Non-Isometric Encoding in Evaporation Models

A non-isometric map can compress an oversized semiclassical interior description into a smaller fundamental Hilbert space and can preserve selected simple observables for typical states. It is not an ordinary quantum encoding: norms and inner products are not preserved globally, normalized postselection is nonlinear, and recovery claims require a specified state set, error norm, and complexity restriction.

Required background. Non-Isometric Encoding Proposals supplies the proposal class. Approximate Recovery and Information–Disturbance supplies the operational test.

Helpful background. Hayden–Preskill Recovery and Decoding Tasks gives the comparison with an isometric channel. Non-Unique JT Matrix-Integral Completion: Fixed-Theory, Disorder-Average, and Ensemble Distinctions constrains averaged claims.

What non-isometry changes

Let

$$V:\mathcal H_{\mathrm{eff}}\longrightarrow\mathcal H_{\mathrm{fund}}, \qquad G=V^\dagger V.$$

An isometry has $G=1$. If $\dim\mathcal H_{\mathrm{eff}}>\dim\mathcal H_{\mathrm{fund}}$, $V$ has a kernel and cannot preserve all inner products. The normalized pure-state rule

$$\lvert\psi\rangle\longmapsto \frac{V\lvert\psi\rangle} {\sqrt{\langle\psi|G|\psi\rangle}}$$

is state-dependent through the denominator and therefore is not a linear quantum channel. A physical implementation must identify a postselection event, its probability, and the full trace-preserving process in which it is embedded.

Application: a random rectangular map

Take $V$ to be an $n\times k$ complex Gaussian matrix with

$$\mathbb E\,V_{ai}^*V_{aj}=\frac{1}{n}\delta_{ij}, \qquad k>n.$$

Then

$$\mathbb E\,(V^\dagger V)=1_k,$$

so for any fixed input $\lvert\psi\rangle$,

$$\mathbb E\,\lVert V\psi\rVert^2=1$$

and the norm concentrates for large $n$. Low-complexity correlators evaluated on a fixed small set of states can therefore look approximately isometric after averaging.

Globally, however, $\operatorname{rank}V\le n<k$, so there exists a normalized $\lvert\chi\rangle$ with $V\lvert\chi\rangle=0$. Hence

$$\lVert V^\dagger V-1\rVert_{\mathrm{op}}\ge1.$$

Average preservation on typical states gives no uniform operator-norm guarantee and says nothing for the atypical kernel. This is the essential adversarial input.

Akers, Engelhardt, Harlow, Penington, and Vardhan propose that computational complexity can protect non-isometric interior codes: feasible observers may be unable to prepare or detect the exceptional states Akers et al. 2024. The conclusion is complexity-bounded and model-dependent, not exact isometry.

Correlators versus inner products

Suppose an effective operator $O$ is represented by a fundamental operator $\widetilde O$ so that averaged simple correlators obey

$$\mathbb E_V\langle\psi| V^\dagger\widetilde O V|\psi\rangle \simeq \langle\psi|O|\psi\rangle .$$

This does not imply $V^\dagger\widetilde O V\simeq O$ in operator norm, nor does it preserve off-diagonal phases for arbitrary superpositions. A valid release statement must name the ensemble, allowed state family, observable complexity, error probability, and norm.

Atypical-input and fixed-map tests

Freeze one realization of $V$ rather than averaging, search for small singular values, and prepare states aligned with their singular vectors. Test norm, inner product, and reference-system entanglement. If a postselection probability becomes state-dependent or exponentially small, include it. Then enlarge the operator family beyond low-complexity probes. Failure under these tests marks the intended boundary of the model rather than a paradox.

Scope and handoff

Non-isometric models can explain how an effective interior description exceeds fundamental dimension without granting exact independent degrees of freedom. They do not by themselves prove unitary evaporation, factorization, or an endpoint. Those demands return on Microscopic Unitarity versus Semiclassical Entropy Calculations.

The chapter overview contains the structure diagram and validity and failure diagram. They are embedded there once so that their shared chapter-level context is not repeated on every article.

For the chapter-wide comparison of assumptions, counterevidence, falsifiers, and claim ceilings, see the claim-domain table.

References

• Akers, C., N. Engelhardt, D. Harlow, G. Penington, and S. Vardhan. “The Black Hole Interior from Non-Isometric Codes and Complexity.” Journal of High Energy Physics 2024, 6 (2024): 155. DOI.