{"id":"62d10d3a-726c-4c3c-83dc-d697f0667917","arxiv_id":"2412.14310","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Equivariant cohomological rigidity holds for compact, connected, four-dimensional Hamiltonian S1-manifolds: every algebra isomorphism of equivariant cohomology rings is induced by an equivariant diffeomorphism.","lead":"For compact, connected four-dimensional symplectic spaces with a Hamiltonian circle action, this paper proves that the equivariant cohomology ring completely determines the space up to an equivariant diffeomorphism. Even more, any algebraic isomorphism between the rings is realized by an actual equivariant diffeomorphism.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Partial-flip construction in Lemmas 4.4–4.7 uses maps that are only equivariant up to inversion of the anti-diagonal circle; as written, the interpolated H cannot be S1-equivariant.","rationale":"The reader's weakest assumption concerned reliance on the unpublished preprint [14]. That is a legitimate external risk, but the most load-bearing problem I find is internal and appears in the present paper's own construction: the partial-flip diffeomorphism in Lemmas 4.4 and 4.7 is asserted to be equivariant with respect to the anti-diagonal circle action, but the explicit formulas (4.5), (4.6), (4.8), and the isotopy φ_t are not equivariant for that action. The maps are equivariant up to the coordinate-exchange automorphism, which acts on the anti-diagonal circle as inversion. This is not a matter of convention: for a generic point, F(λ·x) and λ·F(x) are different quotient classes unless λ = ±1. Since Proposition 4.11, Theorem 1.2, and ultimately Theorem 1.1 all rely on this construction, the proof as written has a gap at its core. The gap is likely repairable by inserting complex conjugations into the maps (the conjugated rotation does commute with the anti-diagonal action and preserves the required endpoint behavior), so I do not recommend rejection; rather, the manuscript should be revised and the corrected equivariance check supplied. I therefore keep a CONDITIONAL verdict, but for a reason different from the reader's. If the authors confirm the conjugation fix and it passes, the remaining external dependency on [14] can be assessed separately.","tokens_in":22251,"tokens_out":61060,"duration_ms":521598,"concrete_test":"Take the n = 3 corner manifold with v2 = (N,1) and a generic representative x with x2 ≠ 0. Set λ = i and compute both F(λ·x) and λ·F(x) from (4.5) and (3.7); they differ unless the second coordinate vanishes. Likewise compute φ_{π/2}(λ·(1,0)) and λ·φ_{π/2}(1,0) for λ = i; the two vectors are (−i, i) and (i, i) up to signs, not equal. Then test the proposed conjugated isotopy ψ_t(z1,z2) = (cos t z1 − sin t \\bar z2, cos t z2 + sin t \\bar z1) directly against the anti-diagonal action; it satisfies ψ_t(λ·z) = λ·ψ_t(z) for all λ ∈ S1 and all (z1,z2). Re-run the proof of Lemma 4.7 with ψ_t and with (4.5)–(4.6) and (4.8) replaced by their conjugated versions; check that (4.9) holds and that the glued map on the quotient is globally S1-equivariant.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The proof of Theorem 1.2 hinges on Proposition 4.11, which requires an orientation-preserving equivariant diffeomorphism realizing a partial flip. The local model is built from corner manifolds with the anti-diagonal circle action A = {(λ, λ^{−1})}. Lemma 4.4 defines F on representatives by (4.5), (x1, …, xn) ↦ (−xn, xn−1, …, x1). This map is only weakly equivariant with respect to the coordinate-exchange automorphism σ of (S1)^2; on A, σ sends (λ, λ^{−1}) to (λ^{−1}, λ). Consequently F does not satisfy F(λ·x) = λ·F(x) for generic points; it satisfies F(λ·x) = σ(λ)·F(x). The same defect appears in the isotopy φ_t in the proof of Lemma 4.7: φ_t(z1,z2) = (cos(πt/2) z1 − sin(πt/2) z2, cos(πt/2) z2 + sin(πt/2) z1) does not commute with the action λ·(z1,z2) = (λz1, λ^{−1}z2). Direct computation for λ = i shows the results differ by a factor of i in the first coordinate whenever sin(πt/2) ≠ 0. Since H must be A-equivariant on all of N, yet equals F on a triangular neighborhood and equals the identity on the complement of a larger neighborhood, no such map H can be produced by this interpolation. Replacing the rotation with a conjugated version, (z1,z2) ↦ (cos t z1 − sin t \\bar z2, cos t z2 + sin t \\bar z1), would restore equivariance and also gives the desired endpoint (−\\bar z2, \\bar z1). But as written, (4.5)–(4.6), (4.8), and Lemma 4.7 lack the conjugation, so the partial-flip construction is unsupported.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims that compact, connected, four-dimensional Hamiltonian S1-manifolds are equivariantly cohomologically rigid in the strong sense: every algebra isomorphism between their S1-equivariant cohomology rings over H^*_{S1}(pt;Z) is induced by an equivariant diffeomorphism. The proof proceeds by introducing a combinatorial invariant, the oriented dull graph, reducing the problem to realizing graph isomorphisms by equivariant diffeomorphisms, and then constructing the required diffeomorphism locally via mirror corner manifolds and a partial-flip operation. The final section uses the dull graph correspondence and a careful analysis of odd-degree equivariant cohomology to promote the equivariant diffeomorphism to one inducing the given algebra isomorphism.","tokens_in":22620,"tokens_out":11292,"duration_ms":98306,"significance":"If correct, Theorem 1.1 is a substantial result: it shows that the equivariant cohomology algebra, with no additional Chern class or moment-map data, completely determines the equivariant diffeotype of a four-dimensional Hamiltonian S1-manifold, and that every algebraic isomorphism is geometric. The paper contains several well-organized ingredients: explicit local models for corner manifolds, a detailed construction of symplectic blowup maps, and a separation of the odd-cohomology case. It also makes creative use of Karshon's decorated-graph classification and the authors' earlier dull-graph framework. However, the central local construction of the partial flip is not currently valid, so the main theorem is not established as written.","major_comments":[{"comment":"The map F is claimed to be equivariant with respect to the anti-diagonal circle A = {(λ, λ^{-1})}. It is not. For A_λ acting by (z1,z2) ↦ (λz1, λ^{-1}z2), direct computation gives F(A_λ·(x1,...,xn)) = (-λ^{-1} xn, x_{n-1},..., x2, λ x1), while A_λ·F(x1,...,xn) = (-λ xn, x_{n-1},..., x2, λ^{-1} x1). These differ for generic points unless λ = λ^{-1}. The proof itself notes that the reversing map is only weakly equivariant with respect to the coordinate-exchange automorphism of (S1)^2, and the additional sign change does not cancel this failure on the anti-diagonal. Consequently, the equivariance assertion in Lemma 4.4 and the local formula (4.6) are unsupported.","section":"Section 4, Lemma 4.4 (Eqs. (4.5)–(4.6))"},{"comment":"The rotation family φ_t(z1,z2) = (cos(πt/2) z1 - sin(πt/2) z2, cos(πt/2) z2 + sin(πt/2) z1) is not S1-equivariant for the stated action λ·(z1,z2) = (λz1, λ^{-1}z2). For example, at t = 1/2 and z = (1,0), φ_t(A_i z) = (i√2/2, i√2/2) whereas A_i φ_t(z) = (i√2/2, -i√2/2). Since the cut-off map g and hence H are built from φ_t, the map H in Lemma 4.7 is not A-equivariant. A conjugated rotation would restore equivariance, but its endpoint is (-\\bar z2, \\bar z1), not (-z2,z1), so it would not agree with (4.8) unless F is also modified. As written, the interpolation construction does not produce the required equivariant diffeomorphism.","section":"Section 4, Lemma 4.7 (formula for φ_t after Eq. (4.8))"},{"comment":"The partial-flip operation is the only place where an equivariant diffeomorphism between manifolds with different orientations of the dull graph is constructed, and it relies directly on Lemma 4.7. Since Lemma 4.7 is invalid as written, the gluing in Proposition 4.11 produces only a diffeomorphism, not an S1-equivariant one. Without Proposition 4.11, the inductive proof of Theorem 1.2 fails, and so do Corollary 1.3 and the equivariant-diffeomorphism part of Theorem 1.1. This is a load-bearing gap.","section":"Section 4, Proposition 4.11 and proof of Theorem 1.2"},{"comment":"The paper delegates two essential reductions to [14], an unpublished preprint by two of the three authors: the correspondence between equivariant cohomology isomorphisms and dull-graph isomorphisms, and the realization of dull-graph isomorphisms by equivariant diffeomorphisms. The proof of Theorem 1.1 and the proof of Theorem 1.2 are therefore conditional on results that are not included, or independently verified, in this manuscript. This is not an internal inconsistency, but it is a serious dependency: the editors should verify the status and correctness of [14], and the authors should either state the dependence more prominently or supply the needed statements.","section":"Sections 2 and 5 (Lemmas 2.6, and the use of [14, Corollary 7.23, Theorem 7.22, Proposition 7.26])"}],"minor_comments":[{"comment":"The bullet list contains a typo: 'mininum' should be 'minimum'.","section":"Section 2, Definition 2.5"},{"comment":"The text around Figure 3.12 contains unpolished drafting notes, for example 'MINIMAL MODELS: generically assume m>n>0.  I can change that.....  Should I indicate what happens if we scale the symplectic form...' These should be removed before submission.","section":"Figure 3.12 and surrounding text"},{"comment":"The formula for ϖ writes its source as [x1, ..., x_{n+1}] although the manifold has n coordinates; the notation should be made consistent.","section":"Section 3, Lemma 3.17 (Eq. (3.19) and preceding display)"},{"comment":"The proof asserts that the constructed map induces an isomorphism of dull graphs, but it does not explicitly verify that the labels (self-intersection e and genus g) are preserved by the gluing. A sentence spelling this out would improve clarity.","section":"Section 4, Proposition 4.11"}],"recommendation":"major_revision","confidential_remarks":"The equivariance failure in Section 4 is localized and appears repairable: replacing the coordinate reversal and rotation by their complex-conjugated versions would restore anti-diagonal equivariance, at the cost of changing the endpoint of the interpolation and the definition of F. Because the fix is not present in the manuscript, I do not recommend acceptance yet. I also urge the editor to check the status of the authors' preprint [14], since Theorem 1.1 is directly contingent on it."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things about this paper. The main theorem — strong equivariant cohomological rigidity for four-dimensional Hamiltonian S1-manifolds — is a natural and significant result, and the paper is on the whole well organized and carefully written. But there is a concrete error in the proof of the partial-flip lemma: the maps used to interpolate between mirror corner manifolds are not S1-equivariant with respect to the anti-diagonal circle action. I checked the computation; the stress-test note is correct. In Lemma 4.4, the map (4.5) sends (z1,z2) to (−z2,z1), which fails to commute with λ·(z1,z2)=(λz1,λ−1z2). It commutes only up to the coordinate-swap automorphism, i.e., up to λ→λ−1. In Lemma 4.7, the rotation φt has the same defect: its first component picks up a λ−1 where it should have a λ. So the interpolation map H cannot be S1-equivariant as written. The fix is straightforward — insert complex conjugations in (4.5) and in φt, so the endpoint becomes (−z̄2,z̄1) — but without that, the partial-flip construction in Proposition 4.11 is unsupported, and Theorem 1.2 does not follow.\n\nWhat the paper does well: the reduction to dull graphs, the local models via corner manifolds, and the careful handling of odd cohomology in Lemma 5.1. The proof of Theorem 1.1 in Section 5 is a model of how to manage orientation-preserving and orientation-reversing cases. If the equivariance gap is closed, the result is a clean resolution of a well-known open problem.\n\nThe other soft spot is the heavy reliance on [14], an unpublished preprint by two of the three authors. The reductions from equivariant cohomology to dull graphs and from graph isomorphisms to diffeomorphisms both come from [14]. That is a genuine dependency, though not fatal; a referee will want to verify [14]'s key results rather than take them on faith.\n\nThe reader's conditional verdict is fair. I would add that the equivariance issue is concrete, not a vague concern. Still, I would absolutely send this to peer review: the result is important, the structure is solid, and the flaw is very likely repairable. My recommendation: major revision, with the authors asked to fix the equivariance and either publish [14] or make its key proofs easily accessible.","headline":"Strong theorem, but the partial-flip construction has an S1-equivariance gap that is likely fixable with a conjugation; worth refereeing in major-revision form.","tokens_in":23223,"tokens_out":7020,"would_cite":false,"duration_ms":51599,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53D35","55N91","53D20","57S15"],"pacs":[],"model":"deepseek-v4-flash","headline":"For compact, connected, four-dimensional Hamiltonian circle manifolds, equivariant cohomology determines the equivariant diffeotype: every algebra isomorphism between their equivariant cohomology rings is induced by an equivariant…","keywords":["symplectic geometry","Hamiltonian torus action","moment map","complexity one","equivariant cohomology","cohomological rigidity","dull graph","circle actions"],"falsifier":"Compute the automorphism group of $H^*_{S^1}(M;\\mathbb Z)$ for a Hirzebruch surface with a Hamiltonian circle action and compare it with the pullback action of its $S^1$-equivariant diffeomorphism group. The theorem predicts the two match; any algebra automorphism not realized by an equivariant diffeomorphism would refute it. Equivalently, any pair of these manifolds with isomorphic equivariant cohomology algebras but non-isomorphic dull graphs would refute Corollary 1.3.","tokens_in":21972,"feed_emoji":"🔄","tokens_out":17415,"duration_ms":138163,"temperature":0.7,"pith_summary":"This paper asks how much of a manifold with a circle symmetry is remembered by its equivariant cohomology. For compact, connected, four-dimensional Hamiltonian circle actions—symplectic manifolds with a circle action generated by a moment map—the equivariant cohomology ring, together with the base ring coming from the circle acting on a point, remembers everything up to equivariant diffeomorphism. The main theorem states that any algebra isomorphism between two such rings is induced by an equivariant diffeomorphism, so no purely algebraic isomorphism is left unrealized geometrically. The proof compresses the ring to a finite labeled graph, the dull graph, and then builds the required diffeomorphisms using local toric models called corner manifolds.","feed_headline":"Equivariant cohomology pins down 4D Hamiltonian circle manifolds","feed_subtitle":"Two such manifolds are equivariantly diffeomorphic exactly when their equivariant cohomology rings are isomorphic.","key_machinery":"The machinery has three layers. First is the dull graph: a finite labeled graph with one vertex per fixed component, an edge for each isotropy $\\mathbb Z_k$-sphere labeled by $k$, thin vertices for isolated fixed points labeled extremal or not, and fat vertices for fixed surfaces labeled by genus and self-intersection. It is the image of the decorated-graph classification [15] after forgetting moment-map values and surface areas, and the earlier result [14] shows the equivariant cohomology algebra determines it. Second is an orientation on the dull graph, directed by the moment map; any two orientations differ by the opposite orientation and finitely many partial flips along free chains. Third is the geometric realization of a partial flip: a triangular neighborhood of a chain of isotropy spheres inside a corner manifold—a noncompact symplectic toric manifold whose moment image is a Delzant polyhedral set in the positive quadrant, a convex region with integer normal vectors whose adjacent pairs form a basis of $\\mathbb Z^2$—is replaced by the mirror corner manifold. The symplectic blowup map identifies the complement of the chain with a subset of $\\mathbb C^2$, where an explicit rotation $\\varphi_t(z_1,z_2)=(\\cos(\\pi t/2)z_1-\\sin(\\pi t/2)z_2,\\cos(\\pi t/2)z_2+\\sin(\\pi t/2)z_1)$, cut off by a radial function, produces an $S^1$-equivariant diffeomorphism between the mirror manifolds.","core_discovery":"The central claim is strong equivariant cohomological rigidity. Let $M$ and $\\widetilde M$ be compact, connected, four-dimensional Hamiltonian $S^1$-manifolds, and let $\\Lambda\\colon H^*_{S^1}(M;\\mathbb Z)\\to H^*_{S^1}(\\widetilde M;\\mathbb Z)$ be an isomorphism of algebras over $H^*_{S^1}(\\mathrm{pt};\\mathbb Z)$. The theorem asserts that an equivariant diffeomorphism $\\widetilde M\\to M$ exists whose pullback is exactly $\\Lambda$. It follows that two such manifolds are equivariantly diffeomorphic if and only if their equivariant cohomology rings are isomorphic as algebras, and that equivariant homeomorphism, equivariant diffeomorphism, algebra isomorphism, and isomorphism of the associated dull graphs are equivalent conditions. The proof shows that the algebra determines the dull graph, that every isomorphism of dull graphs is realized by an orientation-preserving equivariant diffeomorphism, and that the 'partial flip' reversing the orientation along a chain of isotropy spheres can be built from local corner-manifold models.","pith_inferences":["A practical consequence the authors leave implicit: because the dull graph is finite and its labels are only stabilizer orders, genus, and self-intersection, deciding whether two such manifolds are equivariantly diffeomorphic is a finite combinatorial check once the fixed-point data is known.","The same corner-manifold interpolation may transplant to Hamiltonian torus actions with one-dimensional moment image in higher dimensions, provided a decorated-graph classification exists there; the rotation family in $\\mathbb C^2$ itself does not use the four-dimensional assumption beyond the local model.","The result suggests that continuously deforming the moment-map values and surface areas while preserving the dull graph does not change the equivariant diffeotype; this could be tested by explicitly deforming a manifold and tracking whether the constructed diffeomorphism persists.","In the $b_2=2$ case, where orientation-reversing equivariant diffeomorphisms are forced to exist, the algebra may in fact determine the full group of equivariant self-diffeomorphisms up to equivariant isotopy; this is an extension the paper does not state."],"forward_implications":["Equivariant cohomology is a complete equivariant diffeotype invariant for these manifolds: two are equivariantly diffeomorphic exactly when their equivariant cohomology rings are isomorphic as algebras over the equivariant cohomology of a point.","Every algebra isomorphism between such rings is the pullback of an equivariant diffeomorphism, so the algebraic automorphism group of the ring is realized by equivariant symmetries of the manifold.","Equivariant diffeomorphism, equivariant homeomorphism, algebra isomorphism, and dull-graph isomorphism are equivalent for compact, connected, four-dimensional Hamiltonian $S^1$-manifolds (Corollary 1.3).","An orientation-reversing equivariant diffeomorphism exists exactly when the second Betti number is $2$; otherwise every equivalence can be chosen orientation-preserving (Remark 1.4)."],"supporting_citations":[{"why":"Supplies the bridge from equivariant cohomology algebras to dull graphs; Lemma 2.6 realizes orientation-preserving dull-graph isomorphisms by equivariant diffeomorphisms.","marker":"[14]"},{"why":"Provides the decorated-graph classification of compact connected four-dimensional Hamiltonian circle manifolds; Theorem 2.4 turns graph isomorphisms into equivariant symplectomorphisms.","marker":"[15]"},{"why":"Provides the equivariant perfection and localization framework used for fixed-component structure, canonical classes, and the orientation computations in Section 5.","marker":"[3]"},{"why":"Provides the localization formula used to compare integrals over fixed surfaces and to prove the odd-cohomology generation statements in Lemma 5.1.","marker":"[5]"},{"why":"Provides the symplectic reduction construction of toric manifolds, used to build the corner-manifold local models.","marker":"[8]"},{"why":"Establishes uniqueness of the symplectic toric manifold with a given moment image, identifying every corner manifold with the reduced model $M_\\Delta$.","marker":"[16]"},{"why":"Provides the mapping-class-group result used in Lemma 5.2 to realize a prescribed orientation-preserving cohomology isomorphism on a positive-genus fixed surface.","marker":"[9]"}],"fun_headline_variants":["Cohomology ring decides diffeotype of 4D circle actions","Equivariant cohomology is the whole story for these 4D manifolds","4D Hamiltonian S^1-manifolds: cohomology determines all","Circle actions on symplectic 4-manifolds: cohomology says it all","One equivariant cohomology ring, one diffeomorphism type: 4D case"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes prior results that a small labeled graph completely describes each of these manifolds and that any isomorphism between such graphs can be realized by an equivariant diffeomorphism; if either prior classification has a hidden exception, the main theorem would not follow.","fun_headline_variants_meta":{"raw":{"variants":["Cohomology ring decides diffeotype of 4D circle actions","Equivariant cohomology is the whole story for these 4D manifolds","4D Hamiltonian S^1-manifolds: cohomology determines all","Circle actions on symplectic 4-manifolds: cohomology says it all","One equivariant cohomology ring, one diffeomorphism type: 4D case"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000208,"raw_usage":{"total_tokens":1367,"prompt_tokens":871,"completion_tokens":496,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":487,"completion_tokens_details":{"reasoning_tokens":387}},"tokens_in":487,"tokens_out":496,"duration_ms":4178,"temperature":1.0,"reasoning_tokens":387,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:20:54.803430+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the automorphism group of $H^*_{S^1}(M;\\mathbb Z)$ for a Hirzebruch surface with a Hamiltonian circle action and compare it with the pullback action of its $S^1$-equivariant diffeomorphism group. The theorem predicts the two match; any algebra automorphism not realized by an equivariant diffeomorphism would refute it. Equivalently, any pair of these manifolds with isomorphic equivariant cohomology algebras but non-isomorphic dull graphs would refute Corollary 1.3.","supporting_citations":[{"cited_title":"Equivariant cohomology of a complexity-one four-manifold is determined by combinatorial data","cited_arxiv_id":"1912.05647","evidence_quote":"Supplies the bridge from equivariant cohomology algebras to dull graphs; Lemma 2.6 realizes orientation-preserving dull-graph isomorphisms by equivariant diffeomorphisms."},{"cited_title":"Karshon, Periodic Hamiltonian flows on four dimensional manifolds , Memoirs of the Amer","cited_arxiv_id":null,"evidence_quote":"Provides the decorated-graph classification of compact connected four-dimensional Hamiltonian circle manifolds; Theorem 2.4 turns graph isomorphisms into equivariant symplectomorphisms."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the equivariant perfection and localization framework used for fixed-component structure, canonical classes, and the orientation computations in Section 5."},{"cited_title":"Berline and M","cited_arxiv_id":null,"evidence_quote":"Provides the localization formula used to compare integrals over fixed surfaces and to prove the odd-cohomology generation statements in Lemma 5.1."},{"cited_title":"Delzant, Hamiltoniens p´ eriodiques et image convexe de l’application moment , Bulletin de la Soci´ et´ e Math´ ematique de France116 (1988), 315–339","cited_arxiv_id":null,"evidence_quote":"Provides the symplectic reduction construction of toric manifolds, used to build the corner-manifold local models."},{"cited_title":"Karshon and E","cited_arxiv_id":null,"evidence_quote":"Establishes uniqueness of the symplectic toric manifold with a given moment image, identifying every corner manifold with the reduced model $M_\\Delta$."},{"cited_title":"Farb and D","cited_arxiv_id":null,"evidence_quote":"Provides the mapping-class-group result used in Lemma 5.2 to realize a prescribed orientation-preserving cohomology isomorphism on a positive-genus fixed surface."}],"review_version":1}