’” Journal of High Energy Physics 03, 026. arXiv:hep-th/0311269.
• Horowitz, Gary T., and Juan Maldacena. 2004. “The Black Hole Final State.” Journal of High Energy Physics 02, 008. arXiv:hep-th/0310281.
• Lloyd, Seth, and John Preskill. 2014. “Unitarity of Black Hole Evaporation in Final-State Projection Models.” Journal of High Energy Physics 08, 126. DOI. arXiv:1308.4209.