{"id":"697208ab-3c08-462e-a98d-e4aa05f0c473","arxiv_id":"1908.00567","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For any acyclic quiver partitioned into Dynkin subquivers, the cohomological Hall algebra is a tensor product of the subquiver Hall algebras, and Dynkin quiver polynomials multiply to classes of quiver strata.","lead":"This paper gives a new, characteristic-class proof that the cohomological Hall algebra of an acyclic quiver decomposes into tensor products of simpler Hall algebras indexed by a partition of the quiver into Dynkin subquivers. The result itself was known through earlier stability-condition work, but this proof makes the decomposition explicit and shows products of Dynkin quiver polynomials compute classes of new geometric strata.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Poincaré-series comparison in §7.3 rests on Proposition 5.1, a quantum dilogarithm identity quoted from the unpublished [All18] and neither proved nor derived in this paper; if that identity has an error, the proof of Theorem 6.3 would not establish surjectivity.","rationale":"The reader's stated weakest assumption is the rational-singularity conjecture for type E orbit closures, and the reader's rationale also mentions the unpublished quantum dilogarithm identities. I agree that the rational-singularity caveat is a genuine condition on the proof, but I do not think it is the single most load-bearing assumption: the cycle-class identity in Proposition 4.10 is typically a consequence of having a resolution of singularities, so a direct localization argument could plausibly supply the needed identification and degree count without resolving the open rational-singularity conjecture. By contrast, Proposition 5.1 is the only bridge between the Poincaré series of the tensor product and that of H; without it, the equality of (30) and (31) has no support in the paper. The cited source [All18] is an unpublished preprint, and the statement is sufficiently intricate in its signs and q-shifts that a hidden error is plausible. Thus the central claim is conditional on an unverified external identity. Since the reader already reached CONDITIONAL and explicitly flagged the All18 dependence in the rationale, my read does not change the verdict; it shifts the emphasis from the rational-singularity caveat to the missing quantum dilogarithm proof. A targeted computational verification of Proposition 5.1 in a small non-Dynkin example would either remove this obstruction or expose a concrete failure; such a check is feasible with the explicit quantum algebra relations given in Section 5.","tokens_in":26128,"tokens_out":9633,"duration_ms":96822,"concrete_test":"Use a computer algebra system to expand both sides of Proposition 5.1 in the completed quantum algebra for the smallest non-Dynkin admissible case: Q = 1←2←3 with Q• = {A1 at vertex 1, A2 on 2←3}, Reineke order β1 = e1, β2 = e3, β3 = e2+e3, β4 = e2. Multiply truncated series E(y_{e1})E(y_{e2})E(y_{e3}) and E(y_{β1})E(y_{β2})E(y_{β3})E(y_{β4}) through total degree at least 6, using the relation y_{γ1+γ2} = -q^{-⟨γ1,γ2⟩/2} y_{γ1} y_{γ2}, and compare coefficients. Also test a few Q•-partitions m against Proposition 5.2's formula for w_m by comparing with codim(η_m; Repγ) plus the explicit quadratic terms. If equality holds in this and one larger example, the omitted identity is supported; a mismatch of signs or q-powers would invalidate the Poincaré-series comparison in §7.3.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The main theorem is proved by injectivity (Lemma 7.6) plus a Poincaré-series comparison (§7.3). The comparison equates (30) with (31) by invoking Proposition 5.1, the identity E(y_{e_1})···E(y_{e_n}) = E(y_{β_1})···E(y_{β_r}) in the completed quantum algebra Â_Q, quoted from the unpublished preprint [All18]. Neither Proposition 5.1 nor the companion Proposition 5.2 is proved or derived in this paper, and the cited preprint is not available for inspection. The coefficient comparison at (32) then uses Proposition 5.2's explicit formulas for the signs s_m and q-exponents w_m; an error in either the identity or the q-shift/sign would break the dimension count and therefore surjectivity of (22). This is the most load-bearing step: without it, the paper establishes only that the natural map in (22) is injective, not that it is an isomorphism. The rational-singularity caveat in Remark 4.9 is a second condition, but the cycle-class identity in Proposition 4.10 may be replaceable by a direct localization computation, as Remark 7.8 suggests for injectivity; the missing quantum dilogarithm identity is not obviously replaceable within the proof as written. The theorem may well be true, and the paper states that it is implicit in the work of Franzen–Reineke and Davison–Meinhardt, but the proof supplied here is conditional on an unavailable external result.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a tensor decomposition theorem for the cohomological Hall algebra (CoHA) of an acyclic quiver Q with an admissible ordered partition into Dynkin subquivers, none of which is an orientation of E8. The main theorem (Theorem 6.3) states that the CoHA is spanned by certain subalgebras A_β, each isomorphic to the CoHA of A1, indexed by the positive roots of the subquivers, and that the multifactor ∗-multiplication in Reineke order gives an isomorphism from their tensor product to H(Q). As corollaries, it recovers a factorization H(Q) ≅ H(Q1)⊗⋯⊗H(Qℓ) and expresses products of Dynkin quiver polynomials as classes of quiver strata. The proof combines equivariant localization, a direct injectivity argument via restriction to a normal locus, and a Poincaré-series comparison using quantum dilogarithm identities.","tokens_in":26473,"tokens_out":8517,"duration_ms":84554,"significance":"If the proof is completed, the paper gives an explicit topological realization, in the language of characteristic classes, of CoHA decompositions previously obtained through stability conditions by Davison–Meinhardt and Franzen–Reineke. The injectivity argument (Propositions 7.3 and 7.4, Lemma 7.6) is an elegant extension of Rimányi's localization method, and the worked examples are pedagogically useful. The paper also provides a clear link between CoHA products and quiver polynomials. However, the proof as written is conditional: it depends on quantum dilogarithm identities quoted from the author's unpublished preprint [All18], and on the open rational-singularity assumption for type E Dynkin orbit closures. These dependencies are load-bearing; the manuscript is therefore not yet a complete proof of the theorem as stated.","major_comments":[{"comment":"The Poincaré-series comparison that completes the proof of Theorem 6.3 depends entirely on Propositions 5.1 and 5.2, both quoted from the unpublished preprint [All18]. Proposition 5.1 is the identity E(y_{e1})⋯E(y_{en}) = E(y_{β1})⋯E(y_{βr}) in the completed quantum algebra Â_Q, and Proposition 5.2 supplies the signs s_m and q-exponents w_m used in the coefficient comparison at (32). Neither identity is proved or derived in this paper, and [All18] is cited as a 2018 preprint without an arXiv identifier or other public availability. An error in either identity would break the equality of the Poincaré series and hence the surjectivity of (22). This is load-bearing; please include complete proofs of both propositions in the manuscript, or otherwise make the full content of [All18] available for verification.","section":"Sections 5 and 7.3, Propositions 5.1 and 5.2"},{"comment":"Theorem 6.3 as stated claims an unconditional isomorphism for every admissible, ordered, Dynkin subquiver partition with no E8 factor. However, the proof of Proposition 4.10 uses the assumption, acknowledged in Remark 4.9, that type E Dynkin quiver orbit closures have rational singularities. This assumption enters the proof of Proposition 7.4 through the identification of the rational function g in (28) with the Euler class ε_m, and Lemma 7.9 uses the degree of ε_m, namely codim(η_m; Rep_γ), for the degree shift in equation (32). Without the rational-singularity assumption, that degree shift is not established, so the Poincaré-series comparison, and hence surjectivity of (22), rests on an open problem. The abstract and Theorem 6.3 do not list this as a hypothesis. Please either prove the degree shift by a direct localization computation independent of rational singularities, or restate the main theorem and its corollaries as explicitly conditional on this assumption.","section":"Remark 4.9, Proposition 4.10, and Lemma 7.9"},{"comment":"The assertion that each subalgebra A_β is isomorphic to H(A1) is central: it justifies the identification (A_{β_u,m_u})_{k_u} ≅ H^{2k_u}(BGL(C_{m_u})) used in formula (31) to write the Poincaré series of the domain of (22). The proof given for general positive roots is only a sketch: it states that the dependence on variables at other vertices 'cancels thanks to supersymmetry' without presenting the cancellation. This is a nontrivial property of the multifactor multiplication (11). Please provide a complete proof, or reduce the statement to the known Dynkin-quiver case of Rimányi ([Rim13, Theorem 11.3]) by observing that all factors and all products within A_β are supported on a single subquiver Q_j, so the multiplication in question is the CoHA product of that Dynkin subquiver.","section":"Section 6, Proposition 6.1"},{"comment":"The proof asserts that 'there is only one torus fixed point in Σ ... over ν_m' and that all other localization terms vanish after applying ι*_m. This is a crucial geometric claim because it makes formula (28) exact, and it feeds into Lemma 7.9. The claim is not proved; it is only illustrated in Example 7.5. Please supply a proof, for instance by describing the torus fixed points of the quiver flag variety and checking the incidence conditions, or provide a reference that covers this statement in the present acyclic generality.","section":"Section 7.2, proof of Proposition 7.4"}],"minor_comments":[{"comment":"The phrase 'representations of a cyclic quivers' appears to be a typo; it should presumably read 'acyclic quivers'.","section":"Introduction, paragraph after Corollary 6.7"},{"comment":"The theorem's hypotheses omit the rational-singularity assumption from Remark 4.9. Since the proof is conditional on that assumption, the theorem should state it explicitly in its hypothesis list, not only in a remark.","section":"Theorem 6.3 statement"},{"comment":"The notation codim_C(η_m; Rep_γ(Q)) is used without prior definition; please define it when the stratum η_m is introduced in Definition 4.2.","section":"Section 5, equation (18) and Proposition 5.2"},{"comment":"The proof says Proposition 7.3 implies injectivity of (22) because it holds for all Q•-partitions; a sentence explaining that a map between graded vector spaces is injective if every graded summand is injective would make the argument clearer.","section":"Lemma 7.6"}],"recommendation":"major_revision","confidential_remarks":"The paper's main contribution is a clean topological interpretation of known CoHA decompositions; the novelty is modest but real. The central risk is verifiability rather than correctness: the proof relies on the author's unpublished [All18] for two essential identities and on the open rational-singularity problem for type E. I would ask the editor to require that the revision either include proofs of Propositions 5.1 and 5.2 (or make [All18] publicly available) and state the rational-singularity assumption as an explicit hypothesis of the main theorem, or the paper should be revised to remove the dependence on the open problem. The writing is generally clear and the examples are helpful; the main fix is about completeness and transparency."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the decomposition theorem itself is not new, and the paper says so plainly. What is new is the proof route—equivariant localization, quiver strata, and the class of a normal locus—and that topological realization is a real contribution. The concrete corollary (6.7) expressing products of Dynkin quiver polynomials as fundamental classes of quiver strata is a nice payoff.\n\nWhat the paper does well: it is honest about the relation to Davison–Meinhardt and Franzen–Reineke. The geometric setup is careful: the consistency subset, the Euler class identification in Proposition 4.7, and the one-fixed-point-after-restriction computation in Proposition 7.4 are all well explained. The worked example in 7.5 genuinely helps. The author flags the type E rational-singularity issue in Remark 4.9 rather than hiding it.\n\nThe soft spots are real but specific. The Poincaré-series comparison in §7.3 depends on Proposition 5.1, a quantum dilogarithm identity quoted from the unpublished preprint [All18]. Neither that identity nor the formulas in Proposition 5.2 are proved or derived in the paper, and the preprint is not available for inspection. That is load-bearing: without it, surjectivity of the map in (22) is not established, and the paper only proves injectivity. The rational-singularity assumption for type E is a second caveat; it is openly stated, but it is needed for Proposition 4.10 and the degree shift in Lemma 7.9. Both of these concerns are addressable, not fatal. I also found Proposition 6.1 sketched—the cancellation claim is plausible but deserves a detailed proof from a referee.\n\nI would not cite this paper for the decomposition theorem itself, since the theorem is, by the author's own account, already implicit elsewhere. I might cite it for the topological method if I were doing similar localization arguments. It is a reasonable reading-group paper for people working on CoHAs.\n\nBottom line: send it to a serious referee. The right referee should ask for the [All18] material or for the dilogarithm identity to be included, and should push for a cleaner treatment of Proposition 6.1. The thinking is clear and the contribution, while not groundbreaking, is solid enough to warrant peer review.","headline":"A topological proof of CoHA decompositions that the author openly says are implicit in earlier work; the proof method is genuinely different, but it leans on an unpublished quantum dilogarithm identity and an open rational-singularity assumption, so it is a conditional contribution that still deserves a referee.","tokens_in":26989,"tokens_out":2131,"would_cite":false,"duration_ms":24546,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["16G20","55R40","57R20"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the cohomological Hall algebra of an acyclic quiver is a tensor product of exterior algebras indexed by the positive roots of any admissible collection of Dynkin subquivers.","keywords":["Cohomological Hall Algebra","Dynkin quiver","acyclic quiver","quiver polynomial","tensor product decomposition","quiver flag variety","quantum dilogarithm identity","equivariant cohomology"],"falsifier":"Take an acyclic quiver obtained by appending a single vertex to an E6 (or E7) Dynkin quiver along one arrow, choose the admissible partition consisting of that Dynkin subquiver and the new $A_1$ vertex, and compute the twisted Poincaré series of both sides of (22) using the formulas (30) and (31); any $q$-degree in which they differ would refute the claimed isomorphism.","tokens_in":25933,"feed_emoji":"🧩","tokens_out":17647,"duration_ms":156985,"temperature":0.7,"pith_summary":"This paper proves a structural theorem for the cohomological Hall algebra (CoHA) of any acyclic quiver: given an admissible partition of the quiver into Dynkin subquivers—connected subquivers of types A, D, or E whose contraction leaves the quiver acyclic—the whole algebra is a tensor product of very small subalgebras, each isomorphic to the CoHA of the one-vertex, no-arrow quiver $A_1$, which is an exterior algebra on countably many generators. The factors are indexed by the positive roots of the subquivers, taken in a Reineke order (a total order compatible with the quiver's Euler form), and the isomorphism is given by the CoHA's own multiplication. A second formulation regroups the factors to show that the CoHA of the whole acyclic quiver is isomorphic to a tensor product of the CoHAs of the chosen Dynkin subquivers. The paper also shows that products of Dynkin quiver polynomial classes—the equivariant fundamental classes of representation-space orbits—compute the fundamental classes of certain 'quiver strata' in the ambient CoHA. In the Dynkin case the theorem interpolates between the two known decompositions, and for general acyclic quivers it gives an explicit topological realization of decompositions previously obtained by stability-condition methods.","feed_headline":"Quiver Hall algebras split into one-vertex building blocks","feed_subtitle":"Each block is an exterior algebra; Dynkin quiver polynomials supply the structure constants.","key_machinery":"The machinery is a collection of quiver flag varieties—products, over the quiver's vertices, of ordinary flag varieties whose step sizes are the entries of the dimension vectors—and their incidence subvarieties. For dimension vectors $\\gamma_1,\\dots,\\gamma_r$ consistent with the ordered partition $Q_\\bullet$, the consistency subset $\\Sigma_{\\gamma_1,\\dots,\\gamma_r}(Q_\\bullet)$ inside the product of the flag variety and the representation space consists of flags preserved by all arrows lying inside the subquivers $Q_j$. Its equivariant fundamental class equals the Euler class of a tautological bundle $G(Q_\\bullet)$ (Proposition 4.7), and Proposition 4.4 rewrites a multifactor CoHA product as the equivariant pushforward of that Euler class; this is the formula that lets the paper remove all arrows between different subquivers. A desingularization theorem for quiver orbit closures, applied factorwise, shows that the same incidence variety resolves the closure of the quiver stratum $\\eta_m$, so $[\\eta_m]$ is itself a CoHA product of units $1_{\\beta_u}$. Injectivity comes from restricting the product to a normal locus (the affine subspace where each subquiver component is held at a chosen point): the composed map lands in an integral domain and becomes $f_1\\cdots f_r\\cdot\\varepsilon_m$ with $\\varepsilon_m$ a nonzero Euler class (Proposition 7.4). Surjectivity comes from comparing twisted Poincaré series, where a quantum dilogarithm identity $E(y_{e_1})\\cdots E(y_{e_n})=E(y_{\\beta_1})\\cdots E(y_{\\beta_r})$ from the companion paper equates the two sides factor by factor.","core_discovery":"The central claim is Theorem 6.3. For an acyclic quiver $Q$ and an admissible, ordered, Dynkin subquiver partition $Q_\\bullet=\\{Q_1,\\dots,Q_\\ell\\}$ in which no $Q_j$ is an orientation of $E_8$, write the positive roots $\\Phi^+(Q_\\bullet)$ of the union in a Reineke order (a total order compatible with the Euler form) as $\\beta_1,\\dots,\\beta_r$. Then the left-to-right $\\ast$-multiplication induces an isomorphism $\\mathcal{A}_{\\beta_1}\\otimes\\cdots\\otimes\\mathcal{A}_{\\beta_r}\\xrightarrow{\\sim}H(Q)$, where each $\\mathcal{A}_{\\beta_u}$ is the subalgebra generated by polynomials in one equivariant Chern-root variable at a chosen vertex of $\\beta_u$ and is isomorphic to $H(A_1)$. Equivalently, the $\\ast$-multiplication induces an isomorphism $H(Q_1)\\otimes\\cdots\\otimes H(Q_\\ell)\\xrightarrow{\\sim}H(Q)$. A corollary identifies, for every $Q_\\bullet$-partition $m$, the product of the Dynkin orbit classes $[\\Omega_{m_j}(Q_j)]$ with the equivariant fundamental class $[\\eta_m]$ of the quiver stratum $\\eta_m$—the locus where each subquiver component lies in a fixed Dynkin orbit—so the Dynkin quiver polynomials serve as structure constants of the entire acyclic CoHA.","pith_inferences":["An unresolved regularity fact stands between the proof and full generality for E6/E7: if rational singularities for those orbit closures are proved, the paper's argument closes as written; if a counterexample appears, the isomorphism might still hold but would need a new identification of the product of units with the stratum class.","The paper notes that the Dynkin hypothesis on the subquivers may be stronger than needed for the tensor decomposition; a plausible extension is that any admissible partition into acyclic subquivers gives a topological decomposition, though without the quiver-polynomial reading that Corollary 6.7 exploits.","The same incidence-variety formula (Proposition 4.4) suggests an algorithmic route: compute CoHA structure constants for small acyclic quivers by explicit flag-variety integrals, which could test conjectures about non-Dynkin CoHAs beyond the cases treated here.","Since each admissible partition corresponds to a stability condition, a natural converse is to classify the stability conditions for which every semistable factor is isomorphic to $H(A_1)$ and ask whether all of them arise topologically from admissible Dynkin subquiver partitions."],"forward_implications":["Any admissible Dynkin subquiver partition of an acyclic quiver yields an explicit tensor decomposition of its CoHA into $A_1$-type factors, so the same algebra carries many different decompositions parameterized by such partitions.","Grouping consecutive factors by subquiver gives $H(Q)\\cong H(Q_1)\\otimes\\cdots\\otimes H(Q_\\ell)$; the CoHA of an acyclic quiver is therefore built, as an algebra, from the CoHAs of the Dynkin subquivers in the partition.","The fundamental class of every quiver stratum $\\eta_m$ is a left-to-right product of Dynkin quiver polynomial classes, meaning the structure constants of any acyclic CoHA can be approached through the well-developed combinatorics of quiver polynomials.","When $Q$ is itself a Dynkin quiver, the theorem with $Q_\\bullet=\\{Q\\}$ recovers the positive-root decomposition and with singleton subquivers recovers the simple-root decomposition, interpolating between the two previously known presentations.","For general acyclic quivers, the proof realizes the semistable CoHA decompositions of earlier work in explicit topological and characteristic-class terms rather than through stability conditions."],"supporting_citations":[{"why":"Defines the CoHA and its localization-formula multiplication; the $A_1$ exterior-algebra computation is the model for each tensor factor.","marker":"[KS11]"},{"why":"Identifies indecomposable representations of Dynkin quivers with positive roots, so orbit data is encoded by Kostant partitions.","marker":"[Gab72]"},{"why":"Provides the desingularizations of quiver orbit closures by quiver flag varieties used in Proposition 4.8.","marker":"[Rei03]"},{"why":"Describes isotropy subgroups and the restriction map for Dynkin quiver orbits, generalized here to the normal locus of a stratum.","marker":"[FR02]"},{"why":"Establishes the Dynkin CoHA structure theorem and the localization and restriction techniques that the paper extends to acyclic quivers.","marker":"[Rim13]"},{"why":"Proves rational singularities for type A Dynkin orbit closures, a known case of the assumption in Remark 4.9.","marker":"[BZ01]"},{"why":"Proves rational singularities for type D Dynkin orbit closures, the other known case of the assumption in Remark 4.9.","marker":"[BZ02]"},{"why":"Supplies the quantum dilogarithm identity that equates the Poincaré series of the two tensor decompositions.","marker":"[All18]"}],"fun_headline_variants":["CoHA of any acyclic quiver decomposes into Dynkin subalgebras","Dynkin subquiver partitions yield CoHA isomorphisms","Topological isomorphisms decompose acyclic CoHAs","CoHA structure constants from Dynkin quiver polynomials","Every acyclic quiver CoHA is a tensor product of Dynkin pieces"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on an unproved property of the exceptional Dynkin diagrams E6 and E7: the closures of their representation-space orbits must have only rational singularities, a regularity condition known to hold for types A and D but still open for E.","fun_headline_variants_meta":{"raw":{"variants":["CoHA of any acyclic quiver decomposes into Dynkin subalgebras","Dynkin subquiver partitions yield CoHA isomorphisms","Topological isomorphisms decompose acyclic CoHAs","CoHA structure constants from Dynkin quiver polynomials","Every acyclic quiver CoHA is a tensor product of Dynkin pieces"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000821,"raw_usage":{"total_tokens":3640,"prompt_tokens":1042,"completion_tokens":2598,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":658,"completion_tokens_details":{"reasoning_tokens":2507}},"tokens_in":658,"tokens_out":2598,"duration_ms":19808,"temperature":1.0,"reasoning_tokens":2507,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:47:10.435826+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take an acyclic quiver obtained by appending a single vertex to an E6 (or E7) Dynkin quiver along one arrow, choose the admissible partition consisting of that Dynkin subquiver and the new $A_1$ vertex, and compute the twisted Poincaré series of both sides of (22) using the formulas (30) and (31); any $q$-degree in which they differ would refute the claimed isomorphism.","supporting_citations":[],"review_version":1}