{"id":"8de9ad12-d556-491c-8848-e70b5dbd5aff","arxiv_id":"2607.15210","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For every Rényi order p<1/4 or p>3/4, some finite-dimensional quantum channels have non-additive minimum output p-Rényi entropy.","lead":"This paper proves that additivity of minimum output Rényi entropy fails for quantum channels for every Rényi order p in (0,1/4) and for p>3/4, leaving only [1/4,3/4] open in 0<p<1. It uses random projections and free probability to turn two known constructions into explicit intervals, a major step in a central quantum-information problem.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; core argument is internally coherent and the cited external theorems cover the main technical step.","rationale":"The reader identified the strong-convergence-under-local-normalization step as the weakest assumption. That is indeed the most technical and least self-contained part of the proof, since it imports heavy free-probability machinery. However, the application appears sound: Lemma 2.4 is a direct consequence of the strong block-modification theorem for maps whose Choi matrix has spectral projections with partial traces proportional to the identity, which holds for phi_a(X)=Tr(AX)I_k. Prop. A.2 then uses the spectral lower bound m_{k,t}>0 to pass the inverse square root through the strong limit, a standard functional-calculus step for strongly convergent positive matrices with uniform spectral gap. The subsequent Bell-output calculation (Prop. A.3) is elaborate but internally consistent, and the final entropy comparisons in Section 5 check out. The only unproved assertion I find is the p=0 full-rank statement, but the paper explicitly cites [CHL+08] for p=0; since the theorem's existential claim at p=0 is already established there, this does not undermine the main novel result. I therefore do not see a reason to change the reader's ACCEPT verdict.","tokens_in":16650,"tokens_out":34922,"duration_ms":273640,"concrete_test":"Verify Prop. A.2 by an independent free-probability calculation: for k=3, t=1/2, simulate n=2000, compute the empirical spectrum of P_A^{-1/2}P_11 P_A^{-1/2} and check its largest eigenvalue against max_{u in D_{3,1/2}} u_1, and compute 1/(nk)Tr(J_n^2) against formula (59). Agreement within expected O(n^{-1/2}) fluctuations would confirm the local-normalization step.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After reviewing Theorem 1.1 and its proof, I find no load-bearing defect. The two asymptotic constructions (product-conjugate for p>3/4, transpose-complement for p<1/4) are internally consistent: the output-body limit K_{k,t}, the Bell-state limit Z_{k,t}, and the large-k entropy expansions match at the required order, and the threshold comparisons (13) and (48) are correct. The technical pivot is the strong block-modification theorem applied to the locally normalized Choi blocks (Lemma 2.4 and Prop. A.2); the cited theorems [Nec18, ANV16, CM14] apply to this unitarily invariant setting, and the spectral lower bound m_{k,t}>0 justifies passing the inverse square root through the strong limit. The p=0 endpoint is cited to [CHL+08] rather than re-proved for the random ensemble; this is a presentational gap, but the existential claim for p=0 is already established there, so it does not threaten Theorem 1.1.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies additivity of the minimum output p-Rényi entropy of quantum channels. It constructs, for each p in (3/4,∞) and each p in (0,1/4), finite-dimensional projection-induced channels Φ,Ψ with S_p^min(Φ⊗Ψ)<S_p^min(Φ)+S_p^min(Ψ). The high-p construction uses a Haar-random projection, locally normalizes its Choi matrix, pairs the channel with its complex conjugate, and feeds it a maximally entangled state; a free-probability limit identifies the Bell output as an isotropic state and a large-k expansion yields a threshold at p=3/4. The low-p construction pairs a half-rank projection with its transpose-orthogonal complement and uses a rank-deficit lemma to bound the joint output entropy by log(k^2−1), while a one-channel output-body limit gives 2log k−4p/k^2, yielding violation for p<1/4. The unresolved interval for 0<p<1 is thus reduced to [1/4,3/4].","tokens_in":16850,"tokens_out":41991,"duration_ms":310146,"significance":"Assuming correctness, this is a substantial advance: it gives the first explicit uniform intervals on both sides of the von Neumann point for which additivity of minimum output Rényi entropy fails, replacing channel-dependent neighborhoods from continuity arguments. The proof is a coherent combination of strong asymptotic freeness, support-function convergence, and asymptotic entropy expansions. The paper is honest about prior work: the p=0 endpoint and the rank-deficit lemma are attributed to CHL+08. It also reports an improvement of the output-dimension threshold for von Neumann additivity violation from 183 to 182, though this numerical claim is not load-bearing.","major_comments":[],"minor_comments":[{"comment":"The abstract and Section 1 claim the result for 0≤p<1/4, but Theorem 1.1 as stated covers p∈(0,1/4)∪(3/4,∞). The p=0 endpoint is not part of the stated theorem, yet the text later says the unresolved part is reduced to [1/4,3/4], implying 0 is covered. Please reconcile the statement with the abstract and with the actual proof.","section":"Abstract and Theorem 1.1"},{"comment":"Proposition 5.4 asserts the random half-rank construction works for every 0≤p<1/4, but the p=0 case is dismissed with 'The case p=0 follows from [CHL+08].' That reference proves existence of some counterexample, not that the specific half-rank projection-induced pair works. If p=0 is intended to be covered by Theorem 1.1, supply the missing rank argument; otherwise restrict the proposition to p>0.","section":"§5.2, Proposition 5.4"},{"comment":"The overview states that for fixed k≥3 and large n the one-channel outputs are almost surely of full rank. No proof is given. This fact is nontrivial: for k=2 it is false, since a random half-rank subspace of C^n⊗C^2 has a product vector in its kernel with probability one. The proof of the p=0 case (if pursued) should justify the full-rank claim, e.g., by dimension-counting the absence of product k-planes in the kernel.","section":"§1.1 and §5.2"},{"comment":"The sentence 'This limit is strictly larger than 4p precisely when p>3/4' is mathematically incorrect: the limit is 1/(1−p), and 1/(1−p)>4p holds for all p≠1/2 in (0,1). The comparison relevant to (46) is A_p(x^{-1})/(p(x−1)) > 4, i.e., 1/(1−p)>4, which is exactly p>3/4. The conclusion of the lemma is correct, but the stated comparison should be fixed.","section":"Lemma 5.2"},{"comment":"The proof of Proposition 5.4 says 'Almost surely, both Tr_B P_n and Tr_B Q_n are invertible for all sufficiently large n' without proof. Lemma 2.3 covers Tr_B P_n, but the invertibility of Q_n = kI_n − P_n^T needs a separate argument (e.g., that the probability of P_n containing a product k-plane is zero for k≥3 and large n). Please add a sentence or reference.","section":"§5.2"},{"comment":"There are small typos: 'Nechida' should be 'Nechita' in the abstract; 'developped' should be 'developed' in the proof of Lemma 2.4; the notation ⊞^k in Lemma 2.4 is unclear and should be cleaned up.","section":"Throughout"}],"recommendation":"minor_revision","confidential_remarks":"The central proof appears sound; the main issues are presentational. The authors should reconcile the abstract's 0≤p<1/4 claim with Theorem 1.1's (0,1/4) statement and either prove or clearly relegate the p=0 case. The paper should be acceptable after these revisions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a genuine advance and deserves a proper referee. The paper proves that for every p in (0,1/4) and (3/4,∞) there are finite-dimensional projection-induced channels violating additivity of minimum output p-Rényi entropy. That explicit interval statement is new. Previous work only gave an unspecified neighborhood of p=0 and channel-dependent continuity below p=1, and the YY10 preprint was withdrawn. The high-p Bell-state witness and the low-p transpose-complement construction are not new in themselves—the authors say this more honestly than most—but the quantitative free-probability analysis that converts them into uniform endpoints is the real contribution.\n\nWhat the paper does well: the asymptotic machinery is coherent. The one-channel output body K_{k,t}, the Bell-state limit Z_{k,t}, and the large-k entropy expansions all have the right orders, and the threshold comparisons at p>3/4 and p<1/4 are internally consistent. The paper is careful about provenance, noting that the low-p construction comes from CHL+08 and that its own contribution is the quantitative entropy analysis. It also flags YY10 as withdrawn rather than pretending the interval was open. Extending the gap function continuously to p=1 so that the von Neumann point is not singular is a nice touch.\n\nSoft spots, in proportion: the p=0 case is handled too tersely. The claim that one-channel outputs are almost surely full rank is asserted and then deferred to CHL+08, but since Theorem 1.1 includes p=0, that step deserves a direct argument in this random ensemble. The 183→182 output-dimension improvement for the von Neumann case is numerical, not rigorous; it is clearly labelled as witness-specific, so this is minor, but it should not be advertised as a proven threshold. The main technical pivot—strong convergence after local normalization in Lemma 2.4 and Proposition A.2—rests on external theorems from CM14, ANV16, and Nec18. The spectral lower bound m_{k,t}>0 makes the step plausible, and the stress-test pass agrees, but a referee should verify that the inverse-square-root normalization preserves strong convergence in the joint sense needed. I do not see circularity: the target intervals are not assumed anywhere; the asymptotic formulas are computed from the random model.\n\nWho it is for: quantum information researchers working on additivity and random channels; free probabilists will also find the block-modification application useful. It deserves a serious referee and likely publication after tightening the p=0 endpoint.","headline":"Solid quantitative advance: explicit uniform intervals (0,1/4) and (3/4,∞) for violation of minimum-output Rényi entropy additivity, reducing the open range to [1/4,3/4]; the proof relies on external free-probability theorems but the internal asymptotics check out.","tokens_in":17384,"tokens_out":1669,"would_cite":true,"duration_ms":15120,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P45","46L54","60B20","94A17"],"pacs":["03.67.-a","03.67.Hk"],"model":"deepseek-v4-flash","headline":"Minimum output p-Rényi entropy is shown nonadditive for all p>3/4 and all 0≤p<1/4, leaving only [1/4,3/4] open in 0<p<1.","keywords":["quantum channels","minimum output entropy","Rényi entropy","additivity","random projections","free probability","Bell state","nonadditivity"],"falsifier":"Numerically evaluate the support function of the normalized Choi blocks P_A^{-1/2} S_n(a) P_A^{-1/2} for random a and compare to max_{u∈D_{k,t}} Σ a_i u_i; a stable deviation would indicate the strong block-modification theorem is not preserved under local normalization, undermining both proofs. Alternatively, for a fixed p>3/4 such as 0.8, compute the exact finite-n minimum output entropy of the product-conjugate pair and check whether the predicted asymptotic gap appears for some n; if not, the strong-convergence step is false.","tokens_in":16497,"feed_emoji":"⚛️","tokens_out":7595,"duration_ms":64243,"temperature":0.7,"pith_summary":"The paper asks whether two independent noisy quantum channels can amplify noise more than the sum of their individual effects—the additivity question for minimum output Rényi entropy. It proves that additivity fails for every Rényi order p>3/4 and for every 0≤p<1/4: in each case there exist finite-dimensional quantum channels, built from random projections, whose combined minimum output entropy is strictly smaller than the sum of the separate minima. The proof gives explicit asymptotic entropy gaps, and because the arguments extend continuously to p=1, the von Neumann point is shown to be part of the same nonadditive mechanism rather than a singular case. If the theorem is right, the only unsettled orders in 0<p<1 are the interval [1/4,3/4].","feed_headline":"Additivity of output Rényi entropy fails for p>3/4 and p<1/4","feed_subtitle":"Random-projection counterexamples shrink the open range for 0<p<1 to the single interval [1/4,3/4].","key_machinery":"The argument rides on two limiting objects. The first is the one-channel output body K_{k,t} = {X/Tr X : 0≤X≤I_k, Tr c_t(X) ≤ 1/k}, where c_t(u) = (√(t(1−u)) − √(u(1−t)))²; this is the almost-sure Hausdorff limit of the output states of a Haar-random projection-induced channel and its minimum Rényi entropy has a known large-k expansion. The second is the isotropic Bell-state limit Z_{k,t} = r_{k,t} ψ⁺_k + (1−r_{k,t}) I_{k²}/k², with r_{k,t} = k²(1−t)/((k²−1)t + 1−t), obtained by feeding a maximally entangled state through the product of the channel and its conjugate. The rank-defect witness instead uses exact orthogonality P Q^T = 0 between a half-rank projection and its transposed complemen","core_discovery":"The central claim is Theorem 1.1: for every p in (0,1/4) ∪ (3/4,∞), there exist finite-dimensional projection-induced quantum channels Φ and Ψ with S_p^min(Φ⊗Ψ) < S_p^min(Φ)+S_p^min(Ψ). The paper establishes this by computing the large-dimension limits of two random-projection constructions. For p>3/4, it pairs a channel with its complex conjugate and evaluates the product on a maximally entangled input; the output converges to a known isotropic state, and comparing its entropy with the one-channel minimum gives a strict violation whenever A_p(t) > 4p(1−t)/t, which holds for small t precisely when p>3/4. For 0≤p<1/4, it pairs a half-rank projection channel with its transpose-orthogonal compl","pith_inferences":["The two constructions suggest that additivity might fail throughout (0,1) except possibly a middle band; numerical continuation of the asymptotic gaps could indicate whether the endpoints 1/4 and 3/4 are sharp or artifacts of the witnesses.","Because the entropy gaps scale as k^{-2}, an explicit deterministic (non-random) counterexample would need a different mechanism, presumably with more structure than Haar projections.","The rank-defect idea might generalize to other correlated projection pairs, potentially pushing the low-p endpoint toward 1/4 from below or even beyond.","The connection to random-subspace geometry suggests nonadditivity is a generic high-dimensional phenomenon, so one might expect violations for 'most' projection-induced channels in these parameter regimes."],"forward_implications":["For the existence of additivity violations with 0<p<1, only the interval [1/4,3/4] remains open.","The von Neumann point p=1 is covered by the same high-p mechanism via continuous extension; no separate construction is needed in this ensemble, and the Bell-state criterion first detects a violation at output dimension k=182 within this model.","The previously unspecified neighborhood of p=0 is replaced by the explicit range 0≤p<1/4.","Any future universal additivity theorem for quantum channels, if it exists, must have its p-domain contained in [1/4,3/4].","Both violations are asymptotic with gap of order k^{-2}; hence explicit counterexamples require sufficiently large output dimensions, and the phenomena are small but non-perturbative."],"fun_headline_variants":["Output Rényi entropy additivity fails for p>3/4 and p<1/4","Random-projection channels break p-Rényi additivity for two p-ranges","Additivity violation proven for p-Rényi entropy outside [1/4,3/4]","Quantum channel counterexamples narrow open additivity gap","Conjugate and transpose witnesses defeat entropy additivity"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The whole asymptotic proof depends on the assumption that the random projection's block entries, after rescaling by the inverse square root of its partial trace, still converge strongly to the free-probability limit; if that normalization step destroys the convergence, the limiting output body, the Bell-state spectrum, and both entropy comparisons would not be justified.","fun_headline_variants_meta":{"raw":{"variants":["Output Rényi entropy additivity fails for p>3/4 and p<1/4","Random-projection channels break p-Rényi additivity for two p-ranges","Additivity violation proven for p-Rényi entropy outside [1/4,3/4]","Quantum channel counterexamples narrow open additivity gap","Conjugate and transpose witnesses defeat entropy additivity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000825,"raw_usage":{"total_tokens":3474,"prompt_tokens":803,"completion_tokens":2671,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":547,"completion_tokens_details":{"reasoning_tokens":2585}},"tokens_in":547,"tokens_out":2671,"duration_ms":18444,"temperature":1.0,"reasoning_tokens":2585,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T23:52:24.207402+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Numerically evaluate the support function of the normalized Choi blocks P_A^{-1/2} S_n(a) P_A^{-1/2} for random a and compare to max_{u∈D_{k,t}} Σ a_i u_i; a stable deviation would indicate the strong block-modification theorem is not preserved under local normalization, undermining both proofs. Alternatively, for a fixed p>3/4 such as 0.8, compute the exact finite-n minimum output entropy of the product-conjugate pair and check whether the predicted asymptotic gap appears for some n; if not, the strong-convergence step is false.","supporting_citations":[],"review_version":1}