{"id":"529a25a8-bfa3-400c-8021-1abc2ad24a52","arxiv_id":"2501.02358","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper characterizes when discrete best uniform approximants alternate, proves a Sturm oscillation bound for Jacobi-matrix eigenfunctions, and derives monotone Fourier coefficients for polynomials with removed largest zeros.","lead":"This paper proves discrete versions of Chebyshev alternation and Sturm oscillation for functions on finite integer grids. The results give a complete description of when best uniform fits must alternate and solve an extremal problem for polynomials with spectral gaps.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proof of Theorem 1.11 applies Δ-Rolle inequalities to f=V/ψ1, which violates the stated hypothesis f(q+1)=0; no replacement for the actual boundary condition is supplied.","rationale":"Read the paper in good faith. The central claim is Theorem 1.11, the discrete Sturm oscillation theorem, and its proof is built on Lemma 3.7 together with Step 3. Scrutinizing Lemma 3.7, I found that it invokes Lemmas 3.5/3.6 for two discrete Rolle-type comparisons: K(∇g) ≥ K(g) and K(Δ(V/ψ1)) ≥ K(V/ψ1). The first comparison is fine, since g has zeros at −1 and q and the ∇-inequality only needs f(−1)=0. The second comparison is problematic: Lemmas 3.5 and 3.6 are stated with f(q+1)=0, and their Δ-inequalities are proved using that condition. The actual f = V/ψ1 satisfies f(q+1)=f(q) for η≠0, not f(q+1)=0. The paper does not supply a variant of the discrete Rolle theorem for this boundary condition. This is not a matter of unproved standard background such as Favard's theorem or zero interlacing; it is an internal step of the main proof. A small explicit numerical search over a Jacobi system with η≠0 could either expose a counterexample to Lemma 3.7 or give confidence that the missing inequality is true and merely unproved. In either case, the paper as written has a real soft spot in the derivation of the central theorem, so I would move from ACCEPT to CONDITIONAL rather than reject outright.","tokens_in":31411,"tokens_out":30998,"duration_ms":266871,"concrete_test":"Run a numerical search over small q (e.g., q=3) with the Chebyshev Jacobi data α_l=0, β_l=γ_l=ρ_l=1 and η=2, so that f(q+1)=f(q), enumerating V in span{ψ_2,...,ψ_{q+1}} to test whether K(Δ(V/ψ1)) ≥ K(V/ψ1) for K=N,S−,S+. If any V gives a strict violation, Lemma 3.7 is false and Theorem 1.11 requires a different argument; if no violation is found, attempt an analytical proof of the Δ-Rolle inequalities under f(q+1)=f(q) and patch the proof of Lemma 3.7 accordingly.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The proof of Theorem 1.11 rests on Lemma 3.7, which asserts K(V1) ≥ K(V) for K = N, S−, S+. The chain of inequalities in the proof of Lemma 3.7 requires K(Δ(V/ψ1)) ≥ K(V/ψ1), and this is justified by citing Lemmas 3.5/3.6. Those lemmas are stated under the assumption f(q+1)=0, and their Δ-inequalities (3.18)/(3.20) are exactly the parts whose proofs use that Dirichlet condition, via the final clause of Lemma 3.4 and the remark before Lemma 3.6. For f = V/ψ1, the boundary condition (1.12) gives instead f(q+1) = V(q+1)/ψ1(q+1) = ηV(q)/(ηψ1(q)) = f(q), so f(q+1)=0 only in the special case η=0. Thus the monotonicity step K(V1) ≥ K(V) — the engine that yields both m−1 ≤ K(V) and K(V) ≤ n−1 in Step 3 — is not proved for general η ∈ R. The paper neither proves a version of the discrete Rolle inequalities for the boundary condition f(q+1)=f(q) nor shows that the class {V/ψ1} enjoys extra structure that makes them hold. This is a genuine gap in the written proof of the central claim.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves three main results: (i) Theorem 1.4, a characterization of discrete Chebyshev systems (T Z-systems) via the existence of Chebyshev alternance sets in the best uniform approximation of discrete functions; (ii) Theorem 1.11, a discrete Sturm oscillation theorem for eigenfunctions of the discrete Sturm-Liouville problem (1.12), asserting that any nontrivial linear combination V = Σ_{k=m}^n a_k ψ_k satisfies m−1 ≤ S_−(V) ≤ N(V) ≤ S_+(V) ≤ n−1; and (iii) Theorems 1.14 and 1.16, giving monotonicity of Fourier coefficients of polynomials with removed largest zeros and a solution of a Yudin-type extremal problem. The proofs are largely self-contained and rely on detailed determinant and sign-change arguments.","tokens_in":31660,"tokens_out":18850,"duration_ms":150087,"significance":"If the main theorems hold, this is a substantial contribution: Theorem 1.11 supplies a long-sought discrete analogue of Sturm's oscillation theorem, yielding Corollary 1.12 (the eigenfunctions form a T Z-system) and Corollary 1.13 (a discrete Sturm-Hurwitz spectral gap theorem). Theorem 1.4 is a clean characterization of when a discrete best-uniform-approximation problem admits a full alternance set. Theorem 1.14 strengthens earlier results of Cohn-Kumar on nonnegative coefficients, and Theorem 1.16 solves a Yudin-type extremal problem, with the appendix providing explicit determinant sign computations. The proofs are written in full detail and are mostly rigorous. However, the gap identified in the major comment below currently leaves Theorem 1.11 unproved for general boundary parameter η, which is a central claim of the paper.","major_comments":[{"comment":"The proof of Lemma 3.7, which is the engine of Theorem 1.11, applies Lemmas 3.5 and 3.6 to the function f = V/ψ_1 in order to conclude K(Δ(V/ψ_1)) ≥ K(V/ψ_1) for K = N, S_−, S_+. Lemmas 3.5 and 3.6 are explicitly stated under the hypothesis f(q+1)=0, and the parts (3.18) and (3.20) for the forward difference Δf rely exactly on that Dirichlet condition through the final clause of Lemma 3.4. For f = V/ψ_1, the boundary condition in (1.12) gives, for η ≠ 0, f(q+1) = V(q+1)/ψ_1(q+1) = ηV(q)/(ηψ_1(q)) = f(q), not f(q+1)=0. Thus the hypotheses of Lemmas 3.5 and 3.6 are not satisfied, and no substitute argument is provided. This is not a merely technical gap: for example, with q=1, η=1, and the Jacobi coefficients α_l=0, β_l=γ_l=ρ_l=1, the function f=ψ_2/ψ_1 satisfies f(2)=f(1) but S_−(Δf)=0 < 1 = S_−(f), so the announced discrete Rolle inequality fails under the actual boundary condition. Consequently, the monotonicity step K(V_1) ≥ K(V) is unproved for general η, and Theorem 1.11 is not established for the stated range η ∈ R. The proof does work for η=0, but the theorem and its corollaries are claimed for arbitrary real η.","section":"§3.1, Lemma 3.7"}],"minor_comments":[{"comment":"Equation (3.14) appears to have a sign error: the correct identity is d_ν(λ_l−λ_k)ψ_l(ν)ψ_k(ν) = −∇(w_ν{ψ_l(ν)Δψ_k(ν) − ψ_k(ν)Δψ_l(ν)}). This does not affect the later arguments, since only zero counts and sign-change counts of g are used, but it should be corrected.","section":"§3.1, equation (3.14)"},{"comment":"In the proof of Lemma 3.5, the sentence 'Therefore, N(Δf) ≥ N(f)' in the paragraph proving (3.17) should read 'N(∇f) ≥ N(f)', since the inequality being established is for ∇f.","section":"§3.1, proof of Lemma 3.5"},{"comment":"In the second bullet of the proof of Lemma 3.1, the notation 'S−(f, [m,q]_Z)' should likely be 'S−(f, [m,n]_Z)' for consistency with the other terms in the displayed inequality.","section":"§3.1, proof of Lemma 3.1"},{"comment":"In the chain (3.21), the expression '∇g(s)/(d_l ψ_1(s))' uses an undefined index 'l'; it should be 'd_s ψ_1(s)' (or simply a positive factor), since ∇g(s) = d_s ψ_1(s) V_1(s).","section":"§3.1, proof of Lemma 3.7"},{"comment":"There are numerous typographical and grammatical errors (e.g., 'Fourier' for 'Fourier' in a reference, inconsistent notation for intervals). A careful proofreading pass is recommended.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"This is a strong and substantial paper, and the identified gap in Lemma 3.7 is localized yet load-bearing: it affects the proof of the main Sturm oscillation theorem for general boundary parameter η. I would encourage the editor to allow a revision. The authors should either supply a correct discrete Rolle-type inequality for the boundary condition f(q+1)=f(q) (or otherwise justify the application to f=V/ψ_1), or restrict Theorem 1.11 and its corollaries to η=0. The remainder of the paper, including Theorem 1.4, Theorem 1.14, and the determinant computations in the appendix, appears sound to me."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the paper's Theorem 1.4—the iff characterization of discrete Chebyshev systems via alternance—is a clean and correct piece of work, and the determinant proof in Section 4 for the monotone Fourier coefficients is independent and looks solid. Second, the advertised discrete Sturm oscillation theorem (Theorem 1.11) is not actually proved as written. The stress-test note is right: Lemma 3.7 applies the discrete Rolle inequalities to f = V/ψ1, but those lemmas require f(q+1)=0. For the eigenfunctions, the boundary condition gives ψ(q+1)=ηψ(q), so V/ψ1 satisfies f(q+1)=f(q) when η≠0, and when η=0 the ratio is 0/0 at q+1. The paper never proves a Rolle-type inequality for that boundary condition. The step K(Δ(V/ψ1)) ≥ K(V/ψ1) is the engine of Lemma 3.7 and therefore of the whole theorem; without it, the inequalities m−1 ≤ S−(V) and S+(V) ≤ n−1 don't follow from the written argument. This is not a trivial typo: the proof of Theorem 1.11 rests on it, and Corollaries 1.12 and 1.13 inherit the problem.\n\nWhat the paper does well: the equivalence in Theorem 1.4 is a genuine contribution, with useful lemmas about alternating sets, and the application to coefficient monotonicity in Theorem 1.14 is proved by a self-contained determinant argument (Section 4) that does not appear to depend on the Sturm theorem. The Yudin application is a reasonable extension if the coefficient result holds. The writing is careful and the citations are appropriate.\n\nSo my overall take: this is a paper with real value, but the central result is currently unproven. It deserves a serious referee—the problem is nontrivial and the rest of the paper is worth engaging with—but I would not accept it in this form. I'd ask the authors to either prove the Δ-Rolle inequality for the boundary condition f(q+1)=f(q), or restructure the proof to avoid that step. If they do, the paper would be a solid accept.\n\nMy recommendation to you: if you're editor, send it to review, but with clear instructions that the Sturm proof needs scrutiny. If you're reading it for your own work, cite Theorem 1.4; be cautious about Theorem 1.11 until the gap is closed.","headline":"The Chebyshev-system characterization is solid, but the proof of the central discrete Sturm theorem has a real boundary-condition gap that a referee should require the authors to fix.","tokens_in":32240,"tokens_out":10934,"would_cite":true,"duration_ms":97351,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["41A50","39A21","52A40"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves exact discrete analogues of Chebyshev's alternation theorem and Sturm's oscillation theorem for eigenfunctions of Jacobi matrices, and applies them to spectral-gap polynomials.","keywords":["Chebyshev system","best uniform approximation","Sturm oscillation theorem","discrete polynomials","spectral gap problem","Jacobi matrix","T_Z-system","orthogonal polynomials"],"falsifier":"Take a small Jacobi problem, say $q=2$ with concrete positive coefficients such as $\\alpha=(0,0,0)$, $\\beta=\\gamma=(1,1)$, $\\rho=(1,1,1)$, $\\eta=0$, compute $\\psi_1,\\psi_2,\\psi_3$ by the recurrence, and list the sign-change counts of all combinations $\\sum_{k=m}^n a_k\\psi_k$; any combination violating $m-1\\le S_-(V)\\le S_+(V)\\le n-1$ would refute the central theorem.","tokens_in":31178,"feed_emoji":"🔢","tokens_out":9736,"duration_ms":85950,"temperature":0.7,"pith_summary":"On a finite integer grid, the paper finds the precise condition under which best uniform approximation behaves exactly as on an interval: a discrete system of $n$ functions admits alternating-error sets of length $n+1$ for every target function if and only if it is a Chebyshev system on the grid. It then shows that eigenfunctions of a discrete Sturm-Liouville problem form such systems, and proves the discrete Sturm oscillation theorem: every nontrivial combination of eigenfunctions with indices $m$ through $n$ has between $m-1$ and $n-1$ zeros, counting sign changes. This yields a discrete Sturm-Hurwitz spectral gap theorem and, as an application, a monotonicity result for the Fourier coefficients of orthogonal polynomials with their largest zeros removed, which in turn settles an extremal problem for polynomials with a spectral gap.","feed_headline":"Eigenfunction sums obey discrete Sturm oscillation bounds","feed_subtitle":"Chebyshev alternation holds on integer grids, and spectral-gap sums must change sign at least m-1 times.","key_machinery":"The load-bearing objects are the eigenfunctions $\\psi_k(\\nu)=P_\\nu(\\lambda_k)$ of the discrete Sturm-Liouville problem (1.12), generated by a Jacobi matrix with positive off-diagonal coefficients, and the two notions of discrete zero: a first-type zero where $f(\\nu)=0$, and a second-type zero where $f(\\nu-1)f(\\nu)<0$. The argument combines Favard's theorem (positivity of the coefficients produces a positive orthogonal measure), the interlacing of zeros of orthogonal polynomials, a Christoffel-Darboux identity, and a discrete version of Liouville's method: multiplying $V$ by $(\\lambda_1-\\lambda_k)^r$ and letting $r\\to\\pm\\infty$ forces $V_r$ to converge to a single eigenfunction, transferring the known zero count of that eigenfunction to $V$.","core_discovery":"The central claim is a discrete analogue of Sturm's theorem. For the eigenvectors $\\psi_k(\\nu)=P_\\nu(\\lambda_k)$ of the Jacobi Sturm-Liouville problem (1.12), any nontrivial polynomial $V(\\nu)=\\sum_{k=m}^{n} a_k \\psi_k(\\nu)$ with $1\\le m\\le n\\le q+1$ satisfies $m-1\\le S_-(V)\\le N(V)\\le S_+(V)\\le n-1$, where $S_-$ and $S_+$ are the least and largest numbers of sign changes after zero values are replaced, and $N$ counts vanishings together with sign changes between neighbouring points. Together with Theorem 1.4, which characterizes the systems for which every best uniform approximant has a Chebyshev alternance set of length $n+1$ as exactly the Chebyshev ($T_{\\mathbb{Z}}$) systems, this implies that $\\{\\psi_k\\}_{k=1}^n$ is a $T_{\\mathbb{Z}}$-system and that any discrete function with a spectral gap starting at $m$ has at least $m-1$ sign changes.","pith_inferences":["The determinant condition of Theorem 1.4 gives a finite, checkable certificate for a discrete system to be Chebyshev: one only needs to verify that all $n\\times n$ interpolation determinants have one sign, which could be automated for numerical basis selection.","The discrete Sturm-Hurwitz statement may extend to any orthonormal family with a three-term recurrence and positive transfer coefficients, suggesting a general finite-dimensional uncertainty principle in which the size of the spectral gap controls the minimum number of oscillations.","The monotonicity of the normalized coefficients is stronger than positivity and could yield quantitative lower bounds for extremal constants of spectral-gap polynomials, not just the extremal values computed in the two cases covered by Theorem 1.16.","The Krein property in Theorem 1.16 is used only to ensure that squares of basis expansions stay in the nonnegative cone; if it fails, the extremal polynomial may still be extremal, but the present proof would not cover it."],"forward_implications":["Best uniform approximation on a finite grid has an alternating-error set exactly for $T_{\\mathbb{Z}}$-systems, so alternation-based (Remez-type) algorithms are justified precisely for this class.","Eigenfunction systems of discrete Sturm-Liouville problems are $T_{\\mathbb{Z}}$-systems, giving unique best approximants and Chebyshev alternance for such bases.","A discrete spectral-gap theorem holds: if $f=\\sum_{k=m}^{q+1} a_k\\psi_k$, then $f$ has at least $m-1$ sign changes on $[0,q]_{\\mathbb{Z}}$.","In the expansion of $P_{q+1}(\\lambda)$ divided by its $m+1$ largest zero factors, the Fourier coefficients, after normalization by $P_l(b)$, are strictly monotone and positive, strengthening earlier nonnegativity results.","The extremal problem for polynomials with spectral gap and nonnegative coefficients is solved: the extremal value is the $(m+1)$-st zero of the corresponding orthogonal polynomial, with a unique extremizer, whenever the Krein property holds."],"supporting_citations":[{"why":"Supplies Haar's and Chebyshev's theorems and the T0-system uniqueness result that Theorem 1.4 extends to the discrete grid.","marker":"[15]"},{"why":"Gives the determinant-definition of T-systems used to formulate the TZ-system criterion via same-sign determinants.","marker":"[14]"},{"why":"Introduces S-(f), S+(f), generalized zeros, and proves a partial oscillation bound for oscillatory Jacobi matrices that Theorem 1.11 generalizes.","marker":"[9]"},{"why":"Provides the Favard-type theorem by which positivity of the Jacobi coefficients yields the orthogonal polynomial measure used throughout Section 3.","marker":"[5]"},{"why":"Provides Sturm oscillation and comparison theorems and the eigenvalue-count identity (Theorem 1.9) from which the eigenfunction zero counts are derived.","marker":"[23]"},{"why":"Supplies the interlacing of zeros of orthogonal polynomials used in Theorem 1.10 and in the coefficient monotonicity proof.","marker":"[25]"},{"why":"Gives the modern Liouville method for Sturm's theorem that the discrete proof of Theorem 1.11 adapts.","marker":"[3]"},{"why":"Establishes the nonnegativity of the Fourier coefficients in expansion (1.14); Theorem 1.14 strengthens this to strict monotonicity.","marker":"[4]"},{"why":"Solves the spectral-gap extremal problem without the nonnegative-coefficient constraint, providing the upper bound used in Theorem 1.16.","marker":"[13]"},{"why":"Defines the Krein property for products of orthogonal polynomials, the positivity condition that Theorem 1.16 assumes to certify extremizers.","marker":"[16]"}],"fun_headline_variants":["Discrete Sturm: sums must have at least m-1 sign changes","Integer-grid Sturm: sums oscillate at least m-1 times","Spectral-gap sums obey discrete Sturm bounds","Discrete Sturm oscillation: m-1 to n-1 sign changes","Chebyshev systems guarantee Sturm sign changes on integer grids"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole Sturm part assumes that the Jacobi coefficients $\\beta_l,\\gamma_l,\\rho_l$ are positive and the boundary condition has the form $\\psi(q+1)=\\eta\\psi(q)$; if any coefficient changes sign, the interlacing of zeros and the oscillation bounds in Theorem 1.11 need not survive.","fun_headline_variants_meta":{"raw":{"variants":["Discrete Sturm: sums must have at least m-1 sign changes","Integer-grid Sturm: sums oscillate at least m-1 times","Spectral-gap sums obey discrete Sturm bounds","Discrete Sturm oscillation: m-1 to n-1 sign changes","Chebyshev systems guarantee Sturm sign changes on integer grids"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00127,"raw_usage":{"total_tokens":5236,"prompt_tokens":1021,"completion_tokens":4215,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":637,"completion_tokens_details":{"reasoning_tokens":4133}},"tokens_in":637,"tokens_out":4215,"duration_ms":28550,"temperature":1.0,"reasoning_tokens":4133,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:14:09.927782+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a small Jacobi problem, say $q=2$ with concrete positive coefficients such as $\\alpha=(0,0,0)$, $\\beta=\\gamma=(1,1)$, $\\rho=(1,1,1)$, $\\eta=0$, compute $\\psi_1,\\psi_2,\\psi_3$ by the recurrence, and list the sign-change counts of all combinations $\\sum_{k=m}^n a_k\\psi_k$; any combination violating $m-1\\le S_-(V)\\le S_+(V)\\le n-1$ would refute the central theorem.","supporting_citations":[{"cited_title":"Laurent, Approximation et Optimisation, Herman n, Paris, 1972","cited_arxiv_id":null,"evidence_quote":"Supplies Haar's and Chebyshev's theorems and the T0-system uniqueness result that Theorem 1.4 extends to the discrete grid."},{"cited_title":"Karlin and W.J","cited_arxiv_id":null,"evidence_quote":"Gives the determinant-definition of T-systems used to formulate the TZ-system criterion via same-sign determinants."},{"cited_title":"Gantmacher and M","cited_arxiv_id":null,"evidence_quote":"Introduces S-(f), S+(f), generalized zeros, and proves a partial oscillation bound for oscillatory Jacobi matrices that Theorem 1.11 generalizes."},{"cited_title":"Chihara, An Introduction to Orthogonal Polynomials , Gordon and Breach Science Publishers, New York–London–Paris, 1978","cited_arxiv_id":null,"evidence_quote":"Provides the Favard-type theorem by which positivity of the Jacobi coefficients yields the orthogonal polynomial measure used throughout Section 3."},{"cited_title":"Simon, Sturm oscillation and comparison theorems , Amrein, Werner O","cited_arxiv_id":null,"evidence_quote":"Provides Sturm oscillation and comparison theorems and the eigenvalue-count identity (Theorem 1.9) from which the eigenfunction zero counts are derived."},{"cited_title":"Szeg¨ o, Orthogonal Polynomials, AMS, New York, 1959","cited_arxiv_id":null,"evidence_quote":"Supplies the interlacing of zeros of orthogonal polynomials used in Theorem 1.10 and in the coefficient monotonicity proof."},{"cited_title":"Sturm's theorem on zeros of linear combinations of eigenfunctions","cited_arxiv_id":"1706.08247","evidence_quote":"Gives the modern Liouville method for Sturm's theorem that the discrete proof of Theorem 1.11 adapts."},{"cited_title":"Cohn and A","cited_arxiv_id":null,"evidence_quote":"Establishes the nonnegativity of the Fourier coefficients in expansion (1.14); Theorem 1.14 strengthens this to strict monotonicity."},{"cited_title":"Ivanov, Yudin–Hermite extremal problems for polynomials , Math","cited_arxiv_id":null,"evidence_quote":"Solves the spectral-gap extremal problem without the nonnegative-coefficient constraint, providing the upper bound used in Theorem 1.16."},{"cited_title":"Levenshtein, Boundaries for packings of metric spaces and some application s, Problems of Cyber- netics 40 (1983), 43–110","cited_arxiv_id":null,"evidence_quote":"Defines the Krein property for products of orthogonal polynomials, the positivity condition that Theorem 1.16 assumes to certify extremizers."}],"review_version":1}