{"id":"f14a6285-e046-413d-8eaf-ce18cc0be634","arxiv_id":"2502.02868","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A crossed multi-copy pairing of entanglement witnesses detects entangled states that the individual witnesses and the standard same-copy pairing fail to detect.","lead":"This paper uses existing entanglement witnesses on two or more copies of a quantum state, but pairs the subsystems across copies instead of within the same copy, so that a combined measurement can detect states that each witness alone misses. It gives explicit examples where one, two, or three copies behave differently, and applies the idea to multipartite states and entanglement concentration.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Example 1's proof that W is a valid entanglement witness rests on an invalid inference; without an independent nonnegativity check, Observations 1–2 are not rigorously supported as written.","rationale":"The reader's weakest assumption identifies the same load-bearing gap: the witness validity of the constructed operators. The paper's proof in Example 1 contains a genuine logical error, since |ρ14|+|ρ23| > 2√(ρ11ρ44) does not imply |ρ23|^2 > ρ11ρ44. This matters because Observation 1 only constitutes a counterexample to linear detection if W and V are true entanglement witnesses. I also agree with the reader that the central crossed-copy trace calculations appear sound and that the strategy is conceptually valid. An independent Bloch-vector check indicates W, V, and W1 are indeed nonnegative on product states, so the flaw is a missing proof rather than a false construction. Secondary issues, including the sign inconsistency in Appendix E's displayed trace and the normalization concerns in Appendix F, are real but do not overturn the main bipartite existence claim; they reinforce the need for revision without changing the CONDITIONAL recommendation.","tokens_in":17876,"tokens_out":31221,"duration_ms":270561,"concrete_test":"Minimize E(a,b) = 1 − a_x b_x + a_z b_z over unit Bloch vectors a,b for W = I − X⊗X + Z⊗Z; if the minimum is negative, W is not an entanglement witness and Observation 1 fails. Repeat for V = 2|φ+><φ+|^τ2 using its Pauli decomposition ½(I − Z⊗Z + X⊗X − Y⊗Y), and for W1 = I + X⊗X − Y⊗Y. A symbolic or SDP check should return minimum 0 for all three; a negative value would invalidate the central examples.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central construction requires W = I − X⊗X + Z⊗Z to be a genuine entanglement witness. Example 1 tries to prove this by showing that Tr(Wρ)<0 forces a PPT violation. The chain displayed after Eq. (1) runs: |ρ14|+|ρ23| ≥ Re(ρ14)+Re(ρ23) > ρ11+ρ44 ≥ 2√(ρ11ρ44), hence |ρ23|^2 > ρ11ρ44. The last step is invalid: the premise is also compatible with |ρ14| large and |ρ23| small, so no PPT violation of the claimed form follows. The subsequent sentence, 'as W has a negative eigenvalue of −1, W is a well defined entanglement witness by definition [54]', is likewise not a sufficient witness criterion. Since W is the witness used in Observation 1 and Example 2, the paper's written proof does not establish the central existence claim. The claim may be repairable: a Bloch-vector minimization of Tr(W |a><a|⊗|b><b|) suggests the minimum on product states is 0, so W is probably a valid witness, but that argument is absent from the text.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a nonlinear entanglement detection strategy in which, given two copies of a bipartite state, one measures a witness W on AB' and another witness V on BA' (a 'crossed' ordering), rather than the same-copy pairing used in [47]. It claims (Observation 1) that there exist witnesses W, V and a state ρ such that Tr(Wρ) ≥ 0 and Tr(Vρ) ≥ 0, but Tr((W_AB' ⊗ V_BA')ρ^⊗2) < 0. It further claims (Observation 2) that when the two-copy crossed strategy fails, a three-copy crossed ordering can succeed, and gives a Werner-state example with a negativity region w < 0.206. Observations 3 and 4 extend the idea to combining witnesses with positive semidefinite operators and to detecting tripartite entanglement with bipartite witnesses. An entanglement concentration protocol is also sketched.","tokens_in":18055,"tokens_out":39337,"duration_ms":311492,"significance":"If the examples are correct, the paper gives a simple and explicit demonstration that reordering tensor factors of existing witnesses can increase detection power, complementing the trace-polynomial approach of [47]. The two- and three-copy trace calculations in Examples 1–3 are explicit and largely checkable, and the central existence claims are credible. However, the written proof that the operator W in Example 1 is an entanglement witness contains an invalid inference, and the formula in Appendix E is numerically wrong. Because these points are load-bearing for the stated observations, the manuscript requires revision before its claims are fully supported.","major_comments":[{"comment":"The proof that W = I − X⊗X + Z⊗Z is an entanglement witness is invalid as written. From Tr(Wρ) < 0 the text obtains |ρ14| + |ρ23| > ρ11 + ρ44 ≥ 2√(ρ11ρ44), and then concludes |ρ23|^2 > ρ11ρ44. This conclusion does not follow: the displayed inequality can hold with |ρ14| large and |ρ23| = 0, in which case |ρ23|^2 = 0. The subsequent statement that W is a witness because it has a negative eigenvalue is also not a valid witness criterion. Since Observation 1 and Example 2 rely on W being a genuine witness, this gap must be closed. The claim is repairable: for product states, ⟨a|⊗⟨b|W|a⟩⊗|b⟩ = 1 − x_a x_b + z_a z_b ≥ 0 by the Bloch-vector bound |x_a x_b| + |z_a z_b| ≤ 1, so a correct proof can be supplied.","section":"Section II, Example 1, Eq. (1)"},{"comment":"The displayed expression for Tr((W4,AB' ⊗ W3,BC' ⊗ W3,CA')ρ_c^⊗2), namely 2(c/8)^2 + 12(c/8)(8−5c)/24 + 4((1−c)/3)^2, is positive for every c ∈ [0,1], so it cannot imply the stated negativity for c < 0.406. A direct six-qubit trace calculation gives (50c − 25c^2 − 16)/36, which is negative for c < 0.4. The numerical threshold should be corrected, and the formula in Appendix E should be replaced, although the qualitative claim that the crossed ordering detects ρ_c for a range of c remains valid.","section":"Appendix E, Example 5"},{"comment":"The assertion that Tr(Wi,A1B2 ⊗ Wj,B1A2 ρ_w^⊗2) ≥ 0 for all i,j ∈ {1,2,3} is stated without proof or calculation. This is the 'two-copy failure' half of Observation 2 and is essential for the claimed hierarchy two-copy failure / three-copy success. Please provide the verification, or at least summarize the computation in an appendix.","section":"Section II, Example 3"}],"minor_comments":[{"comment":"There are numerous typos, including 'quanutm' in the abstract, 'whi ch' in the introduction, 'sperable' in Figure 4, 'Muliticoy' in reference [46], 'Wotters' in reference [2], and 'Lioyd' in reference [7].","section":"Throughout"},{"comment":"The opening sentence writes W2,B1A3 where the main text and the subsequent calculation use W3,B1A3; please fix this notation.","section":"Appendix C"},{"comment":"Observation 4 is phrased with witnesses Wi for i=1,2,3, but Example 5 uses W3 and W4. Please clarify the indexing so that the statement and the example match.","section":"Observation 4 and Example 5"},{"comment":"The definition of W3 in the multipartite discussion omits the factor 2 that appears in the earlier definition (W3 = 2|ψ+⟩⟨ψ+|τ2). Please make the conventions consistent.","section":"Section II, multipartite discussion"}],"recommendation":"major_revision","confidential_remarks":"The contribution is incremental relative to the trace-polynomial framework of [47]: the crossed-ordering idea is a simple reordering of tensor factors, and the paper's value lies in the explicit examples. The invalid witness proof in Example 1 and the wrong numerical formula in Appendix E are repairable, but they currently undermine the rigor of the two central observations. I recommend major revision rather than rejection, provided the authors supply a correct proof that W is a witness, correct Appendix E, and verify the two-copy failure claim in Example 3."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe crossed multi-copy witness pairing is a genuine new idea. Instead of the usual ordering W_AA' ⊗ V_BB', the authors measure W on AB' and V on BA'. For separable states this is still nonnegative, so it is a valid detection strategy, and the examples show it detects states that both the linear witnesses and the same-copy ordering miss. That is a useful addition to the entanglement-detection toolbox.\n\nThe paper does some things well. The examples are explicit: I spot-checked the trace in Example 1 and it gives −1/2 as claimed. Example 2's imaginary-coefficient state σ is a nice demonstration that the crossed ordering succeeds where the same-copy method of [47] returns a positive value. The three-copy Werner example with threshold w < 0.206 is striking, and the multipartite observations give a plausible route to detecting GHZ and W states with bipartite witnesses.\n\nThe soft spots are real. The main problem is that the paper never properly proves its operators are entanglement witnesses. In Example 1, the chain from |ρ14|+|ρ23| > ρ11+ρ44 ≥ 2√(ρ11ρ44) to |ρ23|^2 > ρ11ρ44 is invalid—the premise only implies something about the sum, not about |ρ23| individually. And the sentence \"as W has a negative eigenvalue of −1, W is a well defined entanglement witness by definition\" is not a sufficient witness criterion. The same flaw appears in Example 3's proof for W1. Observations 1 and 2 rest on W and W1 being valid witnesses, so these proofs matter. I suspect they are repairable—a Bloch-vector minimization over product states likely shows nonnegativity—but that argument is absent.\n\nTwo smaller issues: Appendix E's formula for the three-copy trace appears to have a plus sign where a minus is needed to match the stated threshold c < 0.406, and the entanglement-concentration appendix has normalization steps that need a careful pass.\n\nOverall: the central strategy is interesting and likely correct; the examples are concrete; the flaws are in the proof details, not the idea. I would send this to peer review, but I would ask for a revised version that correctly verifies the witness properties and fixes the appendix errors. Until then, I would not cite it in my own work.","headline":"Crossed multi-copy witness pairing is a genuine new detection idea with explicit examples; proof gaps are real but likely repairable.","tokens_in":18628,"tokens_out":3797,"would_cite":false,"duration_ms":33337,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P42"],"pacs":["04.70.Dy","03.65.Ud","04.62.+v"],"model":"deepseek-v4-flash","headline":"Reordering the subsystems of multiple copies of a state lets existing entanglement witnesses detect entanglement that each witness alone—and the standard aligned multi-copy construction—misses.","keywords":["quantum entanglement","entanglement witness","nonlinear entanglement detection","multi-copy detection","Werner state","GHZ state","W state","entanglement concentration"],"falsifier":"Compute the minimum of $\\mathrm{Tr}(W\\rho)$ and $\\mathrm{Tr}(W_1\\rho)$ over all separable two-qubit states, e.g. by minimizing over product Bloch vectors; if either minimum is negative, the operator is not a witness and the corresponding example collapses, whereas if both are nonnegative the crossed-copy negative expectations stand.","tokens_in":17648,"feed_emoji":"🔀","tokens_out":9888,"duration_ms":82527,"temperature":0.7,"pith_summary":"This paper proposes a nonlinear entanglement detection strategy: take existing entanglement witnesses and apply them to multiple copies of a state with the local subsystems crossed between copies, rather than aligned. The central claim is that this reordering can produce negative expectation values for entangled states on which every witness alone, and the aligned two-copy construction, gives nonnegative values. That is demonstrated for the Bell state $|\\psi^+\\rangle$, for Werner states with mixing parameter $w<0.206$ using three copies, for a Werner-like family using a witness combined with a positive semidefinite operator, and for three-qubit GHZ and W states using two-copy bipartite witnesses. The same crossed layout yields an entanglement concentration protocol. A sympathetic reader would care because the strategy extracts new detection power from already-known witnesses without constructing new ones.","feed_headline":"Crossed-copy witnesses detect entanglement linear tests miss","feed_subtitle":"Reordering two-qubit witnesses across copies yields negative expectations on Bell, Werner, GHZ, and W states.","key_machinery":"The central object is the crossed-copy witness tensor product: for two copies of a bipartite state, instead of the aligned product $W_{AA'}\\otimes V_{BB'}$, one measures $W_{AB'}\\otimes V_{BA'}$, and for $k$ copies one uses orderings such as $W_{1,A_1B_2}\\otimes W_{2,A_2B_3}\\otimes W_{3,B_1A_3}$. The factor order is chosen so that each witness's negative eigenspace is aligned with correlations of the entangled state across copies. The expectation value is automatically nonnegative on separable states because each factor is a witness or a positive semidefinite operator, so negativity of the expectation value is a valid entanglement certificate. The accompanying proofs that the constructed operators are witnesses use positive-partial-transpose type inequalities plus the negative-eigenvalue criterion for witnesses.","core_discovery":"The paper's core discovery is that the tensor product of two entanglement witnesses is not invariant under swapping which subsystems are paired across copies, and this asymmetry can be used for detection. In Observation 1, for $\\rho=|\\psi^+\\rangle\\langle\\psi^+|$ and witnesses $W=I-X\\otimes X+Z\\otimes Z$ and $V=2|\\phi^+\\rangle\\langle\\phi^+|_{\\tau_2}$, the single-copy expectations are $\\mathrm{Tr}(W\\rho)=\\mathrm{Tr}(V\\rho)=1$, yet the crossed product $W_{AB'}\\otimes V_{BA'}$ has expectation $-1/2$ on $\\rho^{\\otimes 2}$. Observation 2 extends the phenomenon to three copies: for the Werner family $\\rho_w=w I/4+(1-w)|\\psi^+\\rangle\\langle\\psi^+|$ with $w<0.206$, every two-copy crossed pairing of the three witnesses $W_1,W_2,W_3$ is nonnegative, but the three-copy ordering $W_{1,A_1B_2}\\otimes W_{2,A_2B_3}\\otimes W_{3,B_1A_3}$ has negative expectation. The paper also shows that one factor in the product may be replaced by any positive semidefinite operator and that the construction detects tripartite entanglement using bipartite witnesses.","pith_inferences":["Our inference: the power of the crossed ordering likely comes from aligning the negative eigenvectors of different witnesses against different tensor factors of the state; a general criterion for when reordering enlarges the detected set could be expressed in terms of the support projections of the witnesses' negative eigenspaces.","Our inference: because nonnegativity on separable states is automatic for any factor permutation, the detection gain is purely about entangled states; one could test whether crossed orderings of a single witness $W\\otimes W$, rather than two different witnesses, ever detect states that $W$ alone misses.","Our inference: the gap in the written proof that $W$ and $W_1$ are witnesses can be settled by a direct numerical minimization over product-state Bloch vectors; if the minima are nonnegative, Observation 1 and the Werner three-copy example survive independently of the flawed inequality chain.","Our inference: a natural experimental extension is to implement the crossed two-copy measurement on photonic or trapped-ion Bell states and record the negative expectation value, which would give a few-copy entanglement certificate without full tomography."],"forward_implications":["Existing witnesses can be reused on few copies of an unknown state to detect entanglement that linear detection misses, as in the Bell state example where both single-copy expectations are $+1$ but the crossed two-copy expectation is $-1/2$.","For Werner states, more copies can activate detection: three-copy crossed ordering detects entanglement for $w<0.206$ even though every two-copy crossed pairing of the same witnesses fails.","A witness combined with a positive semidefinite operator under the crossed ordering detects entanglement for a family of Werner-like states, and tuning the operator extends the detection region toward the PPT boundary, approaching $a>1/\\sqrt{3}$.","Two-qubit witnesses applied to two copies of three-qubit states detect genuine tripartite entanglement of noisy W states for $c<0.406$, a wider range than the standard tripartite W-state witness's $c<0.38$ threshold.","The crossed-copy layout also realizes entanglement concentration: a suitable measurement on the crossed subsystem $BA'$ projects two copies of a full-Schmidt-rank pure state onto the same state, or onto the maximally entangled state, with positive probability."],"supporting_citations":[{"why":"Defines the aligned multi-copy witness strategy that the paper's crossed ordering complements and outperforms in the examples.","marker":"[47]"},{"why":"Justifies that V, built from the partially transposed entangled Bell state |φ+⟩⟨φ+|, is an entanglement witness.","marker":"[52]"},{"why":"Supplies the PPT separability criterion used in the proofs that W and W1 are witnesses.","marker":"[53]"},{"why":"Gives the negative-eigenvalue definition used to certify W and W1 as well-defined witnesses.","marker":"[54]"},{"why":"Introduces the Werner state family on which Observation 2 and the three-copy detection threshold w<0.206 are demonstrated.","marker":"[60]"},{"why":"Provides the standard W-state tripartite witness whose detection range c<0.38 the paper's strategy extends to c<0.406.","marker":"[65]"}],"fun_headline_variants":["Cross-copy witness swap reveals entanglement single tests miss","Nonlinear witness order: multiple copies catch more entanglement","Swap witness pairs across copies to detect hidden entanglement","Two or three copies: witness reordering finds entanglement"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything rests on each constructed operator being a true entanglement witness, never negative on a separable state, and for two of the operators the paper's written proof of that property contains an unjustified step, so as written the witness status of those operators is the load-bearing premise.","fun_headline_variants_meta":{"raw":{"variants":["Cross-copy witness swap reveals entanglement single tests miss","Nonlinear witness order: multiple copies catch more entanglement","Swap witness pairs across copies to detect hidden entanglement","Two or three copies: witness reordering finds entanglement"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000603,"raw_usage":{"total_tokens":2793,"prompt_tokens":901,"completion_tokens":1892,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":517,"completion_tokens_details":{"reasoning_tokens":1839}},"tokens_in":517,"tokens_out":1892,"duration_ms":12545,"temperature":1.0,"reasoning_tokens":1839,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T10:51:37.482238+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the minimum of $\\mathrm{Tr}(W\\rho)$ and $\\mathrm{Tr}(W_1\\rho)$ over all separable two-qubit states, e.g. by minimizing over product Bloch vectors; if either minimum is negative, the operator is not a witness and the corresponding example collapses, whereas if both are nonnegative the crossed-copy negative expectations stand.","supporting_citations":[{"cited_title":"⊗Wn such that Tr( Wρ⊗ k ent)< 0 for some entangled states ρent, while Tr( Wρ⊗ k sep) ≥ 0 for all separable states, where k depends on the local dimension and n","cited_arxiv_id":null,"evidence_quote":"Defines the aligned multi-copy witness strategy that the paper's crossed ordering complements and outperforms in the examples."},{"cited_title":"Kotowski, M","cited_arxiv_id":null,"evidence_quote":"Justifies that V, built from the partially transposed entangled Bell state |φ+⟩⟨φ+|, is an entanglement witness."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the PPT separability criterion used in the proofs that W and W1 are witnesses."},{"cited_title":"Shen, J.M","cited_arxiv_id":null,"evidence_quote":"Gives the negative-eigenvalue definition used to certify W and W1 as well-defined witnesses."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the standard W-state tripartite witness whose detection range c<0.38 the paper's strategy extends to c<0.406."}],"review_version":1}