{"id":"343ed4d8-f706-45d2-ba99-70952c409242","arxiv_id":"2412.11879","paper_version":5,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For every root system, the Witten zeta function vanishes to order at least the rank at negative even integers, and its leading coefficient is a Q-linear combination of Hurwitz zeta values.","lead":"Witten zeta functions add up powers of the sizes of a symmetry group's representations. This paper proves they vanish to order at least the rank at every negative even integer, and describes the leading coefficient through Hurwitz zeta values and the highest root.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"E8 case check in Proposition 7.7 is load-bearing, and the induction's group-theoretic step (Lemma 7.10, Eq. 7.4) is not valid as stated: the quotient does not generally split as Z/qZ ⊕ Z^{n-1}/AZ^{n-1}.","rationale":"The reader correctly identifies the unexpanded E8 case check in Proposition 7.7 as the weakest load-bearing point, and my read agrees with the CONDITIONAL verdict. However, the concern is more acute than a mere omission of routine casework: the proof of Lemma 7.10 relies on an unjustified group-theoretic splitting. The quotient of Z^n by the row lattice of a block matrix [[A,0],[v,q]] is not always isomorphic to Z/qZ ⊕ Z^{n-1}/AZ^{n-1}; a simple 2×2 counterexample gives exponent 4 when the stated LCM formula gives 2. Therefore, even the cases that are 'checked exactly as before' may conceal a failure of the induction step, unless root-system-specific properties prevent such block matrices from occurring. The E8 nodes i=7 and i=8 are the most concerning because they combine q>1 with a sub-root system whose E-set contains numbers >1; a failure there could introduce denominators 8 or 9, which are not in H(E8), directly invalidating the denominator set T in Theorem 1.5. The computational test I propose settles the matter definitively: if no E8 subset yields an exponent outside {1,2,3,4,5,6}, then Proposition 7.7 is true and the proof can be repaired by a Smith-normal-form argument; if such a subset exists, Theorem 1.5 as stated is false. The analytic core of the paper, including Theorem 1.2 on vanishing orders, does not depend on Proposition 7.7 and appears sound, so the verdict should remain CONDITIONAL pending this verification.","tokens_in":23085,"tokens_out":32100,"duration_ms":281805,"concrete_test":"Use a computational algebra system (e.g., SageMath) to enumerate, up to Weyl group action, all 8-element subsets S of positive roots of E8, compute the Smith normal form of the 8×8 coordinate matrix M_S with respect to the simple roots, and record the largest invariant factor (the exponent of L(E8)/Span_Z S). Verify that every value lies in {1,2,3,4,5,6}. As a cheaper check, fix each deleted node i=1,...,8, enumerate subsets S whose first 7 roots normalize into Φ_{∆−{α_i}}, and compute the exponent of the full block matrix directly, testing whether it can exceed lcm(q, level(A)); in particular, check i=7 and i=8 for exponents 9 or 8.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader's conditional is well placed, and the concern is sharper than a missing E8 expansion. Lemma 7.10 asserts, in equation (7.4), that Z^n/rowspan(M) is isomorphic to Z/qZ ⊕ Z^{n-1}/AZ^{n-1} for the block matrix M = [[A,0],[v,q]]. This is false in general: for A=[2], v=[1], q=2, the matrix M has Smith normal form diag(1,4), so the quotient is Z/4Z with exponent 4, while lcm(q, level(A)) = lcm(2,2) = 2. Thus the induction's bound E_i(Φ) = {LCM(q,e)} is not justified by the given argument. For E8, the dangerous cases are i=7 (q=3, E(E6×A1) contains 3) and i=8 (q=2, E(E7) contains 4), where the true exponent could be 9 or 8 respectively, both outside H(E8) = {2,3,4,5,6}; Theorem 1.5's denominator set T would then be wrong. The paper's phrase 'one checks exactly as before' does not resolve this, because the asserted splitting is precisely the step that can fail.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops an analytic method for the Witten zeta function ζΦ(s) of a root system, based on an integral representation whose integrand is a product of Hurwitz zeta functions. The main analytic result, Theorem 1.2, gives a lower bound of rank on the order of vanishing at negative even integers, together with partial results at negative odd integers. The main arithmetic result, Theorem 1.5 (restated as Theorem 5.1 in the body), describes the leading Taylor coefficient at negative integers as a Q-linear combination of products of Hurwitz zeta values with parameters whose denominators divide elements of H(Φ), the coefficients of the highest root. The proof of Theorem 1.2 rests on an elementary pole-order lemma (Lemma 3.1) and on the factorization of the Poincaré polynomial of the Weyl group. The proof of Theorem 1.5 rests on a triangulation lemma (Lemma 7.3) and on a root-lattice exponent computation (Propositions 7.4 and 7.7).","tokens_in":23392,"tokens_out":17466,"duration_ms":160180,"significance":"If the arguments are completed, the paper would settle the Kurokawa–Ochiai conjecture in a strong form for Witten zeta functions of compact simply connected Lie groups and would give the first general arithmetic description of the leading coefficient at negative integers. The analytic lemma is elementary, self-contained, and potentially useful beyond this setting; the derivation of Onodera's A2 formula from the general method is a nice illustration. The paper is clearly written and honestly credits prior and independent work, including Onodera's unpublished results. However, the arithmetic theorem depends on a combinatorial root-system statement whose proof currently contains a load-bearing gap.","major_comments":[{"comment":"The isomorphism asserted in Eq. (7.4) is false in general. The quotient L(Φ)/SpanZ(S) is the cokernel of the block matrix M = [[A,0],[v,q]], and it is not generally isomorphic to Z/qZ ⊕ Z^{n-1}/AZ^{n-1}. For example, take n=2, A=[2], v=[1], q=2. Then M has Smith normal form diag(1,4), so the quotient is Z/4Z, whereas the claimed right-hand side is Z/2Z ⊕ Z/2Z; the exponent is 4, not lcm(q, level(A)) = 2. This is load-bearing: the induction bound E_i(Φ) = {LCM(q,e)} in Eq. (7.3) is not justified by the given argument. For E8, the cases i=7 and i=8 can produce exponents 9 and 8 respectively, both outside H(E8) = {2,3,4,5,6}, so Theorem 5.1 and hence Theorem 1.5 are not established as written.","section":"Section 7.1, E8 step of Proposition 7.7"},{"comment":"The E8 verification in the proof of Proposition 7.7 is incomplete even apart from Lemma 7.10. The text gives details for i=4,5,7 and says 'One checks exactly as before' for the remaining cases. The dangerous cases i=7 (q=3, E(E6×A1) contains 3) and i=8 (q=2, E(E7) contains 4) are exactly the ones where the invalid splitting in Eq. (7.4) would matter. A correct proof of E(Φ)=H(Φ)∪{1} needs either a valid induction step or a direct computation of D(Φ) for the exceptional root systems; the current manuscript provides neither.","section":"Section 7, Lemma 7.3"},{"comment":"The proof of Lemma 7.3 contains an assertion that is not justified and is in fact false for some maximal triangulations. The text claims that in a maximal triangulation with vertices in P, all faces are of the form l_k(x) ∈ Z. This is not true: for B = [[1,0,1],[0,1,1]] and the triangulation of [0,1]^2 by the diagonal from (0,0) to (1,1), the diagonal edge is not contained in any hyperplane l_k ∈ Z, and the hyperplane l_3 = x+y = 1 meets the interior of one triangle without providing a new vertex in P for a refinement. The lemma may be true (for example, via the standard polyhedral subdivision by the hyperplanes l_i ∈ Z), but the supplied maximal-triangulation argument does not prove it. Since Lemma 7.3 is used to control the denominators in Proposition 7.4, this proof gap is load-bearing for Theorem 5.1 as well.","section":"Section 7, Lemma 7.3"}],"minor_comments":[{"comment":"There are corrupted characters in the manuscript, for example 'Leftr⫯g⊸tl⫯ne⇒' in the display after the definition of E(Φ); these should be fixed to normal implication arrows.","section":"Section 1, paragraph after Theorem 1.2"},{"comment":"The sentence 'even in rank 2, it already yields deep conclusions' is followed by a display with several identities; it would help the reader if the variables n, m, and the coefficient extraction notation were defined immediately before the display rather than partly in the text afterward.","section":"Section 7, Example 6.5"},{"comment":"In the B2 example, the triangulation is depicted but not explicitly derived from Lemma 7.3; a short comment on how the vertices with denominator 2 arise from D(B) would improve readability.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The analytic part of the paper (Theorem 1.2 and its proof) appears sound and is a genuine contribution; the problem is concentrated in Section 7. The arithmetic theorem (Theorem 1.5) is not proved as written because of the invalid splitting in Lemma 7.10 and the incomplete E8 case. I would encourage the editor to require a corrected proof of Proposition 7.7 or a direct computation of D(Φ) for the exceptional types before publication. The independent work of Onodera supports the vanishing theorem but not the arithmetic description, so the arithmetic claim should not be carried by that corroboration."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version. The main analytic theorem—high-order vanishing of ζΦ at negative even integers—is real, new, and I believe correct. The arithmetic description of the leading coefficient, Theorem 1.5, is not proved as written: the root-system lemma it depends on has a false step, and this is not just a matter of expanding the E8 check.\n\nWhat the paper does well: it introduces an integral representation using Hurwitz zeta functions and proves a self-contained lemma bounding pole orders of simplex integrals. The deduction of Theorem 1.2 from that lemma is clean: at negative even integers the gamma factor supplies r zeros, the integral can only have poles of order r−n, so ζΦ vanishes to order at least n. The odd-integer cases are read off from the Poincaré polynomial degrees. This part is convincing and independently corroborated by Onodera's A2 work.\n\nThe soft spot is Section 7. Lemma 7.10 claims, in equation (7.4), that Z^n/rowspan([[A,0],[v,q]]) is isomorphic to Z/qZ ⊕ Z^{n-1}/AZ^{n-1}. That is false in general. Take A=[2], v=[1], q=2; the matrix has Smith normal form diag(1,4), so the quotient is Z/4Z, exponent 4, while the claimed direct sum has exponent 2. The induction bound E_i(Φ)={LCM(q,e)} is therefore not justified by the given argument. For E8 this could matter concretely: for i=7 one gets q=3 and e∈{1,2,3}, so the true exponent could be 9; for i=8, q=2 and e∈{1,2,3,4}, so the true exponent could be 8. Neither 9 nor 8 is in H(E8). The sentence 'one checks exactly as before' does not resolve this, because the failure is in the general splitting step, not in the arithmetic of a particular Dynkin diagram.\n\nI want to be clear about proportion: Theorem 1.2 does not use Proposition 7.7, so the central vanishing result stands. Proposition 7.7 might well be true; it just needs a corrected proof or a direct verification for the exceptional types. The paper is worth a serious referee, but the referee should require the author to fix Lemma 7.10 or replace it.","headline":"Solid new vanishing theorem and a genuinely useful analytic method; the arithmetic leading-term theorem currently rests on a false induction lemma in the root-system combinatorics.","tokens_in":23875,"tokens_out":4860,"would_cite":true,"duration_ms":44423,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11M32","17B22","11M35"],"pacs":[],"model":"deepseek-v4-flash","headline":"Witten zeta functions vanish to order at least rank at negative even integers","keywords":["Witten zeta function","root systems","Hurwitz zeta function","vanishing at negative integers","highest root","multiple zeta values","Bernoulli numbers","Weyl group"],"falsifier":"For $\\Phi = E_8$, exhaustively enumerate the levels of all invertible $8 \\times 8$ submatrices of the $8 \\times 16$ integral matrix $M_\\Phi$; if any level lies outside $\\{1,2,3,4,5,6\\}$, then Proposition 7.7 is false and the set $T$ in Theorem 1.5 must be enlarged — alternatively, computing the actual order of vanishing of $\\zeta_{E_8}(s)$ at $s=-2$ and finding it below $8$ would refute the rank bound of Theorem 1.2(a).","tokens_in":22921,"feed_emoji":"🧮","tokens_out":12285,"duration_ms":97300,"temperature":0.7,"pith_summary":"This paper proves that the Witten zeta function $\\zeta_\\Phi(s)$ of a simply connected compact Lie group, identified with its root system $\\Phi$, has a zero of order at least the rank of $\\Phi$ at every negative even integer. It also establishes vanishing at negative odd integers for the root systems $A_n$, $D_n$ with odd $n \\geq 5$, and $E_6$, with explicit lower bounds on the order. A second theorem describes the first non-vanishing Taylor coefficient: after multiplication by a power of $\\pi$, the coefficient lies in the $\\mathbb{Q}$-span of products of Hurwitz zeta values whose second arguments have denominators taken from the coefficients of the highest root. The method is a new analytic treatment of an integral representation of the zeta function, and it recovers a previously isolated formula for the $A_2$ case as a special consequence. If correct, the results confirm the standing conjecture that the function vanishes at $-2$ for every infinite compact group.","feed_headline":"Witten zeta functions vanish to order at least rank","feed_subtitle":"Integral method ties the leading coefficient to Hurwitz zeta values with highest-root denominators.","key_machinery":"The engine is an integral representation of the form $P_\\Phi(s)\\,\\xi_\\Phi(s) = ((2\\pi i)^s/\\Gamma(s))^r I_\\Phi(s)$, where $I_\\Phi(s)$ integrates products of Hurwitz zeta functions over the cube $[0,1]^{r-n}$ and $P_\\Phi(s)$ is the Weyl group's Poincaré polynomial. An elementary lemma (Lemma 3.1) controls meromorphic continuation of simplex integrals of powers of linear forms: the pole order is at most the dimension $n$, and all poles lie at rational numbers. Applying this lemma to the triangulated Hurwitz-zeta integral bounds its pole order by $r-n$. On the arithmetic side, a triangulation of the cube whose vertices have denominators controlled by the levels of submatrices of the matrix $M_\\Phi = ((\\lambda_i, \\alpha_j^\\vee))$ reduces the leading coefficient to a finite sum of Hurwitz zeta products; a root-system-theoretic proposition identifies the resulting denominator set with $H(\\Phi) \\cup \\{1\\}$, the highest-root coefficients together with $1$.","core_discovery":"On the paper's own terms, the central discovery is that the meromorphic behaviour of $\\zeta_\\Phi(s)$ at negative integers is controlled by a pole-counting comparison. When $s$ is a negative even integer, the prefactor $((2\\pi i)^s/\\Gamma(s))^r$ in the integral representation contributes a zero of order $r$, the number of positive roots, while the Hurwitz-zeta integral can have poles of order at most $r-n$, where $n$ is the rank; the quotient therefore vanishes to order at least $n$. The same comparison gives lower bounds at negative odd integers from the number of odd degrees of the Weyl group. For the leading coefficient, the paper shows that $\\pi^{mr}\\,\\zeta_\\Phi^{(n)}(-m)$ lies in the $\\mathbb{Q}$-span of products of Hurwitz zeta values with parameters in $(0,1]$ whose denominators are elements of $H(\\Phi)$, the set of coefficients of the highest root; for classical root systems this reduces to a polynomial in $\\pi^2$ and odd Riemann zeta values.","pith_inferences":["A likely extension of the method is to multivariable Witten zeta functions or twisted versions, wherever an integral representation with a gamma prefactor and a Hurwitz-zeta integrand exists; the same pole-counting should dictate vanishing orders.","If Conjecture 1.3 (that the lower bounds on orders are exact) is true, Theorem 1.5 describes the actual first non-vanishing coefficient; if the orders are larger, the arithmetic description presumably applies to whatever coefficient is first non-zero, giving a family of invariants indexed by the discrepancy.","The appearance of the highest-root coefficients as denominators may reflect a general pattern: for other families of representations, the denominator set of the leading coefficient could be governed by analogous lattice quotient exponents, a checkable hypothesis in explicit examples like minuscule weights.","The only gap between the theorem as written and a fully verified statement is the $E_8$ case check in Proposition 7.7; a direct computer enumeration of the submatrix levels would settle it, and it is a short verification."],"forward_implications":["The stated conjecture that $\\zeta_G(-2)=0$ for every infinite compact group holds for all simply connected compact Lie groups, since the rank is at least $1$.","Every negative even integer is a zero of $\\zeta_\\Phi$ of order at least the rank; for $A_n$, $D_n$ with odd $n \\geq 5$, and $E_6$, the specified negative odd integers are also zeros.","The leading coefficient of $\\zeta_\\Phi$ at a negative even integer $-m$, scaled by $\\pi^{mr}$, is a $\\mathbb{Q}$-linear combination of products of Hurwitz zeta values with second arguments from $H(\\Phi)$.","For the classical root systems $A_n$, $B_n$, $C_n$, $D_n$, the same coefficient is a polynomial in $\\pi^2$ and the odd Riemann zeta values $\\zeta(2j+1)$.","The explicit formula for the $A_2$ leading coefficient is recovered as a special case, giving a conceptual proof of a result that previously stood alone."],"supporting_citations":[{"why":"States the conjecture that $\\zeta_G(-2)=0$ for infinite compact $G$, which the paper's Theorem 1.2 settles for simply connected compact Lie groups.","marker":"[24]"},{"why":"Supplies the integral representation of $\\zeta_\\Phi$ in terms of Hurwitz zeta functions that the new proof starts from.","marker":"[20]"},{"why":"Provides the book-length theory of root-system zeta functions and the factorization of the Poincaré polynomial used in the odd-integer argument.","marker":"[23]"},{"why":"Gives the classical evaluation $\\zeta_\\Phi(2m) \\in \\mathbb{Q}\\pi^{2mr}$, the positive-integer counterpart whose method the paper contrasts with its own.","marker":"[33]"},{"why":"Contains the explicit formula for the $A_2$ leading coefficient that the paper re-derives as a special case.","marker":"[27]"},{"why":"Records the earlier weaker statement about primes dividing the lattice quotient that Proposition 7.7 strengthens to the full denominator set.","marker":"[31]"},{"why":"Supplies the table of highest roots and Weyl degrees from which $H(\\Phi)$ and the odd-degree counts are read off.","marker":"[16]"}],"fun_headline_variants":["Witten zeta zeros: rank determines vanishing order","Witten zeta vanishing: highest root appears in leading term","New integral method sharpens Witten zeta zeros","Witten zeta: leading coefficient reveals highest root"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on an unexpanded case check: for the exceptional root system $E_8$, only some of the eight subcases are shown explicitly, the rest being dismissed with 'checked exactly as before,' and a missed denominator outside $H(\\Phi) \\cup \\{1\\}$ would make the leading-coefficient theorem false.","fun_headline_variants_meta":{"raw":{"variants":["Witten zeta zeros: rank determines vanishing order","Witten zeta vanishing: highest root appears in leading term","New integral method sharpens Witten zeta zeros","Witten zeta: leading coefficient reveals highest root"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000443,"raw_usage":{"total_tokens":2179,"prompt_tokens":814,"completion_tokens":1365,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":430,"completion_tokens_details":{"reasoning_tokens":1300}},"tokens_in":430,"tokens_out":1365,"duration_ms":9810,"temperature":1.0,"reasoning_tokens":1300,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T14:31:46.591296+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For $\\Phi = E_8$, exhaustively enumerate the levels of all invertible $8 \\times 8$ submatrices of the $8 \\times 16$ integral matrix $M_\\Phi$; if any level lies outside $\\{1,2,3,4,5,6\\}$, then Proposition 7.7 is false and the set $T$ in Theorem 1.5 must be enlarged — alternatively, computing the actual order of vanishing of $\\zeta_{E_8}(s)$ at $s=-2$ and finding it below $8$ would refute the rank bound of Theorem 1.2(a).","supporting_citations":[{"cited_title":"Zeros of Witten zeta functions and absolute limit.Kodai Mathematical Journal, 36(3):440–454, 2013","cited_arxiv_id":null,"evidence_quote":"States the conjecture that $\\zeta_G(-2)=0$ for infinite compact $G$, which the paper's Theorem 1.2 settles for simply connected compact Lie groups."},{"cited_title":"On Witten multiple zeta-functions associated with semisimple Lie algebras III.Multiple Dirichlet series, L-functions and automorphic forms, pages 223–286, 2012","cited_arxiv_id":null,"evidence_quote":"Supplies the integral representation of $\\zeta_\\Phi$ in terms of Hurwitz zeta functions that the new proof starts from."},{"cited_title":"Springer, 2023","cited_arxiv_id":null,"evidence_quote":"Provides the book-length theory of root-system zeta functions and the factorization of the Poincaré polynomial used in the odd-integer argument."},{"cited_title":"University of Cologne, Department of Mathematics and Computer Science, Weyertal 86-90, 50931 Cologne, Germany Email address:kau1@smail.uni-koeln.de","cited_arxiv_id":null,"evidence_quote":"Gives the classical evaluation $\\zeta_\\Phi(2m) \\in \\mathbb{Q}\\pi^{2mr}$, the positive-integer counterpart whose method the paper contrasts with its own."},{"cited_title":"A functional relation for Tornheim’s double zeta functions.Acta Arithmetica, 162(4):337–354, 2014","cited_arxiv_id":null,"evidence_quote":"Contains the explicit formula for the $A_2$ leading coefficient that the paper re-derives as a special case."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Records the earlier weaker statement about primes dividing the lattice quotient that Proposition 7.7 strengthens to the full denominator set."},{"cited_title":"Springer Science & Business Media, 2012","cited_arxiv_id":null,"evidence_quote":"Supplies the table of highest roots and Weyl degrees from which $H(\\Phi)$ and the odd-degree counts are read off."}],"review_version":1}