{"id":"eb71a14d-0468-452e-940d-e72d56c9804d","arxiv_id":"1908.03038","paper_version":6,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A proof that Gaussian coherent-state ensembles maximize capacity for Gaussian observables, solving the accessible-information conjecture via ensemble-observable duality.","lead":"This paper proves that the accessible information of a Gaussian quantum ensemble is achieved by a multimode heterodyne measurement. It also shows that the classical capacity of a Gaussian measurement is achieved by a Gaussian ensemble of coherent states, resolving a conjecture from the 1970s.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the central argument holds; only a non-load-bearing typo in the dual-observable change of variables should be corrected.","rationale":"I read the paper as proving two claims: the constrained chi-capacity formula for gauge-covariant Gaussian observables (Theorem 1) and the accessible-information formula for Gaussian ensembles (Theorem 2). The proof of Theorem 1 reduces to the vacuum-attains-minimum-output-entropy result for the noisy heterodyne measurement, which is imported from the published work [2]. This is the same weakest assumption identified by the reader; it is an external reliance rather than an internal inconsistency. The main derivations survive scrutiny: the maximum-entropy bound in Eq. (17) is legitimate for the output distributions of gauge-invariant inputs; the capacity upper bound plus the coherent-state ensemble gives the claimed value; and the duality argument in Theorem 2 uses Proposition 4 correctly, with the dual observable falling into the class covered by Theorem 1. I checked the algebraic step leading to the Gaussian form of the dual observable in some detail. The displayed definition of \\tilde z after Eq. (34) appears to be missing a square root over the factor (\\tilde\\Sigma+I_s); as written, the quadratic form does not complete to the stated Gaussian. However, Eq. (37) and the later use of T = \\sqrt{\\tilde\\Sigma(\\tilde\\Sigma+I_s)} show that the intended K is T\\Sigma^{-1}, and with that replacement the cross-term identity, the determinant identity, and the normalization of the POVM all work. Thus the proof is repairable and the central claim is not undermined. For these reasons I would leave the reader's ACCEPT verdict unchanged, while noting that the typo should be fixed in a revised version.","tokens_in":11652,"tokens_out":51085,"duration_ms":512004,"concrete_test":"Re-derive the passage from Eq. (34) to Eq. (35) with K = T\\Sigma^{-1}, T = \\sqrt{\\tilde\\Sigma(\\tilde\\Sigma+I_s)}: verify the identity A K = \\sqrt{(I+\\tilde\\Sigma^{-1})^{-1}} N^{-1} for A = \\tilde N^{-1} and that the exponent completes to -(u-Kz)^* A (u-Kz); also confirm det(I+(\\tilde N+I)^{-1}\\tilde\\Sigma) = det(\\tilde\\Sigma+I)/\\det(N+I). If the corrected K fails these identities, the upper bound in Theorem 2 would not follow.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After checking the main argument, I find no load-bearing defect. The central claim depends on the generalized Wehrl-type minimum-output-entropy result for the noisy heterodyne POVM (11), imported from [2]; this is an external but published result, so the dependence is a verification burden rather than a demonstrated gap. The duality proof of Theorem 2 is internally consistent once one corrects a typo in the change of variables after Eq. (34): the cross-term algebra requires \\tilde z = \\sqrt{\\tilde\\Sigma(\\tilde\\Sigma+I_s)}\\,\\Sigma^{-1} z, not \\sqrt{\\tilde\\Sigma}(\\tilde\\Sigma+I_s)\\,\\Sigma^{-1} z; Eq. (37) itself uses T = \\sqrt{\\tilde\\Sigma(\\tilde\\Sigma+I_s)}. With that correction, the completion of the square, the determinant identity, and the POVM normalization all check out. This typo does not affect the theorem's statement.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies multimode bosonic Gaussian systems with global gauge symmetry. Theorem 1 gives an explicit formula for the chi-capacity of a gauge-covariant Gaussian observable under a covariance constraint, namely log det(I_s + (N + I_s)^{-1} Sigma), and identifies the Gaussian coherent-state ensemble as an optimizer. Theorem 2 uses a continuous-variable ensemble-observable duality to show that the accessible information of a gauge-invariant Gaussian ensemble equals the same quantity and is attained by any scaled multimode heterodyne measurement. The proofs combine a classical maximum-entropy bound, a Wehrl-type minimum-output-entropy result imported from the author's earlier work, and an infinite-dimensional duality theorem proved in Section 4.","tokens_in":11834,"tokens_out":17414,"duration_ms":171290,"significance":"If correct, the paper settles an old conjecture of Holevo and Belavkin--Stratonovich on accessible information of Gaussian ensembles and extends earlier single-mode heterodyne capacity results to arbitrary multimode gauge-covariant Gaussian observables. The explicit, parameter-free formulas and the infinite-dimensional duality framework (Propositions 3 and 4) are valuable contributions. The main external input is the generalized Wehrl-type entropy-minimization result from [2]; I verified that the coherent-state ensemble in Theorem 1 attains the upper bound directly, so the additional citation to [15] is not load-bearing. The proof is coherent and the remaining issues are local typographical and presentational.","major_comments":[],"minor_comments":[{"comment":"The change of variables following Eq. (34) contains a typo: the definition of \\tilde z should be \\tilde z = \\sqrt{\\tilde\\Sigma(\\tilde\\Sigma + I_s)} \\Sigma^{-1} z, not \\sqrt{\\tilde\\Sigma}(\\tilde\\Sigma + I_s) \\Sigma^{-1} z. With this correction, K equals the matrix T used in Eq. (37), and the completion of the square, the determinant identity, and the normalization all check out.","section":"Sec. 3, proof of Theorem 2 (after Eq. (34))"},{"comment":"The approximation sequence {P_n} should be specified to be finite-rank (or trace-class) projections so that Tr P_n is finite and the normalization in Eq. (48) is well-defined; as written, 'nondecreasing sequence of projections P_n \\uparrow I' does not guarantee this.","section":"Sec. 4, proof of Proposition 4"},{"comment":"The sentence 'Its achievability follows from Proposition 2 of the recent paper [15]' is misleading. The explicit Gaussian coherent-state ensemble (20) attains the upper bound directly once one has the maximum-entropy bound (17) and the Wehrl-type minimum from [2]; Proposition 2 of [15] is not needed for the value of the capacity. This should be clarified so the proof does not appear less self-contained than it is.","section":"Sec. 3, Theorem 1 proof (Eqs. (21)-(24))"},{"comment":"The notation in Eq. (6) is missing a subscript: 'TrρΛD(w)' should read 'Tr ρ_Λ D(w)', consistent with the definition of the gauge-invariant Gaussian state.","section":"Sec. 2, Eq. (6)"},{"comment":"The page range in reference [1] appears garbled ('pp. 1050-2947'); please verify the correct pagination.","section":"References"}],"recommendation":"minor_revision","confidential_remarks":"No concerns beyond the minor revisions listed. The paper leans on the author's own prior results, but those results are published and properly cited, and the central claims do not reduce to their own inputs. The typo in the change of variables after Eq. (34) is cosmetic and easily fixed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this is the real thing: Theorem 2 proves the old conjecture that the accessible information of a gauge-invariant Gaussian ensemble is achieved by multimode heterodyne, with explicit formula log det(I_s + (N+I_s)^{-1} Sigma). Second, the proof is not self-contained; it imports two heavy results from the author's own earlier papers, the generalized Wehrl minimum-output-entropy theorem and the Gaussian optimizer conjecture. That is a verification burden, not a circularity, because those results are published and independent.\n\nWhat's actually new: Theorem 1 gives the classical capacity of an arbitrary gauge-covariant Gaussian observable, attained on a coherent-state Gaussian ensemble; this generalizes one-mode heterodyne results. Theorem 2 uses the ensemble-observable duality of Sec. 4 to turn the accessible-information problem into a capacity problem, then applies Theorem 1. The duality section is a genuine infinite-dimensional formulation, not just finite-dimensional handwaving.\n\nWhere are the soft spots? Two, both minor. (1) The achievability part of Theorem 1 rests on Proposition 2 of the author's 2020 paper with Kuznetsova [15]; the paper says all assumptions are fulfilled but doesn't spell them out. A referee should check that. (2) There is a typo in the change of variables after Eq. (34): the correct substitution is \\tilde z = sqrt{\\tilde Sigma (\\tilde Sigma + I_s)} Sigma^{-1} z, not sqrt{\\tilde Sigma}(\\tilde Sigma + I_s) Sigma^{-1} z. The subsequent algebra, including Eq. (37), uses the correct form, so the theorem statement is unaffected.\n\nThe infinite-dimensional duality arguments in Sec. 4 are sketched—approximation by finite-rank projections and lower semicontinuity of mutual information—but they follow known patterns and look sound. The self-citations are to published theorems, not to re-fitted results, so I don't see a circularity problem.\n\nBottom line: this is a solid paper that settles a 1970s conjecture. It deserves a serious referee and should be accepted after minor corrections. I'd cite it and would bring it to a reading group focused on Gaussian quantum information.","headline":"Holevo settles the Gaussian ensemble accessible-information conjecture by proving a Gaussian optimizer theorem for observables; the proof leans on his own prior Gaussian-optimizer results, but those are external published theorems and the argument holds up.","tokens_in":12261,"tokens_out":1558,"would_cite":true,"duration_ms":15891,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P45","94A17"],"pacs":[],"model":"deepseek-v4-flash","headline":"For phase-insensitive Gaussian systems, the same determinant formula gives the classical capacity of every Gaussian measurement and the accessible information of every Gaussian ensemble, with multimode heterodyne as an optimizer.","keywords":["Gaussian observables","accessible information","heterodyne measurement","coherent states","Gaussian optimizer conjecture","ensemble-observable duality","phase-insensitive systems","continuous-variable quantum information"],"falsifier":"For a single mode with $\\Sigma=1$ and $N=0$, the theorem predicts accessible information $\\log 2$; a numerical search over finite-outcome measurements on a truncated photon-number space that finds any measurement exceeding $\\log 2$ would refute Theorem 2.","tokens_in":11477,"feed_emoji":"","tokens_out":23002,"duration_ms":220622,"temperature":0.7,"pith_summary":"This paper establishes two maximization theorems for continuous-variable quantum information. For any gauge-covariant multimode Gaussian observable—a phase-insensitive measurement built from displaced Gaussian states, including noisy heterodyne detection—the classical capacity under an input covariance constraint is attained by a Gaussian ensemble of coherent states and equals $\\log\\det(I_s+(N+I_s)^{-1}\\Sigma)$. For a phase-insensitive Gaussian ensemble, the accessible information—the maximum mutual information over all possible measurements—is shown to be the same quantity, attained by the multimode heterodyne measurement. This settles a conjecture from the early 1970s and exhibits the optimal measurements as highly non-unique.","feed_headline":"Heterodyne is optimal for phase-insensitive Gaussian ensembles","feed_subtitle":"A single determinant formula gives the classical capacity and settles a decades-old accessible-information conjecture.","key_machinery":"The load-bearing mechanism is the continuous-variable ensemble–observable duality. Given an ensemble with average state $\\bar\\rho_E$, the dual observable is $M'(dz)=\\bar\\rho_E^{-1/2}\\rho_z\\bar\\rho_E^{-1/2}\\pi(dz)$; switching to the dual pair preserves mutual information, so the accessible information is bounded by the capacity of a single observable. For Gaussian ensembles this dual observable is again Gaussian, so Theorem 1 applies; the calculation closes with the determinant identity $\\det(I_s+(\\tilde N+I_s)^{-1}\\tilde\\Sigma)=\\det(I_s+(N+I_s)^{-1}\\Sigma)$, where $\\tilde\\Sigma=\\Sigma+N$. Theorem 1 itself rests on a generalized Wehrl-type minimum-output-entropy theorem: among all states, the vacuum minimizes the output entropy of the noisy heterodyne measurement.","core_discovery":"The central claim is that one determinant formula governs both sides of the measurement duality. For a phase-insensitive (gauge-covariant) Gaussian observable with noise matrix $N$ and input covariance constraint $\\Sigma$, the classical capacity is $C_\\chi=\\log\\det(I_s+(N+I_s)^{-1}\\Sigma)$, attained on the Gaussian ensemble of coherent states with covariance $\\Sigma$. For the phase-insensitive Gaussian ensemble with displacement covariance $\\Sigma$ and thermal noise $N$, the accessible information equals the same number, and it is attained by the multimode heterodyne measurement $M_*(dz)=|z\\rangle\\langle z| d^{2s}z/\\pi^s$, as well as by every nonsingular linear rescaling $D(Kz)|0\\rangle\\langle0|D(Kz)^\\dagger$. This is a global optimality statement, not a local one, and it answers a conjecture dating from the early 1970s.","pith_inferences":["If the duality argument can be extended to squeezed or rotated Gaussian ensembles, a similar determinant formula should govern their accessible information; testing this would show whether phase-insensitive symmetry is essential or merely convenient.","The explicit optimizers suggest a practical benchmark: in an $s$-mode optical setup, heterodyne detection should extract exactly $\\log\\det(I_s+(N+I_s)^{-1}\\Sigma)$ nats from a displaced-thermal ensemble, a value a tabletop experiment could check.","The capacity formula may also serve as an upper bound for information gained by any phase-insensitive receiver used in continuous-variable quantum key distribution, since such receivers are modelled by Gaussian observables.","The high degeneracy of the maximizers hints that small non-Gaussian perturbations of the ensemble may not destroy saturation of the bound; exploring this robustness would delimit how universal the Gaussian-optimizer principle is."],"forward_implications":["The energy-constrained classical capacity of any phase-insensitive Gaussian measurement is now explicitly computable by maximizing $\\log\\det(I_s+(N+I_s)^{-1}\\Sigma)$ over covariance matrices; for diagonal Hamiltonians and noise this reduces to a water-filling formula.","Because the same expression bounds the mutual information of every observable acting on a Gaussian ensemble, heterodyne detection is globally optimal among all measurements, not merely among Gaussian ones.","Every nonsingular linear rescaling of the heterodyne measurement attains the same value, so the optimal measurement is highly degenerate.","Because measurement channels destroy quantum correlations, the additivity obstacle does not arise, so the computed quantity is the true classical capacity of the observable."],"supporting_citations":[{"why":"It proves the minimum-output-entropy-at-vacuum result for noisy heterodyne measurements that supplies the upper bound in Theorem 1.","marker":"[2]"},{"why":"It gives the general solution of the Gaussian optimizer conjecture from which the entropy-minimization result is imported as a limiting case.","marker":"[3]"},{"why":"It states the original conjecture that heterodyne measurement maximizes accessible information of a Gaussian ensemble.","marker":"[6]"},{"why":"It defines the multimode heterodyne measurement as the coherent-state projector observable.","marker":"[13]"},{"why":"It provides the covariance argument that reduces the capacity of irreducibly covariant quantum-classical channels to a maximum-entropy minus minimum-entropy expression.","marker":"[14]"},{"why":"It supplies the achievability result used to show the capacity bound is attained by a Gaussian ensemble.","marker":"[15]"}],"fun_headline_variants":["One determinant formula settles Gaussian capacity and info","Heterodyne proves optimal for Gaussian info","Gaussian capacity and info: heterodyne wins","Coherent states and heterodyne: Gaussian optimizers proven","Determinant formula resolves Gaussian conjecture"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes that the vacuum state, containing no photons, gives the smallest possible output entropy for every noisy heterodyne measurement; if some other state produced smaller output entropy, the claimed capacity formula would understate the true value.","fun_headline_variants_meta":{"raw":{"variants":["One determinant formula settles Gaussian capacity and info","Heterodyne proves optimal for Gaussian info","Gaussian capacity and info: heterodyne wins","Coherent states and heterodyne: Gaussian optimizers proven","Determinant formula resolves Gaussian conjecture"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000374,"raw_usage":{"total_tokens":1924,"prompt_tokens":802,"completion_tokens":1122,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":418,"completion_tokens_details":{"reasoning_tokens":1060}},"tokens_in":418,"tokens_out":1122,"duration_ms":8707,"temperature":1.0,"reasoning_tokens":1060,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:24:53.221169+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a single mode with $\\Sigma=1$ and $N=0$, the theorem predicts accessible information $\\log 2$; a numerical search over finite-outcome measurements on a truncated photon-number space that finds any measurement exceeding $\\log 2$ would refute Theorem 2.","supporting_citations":[{"cited_title":"Majorization and additivity for multimode bosonic Gaussian channels,","cited_arxiv_id":null,"evidence_quote":"It proves the minimum-output-entropy-at-vacuum result for noisy heterodyne measurements that supplies the upper bound in Theorem 1."},{"cited_title":"A Solution of Ga ussian Optimizer Conjecture for Quantum Channels,","cited_arxiv_id":null,"evidence_quote":"It gives the general solution of the Gaussian optimizer conjecture from which the entropy-minimization result is imported as a limiting case."},{"cited_title":"On the Mathematical Theory of Quantum Communic ation Channels, Probl. Inform. Transmission ,","cited_arxiv_id":null,"evidence_quote":"It states the original conjecture that heterodyne measurement maximizes accessible information of a Gaussian ensemble."},{"cited_title":"Quantum statistics of homodyne an d hetero- dyne detection","cited_arxiv_id":null,"evidence_quote":"It defines the multimode heterodyne measurement as the coherent-state projector observable."},{"cited_title":"On the constrained classical capacity of infinite- dimensional covariant channels,","cited_arxiv_id":null,"evidence_quote":"It provides the covariance argument that reduces the capacity of irreducibly covariant quantum-classical channels to a maximum-entropy minus minimum-entropy expression."},{"cited_title":"Information capacity of con tinu- ous variable measurement channel,","cited_arxiv_id":null,"evidence_quote":"It supplies the achievability result used to show the capacity bound is attained by a Gaussian ensemble."}],"review_version":1}