{"id":"f64002fb-8bf0-4453-8588-1feb6a498296","arxiv_id":"2501.18132","paper_version":4,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"In complex projective 4-space, the only algebraically skew curves are the rational normal curve and the elliptic normal curves; in 3-space, the only one is the twisted cubic.","lead":"This paper classifies smooth complex curves that can be embedded in projective space so that tangent lines at any two distinct points never intersect, and it proves general bounds on the smallest possible ambient dimension. The main result says that in four-dimensional projective space, only the rational normal curve and the elliptic normal curves are algebraically skew, a notion that connects projective geometry with the classical skew-loop problem in differential topology.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central enumeration Lemma 4.11 remains unverified; the false Lemma 3.1(ii) is real but the specific P3 identity it purports to justify is actually correct, so the load-bearing risk is the P4 count.","rationale":"After going through the blowup computations, I find the reader's suspicion about Lemma 3.1(ii) partially justified but mislocated. The lemma as printed is dimensionally wrong and false; however, the pushforward identity used in Section 3.11 for σ21 in Gr(2,4) is correct: using the diagonal class [Δ]=Σ σ_a⊗σ_a^* and multiplying by σ21⊗1 yields exactly σ21⊗σ22+σ22⊗σ21. So Theorem 3.15 is not necessarily invalidated. The P4 classification depends instead on Lemma 4.11. That formula is derived through a long chain of blowup and excess-intersection calculations with several OCR/typing slips; no full independent derivation is given, and the Macaulay2 verification covers only (d,g)=(8,5). Because the classification theorem is an if-and-only-if statement, a single erroneous coefficient would change the allowed pairs (g,d) and hence the list of curves. The author notes Ciliberto independently obtained Lemma 4.11, which is encouraging, but that external statement is not checked in the preprint. I therefore recommend keeping the CONDITIONAL verdict; the proposed test on rational curves would settle whether the enumeration formula is actually correct.","tokens_in":32214,"tokens_out":30837,"duration_ms":289164,"concrete_test":"Run a Macaulay2/Schubert2 script on a smooth rational quintic curve in P4: compute the number of pairs of distinct points whose tangent lines intersect. Lemma 4.11 predicts d∨=2·5+0-2=8, so count = 8^2 - 10·8 + 24 = 8. If the computed number is not 8, the enumeration formula and Theorem 4.13 fail. A second data point at degree 6 genus 0 predicts count = 10^2 - 10·10 + 24 = 24; matching both would substantially certify the formula.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim reduces to Lemma 4.11: the number of nonskew tangent-line pairs of a degree-d genus-g curve in P4 is (2d+2g-2)^2 - 20d - 44g + 44. If this count is wrong, Theorem 4.13 collapses. The written proof of that count is a long unverified Chow-ring computation. The reader flagged Lemma 3.1(ii) as the weak link; that lemma is indeed false as stated (for G=P1 it predicts i_*h=h⊗1+1⊗h rather than h⊗h, and its RHS has codim |p| rather than |p|+dim G). But the specific identity the author uses in Lemma 3.11, i_*σ21 = σ21⊗σ22 + σ22⊗σ21 in Gr(2,4), is actually correct when derived from the diagonal class, so the P3 genus-zero theorem may survive. Thus the real load-bearing concern is not that one lemma but whether the elaborate P4 computation leading to Lemma 4.11 has no hidden sign or coefficient error. The only reported check is the single case (d,g)=(8,5), which is insufficient support for a classification statement.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper defines algebraically skew embeddings of smooth projective varieties and studies the minimal skew embedding dimension msdim X. It proves general bounds 3n ≤ msdim X ≤ 4n+1, then specializes to curves. The main technical tool is a blowup of G×G along the diagonal and an excess-intersection computation on the proper transforms of the degeneracy locus D1 and of Γ = γ(X)×γ(X). The claimed results are that the only algebraically skew curve in P3 is the twisted cubic (Theorem 3.15), and that the only algebraically skew curves in P4 are the rational normal curve and the elliptic normal curves (Theorem 4.13). The paper also applies the method to scrolls and includes a Macaulay2 verification of one numerical case, Example 4.12.","tokens_in":32420,"tokens_out":28753,"duration_ms":295485,"significance":"If the central enumeration Lemma 4.11 is correct, the classification of algebraically skew curves in P3 and P4 is a clean and natural result, and the blowup/excess-intersection approach offers a genuinely different method from the Terracini-locus and Porteous computations of [Cil24]. The paper is self-contained, has no fitted parameters, and provides an independent machine check in Example 4.12; the general bounds in Section 2 are also a useful contribution. However, the written derivation contains a demonstrably false lemma in the intersection calculus and several unclear steps in the classification argument, so the main classification is not yet established as written.","major_comments":[{"comment":"Lemma 3.1(ii) is false as stated. Since the diagonal inclusion i: ΔG → G×G has codimension dim G, the pushforward i_*(σ_p) must land in codimension |p| + dim G, not |p|. For G = Gr(2,4) the displayed rule would place i_*(σ_{21}) in codimension 3, whereas the correct class has codimension 7; the identity actually used in Lemma 3.9 and Lemma 3.11, i_*(σ_{21}) = σ_{21}⊗σ_{22} + σ_{22}⊗σ_{21}, has total codimension 7. The proof given for the lemma confuses the condition σ_aσ_b = σ_p with the Künneth decomposition of the diagonal class. Because the same pushforward rule is used in the blowup ring structure underlying the P4 computation leading to Lemma 4.11, this is a load-bearing error and must be replaced by the correct diagonal-class computation.","section":"§3.1, Lemma 3.1(ii)"},{"comment":"The dual-degree formula d∨ = 2d + 2g − 2 is a load-bearing input, used in Lemmas 3.9, 4.2, and 4.11, but the proof is not acceptable as written: it cites [GKZ94, p.61] and simultaneously asserts that the formula printed there is a typo, without reproducing the correction. Moreover, the proof writes the point class of Gr(1,N) as σ_{22}, which is correct only for N = 3; for N = 4 the point class is σ_{33}. Please supply a direct proof of the dual-degree formula (for instance via the Gauss map or Plücker-type class formulas) or a correct, unambiguous citation.","section":"§3.4, Lemma 3.10"},{"comment":"The final classification contains a scrambled and insufficient exclusion of impossible pairs. After the algebra, the author obtains (g,d) ∈ {(0,4),(1,5),(2,5),(5,4)}. The next sentence says that 'the cases (5,2),(4,5) are in fact not possible', which does not match this list; presumably the intended exclusions are (g,d) = (2,5) and (g,d) = (5,4). Even with that correction, the appeal to [GH94, p.253] is not a proof: the nonexistence of nondegenerate smooth genus-2 degree-5 and genus-5 degree-4 curves in P4 should be shown directly, for example by Riemann–Roch and Clifford's theorem. The derivation of the bounds 6 ≤ d∨ ≤ 16 and d ≤ 5 should also be written out explicitly. As it stands, Theorem 4.13 is not fully proven.","section":"§4.8, Theorem 4.13"},{"comment":"The numerical heart of the paper is Lemma 4.11, which converts the Chow-ring computation into the count (2d+2g−2)^2 − 20d − 44g + 44. The proof is a long sequence of blowup and Chern-class calculations whose only reported check is the single case (d,g) = (8,5) in Example 4.12. Given that Lemma 3.1(ii) is false and that the later classification step is scrambled, this computation is not verifiable as written. In addition, the step from the excess-intersection identity to 'the number of nonskew pairs' implicitly assumes that all off-diagonal contributions are reduced, transverse points; this should be stated and justified. I ask for a fully displayed calculation, cross-checked against [Cil24] or verified in additional examples, before the classification can be accepted.","section":"§4.3–4.8, Lemma 4.11"}],"minor_comments":[{"comment":"There are several typos: 'sekw' should be 'skew', and the abstract contains 'minim al'.","section":"§1, Definition 1.2"},{"comment":"The sentence 'Since there is no elliptic curves of degree ≤ 2 in P4, the case of g = 1 is impossible' is unclear: the preceding equation gives d∨ = 5 for g = 1, not d∨ ≤ 2, and an elliptic scroll of degree 5 in P4 is not obviously ruled out. Please rewrite the exclusion.","section":"§4.2, Theorem 4.2"},{"comment":"The angle-bracket notation for numerical classes, written with '/llbracket' and '/rrbracket', is rendered inconsistently and is not defined cleanly; please define it once and use it uniformly.","section":"§2.1, Notation"},{"comment":"In the last sentence of the proof, 'To concludes the proof' should be 'To conclude the proof'.","section":"§3.4, Lemma 3.10"}],"recommendation":"major_revision","confidential_remarks":"The author acknowledges in the introduction that Lemma 4.11 coincides with a result of Ciliberto. This makes the manuscript's contribution an independent method rather than a new classification per se, so the burden of a fully correct computational proof is especially high. The false Lemma 3.1(ii) and the opaque exclusion in Theorem 4.13 are the main obstacles; both appear fixable within the manuscript's scope, but the central computation must be reworked or cross-verified before the paper can be accepted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Hi — short take: this is a real research paper with a plausible and interesting classification and a genuinely new technique, but the key P4 enumeration is not yet demonstrated as written, and one stated lemma is false. I would send it to a referee; I would not yet rely on the P4 theorem.\n\nWhat is new: the 3n lower bound for algebraically skew embeddings, the blowup-of-Gauss-images method, and the classifications — twisted cubic in P3, rational normal and elliptic normal curves in P4. The paper is also straight about the overlap with Ciliberto: Lemma 4.11 is the same count as the Terracini-locus computation in [Cil24], and the author reports a Macaulay2 check for one genus-5 canonical curve giving 240. That is real evidence and the right level of honesty.\n\nThe soft spots are equally real. Lemma 3.1(ii), as stated, cannot be right: diagonal pushforwards have codimension |p| + dim G, not |p|, and the displayed sum is dimensionally inconsistent. The paper then uses this lemma in the P3 computation. However, the particular pushforward actually used for Gr(2,4), σ21 ↦ σ21⊗σ22 + σ22⊗σ21, is correct and can be checked directly, so the P3 theorem may survive a repaired lemma. The bigger problem is the P4 count: Lemma 4.11 is the output of a long Chow/excess-intersection computation, and the only independent check reported is a single (d,g)=(8,5) example. One Macaulay2 point is not enough to support a classification statement. Theorem 4.13 also has a scrambled and under-explained exclusion of the remaining degree-genus pairs; the GH page reference is a pointer, not a proof of nonexistence. Lemma 3.10's GKZ quote is sloppy, though the formula itself is standard.\n\nIf Ciliberto's independent computation is sound, the enumeration is probably right; if it is not, the paper's central theorem has no adequate verification yet. The author should repair Lemma 3.1, expand the case exclusion in Theorem 4.13, and either get an independent check of Lemma 4.11 or write out the full computation in enough detail to be checked. A referee with Terracini-loci and Schubert calculus background can do that. Worth sending out.","headline":"A plausible classification of skew curves in P3 and P4 with a newly developed blowup method, but the load-bearing P4 count is under-verified and one stated lemma is false; it deserves serious refereeing.","tokens_in":32999,"tokens_out":7347,"would_cite":true,"duration_ms":80845,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14H50","14N05","14N10","14C17"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that in P4 the only algebraically skew curves are the rational normal curve and the elliptic normal curves, and that in P3 the only such curve is the twisted cubic.","keywords":["algebraically skew embedding","skew curves","Terracini loci","Gauss map","excess intersection","blowup","rational normal curve","elliptic normal curves"],"falsifier":"Compute the number of non-skew ordered pairs of tangent lines for the canonical genus 5 curve in P4 by an independent method: the paper predicts 240. Alternatively, test Lemma 3.1(ii) on Gr(2,4): the class i_*(σ21) must lie in codimension 7, while the stated formula places it in codimension 3, so the formula as written is false and the count built on it needs correction or a new justification.","tokens_in":31966,"feed_emoji":"📐","tokens_out":6315,"duration_ms":61731,"temperature":0.7,"pith_summary":"This paper studies algebraically skew embeddings: smooth complex projective varieties whose embedded tangent spaces at any two distinct points are disjoint, with an extra condition ruling out infinitesimal tangencies. It proves that any n-dimensional variety admits such an embedding into $P^{{4n+1}}$ and none into a projective space of dimension below 3n. For curves the classification is complete: in P3 the only algebraically skew curve is the twisted cubic, and in P4 the only ones, up to linear equivalence, are the rational normal curve and the elliptic normal curves. A sympathetic reader should care because this settles the projective analogue of the classical skew-loop question for curves in the two lowest nontrivial ambient dimensions, and the counting formula it produces gives a concrete enumerative invariant for every degree and genus.","feed_headline":"Only two curve families in P4 have pairwise skew tangents","feed_subtitle":"A projective geometry proof classifies the twisted cubic and elliptic normal curves as the sole algebraically skew curves.","key_machinery":"The main tool is the degeneracy locus D_1 = {(U,W) : dim(U∩W) ≥ 1} inside the product Gr(n,N)×Gr(n,N), together with the product Γ = C×C of a curve's Gauss image with itself; skewness is exactly the statement that D_1 meets Γ only along the diagonal. Because D_1 is singular along the diagonal, the paper blows up the diagonal of the Grassmannian product, writes the Chow-ring classes of the proper transforms, and for P4 performs a second blowup along the diagonal curve to resolve tangency. The auxiliary incidence variety \\hat D_1 ≅ P(Q) ×_{P4} P(Q) (with Q the universal quotient bundle on P4) makes the Chow ring of the blown-up degeneracy locus computable, and the symmetric excess-intersection formula converts the class computation into the numerical count of non-skew pairs.","core_discovery":"The central claim is that algebraic skewness of a curve can be detected by an excess-intersection count of its tangent lines. For a smooth degree d genus g curve in P4 whose Gauss map is an isomorphism onto its image, the number of ordered pairs of distinct points whose embedded tangent lines meet is (2d+2g−2)^2 − 20d − 44g + 44, counting degenerate third osculating planes. Algebraic skewness means this number is zero, and solving the resulting equation with the constraints on degree and genus leaves exactly the rational normal curve (d=4, g=0) and the elliptic normal curves (d=5, g=1) in P4, together with the twisted cubic in P3. The proof also establishes the general bounds 3n ≤ msdim X ≤ 4n+1 for an n-dimensional smooth projective variety X.","pith_inferences":["If Lemma 3.1(ii) is corrected rather than discarded, the classification may survive, but the enumerative constant in Lemma 4.11 would likely change; the paper's numerical predictions are the place to test the repair.","The same double-blowup strategy should apply to curves in P5 and beyond, where the relevant degeneracy loci are D_r with r ≥ 1 and the expected answers may be families rather than single curves.","The coincidence of the genus 5 count with a Steiner complex cardinality suggests a hidden theta-characteristic interpretation of non-skew tangent pairs for canonical curves, which the paper leaves open.","The real totally skew surface constructed from an affine elliptic normal curve in R8 may be a test case for whether the algebraic classification has smooth counterparts in low-dimensional real Euclidean spaces."],"forward_implications":["In P3, a smooth spatial curve with pairwise disjoint embedded tangent lines must be the twisted cubic; no other degree or genus is possible.","In P4, algebraic skewness forces either the degree 4 rational normal curve or a degree 5 elliptic normal curve, up to linear equivalence.","Every n-dimensional smooth projective variety has an algebraically skew embedding into P^{4n+1}, and none exists in P^N with N < 3n.","For any smooth degree d genus g curve in P4 with nondegenerate Gauss map, the number of non-skew ordered pairs of tangent lines is (2d+2g−2)^2 − 20d − 44g + 44; for the canonical genus 5 curve this is 240.","Generic skew scrolls in P3 and P4 must be rational normal scrolls, so non-rational scrolls always contain intersecting rulings."],"supporting_citations":[{"why":"Introduces Terracini loci and establishes skewness of high-degree rational curves, giving the upper-bound side.","marker":"[BC21]"},{"why":"Studies Terracini loci of curves in P3 and canonical curves, providing the comparison for the P3 classification.","marker":"[BC23]"},{"why":"Independently counts Terracini schemes for curves in P4, the same enumerative quantity as Lemma 4.11.","marker":"[Cil24]"},{"why":"Supplies the blowup Chow-ring calculus, Porteous formula, and excess-intersection formula used throughout.","marker":"[EH16]"},{"why":"Provides the degeneracy-locus formula for the Schubert class of D1 and the general intersection-theoretic framework.","marker":"[Ful98]"},{"why":"Gives the degree d∨ = 2d + 2g − 2 of the dual curve, the key substitution in the counting formula.","marker":"[GKZ94]"},{"why":"Supplies the first Chern class formulas for projective bundles and the classification facts excluding impossible degrees and genera in P4.","marker":"[GH94]"}],"fun_headline_variants":["Skew tangents: only twisted cubic and elliptic normal curves qualify","Curves with disjoint tangent lines: just two families exist","Algebraically skew curves: rational normal and elliptic normal only","Only two curve families have pairwise skew tangents in P4","Twisted cubic and elliptic normal: the only skew-tangent curves"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing step is the paper's formula for pushing a Schubert class forward along the diagonal inclusion of a Grassmannian; as stated that formula gives a class of the wrong codimension in a case used later, and every subsequent intersection count inherits it.","fun_headline_variants_meta":{"raw":{"variants":["Skew tangents: only twisted cubic and elliptic normal curves qualify","Curves with disjoint tangent lines: just two families exist","Algebraically skew curves: rational normal and elliptic normal only","Only two curve families have pairwise skew tangents in P4","Twisted cubic and elliptic normal: the only skew-tangent curves"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000194,"raw_usage":{"total_tokens":1294,"prompt_tokens":829,"completion_tokens":465,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":445,"completion_tokens_details":{"reasoning_tokens":378}},"tokens_in":445,"tokens_out":465,"duration_ms":5698,"temperature":1.0,"reasoning_tokens":378,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T00:37:17.177885+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the number of non-skew ordered pairs of tangent lines for the canonical genus 5 curve in P4 by an independent method: the paper predicts 240. Alternatively, test Lemma 3.1(ii) on Gr(2,4): the class i_*(σ21) must lie in codimension 7, while the stated formula places it in codimension 3, so the formula as written is false and the count built on it needs correction or a new justification.","supporting_citations":[],"review_version":1}