Universal quantum resource distillation via composite generalised quantum Stein's lemma

• Quantum resource manipulation converts many copies of one resourceful object into as many copies as possible of another, while hypothesis testing discriminates between quantum-state assumptions.Both tasks are benchmarked by optimal asymptotic rates, with quantum entropies providing a unifying quantitative description.Source: Quantum resources and hypothesis testing · p. 1
• Quantum Stein’s lemma characterises asymptotic discrimination between ρ^⊗n and σ^⊗n using the quantum relative entropy D(ρ∥σ), introduced by Umegaki.The formal treatment was initiated by Hiai and Petz [1].Source: Quantum resources and hypothesis testing · p. 1
• Entanglement distillation converts weakly entangled states into pure entanglement, with pure-state rates given by entropy of entanglement and mixed-state measures based on relative entropy of entanglement [6] [7].These entropic connections extend beyond distillation to broader entanglement manipulation problems.Source: Quantum resources and hypothesis testing · p. 1
• Under asymptotically non-entangling operations, the regularised relative entropy of entanglement D∞(ρ∥SEP) uniquely determines transformation rates, including distillable entanglement into a maximally entangled state.For general resource theories satisfying the Brandão–Plenio axioms, the corresponding D∞(·∥S) similarly determines asymptotic transformation rates.Source: Quantum resources and hypothesis testing · pp. 1–2
• The resource-manipulation rate depends on a hypothesis test distinguishing ρ^⊗n from free states, keeping type I error below ε while minimising the type II error β_ε.The generalised quantum Stein’s lemma identifies the exponential decay rate with the regularised relative entropy, after earlier proof gaps were resolved by complete solutions.Source: Quantum resources and hypothesis testing · p. 2

Summary of results

• The framework addresses the practical limitation that existing optimal tests and resource transformations require exact knowledge of the source state.Tomography-based estimation can sacrifice copies and reduce the transformation rate, motivating a state-independent construction.Source: Summary of results · pp. 2–3
• Universal protocols can distill any target state σ from any input state ρ at an optimal rate without prior information about ρ.The protocol depends on the target state, while the input state remains unknown.Source: Summary of results · p. 2
• For entanglement, the result yields a universal non-entangling protocol converting any noisy entangled state into singlets at a rate governed by the regularised relative entropy of entanglement.The paper therefore extends universal resource manipulation beyond previously studied special cases and thermodynamic resource theories.Source: Summary of results · p. 2
• The composite testing problem uses a null hypothesis of i.i.d. states ρ^⊗n with ρ∈R1 and an alternative consisting of free states S_n.The base set R1 is treated as completely arbitrary, while S satisfies the Brandão–Plenio structural assumptions.Source: Summary of results · p. 3
• Universal tests keep type I error below ε for every n while the type II error decays exponentially to zero.This supplies the hypothesis-testing ingredient needed for universal resource transformations.Source: Summary of results · p. 3
• The construction refines the blurring technique from so it can handle the composite variant of the generalized quantum Stein’s lemma.The extension provides a route beyond methods that cannot directly produce universal resource transformations for general mixed-state entanglement.Source: Summary of results · pp. 2–3

Problem

• Composite hypothesis testing receives an unknown n-copy state from either R_n or S_n and seeks a measurement that distinguishes the two alternatives.The null and alternative are sets of states on H^⊗n, and the measurement may act jointly on the full n-copy system.Source: QUANTUM HYPOTHESIS TESTING · p. 3

Asymptotic performance

• The Stein exponent characterizes the asymptotic decay of the minimum type-II error at a fixed type-I error threshold.The framework also defines a strong converse Stein exponent for the complementary asymptotic regime.Source: QUANTUM HYPOTHESIS TESTING · p. 3

Setting

• The paper studies a composite i.i.d. null, R_n={ρ^⊗n:ρ∈R_1}, while the alternative S_n may have a structure far beyond i.i.d.This setting is motivated by resource theories, where S_n represents free states and R_1 can model a neighborhood of an unknown state.Source: QUANTUM HYPOTHESIS TESTING · p. 3

Prior results and obstacle

• For composite i.i.d. extensions with simple alternatives, the expected Stein exponent is the worst-case exponent over the states in the composite null.Those methods do not apply when the alternative is fully composite, because even a simple i.i.d. null can face substantial difficulty against such alternatives [10].Source: QUANTUM HYPOTHESIS TESTING · p. 4

Theorem

• The main theorem establishes the conjectured composite i.i.d. generalised quantum Stein’s lemma for alternatives satisfying convexity, full-rank feasibility, subsystem tracing, tensor products, and permutation closure.The result applies to any single-copy base set R_1 on a finite-dimensional Hilbert space.Source: MAIN RESULT #1: COMPOSITE I.I.D.… · p. 4

Problem

• The universal-testing problem differs from fixed-state testing because the null is a generic mixture of i.i.d. states, requiring one test to work for every ρ∈R_1.This is the operational distinction between a test tailored to one known state and a universal test over the whole base set.Source: MAIN RESULT #1: COMPOSITE I.I.D.… · p. 4

Reduction

• A quasi-concavity property of hypothesis-testing relative entropy removes convex combinations by replacing them with ρ^(n)^⊗n, at the cost of asymptotically negligible error terms.The argument uses one-shot relative entropies and weak/strong converse duality with smooth max-relative entropy.Source: MAIN RESULT #1: COMPOSITE I.I.D.… · p. 4

Technical obstacle

• Existing generalised Stein formulations are not robust when the i.i.d. base state fluctuates with n, so the proof requires a strengthened lemma.The fluctuating sequence may approach a limiting state, but the state used at each n is not fixed.Source: MAIN RESULT #1: COMPOSITE I.I.D.… · p. 4

Proof strategy

• The refined blurring method adds controlled noise to tame fluctuating sources, while dynamical bosonic lifting embeds the n-dependent sequence into an infinite-dimensional bosonic space.The n-dependent embedding maps the reference state to a vacuum-like representation and, together with the refined analysis, proves Theorem 1.Source: MAIN RESULT #1: COMPOSITE I.I.D.… · pp. 4–5

Blurring construction

• The blurring construction uses a small noisy extension, random permutation, and partial trace to return the system to n subsystems.The parameter δ controls the number of added noisy subsystems.Source: MAIN RESULT #1: COMPOSITE I.I.D.… · p. 4

Universal transformations

• Universal protocols attain the optimal resource-transformation rate D∞(ρ∥S) for any unknown input state ρ, using asymptotically resource non-generating operations.The result assumes the Brandão–Plenio axioms and a compatible tomographically complete measurement set.Source: MAIN RESULT #2: UNIVERSAL… · p. 5

Optimality

• The universal rate equals the best rate achievable when the input state is fully known, despite the protocol being designed without that knowledge.This closes the gap between universal and state-tailored resource manipulation rates.Source: MAIN RESULT #2: UNIVERSAL… · p. 5

Entanglement distillation

• For entanglement, the universal protocol distills singlets from any noisy entangled state at the regularised relative entropy of entanglement rate.The construction uses the entanglement-distillation protocol based on non-entangling operations and the generalised quantum Stein’s lemma.Source: MAIN RESULT #2: UNIVERSAL… · p. 5

Protocol construction

• The construction estimates the unknown state by sacrificing a fraction μ of the copies, then applies a resource-distillation protocol to the remaining (1−μ)n copies.Tomography succeeds with probability tending to one, and μ can be chosen arbitrarily small in the asymptotic analysis.Source: MAIN RESULT #2: UNIVERSAL… · p. 6

Resource non-generation

• Compatible measurements preserve freeness after conditioning, so combining tomography with the Brandão–Plenio protocol remains asymptotically resource non-generating.The proof uses convexity of generalized robustness and compatibility of the measurement with the free-state set.Source: MAIN RESULT #2: UNIVERSAL… · p. 6

Appendix roadmap

• The appendix organizes the proof around technical lemmas, a black-box reduction, and an extension of blurring techniques for the composite i.i.d. Stein problem.Appendix A supplies one-shot results, Appendix B performs the reduction, and Appendix C contains the main proof strategy.Source: — Appendix — · p. 9

Proof ingredients

• The proof controls smoothing with purified distance and restricts the optimization, without loss of generality, to normalized states.The derivation uses Cauchy–Schwarz, convexity, a reversed Cauchy–Schwarz inequality, and the triangle inequality for max-relative entropy.Source: Appendix A · pp. 9–11

Fidelity decomposition

• Lemma 3 establishes a decomposition property for an ensemble state, using strong concavity of fidelity for one direction and Uhlmann’s theorem for the converse.The proof constructs purifications on a common ancillary system and uses data processing under dephasing and partial trace.Source: Appendix A · p. 9

Approximate quasi-concavity

• Approximate quasi-concavity bounds the smooth max-relative entropy of a convex combination using the worst-case ensemble element plus an additive penalty.The bound applies to finite ensembles and smoothing parameters satisfying ε+μ≤1.Source: Appendix A · p. 9

Hypothesis-testing consequence

• Corollary 5 transfers the one-shot bound from smooth max-relative entropy to hypothesis-testing relative entropy through weak/strong converse duality inequalities.The result holds for finite ensembles and parameters 0<δ<ε<1, after optimizing over an auxiliary smoothing parameter.Source: Appendix A · pp. 11–12

Symmetric setting

• Proposition 6 applies the reduction to permutation-symmetric state sets whose convex hull is generated by a restricted symmetric subset.The proposition assumes finite-dimensional systems and permutation invariance under tensor-factor permutations.Source: Appendix B · p. 13

Finite discretisation

• Carathéodory’s theorem represents each state in the relevant convex hull as a finite mixture of states from the restricted set.Permutation symmetry and the finite-dimensional affine-space structure yield a polynomial bound on the number of mixture elements.Source: Appendix B · pp. 13–14

Black-box reduction

• The reduction applies the one-shot Corollary 5 to convex combinations of i.i.d. states and incurs only a logarithmic finite-size correction.The correction is controlled using the polynomial bound on the decomposition size.Source: Appendix B · p. 14

Asymptotic exponents

• After dividing by n and taking limit infima or suprema, the finite-size correction vanishes and yields the desired direct and strong-converse Stein-exponent bounds.Monotonicity in the type-I error parameter completes the limiting argument, together with the converse inequalities.Source: Appendix B · p. 14

Reduction

• The first hypothesis-testing argument reduces the optimisation to i.i.d. states ρ^⊗n with ρ∈R1, independently of the alternative hypothesis at that stage.The reduction is achieved through the quantum blurring lemma and unitary-invariance arguments, including a reduction from mixed to pure states.Source: Appendix C · pp. 15–17

Quantum blurring

• A strengthened asymptotic quantum blurring lemma handles sequences of varying base states by exploiting permutation symmetry and extracting an infinite subsequence.The construction uses permutationally symmetric states, compactness, and a refined blurring map compatible with the Brandão–Plenio axioms.Source: Appendix C · pp. 15, 18

Theorem 1

• Theorem 1 establishes the composite generalised quantum Stein result for any alternative-hypothesis sequence satisfying all five Brandão–Plenio axioms.The proof uses convexity and the strong-converse form of the generalised quantum Stein’s lemma to connect the relevant asymptotic quantities.Source: Appendix C · p. 17

Continuity and limits

• The proof resolves the mismatch between ρ_n^⊗n and ρ^⊗n through a stronger continuity property of the relative entropy distance, rather than ordinary asymptotic continuity.This continuity argument permits passage to the limit and replacement of the topological closure of R1 by R1 itself.Source: Appendix C · pp. 19–20

Consequences

• The resulting contradiction completes the proof, and Corollary 8 supplies the consequence used for universal quantum resource transformations under the Brandão–Plenio axioms.The corollary applies to sequences of free-state sets satisfying the five axioms and to trace-norm neighbourhoods of a state.Source: Appendix C · p. 20