{"id":"72629589-94a0-476c-89f4-ffe16cdd4260","arxiv_id":"2411.18067","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Complements of branch curves of generic projections of complete intersection surfaces are CAT(0), while the universal families for E6, E7, and E8 singularities are not CAT(0).","lead":"This paper studies when the fundamental group of the complement of a plane curve, or of a universal family attached to a surface singularity, admits a CAT(0) geometric structure. It gives new positive results for branch-curve complements of complete intersection surfaces and new negative results for the universal families of E6, E7, and E8 singularities.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Theorem 4.8 asserts, without proof, that the center of A_R gives a central direct-factor splitting of the finite-index subgroup G'; if the boundary multitwist acts nontrivially on π1(C_R), the reduction to H' and the non-CAT(0) conclusion for π1(V_R) do not follow.","rationale":"I agree with the reader that the central-splitting step in Theorem 4.8 is the weakest point in the paper's main negative claim. The step is plausibly true — a boundary multitwist should act trivially on the fundamental group of the open surface — but the written proof does not establish it, and the justification offered ('since Z(A_R) is free') is not the correct reason. The concern is load-bearing because both the Bowers–Ruane reduction and the application of Theorem 2.5 to H' depend on the direct-product decomposition G'=H'×σ(Z). I would keep the verdict CONDITIONAL: the gap is repairable, but the authors should expand this argument (or cite a proof that boundary twists act trivially on π1) before the paper is accepted. The secondary issue in Proposition 3.2, where the statement uses π1(Bl_O(P^2)\\C) but the proof treats Bl_O(P^2)\\(C∪L), is a clear typo and does not affect the main theorems once corrected. I did not find a more serious objection to Theorem 3.6 or to the proof of Corollary 4.5; those parts appear sound, and Theorem 3.6 is supported by Robb's presentation.","tokens_in":22571,"tokens_out":23452,"duration_ms":222332,"concrete_test":"For R=E6, take the standard geometric representation of A_E6 on the Milnor fiber (e.g., the six Dehn-twist generators in Figure 4) and compute the image of a generator of the center (the Garside element Δ^2) in Out(F_6) by acting on a free generating set of π1(C_R). If this outer automorphism is trivial, the central splitting in Theorem 4.8 is justified; if it is nontrivial, the claimed direct-product decomposition and the non-CAT(0) proof for π1(V_R) collapse and need to be reworked.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Theorem 4.8, after passing to the finite-index cover G' = (π_R)_*^{-1}(A'), the proof states that the center Z(A_R) acts by a boundary multitwist on the fiber C_R, and then immediately concludes 'there is a splitting σ:Z→G' such that σ(Z) is central in G'', hence G' = H'×σ(Z). This is the load-bearing step in the negative result: the subsequent Bowers–Ruane argument and the application of Theorem 2.5 to H' both require the center to split off as a direct factor. The written justification is incomplete. A boundary multitwist is central in Mod(C_R) fixing boundary pointwise, but that alone does not imply that a section of the extension is central in G' or that the extension splits. One must show that the induced outer automorphism of π1(C_R) ≅ F_g(R) is trivial, so a lift of the central generator can be adjusted by an element of F_g(R) to commute with the fiber subgroup; the split then follows because F_g(R) has trivial center for g(R)≥2. The phrase 'since Z(A_R) is free' supplies neither of these facts. If the boundary multitwist represented a nontrivial class in Out(F_g(R)), the monodromy of the extension would not factor through P'=ker α, and the reduction G'=H'×Z would fail, invalidating the non-CAT(0) conclusion for V_R.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies whether fundamental groups of plane curve complements and of certain universal deformation families are CAT(0). In the positive direction, it proves that for a generic projection of a smooth complete intersection surface, the complement of the branch curve in P^2 has CAT(0) fundamental group, by showing that the groups ~B_d and their quotients appearing in Robb's theorem are virtually abelian for d ≥ 4. It also claims that complements with finite monodromy admit CAT(0) fundamental groups and, in fact, finite-volume nonpositively curved Riemannian metrics. In the negative direction, it uses Zhu's LIP obstruction to show that the complement of the 3-cuspidal quartic together with singular fibers is not CAT(0), and that the fundamental group of the universal family over the semi-universal deformation of E6, E7, and E8 simple singularities is not CAT(0).","tokens_in":22872,"tokens_out":11905,"duration_ms":112328,"significance":"If the results are correct, the paper supplies new positive and negative examples in the study of CAT(0) groups among quasi-projective fundamental groups, and it connects the longstanding question of whether braid groups are CAT(0) with plane-curve complements. The proof of Theorem 3.6 is short and checkable, and the explicit braid-monodromy computation for the 3-cuspidal quartic is a useful concrete contribution. The paper is transparent about its reliance on external results of Robb, Wajnryb, Zhu, Calvez--Wiest, and others, and it contains no fitted parameters or circular reasoning. However, two load-bearing points need repair: the statement of Proposition 3.2 does not match its proof, and the proof of Theorem 4.8 contains an unjustified splitting argument.","major_comments":[{"comment":"The proposition is stated for π1(BlO(P^2) \\ C), but the proof uses that BlO(P^2) \\ C is the total space of a locally trivial punctured-sphere bundle over a punctured sphere. That bundle statement is only true after deleting the singular fibers L, i.e. for BlO(P^2) \\(C ∪ L), as established in Section 2.1. The proof of Lemma 3.4 and the sentence 'we know that BlO(P^2 \\ C) is the total space of a locally trivial S_{0,n+1}-bundle over S_{0,m+1}' are therefore inconsistent with the proposition's hypothesis. Moreover, the proposition's second assertion also switches notation to 'BlO(P^2 ∖ C)', which is not a standard blow-up. The statement should be corrected to concern BlO(P^2) \\(C ∪ L), in line with Theorem A(1); as written, the proposition is not a valid theorem about π1(BlO(P^2) \\ C).","section":"Section 3.1, Proposition 3.2"},{"comment":"The proof asserts that because the center Z = Z(A_R) acts by a boundary multitwist and 'since Z(A_R) is free, there is a splitting σ: Z → G′ such that σ(Z) is central in G′.' This does not follow. The freeness of Z as an abstract group does not split an arbitrary extension, and centrality of the boundary multitwist in Mod(C_R) does not by itself imply that the associated outer automorphism of π1(C_R) ≅ F_{g(R)} is trivial. To justify the direct-product decomposition G′ = H′ × σ(Z), one must show that the monodromy of Z is trivial in Out(F_{g(R)}), then use the trivial center of F_{g(R)} to adjust a lift of the generator of Z to a central element. This is load-bearing: the subsequent Bowers–Ruane reduction and the application of Theorem 2.5 to H′ both require the splitting. The gap is local and likely fixable, but the argument as written is incomplete.","section":"Section 4.3, proof of Theorem 4.8"}],"minor_comments":[{"comment":"The notation for L is inconsistent: earlier L = ∪_{i=1}^4 L_i, but the displayed extension before Theorem 4.2 writes 'π1(BlO(P^2) ∖ (C ∪_{i=1}^3 L_i))'. Since the base is P^1 with four punctures, the sum should be over all four singular fibers.","section":"Section 4.1"},{"comment":"The equation of the 3-cuspidal quartic has mismatched parentheses: 'C = {F([x ∶ y ∶ z] = ... = 0}' is missing a closing brace/parenthesis. This is purely typographical but should be fixed.","section":"Section 4.1"},{"comment":"In the discussion of L0, the phrase 'setting τ2 = 1 and τ1τ2τ3 = 1' appears to be a typo; with four singular fibers one expects the relations τ2 = τ3 = 1 and τ1τ4 = 1 coming from the product relation. The intended computation is clear, but the displayed relation is confusing.","section":"Section 4.1"},{"comment":"The sentence 'The fundamental group π1(F1 ∖ C) where F1 is the blow-up of P2 at P has expression similar to (12)' uses P without definition and conflicts with the point O elsewhere; the notation should be unified.","section":"Appendix A.2.1"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this one has real content. The genuinely new thing is Theorem 3.6: ~B_d, the quotient of the braid group appearing in Robb's theorem, is virtually abelian, so the branch locus of a generic projection of a smooth complete intersection surface has CAT(0) complement. That's a clean argument from Robb's presentation, and it gives a real infinite family. The other main result, Theorem B, applies Zhu's LIP obstruction to the universal families of E6/E7/E8, using Wajnryb's non-injectivity. That is a natural and correct way to obtain non-CAT(0) quasi-projective groups. The paper also computes the 3-cuspidal quartic example explicitly, which is useful.\n\nThe soft spots are real but mostly cosmetic. Proposition 3.2 is stated for π1(Bl_O(P^2)\\C), but the fibration and the finite-monodromy hypothesis only make sense for C∪L, the complement of the singular fibers. That's a typo in the statement, not a conceptual issue.\n\nThe more substantive wart is in Theorem 4.8. The paper wants to split the infinite cyclic center of A_R off as a direct factor of a finite-index subgroup G' ≤ π1(V_R). It says the center acts by a boundary multitwist, and then 'since Z(A_R) is free' there is a central splitting. The stress-test note worries that a boundary multitwist might be nontrivial in Out(F_g). I don't think that specific worry lands: a boundary-parallel Dehn twist on a surface with boundary fixes a core of the collar and induces the identity on π1, so the action on the fiber is trivial. But the step is still compressed. Trivial action on the fiber gives a splitting over Z; it does not by itself imply the section can be chosen to centralize the rest of G'. The paper should spell out the direct-product decomposition. I suspect the claim is true, but as written it's a jump.\n\nOverall, the central assertions look correct, the novelty is real, and the citation pattern is fine—they lean on Robb, Wajnryb, Zhu, Calvez-Wiest, which are the right sources. The paper would benefit from a careful referee focusing on Theorem 4.8 and the Proposition 3.2 statement. I'd send it out; it deserves a serious referee.","headline":"Solid paper with two genuinely new families; the E6/E7/E8 negative result has a compressed central-splitting step that needs a fix, but the stress-test concern about nontrivial outer monodromy is likely a red herring.","tokens_in":23468,"tokens_out":11972,"would_cite":true,"duration_ms":110330,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20F65","14H50","32S25","57M07"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that complements of branch curves of generic complete intersections have CAT(0) fundamental groups, and that the universal families for E6, E7, and E8 simple singularities do not.","keywords":["CAT(0) group","plane curve complement","braid monodromy","complete intersection surface","Artin group","simple singularity","nonpositive curvature","geometric monodromy"],"falsifier":"Compute the action of the center $Z(A_R)$ on $\\pi_1(C_R)$ through the monodromy of the bundle $V_R \\to U_R$; if the boundary multitwist does not generate a direct $\\mathbb{Z}$ factor of a finite-index subgroup of $\\pi_1(V_R)$, the E6/E7/E8 proof loses its reduction. Independently, constructing a proper cocompact isometric action of $\\pi_1(V_R)$ on any CAT(0) space for a single E-type would refute the theorem.","tokens_in":22341,"feed_emoji":"📐","tokens_out":11515,"duration_ms":90094,"temperature":0.7,"pith_summary":"The paper asks which fundamental groups of plane-curve complements and their deformation families can carry CAT(0) geometry, a question motivated by the conjecture that braid groups are CAT(0). It answers positively for branch loci of generic projections of smooth complete intersection surfaces: the affine and projective complement groups are CAT(0), and the proof goes through the virtual abelianness of the quotients $\\tilde B_d$ of braid groups. It answers negatively for the universal families over the semi-universal deformations of the simple singularities $E_6$, $E_7$, and $E_8$: those fundamental groups are not CAT(0). A separate computation shows that deleting singular fibers from the 3-cuspidal quartic projection produces a free-by-free group with infinite monodromy that is not CAT(0).","feed_headline":"Branch-curve groups get CAT(0) geometry; E6/E7/E8 don't","feed_subtitle":"Generic complete-intersection branch complements are virtually abelian; universal E6/E7/E8 deformation families fail the monodromy test.","key_machinery":"The main obstruction is Property LIP: a group $Q$ has Property LIP if every infinite normal subgroup contains an infinite finitely generated subgroup with finite centralizer, and acylindrically hyperbolic groups have it. The load-bearing theorem states that if a CAT(0) group $G$ is an extension $1 \\to R \\to G \\to Q \\to 1$ with $R$ finitely generated and centerless and $Q$ having Property LIP, then the algebraic monodromy $Q \\to \\mathrm{Out}(R)$ has finite image or finite kernel. The positive result uses the opposite mechanism: the cited presentation of $\\tilde P_{0,d}$ exhibits it as a central extension of $\\mathbb{Z}^{d-1}$ by $\\mathbb{Z}/2$, so $\\tilde B_d$ is virtually free abelian.","core_discovery":"The central assertion is that CAT(0) behavior in these families is governed by the monodromy of a fibration by punctured spheres or affine curves. On the positive side, the group $\\tilde B_d = B_d / \\langle\\![x_2, (x_3x_1)^{-1}x_2(x_3x_1)]\\!\\rangle$ is virtually abelian for every $d \\ge 4$; since this group is isomorphic to $\\pi_1(\\mathbb{C}^2 \\setminus C_a)$ and its quotient is $\\pi_1(\\mathbb{P}^2 \\setminus C)$, both complement groups act properly and cocompactly on a CAT(0) space. On the negative side, for each $R \\in \\{E_6, E_7, E_8\\}$ the extension $1 \\to F_{g(R)} \\to \\pi_1(V_R) \\to A_R \\to 1$ is argued to be non-CAT(0): after a finite-index cover splits off the center of the Artin group $A_R$ as a $\\mathbb{Z}$ factor, the residual extension has a geometric monodromy with an infinite kernel, which contradicts the Property LIP dichotomy for CAT(0) extensions.","pith_inferences":["The same LIP dichotomy plausibly applies to any singularity family whose geometric monodromy is known to have an infinite kernel, so the E6/E7/E8 technique is not tied to that specific root system.","For $A_n$ and $D_n$, where geometric monodromy is injective, the paper's obstruction disappears; determining whether $\\pi_1(V_R)$ is CAT(0) there would separate the two monodromy regimes.","A testable extension is to compute the action of the center of $A_R$ on the fiber free group directly; if that action is not a boundary multitwist that splits as a direct factor, the product structure used in the negative proof would need revision.","On the positive side, the virtual abelianness of $\\tilde B_d$ suggests these branch-curve complements are as flexible as possible among CAT(0) groups, potentially admitting flat-space actions rather than only general CAT(0) actions."],"forward_implications":["For every branch curve of a generic projection of a smooth complete intersection surface, both $\\pi_1(\\mathbb{C}^2 \\setminus C_a)$ and $\\pi_1(\\mathbb{P}^2 \\setminus C)$ admit proper cocompact actions on Euclidean space.","The groups $\\tilde B_d$ are virtually free abelian for all $d \\ge 4$, so this infinite family of braid-group quotients is CAT(0) even though the full braid-group conjecture remains open.","The 3-cuspidal quartic example shows that a free-by-free group with infinite monodromy and infinite kernel can fail to be CAT(0), even when the underlying curve complement is finite and CAT(0).","The E6, E7, and E8 universal families supply new non-CAT(0) quasi-projective groups whose base is an Artin group rather than a surface group."],"supporting_citations":[{"why":"Supplies the presentation and center structure of $\\tilde B_d$ and identifies $\\pi_1(\\mathbb{C}^2 \\setminus C_a)$ with it.","marker":"[Rob97]"},{"why":"Provides Property LIP and the dichotomy theorem limiting algebraic monodromy in CAT(0) extensions.","marker":"[Zhu23]"},{"why":"Proves that geometric monodromy representations of the E6, E7, and E8 Artin groups have an infinite kernel.","marker":"[Waj99]"},{"why":"Shows geometric monodromy is injective for the An and Dn cases, marking the contrast for the negative result.","marker":"[PV92]"},{"why":"Shows the quotient of an Artin group by its center is acylindrically hyperbolic, yielding Property LIP for the base.","marker":"[CW17]"},{"why":"Gives the product criterion used to reduce CAT(0)ness of $G'$ to the core extension $H'$.","marker":"[BR96]"},{"why":"Provides the explicit braid monodromy computation for the 3-cuspidal quartic example.","marker":"[CW04]"},{"why":"Supplies the braid-monodromy images for smooth and nodal curves used in the non-CAT(0) complement corollary.","marker":"[Moi81]"}],"fun_headline_variants":["Curve complement groups are CAT(0); E6/E7/E8 are not","CAT(0) for branch loci, failure for E6/E7/E8 families","Braid groups inspired: complements CAT(0), universals not","Generic complements get CAT(0), E6/E7/E8 don't","CAT(0) geometry: yes for complements, no for E6/E7/E8"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The non-CAT(0) conclusion for E6/E7/E8 depends on the claim that the infinite cyclic center of the Artin group acts on the free-group fiber only through a boundary multitwist and splits off as a direct factor of a finite-index subgroup; if that splitting is not genuine, the reduction to a non-CAT(0) core extension fails.","fun_headline_variants_meta":{"raw":{"variants":["Curve complement groups are CAT(0); E6/E7/E8 are not","CAT(0) for branch loci, failure for E6/E7/E8 families","Braid groups inspired: complements CAT(0), universals not","Generic complements get CAT(0), E6/E7/E8 don't","CAT(0) geometry: yes for complements, no for E6/E7/E8"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000214,"raw_usage":{"total_tokens":1410,"prompt_tokens":912,"completion_tokens":498,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":528,"completion_tokens_details":{"reasoning_tokens":392}},"tokens_in":528,"tokens_out":498,"duration_ms":4235,"temperature":1.0,"reasoning_tokens":392,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:35:30.464134+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the action of the center $Z(A_R)$ on $\\pi_1(C_R)$ through the monodromy of the bundle $V_R \\to U_R$; if the boundary multitwist does not generate a direct $\\mathbb{Z}$ factor of a finite-index subgroup of $\\pi_1(V_R)$, the E6/E7/E8 proof loses its reduction. Independently, constructing a proper cocompact isometric action of $\\pi_1(V_R)$ on any CAT(0) space for a single E-type would refute the theorem.","supporting_citations":[],"review_version":1}