{"id":"d99de8e4-1b91-4337-a24f-c83418fefae9","arxiv_id":"2608.10973","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Morphisms from smooth projective varieties to P^1 can have singular fibers that are topologically invisible, yielding counterexamples to the Fernandez de Bobadilla-Kollar, Kollar-Pardon, Kotschick, and Schreieder conjectures.","lead":"This paper constructs explicit maps from smooth complex projective varieties to the projective line whose fibers are singular in the algebraic sense yet invisible to the topology of the total space. The examples refute four published conjectures on equisingularity, holomorphic one-forms, universal covers, and Aomoto complexes.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"PL-sphere claim for local model (13) is asserted over a non-isolated singularity via Brieskorn's Satz 1 without the required reduction; item (iii) hangs on it.","rationale":"The reader's weakest_assumption identifies exactly this step, and I agree. This is the only place where a central property (PL fibers) depends on a deferred citation to a theorem whose hypotheses are not visibly met. The rest of the construction—Z-homology fiber bundle in Section 2, the blow-up lifting in Section 3, the h-cobordism smoothing in Section 4—is internally consistent and the cited statements (BBD, Ehresmann, Smale, Cerf) are standard. The gap is not fatal, since the missing computation is short and almost certainly true: x^2+y^3-u'w is a Thom-Sebastiani sum of the cusp x^2+y^3 with the node u^2+v^2, and the resulting link is a simply connected rational homology 5-sphere, hence a standard S^5. But because the paper does not write this reduction, the proof of Theorem 1(iii) is incomplete as printed. Thus ACCEPT should be made CONDITIONAL on adding this verification (or an explicit citation to the precise Brieskorn criterion for exponents (2,3,2,2)). This matches the reader's medium correctness risk.","tokens_in":22964,"tokens_out":24446,"duration_ms":220788,"concrete_test":"Carry out the reduction of (13): set u' = a+b, w = a-b (then -u'w = b^2 - a^2, and after b -> ib this is a^2+b^2), so the transverse singularity is x^2+y^3+a^2+b^2=0 in C^4, a Pham-Brieskorn hypersurface with exponents (2,3,2,2). Using the Pham/Brieskorn formula, compute the Milnor monodromy eigenvalues: they are exp(2pi i (1/2+1/2+1/2+b/3)) for b=1,2, i.e. primitive sixth roots, giving characteristic polynomial t^2-t+1 and chi(1)=1. Since the link of an isolated hypersurface singularity in C^4 is simply connected (Milnor), the link is a homotopy S^5; by Smale's h-cobordism theorem (and Kervaire-Milnor Theta_5=0) it is diffeomorphic to the standard S^5. Then confirm that the link of a point on the v-axis in (13) is the join S^5*S^1, hence a standard PL S^7.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The proof of Theorem 1(iii) (all fibers are PL-manifolds) rests on the assertion, in the proof of Theorem 1 near equation (13), that the link of every point of the singular locus of the local model {x^2+y^3-u'w=0} in A^5 is a standard PL sphere, justified by [Bri66, Satz 1] and a join decomposition with a circle. The singular locus of (13) is the v-axis, so the validity requires two subclaims: (a) the transverse isolated singularity x^2+y^3-u'w=0 in C^4 has link S^5; (b) the link in (13) is the join of that S^5 with the S^1 from the v-direction. Subclaim (b) is standard, but (a) is not a direct reading of Satz 1, since (13) is not a Pham-Brieskorn hypersurface as printed. One must first make the linear change -u'w = u^2-v^2 (or u^2+v^2), reducing to x^2+y^3+u^2+v^2 with exponents (2,3,2,2), and then check that this exponent vector satisfies the Brieskorn sphere criterion. The paper does not supply this check or the resulting characteristic polynomial of the Milnor monodromy (which is t^2-t+1, so chi(1)=1). Without it, the link could in principle be a nonstandard homotopy 5-sphere or have nontrivial H_2, in which case the singular fibers would not be locally cones over PL spheres and item (iii) would fail. This is the single most load-bearing step for the 'looks topologically smooth' conclusion.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs explicit morphisms from smooth complex projective varieties to the projective line that have singular fibers while being topologically indistinguishable from smooth fibrations: the direct images of the constant integral sheaf are constant, the map is a homotopy fiber bundle with simply connected fibers, all fibers are PL-manifolds, and the map is homotopic to a C-infinity fiber bundle, yet it is not a C^0 fiber bundle. The same construction is used to give counterexamples to four conjectures: the smoothness conjecture of Fernández de Bobadilla–Kollár, a question of Kollár–Pardon on universal covers, Kotschick's conjecture on one-forms, and Schreieder's conjecture on Aomoto complexes. The paper also proves a positive equisingularity theorem for fibrations over Riemann surfaces and a rigidity result in dimension three.","tokens_in":23319,"tokens_out":19729,"duration_ms":198659,"significance":"If the results are correct, they settle several open problems and show that singular fibers of morphisms from smooth projective varieties can be invisible to the topology of the total space. The main construction is explicit and relatively low-dimensional (dimension five), and the paper connects it to substantial existing machinery: Milnor and Brieskorn theory, Smale and Cerf smoothing theory, blow-up arguments, and the decomposition theorem. The paper is well organized and the main lines of the proofs are credible. However, a few load-bearing justifications are asserted rather than demonstrated; they appear repairable, but they must be addressed before the main claims are fully established.","major_comments":[{"comment":"The conclusion that the singular fibers are PL-manifolds depends on the assertion that the link of every point of the singular locus of the local model {x^2+y^3-u'w=0} in A^5 is a standard PL sphere, justified by [Bri66, Satz 1]. This is not a direct application of that theorem as written: (13) has a one-dimensional singular locus (the v-axis), and the equation is not in Pham-Brieskorn form. The missing steps are (a) reducing the transverse isolated singularity x^2+y^3-u'w=0 in C^4, after the linear change u'w = u^2+v^2 (or u^2-v^2), to the Pham-Brieskorn singularity x^2+y^3+u^2+v^2 with exponents (2,3,2,2), and (b) verifying the Brieskorn sphere criterion for this exponent vector, for instance by computing the characteristic polynomial of the Milnor monodromy and checking chi(1)=±1. Without this verification, item (iii) of Theorem 1 is not established. Please supply the reduction and the check.","section":"Proof of Theorem 1, near Eq. (13)"},{"comment":"In the proof of Proposition 20, the transition from Step 3 to Step 4 asserts that 'every Phi_t preserves F times {1}', and this is used to conclude that Psi' is fiber-preserving on the right vertical face, i.e., g(Psi'(z,1,y)) = (1,y). The pseudo-isotopy group P(F) defined at the beginning of Step 3 consists of diffeomorphisms of F times I fixing F times {0}; Cerf's theorem does not imply that the connecting path can be chosen to preserve the opposite face F times {1}. This matters because the right-face condition is needed to glue ~g_Q to g on the boundary and to obtain a globally defined submersion. Please either prove that the path can be chosen in the subgroup preserving F times {1} (for example by a relative version of Cerf's theorem), or modify the argument by prescribing the top-face trivialization directly in Theorem 21.","section":"Proposition 20, Steps 3 and 4"},{"comment":"The proof of Theorem 4 uses the claim that R^4 tilde h_* Z fails to be locally constant at every point of tilde E lying over the branch locus of p_C: C -> E, and this is asserted without proof. This does not follow from Corollary 37, which only shows that the rational direct image is a trivial local system; the failure is an integral-coefficient phenomenon. Since Proposition 32 is then applied to these integral sheaves to conclude that H_i(tilde X, Z) is not finitely generated, the integral-coefficient statement needs a direct justification, for instance by computing the specialization map F -> F/<sigma_1> on integral cohomology and showing that it is not an isomorphism. Please add this argument.","section":"Section 8.4, proof of Theorem 4"}],"minor_comments":[{"comment":"The cases n=4,5 are deferred to Example 23, but the embedding construction is not spelled out there for those cases; please make the reduction explicit or state precisely how Example 23 supplies the required embedding into a rationally connected fourfold bundle.","section":"Corollary 16"},{"comment":"The claim that P' is smooth along the image of V is used to lift the embedding V -> P' to V -> P; a short local-coordinate verification of this smoothness would make the argument easier to check.","section":"Example 23"},{"comment":"Remark 25 states without proof that the singular fibers admit C-infinity manifold structures agreeing with the algebraic structure outside small neighborhoods; since this remark is not needed for the main theorems, it would be helpful either to indicate the argument or to mark it as a separate open point.","section":"Remark 25"},{"comment":"The statement of Theorem 26 says b_1(X)=2, but this is justified only through the assertion that f induces an isomorphism on fundamental groups; please state that deduction explicitly at the statement or immediately after the construction of X.","section":"Theorem 26 and Corollary 27"}],"recommendation":"major_revision","confidential_remarks":"This is a strong paper that likely deserves publication after revision. The main risk is the PL-sphere justification in the proof of Theorem 1; if the authors cannot supply the missing Brieskorn reduction, the 'looks topologically smooth' conclusion would collapse. I also recommend asking for a rigorous proof of the integral-coefficient local-system failure in Section 8.4 and for a fix of the Cerf step in Proposition 20. All three issues appear local and repairable, so I do not recommend rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe paper to know about: Corrêa, Kollár, Schreieder, and Wang construct explicit morphisms from smooth projective fivefolds to P^1 that have singular fibers but are topologically undetectable: the higher direct images are constant, the map is a homotopy fiber bundle with simply connected fibers, all fibers are PL-manifolds, and the map is homotopic to a C^∞ fiber bundle, yet it is not a C^0 fiber bundle. The same construction gives counterexamples to the FdBK smoothness conjecture, Kollár–Pardon's universal cover question, Kotschick's conjecture, and Schreieder's Aomoto conjecture. This is a genuinely new result and, if correct, resets expectations in equisingularity theory.\n\nWhat I think is strong: The construction is explicit and not over-engineered. The quotient-plus-blow-up strategy is natural, and the proof of the C^∞ smoothing up to homotopy (Proposition 20) is a useful contribution in itself, carefully using Smale's h-cobordism theorem and Cerf's pseudo-isotopy. The applications are clean. I also appreciate that the paper proves a sharp positive result in dimension three (Theorem 40), which calibrates how surprising the counterexamples are.\n\nWhere the soft spots are: The claim that the singular fibers are PL-manifolds (Theorem 1(iii)) rests on a short citation of Brieskorn's Satz 1 in the local model (13), which is a non-isolated singularity. The stress-test is right that the reduction to the isolated transverse singularity and the Brieskorn criterion are not spelled out. I checked the missing step: after the linear change -u'w = u^2+v^2, the transverse singularity is Pham–Brieskorn with exponents (2,3,2,2), whose characteristic polynomial is t^2 - t + 1, so χ(1)=1; the link is a homotopy 5-sphere, hence a standard PL sphere by the generalized Poincaré conjecture in that dimension. So the assertion is true, but a referee should ask for a one-paragraph justification. This is a minor exposition gap, not a load-bearing error.\n\nThe paper is honest with its citations, including refuting a conjecture by one of the authors, and the AI disclosure is a non-issue.\n\nFor peer review: recommend sending to a serious referee. The paper is important and the mathematics appears correct; the referee's main job should be to tighten the proof of item (iii) and to double-check the h-cobordism gluing, not to look for a fatal flaw.\n\nBest,","headline":"A high-quality counterexample paper that settles four open questions in equisingularity theory; the main proof is solid, with one terse spot in the PL-sphere claim that is fixable.","tokens_in":23863,"tokens_out":23347,"would_cite":true,"duration_ms":201506,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14D06","14F45","32Q55","32S15","55R10"],"pacs":[],"model":"deepseek-v4-flash","headline":"Singular fibers of a morphism from a smooth projective variety need not be detectable by the topology of the total space.","keywords":["topology of varieties","vanishing cycle","equisingularity","fiber bundle","smoothness conjecture","universal cover","holomorphic one-form","Aomoto complex"],"falsifier":"Compute the link of the singular locus of $\\{x^2+y^3-u'w=0\\}$ in $\\mathbb{A}^5$ directly and check whether every such link is PL-homeomorphic to a standard sphere; a single non-spherical link would invalidate the claim that all fibers are piecewise-linear manifolds.","tokens_in":22783,"feed_emoji":"","tokens_out":19086,"duration_ms":145149,"temperature":0.7,"pith_summary":"This paper constructs explicit morphisms $g: Y \\to \\mathbb{P}^1$ from smooth complex projective varieties in which some fibers are singular yet the singularity is invisible to the topology of $Y$. The maps satisfy the strongest regularity conditions short of smoothness: the direct images $R^i g_* \\mathbb{Z}$ are constant sheaves, $g$ is a homotopy fiber bundle with simply connected fibers, every fiber is a piecewise-linear manifold, and $g$ is homotopic to a $C^\\infty$ fiber bundle. Despite this, $g$ is not a $C^0$ fiber bundle, so the singular fibers are topologically indistinguishable from smooth ones without being homeomorphic to them. The same construction gives negative answers to four open problems about equisingularity, universal covers, real versus holomorphic one-forms, and the exactness of Aomoto complexes.","feed_headline":"Singular fibers can hide from the topology of a smooth variety","feed_subtitle":"A 5-dimensional construction answers four open questions about when singular fibers must be visible.","key_machinery":"The load-bearing mechanism is a quotient-and-blow-up recipe. One starts with a family $Z \\to \\Delta$ whose Milnor monodromy $T$ satisfies $\\det(T-I) = \\pm 1$, quotients $Z \\times A$ by a diagonal finite group action where $A$ is an abelian variety, and then blows up the resulting family inside a smooth projective fibration with simply connected fibers. The two identities that make this work are the blow-up decomposition $R\\pi_* \\mathbb{Z}_M \\simeq \\mathbb{Z}_M \\oplus \\bigoplus_{r=1}^{c-1} \\iota_* \\mathbb{Z}_Z[-2r]$, which preserves the homology-fiber-bundle property in both directions, and the eigenvalue computation $\\chi_{\\vec c}(1) = \\pm 1$ for the monodromy of a sum of powers, which guarantees that the quotient by the abelian variety does not change the homology of the fibers. The final smoothing step uses the h-cobordism theorem to trivialize the cobordism over a square and a pseudo-isotopy theorem to make the trivialization compatible with the projection, producing the $C^\\infty$ fiber bundle homotopic to $g$.","core_discovery":"The central discovery is a five-dimensional morphism $g: Y \\to \\mathbb{P}^1$, with $Y$ smooth and projective, that has singular fibers but is a topological fiber bundle in almost every sense: the local systems $R^i g_* \\mathbb{Z}$ are constant, all fibers are simply connected and are piecewise-linear manifolds, and $g$ is homotopic to a locally trivial $C^\\infty$ fiber bundle that agrees with $g$ outside a small neighborhood of the singular fibers. Yet $g$ is not a $C^0$ fiber bundle, because the Milnor fiber over a critical value has $H^3 \\cong \\mathbb{Z}^2$, so vanishing cycles appear even though they do not change the homology of the total space. The construction begins with a rational elliptic surface with six type-II singular fibers, forms a diagonal quotient with an elliptic curve under a sixth-order automorphism, and blows up the resulting threefold inside a rationally connected fivefold fibration over $\\mathbb{P}^1$; the blow-up makes the fibers simply connected, upgrading the homology fibration to a homotopy fibration, while the quotient makes the monodromy unimodular so the homology does not jump.","pith_inferences":["The quotient-and-blow-up template may generalize to other starting families and abelian varieties, potentially producing invisible singularities in every dimension at least five, or with singular loci of higher dimension.","Since the counterexample to the 1-form conjecture is rationally connected, the paper leaves open whether the conjecture holds for varieties of general type or with nef canonical class; the same two questions are posed in the paper, and testing them would require new constructions.","The exactness of the Aomoto complex for all finite étale covers makes the constructed variety a natural test space for other proposed topological detectors of holomorphic one-forms without zeros, such as restrictions on the Albanese or on the cohomology ring.","The unproved step in the PL-manifold argument could be checked numerically or conceptually by computing the link of the non-isolated singular locus of the local model $\\{x^2+y^3-u'w=0\\}$; if a non-spherical link exists, the main theorem's PL conclusion would fail even if the other conclusions survive."],"forward_implications":["There are morphisms from smooth projective fivefolds to $\\mathbb{P}^1$ that are homotopy fiber bundles with simply connected fibers but are not $C^0$ fiber bundles, so the singular locus can be invisible to every cohomological or homotopical invariant of the total space.","Taking the fiber product of such a map with an elliptic curve $E \\to \\mathbb{P}^1$ yields a smooth projective fivefold $X$ whose Albanese map is a homotopy fiber bundle and is homotopic to a $C^\\infty$ fiber bundle but is not a submersion; consequently the universal cover of $X$ has the homotopy type of a finite CW complex even though $X$ is not smoothly fibered over $E$.","On that fivefold every real cohomology class in $H^1(X,\\mathbb{R})$ is represented by a closed one-form without zeros, yet the harmonic representative of every such class has a zero and no holomorphic one-form is nowhere zero, giving a negative answer to the 1-form conjecture.","There is a smooth projective fivefold for which every Aomoto complex $(H^*(X',\\mathbb{C}), \\wedge \\omega)$ is exact for every finite étale cover $X' \\to X$ and every nonzero holomorphic one-form $\\omega$, yet every real closed one-form on $X$ has a zero, so exactness of all Aomoto complexes does not force a nowhere-zero holomorphic one-form.","In dimension three, the paper proves the opposite rigidity: a morphism with simply connected fibers from a smooth projective threefold to $\\mathbb{P}^1$ that is a $\\mathbb{Q}$-homology fiber bundle must be smooth, leaving dimension four as the only open case."],"supporting_citations":[{"why":"Supplies the theorem that links of certain isolated hypersurface singularities are topological spheres, used to identify the singular fibers as PL-manifolds.","marker":"[Bri66]"},{"why":"Provides the Milnor fiber topology and the existence of vanishing cycles, used to show the morphism is not a $C^0$ fiber bundle.","marker":"[Mil68]"},{"why":"Supplies the pseudo-isotopy theorem used in Proposition 20 to adjust the h-cobordism trivialization so the smoothing agrees with the original map near the boundary.","marker":"[Cer70]"},{"why":"Supplies the h-cobordism theorem that trivializes the cobordism over the square, producing the locally trivial $C^\\infty$ bundle homotopic to $g$.","marker":"[Sm62]"},{"why":"Gives the eigenvalue formula for monodromy on Milnor fibers of sums of powers, used to verify the condition $\\chi(1) = \\pm 1$ in the quotient construction.","marker":"[Pha65]"},{"why":"Establishes uniqueness of subanalytic triangulations up to PL homeomorphism, used to glue the local PL-manifold structures on the singular fibers.","marker":"[SY84]"},{"why":"Provides the rational cohomology torus construction that underlies the blow-up example disproving the Aomoto-complex conjecture.","marker":"[DJL17]"},{"why":"Supplies the decomposition theorem used to show the derived pushforward of the constant sheaf splits into shifted local systems, a key step for exactness of Aomoto complexes.","marker":"[BBD82]"}],"fun_headline_variants":["Topologically hidden singular fibers refute four conjectures","Vanishing cycles with constant homology: a 5D counterexample","Singular fibers look smooth topologically but aren't a fiber bundle","Morphism hides singular fibers from topology, settling four questions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof that the bad fibers are still nice manifolds relies on the assumption that a certain non-isolated singular point in the local model has a link that is a standard sphere, even though the cited sphere theorem was proved only for isolated singularities.","fun_headline_variants_meta":{"raw":{"variants":["Topologically hidden singular fibers refute four conjectures","Vanishing cycles with constant homology: a 5D counterexample","Singular fibers look smooth topologically but aren't a fiber bundle","Morphism hides singular fibers from topology, settling four questions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000849,"raw_usage":{"total_tokens":3645,"prompt_tokens":850,"completion_tokens":2795,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":466,"completion_tokens_details":{"reasoning_tokens":2725}},"tokens_in":466,"tokens_out":2795,"duration_ms":30722,"temperature":1.0,"reasoning_tokens":2725,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:04:02.334493+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the link of the singular locus of $\\{x^2+y^3-u'w=0\\}$ in $\\mathbb{A}^5$ directly and check whether every such link is PL-homeomorphic to a standard sphere; a single non-spherical link would invalidate the claim that all fibers are piecewise-linear manifolds.","supporting_citations":[],"review_version":1}