{"id":"ad201ff8-aa57-49f0-88b5-4774b1428c7f","arxiv_id":"2510.27311","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Invariant cohomology Poincaré polynomials of quaternionic reflection arrangements coincide with the complex cases except for imprimitive groups with non-cyclic K/H, where P(t^{1/3}) = 1+2t+...+2t^{n-2}+5t^{n-1}+4t^n.","lead":"This paper computes the invariant Poincaré polynomials for the rational cohomology of complements of quaternionic reflection arrangements, extending prior work for complex reflection groups. It finds that only one new polynomial family appears beyond the complex cases, for imprimitive groups with non-cyclic quotient K/H.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Rank-2 primitive groups break the completeness claim: Prop 3.7 needs the hyperplane-orbit count a, but no a-values are computed for these groups, including the new groups flagged from [Wal25]/[Tay25].","rationale":"The stress-test confirms the central claim is internally consistent where it is actually computed: I spot-checked Lemma 5.3's orbit counts (for K/H = C_m even, the diagonal action on ratio cosets x ↦ x + 2c together with inversion yields exactly 3 hyperplane orbits; for C_2×C_2, exactly 5), and Theorem 5.4/Cor 5.5 are consistent with (3.6). The reader's weakest assumption (Prop 2.8) is real but well-defended: the sign computation is essentially forced by H-linearity of the localization maps (positive real determinant), with only the scope of [LS01, Cor. 5.6] as residual risk. The least secure condition is instead the completeness of the rank-2 primitive case: §6 asserts Prop 3.7 covers it, but Prop 3.7 is only a formula in the unknown parameter a, and §2.1 states the rank-2 classification is actively being revised ([Wal25], [Tay25]). The reader's rationale mentions 'rank-2 orbit counts that are not displayed or proved' as a secondary gap; I promote it to the primary objection because it bears directly on the headline classification and is not answerable from the text. Since the concern is a missing case analysis rather than an identified error, the reader's CONDITIONAL verdict stands (UNCHANGED); the fix is a short computational table of a-values for rank-2 primitive groups.","tokens_in":25312,"tokens_out":58131,"duration_ms":472972,"concrete_test":"For every primitive irreducible quaternionic reflection group of rank 2 — the six exceptional groups of [Coh80] not listed in Table 3, the infinite rank-2 primitive families, and the new rank-2 groups constructed in [Wal25]/[Tay25] — compute the hyperplane-orbit count a = |A/G| (equivalently |L(A)_1/G|) with the OSCAR method of §6; for this count only the permutation action on the reflecting hyperplanes is needed, not the full character computation. Then check a ∈ {1,2,3,5}: values 1–3 give the complex types of Theorem 3.10; value 5 gives the Cor 5.5(2) family; any a ∉ {1,2,3,5} gives a polynomial 1 + a t^3 + (a−1)t^6 realized in the quaternionic setting but not in the complex case and not in the claimed new family, disproving the abstract. A positive result closes the gap by adding a table of rank-2 primitive orbit counts analogous to Lemma 5.3.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is a full classification: 'only one additional family of new types of Poincaré polynomials occurs in the quaternionic setting' (Abstract; Intro; Cor. 5.5(2)). Coverage in the paper: complex-reducible groups (Thm 3.10), imprimitive groups for all n ≥ 2 (Cor 5.5, base case Lemma 5.3), and primitive groups with n > 2 (Table 3). The remaining case — primitive groups with n = 2 — is dismissed in §6: 'almost all of these groups act on a vector space of quaternionic dimension n = 2, so are covered by Proposition 3.7.' This dismissal is not substantiated. Prop 3.7 only states P(A,G;t) = 1 + a t^3 + (a−1)t^6 where a is the number of G-orbits on the hyperplane set A; it provides no value for a. The paper never computes or bounds a for the six exceptional rank-2 primitive groups (13 exceptional groups in [Coh80, §2.1], minus the seven in Table 3), for the infinite rank-2 primitive families, or for the additional rank-2 groups that §2.1 itself flags: 'Waldron [Wal25] and Taylor [Tay25] independently revise part of this classification and notably construct further groups in rank 2.' At rank 2 the complex types correspond to a ∈ {1,2,3} (Thm 3.10) and the claimed single new family has a = 5 (Cor 5.5(2), n=2); hence any rank-2 primitive group with a = 4 (or a ≥ 6) would yield a polynomial outside both lists and falsify the headline. This is a concrete missing case in a classification statement, not a style issue. By contrast, the reader's candidate (Prop 2.8) is better supported than its prominence suggests: the map π_i is H-linear, so as an isomorphism of right-H-vector spaces its real determinant is automatically positive, and the explicit determinant d_i = N(α_i)^2 > 0 confirms ε = +1; the residual risk there is only the exact hypotheses of [LS01, Cor. 5.6], which the independent proof in the paper addresses.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the rational cohomology of complements of quaternionic reflection arrangements, with focus on the Poincaré polynomial P(A,G;t) of the G-invariants in H*(M(A);Q). It extends Douglas–Pfeiffer–Röhle's work from complex reflection groups to quaternionic reflection groups. The authors prove an Orlik–Solomon presentation for complements of quaternionic hyperplane arrangements (Proposition 2.8), establish analogues of Brieskorn's lemma and the induction formula (3.6), treat the complex-reducible case by reduction to [DPR25], and analyze the imprimitive groups Gn(K,H) in detail using the Dowling lattice description, producing an explicit list of orbit representatives and the resulting Poincaré polynomials in Theorem 5.4 and Corollary 5.5. The primitive groups of quaternionic dimension >2 are handled by OSCAR computations summarized in Table 3, and explicit bases are discussed in Section 7. The headline claim is that, across all quaternionic reflection groups, only one new family of Poincaré polynomials occurs that is not realised in the complex case, namely the imprimitive family in Corollary 5.5(2).","tokens_in":25746,"tokens_out":4582,"duration_ms":46890,"significance":"If the classification claim is correct, this is a valuable structural result: it shows that the invariant Poincaré polynomials for quaternionic reflection arrangements are almost all controlled by the complex case, with a single explicit new imprimitive family. The paper contains several strong components: an independent proof of the Orlik–Solomon presentation for quaternionic arrangements; a new Dowling-lattice description of intersection lattices of imprimitive groups; explicit orbit representatives; and explicit inductive dimension computations. The imprimitive part of the argument is coherent and appears internally sound. However, the global completeness claim is not supported by the evidence presented for primitive groups of rank 2, which are explicitly excluded from the detailed analysis. The higher-dimensional primitive entries also rely entirely on OSCAR computations without shipped code or certificates, limiting verifiability.","major_comments":[{"comment":"The Abstract and Introduction state that 'only one additional family of new types of Poincaré polynomials occurs in the quaternionic setting.' This is a global classification statement over all quaternionic reflection groups. Yet §2.1 restricts attention to dim V > 2, saying the rank-2 case is 'trivially handled by Proposition 3.7', and §6 repeats that 'almost all of these groups act on a vector space of quaternionic dimension n = 2, so are covered by Proposition 3.7.' Proposition 3.7 only gives P(A,G;t) = 1 + a t^3 + (a−1)t^6, where a is the number of G-orbits on the hyperplane set A; it does not compute or bound a. In the complex case, Theorem 3.10 gives a ∈ {1,2,3}; the new imprimitive family in Corollary 5.5(2) has a = 5 at n = 2. Thus any primitive rank-2 group with a = 4 or a ≥ 6 would produce a polynomial outside both lists and contradict the headline. Since §2.1 itself cites [Wal","section":"§2.1, §6, Abstract"},{"comment":"The seven primitive groups of dimension n > 2 are the only remaining non-complex primitive cases, and Table 3 lists their Poincaré polynomials as OSCAR computations. The text describes the algorithm (NBC basis, character computation) but provides no code, no input matrices, and no certificates. Since these seven entries are load-bearing for the primitive part of the classification, a referee cannot verify them from the manuscript alone. Please supply a documented script or data file (or, failing that, full generator matrices and intermediate data such as NBC basis sizes and character tables) so that the entries in Table 3 and Table 4 can be independently checked.","section":"§6, Table 3 and Table 4"}],"minor_comments":[{"comment":"Typo: 'Brieskorn's Lemma for quaternionic arrngements' should be 'arrangements'.","section":"§2.3"},{"comment":"The table is difficult to read: the rows labelled 'else' are repeated three times, and the first three data columns are all '1' for every case. Please restructure the table so that the dependence on [K:H], n, and K/H is immediately transparent, for example by splitting into three separate cases as in Corollary 5.5.","section":"§5, Table 2"},{"comment":"The notation 'P(A(G),G;t)(t^{1/3})' near Table 3 is confusing; the paper elsewhere uses P(A(G),G;t^{1/3}) consistently. Please use a single notational convention.","section":"§6"},{"comment":"The table lists hyperplanes for bases but does not explain how the particular parabolic subgroups in the second column were selected from [BST23, §7.2]. A sentence identifying the parabolic type in terms of the group's classification would help reproducibility.","section":"§7.2, Table 4"}],"recommendation":"major_revision","confidential_remarks":"The central weakness is the rank-2 primitive gap in an otherwise solid paper. The imprimitive and complex-reducible parts appear sound, and the OS presentation proof is a useful addition. I would not reject on the present evidence because the gap may be fixable by either computing the missing orbit counts or narrowing the stated scope. The abstract's global wording, however, is currently unsupported. The OSCAR computations should also be made available if the primitive table is to be used as evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take: the paper's real contribution is the imprimitive computation and the Dowling-lattice description of parabolic subgroups for quaternionic imprimitive groups. Those parts look careful and correct. But the headline assertion — 'only one additional family of new types of Poincaré polynomials occurs' — is not supported, because the rank-2 primitive groups are dismissed without computing the orbit count that Proposition 3.7 leaves open.\n\nWhat's new: the inductive machinery from DPR25 is transplanted to quaternionic arrangements using the Orlik–Solomon presentation for the cohomology (Prop 2.8). The part I checked most closely, the sign computation in Prop 2.8, seems fine: the relevant maps are H-linear, so the real determinant is positive, and the explicit determinant di = N(αi)^2 confirms ε=+1. The imprimitive analysis for K non-cyclic is the meat of the paper. The orbit representatives, the conjugacy classification, and the inductive table in Theorem 5.4 are coherent, and the new family in Cor 5.5(2) follows genuinely from the orbit-count for G2(D2d, C2d). The Dowling lattice isomorphism for parabolic subgroups (Proposition 4.6) is a genuinely new structural result, and the basis construction in Section 7 is a nice extension of DPR25's.\n\nSoft spots: first, the rank-2 primitive gap. Proposition 3.7 gives P = 1 + a t^3 + (a−1)t^6 in terms of a, the number of G-orbits on hyperplanes. The paper says these groups are 'covered' by that proposition and never computes a. But the abstract's completeness claim requires knowing that only a ∈ {1,2,3} for complex groups and a=5 for the new imprimitive family appear. There are 13 exceptional primitive rank-2 groups plus infinite families, and §2.1 itself flags that Waldron and Taylor add further rank-2 groups. No a-value is computed or bounded for any of them. If one of those groups has a=4 or a≥6, the abstract's claim is false. That is not a style concern; it's an unhandled case in the classification.\n\nSecond, Tables 3 and 4 come from OSCAR computations, but no code or certificates are shipped. For a completeness claim, that's a reproducibility issue, though likely addressable.\n\nVerdict: the imprimitive part is worth keeping; the global classification needs more work. This paper is for people working on quaternionic reflection arrangements or invariant cohomology — they'll find the imprimitive computation valuable. Send it to a serious referee: the mathematical content deserves scrutiny, but the referee should ask for the rank-2 orbit counts or a revised claim. I would not desk-reject.","headline":"Solid imprimitive computation saddled to an overbroad completeness claim; the rank-2 primitive cases are not actually handled.","tokens_in":26296,"tokens_out":4769,"would_cite":false,"duration_ms":39554,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["52C35","20F55","14N20","05E18"],"pacs":[],"model":"deepseek-v4-flash","headline":"For quaternionic reflection groups, the invariant cohomology Poincaré polynomials are the complex ones plus exactly one new imprimitive family, and explicit bases for the invariants are given.","keywords":["quaternionic reflection groups","reflection arrangements","cohomology of arrangement complements","Poincaré polynomials","invariants","Dowling lattices","Orlik-Solomon algebra","imprimitive reflection groups"],"falsifier":"Take the rank-2 imprimitive group G_2(D_4,C_4), where K/H is a Klein four-group; the paper predicts dim H^3(M(A);Q)^G = 5 and dim H^6(M(A);Q)^G = 4, giving Poincaré polynomial 1+5t^3+4t^6. A direct computation of these invariant dimensions from the generators and relations, without invoking the inductive identity, would either confirm the sign convention in Proposition 2.8 or expose a shift.","tokens_in":25204,"feed_emoji":"📐","tokens_out":9388,"duration_ms":90878,"temperature":0.7,"pith_summary":"The paper asks what the group G-invariants look like in the rational cohomology of the complement of the reflection arrangement of a quaternionic reflection group. It establishes that, apart from one infinite family, every invariant Poincaré polynomial that occurs in the quaternionic setting already occurs for complex reflection groups. The one genuinely new family comes from imprimitive groups G_n(K,H) for which K is non-cyclic, [K:H] and n are even, and K/H is not cyclic; there the Poincaré polynomial in t^{1/3} is 1+2t+...+2t^{n-2}+5t^{n-1}+4t^n. The proof runs by induction using an Euler-characteristic identity and a Dowling-lattice description of the intersection lattice, and it supplies explicit bases of the invariant ring. If correct, this completes the classification of these invariant Poincaré polynomials for all irreducible quaternionic reflection groups of rank at least two.","feed_headline":"Quaternionic groups yield one new invariant polynomial family","feed_subtitle":"The full list of invariant Poincaré polynomials now matches the complex one barring a single imprimitive family.","key_machinery":"The engine is the Euler-characteristic identity (3.6): the alternating sum, over orbit representatives X of the intersection lattice, of dim H^{3 rk(X)}(M(A_X);Q)^{N_G(X)} vanishes. It is obtained from a G-equivariant degree-3 differential ∂ and multiplication operator μ satisfying ∂μ+μ∂=id, and it reduces the Poincaré polynomial of H^*(M(A))^G to top-degree invariant dimensions in proper local arrangements, which are known by induction. The companion combinatorial tool is the isomorphism L(A(G_n(K,H))) ≅ D_n(K), where D_n(K) is the Dowling lattice of partial K-partitions of {1,...,n}; this gives the orbit representatives of parabolic subgroups that feed the induction. For the seven primitiv","core_discovery":"The paper proves that the Poincaré polynomial P(A(G),G;t) of the graded ring H^*(M(A(G));Q)^G is determined, for every irreducible quaternionic reflection group G, by a short list of four complex patterns together with one additional imprimitive pattern. Complex-reducible groups simply give P(A_C,G;t^3), reproducing the four complex types. Among genuinely quaternionic groups, imprimitive groups G_n(K,H) fall into three cases according to parity of [K:H] and n and cyclicity of K/H; the non-cyclic quotient case produces the new polynomial 1+2t+...+2t^{n-2}+5t^{n-1}+4t^n after substituting t^{1/3}. The seven primitive groups of quaternionic rank greater than two have polynomials 1+t, 1+t+t^3+t^","pith_inferences":["Inference: because H^*(M(A);Q)^G is isomorphic to H^*(M(A)/G;Q) by transfer, the new polynomial 1+5t^3+4t^6 in rank 2 predicts the rational Betti numbers of the orbit space of H^2 minus the arrangement by G_2(D_4,C_4); this quotient is a concrete space whose cohomology could be checked independently.","Inference: the near-identity with the complex list suggests a structural explanation beyond the computation, perhaps a uniform 'parabolic induction' formula governed by the lattice of parabolic subgroups; the quaternionic imprimitive case would then be the only place where the orbit classification differs.","Inference: the Dowling-lattice identification may transfer matroid-theoretic tools to this setting, since the Möbius numbers |μ(X)| control local top-degree dimensions; this could recast the computation purely as a statement about Dowling lattices and potentially extend to other classes of symplectic reflection groups.","Inference: the explicit bases given in the paper make it testable whether the invariant ring is generated by a small explicit set of degree-3 and top-degree invariants; deriving such a presentation would describe the full ring structure, not just its Poincaré polynomial."],"forward_implications":["For any complex-reducible quaternionic reflection group, the invariant Poincaré polynomial is exactly the complex one with t replaced by t^3, so the four polynomial types known from the complex classification reappear unchanged.","For imprimitive groups with non-cyclic K, the invariant Poincaré polynomial in t^{1/3} is 1+2t+...+2t^{n-1}+t^n when [K:H] or n is odd; it becomes 1+2t+...+2t^{n-2}+3t^{n-1}+2t^n when both are even and K/H is cyclic; and the new family 1+2t+...+2t^{n-2}+5t^{n-1}+4t^n when K/H is non-cyclic.","The seven primitive non-complex groups in dimension greater than two have invariant Poincaré polynomials 1+t, 1+t+t^3+t^4, or 1+t+t^4+t^5 in t^{1/3}, so no further new polynomials arise outside the imprimitive family.","Explicit bases exist: in the imprimitive case the invariant top-degree classes are averages ϵ_G(h_1 h_2^ξ h_3 ... h_k) with ξ running over representatives of K/H in the exceptional cases, and the primitive case has explicit hyperplane products listed in the paper.","The inductive identity reduces the computation to orbit representatives of parabolic subgroups, so the same method applies uniformly to all imprimitive quaternionic reflection arrangements once the orbit classification is known."],"fun_headline_variants":["Quaternionic invariants: one new Poincaré family","One extra invariant family in quaternionic cohomology","Quaternionic reflection groups: only one new invariant pattern","Poincaré polynomials: quaternionic case adds a single family","New invariant polynomial family from quaternionic groups"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the cohomology ring of the complement of any quaternionic arrangement is presented by hyperplane-indexed degree-3 generators with relations from minimal dependent sets all carrying '+' signs, and that the group acts by permuting those generators; the paper's determinant-positivity argument is what guarantees the signs, and every later dimension count uses this presentation.","fun_headline_variants_meta":{"raw":{"variants":["Quaternionic invariants: one new Poincaré family","One extra invariant family in quaternionic cohomology","Quaternionic reflection groups: only one new invariant pattern","Poincaré polynomials: quaternionic case adds a single family","New invariant polynomial family from quaternionic groups"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000257,"raw_usage":{"total_tokens":1465,"prompt_tokens":846,"completion_tokens":619,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":590,"completion_tokens_details":{"reasoning_tokens":537}},"tokens_in":590,"tokens_out":619,"duration_ms":5561,"temperature":1.0,"reasoning_tokens":537,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T06:58:45.659880+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the rank-2 imprimitive group G_2(D_4,C_4), where K/H is a Klein four-group; the paper predicts dim H^3(M(A);Q)^G = 5 and dim H^6(M(A);Q)^G = 4, giving Poincaré polynomial 1+5t^3+4t^6. A direct computation of these invariant dimensions from the generators and relations, without invoking the inductive identity, would either confirm the sign convention in Proposition 2.8 or expose a shift.","supporting_citations":[],"review_version":1}