{"id":"bd5011b1-b7a3-4341-bd1d-8a20417dbe57","arxiv_id":"2502.09345","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"One-shot dynamic coherence cost and distillation for the quantum Fourier transform are bounded by log-robustness and hypothesis-testing relative entropy, with a catalytic extension under approximately free superchannels.","lead":"This paper derives one-shot bounds for converting the quantum Fourier transform into arbitrary quantum channels using free superchannels, and for distilling quantum Fourier transforms from channels. It also bounds the catalytic cost of this conversion when the free operations may generate a small amount of coherence.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 13's upper bound uses conflicting definitions of ε′: Lemma 12 requires ε′ = ε²/(2|A0|²), but Theorem 13 sets ε′ = ε²/(2|A0|), so the constructed protocol need not meet Definition 10's error tolerance.","rationale":"The paper's central claim is that one-shot dynamic coherence cost and distillation are controlled by log-robustness and hypothesis-testing relative entropy, with catalytic bounds in Theorem 13. The most load-bearing condition for the catalytic upper bound is that Lemma 12's approximation Nεᵃ, which is within ε of N_A, can be fed into the δ-MISC construction of Eq. (81). Lemma 12 obtains this via the chain ½‖Nεᵃ−N_A‖⋄ ≤ |A0|√(2ε′), so ε′ must be ε²/(2|A0|²) for the final inequality to give exactly ε. Theorem 13 states ε′ = ε²/(2|A0|), which is off by a factor of |A0| and yields only ε√|A0|. This is not a stylistic issue: it breaks the feasibility condition in Definition 10, so the upper bound (74) is unproven as written. The reader's chosen weakest assumption, additivity of channel log-robustness, is less compelling: Eq. (77) only needs LR(F_d⊗F_l) = log d² + log l² for QFT tensor products. This can be checked directly from the Choi state (uniform coefficients of magnitude 1/(dl) over (dl)² basis elements), so it is not a fragile unproven premise. The ε′ mismatch is more central and more concrete. A single recomputation of the final chain in Lemma 12 with the theorem's ε′ settles whether the concern lands. If ε′ is corrected to ε²/(2|A0|²), the upper bound proof goes through; otherwise an alternative argument is required. The paper otherwise has reasonable internal support: Theorems 1 and 3 have explicit constructions, and the cost/distillation bounds are stated with proofs. The definitional slip in Definition 3 (Δ_B∘Θ vs Θ∘Δ_A) is real but less central, since the intended commutation relation is used consistently in Theorem 3. Overall, the appropriate disposition remains conditional: accept with the ε′ inconsistency resolved.","tokens_in":21702,"tokens_out":34328,"duration_ms":343705,"concrete_test":"Recompute the final inequality of Lemma 12 with ε′ = ε²/(2|A0|), the value stated in Theorem 13. For any |A0| > 1, the chain yields ½‖Nεᵃ−N_A‖⋄ ≤ ε√|A0| rather than ε. Concretely, take |A0| = 2 and ε = 0.1: Lemma 12 requires ε′ = 0.00125, while Theorem 13 uses ε′ = 0.0025, so the constructed Nεᵃ is only guaranteed to be 0.141-close, violating Definition 10's 0.1 tolerance. If the authors replace ε′ by ε²/(2|A0|²) in Theorem 13, the proof goes through; if they keep ε′ = ε²/(2|A0|), an additional argument is needed.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The most load-bearing gap is an internal inconsistency in Theorem 13's upper bound. Lemma 12 defines ε′ = ε²/(2|A0|²) and proves, in Eq. (73), that the constructed Nεᵃ obeys ½‖Nεᵃ−N_A‖⋄ ≤ |A0|√(2ε′) = ε. Theorem 13, however, states and uses ε′ = ε²/(2|A0|). Substituting that value gives only ½‖Nεᵃ−N_A‖⋄ ≤ |A0|√(2ε′) = ε√|A0|, which exceeds ε whenever |A0| > 1. Since Definition 10 requires the simulating channel N′ to satisfy ½‖N′−N‖⋄ ≤ ε, the superchannel Θ constructed in the proof (Eq. (81)) may not be a feasible catalytic-cost protocol for the stated error. Thus the claimed inequality (74) is not established as written. This issue is distinct from the additivity point in Eq. (77): there the proof only needs LR(F_d⊗F_l) = log d² + log l² for QFT channels, which follows directly from the uniform Choi coefficients and is not fragile.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a dynamic resource theory of quantum coherence in which classical channels are the free channels, quantum Fourier transform (QFT) channels are the golden units, and the free superchannels are maximally incoherent superchannels (MISC) and dephasing-covariant incoherent superchannels (DISC). It defines one-shot dynamic coherence cost and distillation, proves bounds relating the cost to the (smoothed) log-robustness and dephasing log-robustness (Theorems 1 and 3), and bounds distillation by hypothesis-testing relative-entropy quantities (Theorems 6 and 9). It then introduces δ-MISC operations and states two-sided bounds for the one-shot catalytic dynamic coherence cost of a channel (Theorem 13).","tokens_in":21979,"tokens_out":34869,"duration_ms":352623,"significance":"If the results hold, they give the channel log-robustness, dephasing log-robustness, and the respective hypothesis-testing quantities a direct operational meaning as one-shot dynamic coherence cost and distillation measures, with the QFT channel singled out as the natural golden unit. Theorems 1, 3, 6, and 9 are supported by explicit measure-and-prepare superchannel constructions and standard data-processing arguments, and the smoothed bounds are of the expected form. The catalytic section is the most ambitious part, but Theorem 13 as written contains an internal parameter inconsistency and a quantifier gap; these are local in nature and appear fixable, but they currently prevent the paper's central catalytic claim from being accepted as proven.","major_comments":[{"comment":"There is a load-bearing inconsistency in the definition of ε′. Lemma 12 defines ε′ = ε²/(2|A0|²) and, in Eq. (73), proves ½‖Nε_A − N_A‖⋄ ≤ |A0|√(2ε′) = ε. Theorem 13, however, states and uses ε′ = ε²/(2|A0|), both in the theorem statement and in the final line of its proof. Substituting this larger ε′ into Eq. (73) gives only ½‖Nε_A − N_A‖⋄ ≤ |A0|√(2ε′) = ε√|A0|, which exceeds ε whenever |A0| > 1. Since Definition 10 requires the simulating channel N′_A to satisfy ½‖N′_A − N_A‖⋄ ≤ ε, the superchannel Θ constructed in Eq. (81) may not be a feasible catalytic-cost protocol for the stated error, and the upper bound in Eq. (74) is not established as written. The proof can be repaired by using Lemma 12's ε′ = ε²/(2|A0|²) consistently throughout Theorem 13.","section":"Sec. 5, Theorem 13 vs. Lemma 12"},{"comment":"The lower-bound proof in Eq. (77) starts with an optimal protocol for the catalytic cost and fixes the catalyst size l of that optimal protocol. The theorem statement, however, asserts that there exists one l with l² ≥ 1 + 1/δ for which both the upper and lower bounds hold. No argument is given that the optimal protocol can be chosen with such an l, nor that the lower bound for the l used in the upper-bound construction follows from Eq. (77). As written, the two-sided quantified statement of Theorem 13 is therefore not proven; the theorem needs to be rephrased so that the existence of l in the upper bound and the validity of the lower bound are related by an explicit argument.","section":"Sec. 5, Theorem 13, lower bound"},{"comment":"In the upper-bound proof, Eq. (78) writes M_{ε′}^{AB} ≤ 2^{LR_{ε′}(N^ε_A ⊗ F_l)} P_{AB}, and Eq. (79) keeps the smoothing parameter centered at N^ε_A. Lemma 12, however, gives LR(M_{ε′}^{AB}) ≤ LR_{ε′}(N_A ⊗ F_l), with the smoothing centered at the original channel N_A, not at the approximate channel N^ε_A. The inequality with N^ε_A in the exponent therefore does not follow from Eq. (60). The chain is repaired by replacing N^ε_A with N_A in the exponents of Eqs. (78) and (79), after which the subsequent use of p ≥ 1 − 2ε′ is valid.","section":"Sec. 5, Eq. (78)–(79)"}],"minor_comments":[{"comment":"The defining equation of DISC should read Δ_B ∘ Θ = Θ ∘ Δ_A; as printed, Δ_B ∘ Θ = Θ ∘ Δ_B is not type-correct for Θ: A → B.","section":"Definition 3, Eq. (2)"},{"comment":"The first coefficient in the expression for Θ ∘ Δ[N] should be (Tr(J_{F_{d0}}J_{Δ[N]}) − 1/d0²), not (1 − Tr(J_{F_{d0}}J_{Δ[N]})); with the printed coefficient the displayed equality to Δ[N_ε] is incorrect.","section":"Proof of Theorem 3, Eq. (22)"},{"comment":"The notation Ω_B[\\tilde M_{ε′}^{AB}] should be explicitly defined as the action of the B-side superchannel (id_A ⊗ Ω_B) on the bipartite channel; as written, Ω_B by itself is defined only on channels with input/output systems B0/B1, not on channels with input A0B0 and output A1B1.","section":"Lemma 12, Eq. (64)–(67)"},{"comment":"The claim that d_n ≥ 2 in the regularization proof is false when LR^ε(N^{⊗n}) = 0, e.g., for classical N; in that case the expression log((d_n/(d_n−1))²) is undefined. This case should be handled separately.","section":"Corollary 4"},{"comment":"There is a typo 'to easure' for 'to ensure', and the displayed bound '1/l²+1 ≤ δ' should read '1/(l²−1) ≤ δ' to match the preceding derivation.","section":"End of Theorem 13 proof"},{"comment":"The proof cites Ref. [53] for additivity of the channel log-robustness under tensor products. For the specific tensor product F_d ⊗ F_l used here, additivity follows directly from the uniform Choi coefficients of QFT channels, so the general additivity citation is not a blocker; a direct one-line proof would make the argument self-contained.","section":"Eq. (77), additivity citation"}],"recommendation":"major_revision","confidential_remarks":"The paper contains several presentation-level typos (Eq. (2), Eq. (22), Eq. (78)) that currently make verification more difficult than necessary. The main technical issue is concentrated in Theorem 13: the ε′ parameter mismatch is a genuine internal inconsistency, and the quantifier structure of the two-sided catalytic bound needs to be made precise. These problems appear fixable without changing the overall framework, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is a genuine contribution to dynamic coherence resource theory. The one-shot cost and distillation bounds under MISC and DISC are clean, correctly use standard superchannel techniques, and give the QFT channel a clear operational role. The genuinely new material is the catalytic cost under δ-MISC (Theorem 13), which is the reason to read the paper.\n\nThe catalytic section has a load-bearing flaw. Lemma 12 is stated and proved with ε′ = ε²/(2|A0|²), and its final inequality uses |A0|√(2ε′) = ε. Theorem 13, however, sets ε′ = ε²/(2|A0|). Substituting that value gives |A0|√(2ε′) = ε√|A0|, which is larger than ε for |A0| > 1. So the superchannel Θ constructed in the proof of Theorem 13 does not, as written, guarantee the error tolerance required by Definition 10. The theorem may be salvageable by redefining ε′ consistently, but as stated the proof does not establish the claimed upper bound. This is not a minor typo; it is the central new result.\n\nThe reader's concern about additivity of log-robustness under tensor products is, on inspection, not fragile. The proof only needs LR(F_d ⊗ F_l) = log d² + log l², which follows from the uniform Choi coefficients of QFT channels. So that part is fine.\n\nOther soft spots are smaller. The DISC definition in Eq. (2) as written is a type error: it should read ∆_B ∘ Θ = Θ ∘ ∆_A, not ∆_B on both sides. The paper also underplays its proximity to the general one-shot yield-cost relations in Refs. [51,63]; the catalytic result is the clear point of separation. The proof of Theorem 13's upper bound uses a somewhat crude +2 estimate, but that is consistent with the prior art and not a concern.\n\nThe paper is written clearly and the author engages honestly with the existing literature, including noting an alternative state-based formulation in Ref. [53]. The framework of taking classical channels as free and QFT as the golden unit is well motivated.\n\nShould it go to peer review? Yes. The core ideas are sound, the framework will be useful to the resource-theory community, and the main theorem is very likely repairable. A careful referee should ask for the ε′ inconsistency to be fixed, the DISC definition corrected, and the relation to Refs. [51,63] made more explicit. After those changes the paper would be a solid addition. I would not cite it in its current form, but I would bring it to a reading group to discuss the catalytic construction.","headline":"A solid one-shot dynamic coherence paper whose main new theorem currently has a fixable but real error in the epsilon-prime bookkeeping; the rest of the framework is sound and worth refereeing.","tokens_in":22474,"tokens_out":5388,"would_cite":false,"duration_ms":51019,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the one-shot dynamic coherence cost and distillation of quantum channels are bounded by the channel's log-robustness and hypothesis-testing relative entropy, with catalytic cost also controlled by log-robustness.","keywords":["quantum resource theory","dynamic coherence","quantum channels","superchannels","quantum Fourier transform","one-shot manipulation","catalysis","log-robustness"],"falsifier":"Compute the channel log-robustness $\\widehat{LR}(\\mathcal{N}\\otimes \\mathcal{F}_l)$ for a non-classical channel $\\mathcal{N}$ and a QFT channel $\\mathcal{F}_l$ in dimension 2 or 3; if it is strictly less than $\\widehat{LR}(\\mathcal{N}) + \\log l^2$, the additivity assumption behind the lower bound of Theorem 13 is violated.","tokens_in":21518,"feed_emoji":"⚛️","tokens_out":11729,"duration_ms":94793,"temperature":0.7,"pith_summary":"The paper develops a dynamic resource theory of quantum coherence in which classical channels are free and two classes of free superchannels, maximally incoherent superchannels (MISC) and dephasing-covariant incoherent superchannels (DISC), are allowed. Its central claim is that the one-shot cost of using quantum Fourier transform (QFT) channels to simulate an arbitrary quantum channel is bounded above and below by the channel's log-robustness of coherence, while the one-shot distillation of QFT from a channel is bounded by its hypothesis-testing relative entropy. The paper further shows that when a catalyst is allowed and the free superchannels are relaxed to $\\delta$-MISC, the catalytic cost remains controlled by the same log-robustness measure. If these bounds are correct, coherence measures for quantum channels acquire a concrete operational meaning: they tell how many QFT 'golden units' are needed to implement a given operation, or can be extracted from it.","feed_headline":"Channel simulation cost is governed by coherence log-robustness","feed_subtitle":"New bounds tie channel coherence measures to how many QFT gates are needed, even when a catalyst is used.","key_machinery":"The central object is the quantum Fourier transform (QFT) channel $\\mathcal{F}_d$, chosen as the golden unit because it has maximal log-robustness $\\log d^2$ and, under MISC, can generate any channel. The argument is carried by two families of free superchannels: MISCs, which map classical channels to classical channels, and DISCs, which commute with the dephasing superchannel. The key constructions are explicit superchannels (Eqs. (15), (21), and (81)) that use the overlap of a channel's Choi matrix with the QFT's Choi matrix to interpolate between a target channel and a classical channel, so the resource cost is expressed directly through the feasibility condition defining log-robustness. The proofs also rely on the Choi-matrix representation of channel max-relative entropy and on the additivity of channel log-robustness under tensor products, the latter being assumed from prior work.","core_discovery":"The paper proves explicit two-sided bounds for the one-shot dynamic coherence cost and one-sided bounds for distillation. For MISC, the cost satisfies $\\widehat{LR}^\\epsilon(\\mathcal{N}) \\le c^{(1)}_{\\epsilon,\\mathrm{MISC}}(\\mathcal{N}) < \\widehat{LR}^\\epsilon(\\mathcal{N}) + \\log(d_0/(d_0-1))^2$, and under DISC the identical form holds with the dephasing log-robustness $\\widehat{LR}^\\epsilon_\\Delta$ in place of $\\widehat{LR}^\\epsilon$. For distillation, it proves $d^{(1)}_{\\epsilon,\\mathrm{MISC}}(\\mathcal{N}) \\le C_H^{2\\epsilon}(\\mathcal{N})$ and $d^{(1)}_{\\epsilon,\\mathrm{DISC}}(\\mathcal{N}) \\le C_{H,\\Delta}^{2\\epsilon}(\\mathcal{N})$, where $C_H$ and $C_{H,\\Delta}$ are the hypothesis-testing relative entropy of coherence and its dephasing variant. In the catalytic setting with $\\delta$-MISC, Theorem 13 gives $\\widehat{LR}^{\\epsilon'}(\\mathcal{N}\\otimes \\mathcal{F}_l) - \\log[l^2(1-2\\epsilon')] + 2 \\ge \\tilde{c}^{(1)}_{\\epsilon,\\delta}(\\mathcal{N}) \\ge \\widehat{LR}^\\epsilon(\\mathcal{N}\\otimes \\mathcal{F}_l) - \\log[l^2(1+\\delta)]$, with $\\epsilon' = \\epsilon^2/(2|A_0|)$. These are obtained by constructing explicit superchannels that achieve the bounds and by using monotonicity of the resource measures under free superchannels. If correct, the results give a direct operational interpretation of the log-robustness, dephasing log-robustness, and hypothesis-testing relative entropy of quantum channels.","pith_inferences":["The paper leaves open whether the gap between the upper and lower catalytic bounds can be closed by optimizing the catalyst dimension $l$; an explicit optimization over $l$ would sharpen the resource-theoretic picture.","Because the QFT is used as a universal golden unit, the same machinery might transfer to other channel resource theories (e.g., magic channels or imaginarity) by substituting the corresponding golden channel, although the paper does not explore this.","A direct numerical test of additivity of channel log-robustness in low dimensions would independently probe the main assumption behind the catalytic lower bound.","The explicit superchannel constructions are, in principle, algorithmic recipes for simulating any channel with QFT resources, but their circuit complexity is not addressed here."],"forward_implications":["The log-robustness of a quantum channel becomes an operationally meaningful resource measure: up to small gaps, it determines how many QFT channels are needed to simulate the channel in one shot.","The regularized cost equals the asymptotic log-robustness (Corollary 4), so the theory is asymptotically exact in the cost direction.","The hypothesis-testing relative entropy of coherence limits one-shot QFT distillation, providing a monotone that quantifies distillable dynamic coherence.","Allowing a catalyst and slightly coherence-generating superchannels ($\\delta$-MISC) does not change the governing resource measure: the cost is still controlled by log-robustness.","The dephasing versions of these measures are singled out when DISC superchannels are used, giving a quantitative distinction between the two free-operation sets."],"supporting_citations":[{"why":"Defines the dynamic coherence resource theory, the MISC and DISC superchannels, and provides the additivity of channel log-robustness used in Theorem 13.","marker":"[53]"},{"why":"Supplies the framework of superchannels and the diamond-norm contractivity used in the explicit constructions.","marker":"[21]"},{"why":"Establishes the Choi-matrix representation of channel max-relative entropy used to express log-robustness.","marker":"[75]"},{"why":"Provides the diamond-norm-to-Choi-norm bound and the notion of $\\delta$-MISC-style free superchannels used in Lemma 12 and Theorem 13.","marker":"[34]"},{"why":"Gives the characterization of maximally incoherent operations used to show QFT channels can generate any channel via a MISC (Appendix A).","marker":"[15]"},{"why":"Provides the hypothesis-testing relative entropy and its data-processing inequality, which underpin the distillation bounds.","marker":"[72]"}],"fun_headline_variants":["One-shot coherence cost pinned by log-robustness","Dynamic coherence: cost bounds via log-robustness","Catalytic bounds tie channel cost to coherence","One-shot dynamic coherence: tight cost bounds","Coherence log-robustness sets channel cost"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The results hinge on the assumption that the coherence log-robustness of two channels combined in parallel is simply the sum of their individual log-robustnesses; this is used without proof in the catalytic cost bound.","fun_headline_variants_meta":{"raw":{"variants":["One-shot coherence cost pinned by log-robustness","Dynamic coherence: cost bounds via log-robustness","Catalytic bounds tie channel cost to coherence","One-shot dynamic coherence: tight cost bounds","Coherence log-robustness sets channel cost"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000244,"raw_usage":{"total_tokens":1638,"prompt_tokens":1157,"completion_tokens":481,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":773,"completion_tokens_details":{"reasoning_tokens":407}},"tokens_in":773,"tokens_out":481,"duration_ms":4845,"temperature":1.0,"reasoning_tokens":407,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T21:51:58.652737+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the channel log-robustness $\\widehat{LR}(\\mathcal{N}\\otimes \\mathcal{F}_l)$ for a non-classical channel $\\mathcal{N}$ and a QFT channel $\\mathcal{F}_l$ in dimension 2 or 3; if it is strictly less than $\\widehat{LR}(\\mathcal{N}) + \\log l^2$, the additivity assumption behind the lower bound of Theorem 13 is violated.","supporting_citations":[{"cited_title":"Saxena, E","cited_arxiv_id":null,"evidence_quote":"Defines the dynamic coherence resource theory, the MISC and DISC superchannels, and provides the additivity of channel log-robustness used in Theorem 13."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the framework of superchannels and the diamond-norm contractivity used in the explicit constructions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the Choi-matrix representation of channel max-relative entropy used to express log-robustness."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the diamond-norm-to-Choi-norm bound and the notion of $\\delta$-MISC-style free superchannels used in Lemma 12 and Theorem 13."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the characterization of maximally incoherent operations used to show QFT channels can generate any channel via a MISC (Appendix A)."},{"cited_title":"Wang and R","cited_arxiv_id":null,"evidence_quote":"Provides the hypothesis-testing relative entropy and its data-processing inequality, which underpin the distillation bounds."}],"review_version":1}