{"id":"4f9298da-174a-4df5-83bb-caf670963b94","arxiv_id":"1909.01906","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For inclusions of II₁ factors and finite-dimensional algebras, the logarithmic Pimsner-Popa index equals the supremum, over all states and all Rényi parameters p in [1/2,∞], of the sandwiched Rényi relative entropy to the subalgebra.","lead":"This paper proves that a classic number from operator algebra theory, the subfactor index measuring the relative size of two nested algebras, is exactly the largest quantum information distance from the larger algebra to the smaller one, for every Rényi parameter p. The identity yields new bounds on how quickly open quantum systems lose coherence, uniformly in the ancilla dimension.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4.1's additive bound D(T_tρ||N) ≤ 2e^{-λt}+D_2(ρ||N)/2 is false; the proof's final inequality fails for states with large D_2. The abstract's exponential bound is the correct statement, so the decoherence-time estimate needs re-derivation.","rationale":"The reader's weakest assumption was the Pimsner-Popa equality (12), a genuine scope limitation but not an error within the stated classes of II₁ factors and finite-dimensional algebras. My stress-test found a more concrete internal inconsistency: the additive bound in Theorem 4.1 is false, as demonstrated by an elementary finite-dimensional example at t=0. This does not invalidate Theorem A or the abstract's exponential decay bound—the proof is repaired by using 2log(1+x)≤2x—but it means the paper as written contains a false theorem in the decoherence section. Since the intended result is recoverable with the correction already present in the abstract, the verdict remains CONDITIONAL (unchanged), with the additional required revision of Theorem 4.1 and Corollary 4.2. I partially agree with the reader: both of us flag the decoherence section, but the reader treated (24) as a display error, whereas it is mathematically false; conversely, I do not dispute the cited equality (12) within the stated scope.","tokens_in":28019,"tokens_out":23140,"duration_ms":194919,"concrete_test":"Fix n=100 and take M=M_100 with normalized trace, N=C1, and T_t=e^{-t}(id−E)+E (self-adjoint, λ=1, fixed-point algebra C1). Let ρ=100|ψ><ψ| be the pure-state density. Evaluate the claimed inequality (24) at t=0: D(ρ||N)=log 100≈4.605 and D_2(ρ||N)=log 100≈4.605, so the right-hand side is 2+4.605/2=4.303, violating (24). This directly confirms the proof's faulty step 2log(1+ab)≤2a+log b at a=1, b=100.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In §4, Theorem 4.1 asserts (24): D(T_tρ||N) ≤ 2e^{-λt}+D_2(ρ||N)/2. The proof uses D(T_tρ||N) ≤ 2log‖T_tρ‖_{L^2_1} ≤ 2log(1+e^{-λt}‖ρ‖_{L^2_1}) and then claims this is ≤ 2e^{-λt}+D_2/2. Since D_2(ρ||N)=2log‖ρ‖_{L^2_1}, this last step asserts 2log(1+ab) ≤ 2a + log b with a=e^{-λt}∈[0,1] and b=‖ρ‖_{L^2_1}≥1. That inequality is false: take a=1, b=100, giving LHS≈9.21 and RHS≈6.61. A concrete counterexample to the stated theorem: let M=M_n with normalized trace, N=C1, ρ=n|ψ><ψ|, and T_t=e^{-t}(id−E)+E, a self-adjoint quantum Markov semigroup with spectral gap 1 and N as fixed-point algebra. At t=0, T_0ρ=ρ, D(ρ||N)=log n, and D_2(ρ||N)=log n, so (24) requires log n ≤ 2+(log n)/2, which fails for n>e^4≈54.6. The abstract's bound D ≤ 2exp(−λt+D_2/2) is correct and is obtained by replacing the final step with 2log(1+ab)≤2ab=2exp(−λt+D_2/2); this also yields the stated decoherence time (5). Thus the application survives, but Theorem 4.1 and Corollary 4.2 (25) as printed are false and need correction.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the Pimsner-Popa index λ(M:N) for inclusions N⊂M of finite von Neumann algebras and connects it to sandwiched Rényi relative entropies. The central result, Theorem 3.1, states that for inclusions of II₁ factors or finite-dimensional von Neumann algebras, and for every 1/2≤p≤∞, one has −log λ(M:N) = sup_ρ D_p(ρ||E(ρ)) = sup_ρ D_p(ρ||N), where E is the trace-preserving conditional expectation. The authors introduce D_p(M||N) and a completely bounded version D_{p,cb}(M||N), prove that the latter equals log[M:N] for finite subfactors and satisfies an additivity property, and apply these quantities to bound the decoherence time of self-adjoint quantum Markov semigroups in terms of the spectral gap and D_{2,cb}. The paper is written so that the main structural result is accessible to a quantum information audience, with the more technical operator-space material placed in an appendix.","tokens_in":28226,"tokens_out":8030,"duration_ms":84012,"significance":"If correct, Theorem 3.1 is a substantial conceptual contribution: it identifies the Pimsner-Popa index with a maximal Rényi-type divergence from the algebra to a subalgebra, uniformly in p, and gives an information-theoretic interpretation of the index. The completely bounded relative entropy D_{p,cb}, with its additivity and its equality to log[M:N] for finite subfactors, is a useful new invariant. The paper is also commendably explicit about its dependency on the Pimsner-Popa structural equality (12), whose converse direction is noted to be open in general, and it provides a concrete finite-dimensional formula (14). The decoherence-time application is interesting and potentially important. However, the main application theorem as printed in Section 4 contains a false inequality, so the paper needs a substantive correction before acceptance.","major_comments":[{"comment":"The asserted bound D(T_t(ρ)||N) ≤ 2e^{-λt} + D_2(ρ||N)/2 is false. The proof reduces the last step to 2 log(1+ab) ≤ 2a + log b with a=e^{-λt}∈[0,1] and b=‖ρ‖_{L^2_1(N⊂M)}≥1; this inequality is not true, e.g. a=1, b=100 gives 2log(101)≈9.21 > 6.61. A concrete state counterexample is M=M_n with normalized trace, N=C1, ρ=n|ψ><ψ|, and T_t=e^{-t}(id−E)+E, which is a self-adjoint quantum Markov semigroup with spectral gap 1. At t=0 one has D(ρ||N)=D_2(ρ||N)=log n, so (24) would require log n ≤ 2 + log n/2, which fails for every n>e^4≈54.6. The abstract's exponential bound, D(T_tρ||N) ≤ 2exp(−λt + D_2(ρ||N)/2), is the correct statement and follows by replacing the false step with 2log(1+ab)≤2ab. Corollary 4.2's displayed inequality (25) is false for the same reason, although the decoherence-time estimate (26) can still be recovered from the corrected exponential bound. This is a load-bearing error in the application section and must be fixed.","section":"Section 4, Theorem 4.1, Eq. (24)"},{"comment":"The proof uses the estimate ‖T_t−E : L^2_1(N⊂M)→L^2_1(N⊂M)‖ ≤ e^{-λt}, imported from Lemma 3.12 of the authors' own preprint [16]. This lemma is central to both Theorem 4.1 and Theorem 4.6, but it is neither stated nor proved in the present paper. Since the current theorem statement must be revised in any case, the authors should make the dependence on this result explicit and either state the lemma in full or provide a proof in an appendix, so that the corrected decoherence bound is self-contained.","section":"Section 4, proof of Theorem 4.1"}],"minor_comments":[{"comment":"The phrase 'a new notation of relative entropy' should be 'a new notion of relative entropy', and 'egordic theory' should be 'ergodic theory'.","section":"Abstract and Introduction"},{"comment":"The displayed derivation of Eq. (11) contains erroneous reciprocal and exponent placement (e.g. '(sup ... )^{-1}' and the final '−1'); the stated conclusion is correct, but the chain of equalities as written does not parse and should be rewritten.","section":"Section 3, Eq. (11)"},{"comment":"In the statement of Theorem 4.6, the map on M_n(M) should be id⊗T^k and the target subalgebra should be M_n(N); as written, 'D(T^k(ρ)||M_n(N))' applies T^k, defined on M, to a density in M_n(M).","section":"Section 4, Theorem 4.6"},{"comment":"There are several typographical slips, including 'gourp von Neumann algebra', 'quantum Makrov map', and 'correpsonds'; these should be corrected in a final revision.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The central Theorem 3.1 appears sound and the tensorization/additivity results are valuable. The main obstacle to acceptance is the false inequality in Theorem 4.1, which is fortunately localized: the exponential bound in the abstract is the correct statement and the decoherence-time application can be repaired. I would be willing to look at a revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Core result: for II1 factors and finite-dimensional inclusions, -log λ(M:N) = sup_ρ D_p(ρ||N) for every p in [1/2,∞]. That is genuinely new and the proof is largely sound. The p=∞ case is a rewrite of the definition, and the p=1/2 lower bound via Hölder plus the cited Pimsner-Popa equality (12) gives the sandwich. The cb-version D_{p,cb}=log[M:N] and the additivity in Theorem 3.9 are solid additions. The paper is also honest that the structural equality is open in general, which is exactly why Theorem A is restricted to those two classes.\n\nThe big problem is Section 4. Theorem 4.1's statement (24) is false. The proof's last step claims 2log(1+e^{-λt}||ρ||_{L^2_1}) ≤ 2e^{-λt}+D_2/2, but with D_2/2 = log||ρ||_{L^2_1} this is the elementary false inequality 2log(1+ab) ≤ 2a+logb. Concrete counterexample: M_n with normalized trace, N=C1, ρ=n|ψ><ψ|, t=0; (24) demands log n ≤ 2 + (log n)/2, violated for n>e^4. The abstract and Theorem B use the correct exponential bound D≤2exp(-λt+D_2/2), and that version does imply the stated decoherence time. So the application survives but Theorem 4.1 and Corollary 4.2 as printed need correction.\n\nMinor issues: display errors (e.g. log 126 vs 124 in Remark 3.5, missing logs in (15), a spurious -Σp_i log p_i term) and unfilled placeholders. None affect Theorem A. The decoherence argument also imports an estimate from the authors' own preprint [16], which I have not independently checked, but that is secondary to the correctable error above.\n\nWho this is for: anyone working on subfactor index as an information quantity, or on decoherence estimates from spectral gaps. The main theorem deserves a serious referee; this should not be desk-rejected. But I would not accept the current version without fixing Theorem 4.1. Send it back for major revision.","headline":"Solid index-entropy core with a false decoherence theorem as printed; the exponential bound in the abstract is right and fixes it.","tokens_in":29122,"tokens_out":3457,"would_cite":true,"duration_ms":32970,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46L53","46L60","46L37","46L51"],"pacs":[],"model":"deepseek-v4-flash","headline":"The Pimsner-Popa index equals the maximal sandwiched Renyi relative entropy to a subalgebra.","keywords":["Pimsner-Popa index","sandwiched Renyi relative entropy","subfactors","conditional expectations","quantum Markov semigroups","decoherence time","amalgamated Lp spaces","completely bounded relative entropy"],"falsifier":"Compute the supremum of $D_{1/2}(\\rho\\|N)$ for the explicit finite-dimensional inclusion $M_6$ with $N=M_2$ embedded with multiplicity 2; the paper predicts the value $\\log 6$. A numerical optimization that returns any strictly smaller supremum would refute Theorem 3.1, since monotonicity in $p$ makes $p=1/2$ the most sensitive endpoint.","tokens_in":27499,"feed_emoji":"🧮","tokens_out":12672,"duration_ms":104514,"temperature":0.7,"pith_summary":"This paper claims that for an inclusion $N\\subset M$ of II$_1$ factors or finite-dimensional von Neumann algebras, the Pimsner-Popa index is an information-theoretic quantity: $\\log\\lambda(M:N)$ equals the maximal sandwiched Renyi relative entropy from a state of $M$ to the subalgebra $N$, for every Renyi parameter $1/2\\le p\\le\\infty$. The equality is $-\\log\\lambda(M:N)=\\sup_\\rho D_p(\\rho\\|N)=\\sup_\\rho D_p(\\rho\\|E(\\rho))$, with $E$ the trace-preserving conditional expectation. If correct, the paper unifies the subfactor index with quantum information measures such as Renyi conditional entropy, coherence, and asymmetry, and shows the index does not depend on $p$. It then defines completely bounded versions of this relative entropy, proves additivity, and derives a dimension-independent decoherence-time estimate for self-adjoint quantum Markov semigroups from a spectral-gap condition.","feed_headline":"Subfactor index is a maximal quantum divergence","feed_subtitle":"For II1 factors and finite algebras, every Renyi p-divergence from M to N peaks at the index.","key_machinery":"The load-bearing identity is the Pimsner-Popa structural equality $\\lambda(M:N)=\\inf\\{\\|E(e)\\|_\\infty : e\\text{ projection in }M\\}$, where $E$ is the trace-preserving conditional expectation and $\\lambda(M:N)=\\max\\{\\lambda:\\lambda\\rho\\le E(\\rho)\\ \\forall\\rho\\ge 0\\}$ is the Pimsner-Popa index. The proof of Theorem 3.1 applies this equality only at the $p=1/2$ step: a Holder estimate gives $D_{1/2}(\\rho\\|N)\\ge -\\log\\|E(\\mathrm{supp}\\,\\rho)\\|_\\infty$, so the supremum reaches $-\\log\\lambda(M:N)$, while monotonicity in $p$ and the $p=\\infty$ identity $-\\log\\lambda(M:N)=\\sup_\\rho D_\\infty(\\rho\\|E(\\rho))$ provide the reverse direction. An extremal projection $e_0$ with $E(e_0)$ a scalar multiple of a projection certifies that the maximum is attained. For the decoherence application, the additional mechanism is an amalgamated $L_p$-norm estimate imported from [16] that transfers an $L^2$ spectral-gap contraction to the conditional space $L^2_1(N\\subset M)$.","core_discovery":"The central discovery, stated as Theorem 3.1, is that for $N\\subset M$ an inclusion of II$_1$ factors or finite-dimensional von Neumann algebras, the supremum over densities of the sandwiched Renyi relative entropy to the subalgebra is exactly $-\\log\\lambda(M:N)$, and the same value is obtained when the second argument is $E(\\rho)$ instead of the infimum over $\\sigma\\in N$. Because sandwiched entropies are monotone in $p$, it suffices to prove the lower bound at $p=1/2$, where the divergence is fidelity-based; the upper bound follows from the identity $-\\log\\lambda(M:N)=\\sup_\\rho D_\\infty(\\rho\\|E(\\rho))$. The paper also introduces $D_p(M\\|N)=\\sup_\\rho D_p(\\rho\\|N)$ and its completely bounded version $D_{p,cb}(M\\|N)$, and shows $D_{p,cb}(M\\|N)=\\log[M:N]$ for finite subfactors. For self-adjoint quantum Markov semigroups with spectral gap $\\lambda$, it proves $D(\\mathrm{id}\\otimes T_t(\\rho)\\|M_n(N))\\le 2e^{-\\lambda t}+D_2(\\rho\\|M_n(N))/2$, giving the uniform decoherence-time bound $t\\ge \\lambda^{-1}(2\\log(2/\\epsilon)+D_{2,cb}(M\\|N)/2)$.","pith_inferences":["I infer that if the Pimsner-Popa structural equality were established for a wider class of inclusions, Theorem 3.1 would extend verbatim, making the subfactor index an information-theoretic invariant for general finite von Neumann algebras.","The additivity of $D_{cb}$ suggests using $e^{D_{p,cb}}$ as a tensor-stable, computable proxy for the index in inclusions where the index is not defined, potentially connecting entropy inequalities to classification questions.","The decoherence bound is likely not tight: for concrete dephasing channels one can numerically compare the true worst-case trace distance with the paper's exponential bound and test whether the $D_{2,cb}$ term overestimates the transient time.","A testable extension is to replace the spectral-gap assumption by a complete logarithmic Sobolev inequality for the same semigroup; the paper's framework predicts the same form of estimate with $D_{2,cb}$ replaced by $D_{1,cb}$."],"forward_implications":["For finite subfactors, the completely bounded relative entropy recovers exactly the subfactor index: $D_{p,cb}(M\\|N)=\\log[M:N]$ for every $1/2\\le p\\le\\infty$, so the index is a divergence maximum and not merely a coupling constant.","The completely bounded relative entropy is additive: $D_{cb}(M_1\\otimes M_2\\|N_1\\otimes N_2)=D_{cb}(M_1\\|N_1)+D_{cb}(M_2\\|N_2)$, so composite systems behave like independent information reservoirs.","Every finite-dimensional self-adjoint quantum Markov semigroup whose decoherence-free subalgebra is commutative has an $\\epsilon$-separability time that is independent of the dimension of any auxiliary system entangled with $M$.","For Schur-multiplier dephasing semigroups on $M_m$, the uniform decoherence time is at most $\\lambda^{-1}(2\\log(2/\\epsilon)+m/2)$, with $\\lambda$ the smallest nonzero decoherence rate."],"supporting_citations":[{"why":"Supplies the Pimsner-Popa structural equality and the finite-dimensional index formula that are the only non-trivial input in Theorem 3.1.","marker":"[31]"},{"why":"Defines the subfactor index and its multiplicativity, used to identify $D_{p,cb}=\\log[M:N]$ for II$_1$ factors.","marker":"[22]"},{"why":"Provides the amalgamated-norm estimate that converts $L^2$ spectral-gap contraction into conditional-entropy decay in Theorem 4.1 and Corollary 4.2.","marker":"[16]"},{"why":"Establishes the amalgamated $L_p$-space framework and duality used to express $D_p(\\rho\\|N)$ as a norm.","marker":"[23]"},{"why":"Introduces the sandwiched Renyi relative entropy for matrix algebras and its basic properties used throughout.","marker":"[28]"},{"why":"Develops sandwiched Renyi entropies and their data-processing and monotonicity properties, foundational for the $D_p$ formalism.","marker":"[41]"},{"why":"Supplies the quantum Pinsker inequality that converts entropy decay into the stated trace-distance decoherence time.","marker":"[40]"}],"fun_headline_variants":["Subfactor index is the maximal Renyi divergence","Renyi entropy to subalgebra caps at index","Index sets ceiling on quantum divergence","Subalgebra relative entropy peaks at index"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof leans on a structural fact about the Pimsner-Popa index, that it is determined by the worst projection in $M$ rather than by all positive operators; this is known only for II$_1$ factors and finite-dimensional algebras, and if it fails, only the inequality $D_p(M\\|N)\\le -\\log\\lambda(M:N)$ remains.","fun_headline_variants_meta":{"raw":{"variants":["Subfactor index is the maximal Renyi divergence","Renyi entropy to subalgebra caps at index","Index sets ceiling on quantum divergence","Subalgebra relative entropy peaks at index"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000175,"raw_usage":{"total_tokens":1269,"prompt_tokens":915,"completion_tokens":354,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":531,"completion_tokens_details":{"reasoning_tokens":299}},"tokens_in":531,"tokens_out":354,"duration_ms":3937,"temperature":1.0,"reasoning_tokens":299,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T05:10:11.150485+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the supremum of $D_{1/2}(\\rho\\|N)$ for the explicit finite-dimensional inclusion $M_6$ with $N=M_2$ embedded with multiplicity 2; the paper predicts the value $\\log 6$. A numerical optimization that returns any strictly smaller supremum would refute Theorem 3.1, since monotonicity in $p$ makes $p=1/2$ the most sensitive endpoint.","supporting_citations":[{"cited_title":"Entropy and index for subfactor s","cited_arxiv_id":null,"evidence_quote":"Supplies the Pimsner-Popa structural equality and the finite-dimensional index formula that are the only non-trivial input in Theorem 3.1."},{"cited_title":"Index for subfactors","cited_arxiv_id":null,"evidence_quote":"Defines the subfactor index and its multiplicativity, used to identify $D_{p,cb}=\\log[M:N]$ for II$_1$ factors."},{"cited_title":"Fisher Information and Logarithmic Sobolev Inequality for Matrix Valued Functions","cited_arxiv_id":"1807.08838","evidence_quote":"Provides the amalgamated-norm estimate that converts $L^2$ spectral-gap contraction into conditional-entropy decay in Theorem 4.1 and Corollary 4.2."},{"cited_title":"Mixed-norm inequalities and operator space L p embedding theory","cited_arxiv_id":null,"evidence_quote":"Establishes the amalgamated $L_p$-space framework and duality used to express $D_p(\\rho\\|N)$ as a norm."},{"cited_title":"On quan- tum r´ enyi entropies: A new generalization and some properties","cited_arxiv_id":null,"evidence_quote":"Introduces the sandwiched Renyi relative entropy for matrix algebras and its basic properties used throughout."},{"cited_title":"Strong converse for the classical capacity of entanglement-breaking and hadamard channels via a sandwiched r´ enyi relative entropy","cited_arxiv_id":null,"evidence_quote":"Develops sandwiched Renyi entropies and their data-processing and monotonicity properties, foundational for the $D_p$ formalism."}],"review_version":1}