Source-linked AI summary
Universal recovery maps and approximate sufficiency of quantum relative entropy
Marius Junge, Renato Renner, David Sutter, Mark M. Wilde, Andreas Winter
TL;DR
The paper addresses how to strengthen quantum relative-entropy data processing with a recovery-based remainder that characterizes approximate sufficiency. It constructs an explicit recovery map depending only on σ and N, proves its universal recovery guarantee, and connects the result to approximate quantum error correction. The construction is presented for separable Hilbert spaces and states supported on σ.
Problem
Existing relative-entropy recovery results did not establish a recovery map independent of ρ, leaving universality unresolved for approximate sufficiency.
Method
The paper constructs an explicit recovery map Rσ,N using finite-dimensional results and a limiting argument to extend the theorem to separable Hilbert spaces.
Results
The universal map recovers σ exactly and satisfies the strengthened recovery inequality for every density operator ρ with supp(ρ) ⊆ supp(σ).
Takeaways & Limitations
The result provides an information-theoretic characterization of approximate quantum error correction and explicit remainder terms for several entropy inequalities.
Takeaways & Limitations
The theorem assumes the inverse σ^-1 is taken on σ’s support and applies to states ρ supported within σ.
Abstract
from arXiv · showhide
The data processing inequality states that the quantum relative entropy between two states $ρ$ and $σ$ can never increase by applying the same quantum channel $\mathcal{N}$ to both states. This inequality can be strengthened with a remainder term in the form of a distance between $ρ$ and the closest recovered state $(\mathcal{R} \circ \mathcal{N})(ρ)$, where $\mathcal{R}$ is a recovery map with the property that $σ= (\mathcal{R} \circ \mathcal{N})(σ)$. We show the existence of an explicit recovery map that is universal in the sense that it depends only on $σ$ and the quantum channel $\mathcal{N}$ to be reversed. This result gives an alternate, information-theoretic characterization of the conditions for approximate quantum error correction.
1 Introduction
The paper frames approximate sufficiency through recoverability: strengthened data-processing inequalities should quantify how well states can be recovered after a channel. It establishes an explicit recovery map depending only on σ and the channel, not on ρ.
- Motivation: Data processing makes quantum relative entropy non-increasing under trace-preserving completely positive maps.Equality for all states characterizes sufficiency when a recovery map reverses the channel on both states.
- Contribution: The Petz recovery map is completely positive and trace non-increasing, while the paper’s result provides a universal construction with stronger scope.The finite-dimensional Petz map is unique on the support of N(σ).
- Motivation: Approximate sufficiency asks whether a channel admits one recovery map whose trace-distance error is at most ε for every state in a specified set.The recovery condition is 1/2 ∥ρ − (Rε ◦ N)(ρ)∥1 ≤ ε.
- Motivation: A strengthened data-processing inequality could add a non-negative recovery-distance remainder and thereby characterize approximate sufficient statistics information-theoretically.This is the central question motivating the paper.
- Prior work: Prior relative-entropy recovery results supplied rotated Petz-type maps but did not establish universality, meaning independence from ρ.The cited approaches produced recovery maps satisfying the relevant inequalities without proving that one map works independently of the input state.
- Contribution: The paper proves an explicit universal recovery map Rσ,N for any non-negative σ and channel N, with recovery valid for all ρ supported on σ.Universality implies (Rσ,N ◦ N)(σ)=σ.
2 Main results
The main result constructs a universal recovery map from σ and N, expressed as a mixture of rotated Petz maps, and establishes its recovery and structural properties. The result also characterizes equality in data processing through exact recoverability.
- Universal recovery map: Theorem 2.1 constructs a recovery map Rσ,N for σ, ρ supported on σ, and any quantum channel N.The map is defined using the Petz recovery map and rotated recovery maps.
- Equality and recovery: If data processing preserves relative entropy exactly, continuity implies exact recovery by the rotated Petz maps.The relevant condition is D(ρ∥σ) = D(N(ρ)∥N(σ)).
- Universal recovery map: The recovery map is universal because it depends on σ and N but not on ρ.This universality is stated as a functoriality property of Rσ,N.
- Recovery properties: Rσ,N perfectly reconstructs σ from N(σ).Every rotated Petz map in the construction recovers σ, so their convex combination does as well.
- Recovery properties: When N is the identity, Rσ,N projects onto the support of σ and equals the identity channel when σ is faithful.This is the normalization property of the recovery map.
- Recovery properties: The recovery map is stable under adjoining a faithful reference state and obeys parallel and serial composition rules.These are additional functoriality properties established for Rσ,N.
3 Proof of Theorem 2.1
The proof first establishes the recovery inequality in finite dimensions using a Rényi relative-entropy difference and Hirschman’s three-line theorem, then extends it to separable Hilbert spaces by approximation. Convergence of projected states, channels, Petz maps, and fidelities completes the infinite-dimensional argument.
- Finite-dimensional proof: The finite-dimensional proof combines a Rényi generalization of relative entropy difference with Hirschman’s improvement of the Hadamard three-line theorem.These are identified as the two main ingredients of the proof.
- Finite-dimensional proof: The operator-valued function G is placed in the hypotheses of the three-line lemma, yielding bounds that imply the target inequality.The argument applies norm bounds for G and then takes limits in the Rényi parameter.
- Extension to separable spaces: The proof extends from finite-dimensional to separable Hilbert spaces through finite-rank projections and a limiting argument.The extension uses projected sequences that converge to the identity operators.
- Extension to separable spaces: Lower semicontinuity of relative entropy allows the projected inequalities to pass to the infinite-dimensional limit.The limiting steps conclude the desired assertion for separable Hilbert spaces.
- Convergence of recovery maps: The finite-dimensional Petz recovery maps converge to the infinite-dimensional Petz map under the projection scheme.This convergence is established through Lemma 3.6 and related operator estimates.
- Convergence of recovery maps: For every t, the rotated recovery maps also converge, and continuity of fidelity transfers the finite-dimensional result to the limit.Serial concatenation preserves the relevant weak convergence, while fidelity is continuous in its inputs.
4 Universal and explicit refinements of other entropy inequalities
The universal recovery-map result yields explicit remainder terms for strong subadditivity, conditional-entropy concavity, and joint convexity of relative entropy. These applications preserve a recovery map that depends only on the relevant marginal or reference operator.
- Strong subadditivity: Theorem 2.1 yields a strengthened strong-subadditivity result with a universal recovery map.The recovery map is obtained by applying the theorem to the partial-trace channel.
- Concavity of conditional entropy: The strong-subadditivity corollary gives a universal remainder term for concavity of conditional entropy.Earlier remainder terms cited for this inequality were either conjectured or neither universal nor explicit.
- Strong subadditivity: The resulting recovery map for strong subadditivity is explicit and depends only on ρAB.
- Joint convexity of relative entropy: Theorem 2.1 also implies a universal remainder term for joint convexity of relative entropy.The paper contrasts this with a prior result that was neither universal nor explicit.
5 Approximate quantum error correction
The paper extends relative-entropy characterizations of perfect quantum error correction to the approximate setting. Approximate recovery is possible exactly when distinguishability from the codespace projector decreases only slightly under the noisy channel.
- Perfect error correction: Perfect error correction is equivalent to preserving D(ρ∥Π) for every codespace state under the channel.This is also equivalent to recovery by the Petz map for every state in the codespace.
- Approximate error correction: The paper asks whether this exact relative-entropy characterization has a robust analogue for approximate error correction.The question is motivated by prior information-theoretic conditions for perfect error correction and existing work on approximate correction.
- Approximate error correction: Theorem 2.1 establishes necessary and sufficient information-theoretic conditions for approximate quantum error correction.The condition is expressed through how much pairwise distinguishability between ρ and Π decreases under N.
- Approximate error correction: If the relative-entropy condition in Corollary 5.1 holds for every codespace state, then every state can be approximately recovered.The corollary also gives a converse: approximate recoverability implies the corresponding condition with parameter ε.
- Proof ingredients: The proof uses monotonicity of relative entropy under the recovery map and the Fannes–Audenaert inequality to relate entropy loss to recovery distance.It also uses that the recovery map fixes Π and that the relevant states are supported on Π.
6 Conclusion
The work constructs an explicit recovery map that depends only on σ and the channel N. Its universality ensures that σ is recovered exactly and makes the construction independent of the input state ρ.
- Conclusion: For every non-negative operator σ and channel N, the paper constructs an explicit universal recovery map Rσ,N.The map satisfies the stated recovery relation for every density operator supported on σ.
- Conclusion: Universality implies that Rσ,N∘N fixes σ, while the map remains independent of ρ.The authors contrast this explicit, state-independent construction with prior work.
A Proof of Lemma 3.2
The proof of Lemma 3.2 applies a three-line theorem to a holomorphic operator-valued interpolation constructed from the singular value decomposition of X. Boundary norm identities and Hölder’s inequality then yield the target bound.
- Analytic tool: Hirschman’s strengthening of the Hadamard three-line theorem is recalled as the analytic tool underlying the argument.The proof later invokes the resulting lemma to bound the interpolation expression.
- Interpolation construction: For fixed θ, the proof introduces Hölder conjugates qθ, q0, and q1 and chooses an operator X with a suitable decomposition.The singular value decomposition is written as X = UD^(1/qθ)V, with tr(D) = 1.
- Interpolation construction: The construction defines X(z) so that it is holomorphic in the strip, continuous on its boundary, and satisfies X(θ) = X.These properties make X(z) admissible for applying Lemma A.1.
- Final estimate: Applying Hölder’s inequality together with the boundary identities ∥X(it)∥q0 = 1 and ∥X(1 + it)∥q1 = 1 produces the required estimate.The resulting expression is then bounded above using the stated inequalities to obtain (27).