{"id":"85513070-47be-4476-8475-908b0ee8d412","arxiv_id":"2411.16359","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Exact L2 Bernstein-Markov factors are determined for generalized Hermite and Gegenbauer weights, for ordinary derivatives and Dunkl operators, with extremal polynomials identified.","lead":"This paper finds the exact largest possible ratios between the size of a polynomial's derivative and the size of the polynomial itself, for generalized Hermite and Gegenbauer weights. It also solves the same sharp-constant problem for Dunkl operators, a reflection-symmetric version of derivatives used in special function theory.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Exact constants survive scrutiny; the genuine flaw is non-uniqueness at eigenvalue coincidences, not a lowering of the claimed maxima.","rationale":"The reader's weakest assumption was that the classification of polynomial solutions might miss mixed-parity solutions and thereby lower the claimed maxima. I checked this by tracing the recurrences in Propositions 1-4. The operator L splits into even and odd parts, and the recurrence fixes all coefficients of a given parity once the leading coefficient and M are known; the condition a_{s+1}=a_{s+2}=0 forces M to be one of the lambda_s. Hence no genuinely new eigenvalue can arise from mixed-parity solutions. The exact constants in Theorems 1-4 are therefore safe. What does happen is that at parameter values where lambda_s=lambda_r for s != r, the eigenspace has dimension greater than one, so scalar-multiple uniqueness claims fail. For example, in the Gegenbauer Dunkl equation with lambda=1, mu=5/2, C_1 and C_2 both solve with M^2=18. The paper's Proposition 1, Remark 3, Tables 3-4, and Theorem 5 equality statements are consequently overclaimed, exactly as the reader suspected regarding degeneracies. This warrants a conditional acceptance with a request to rephrase uniqueness assertions as eigenspace descriptions, but it does not change the main numerical results. The verdict should remain CONDITIONAL, with no adjustment needed from the reader's assessment.","tokens_in":23208,"tokens_out":29280,"duration_ms":245436,"concrete_test":"Set lambda=1, mu=5/2, n=2 in the Gegenbauer Dunkl equation (2.7). Compute C_1 and C_2 explicitly, verify L[C_1]=L[C_2]=0 with M^2=18, and evaluate the Rayleigh quotient for p=C_1+C_2. If the quotient equals sqrt(18) while the eigenvalue/determinant maxima also equal sqrt(18), the uniqueness overclaim is confirmed and the exact constant is unaffected. Additionally, repeat for the Hermite Dunkl case lambda=1/2, n=2, where H_1 and H_2 share M^2=4.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The exact Bernstein-Markov constants in Theorems 1-4 are not endangered by the reader's classification worry. The coefficient recurrences in Propositions 1-4 force every polynomial solution to have M^2 equal to one of the listed lambda_s^2: even and odd subspaces decouple under L, and within each parity the first-order recurrence determines the coefficients uniquely. Thus no mixed-parity solution can produce a new, larger eigenvalue. The real error is the uniqueness clause. When two eigenvalues coincide, arbitrary linear combinations of the corresponding generalized polynomials are also solutions. Example: for the Gegenbauer Dunkl equation (2.7) with lambda=1, mu=5/2, lambda_1^2=18 and lambda_2^2=18, so p=aC_1+bC_2 solves with M^2=18 although Proposition 1 asserts p=cC_s. This makes Remark 3, Tables 3-4, and the 'if and only if' in Theorem 5 overclaimed at such degenerate parameter choices. It does not raise M_n, so the central exact-value claims stand.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives exact L2 Bernstein-Markov factors for generalized Hermite weights |x|^{2λ}e^{-x^2} and generalized Gegenbauer weights |x|^{2λ}(1-x^2)^{μ-1/2}, for both the ordinary derivative and the Dunkl operator. The main results (Theorems 1-4) express the sharp constant as the square root of the largest eigenvalue of an explicitly given differential or differential-difference operator, with the extremal polynomials identified as generalized Hermite or Gegenbauer polynomials. The proofs rest on two duality lemmas that reformulate the supremum as an eigenvalue problem, followed by coefficient-recurrence arguments and determinant formulas for the odd-degree cases. The paper also contains a characterization-type inequality for the generalized orthogonal polynomials (Theorem 5) and several worked examples.","tokens_in":23377,"tokens_out":9630,"duration_ms":107183,"significance":"If correct, the paper closes a gap in the literature: for λ>0 the ordinary-derivative Hermite case was previously known only via bounds, and the Dunkl-operator cases appear to be new. The method is attractive and self-contained in its main steps: the duality lemmas are clean, the coefficients of the generalized polynomials are derived from explicit recurrences, and no parameters are fitted. The exact constants themselves survive scrutiny. However, the paper's extremal-polynomial classification and its equality characterization are overstated, because the relevant eigenvalue problems have nontrivial degeneracies at certain parameter values. The central numerical constants are not endangered, but the statements describing all extremal polynomials must be weakened and the degeneracy cases recorded.","major_comments":[{"comment":"The uniqueness claim that a nontrivial polynomial solution p of (2.7) must be a scalar multiple of a single C_s is false when the eigenvalues in (2.6) coincide for indices of different parity. For example, with λ=1 and μ=5/2, one has λ_1^2=λ_2^2=18, so p = a C_1^{(5/2,1)} + b C_2^{(5/2,1)} solves (2.7) with M^2=18 but is not proportional to C_s for any single s. The proof's deduction 'a_{s-1}=a_{s-3}=...=0' on page 8 is not justified because the recurrence coefficient can vanish exactly when M^2 equals the opposite-parity eigenvalue. The same defect occurs in Proposition 3 for the Hermite equation (2.12); for λ=1/2, λ_1^2=λ_2^2=4, so H_1 and H_2 are both solutions for the same M^2.","section":"Section 2, Proposition 1"},{"comment":"The degeneracy propagates to the extremal-polynomial statements. In Theorem 3 and Table 3, for even n and λ=1/2, both H_{n-1} and H_n attain the value M_n^2=2n, so the full extremal set includes their linear combinations, not only cH_n as stated. Likewise, in Theorem 4(ii) at the boundary (2λ-1)(2μ-1)=4 with n=2, λ_2^2=λ_1^2, so additional extremal polynomials exist beyond cC_2. The equality characterization in Theorem 5 should be 'p belongs to the span of {C_s : λ_s^2=λ_n^2}' rather than 'p=cC_n'. None of this lowers the exact constants in Theorems 1-4: for any nontrivial solution, the highest-degree parity component is itself a solution, and the leading-coefficient argument still forces M^2 to be one of the listed eigenvalues. The constants are therefore sound, but the uniqueness and 'if and only if' claims require correction.","section":"Theorems 3-4"}],"minor_comments":[{"comment":"The statement says 'Let λ>0 and μ>-1/2' but the equation (2.13) contains no μ; the condition on μ is extraneous and should be removed.","section":"Proposition 4"},{"comment":"Reference [16] misspells the author's name as 'Schimidt'; it should be 'Schmidt'.","section":"References"},{"comment":"The column header 'λ(λ = -μ)' is potentially confusing because the table also allows positive μ; a short note describing the parameter ranges used in the computations would improve readability.","section":"Table 2"},{"comment":"The determinant definitions of F_{m+1}(t) and G_{m+1}(t) are terse; stating the matrix size and the range of indices explicitly in the theorem statements would help readers verify the Cramer-rule step.","section":"Theorem 1 and Theorem 2"}],"recommendation":"major_revision","confidential_remarks":"The reader's conditional verdict is accurate, and the skepticism about the uniqueness claims is confirmed, but the diagnosis is localized: the exact Bernstein-Markov constants are not endangered, and the flaw appears fixable by weakening the extremal uniqueness statements and recording the degeneracy cases. I would advise the editor that the revision should focus on Propositions 1 and 3, Remark 3, Tables 3-4, and Theorem 5; the main theorems' numerical values can stand. There is no circularity or parameter-fitting concern in the paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the exact constants are right, but the extremal-polynomial claims are overreaching at a few degenerate parameter values. The paper deserves a serious referee, and a revision that fixes the uniqueness statements.\n\nWhat's new: exact L2 Bernstein-Markov factors for generalized Hermite and Gegenbauer weights, both for the ordinary derivative and for Dunkl operators. For λ>0 and ordinary derivative, only two-sided bounds were known (Draux–Kaliaguine), so Theorems 1–2 are a real step. Theorems 3–4 for Dunkl operators are new and reduce to known Schmidt and Guessab–Milovanovic results at λ=0. The proofs are clean: duality lemmas convert the extremal ratio into an eigenproblem, and coefficient recurrences from Ben Cheikh–Gaied identify the eigenvalues. I checked the algebra in the propositions; the eigenvalue computation is sound, and the determinant formulas for the odd cases look right.\n\nThe problem is the uniqueness clauses. Propositions 1 and 3 claim every polynomial solution is a scalar multiple of one generalized Gegenbauer or Hermite polynomial. That is false when two eigenvalues coincide across parities. Example: for the Dunkl Gegenbauer equation with λ=1, μ=5/2, both C_1 and C_2 have λ^2=18, so any linear combination solves. The same happens in the Dunkl Hermite case at λ=1/2, where H_{n-1} and H_n are both extremal for even n. This makes the extremal-polynomial statements in Theorem 3 (Table 3), Theorem 4(ii)–(iii) and its Table 4, and the “if and only if” in Theorem 5 overclaimed at those boundaries. The constants are not endangered: taking the max over eigenvalues still gives the right M_n, so the main theorems survive. But the characterization of extremizers is wrong as stated, and some reader will trip on it.\n\nOne thing the reader's report worried about—the reliance on the external classification of polynomial solutions—doesn't bother me. Even/odd subspaces decouple, and the first-order recurrences force the eigenvalue list; the stress-test note says the same and I agree.\n\nBottom line: worth a serious referee. Ask the authors to rework Propositions 1 and 3 to describe the solution space at eigenvalue coincidences, and to qualify the equality cases in Theorems 3–5 accordingly. The main constants are a solid contribution.","headline":"Exact L2 Bernstein-Markov constants are correct, but the extremal-polynomial uniqueness claims fail at eigenvalue coincidences; the paper is worth refereeing after a revision.","tokens_in":23948,"tokens_out":7973,"would_cite":true,"duration_ms":68678,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["33C45","41A17","41A44","42C05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper determines the exact L2 Bernstein–Markov constants for generalized Hermite and Gegenbauer weights, for both the ordinary derivative and the Dunkl operator, and identifies the polynomials that attain them.","keywords":["L2 Bernstein-Markov inequalities","generalized Hermite weight","generalized Gegenbauer weight","Dunkl operator","extremal polynomials","exact Bernstein-Markov constants","orthogonal polynomials"],"falsifier":"For a fixed $\\lambda>0$ and an odd degree $n$ (for instance $\\lambda=1$, $n=3$), directly compute the largest eigenvalue of the matrix with entries $\\int_{\\mathbb{R}} p_i' p_j' w_\\lambda\\,\\mathrm{d}x$ relative to $\\int_{\\mathbb{R}} p_i p_j w_\\lambda\\,\\mathrm{d}x$ over a basis of $\\mathcal{P}_3$; the claimed constant is $\\sqrt{\\nu_2}$ from Example 1, and any ratio exceeding that value, or a largest eigenvalue differing from $\\nu_2$, would refute Theorem 1.","tokens_in":73,"feed_emoji":"📐","tokens_out":18800,"duration_ms":190012,"temperature":0.7,"pith_summary":"The paper determines, for every degree $n$, the sharp constant in the weighted-$L_2$ comparison between a polynomial and its derivative (or Dunkl derivative), for two families of even weights: the generalized Hermite weight $|x|^{2\\lambda}e^{-x^2}$ on the real line and the generalized Gegenbauer weight $|x|^{2\\lambda}(1-x^2)^{\\mu-1/2}$ on $[-1,1]$. For the ordinary derivative the constant is $\\sqrt{2n}$ when $n$ is even (the same as for the classical Hermite weight), and for odd $n$ it is the largest positive root of an explicit determinant; in the Gegenbauer case it is the maximum of such a root and $\\sqrt{n(n+2\\lambda+2\\mu)}$ (or a shifted version for odd $n$). For the Dunkl operator the constants collapse to simple closed forms: $\\sqrt{2n}$, $\\sqrt{2(n+2\\lambda-1)}$, or $\\sqrt{2(n+2\\lambda)}$ for Hermite weights depending on parity and on $\\lambda$, and $\\sqrt{n(n+2\\lambda+2\\mu)}$ with stated corrections for Gegenbauer weights. In every case the extremal polynomials are identified as generalized Hermite or Gegenbauer polynomials, or odd polynomials from a specified linear system, so the inequalities are sharp.","feed_headline":"Exact derivative bounds found for Hermite and Gegenbauer weights","feed_subtitle":"Sharp constants: √(2n), determinant roots, and Dunkl corrections.","key_machinery":"The argument rests on a duality lemma (Lemmas 1 and 2) that equates the Bernstein–Markov factor with the largest $M>0$ for which a differential equation has a nontrivial polynomial solution: for the ordinary derivative, the integral equation system $\\int_I \\{A p'' + C p' + (2\\lambda/x) p' + M^2 p\\} q\\, W_\\lambda = 0$ for all $q\\in\\mathcal{P}_n$, and for the Dunkl operator the differential equation $A D_\\lambda^2 p + B D_\\lambda p + M^2 p = 0$, with coefficients $A,B,C$ from Table 1. Propositions 1–4 classify all polynomial solutions of these singular equations, using coefficient recurrences (2.4) and (2.9) from [4], as scalar multiples of the generalized Gegenbauer polynomials $C_s^{(\\mu,\\lambda)}$ or generalized Hermite polynomials $H_s^\\lambda$, with eigenvalues given by (2.6) and (2.11). The parity decomposition of the extremal polynomial then forces either an even solution, which is a generalized orthogonal polynomial, or an odd one; in the odd case the constant appears as the largest positive root of an explicit determinant $F$ or $G$ built from the even moments of the weight.","core_discovery":"The paper's central claim is that the extremal problem $M_n^2(L_2(W_\\lambda), D) = \\sup_{p\\in\\mathcal{P}_n} \\|D p\\|_{L_2(W_\\lambda)}^2 / \\|p\\|_{L_2(W_\\lambda)}^2$ has exact, explicitly computable values. For the generalized Hermite weight $w_\\lambda(x)=|x|^{2\\lambda}e^{-x^2}$ on $\\mathbb{R}$ with $D=\\mathrm{d}/\\mathrm{d}x$, $M_n=\\sqrt{2n}$ for even $n$, while for odd $n$ $M_n=\\sqrt{\\nu_{(n+1)/2}}$ where $\\nu_m$ is the largest positive root of the determinant $F_m(t)=\\det\\{(2j+1)(2j+2\\lambda)d_{2i+2j}+(t-4j-2)d_{2i+2j+2}\\}_{i,j=0}^{m-1}$ with moments $d_{2s}=\\Gamma(s+\\lambda+1/2)$. For the generalized Gegenbauer weight $w_{\\lambda,\\mu}(x)=|x|^{2\\lambda}(1-x^2)^{\\mu-1/2}$ on $[-1,1]$ with $D=(1-x^2)^{1/2}\\mathrm{d}/\\mathrm{d}x$, the constant is the maximum of $\\sqrt{\\nu}$ and $\\sqrt{n(n+2\\lambda+2\\mu)}$ for even $n$ (and of $\\sqrt{\\nu}$ with $\\sqrt{(n-1)(n+2\\lambda+2\\mu-1)}$ for odd $n$), where $\\nu$ is the largest positive root of the analogous determinant $G_m$ built from the $\\beta$ moments $c_{2s}=\\Gamma(s+\\lambda+1/2)\\Gamma(\\mu+1/2)/\\Gamma(\\lambda+\\mu+s+1)$. For the Dunkl operator $D_\\lambda$, the constants are exactly $\\sqrt{2n}$, $\\sqrt{2(n+2\\lambda-1)}$, or $\\sqrt{2(n+2\\lambda)}$ in the Hermite case depending on parity and on whether $\\lambda\\le 1/2$, and for the Gegenbauer case $\\sqrt{n(n+2\\lambda+2\\mu)+4\\lambda\\mu}$ for odd $n$, plus the stated $\\sqrt{n(n+2\\lambda+2\\mu)}$ or $\\sqrt{n(n+2\\lambda+2\\mu)+2(n_0-n)}$ alternatives for even $n$ when $(2\\lambda-1)(2\\mu-1)>4$, with $n_0=(\\lambda-1/2)(2\\mu-1)$. Each constant is attained by the stated generalized orthogonal polynomial (or, in the odd ordinary-derivative cases, by an odd polynomial solving the displayed linear system), so all displayed inequalities are sharp.","pith_inferences":["The parity-splitting determinant construction suggests a general recipe for any even weight: restrict to the odd subspace, and the largest eigenvalue of the resulting moment matrix gives the Bernstein–Markov constant whenever the even subspace is governed by a classified orthogonal family.","For the generalized Hermite weight, one could test numerically whether the odd-$n$ roots $\\nu_{(n+1)/2}$ approach $2n$ as $n\\to\\infty$; the paper does not address the size of the parity gap.","The same duality lemma is specific to $L_2$; for $L_q$ with $q\\ne2$ no such eigenvalue/determinant characterization is known, so the exact-constant phenomenon described here is likely special to the Hilbert-space setting."],"forward_implications":["The classical constants are recovered as special cases: $\\lambda=0$ in Theorem 3 gives Schmidt's $\\sqrt{2n}$ for the Hermite weight, and $\\lambda=0$ in Theorem 4 gives the Gegenbauer constant $\\sqrt{n(n+2\\mu)}$.","For the ordinary derivative and even $n$, the generalized Hermite constant is $\\sqrt{2n}$, independent of the singularity parameter $\\lambda$.","For odd $n$, the exact constant is the largest positive root of a determinant of size $(n+1)/2$ (ordinary derivative), so it is computable in finitely many algebraic operations but not a simple closed formula in general.","In the Dunkl/Gegenbauer case, the extremal polynomial switches from $C_n^{(\\mu,\\lambda)}$ to $C_{n-1}^{(\\mu,\\lambda)}$ when $(2\\lambda-1)(2\\mu-1)>4$ and $n$ is below the threshold $n_0=(\\lambda-1/2)(2\\mu-1)$.","The quadratic inequalities of Theorem 5, with their equality cases, characterize the generalized Gegenbauer and Hermite polynomials as the unique extremizers, extending the $\\lambda=0$ results."],"supporting_citations":[{"why":"Supplies the coefficient recurrences (2.4) and (2.9) and the uniqueness of polynomial solutions used in Propositions 1–4 to classify extremal polynomials.","marker":"[4]"},{"why":"Establishes the classical Hermite constant sqrt(2n) that Theorem 1 extends and that the even-n case reproduces.","marker":"[16]"},{"why":"Gives the two-sided bounds (1.4) for the generalized Hermite weight that force the parity of the extremal polynomial in the proof of Theorem 1.","marker":"[6]"},{"why":"Determines the Jacobi-weight constant sqrt(n(n+α+β+1)) that Theorem 2 generalizes to λ>0.","marker":"[10]"},{"why":"Provides the Dunkl operator definition and the orthogonal polynomial framework for the generalized Hermite and Gegenbauer weights.","marker":"[7]"},{"why":"Supplies the λ=0 extremal inequalities that Theorem 5 reduces to in Corollary 6.","marker":"[3]"}],"fun_headline_variants":["Exact Bernstein-Markov constants for generalized Hermite and Gegenbauer","Sharp L2 derivative bounds: exact values and extremal polynomials","Dunkl operator case solved: exact L2 constants","Determinant roots yield exact extremal constants","Generalized weights: exact L2 Bernstein-Markov inequalities"],"cache_read_input_tokens":26112,"weakest_assumption_plain":"The full classification of polynomial solutions of the singular differential equations in Propositions 1–4 is taken from the coefficient recurrences (2.4) and (2.9) of reference [4]; if that classification missed any mixed-parity or additional polynomial solutions, the claimed maxima could be too low.","fun_headline_variants_meta":{"raw":{"variants":["Exact Bernstein-Markov constants for generalized Hermite and Gegenbauer","Sharp L2 derivative bounds: exact values and extremal polynomials","Dunkl operator case solved: exact L2 constants","Determinant roots yield exact extremal constants","Generalized weights: exact L2 Bernstein-Markov inequalities"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000334,"raw_usage":{"total_tokens":2116,"prompt_tokens":1469,"completion_tokens":647,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":1085,"completion_tokens_details":{"reasoning_tokens":563}},"tokens_in":1085,"tokens_out":647,"duration_ms":6271,"temperature":1.0,"reasoning_tokens":563,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:21:44.310791+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a fixed $\\lambda>0$ and an odd degree $n$ (for instance $\\lambda=1$, $n=3$), directly compute the largest eigenvalue of the matrix with entries $\\int_{\\mathbb{R}} p_i' p_j' w_\\lambda\\,\\mathrm{d}x$ relative to $\\int_{\\mathbb{R}} p_i p_j w_\\lambda\\,\\mathrm{d}x$ over a basis of $\\mathcal{P}_3$; the claimed constant is $\\sqrt{\\nu_2}$ from Example 1, and any ratio exceeding that value, or a largest eigenvalue differing from $\\nu_2$, would refute Theorem 1.","supporting_citations":[{"cited_title":"Ben Cheikh, M","cited_arxiv_id":null,"evidence_quote":"Supplies the coefficient recurrences (2.4) and (2.9) and the uniqueness of polynomial solutions used in Propositions 1–4 to classify extremal polynomials."},{"cited_title":"Preussischen Akad","cited_arxiv_id":null,"evidence_quote":"Establishes the classical Hermite constant sqrt(2n) that Theorem 1 extends and that the even-n case reproduces."},{"cited_title":"Draux, V","cited_arxiv_id":null,"evidence_quote":"Gives the two-sided bounds (1.4) for the generalized Hermite weight that force the parity of the extremal polynomial in the proof of Theorem 1."},{"cited_title":"Guessab, G.V","cited_arxiv_id":null,"evidence_quote":"Determines the Jacobi-weight constant sqrt(n(n+α+β+1)) that Theorem 2 generalizes to λ>0."},{"cited_title":"Dunkl, Y","cited_arxiv_id":null,"evidence_quote":"Provides the Dunkl operator definition and the orthogonal polynomial framework for the generalized Hermite and Gegenbauer weights."},{"cited_title":"Agarwal, G.V","cited_arxiv_id":null,"evidence_quote":"Supplies the λ=0 extremal inequalities that Theorem 5 reduces to in Corollary 6."}],"review_version":1}