{"id":"cedcd096-c41c-49bd-9364-76a595356c79","arxiv_id":"2608.12135","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"For a rank-three qutrit-qutrit subspace, the exact PPT entanglement distillation rate is strictly below the min-Rains relative entropy, so the additive min-Rains bound is provably not tight.","lead":"Exact entanglement distillation from noisy quantum states is not always governed by the popular min-Rains relative entropy: the paper constructs a three-dimensional example where the formula strictly overestimates the achievable rate. This settles a question open since 2016 and rules out the min-Rains bound as a closed-form answer for zero-error distillation under PPT operations.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; the central separation proof is sound and the exact arithmetic checks out.","rationale":"The paper resolves an open problem with an explicit example. I re-derived the key inequalities: Lemma 2.1 is correct; Theorem 2.2's Hölder step is valid; the exact rational margins in Theorem 3.1 are narrow but positive. The decomposition typo in K is real but cosmetic: the printed b2 makes the two-term expansion not equal K, yet the asserted singular values are those of the corrected expansion, so the bound stands. The only assumption brought from outside is the Wang-Duan SDP characterization, which is a standard published result and not in doubt. The verdict ACCEPT is appropriate; no change needed.","tokens_in":12129,"tokens_out":28514,"duration_ms":239939,"concrete_test":"Use a CAS (e.g., sympy) to recompute the trace norm of the matrix K in Eq. (3.21) directly from its entries, without relying on the rank-one decomposition; confirm the two nonzero singular values are (2 beta^2 + z^2)^{1/2} and (2 alpha^2 + z^2)^{1/2} with the stated rational bounds, and verify the product inequality (391/250)*(6393/10000) < 1. Also check that the corrected b2 reproduces K.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central argument is internally consistent and the exact rational certificates check out. The proof of Theorem 3.1 does not contain a flaw that threatens the conclusion. The only external premise is the Wang-Duan one-shot characterization (Eq. (2.2)), which is a published, standard result; reliance on it is appropriate. I did find a minor typo in the decomposition of K in Eqs. (3.22)-(3.23): the second factor b2 should be (-iz/2, alpha, iz/2) rather than (iz/2, alpha, -iz/2) for the sum to reproduce Eq. (3.21). Because the singular values stated in Eq. (3.25) are unchanged under this correction, the final upper bound log ||T^Gamma||_1 < log(391/250) is unaffected.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper addresses the asymptotic tightness of the min-Rains relative entropy as a single-letter upper bound on the regularized exact PPT distillable entanglement. It introduces a 'range-supported tensor witness' T satisfying PT=T and TrT=1, and proves (Theorem 2.2) that for every n, W0(P^⊗n) ≥ ||T^Γ||_1^{-n}, giving E^∞_{0,PPT}(ρ) ≤ log||T^Γ||_1 for every state with support projection P. Pairing such a witness with a feasible dual certificate R ≥ P for the min-Rains SDP yields a strict separation criterion (Corollary 2.3). The main construction is an explicit rank-three qutrit–qutrit projection P (Section 3) with rational certificates T and R satisfying ||T^Γ||_1 < 391/250 and ||R^Γ||_∞ = 6393/10000, separated by the exact margin 337/3910000. Consequently every state supported on P has E^∞_{0,PPT}(ρ) < R_min(ρ), answering in the negative the question left open in [WD17] and [BDWW19].","tokens_in":12168,"tokens_out":20386,"duration_ms":179187,"significance":"Assuming the standard Wang–Duan one-shot characterization (Eq. (2.2)), the result is sound and settles an open problem in entanglement theory. The significance is twofold: it disproves the conjectured exactness of the min-Rains relative entropy for exact PPT distillation, and it introduces a new tensor-stable technique (the range-supported witness) that converts the fixed-action rigidity of feasible distillation effects into a single-letter bound. The counterexample is unusually verifiable: all certificates are explicit rational matrices, the semidefinite comparisons are reduced to 2×2 determinants and elementary eigenvalue computations, and the final margin is an exact rational number. This makes the proof easy to audit and the result likely to be influential for further studies of asymptotic exact entanglement manipulation.","major_comments":[],"minor_comments":[{"comment":"The sentence 'where (i) is the strong duality established in Ref. [WD17] follows from the standard Slater condition [VB96]' is grammatically garbled; it should read, for example, 'where (i) is the strong duality, established in Ref. [WD17] via the standard Slater condition [VB96].'","section":"Section 2 (after Eq. (2.9))"},{"comment":"The decomposition of K is correct as printed: with b2 = (iz/2, α, -iz/2)^T one has b2† = (-iz/2, α, iz/2), and a1 b1† + a2 b2† reproduces Eq. (3.21). Displaying b2† explicitly would remove any doubt for readers; the flagged 'typo' does not affect the singular values in Eq. (3.25) or the final bound.","section":"Section 3.1, Eqs. (3.22)–(3.23)"},{"comment":"The proof relies on the Wang–Duan one-shot characterization without re-derivation. This is an appropriate citation of a standard published result, but a one-sentence reminder that all subsequent conclusions are conditional on Eq. (2.2) would help the reader distinguish the internal proof from the imported theorem.","section":"Section 2, Eq. (2.2)"}],"recommendation":"minor_revision","confidential_remarks":"The manuscript is suitable for the journal's scope. The central claim is sound, the exact rational certificates are convincing, and the construction is transparent. The only external premise is the Wang–Duan one-shot characterization, which is standard and correctly cited. I have no concerns about the citation pattern, the LLM-assisted construction disclosure, or the novelty of the result."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know about arXiv:2608.12135. First, the separation is genuine: the min-Rains relative entropy is not tight for exact PPT distillation, and the proof is exact, with rational arithmetic closing a gap of 337/3910000. Second, the paper's real contribution is the method, not just the scalar answer. The observation that every feasible exact-distillation effect acts as the identity on the input support (Lemma 2.1) is simple and powerful, and the tensor-stable range-supported witness (Theorem 2.2) turns that rigidity into a blocklength-robust upper bound. That is the part people will build on.\n\nThe counterexample is well chosen: a rank-three qutrit subspace built from three weighted cycle edges. The certificates T and R are explicit, and the verification is complete: R-P positivity is checked block by block with positive determinants, the partial transpose spectrum of R is computed via the characteristic polynomial, and the singular values of T^Gamma are bounded by rationals. I spot-checked the key claims and found the arithmetic consistent. The norm margin is tiny but rigid, with no fitted slack covering a gap.\n\nThe weak spots are modest. The paper imports the Wang-Duan characterization W0(P)=min{||E^Gamma||_infinity : P<=E<=1} without re-deriving it; that is standard and published, so it is not a defect, but a referee should verify that the stated version applies to arbitrary states with support P, including the rounding relations in Eqs. (2.2)-(2.4). The handwritten check of the singular-value decomposition of K in Eqs. (3.22)-(3.23) contains a small typo: the vector b2 should have (-iz/2, alpha, iz/2), not (iz/2, alpha, -iz/2). It changes nothing, since the two rank-one terms still have orthogonal input/output spaces and the singular values stated in Eq. (3.25) are unaffected. Still, it should be fixed. The absence of code for the numerical search is minor; Section 4 reconstructs the search transparently, and the final proof is self-contained.\n\nWho is this for? Anyone working on Rains bounds, SDP relaxations of distillable entanglement, or the asymptotic structure of exact entanglement manipulation. It answers an explicit open question from WD17 and BDWW19, so it deserves a serious referee. I would send it to a good quant-ph referee with a request to re-verify the arithmetic in Section 3 and to check whether the non-Hermitian witness technique extends to other supports. After a minor revision fixing the typo and perhaps adding a short code snippet for the certificates, it should be accepted.","headline":"Exact rational certificates show the min-Rains relative entropy is genuinely not tight for exact PPT distillation, via a simple rigidity lemma and a tensor-stable witness worth building on.","tokens_in":12811,"tokens_out":2012,"would_cite":true,"duration_ms":18649,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P40","81P45"],"pacs":["03.67.-a","03.67.Mn"],"model":"deepseek-v4-flash","headline":"The additive min-Rains relative entropy is not an exact formula for regularized zero-error PPT entanglement distillation.","keywords":["exact entanglement distillation","PPT operations","min-Rains relative entropy","regularized distillable entanglement","range-supported tensor witness","qutrit counterexample","semidefinite programming","entanglement theory"],"falsifier":"Evaluate the semidefinite program for $W_0(P)$ on the projection in Eq. (3.2) with exact rational arithmetic. The theorem forces $W_0(P)\\ge 250/391\\approx 0.639386$, while the min-Rains certificate gives $M(P)\\le 6393/10000=0.6393$; the two one-shot values differ in the fourth decimal, so an exact or high-precision SDP directly checks the one-shot gap. If the computed $W_0(P)$ were no larger than $M(P)$, the range-witness inequality would be false and the regularization argument would collapse.","tokens_in":11813,"feed_emoji":"⚛️","tokens_out":11120,"duration_ms":110948,"temperature":0.7,"pith_summary":"Exact entanglement distillation asks how many maximally entangled pairs can be produced with zero error, and under operations that preserve positivity of partial transpose a natural candidate formula has been the additive min-Rains relative entropy. This paper proves that formula is not tight: there is a three-dimensional subspace of two qutrits for which every state with that support has regularized exact distillation rate strictly below the min-Rains value. The separation comes from a constraint that the min-Rains relaxation drops: any feasible one-shot effect must act as the identity on the support of the input state. The paper establishes the gap with exact rational certificates, so the conclusion does not rest on numerical approximation.","feed_headline":"Min-Rains bound is not exact for zero-error distillation","feed_subtitle":"A single three-dimensional qutrit subspace keeps the gap at every blocklength.","key_machinery":"The load-bearing mechanism is Lemma 2.1 together with a range-supported tensor witness. Lemma 2.1 says that any operator $E$ with $P\\le E\\le 1$ must satisfy $EP=PE=P$, so $E$ is fixed to the identity on the support and has no support-complement coherences. Given any $T$ with $PT=T$ and $\\operatorname{Tr}T=1$, Hölder's inequality then yields $W_0(P^{\\otimes n})\\ge \\|T^{\\Gamma}\\|_1^{-n}$, so the one-shot SDP value cannot drop faster than an exponential with exponent $\\log\\|T^{\\Gamma}\\|_1$. The paper constructs $T$ as three rank-one blocks on the edge subspaces and a three-by-three coherence block after partial transpose, computes $\\|T^{\\Gamma}\\|_1$ from its two singular values, and pairs it with a feasible dual certificate $R$ for the min-Rains SDP to get the strict ordering. The witness is allowed to be non-Hermitian, which is why it can detect the identity action missed by Hermitian dual relaxations.","core_discovery":"The central discovery is an explicit rank-three projection $P$ in $\\mathbb{C}^3\\otimes\\mathbb{C}^3$, spanned by three two-edge vectors, such that for every state $\\rho$ with support $P$, $$$E^{{\\infty}}$_{0,\\mathrm{PPT}}(\\rho) \\le \\log\\|$T^{{\\Gamma}}$\\|_1 < \\log\\frac{391}{250} < -\\log\\frac{6393}{10000} \\le R_{\\min}(\\rho).$$ The operator $T$ is a non-Hermitian 'range-supported tensor witness' with $PT=T$ and $\\operatorname{Tr}T=1$; its partial-transpose trace norm is computed exactly from two singular values. The Hermitian operator $R$ used for the lower side dominates $P$, but has an eigenvalue larger than one, which is why the min-Rains program misses the effect constraint. Because the constants are rational and the comparisons are exact, the strict inequality is a theorem, not a numerical accident. The conclusion is that the discarded constraint $E\\le 1$ survives arbitrary tensor powers, and the min-Rains relative entropy does not give the exact regularized rate.","pith_inferences":["The range-witness test suggests a general search strategy for other counterexamples: for a candidate support, optimize $\\|T^{\\Gamma}\\|_1 M(P)$ over operators $T$ with $PT=T$ and $\\operatorname{Tr}T=1$; a value below one certifies a separation, and this paper's proof only requires finding one such pair.","The reciprocal weight pattern and the quarter-turn phase that make the coherence block simple hint that larger gaps may be attainable by weighted cycle supports with more edges or different phases, though the optimum for the present family is not proved.","If exact LOCC distillation ever gets a support-only one-shot characterization analogous to $W_0$, the same identity-on-support rigidity would likely force the corresponding relaxed entropy to fail tightness as well, in any dimension where such a witness exists.","The paper leaves open the exact value of $E^{\\infty}_{0,\\mathrm{PPT}}(\\rho_{\\star})$ for the constructed support; the certified upper and lower bounds differ by about $2\\times 10^{-4}$ bits, so pinning down the value might reveal the next constraint an exact formula must include."],"forward_implications":["For every state supported on the constructed qutrit subspace, the regularized exact PPT distillable entanglement is at most $\\log(391/250)$, a number strictly smaller than the min-Rains value $-\\log(6393/10000)$.","The one-shot constraint $E\\le 1$ is asymptotically relevant: the strict one-shot gap between $W_0(P)$ and $M(P)$ does not disappear after tensor powers.","The min-Rains relative entropy cannot be used as the closed-form formula for exact PPT distillable entanglement; any exact formula must encode the identity action on the input support.","Because the bound holds for every state with support $P$, the separation is a geometric property of the subspace, not of a particular density matrix inside it."],"supporting_citations":[{"why":"Supplies the support-only semidefinite characterization $W_0(P)$ of one-shot exact PPT distillation, the setting of the whole argument.","marker":"[WD16]"},{"why":"Introduces the additive min-Rains relative entropy, proves the strong duality behind $M(P)$ and its multiplicativity, and leaves open the tightness question answered here.","marker":"[WD17]"},{"why":"Names the min-Rains relative entropy and explicitly poses the asymptotic tightness question in Section 3.4, the problem resolved by this paper.","marker":"[BDWW19]"},{"why":"Relates the one-shot SDP value to the floor of $W_0(P)^{-1}$ and to the regularized rate, used in converting the witness bound into a statement about distillation.","marker":"[FWTD19]"},{"why":"Defines the partial-transpose-relaxed set and the Rains bound from which the min-Rains quantity is derived.","marker":"[Rai99a]"}],"fun_headline_variants":["Exact PPT distillation resists min-Rains formula","Rank-three gap sinks min-Rains bound","Zero-error rate outperforms min-Rains bound","Min-Rains entropy fails for exact distillation","Tensor-stable rigidity beats min-Rains bound"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument rests on the imported one-shot characterization that exact PPT distillation from a state is determined by the semidefinite program $W_0(P)=\\min\\{\\|E^{\\Gamma}\\|_\\infty: P\\le E\\le 1\\}$ for the support projection $P$; if that characterization were not exact for actual PPT operations, the strict gap proved here would be a statement about the program rather than about distillation.","fun_headline_variants_meta":{"raw":{"variants":["Exact PPT distillation resists min-Rains formula","Rank-three gap sinks min-Rains bound","Zero-error rate outperforms min-Rains bound","Min-Rains entropy fails for exact distillation","Tensor-stable rigidity beats min-Rains bound"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000835,"raw_usage":{"total_tokens":3651,"prompt_tokens":963,"completion_tokens":2688,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":579,"completion_tokens_details":{"reasoning_tokens":2618}},"tokens_in":579,"tokens_out":2688,"duration_ms":19039,"temperature":1.0,"reasoning_tokens":2618,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:16:42.414358+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the semidefinite program for $W_0(P)$ on the projection in Eq. (3.2) with exact rational arithmetic. The theorem forces $W_0(P)\\ge 250/391\\approx 0.639386$, while the min-Rains certificate gives $M(P)\\le 6393/10000=0.6393$; the two one-shot values differ in the fourth decimal, so an exact or high-precision SDP directly checks the one-shot gap. If the computed $W_0(P)$ were no larger than $M(P)$, the range-witness inequality would be false and the regularization argument would collapse.","supporting_citations":[],"review_version":1}