The restricted Hitchin map of wobbly vector bundles

Clemens Nollau

REVIEW 2 major objections 4 minor 13 references

The restricted Hitchin map of wobbly vector bundles

T0 review · 2 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read For a general wobbly rank-2 vector bundle, the restricted Hitchin map is dominant and generically finite, with an explicit degree formula depending on the stratum of the wobbly locus.

desk verdict Genuine new results on wobbly bundles and the restricted Hitchin map, but the degree formula in Theorem C(ii) currently rests on an unverified cohomology hypothesis. read the letter →

arxiv 2607.21457 v1 pith:D2JMG4VQ submitted 2026-07-23 math.AG

Clemens Nollau This is my paper · ORCID

classification math.AG MSC 14H6014D20

keywords restrictedHitchinmapwobblyvectorbundlesbadlinesubbundlenilpotenttwistedendomorphismverystablespectralcurvesBrill-Noethertheorydegreeformula

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper studies the restricted Hitchin map of a stable rank-2 vector bundle on a smooth projective curve, which sends a trace-free twisted endomorphism to its determinant quadratic differential. Its central claim is that for a general wobbly bundle—one that admits a nonzero nilpotent twisted endomorphism—this map is dominant and generically finite, with degree 2^{3g−3} − 2^{2g−2k−λ+1}. This answers a long-open dominance question on a dense open subset of each irreducible component of the wobbly locus. The key structural input is that a general wobbly bundle has a unique bad line subbundle, which collapses the indeterminacy of the projectivized map to a single point and makes the blowup computation tractable.

What carries the argument

The central object is the differential of the restricted Hitchin map at a nilpotent twisted endomorphism φ. It factors as multiplication by the induced section φ₁₂ composed with the restriction map c_ξ from trace-free endomorphisms to Hom(L₁, L₂⊗K_C), where L₁ is the kernel of φ and L₂ = V/L₁. Lemma 2.3, imported from an earlier paper, identifies im(c_ξ) with the kernel of the cup product ξ∪(−) under the hypothesis h⁰(L₁^∨⊗V)=1; this determines the codimension m = 2k+g−4+λ that enters the degree formula. Two blowups of the projectivized map—first at the unique nilpotent class [φ], then along the residual indeterminacy P(ker c_ξ / ⟨φ⟩)—resolve the rational map, and the divisor class 2H−E₁−E₂

What would settle it

Compute h⁰(L₁^∨⊗V) for a general extension in the stratum W_k: if it is greater than 1 on a dense open set, the equality im(c_ξ) = ker(ξ∪(−)) fails and the stated degree formula cannot hold. Equivalently, exhibit a wobbly bundle V satisfying the conditions of Lemma 4.4 for which some integral quadratic differential η has infinitely many preimages under h_V; Theorem C(i) would then be false.

Watch

Extended reading notes

Core claim

The main discovery is that the restricted Hitchin map of a general wobbly vector bundle in the stratum W_k is not only dominant but generically finite of degree 2^{3g−3} − 2^{2g−2k−λ+1}; in the special case (k, λ) = (g, 0) this reads 2^{3g−3} − 2. The proof blows up the projectivized restricted Hitchin map twice, using the differential formula tr(ψφ) = φ₁₂ c_ξ(ψ) and a lemma identifying the image of c_ξ with the kernel of the cup-product map ξ∪(−). For a general wobbly bundle the unique nilpotent twisted endomorphism is the only source of indeterminacy, and the degree is computed from an intersection number on the second blowup. The paper also proves that on a non-hyperelliptic curve every s

Load-bearing premise

The degree formula rests on Lemma 2.3, which assumes h⁰(L₁^∨⊗V)=1 for the extensions under consideration; if that dimension exceeds 1 on any of the bundles constructed in Lemma 4.4, the image of c_ξ is larger than claimed and the degree computation collapses.

Editorial extensions

If this is right

Dominance of the restricted Hitchin map holds for general wobbly bundles in every irreducible component of the wobbly locus, resolving Question 1.1 on a dense open subset.
For an integral spectral curve η, the fiber h_V^{-1}(η) is finite, giving isolated line bundles on the spectral curve whose pushforward is V and hence non-emptiness of relevant Brill-Noether loci.
The degree 2^{3g−3} − 2^{2g−2k−λ+1} is strictly smaller than the very stable degree 2^{3g−3}, quantifying exactly how wobbliness lowers the degree of the restricted Hitchin map.
A general wobbly vector bundle has a unique bad line subbundle, a structural simplification that makes the geometry of the wobbly locus more tractable.
For non-hyperelliptic curves, every stable rank-2 bundle has a twisted endomorphism with smooth spectral curve, giving a nonzero quadratic differential with simple zeros in the image of h_V.
The reducible spectral curve fibers of a wobbly bundle are positive dimensional, confirming that the positive-dimensional fibers of h_V are concentrated over reducible spectral curves.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The same blowup-and-intersection technique could plausibly extend to the remaining wobbly strata W_k with k below the range covered here, where the bad line subbundle may not be unique and the indeterminacy locus is larger.
The explicit degree formula gives a concrete enumerative prediction: the number of preimages of a general integral quadratic differential under the restricted Hitchin map of a general wobbly bundle is exactly the stated power-of-two expression, which could be checked computationally for low-genus curves.
If the result extends to higher-rank wobbly bundles, the structure of nilpotent twisted endomorphisms and the degree of the restricted Hitchin map would likely be governed by analogous cup-product kernels, suggesting a uniform pattern across ranks.
The link to Brill-Noether theory of spectral curves suggests one should test whether the isolated line bundles produced by the dominant restricted Hitchin map have the expected number of sections, potentially yielding new non-emptiness results for spectral curves over a general wobbly base.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: the paper has real content—Theorem A, Theorem B, and the overall strategy for Theorem C are worth taking seriously—but the central numerical claim is not established as written.

What is genuinely new: Theorem A gives a smooth spectral curve for every stable rank-2 bundle on a non-hyperelliptic curve, and the proof via base-point analysis is convincing. Theorem B, the uniqueness of the bad line subbundle for general wobbly bundles, is a useful structural result and the dimension count is reasonable. The double-blowup setup and the intersection-theoretic degree computation are a credible framework; the special case (k,λ)=(g,0) comes out consistently.

The main soft spot: the proof of Theorem C(ii) applies Lemma 2.3 to the bundles constructed in Lemma 4.4 without verifying the hypothesis h^0(L1∨⊗V)=1. Lemma 4.4 as stated assumes h^0(L)=1, which is impossible for large k because deg L = 2k+λ−2 exceeds g. That looks like a typo—presumably h^0(L∨⊗K)=1 is meant—but even with that correction the hypothesis of Lemma 2.3 does not follow automatically. The bundle sits in 0→O→L1∨V→L→0, so h^0(L1∨V)=1 requires the connecting map δ:H^0(L)→H^1(O) to have rank h^0(L). For large k, h^0(L) can be as large as g, so a genericity argument is needed and none is given. The stress-test note overstates the point when it says the failure is independent of δ, since δ can indeed be injective in that range, but the burden is on the authors to prove it. This directly affects the codimension m that feeds into the degree formula, so Theorem C(ii) is not proved for the full stated range.

Smaller issues: the inference in Theorem C(i) that finiteness implies dominance is actually valid for a morphism between equal-dimensional spaces, but the proof that h'' is finite needs the projection to have finite restriction, which requires the image to miss the center of projection; that is not shown. There are also mechanical errors—the dimension count N=3g−3 should be 3g−4 in the projectivized setting, and the section numbering conflicts—but these are easily fixed.

Who it is for: people working on the Hitchin fibration, wobbly bundles, or Brill–Noether theory of spectral curves. It deserves a serious referee, but not acceptance as-is. The referee should focus on Lemma 4.4 and the verification of the Lemma 2.3 hypothesis, and on the finiteness argument in Theorem C(i). If those are repaired, this would be a solid paper.

Referee Report

2 major / 4 minor

Summary. The paper studies the restricted Hitchin map h_V of a stable rank-2 vector bundle V on a smooth projective curve C. For a general wobbly bundle in the stratum W_k it claims that V has a unique bad line subbundle (Theorem B), that h_V is dominant and generically finite, and that its degree is 2^{3g-3} - 2^{2g-2k-\lambda+1} (Theorem C). The proof resolves the indeterminacy of the projectivized map by a double blow-up and computes the degree by intersection theory. The paper also proves that every stable rank-2 bundle on a non-hyperelliptic curve admits a twisted endomorphism with smooth spectral curve (Theorem A).

Significance. If fully established, the main theorem would give the first general answer to the dominance question for the restricted Hitchin map on a dense open subset of each irreducible component of the wobbly locus, and it would produce a degree formula contrasting with the very stable case. The double-blown-up intersection-theoretic approach, the use of a unique bad subbundle, and the dimension counts for strata are promising ideas. However, the numerical claim currently rests on an unverified—and in part impossible—hypothesis about h^0(L_1^\vee\otimes V), so the significance is conditional on a substantive repair.

major comments (2)

[§4, Lemma 4.4 and §5, Theorem C(ii)] The application of Lemma 2.3 requires h^0(L_1^\vee\otimes V)=1, but this is never verified and is incompatible with the stated range. For V an extension of L_2 by L_1 with L_2=L_1\otimes L, the sequence 0\to O_C\to L_1^\vee\otimes V\to L\to 0 gives h^0(L_1^\vee\otimes V)=1+\dim\ker\delta, where \delta:H^0(L)\to H^1(O_C). Lemma 4.4 assumes H^0(L)=1, but deg L=2k+\lambda-2, so for k>(g+2-\lambda)/2 Riemann-Roch gives h^0(L)\ge \deg L+1-g\ge 2; for k=g-\lambda and g\ge3 this is at least g-1. Thus such an L does not exist in most of the stated range. Even when h^0(L)=1 is possible, the proof never shows that \delta is nonzero for the chosen extension. Consequently the codimension m=g+2k+\lambda-4 used in Lemma 7.2 is not justified, and both Theorem C(ii) and the base-point-free argument using (10) in Theorem C(i) collapse.
[§5/§7, proof of Theorem C(ii)] The projective dimension is misstated. PE=P H^0(End_0(V)\otimes K_C) has dimension 3g-4, not 3g-3 as written. In the special case (k,\lambda)=(g,0) the computation \int(2H-E_1)^{3g-4}=2^{3g-4}-1 uses the correct dimension. If Lemma 7.2 is applied with N=3g-3, one obtains deg h''=2^{3g-3}-2^{2g-2k-\lambda+1}, and after the stated doubling this is not Theorem C(ii). With N=3g-4 the formula gives 2^{3g-4}-2^{2g-2k-\lambda}, and doubling reproduces the theorem. This may be a typo, but as written the numerical derivation is internally inconsistent.

minor comments (4)

[§1, Definition 1.4 and following] The definition of Z_k says Z_k\subset Pic^{1-g}(C), but for a bad line subbundle of degree 1-k the correct degree is 1-k. This typo propagates and makes the parametrization of W^0_k hard to follow.
[§1, Theorem B statement] “Let V be a be a general element” should read “Let V be a general element.”
[§1, final paragraph] The proof of Theorem C is said to be split between two occurrences of “section 7”; the second should refer to the analysis of the blowup (section 7) and the first presumably to section 5.
[§4, Lemma 4.4 proof] The notation W^0_d(C) is used both for effective divisor loci and for line bundles with h^0>0; please clarify the convention at first use.

Circularity Check

0 steps flagged · score 2.0 of 10

No substantive circularity: the main degree formula is derived by intersection theory, not fitted; the only self-citation is auxiliary.

full rationale

Theorem C's degree formula is not equivalent to an input. The finiteness/dominance claim is obtained by resolving the indeterminacy of the projectivized restricted Hitchin map on a double blowup, and the degree is computed as the top self-intersection number (2H - π_2^*E_1 - E_2)^N with N = 3g-3. The entry m = 2k+g-4+λ is obtained from Lemma 2.3 plus Riemann-Roch, and the final exponent 2g-2k-λ+1 is the arithmetic consequence N-m; nothing is fitted to force that value. Lemma 2.3 is imported from [2, eq. (2.5)], an external source with no author overlap, so it is not a self-citation. The only self-citation that carries mathematical weight is [9, Lemma 4.10], used in Proposition 6.1 to identify an extension class in the description of fibers over reducible spectral curves; that proposition is not used in the proof of Theorem C, so the self-citation is not load-bearing. The manuscript's most serious issue is a potential correctness gap rather than circularity: Lemma 2.3 assumes h^0(L_1^∨ ⊗ V) = 1, while Lemma 4.4 only enforces H^0(L) one-dimensional for L = L_1^∨ ⊗ L_2; for k near g-λ, h^0(L) is generically greater than 1, so the hypothesis may fail and the codimension m entering the degree formula would be unsupported. However, a missing or unverified hypothesis is not the same as the conclusion being built into the assumptions: the degree formula remains a genuine computed prediction from an external lemma. Score 2 reflects one minor, non-load-bearing self-citation, not circularity in the central derivation.

Assumptions & free parameters 0 free parameters · 10 assumptions · 0 invented entities

The central claims rest on standard algebraic-geometry tools plus several deep cited theorems (BNR, Martens, finiteness of maximal subbundles) and one technical lemma imported from the author's own earlier paper [9, Lemma 4.10]. No free parameters and no invented entities. The heaviest unproved assumption is Lemma 2.3, which feeds directly into the degree formula; the non-hyperelliptic hypothesis in Theorem A is explicit but also load-bearing.

assumptions (10)

domain assumption BNR correspondence: for an integral spectral curve C̃=V(τ²−η), h^{-1}(η) is in bijection with rank-1 torsion-free sheaves on C̃ pushing forward to Higgs bundles of determinant Λ.
Invoked in the Introduction to connect fibers of the restricted Hitchin map to line bundles on spectral curves; taken from [4,12], not proved in the text.
domain assumption Lemma 2.3: im(c_ξ) = ker(ξ∪(−)), imported from [2, equation (2.5)]; requires h^0(L_1^∨⊗V)=1.
Load-bearing for the codimension m of P(ker c_ξ) that enters the degree formula; the paper quotes [2] without proof and without verifying h^0(L_1^∨⊗V)=1 for the bundles in Lemma 4.4.
domain assumption On a non-hyperelliptic curve of genus g≥2, h^0(O_C(2q))=1 for every point q.
Used critically in the proof of Theorem A in both the nilpotent and non-nilpotent cases to force det ρ to be constant and to rule out L_+ ≅ L_−.
domain assumption Marten's theorem [8, Theorem 1]: dim W^r_d(C) ≤ d − 2r − 1 under the stated conditions.
Used in Lemma 4.2 to bound dim R^k_t ≤ 2t−2(g−1)−2(k−1).
standard math Clifford's theorem and the emptiness W^{k−1}_{2t−2(g−1)}(C)=∅ for k>t+2−g.
Used in Lemma 4.2 to prove Q^k_t is empty in the relevant range.
standard math r(E) ≤ ½ r(2E) for a divisor E (byproduct of the proof of Clifford's theorem, [1, III, §1]).
Used in Lemma 4.2 to estimate the dimension of the fiber of Q^k_t → R^k_t.
domain assumption A stable rank-2 bundle has only finitely many maximal line subbundles ([6, Proposition 4.2]).
Used in the proof of Theorem A to show that H^0(End_0(V)⊗O(q)) is nonzero for only finitely many points q.
standard math Bertini's theorem: a general member of a base-point-free linear system on a smooth variety is smooth ([5, Theorem 6.3]).
Used in the proof of Theorem A to conclude existence of φ with smooth spectral curve.
standard math Serre duality, Riemann–Roch, and the stability consequence h^0(End_0 V)=h^1(End_0 V⊗K)=0.
Used in Lemma 3.1 to identify im(ev_q) via the residue pairing; standard tools throughout.
domain assumption [9, Lemma 4.10]: for a line bundle M on the reducible spectral curve, the extension class of ψ_*M is δ(ξ̃).
Used in Proposition 6.1 to determine fibers over square differentials; cited from the author's earlier paper rather than proved here.

how reviews work

0 comments

Cite this review

Pith. "Pith review of The restricted Hitchin map of wobbly vector bundles." pith.science (2026). https://pith.science/paper/D2JMG4VQ

@misc{pith2026260721457,
  author       = {Pith},
  title        = {Pith review of: The restricted Hitchin map of wobbly vector bundles},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D2JMG4VQ}},
  note         = {Machine review of arXiv:2607.21457}
}

abstract

This article studies the restricted Hitchin map $h_V$ of a stable rank 2 vector bundle $V$ on a smooth projective curve $C$. This map associates to a trace-free twisted endomorphism $\varphi : V \to V \otimes K_C$ its determinant, which is a quadratic differential. We show that if $V$ is a general wobbly vector bundle, then it has a single nilpotent twisted endomorphism, up to scalars. As a consequence we show that $h_V$ is generically finite and we compute its degree. Furthermore, we show that if $C$ is not hyperelliptic, then the image of $h_V$ contains a quadratic differential with simple zeros. This is equivalent to saying that there is a smooth spectral curve associated to a twisted endomorphism of $V$.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 1 canonical work pages

[1]

Arbarello et al.Geometry of Algebraic Curves I

E. Arbarello et al.Geometry of Algebraic Curves I. Springer New York, NY, 1984

1984
[2]

Dinh and J

D. Dinh and J. Teschner.Classical limit of the geometric Langlands correspondence for SL(2,C). 2025. arXiv:2312.13393 [math.DG]

arXiv 2025
[3]

Eisenbud and J

D. Eisenbud and J. Harris.3264 and All That: A Second Course in Algebraic Geometry. Cambridge University Press, 2016

2016
[4]

The Self-Duality Equations on a Riemann Surface

N. J. Hitchin. “The Self-Duality Equations on a Riemann Surface”. In:Proceedings of the London Mathematical Societys3-55.1 (), pp. 59–126
[5]

Jouanolou.Th´ eor` emes de Bertini et applications

J-P. Jouanolou.Th´ eor` emes de Bertini et applications. Birkh¨ auser, 1983. 20

1983
[6]

Maximal subbundles of rank two vector bundles on curves

H. Lange and M.S. Narasimhan. “Maximal subbundles of rank two vector bundles on curves”. In:Math. Ann.266 (1983), pp. 55–72

1983
[7]

Larson and S

H. Larson and S. Vemulapalli.Brill–Noether theory of smooth curves in the plane and on Hirzebruch surfaces. 2024. arXiv:2408.12678 [math.AG]

arXiv 2024
[8]

On the varieties of special divisors on a curve

H. Martens. “On the varieties of special divisors on a curve.” In:Journal f¨ ur die reine und angewandte Mathematik227 (1967), pp. 111–120

1967

Show all 13 references

[9]

Nollau.Towards Brill-Noether Theory for Spectral Curves

C. Nollau.Towards Brill-Noether Theory for Spectral Curves. 2025. arXiv:2408.11404 [math.AG]

2025
[10]

Pal and C

S. Pal and C. Pauly. In:Advances in Geometry21.4 (2021), pp. 473–482.doi:doi: 10.1515/advgeom-2021-0020

2021 doi
[11]

Very stable bundles and properness of the Hitchin map

C. Pauly and A. Pe´ on-Nieto. “Very stable bundles and properness of the Hitchin map”. In:Geometriae Dedicata198 (2019). 5 pages, pp. 143–145

2019
[12]

Spectral curves and the gener- alised theta divisor

S. Ramanan, M.S. Narasimhan, and A. Beauville. “Spectral curves and the gener- alised theta divisor.” In:Journal f¨ ur die reine und angewandte Mathematik398 (1989), pp. 169–179

1989
[13]

Sawin.Surjectivity of a restricted Hitchin map

W. Sawin.Surjectivity of a restricted Hitchin map. MathOverflow. comment on a ques- tion of the author.url:https://mathoverflow.net/q/477473. 21

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

Arbarello et al.Geometry of Algebraic Curves I

E. Arbarello et al.Geometry of Algebraic Curves I. Springer New York, NY, 1984

1984

[2] [2]

Dinh and J

D. Dinh and J. Teschner.Classical limit of the geometric Langlands correspondence for SL(2,C). 2025. arXiv:2312.13393 [math.DG]

arXiv 2025

[3] [3]

Eisenbud and J

D. Eisenbud and J. Harris.3264 and All That: A Second Course in Algebraic Geometry. Cambridge University Press, 2016

2016

[4] [4]

The Self-Duality Equations on a Riemann Surface

N. J. Hitchin. “The Self-Duality Equations on a Riemann Surface”. In:Proceedings of the London Mathematical Societys3-55.1 (), pp. 59–126

[5] [5]

Jouanolou.Th´ eor` emes de Bertini et applications

J-P. Jouanolou.Th´ eor` emes de Bertini et applications. Birkh¨ auser, 1983. 20

1983

[6] [6]

Maximal subbundles of rank two vector bundles on curves

H. Lange and M.S. Narasimhan. “Maximal subbundles of rank two vector bundles on curves”. In:Math. Ann.266 (1983), pp. 55–72

1983

[7] [7]

Larson and S

H. Larson and S. Vemulapalli.Brill–Noether theory of smooth curves in the plane and on Hirzebruch surfaces. 2024. arXiv:2408.12678 [math.AG]

arXiv 2024

[8] [8]

On the varieties of special divisors on a curve

H. Martens. “On the varieties of special divisors on a curve.” In:Journal f¨ ur die reine und angewandte Mathematik227 (1967), pp. 111–120

1967