{"id":"ca6cc2f7-e0b4-48e7-811b-764534b42b3b","arxiv_id":"2607.14961","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For normalised arithmetic functions, the near-diagonal coefficients of the D'Arcais polynomials are eventually log-concave in horizontal, vertical, and a new skew sense.","lead":"This paper defines a new 'skew' log-concavity condition and shows that for many arithmetic functions, the near-top coefficients of D'Arcais polynomials eventually satisfy horizontal, vertical, and skew log-concavity. It provides asymptotic support for a conjecture in combinatorial number theory.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: Theorem 2.1's reliance on Proposition 4.3(c) is secure for κ=2, including for negative-valued g.","rationale":"The reader's weakest_assumption flagged Proposition 4.3(c) as fragile because the induction is sketched and partly incorrect for other κ. On close inspection, the κ=2 case used by Theorem 2.1 is not fragile: the difference recurrence has a unique highest-degree term (m=2) whose leading coefficient is determined solely by g(2) and the leading coefficient of c_{k−1}. Negative values of g(m) for m>2 cannot cancel this term because their degrees are strictly lower. Therefore the polynomiality and leading-coefficient formula hold for all normalised Z-valued g with g(2)≠0. The auxiliary errors noted by the reader (Lemma 4.4, Prop 4.3(a) for κ≥3, Theorem 2.2(a) hypothesis) are real but do not bear on the main theorem. The central claim of Theorem 2.1 is mathematically sound, and the existing conditional verdict is appropriate only because of those auxiliary issues, not because of a flaw in the main proof. Hence no change to the reader's verdict is needed.","tokens_in":6329,"tokens_out":26080,"duration_ms":202900,"concrete_test":"Verify computationally for a test g with g(2)=1, g(3)=−10^6, g(4)=10^12: compute c_1,c_2,c_3 via recurrence (3.1) and confirm the leading coefficient of c_3 is 1/48 (=g(2)^3/(2^3·3!)), which would confirm Prop 4.3(c) for negative-valued g.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No significant objection identified. The central theorem depends on Proposition 4.3(c): for κ=2, c_k(n)=a^g_n(n−k) is a polynomial of degree 2k with leading coefficient g(2)^k/(2^k k!). Although the proof is sketched, the κ=2 case is robust. From (3.1), D_k(n)=c_k(n)-c_k(n-1)=Σ_{m=2}^{k+1} (n−1)!/(n−m)! g(m) c_{k+1−m}(n−m). By induction each c_r has degree 2r; the m=2 term has degree 1+2(k−1)=2k−1, while all m>2 terms have degree m−1+2(k+1−m)=2k+1−m≤2k−2. Thus no cancellation can affect the leading coefficient, which is g(2)^k/(2^{k−1}(k−1)!); Lemma 4.1 then gives lc(c_k)=g(2)^k/(2^k k!). This holds for arbitrary Z-valued g with g(2)≠0, negative values included. The auxiliary errors (Lemma 4.4 constant, Prop 4.3(a) for κ≥3, Theorem 2.2(a) hypothesis) do not affect Theorem 2.1.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies log-concavity properties of the D'Arcais polynomials P_n^g(X) and their normalised versions A_n^g(X)=n!P_n^g(X). It introduces a new notion of 'skew log-concavity' and proves two main results. Theorem 2.1 asserts that for any normalised Z-valued arithmetic function g with g(2)≠0, the polynomials A_n^g and P_n^g are eventually horizontally and vertically log-concave at the near-top coefficient positions n−k. Theorem 2.2 asserts eventual skew log-concavity for N-valued g with a gap condition, and everywhere for the Dirichlet identity g=e. The proofs are based on the recurrence (3.1) and on a polynomiality/leading-coefficient analysis of the top coefficients c_k(n)=a_n^g(n−k), together with a comparison of leading coefficients.","tokens_in":6674,"tokens_out":16039,"duration_ms":122629,"significance":"If Theorem 2.1 is correct, it is a clean and reasonably general asymptotic log-concavity statement for D'Arcais polynomials, extending earlier asymptotic results for σℓ in a natural direction. The leading-coefficient comparison idea is simple and potentially reusable. The paper also proposes a new notion, skew log-concavity. However, the submitted manuscript contains several false or internally inconsistent auxiliary statements, and the proof of the load-bearing Proposition 4.3 is only sketched. The central theorem appears sound, but the paper needs substantial revision before the claims are presented correctly.","major_comments":[{"comment":"The hypothesis is self-contradictory as written: it requires g(κ)≠0 and g(m)=0 for 2≤m≤κ. For every κ≥2 this includes m=κ and forces g(κ)=0, so only κ=1 is admissible, where the range is empty and the hypothesis holds for all normalised g. The intended condition is presumably κ≥2 and g(m)=0 for 2≤m≤κ−1, matching Proposition 4.3(a). As stated, the theorem is not supported by the proof, which invokes Proposition 4.3(a).","section":"§2.2, Theorem 2.2(a)"},{"comment":"The asserted degree formula fails for k<κ−1. In the displayed difference, the summation over m=κ,…,k+1 is empty, so a_n^g(n−k) is identically zero, not a polynomial of degree k+⌊k/(κ−1)⌋. For example, when κ=3 and k=1, the formula gives degree 1, but a_n^g(n−1)=g(2)(n−1)n/2=0 if g(2)=0. The statement needs the restriction k≥κ−1, or a separate zero-polynomial case. This does not invalidate Theorem 2.1, which only uses κ=2, but it is a false statement used in the proof of Theorem 2.2(a).","section":"§4.1, Proposition 4.3(a)"},{"comment":"This is the load-bearing step for Theorem 2.1, but the proof is only a sketch. The sentence 'by virtue of Lemma 4.2' hides the induction, and Lemma 4.2 itself is asserted as 'easily verifiable'. Since the leading-coefficient comparison in §4.2 depends on the exact leading coefficient g(2)^k/(2^k k!), the proof should contain a complete induction, including a demonstration that no cancellation occurs in the recurrence for κ=2 and arbitrary Z-valued g. This is especially important because the manuscript itself notes that cancellations may occur for other κ.","section":"§4.1, Proposition 4.3(c)"},{"comment":"The stated leading coefficient is false. For f(X)=a_dX^d+…, the coefficient of X^{2d−2} in f(X)^2−f(X+1)f(X−1) is d a_d^2, not 2d a_d^2. For example, when d=1, the expression is a_1^2, not 2a_1^2. The qualitative conclusion (eventual log-concavity) still holds because the correct coefficient is positive, but the lemma as written is incorrect and must be corrected.","section":"§4.3, Lemma 4.4"}],"minor_comments":[{"comment":"The statement should explicitly restrict k and n so that the coefficients n−k, n−k+1, and n−k−1 are in the natural ranges for the polynomials; otherwise the log-concavity inequalities may refer to undefined coefficients for small n.","section":"§2, Theorem 2.1"},{"comment":"The displayed formula for a_n^g(k)/n! is typeset densely; the summation over λ∈S_{l(µ)} and the nested indices could be reformatted for readability. This is a presentational issue only.","section":"§3.3, Theorem 3.1"},{"comment":"In the displayed definition of F_n(x), the index K should be lowercase k, and the condition 'for n±1' is ambiguous; it should say 'for n−1 and n+1'.","section":"§5.2"},{"comment":"The entries for general g are useful, but the table would benefit from a sentence indicating that they follow from the recurrence (3.1) and perhaps a sample derivation for the general-g row.","section":"§5.1, Table"},{"comment":"The notation n_0 in the definition of log-concavity at n_0 is used only once; it may be simpler to consistently write 'at m' or 'at n'.","section":"§1.2"}],"recommendation":"major_revision","confidential_remarks":"The main theorem (Theorem 2.1) appears mathematically sound; the referee independently verified the κ=2 induction behind Proposition 4.3(c). However, the manuscript contains several false or contradictory statements in auxiliary results, most notably Theorem 2.2(a) and Proposition 4.3(a), and the proof of the central proposition is too sketchy. These are fixable within the manuscript's scope, but a careful revision is required before the paper is acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: Theorem 2.1 is correct and is the real content of the paper. For any normalised Z-valued g with g(2)≠0, the near-diagonal coefficients a_n(n−k) are polynomial in n of degree 2k with the stated leading coefficient, and that gives eventual horizontal and vertical (or diagonal) log-concavity. The κ=2 case of Proposition 4.3 is the load-bearing step and it holds, including for negative values of g, as your stress-tester confirms. This is a genuine extension of Starr's asymptotic result to arbitrary g with g(2)≠0, and it is proved cleanly from the recurrence; no new heavy machinery.\n\nThe paper also introduces 'skew' log-concavity and proves an eventual statement for N-valued g under an initial-zero condition. That's a reasonable notion, but the paper doesn't clearly separate it from what Theorem 2.1 already proves: the proof of 'vertical log-concavity at n−k' uses c_k(n)^2 vs c_k(n+1)c_k(n−1), which is exactly the skew condition. Either the terminology needs to be reconciled or the authors should say explicitly that Theorem 2.1 covers the skew case for g(2)≠0 and Theorem 2.2 covers other κ for nonnegative g.\n\nSoft spots, in order of seriousness:\n\n1. Proposition 4.3(a) is false as stated for k < κ−1: the polynomial is constant, not degree k+floor(k/(κ−1)). The error is off-diagonal and doesn't touch Theorem 2.1, but Theorem 2.2(a) relies on it, so it has to be fixed.\n\n2. Lemma 4.4 has the wrong leading coefficient: it should be d a_d^2, not 2d a_d^2 (check f(X)=X^2). The positivity argument still works, so nothing downstream collapses.\n\n3. Theorem 2.2(a) allows κ=1, which makes the hypothesis vacuous; it should require κ≥2, with the identity case handled by (b). Also the proof of Proposition 4.3 is only sketched; for κ≥3 the degree formula needs a careful induction, especially since cancellations can occur if g is not nonnegative.\n\nNone of these damage the main theorem. The citation pattern is fine: Heim–Neuhauser's formula is used as input, not as a fitted result. The paper is short and mostly transparent.\n\nFor the right reader—anyone working on D'Arcais/partition polynomial log-concavity—it's worth a serious referee. Send it to review; the referee should ask for corrections to the auxiliary statements and a clearer separation of vertical vs skew, but the central result is in good shape.","headline":"Solid eventual log-concavity theorem for normalised D'Arcais polynomials; the new skew notion is mostly a repackaging, and the auxiliary lemmas have small errors that need fixing.","tokens_in":7083,"tokens_out":14068,"would_cite":true,"duration_ms":108664,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11F20","05A20"],"pacs":[],"model":"deepseek-v4-flash","headline":"For any normalised integer-valued arithmetic function g with g(2)≠0, the D'Arcais polynomials are eventually log-concave at every fixed distance from the top coefficient, both horizontally and vertically.","keywords":["D'Arcais polynomials","log-concavity","arithmetic functions","Dedekind eta function","Nekrasov–Okounkov polynomials","polynomial representation","skew log-concavity","asymptotic coefficients"],"falsifier":"Compute a_g_n(n−2) for a normalised integer-valued g with g(2)≠0 but with negative values, e.g. g(2)=1, g(3)=−1, using the recurrence (3.1) for n=1,...,10, and compare with the predicted polynomial (n−2)(n−1)n(3g(2)^2(n−3)+8g(3))/24. If the values deviate from this polynomial, or if the degree drops, Proposition 4.3(c) is false; if they match, the load-bearing premise survives for that test.","tokens_in":6263,"feed_emoji":"📈","tokens_out":10318,"duration_ms":74971,"temperature":0.7,"pith_summary":"This paper proves that, for any normalised integer-valued arithmetic function g with g(2)≠0, the D'Arcais polynomials satisfy log-concavity at each fixed distance k from the top coefficient, once the degree n is large enough. Both the horizontal inequality (within a single polynomial) and the vertical inequality (across consecutive degrees) hold eventually. The proof reduces these inequalities to a single observation: the top coefficient a_g_n(n−k) is exactly a polynomial in n, with a leading coefficient that makes the comparison positive. The paper also introduces a new skew log-concavity and proves it for functions whose first non-zero value beyond 1 occurs late, and everywhere for the trivial function.","feed_headline":"Near-top D'Arcais coefficients eventually log-concave","feed_subtitle":"A polynomial-fit proof shows both horizontal and vertical log-concavity near the top for all large n.","key_machinery":"The engine is the polynomial representation of near-top coefficients (Proposition 4.3): for fixed k, a_g_n(n−k) is a polynomial in n of degree 2k (when g(2)≠0) with leading coefficient g(2)^k/(2^k k!). This turns a log-concavity inequality at n−k into a comparison of leading coefficients of polynomials in n, so eventual validity is decided by a positive number. The paper also uses this representation for the new skew-log-concavity condition, where the same polynomiality (under a non-negativity assumption) yields eventual log-concavity via a general lemma about polynomial log-concavity.","core_discovery":"The central claim is Theorem 2.1: if g is a normalised Z-valued arithmetic function with g(2)≠0, then for every fixed k there exist thresholds n_h and n_v such that the coefficient sequence of A_g_n (and hence P_g_n) is horizontally log-concave at n−k for all n≥n_h, and vertically log-concave at n−k for all n≥n_v. The proof shows that a_g_n(n−k) equals a polynomial c_k(n) of degree exactly 2k with leading coefficient g(2)^k/(2^k k!), so the difference c_k(n)^2 − c_{k+1}(n)c_{k−1}(n) has positive leading coefficient; eventually the sign is right. The same leading-coefficient argument handles the vertical direction because shifting n by ±1 does not change the leading term.","pith_inferences":["If Proposition 4.3(c) can be made fully rigorous for κ=2 (the paper itself notes the induction is sketched), Theorem 2.1 is unconditional; a natural next step is to compute explicit thresholds for concrete functions like the sum-of-divisors function to test sharpness.","The leading-coefficient method suggests that similar eventual log-concavity results may hold for other sequences of polynomials defined by exponential generating products, as long as a polynomial representation of the relevant coefficients is available.","The paper leaves open an equivalence condition between vertical/horizontal and skew log-concavity; proving such an equivalence would give a unified framework for all three notions.","For functions with a later first non-zero value (κ≥3), the paper's own remark shows cancellations can occur when g takes negative values, so the non-negativity assumption in the skew-log-concavity result is likely essential rather than technical."],"forward_implications":["For every normalised integer arithmetic function with nonzero value at 2, both horizontal and vertical log-concavity hold eventually at every fixed distance k from the top coefficient.","The same conclusion transfers from the renormalised polynomials A_g_n to the standard D'Arcais polynomials P_g_n; horizontal log-concavity is in fact equivalent for the two families.","The existence of thresholds n_h(g,k) and n_v(g,k) is established, though explicit values are not given; the result is qualitative.","For non-negative functions whose first non-zero value beyond 1 occurs at κ≥2, the new skew log-concavity holds eventually; for the trivial function e it holds with no threshold.","The paper's computations give closed forms for the top three coefficients of A_g_n for a general g, illustrating the polynomial pattern."],"fun_headline_variants":["D'Arcais near-top coefficients eventually log-concave","Proven: D'Arcais polynomials log-concave near top, both ways","Log-concavity proven for D'Arcais coefficients near the top","D'Arcais: near-top log-concavity, horizontal and vertical, eventually","Polynomial fit shows D'Arcais log-concavity near top for large n"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The proof relies on the claim that, for g(2)≠0, the top coefficient a_g_n(n−k) is exactly a polynomial in n of degree 2k with leading coefficient g(2)^k/(2^k k!); if this polynomial description fails, the leading-coefficient comparison that forces eventual log-concavity collapses.","fun_headline_variants_meta":{"raw":{"variants":["D'Arcais near-top coefficients eventually log-concave","Proven: D'Arcais polynomials log-concave near top, both ways","Log-concavity proven for D'Arcais coefficients near the top","D'Arcais: near-top log-concavity, horizontal and vertical, eventually","Polynomial fit shows D'Arcais log-concavity near top for large n"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000412,"raw_usage":{"total_tokens":1914,"prompt_tokens":635,"completion_tokens":1279,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":379,"completion_tokens_details":{"reasoning_tokens":1173}},"tokens_in":379,"tokens_out":1279,"duration_ms":9943,"temperature":1.0,"reasoning_tokens":1173,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T00:36:52.929618+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute a_g_n(n−2) for a normalised integer-valued g with g(2)≠0 but with negative values, e.g. g(2)=1, g(3)=−1, using the recurrence (3.1) for n=1,...,10, and compare with the predicted polynomial (n−2)(n−1)n(3g(2)^2(n−3)+8g(3))/24. If the values deviate from this polynomial, or if the degree drops, Proposition 4.3(c) is false; if they match, the load-bearing premise survives for that test.","supporting_citations":[],"review_version":1}