{"id":"e2400885-5168-445a-b174-bbdd3e98c783","arxiv_id":"2607.13794","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Sharp complete (matrix-valued) functional calculus bounds for rho-contractions, with constants depending on the smallest or largest singular value of F(0).","lead":"This paper proves sharp bounds for how much matrix-valued analytic functions can distort rho-contractions, depending only on the value of the function at zero. It completes earlier scalar results by Drury and Schwenninger-de Vries and gives the refined analogue of the Okubo-Ando theorem.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Prop. 3.5 non-affine factorization is algebraically wrong: the scalar (1−s†)+(s−s†) belongs in the denominator, not the numerator.","rationale":"I read the paper in good faith. The overall strategy is coherent—reduce to F(0)=0 via Potapov–Möbius transforms, prove stability of C_ρ^(n), and verify a 2×2 positivity criterion—and the scalar case gives confidence. However, the AI-assisted algebraic verification in Proposition 3.5 is the linchpin: if W(t)≥0 fails, Theorems 1.1 and 1.2 collapse. On inspection, the printed factorization on the second branch is not an identity. This is an actual algebraic error, not just a missing expansion. The error is fixable and the corrected computation still yields positive semidefiniteness, so the central claim likely survives. The reader's conditional verdict is therefore unchanged, but the proof needs revision. Agreement: agree—the reader identified the same weakest point.","tokens_in":13912,"tokens_out":33606,"duration_ms":260433,"concrete_test":"Independently expand W(t) for the non-affine branch: substitute b_{ρ,K}(t), δ=ρK, x=−1 into the 2×2 W(t) of Lemma 3.3 and compare with the factorization in Proposition 3.5. A simple check with ρ=2, K=3/2, t=16/25 reveals the printed RHS is Q² times too small, where Q=(1−s†)+(s−s†); the correct denominator is (ρK−1)(1−s)Q. After correcting, verify that the prefactor is positive for all s†<s<1, so W(t)≥0 follows.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorems 1.1 and 1.2 rests on Proposition 3.5, and the least secure step is the claimed factorization of W(t) on the non-affine branch (ρ>1, t>s†²). Direct expansion from the definition of W in Lemma 3.3 with δ=ρK, x=−1, and the printed b_{ρ,K} gives W(t) = 1/(d(1−s)) w(s)w(s)^*, where d=2(K−1)−(ρK−1)(1−s). But the paper states W(t) = ((1−s†)+(s−s†))/((ρK−1)(1−s)) w(s)w(s)^*. Since d=(ρK−1)((1−s†)+(s−s†)), the printed scalar is off by a factor of ((1−s†)+(s−s†))². Numerically, at ρ=2, K=1.5, t=0.64, the left side is [[1/12,1/6],[1/6,1/3]], while the printed right side is [[3/400,3/200],[3/200,3/100]]. Thus the displayed equality in Proposition 3.5 is false. The corrected factorization still yields W(t)≥0 on the branch (the prefactor is positive), so the theorem may be salvageable, but the proof as written contains an incorrect algebraic identity in the central step. This is precisely the AI-assisted computation flagged by the reader.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves complete (matrix-valued) functional calculus bounds for ρ-contractions. For F in M_n(O(D)) with ||F||_∞ ≤ 1, Theorem 1.1 bounds ||F(T)|| by c_ρ(||F(0)||^2) for ρ ≤ 1 and by c_ρ(m(F(0))^2) for ρ > 1, refining the Okubo–Ando complete spectral set bound. The more precise Theorem 1.2 gives an operator upper bound F(T)^*F(T) ≤ b_{ρ,K}(A^*A), where b_{ρ,K} is an explicit piecewise function. The proof introduces a class C_ρ^(n) of matrix-intertwined ρ-contractions and uses Potapov–Möbius transforms; the key algebraic step is Proposition 3.5, which verifies a positivity criterion for W(t). Section 5 addresses sharpness, including a lower bound for any admissible h and an optimality theorem under natural assumptions.","tokens_in":14220,"tokens_out":11129,"duration_ms":87686,"significance":"If the main theorems hold, this is a substantial refinement of the Okubo–Ando bound in the complete setting, and the first sharp complete functional calculus bound for ρ-contractions with a data-dependent constant. The proof strategy via Potapov–Möbius transforms and the class C_ρ^(n) is natural and likely correct after the algebraic correction noted below. The paper is transparent about AI-assisted computations, which is commendable but also means that all such identities require independent verification. The sharpness discussion in Section 5 is valuable, though it rests on a nontrivial computation that should be made fully checkable.","major_comments":[{"comment":"The displayed factorization of W(t) is algebraically false. Direct expansion from Lemma 3.3 with δ=ρK and x(t)=−1 gives W(t) = 1/(d(1−s)) w(s)w(s)^*, where d=2(K−1)−(ρK−1)(1−s). The printed formula instead has the prefactor ((1−s†)+(s−s†))/((ρK−1)(1−s)). Since d=(ρK−1)((1−s†)+(s−s†)), the printed prefactor is off by a factor ((1−s†)+(s−s†))². For ρ=2, K=1.5, t=0.64, the left side is [[1/12,1/6],[1/6,1/3]] while the printed right side is [[3/400,3/200],[3/200,3/100]]. The corrected prefactor is positive on the stated branch, so the conclusion W(t)≥0 remains true, but the proof as written contains an incorrect equality at a load-bearing step. The authors should correct the identity and re-check that no later argument depends on the erroneous prefactor.","section":"Proposition 3.5, non-affine branch"},{"comment":"The step from the Schur complement to inequality (10), and the subsequent claim that the last summand of (10) can be made to vanish by choosing λ∈[0,∞), is a substantial algebraic simplification that is not shown. Since this is the basis for the sharpness claim, the paper should provide the full expansion or a verifiable symbolic computation. The present text is not independently checkable from the material given, especially in light of the AI-assisted provenance of part of this construction.","section":"Theorem 5.3, proof"}],"minor_comments":[{"comment":"The displayed matrix A is malformed. From the subsequent compression of h(A^*A), A should be a 2×2 diagonal matrix with singular values √t and √t0 (e.g., diag(√t,√t0) or diag(√t0,√t)), not the matrix shown in the text.","section":"Theorem 5.3, proof"},{"comment":"The sentence 'the spectral radius of an operator of class C_ρ is at most ρ/(2−ρ)' is imprecise as stated; the cited bound applies for ρ<1, whereas for ρ≥1 the spectral radius is at most 1 by the unitary dilation. Consider clarifying the distinction.","section":"Lemma 2.3"},{"comment":"The notation h(F(0)^*F(0)) for a matrix-valued argument should be explicitly defined as the continuous functional calculus applied to the normal matrix F(0)^*F(0), since h is a scalar function.","section":"Proposition 5.2"}],"recommendation":"major_revision","confidential_remarks":"The false identity in Proposition 3.5 is precisely the type of error that the AI-assistance disclosure should have flagged. I believe the main theorem is correct and the paper will be suitable for publication once the algebraic identities are corrected and verified. The editor may wish to ask the authors to include a machine-checked expansion of the corrected factorization and of the simplification leading to (10)."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the matrix-valued refinement is real and the main theorem is probably correct, but the proof of Proposition 3.5 contains a wrong factorization of W(t) on the non-affine branch. The stress-test is right. The paper prints W(t) = ((1−s†)+(s−s†))/((ρK−1)(1−s)) w(s)w(s)*; direct expansion with the definition of W in Lemma 3.3 gives W(t) = 1/[(ρK−1)(1−s)((1−s†)+(s−s†))] w(s)w(s)*. The factor belongs in the denominator, not the numerator. Numerical check at ρ=2, K=1.5, t=0.64: the left side is [[1/12,1/6],[1/6,1/3]], while the printed right side is [[3/400,3/200],[3/200,3/100]].\n\nThis is not fatal: the corrected prefactor is still positive on that branch, so W(t)≥0 and Proposition 3.5 goes through. But as written, the central algebraic step of the proof is false, and the paper acknowledges those computations were partly ChatGPT-generated. A referee needs to see this fixed and the corrected identity verified line by line.\n\nWhat is genuinely good: Theorems 1.1 and 1.2 give the first matrix-valued sharp refinement of Okubo–Ando, with the dependence on m(F(0)) for ρ>1. The Potapov–Möbius reduction and the new class C_rho^(n) are natural. The sharpness discussion is honest, especially the lower bound in Proposition 5.2 and the conditional nature of Theorem 5.3. The citation pattern is appropriate; the scalar result [11] is properly credited and the extension is not a repackaging.\n\nSoft spots beyond the factorization: Theorem 5.3's \"simplifying yields\" step is compressed and also AI-assisted; it deserves expansion before publication. Lemma 5.4's proof is fine. The claim on the affine branch in Proposition 3.5, equation (7), is a direct computation I did not verify but no red flags.\n\nVerdict: This should go to peer review, not desk reject. The result is significant enough and likely correct, but the authors must fix the Proposition 3.5 identity and provide a human-checked derivation of the non-affine factorization. My recommendation: engage with the work, bring it to reading group, and cite it once the corrected version is out.","headline":"The matrix-valued bound is a real advance, but Proposition 3.5 has a wrong factorization in the central positivity argument—fixable, but the proof as printed is not correct.","tokens_in":14758,"tokens_out":8603,"would_cite":true,"duration_ms":83317,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["47A60","47A20","47A25","47A30","47A63"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper establishes that for any ρ-contraction T on a Hilbert space, a contractive matrix-valued analytic function F satisfies an explicit sharp bound determined solely by the matrix F(0).","keywords":["ρ-contraction","functional calculus","complete bound","Potapov–Möbius transform","smallest singular value","unitary dilation","spectral set","matrix-valued analytic function"],"falsifier":"A direct computer-algebra check of the 2×2 matrix W(t) from Lemma 3.3 with b = b_{ρ,K}, δ = ρK, and the proposed x(t) would settle the key step: for example, take ρ = 2, K = 1.5, and t = 0.9 (which lies on the non-affine branch) and verify whether det(W(t)) is non-negative; a negative determinant would contradict Proposition 3.5 and hence Theorem 1.2. Verifying identity (7) for generic ρ, K, and t on the affine branch would test the affine part.","tokens_in":13741,"feed_emoji":"📐","tokens_out":4792,"duration_ms":63120,"temperature":0.7,"pith_summary":"This paper proves sharp complete functional calculus bounds for ρ-contractions: for every contractive matrix-valued analytic function F on the unit disc, the norm of F(T) is bounded by an explicit decreasing function of the data at zero. When ρ is at most 1, the bound depends on the operator norm of F(0); when ρ exceeds 1, it depends on the smallest singular value of F(0). The result is the matrix-valued refinement of the scalar bounds of Drury and of Schwenninger and de Vries, and it sharpens the classical complete ρ-spectral set bound of Okubo and Ando. The authors also show that the main estimate is optimal, both at the level of the norm inequality and, under natural assumptions, at the level of the underlying operator inequality.","feed_headline":"Every ρ-contraction obeys a sharp matrix calculus bound","feed_subtitle":"The complete version of the Okubo–Ando inequality depends only on F(0) — or, for ρ>1, its smallest singular value.","key_machinery":"The argument combines Potapov–Möbius transforms M_A on the unit ball of M_n(C) with the new operator class C_ρ^{(n)}: ρ-contractions that remain of class C_ρ after multiplication by arbitrary pairs of contractive scalar matrices. Lemma 2.1 shows that composing a ρ-contraction with a matrix-valued function vanishing at 0 produces an operator in C_ρ^{(n)}. Lemma 3.3 reduces the operator inequality M_A(T)*M_A(T) ≤ B to the positivity of a 2×2 matrix W(t) on the spectrum of A*A, and Proposition 3.5 exhibits the explicit b_{ρ,K} for which this positivity holds, choosing the auxiliary parameter x(t) to cancel off-diagonal terms on the affine branch and setting x(t) = −1 on the non-affine branch. T","core_discovery":"Theorems 1.1 and 1.2 give the sharp complete bound: for T of class C_ρ and ||F||_∞ ≤ 1, ||F(T)|| ≤ c_ρ(||F(0)||²) when ρ ≤ 1 and ||F(T)|| ≤ c_ρ(m(F(0))²) when ρ > 1, where c_ρ(t) = (ρ/2)(1−t) + sqrt(ρ²(1−t)²/4 + t) and m is the smallest singular value. At the operator level, F(T)*F(T) is bounded above by an explicit function b_{ρ,K}(A*A), defined piecewise with an affine branch and, for ρ > 1, a non-affine branch. The bound is attained for every admissible matrix F(0) by an explicit example, and Theorem 5.3 shows that among bounds of the form h(F(0)*F(0)) with the natural normalization, the function b_{ρ,K} is optimal.","pith_inferences":["The switch from the operator norm of F(0) for ρ ≤ 1 to the smallest singular value for ρ > 1 suggests a geometric interpretation: for ρ > 1, the functional calculus is controlled by the lowest-rank direction of the initial value, which may have consequences for numerical-radius inequalities for matrix-valued functions.","The piecewise definition of b_{ρ,K} indicates a genuine dichotomy that might reflect two distinct regimes in the action of Potapov–Möbius transforms; a natural testable extension would be to determine whether an optimal bound with a smoother function could exist if one drops the assumption that the bound depends only on F(0)*F(0).","The matrix-positivity criterion of Lemma 3.3 is phrased spectrally and might adapt to commuting tuples of ρ-contractions or to functions of several variables, where sharp complete bounds are still largely open.","For ρ < 1, the bound depends on the largest singular value and the function c_ρ is increasing, so the worst case occurs when F(0) = 0; the symmetry between ρ and 1/ρ suggests a duality relation that could be made explicit by comparing the two branches of b_{ρ,K}."],"forward_implications":["If Theorem 1.1 is correct, the Okubo–Ando complete ρ-spectral set bound follows as a corollary, and combining it with Paulsen's similarity theorem yields a new proof of the Okubo–Ando similarity result.","For n = 1, the bound reduces to the scalar refined estimate ||f(T)|| ≤ c_ρ(|f(0)|²), recovering the earlier results of Drury and of Schwenninger and de Vries.","For ρ = 2, the class C_2 consists of operators with numerical radius at most 1, so the theorem provides a matrix-valued von Neumann-type inequality with constant c_2(m(F(0))²), which can be strictly smaller than the classical constant 2.","Proposition 5.1 shows that the constants in Theorem 1.1 cannot be improved for any fixed matrix A = F(0), settling sharpness of the norm-level bound.","Theorem 5.3 shows that the operator-level bound b_{ρ,K}(A*A) is optimal among all bounds of the form h(F(0)*F(0)) that are continuous, normalized by h(t₀) = K², and valid for all matrix-valued functions with the prescribed value at zero."],"fun_headline_variants":["Optimal matrix bound for ρ-contractions' functional calculus","Sharpest calculus bound for ρ-contractions, explicitly attained","ρ-contractions: complete sharp matrix bound now proven","Matrix functional calculus: sharp bound for ρ-contractions"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The claimed operator bound follows only if the algebraic factorisations in Proposition 3.5 — the identity (7) on the affine branch and the w(s)w(s)* splitting on the non-affine branch — are correct; these steps are asserted with computations not fully expanded in the text, and the proof of Theorem 1.2 depends on them.","fun_headline_variants_meta":{"raw":{"variants":["Optimal matrix bound for ρ-contractions' functional calculus","Sharpest calculus bound for ρ-contractions, explicitly attained","ρ-contractions: complete sharp matrix bound now proven","Matrix functional calculus: sharp bound for ρ-contractions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000326,"raw_usage":{"total_tokens":1611,"prompt_tokens":642,"completion_tokens":969,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":386,"completion_tokens_details":{"reasoning_tokens":904}},"tokens_in":386,"tokens_out":969,"duration_ms":30244,"temperature":1.0,"reasoning_tokens":904,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T03:42:18.587185+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct computer-algebra check of the 2×2 matrix W(t) from Lemma 3.3 with b = b_{ρ,K}, δ = ρK, and the proposed x(t) would settle the key step: for example, take ρ = 2, K = 1.5, and t = 0.9 (which lies on the non-affine branch) and verify whether det(W(t)) is non-negative; a negative determinant would contradict Proposition 3.5 and hence Theorem 1.2. Verifying identity (7) for generic ρ, K, and t on the affine branch would test the affine part.","supporting_citations":[],"review_version":1}