{"id":"b39bb30e-4a5e-49b0-8274-0cd1acc60d5c","arxiv_id":"2608.03841","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Crouzeix's conjecture is proved: for any bounded operator A and any rational f, ||f(A)|| ≤ 2 sup_{W(A)} |f|.","lead":"This paper proves Crouzeix's conjecture, a long-standing 2004 problem in operator theory, by showing that every matrix or operator obeys a universal factor-2 bound for polynomial functions. The proof is short and depends on a new perturbation lemma for 2-dilations applied to the standard double-layer potential representation.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proof depends on unproved, possibly mis-stated bounded antilinear map α for the double-layer potential; the displayed formula omits conjugation, and the uniform boundedness of E_n relies on an α-bound not established in the note.","rationale":"Lemma 1 itself appears correct: the contraction argument with the top singular vector is valid, and the uniform boundedness of E_n together with ‖V^*Q^{*n}V‖≤1 implies uniform boundedness of T^n, so the limit step in (3) is justified. The proof of Theorem 3 therefore hinges entirely on the imported double-layer machinery, specifically the existence of a bounded antilinear α with the stated identity. The reader's weakest assumption already identified these cited premises, and I agree that they are the soft spot. I sharpen this: the note's displayed formula for α is plainly missing a conjugation, and the boundedness of the corrected Cauchy-type map from A(Ω) to A(Ω) is nontrivial and questionable for merely continuous boundary data. Without that bound, the written proof does not establish the uniform boundedness required by Lemma 1. Since the conjecture is independently claimed in [12] and the gap may be repairable via Crouzeix's known estimate, the appropriate verdict is CONDITIONAL rather than rejection: the paper should either prove or precisely cite the corrected α bound, or explicitly substitute the classical uniform bound. This does not change the central mathematical claim's plausibility, but it does change the confidence in the proof as written.","tokens_in":5054,"tokens_out":37698,"duration_ms":415402,"concrete_test":"Independently verify the α identity and boundedness from [20] for a non-circular smooth convex domain (e.g., an ellipse). In particular, fix the displayed definition to α(f)=C(\\bar f), where C is the Cauchy transform, and estimate sup_{z∈Ω}|α(φ^n)(z)| for φ a Riemann map and n=1,2,3. If sup_n‖α(φ^n)‖∞ is not bounded by a constant times ‖φ^n‖∞, the note's uniform-boundedness argument fails; then check whether replacing it with Crouzeix's classical bound ‖2Φ(f^n)-f(A)^n‖≤9.08‖f^n‖∞ repairs the gap.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central reduction to Lemma 1 requires that, for every f∈A(Ω), the operators E_n=α(f^n)(A) are well-defined, uniformly bounded, and commute with T=f(A). The note obtains this from a cited fact: there exists a bounded antilinear α:A(Ω)→A(Ω) with 2Φ(f)-f(A)=α(f)(A)^*. This is the true load-bearing point. As printed, the precursor formula is inconsistent: α(f)=1/(2πi)∫ f(σ)(σ-·)^{-1}dσ is the Cauchy integral of f, hence equals f, making α linear, not antilinear. For scalar A=z in the unit disk this would force 2Φ(f)-f(z)=\\overline{f(z)}, which fails for f(z)=z. The intended formula must use \\overline{f(σ)}. With the corrected formula, α maps A(Ω) to the Cauchy transform of \\bar f; boundedness of this map from A(Ω) to A(Ω) with the sup norm is not a routine fact and is false for the Cauchy transform on C(∂Ω) in general. The proof literally uses ‖E_n‖≤‖θ‖‖α‖‖f^n‖∞; if the corrected α is unbounded, that inequality supplies no uniform bound. Remark 4(1) notes that Crouzeix's older 9.08 estimate already gives sup_n‖E_n‖<∞, but the written proof does not invoke it. Thus the theorem is conditional on the cited α facts, and the note contains an apparent misstatement that should be resolved before the proof is accepted as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims a proof of Crouzeix's conjecture: for every bounded operator A on a Hilbert space, the numerical range W(A) is a 2-spectral set. The argument is built around Lemma 1, a finite-dimensional perturbation lemma: if T admits a contraction Q and an isometry V such that E_n = 2 V* Q^{*n} V - T^{*n} are uniformly bounded and commute with T, then ||T|| ≤ 2. The proof of Lemma 1 is algebraically explicit and appears sound. The authors then apply Lemma 1 to T = f(A) in the operator-valued double-layer potential framework. For a smoothly bounded convex Ω ⊃ W(A), they use a bounded antilinear map α with 2Φ(f) - f(A) = α(f)(A)^*, so that E_n = α(f^n)(A), yielding uniform boundedness and commutativity. The note concludes ||f(A)|| ≤ 2 for all f with ||f||∞ ≤ 1.","tokens_in":5434,"tokens_out":10887,"duration_ms":125265,"significance":"If the proof is correct, this settles a major open problem in operator theory with the sharp constant 2. The main conceptual contribution is Lemma 1, which is a clean and essentially self-contained perturbation argument; its proof is a strength of the paper. The application to the double-layer potential is elegant and short. However, the note is not self-contained: it relies on substantial machinery from [19,20], and the displayed formula for the key antilinear map α contains an apparent misstatement. These caveats do not necessarily invalidate the claim, but they make verification difficult and are load-bearing for the main theorem.","major_comments":[{"comment":"The displayed formula α(f) = 1/(2πi) ∫_{∂Ω} f(σ)(σ-·)^{-1} dσ is not antilinear: by the Cauchy integral formula this expression equals f for f ∈ A(Ω), and it is not the Cauchy transform of \\bar f. The correct formula should presumably involve \\overline{f(σ)} in the integrand. This is not merely cosmetic: the uniform boundedness of E_n in the proof is obtained from ∥E_n∥ ≤ ∥θ∥ ∥α∥ ∥f^n∥∞, and the commutativity of E_n with T uses the representation E_n = α(f^n)(A). As printed, the identity 2Φ(f)-f(A) = α(f)(A)^* is not verifiable from the stated definition. The authors should correct the formula and give a precise statement (or reference) for the boundedness of the intended α.","section":"Theorem 3 proof, formula for α"},{"comment":"The proof of Theorem 3 begins with a reduction to H = C^d and to a smoothly bounded open convex Ω containing W(A), citing [19,20]. This is a substantial step, and the rest of the proof depends on the double-layer machinery from [20]. The note also implicitly uses a bounded homomorphism θ:A(Ω)→L(H) in the estimate ∥E_n∥ ≤ ∥θ∥ ∥α∥; its boundedness is not stated. For a result of this importance, the authors should either state the precise reduction proposition from [19,20] or indicate exactly which theorem in those references supplies it. I am not asking for a reproduction of [20], but the current dependence is too opaque.","section":"Reduction to finite dimensions and convex Ω"}],"minor_comments":[{"comment":"The map θ is introduced only in the sentence about commutativity; its boundedness and norm should be stated before the inequality ∥E_n∥ ≤ ∥θ∥ ∥α∥ is used.","section":"Theorem 3 proof"},{"comment":"If the intended formula is indeed the Cauchy transform of \\bar f, please add the missing overline and clarify the orientation of dσ.","section":"Displayed formula"},{"comment":"There is a minor typographical issue: 'α:A → Aa unital' should read 'α:A → A, a unital'.","section":"Remark 4(4)"},{"comment":"Reference [12] is to a preprint server; if a stable published version exists, it would be helpful to cite it.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The proof depends heavily on [19,20], both coauthored by the second author. This is not by itself improper, but for a claim of this magnitude the crucial antilinear map α should be stated accurately and its boundedness should be either proved in an appendix or quoted as a theorem with the exact definition. The apparent misstatement in the displayed formula creates a real obstacle to verification. I would also note that an independent proof has appeared as [12]; this does not affect the technical assessment."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe thing to know: this note proves Crouzeix's conjecture with a new perturbation lemma for 2-dilations. The conjecture was already settled independently by Jin [12], so the theorem statement isn't new, but the proof strategy is. Lemma 1 is the genuine novelty, and I checked its algebra carefully—it's correct. The contraction argument from the top singular vector, the completing-the-square step, and the κ>2 contradiction all work. The uniform-boundedness assumption on E_n is used to get power-boundedness of T, which is the key step.\n\nWhat the paper does well: it isolates a clean, abstract condition under which a dilation gives a 2-bound, and then shows the double-layer machinery supplies exactly that condition. That's a nice reduction. The writing is compact and honest; they acknowledge the independent proof and the limitations (no completely bounded case).\n\nThe soft spots are the imported machinery. The proof of Theorem 3 depends on two cited facts from [19,20]: the reduction to finite dimensions and the existence of a bounded antilinear α with 2Φ(f)−f(A)=α(f)(A)^*. Those aren't proved in the note. That's fine for a research announcement, but it means the result isn't self-contained.\n\nThere is also a real typo: the displayed formula for α says α(f)= (1/2πi)∫ f(σ)(σ−·)^{-1}dσ, which is the Cauchy transform of f, not of \\bar f. It contradicts the statement that α is antilinear. The intended formula obviously should have \\overline{f(σ)}. The stress-test worry that the corrected α might be unbounded on A(Ω) doesn't hold up—this is a known bounded operator in the double-layer theory, and the note cites [20] for it. The test's counterargument about the Cauchy transform on C(∂Ω) is beside the point, because α acts on A(Ω), not on all continuous functions.\n\nSo the proof as written is conditional on cited facts and has a fixable misstatement, but I don't see a load-bearing flaw. The central lemma is correct and the application is plausible. Given the independent proof, the marginal impact is smaller than it would have been a year ago, but this is still a serious piece of work that deserves a referee.\n\nRecommendation: send it to peer review. The typo should be corrected and the α facts stated more precisely, but the core argument is worthy of scrutiny.","headline":"A short, credible proof of Crouzeix's conjecture whose real contribution is Lemma 1; the application leans on cited double-layer machinery, and there's a fixable typo in the α formula.","tokens_in":5919,"tokens_out":15265,"would_cite":true,"duration_ms":147467,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["47A25","47A12","15A60"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims a full proof of Crouzeix's conjecture: every bounded operator on a Hilbert space has a numerical range that is a 2-spectral set.","keywords":["Crouzeix's conjecture","numerical range","2-spectral set","double-layer potential","functional calculus","operator theory","Hilbert space"],"falsifier":"A direct numerical check of the identity 2Φ(f)−f(A)=α(f)(A)^* for a non-normal 2×2 or 3×3 matrix and a rational f with poles outside W(A); any mismatch would break the proof. Independently, an exhaustive search for a matrix with ||f(A)|| > 2 sup_{W(A)}|f| would refute the conjecture itself.","tokens_in":4949,"feed_emoji":"📐","tokens_out":6426,"duration_ms":64569,"temperature":0.7,"pith_summary":"This paper claims a proof of Crouzeix's conjecture: for every bounded operator A on a Hilbert space, the numerical range W(A) is a 2-spectral set, meaning ||f(A)|| ≤ 2 sup_{z∈W(A)}|f(z)| for every rational function f with poles off the closure of W(A). The conjecture dates to 2004 and previously the best universal constant was 1+√2, with the optimal 2 known for special classes. The proof combines a simple perturbation lemma for 2-dilations with the double-layer potential representation of f(A). If correct, it settles the conjecture with the sharp constant and shows the earlier abstract operator-theoretic program reaches the claimed bound.","feed_headline":"Crouzeix's conjecture proved with sharp constant 2","feed_subtitle":"A perturbation lemma plus double-layer potential settles the 2004 bound for every Hilbert-space operator.","key_machinery":"The double-layer potential P_Ω(σ) = (1/π) Re(n_Ω(σ)(σI - A)^{-1}) on the boundary of a smooth convex region Ω containing W(A), together with the induced map Φ(f) = ½∫_∂Ω f P_Ω |dσ|. The proof uses the identity 2Φ(f) - f(A) = α(f)(A)*, where α is a bounded antilinear Cauchy transform, to pair the functional calculus with Lemma 1. Lemma 1 is the other central mechanism: a finite-dimensional perturbation lemma showing that uniform boundedness of the E_n forces ||T|| ≤ 2.","core_discovery":"On the paper's own terms, the central result is Theorem 3: the numerical range of any bounded Hilbert-space operator is a 2-spectral set, confirming Crouzeix's 2004 conjecture with the optimal constant 2. The proof works by first establishing Lemma 1, a perturbation statement: if an operator T has a contractive dilation Q so that the operators E_n = 2V*Q*^n V - T*^n are uniformly bounded and commute with T, then ||T|| ≤ 2. For T = f(A), the double-layer potential gives a natural dilation Q (multiplication by f on a boundary L^2-space), an isometry V, and the commutativity follows from the identity 2Φ(f) - f(A) = α(f)(A)*. Uniform boundedness of E_n follows from the boundedness of the holomor","pith_inferences":["If the abstract theorem in Remark 4(4) extends cleanly, similar 2-spectral-set bounds might hold in other uniform algebras, and the commutativity assumption in Lemma 1 becomes the natural bottleneck to attack.","The same perturbation lemma could be tried on the completely bounded version of Crouzeix's conjecture; the paper notes its proof does not directly apply, but relaxing the commutativity assumption is a concrete starting point.","The proof suggests that any sequence of matrices approaching the constant 2 must make the E_n's nearly non-uniformly bounded or the commutator condition barely satisfied, which could guide a search for extremal cases.","Remark 2's inequality hints that if the sign of ℜ⟨E_1Tx,x⟩ could be controlled, an even simpler proof of the same bound might exist; this thread is left implicit in the paper."],"forward_implications":["Crouzeix's conjecture holds for all bounded operators on Hilbert space, making 2 the sharp universal bound for rational functions of such operators.","By the generalization cited in the paper, the result extends to closed unbounded operators whose numerical range contains their spectrum.","The proof yields an abstract theorem: any unital bounded homomorphism θ on a commutative uniform algebra with a unital antilinear α satisfying the stated positivity condition has ||θ|| ≤ 2.","The proof does not use the contractivity of α, so the earlier (1+√2) route is not a necessary ingredient; it also avoids extremal functions and measures."],"supporting_citations":[{"why":"Crouzeix's original statement of the conjecture and the proof for d=2; it is the result being settled.","marker":"[5]"},{"why":"Established the functional-calculus framework and the bound 11.08; supplies the reduction and the unital Φ used here.","marker":"[6]"},{"why":"Refined the double-layer potential method to give the 1+√2 bound, the predecessor the proof extends.","marker":"[8]"},{"why":"Introduced the operator-valued double-layer potential from which the representation is built.","marker":"[9]"},{"why":"Cited for the reduction to finite-dimensional matrices and smoothly bounded convex sets replacing W(A).","marker":"[19]"},{"why":"Cited for the standard double-layer setup, including the existence and boundedness of the antilinear map α.","marker":"[20]"}],"fun_headline_variants":["Crouzeix's conjecture proven, optimal constant 2","Sharp 2-spectral set proof for every operator","Perturbation lemma solves Crouzeix's conjecture","Double-layer potential nails Crouzeix's bound"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The proof's load-bearing premises are two cited results it does not re-prove: that the general Hilbert-space case reduces to finite-dimensional matrices with W(A) replaced by a smoothly bounded convex set, and that the double-layer potential yields a bounded antilinear map α satisfying 2Φ(f)−f(A)=α(f)(A)^*.","fun_headline_variants_meta":{"raw":{"variants":["Crouzeix's conjecture proven, optimal constant 2","Sharp 2-spectral set proof for every operator","Perturbation lemma solves Crouzeix's conjecture","Double-layer potential nails Crouzeix's bound"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00058,"raw_usage":{"total_tokens":2494,"prompt_tokens":591,"completion_tokens":1903,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":335,"completion_tokens_details":{"reasoning_tokens":1848}},"tokens_in":335,"tokens_out":1903,"duration_ms":15883,"temperature":1.0,"reasoning_tokens":1848,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T11:16:59.399160+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct numerical check of the identity 2Φ(f)−f(A)=α(f)(A)^* for a non-normal 2×2 or 3×3 matrix and a rational f with poles outside W(A); any mismatch would break the proof. Independently, an exhaustive search for a matrix with ||f(A)|| > 2 sup_{W(A)}|f| would refute the conjecture itself.","supporting_citations":[{"cited_title":"Crouzeix","cited_arxiv_id":null,"evidence_quote":"Crouzeix's original statement of the conjecture and the proof for d=2; it is the result being settled."},{"cited_title":"Crouzeix","cited_arxiv_id":null,"evidence_quote":"Established the functional-calculus framework and the bound 11.08; supplies the reduction and the unital Φ used here."},{"cited_title":"Crouzeix and C","cited_arxiv_id":null,"evidence_quote":"Refined the double-layer potential method to give the 1+√2 bound, the predecessor the proof extends."},{"cited_title":"Delyon and F","cited_arxiv_id":null,"evidence_quote":"Introduced the operator-valued double-layer potential from which the representation is built."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Cited for the reduction to finite-dimensional matrices and smoothly bounded convex sets replacing W(A)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Cited for the standard double-layer setup, including the existence and boundedness of the antilinear map α."}],"review_version":1}