{"id":"cd8fceae-f851-4621-80d7-949e145290df","arxiv_id":"2608.13551","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"Every positive-partial-transpose channel is claimed to become entanglement-breaking after finitely many iterations, but the paper's additional uniform bound of 3 for a large family is false as stated.","lead":"This mathematics paper claims to settle that a broad class of quantum channels, those with positive partial transpose, must eventually stop creating entanglement when applied repeatedly. The first conclusion appears well argued, but one of the paper's advertised secondary results is false under its own definitions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 5.5's proof fails: the ⊤∘SP2 summand's outer factor Φ2∘⊤∘Ad_L^* is not shown to be PPT, so Theorem 5.1 cannot be applied there.","rationale":"The reader's verdict rejects the paper based on the example identity∘transpose∘identity, but the identity map on M_d is not PPT for d≥2 because its Choi matrix is the maximally entangled state, whose partial transpose is the swap with a negative eigenvalue. That specific counterexample is invalid. Nevertheless, the proof of Theorem 5.5 contains a real gap: the outer factor in the ⊤∘SP2 summands is asserted to be PPT without justification, and a concrete PPT map Φ2 and isometry L^* show it can fail to be CP. This undermines the advertised second contribution (Theorem 5.5 and Theorem 1.2) and justifies rejection. The central Theorem 1.1 is not directly affected, but it depends on the external [HRSF20, Theorem 3.14] whose hypotheses for non-trace-preserving maps remain unverified; I flag this as a secondary concern. Overall the submitted manuscript contains an unproved main ancillary result, so it cannot be accepted as is.","tokens_in":13336,"tokens_out":33746,"duration_ms":283201,"concrete_test":"Compute the Choi matrix of the map Φ2∘⊤∘Ad_{L^*} for d=3, Φ2(X)=PXP with P=diag(1,1,0), and L^*=[I_2;0]. The Choi matrix equals (I_2⊗(Φ2∘⊤∘Ad_{L^*}))(|Ω⟩⟨Ω|), which is the swap on C^2⊗C^2 embedded in C^2⊗C^3; its eigenvalues include -1, so the map is not CP. This settles the proof gap. To test whether Theorem 5.5(1) itself is false, run a numerical search over PPT maps Φ1,Φ2∈M_3 and Λ∈DSP_2, checking whether Φ2∘Λ∘Φ1 has a non-separable Choi matrix; a positive finding would disprove the theorem.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In the proof of Theorem 5.5(1), the expansion of Φ2∘Λ∘Φ1 contains summands of the form (Φ2∘⊤∘Ad_{L_{l,β}^*})∘(Ad_{L_{r,β}}∘Φ1). The text asserts that 'all the summands above are compositions of PPT maps factoring through M2.' While the inner factor Ad_{L_{r,β}}∘Φ1 is PPT (as a left composition of the PPT map Φ1 with the CP map Ad_{L_{r,β}}), the outer factor Φ2∘⊤∘Ad_{L_{l,β}^*} need not be PPT. A concrete instance: take d=3, L_{l,β}^* = [I_2;0] (a 3×2 isometry embedding the qubit in the top-left block), and Φ2(X)=PXP with P=diag(1,1,0). Φ2 is CP and co-CP (since (PXP)^T = P X^T P) and hence PPT. Then Φ2∘⊤∘Ad_{L^*}(X) equals the transpose of X embedded into M_3, i.e., [[X^T,0],[0,0]]. This map has Choi matrix equal to the embedded swap on C^2⊗C^3, whose spectrum contains negative eigenvalues, so it is not CP and hence not PPT. Thus the claim that every summand is a composition of PPT maps through M_2 is false, and Theorem 5.1 does not apply to that summand. The advertised Theorem 5.5 and its consequence Theorem 1.2 are therefore not established by the given proof.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims two main results. First, Theorem 1.1 states that every completely positive and completely copositive (PPT) linear map on M_d has finite entanglement-breaking index, removing full-rank, unital, and trace-preserving assumptions. The proof in Section 4 proceeds by induction on dimension: singular right or left Perron eigenmatrices are handled by a support-splitting decomposition developed in Section 3, and the full-support case is delegated to a quoted theorem of Hanson–Rouzé–Stilck França (Theorem 2.3). Second, Section 5 introduces cones DSP_k = SP_k + ⊤∘SP_k and claims, via Theorem 5.5, that the triples (PPT, DSP2, PPT) and the fivefold tuple (PPT, DSP3, PPT, DSP3, PPT) are EB-composable, yielding uniform EB-index bounds 3 and 5 for PPT maps in DSP2 and DSP3 respectively.","tokens_in":13618,"tokens_out":11906,"duration_ms":128604,"significance":"If correct, Theorem 1.1 would be a substantial improvement over earlier eventual-entanglement-breaking results: it would settle the qualitative finite-index problem for all PPT maps without any ancillary assumption. The Perron-support splitting argument in Section 3 is elegant and the induction in Section 4 is coherent and plausible. However, the secondary results in Section 5 are not merely unproven; Theorem 5.5(1) is false as stated, and the proof of Theorem 5.5 contains a concrete composability error involving the transpose map. The advertised consequences in Theorem 1.2 are therefore unsupported. The main theorem deserves serious consideration, but the manuscript in its current form cannot be accepted.","major_comments":[{"comment":"The statement is false. Take d=2, let Φ1 = Φ2 = id, and let Λ = ⊤. Since id ∈ SP2, the transpose map satisfies ⊤ = ⊤∘id ∈ ⊤∘SP2 ⊆ DSP2. Also id is PPT. Yet Φ2 ∘ Λ ∘ Φ1 = ⊤, which is not completely positive for d=2 and hence is not entanglement breaking. Thus the triple (PPT, DSP2, PPT) is not EB-composable. This directly invalidates the claimed derivation of Theorem 1.2(1). Note that this counterexample does not refute Theorem 1.2(1) itself, because ⊤ is not PPT; but the proof mechanism claimed for that theorem is false.","section":"§5, Theorem 5.5(1)"},{"comment":"The proof's grouping of summands is invalid. For the transpose summands, the outer factor Φ2 ∘ ⊤ ∘ Ad_{L_{l,β}^*} need not be PPT: composing a PPT map with the transpose map does not preserve complete positivity. A concrete instance is d=3, L_{l,β}^* = [I_2; 0] (embedding M_2 into the top-left block of M_3), and Φ2(X) = PXP with P = diag(1,1,0). Then Φ2 ∘ ⊤ ∘ Ad_{L^*}(X) equals the transpose of X embedded into M_3; its Choi matrix is the embedded swap on C^2 ⊗ C^3, which has negative eigenvalues and is therefore not CP. Hence the assertion that all summands are compositions of PPT maps factoring through M_2 is false, and Theorem 5.1 cannot be applied to that summand.","section":"§5, proof of Theorem 5.5(1), after Eq. (11)"},{"comment":"The same composability error appears in the proof of part (2). The four displayed 'middle linear maps' are asserted to be PPT maps on M_3, but the two types involving ⊤, for example Ad_{L_{r,β'}^{(2)}} ∘ Φ2 ∘ ⊤ ∘ Ad_{L_{l,β}^{(1)*}}, are not generally PPT. Taking Φ2 = id makes an identity-composed middle factor contain the transpose map, which is not CP and hence not 2-superpositive. Consequently the application of part (1) to these summands is unjustified, and Theorem 1.2(2) is not established by the given argument.","section":"§5, proof of Theorem 5.5(2)"}],"minor_comments":[{"comment":"The reference to 'Theorem??' before Eq. (2) is a placeholder; it should be replaced by the actual theorem number (presumably Theorem 5.5(1)).","section":"§1.2 and §5"},{"comment":"There are several typographical errors, including 'Furhtermore' (Corollary 4.1), 'EB-comosable' (Theorem 5.5(2)), 'trasnpose' (Section 2), 'eventaully' (Definition of EB index), and 'entangelment-breaking' (proof of Theorem 5.5(2)). These should be corrected.","section":"Throughout"},{"comment":"The definition of the rectangular matrices K_l, K_r, L_l, L_r is ambiguous: the text says 'd×k rectangular matrices' and also writes 'M_{k,d}'. Please fix the orientation so that the factorizations K = K_l^* K_r and L = L_l^* L_r have the stated ranks.","section":"§5, proof of Theorem 5.5"},{"comment":"The proof of the full-support case relies entirely on the quoted Theorem 2.3 from [HRSF20]. Please verify that the statement reproduced here exactly matches the original theorem, and add a sentence confirming that no unital or trace-preserving hypothesis is needed; the induction in Theorem 1.1 depends on this black-box result.","section":"§4, Theorem 1.1, Case 4"},{"comment":"The extremality argument for id_d ∉ DSP_k for k < d is terse. Consider spelling out why the two summands Ψ1 and ⊤∘Ψ2 must both be scalar multiples of id_d before concluding that ⊤ would be a scalar multiple of id_d.","section":"§5, Example 5.4(1)"}],"recommendation":"major_revision","confidential_remarks":"The main theorem (Theorem 1.1) appears plausible and potentially publishable, but the false Theorem 5.5(1) is a serious defect that must not remain in the paper. If the authors can remove or correct the Section 5 claims and keep Theorem 1.1 as the sole advertised result, the paper may be acceptable after a careful revision; otherwise rejection may be necessary."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the part that matters: Theorem 1.1 is likely correct. The support-splitting argument in Section 3 is a genuine technical improvement over RJP18 and HRSF20, and the induction in Section 4 handles the singular Perron cases cleanly. The only real external dependency is Theorem 2.3 (HRSF20's full-support result), which is quoted as a black box; if that theorem requires trace-preservation or normalization that isn't exactly satisfied here, the induction would have a hole, but I don't see evidence of that from the text. The base case and the Cayley-Hamilton case are fine. I read Section 4 three times and didn't find a hole.\n\nThe trouble is Section 5. The proof of Theorem 5.5(1) claims that every summand in the expansion of Φ2∘Λ∘Φ1 is a composition of PPT maps through M2. For the Σ Ad_K terms that's true: both factors are PPT and factor through M2. For the Σ ⊤∘Ad_L terms it's not. The inner factor Ad_{L_r}∘Φ1 is PPT, but the outer factor Φ2∘⊤∘Ad_{L_l^*} need not be PPT. The stress-test example is correct: with d=3, L_l^*=[I2;0] and Φ2(X)=PXP, P=diag(1,1,0), that map is the transpose embedded in the top-left corner, which is not CP. So the proof step fails. Note this doesn't show the theorem is false; it just shows the proof doesn't work. The reader's counterexample with d=2 is wrong because the identity map is not PPT on M2. So the Section 5 result may be true, but it's unproven as written.\n\nSince the abstract and intro advertise the uniform bound 3 and the PPT-cubed evidence as main results, the paper can't be accepted as is. But the main theorem is substantial and the author has made a real advance on the finite-index problem. A serious referee should get a chance to sort out Section 5. The paper deserves peer review, not desk rejection.","headline":"Theorem 1.1 is likely correct and closes the full-rank gap; Section 5's proof has a genuine hole, and the reader's counterexample to it misfires because identity is not PPT.","tokens_in":14242,"tokens_out":13183,"would_cite":true,"duration_ms":123031,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P45","15B48","47B65"],"pacs":["03.67.-a"],"model":"deepseek-v4-flash","headline":"Repeatedly applying any PPT quantum channel eventually produces an entanglement-breaking channel.","keywords":["entanglement-breaking index","PPT maps","eventually entanglement breaking","Perron–Frobenius theory","Schmidt number","superpositive maps","PPT-squared conjecture","quantum channels"],"falsifier":"A single PPT map on some $M_d$ whose Choi matrix is PPT but whose iterates $\\Phi^n$ all remain entangled for every $n$ would refute Theorem 1.1; a concrete search would be to compute the Choi matrices of high powers for a parametrized family of PPT maps with singular Perron eigenmatrices and test separability numerically.","tokens_in":13042,"feed_emoji":"🔗","tokens_out":11930,"duration_ms":116833,"temperature":0.7,"pith_summary":"This paper proves that every completely positive and completely copositive linear map on a finite-dimensional matrix algebra—the class of PPT maps—has finite entanglement-breaking index. Equivalently, for every PPT map $\\Phi : M_d \\to M_d$ there is an integer $N$ such that $\\Phi^n$ is entanglement breaking for every $n \\ge N$, meaning the Choi matrix of the iterate is separable. The result removes the unital, trace-preserving, and full-rank assumptions that earlier results needed. The proof splits a PPT map with a singular Perron eigenmatrix into two lower-dimensional PPT blocks and uses induction, while the full-support case is imported from a known theorem. A second set of results gives uniform bounds: PPT maps in $\\mathrm{DSP}_2$ become entanglement breaking at the third iterate, and those in $\\mathrm{DSP}_3$ at the fifth, independently of dimension.","feed_headline":"Every PPT channel eventually breaks entanglement","feed_subtitle":"Finite entanglement-breaking index now holds for all PPT maps, with no unital, trace-preserving, or full-rank assumptions.","key_machinery":"The central object is the Perron–Frobenius support splitting of a PPT map. Given a positive eigenmatrix $\\rho$ of $\\Phi$, the paper writes $\\Phi = A + B$ with $A$ supported on the range of $\\rho$ and $B$ on its orthogonal complement, so that $B \\circ A = 0$; powers of $A$ and $B$ then factor through the restricted corner maps $\\Phi_\\rho$ and $\\Phi_{\\rho^\\perp}$. The key identity is $n_{\\mathrm{EB}}(\\Phi) \\le n_{\\mathrm{EB}}(\\Phi_\\rho) + n_{\\mathrm{EB}}(\\Phi_{\\rho^\\perp}) + 1$, which drives the induction on dimension. A complementary estimate bounds the Schmidt number of the Choi matrix of $\\Phi$ by $\\max(\\operatorname{rank} \\rho, d - \\operatorname{rank} \\rho)$, forcing the map into $\\mathrm{SP}_{d-1}$ when $\\rho$ is singular.","core_discovery":"The central claim is Theorem 1.1: for every linear map $\\Phi : M_d \\to M_d$ that is completely positive and completely copositive, there exists an integer $N$ such that $\\Phi^n \\in \\mathrm{EB}$ for all $n \\ge N$. The proof proceeds by induction on dimension, using Perron–Frobenius theory to find a positive eigenmatrix $\\rho$. When $\\rho$ is singular, the support splitting of Section 3 factors the iterates through two strictly smaller PPT maps, reducing the problem; the full-support case is covered by a previously known theorem. The paper also proves that PPT maps in the classes $\\mathrm{DSP}_2$ and $\\mathrm{DSP}_3$ satisfy the dimension-independent bounds $n_{\\mathrm{EB}}(\\Phi) \\le 3$ and $n_{\\mathrm{EB}}(\\Phi) \\le 5$, respectively.","pith_inferences":["The recursion underlying Theorem 1.1 suggests an algorithmic way to compute an explicit upper bound on $N(\\Phi)$: repeatedly split along singular Perron eigenmatrices until reaching full-support blocks; the paper states the bound but does not optimize or implement this recursion.","If every PPT map on $M_d$ were eventually shown to lie in $\\mathrm{DSP}_2$ or $\\mathrm{DSP}_3$ (the paper's Question 5.6), Theorem 1.2 would supply a dimension-independent uniform bound, settling the open uniformity question affirmatively for all $d$.","The full-support case is imported as a black box; replacing it with a constructive proof would make the entanglement-breaking time of concrete channels computable in practice.","A natural numerical test is to generate random PPT channels on $M_4$ or $M_5$, check membership in $\\mathrm{DSP}_2$ by semidefinite programming, and verify $\\Phi^3 \\in \\mathrm{EB}$; failures would indicate where a uniform bound could break."],"forward_implications":["Every PPT channel is eventually entanglement breaking: no PPT map can preserve entanglement indefinitely under iteration.","For a completely positive map, eventual PPT-ness is equivalent to eventual entanglement-breaking, with the index bound $n_{\\mathrm{EB}}(\\Phi) \\le n_{\\mathrm{PPT}}(\\Phi) \\, n_{\\mathrm{EB}}(\\Phi^{n_{\\mathrm{PPT}}})$.","Compositions of PPT maps through a smaller intermediate algebra are eventually entanglement breaking, with $n_{\\mathrm{EB}}(\\Phi_2 \\circ \\Phi_1) \\le 1 + n_{\\mathrm{EB}}(\\Phi_1 \\circ \\Phi_2) < \\infty$.","Any PPT map in $\\mathrm{DSP}_2$ satisfies $\\Phi^3 \\in \\mathrm{EB}$, and any PPT map in $\\mathrm{DSP}_3$ satisfies $\\Phi^5 \\in \\mathrm{EB}$, uniformly in dimension.","The inclusion $\\mathrm{PPT} \\circ \\mathrm{DSP}_2 \\circ \\mathrm{PPT} \\subseteq \\mathrm{EB}$ gives a large, explicitly parameterized family of triples of maps whose composition is immediately entanglement breaking."],"supporting_citations":[{"why":"Supplies Theorem 2.3, the full-support case that completes Case 4 of the induction.","marker":"[HRSF20]"},{"why":"Proved finite EB index for unital and trace-preserving PPT channels, the base result that Theorem 1.1 extends.","marker":"[RJP18]"},{"why":"Introduced the entanglement-breaking index and its iteration properties, used throughout.","marker":"[LG15]"},{"why":"Shows PPT maps composed through a qubit are entanglement breaking, the key input for the DSP_2 composability theorem.","marker":"[Hen26]"},{"why":"Proved the qutrit PPT-squared case and supplied the fact that PPT maps on M_3 are 2-superpositive, used in Theorem 5.5.","marker":"[CYT19]"},{"why":"Proved the qutrit case and set up the EB-composability framework used to define the DSP_k classes.","marker":"[CMHW19]"},{"why":"Defines k-superpositive cones and their Schmidt-number characterization, grounding the classes DSP_k.","marker":"[SSrZ09]"},{"why":"Defines the Schmidt number, the measure through which Theorem 3.8 bounds the entanglement dimensionality of a PPT map.","marker":"[TH00]"},{"why":"Provides the Perron–Frobenius theorem for positive maps used to produce Perron eigenmatrices in Theorem 2.2.","marker":"[EHK78]"}],"fun_headline_variants":["All PPT maps break entanglement in finite steps","Finite entanglement-breaking index for every PPT channel","PPT channels: entanglement dies in finitely many uses","Every PPT map has bounded entanglement-breaking index","PPT maps eventually disentangle, no assumptions needed"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument leans on a quoted theorem saying that a PPT map whose two Perron eigenmatrices are strictly positive has finite entanglement-breaking index; if that theorem has hidden assumptions, or fails for non-trace-preserving maps, the induction collapses.","fun_headline_variants_meta":{"raw":{"variants":["All PPT maps break entanglement in finite steps","Finite entanglement-breaking index for every PPT channel","PPT channels: entanglement dies in finitely many uses","Every PPT map has bounded entanglement-breaking index","PPT maps eventually disentangle, no assumptions needed"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000158,"raw_usage":{"total_tokens":1147,"prompt_tokens":791,"completion_tokens":356,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":407,"completion_tokens_details":{"reasoning_tokens":286}},"tokens_in":407,"tokens_out":356,"duration_ms":4005,"temperature":1.0,"reasoning_tokens":286,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:22:01.022267+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A single PPT map on some $M_d$ whose Choi matrix is PPT but whose iterates $\\Phi^n$ all remain entangled for every $n$ would refute Theorem 1.1; a concrete search would be to compute the Choi matrices of high powers for a parametrized family of PPT maps with singular Perron eigenmatrices and test separability numerically.","supporting_citations":[],"review_version":1}