{"id":"eddcb285-6bac-44de-b11d-34bce4be7970","arxiv_id":"2507.13927","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A general Fano hypersurface of degree d in P^n has rational curves of degree e with balanced restricted tangent bundle exactly when e exceeds (n-1)/(n+1-d), and quadrics have a separate parity rule.","lead":"This paper finds the exact condition on degree e, degree d, and dimension n under which a general Fano hypersurface contains a rational curve with balanced restricted tangent bundle. The result completes a classification that was previously known only for special degrees, and it also gives explicit example hypersurfaces for low degrees.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 7.2's all-degrees conclusion relies on an unproved existence of a degree n−1 curve with perfectly balanced restricted tangent bundle; Theorems 5.1–7.1 only produce balanced, not perfect, splittings for d≥5.","rationale":"The reader's weakest-assumption analysis correctly identified the terse rank checks for the kernel matrices KF, and those checks are indeed essential for Theorems 5.1, 6.1, and 7.1. However, the present stress-test finds an additional, more structurally load-bearing gap: the perfectly balanced degree n−1 curve used as the generator in the gluing argument of Theorem 7.2 is never constructed or cited. This is not a mere omitted computation; it is an unproved existence assertion on which the extension to all degrees e > max{2d−2, n} depends. Theorems 3.8 and 3.10 cover only e ≤ 2d−2, and Theorem 7.1 concerns e=n; neither yields a perfectly balanced bundle of degree n−1 for general d≥5. The gap can likely be filled by an induction step analogous to Proposition 3.2, but as written the proof is incomplete. Because there is no evidence of falsehood, the appropriate verdict remains conditional rather than reject; the condition should now include verifying the perfect balanced base curve, in addition to the rank checks. The paper's main geometric framework and explicit computations for d=2,3,4 are coherent and credible, so the concern is confined to the final gluing mechanism and its missing premise.","tokens_in":32517,"tokens_out":14233,"duration_ms":161043,"concrete_test":"For the first nontrivial unproved case, take d=5 and n=6. Construct a general quintic hypersurface X⊂P^6 containing the degree-5 rational normal curve C, using the polynomial F chosen in the pattern of Theorem 7.1, and compute the splitting of T_X|C by the kernel-matrix method of Section 2.4 (this can be done in Macaulay2, as the author suggests). If the result is T_X|C ≅ O(2)^5, the perfect balanced base curve required by Theorem 7.2 exists in this case; if any summand has degree different from 2, the gluing step fails and Theorem 1.1(2) is not established for large e. An independent check would also be to compare with the splitting obtained from the induction in Proposition 3.2 starting from the e=n case in Theorem 7.1, verifying that the extension class produces a perfectly balanced bundle for e=n−1.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The final step of the proof of Theorem 1.1(2) is Theorem 7.2, which glues a curve C1 of degree n−1 with perfectly balanced restricted tangent bundle T_X|C1 ≅ O(n+1−d)^{n−1} to balanced curves C2 of degrees e in an interval of length n−1, thereby obtaining all larger degrees. This C1 is asserted with no proof and no citation. It is not supplied by Theorems 5.1, 6.1, or 7.1: for d=5, Theorem 7.1's e=n splitting is O(3)^2 ⊕ O(2)^3 for n=6, which is balanced but not perfectly balanced, and no splitting is given for e=n−1=5. For general d≥5 and n large, e=n−1 can lie outside the ranges covered by Theorems 3.8 and 3.10 (e.g., d=5, n=10 has e=9 > 8 = 2d−2). Lemma 2.14, as stated, requires one component to be perfectly balanced in order to preserve balancedness under gluing, so the degree n−1 perfect curve is genuinely load-bearing: without it, the gluing construction does not generate all degrees e > max{2d−2, n}, and the full classification claimed in Theorem 1.1(2) is unsupported. This is a logical gap in the argument, distinct from the rank-check verifications flagged by the reader, and it is the single most fragile link in the proof of the central claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the splitting type of the restricted tangent bundle T_X|C for rational curves C of degree e on a smooth degree d hypersurface X ⊂ P^n. The main theorem (Theorem 1.1) asserts that for 3 ≤ d ≤ n the bundle is never balanced when e ≤ (n−1)/(n+1−d), and that a general X contains degree-e rational curves with balanced T_X|C for every e above that threshold; quadrics are treated separately in Theorem 1.2. The proof combines the Euler and tangent-bundle exact sequences for rational normal curves, explicit kernel matrices for the maps δ_F, an induction in the ambient dimension n (Proposition 3.2), known balanced normal-bundle results, and a gluing/specialization lemma (Lemma 2.14).","tokens_in":32792,"tokens_out":20105,"duration_ms":228813,"significance":"If the main theorem is correct, it gives a complete numerical classification for balanced restricted tangent bundles on general Fano hypersurfaces, complementing recent results on normal bundles and addressing a question raised by Ran in the rational-curve case. The paper is constructive: for d = 2, 3, 4 and for e = n ≥ 2d−2 it produces explicit polynomials defining hypersurfaces with prescribed balanced splitting, and it connects the splitting to modular interpolation. The exposition is careful about exact sequences and the induction framework. However, the proof of the final gluing theorem contains a genuine gap in the diagonal case n = d, and several linear-algebra rank checks are asserted rather than proved; these issues currently prevent acceptance.","major_comments":[{"comment":"The proof asserts the existence of a degree n−1 rational curve C1 with perfectly balanced restricted tangent bundle T_X|C1 ≅ O(n+1−d)^{n−1}. This is false for n = d, the diagonal case included in Theorem 7.2 and in Theorem 1.1(2): Proposition 2.13 gives the obstruction threshold (n−1)/(n+1−d) = n−1, so a degree n−1 curve cannot have balanced restricted tangent bundle at all. For n > d the assertion is true and follows from the balanced curves already produced in the preceding interval together with the fact that for e = n−1 the slope n+1−d is an integer; however, this implication is not stated. The gluing step is load-bearing for the 'every e > ...' claim when n = d, so the proof of Theorem 1.1(2) is incomplete in that case. Please supply a correct argument for n = d or adjust the scope of the statement.","section":"§7.2 (proof of Theorem 7.2)"},{"comment":"The splitting type T_X|C ∼= O(n+2−d)^{n−d+1} ⊕ O(n+1−d)^{d−2} is obtained by showing that the explicit matrix K_F is injective. The final step of the proof, after the Gauss-Jordan reduction, reads: 'This can be shown by using the diagonal of 1's and induction.' No induction is exhibited and no determinant computation for the reduced (d+1)×(d+1) block is given. Since Theorem 7.1 supplies the existence part for every e ≥ 2d−2 when n is large, this is a genuinely load-bearing linear-algebra check. The same pattern occurs in Theorems 5.1 and 6.1, where the rank assertions are delegated to 'Gauss-Jordan elimination' or 'computing minors' without a complete argument. Please provide a complete proof, for example a general lemma showing that the relevant minors are nonzero, or a reproducible computer verification.","section":"§7.1 (proof of Theorem 7.1)"},{"comment":"The sentence 'In both cases, we have 1 ≤ μ(N_C/X) ≤ 3' is not true for arbitrary n when e = d+1; for instance, d = 3, e = 4, n = 10 gives μ(N_C/X) = 30/8 = 3.75. The intended argument appears to be that the inequality holds in the base case n = e and that Proposition 3.2 provides the induction for larger n, but as written the proof seems to apply Corollary 2.11 outside its range. Please rewrite this step so that the base case and the induction are explicit.","section":"§3 (proof of Theorem 3.8)"}],"minor_comments":[{"comment":"The displayed relation '−t·C_i + s·C_{e+1}' is likely a typo; it should presumably involve C_{i+1} rather than C_{e+1}.","section":"§4.7 (Theorem 4.7, odd case)"},{"comment":"The phrase 'floor threshold' is used informally; please define it as ⌊(n−1)/(n+1−d)⌋ at first use.","section":"§7.2"},{"comment":"Several displayed matrices contain formatting artifacts, such as 's 3' and 's 2' in place of s^3 and s^2, and the large matrices in Theorem 5.1 and Theorem 7.1 would benefit from a careful proofreading pass.","section":"§5.1 and §7.1"},{"comment":"The proof says 'Since B has summands of degree larger than a and r ≥ 1, K contains a column of degree at least one, which can be chosen as the first column of N · J1'; this is not fully precise, and the decomposition of entries would be clearer if written out as in Lemma 3.6.","section":"§3.5 (Lemma 3.5)"}],"recommendation":"major_revision","confidential_remarks":"The diagonal case n = d in Theorem 7.2 is the crux of the remaining difficulty; if a correct argument can be supplied, the paper would be publishable. I would also encourage the author to include a Macaulay2 script or an ancillary file verifying the kernel-matrix rank checks for the displayed families, since those checks are long and are central to the proof."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real advance, but the proof of the all-degrees statement is incomplete in the boundary case d = n.\n\nWhat's new: Theorem 1.1 gives the full threshold for d < n, sharpening Ran's existence in arithmetic progressions, and Theorem 1.2's odd-degree quadric obstruction is genuinely new. The inductive Proposition 3.2 is a useful tool, and the explicit splitting tables for d = 2,3,4 are concrete and checkable.\n\nWhere it's good: the exact sequences and induction are coherent, the interpolation interpretation is clearly explained, and the paper is honest about which parts are explicit and which are left as “Gauss-Jordan elimination.”\n\nSoft spots, in proportion:\n\n1. Minor: the rank checks for K_F in Theorems 5.1, 6.1, and 7.1 are summarized rather than demonstrated. Since the splitting conclusions are literally read off those matrices, a referee will need to verify a lot of linear algebra. This is fillable but real.\n\n2. More serious but fixable: Theorem 7.2 asserts the existence of a degree n−1 curve with perfectly balanced restricted tangent bundle, with no proof. For d < n this is actually a corollary of the earlier balanced results: the slope e(n+1−d)/(n−1) at e = n−1 equals the integer n+1−d, so any balanced bundle there is automatically perfectly balanced. The paper never says this, so the reader is left to reconstruct the argument.\n\n3. The genuine gap: for d = n, that degree n−1 curve is obstructed by the paper's own Proposition 2.13 (slope 1, so no balanced T_X). Thus the gluing step in Theorem 7.2 cannot be run for d = n, and Theorem 1.1 is unproved for hypersurfaces of degree equal to the ambient dimension, for degrees e > 2n−2. This is load-bearing, not routine verification.\n\nOverall: for 3 ≤ d < n the paper is strong and likely correct. The d = n case needs either a new argument or a corrected statement. It deserves a careful referee, who should be asked to check the generation step and the matrix ranks. I'd bring it to a reading group because the threshold phenomenon is interesting and the quadric parity is a nice counterpoint.","headline":"Strong classification of balanced restricted tangent bundles for d < n, with a genuine gap in the generation step for d = n and terse matrix-rank checks.","tokens_in":33390,"tokens_out":9946,"would_cite":true,"duration_ms":102869,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14H60","14J45","14J70","14G17","14N25","14Q15"],"pacs":[],"model":"deepseek-v4-flash","headline":"For smooth Fano hypersurfaces, a rational curve of degree $e$ has balanced restricted tangent bundle exactly when $e$ exceeds $(n-1)/(n+1-d)$, with quadrics split by parity.","keywords":["rational curves","restricted tangent bundles","hypersurfaces","curve interpolation","balanced vector bundles","splitting type","Birkhoff-Grothendieck decomposition","rational normal curves"],"falsifier":"Take an allowed triple from Theorem 7.1, for instance $d=5$ and $n=10$, write the displayed matrix $K_F$ with $t=1$, and compute its rank symbolically at $s=1$; if the rank comes out below $n-1=9$, the claimed balanced splitting $T_X|_C \\cong O(7)^6\\oplus O(6)^3$ is false. The same computation can be repeated for the matrices displayed in Theorems 5.1 and 6.1 at their stated parameter values.","tokens_in":32251,"feed_emoji":"📐","tokens_out":12929,"duration_ms":131205,"temperature":0.7,"pith_summary":"This paper determines exactly when a general smooth Fano hypersurface $X\\subset\\mathbb{P}^n$ of degree $d$ ($3\\le d\\le n$) contains rational curves whose restricted tangent bundle $T_X|_C$ is balanced, meaning it splits as a direct sum of line bundles whose degrees differ by at most one. The result is a clean threshold: for degree-$e$ rational curves, $T_X|_C$ is never balanced when $e \\le (n-1)/(n+1-d)$, and a general $X$ contains balanced examples for every $e > (n-1)/(n+1-d)$. Since a balanced restricted tangent bundle is exactly what lets a curve interpolate the maximum possible number of points, the theorem pinpoints, for every Fano hypersurface, the curve degrees at which maximal interpolation begins. Quadrics form the one parity exception: only even-degree rational curves are balanced there, with odd-degree curves one step short, and the paper also produces explicit hypersurfaces realizing each splitting type for rational normal curves.","feed_headline":"Balanced tangent bundles appear exactly above a simple degree ratio","feed_subtitle":"For degree-d hypersurfaces in P^n: balanced splitting and maximal interpolation start when e > (n-1)/(n+1-d).","key_machinery":"The carrying object is the explicit kernel matrix $K_F$ of the map $\\delta_F=\\psi_F\\circ\\beta: O(e+1)^e\\oplus O(e)^{n-e}\\to O(de)$, whose cokernel is $T_X|_C$. After fixing a degree-$d$ polynomial $F$ defining the hypersurface, the normal-bundle map $\\psi_F$ is computed from the quadratic generators of the ideal of a rational normal curve $C$, and $\\beta$ is the fixed quotient map coming from the Euler sequence. The kernel is written as a matrix whose columns are explicit column relations of $\\delta_F$; proving $K_F$ has maximal rank at every point of $\\mathbb{P}^1$ identifies $T_X|_C$ and its splitting type. An extension proposition then lifts a balanced hypersurface $Y\\subset\\mathbb{P}^{n-1}$ to a balanced $X\\subset\\mathbb{P}^n$ by realizing any extension of $O(e)$ by $T_Y|_C$, using explicit factorization matrices $J_0,J_1,J_2$, and a gluing lemma for vector bundles on trees of rational curves extends the result to all degrees above the threshold.","core_discovery":"The central claim is Theorem 1.1: for a smooth Fano hypersurface $X\\subset\\mathbb{P}^n$ of degree $d$ with $3\\le d\\le n$, the restricted tangent bundle $T_X|_C$ of any degree-$e$ rational curve $C$ is never balanced when $e \\le (n-1)/(n+1-d)$, and a general hypersurface contains rational curves of degree $e$ with balanced $T_X|_C$ for every $e > (n-1)/(n+1-d)$. The paper works over an algebraically closed field of characteristic not dividing $e$. Quadrics are classified separately: for every even $e\\ge2$ there are degree-$e$ rational curves with $T_X|_C \\cong O(e)^{n-1}$, while odd-degree curves always have the unbalanced splitting $O(e-1)\\oplus O(e)^{n-3}\\oplus O(e+1)$. For rational normal curves of degree $e\\le n$, explicit splitting types are listed for $d=2,3,4$, and for general $d$ balanced splitting is exhibited when $e\\ge 2d-2$.","pith_inferences":["An implicit consequence is that the same kernel-matrix formalism could decide balancedness for curves that are not rational normal curves, since the open condition of balancedness would propagate from the explicit examples to general deformations in the same Hilbert scheme.","A testable extension is to compute, for each $(d,n)$, the maximal minors of $K_F$ symbolically; this would supply an independent certificate of the asserted splitting types without relying on the terse Gauss-Jordan descriptions in the paper.","The quadric parity phenomenon is likely a shadow of a general divisibility rule for homogeneous spaces, as in the Grassmannian case the paper cites; the ruled-surface reduction in the quadric proof gives a model for proving such parity obstructions elsewhere.","One could probe stability of the threshold by asking whether $e=(n-1)/(n+1-d)$ also governs interpolation for Fano complete intersections of higher codimension, where the index, not the hypersurface degree, should play the analogous role."],"forward_implications":["The classification of triples $(e,d,n)$ for which a general Fano hypersurface contains a balanced restricted tangent bundle is complete for $d\\ge3$: the only obstruction is the slope bound $e \\le (n-1)/(n+1-d)$.","For every $e>(n-1)/(n+1-d)$, a general hypersurface contains rational curves that interpolate $\\lfloor e(n+1-d)/(n-1)\\rfloor+1$ general points, the maximum possible from the slope of $T_X|_C$.","On quadrics, even-degree curves reach the perfectly balanced splitting $O(e)^{n-1}$, while odd-degree curves stop one step short; consequently deformations of odd-degree curves interpolate $e$ points rather than $e+1$.","The extension proposition converts any explicit balanced example into a family of examples in all higher ambient dimensions, and the gluing lemma converts the finite interval $(n-1)/(n+1-d) < e \\le (n-1)/(n+1-d)+(n-1)$ into every larger degree by adding degree $n-1$ curves.","Explicit polynomials $F$ are provided for rational normal curves with $e\\le n$ when $d\\le4$, and for $e\\ge 2d-2$ in general degree $d$, so the claimed splitting types are checkable examples rather than existence statements alone."],"supporting_citations":[{"why":"supplies the normal-bundle description of rational normal curves and the surjectivity result that lets every map between the relevant bundles be induced by some hypersurface, feeding the construction of $\\delta_F$ and the balanced normal-bundle input.","marker":"[CR19]"},{"why":"supplies the splitting of the restricted tangent bundle of projective space along a rational normal curve and the quotient map from the Euler sequence used to write $\\delta_F=\\psi_F\\circ\\beta$.","marker":"[CR18]"},{"why":"defines modular interpolation for curves on Fano hypersurfaces, proves balanced restricted tangent bundles in certain arithmetic progressions, and supplies a theorem used to get balanced normal bundles for the degrees $d\\le e\\le 2d-2$.","marker":"[Ran24a]"},{"why":"provides the ruled-surface construction for quadrics that underlies the parity reduction showing odd-degree curves cannot interpolate the expected number of points.","marker":"[Kol18]"},{"why":"computes normal bundles of rational normal curves on hypersurfaces and provides the balanced normal-bundle examples that, with the splitting criterion, give the base cases for degrees $d=2,3,4$.","marker":"[Mio25]"},{"why":"gives the specialization theorem for vector bundles on trees of rational curves used in Lemma 2.14 to glue balanced curves into balanced curves of higher degree.","marker":"[Smi23]"},{"why":"supplies the criterion for specialization of vector bundles on $\\mathbb{P}^1$, which makes balancedness an open condition and turns explicit examples into statements about general hypersurfaces.","marker":"[EH16]"},{"why":"shows the restricted tangent bundle of a general rational curve in $\\mathbb{P}^d$ is balanced, an input for the induction on ambient dimension over intermediate degree ranges.","marker":"[Ram90]"}],"fun_headline_variants":["A sharp ratio decides when rational curves get balanced tangents","For quadrics only even-degree curves have balanced tangents","Explicit balanced tangents for rational normal curves","Sharp threshold for balanced tangent splitting on hypersurfaces"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The classification rests on the claim that each explicitly written kernel matrix $K_F$ has full rank at every point of the parameter line $\\mathbb{P}^1$; the paper justifies this with terse Gauss-Jordan descriptions and, for $d\\ge4$, with an induction that is not written out, so a single unverified rank check failing at some $(d,e,n)$ would change the claimed splitting type.","fun_headline_variants_meta":{"raw":{"variants":["A sharp ratio decides when rational curves get balanced tangents","For quadrics only even-degree curves have balanced tangents","Explicit balanced tangents for rational normal curves","Sharp threshold for balanced tangent splitting on hypersurfaces"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00073,"raw_usage":{"total_tokens":3220,"prompt_tokens":850,"completion_tokens":2370,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":466,"completion_tokens_details":{"reasoning_tokens":2306}},"tokens_in":466,"tokens_out":2370,"duration_ms":16917,"temperature":1.0,"reasoning_tokens":2306,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:15:18.995916+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take an allowed triple from Theorem 7.1, for instance $d=5$ and $n=10$, write the displayed matrix $K_F$ with $t=1$, and compute its rank symbolically at $s=1$; if the rank comes out below $n-1=9$, the claimed balanced splitting $T_X|_C \\cong O(7)^6\\oplus O(6)^3$ is false. The same computation can be repeated for the matrices displayed in Theorems 5.1 and 6.1 at their stated parameter values.","supporting_citations":[],"review_version":1}