Skip to Content
MathematicsMathematics
  • Editor’s Choice
  • Article
  • Open Access

14 June 2026

10 Pages

On the Resonance Varieties of Vector Bundles on Curves

Department of Mathematics, University of Trento, 38123 Trento, Italy
†
The author is a member of Gruppo Nazionale per le Strutture Algebriche e Geometriche e loro Applicazioni of Istituto di Alta Matematica, 00185 Rome, Italy.

Abstract

We use the resonance (Papadima, Sociu, Aprodu and others) to study vector bundles E on a smooth curve X of genus g. Our key observation is that if we know it for a specific integer c, then we get strong information on X and E. Fix an integer c ≥ g + 1 . Take genus g curves X and Y and vector bundles E on X and F on Y with the same ranks and degrees. Our main result is that if E and F are sufficiently positive and they have birational resonance in degree c, then X and Y have isomorphic Jacobians. We also study the resonance for a singular curve with arithmetic genus 1.
MSC:
14H60; 14H20; 14J60

1. Introduction

We recall the linear algebra needed to give our definitions ([1], §2). Let V be a finite-dimensional complex vector space and K ⊂ ⋀ 2 V a linear subspace. Let G 2 ( V ∨ ) denote the Grassmannian of all two-dimensional linear subspace of V ∨ embedded in P ⋀ 2 ( V ∨ ) by the Plücker embedding. Set K ⊥ : = ( ⋀ 2 V ) / K ⊆ ⋀ 2 V ∨ . The vector space K ⊥ is called the orthogonal complement of K. The (projectivized) resonance of the pair ( V , K ) is the following subvariety of P V ∨ :
R ( V , K ) : = { [ a ] ∈ P V ∨ ∣ ∃ b ∈ V ∨ such that 0 ≠ a ∧ b ∈ K ⊥ } .
Thus R ( V , K ) is a closed algebraic subset of P V ∨ (it may be empty or reducible). As in [2,3] we take as V the linear space H 0 ( X , E ) , where X is a complex projective variety and E is a vector bundle on X of rank r > 1 . Let
d 2 : ⋀ 2 H 0 ( X , E ) → H 0 ( X , ⋀ 2 E )
denote the second determinantal map. We take K ⊥ : = ker ( d 2 ) and call R ( E ) : = R ( V , K ) the resonance variety of E. Let G 2 ( E ) be the Grassmannian of the two-dimensional linear subspaces of V. Set G ( E ) : = G 2 ( E ) ∩ P K ⊥ . The set G ( E ) is the set of all two-dimensional linear subspace of H 0 ( X , E ) spanning a rank 1 subsheaf of E. The rank 1 subsheaves of E obtained in this way are exactly the rank 1 saturated subsheaves R of E with h 0 ( X , R ) ≥ 2 ([2], §2). If we require that E is globally generated and with no trivial factor, then these spanned rank 1 subsheaves are the spanned rank 1 sheaves R such that E / R has no torsion. If we take a line bundle L on X with h 0 ( X , L ) ≥ 2 , and an injective map f : L → E , then it is only the saturation of f ( L ) that counts in G ( E ) . Because of the origin of the research on this topic and its connection with Hodge theory, usually X is assumed to be smooth, but this is not a necessity and often not used. In this paper (as in [2], §4.2, and in [3]) we take as X a curve, except in Section 4, where X is a smooth one. In many papers on this topic, e.g., [1,2,4,5,6,7], X is a smooth complex projective manifold, either because they study the Hodge theory of X or because they use the homological properties of regular local rings.
Let X be a smooth projective curve of genus g ≥ 0 and E a rank r > 1 vector bundle with h 0 ( X , E ) not too small. Since X is smooth, all rank 1 sheaves are line bundles. We use a minimal modification of the notation in [2], Remark 4.1 (we only use saturated rank 1 sheaves). Let G ( E ) be the set of all saturated line bundles L of E with h 0 ( X , L ) ≥ 2 . We get a set-theoretic map G ( E ) → ⨆ c Pic c ( X ) and we call W ( E ) ⊂ ⨆ c Pic c ( X ) its image. This is not an algebraic map, but M. Aprodu and C. Spiridon gave a finite stratification of G ( E ) , the flattening stratification intersected with the degree one, such that the restriction to each stratum is an algebraic map. For each c ∈ Z set W c ( E ) : = W ( E ) ∩ Pic c ( X ) . Let G c ( E ) be the subset of G ( E ) with image in W c ( E ) .
The main point of our paper is that if one knows a specific W c ( E ) or G c ( E ) instead of knowing the union of all such pieces, then we may get a lot of geometric information on X or E. The main aim of this paper is to show that for very positive bundles E knowing a single W c ( E ) is sufficient to almost reconstruct X and fully reconstruct X for g = 1 . To make rigorous the words ‘very positive line bundle’ we need the following notation. Let E be a rank r vector bundle. The slope μ ( E ) is the integer deg ( E ) / r . By definition the bundle E is semistable (resp. stable) if and only if either r = 1 or it has no proper subbundle F with μ ( F ) > μ ( E ) (resp. μ ( F ) = μ ( E ) ). If E is semistable, set μ + ( E ) = μ − ( E ) = μ ( E ) . If E is not semistable let μ + ( E ) be the maximal slope of a proper subbundle of E and let μ − ( E ) be the minimal slope of a quotient of E. It is well-known that μ + ( E ) and μ − ( E ) are finite rational numbers, that μ + ( E ) ≥ μ ( E ) ≥ μ − ( E ) and that these inequalities are strict if E is not semistable [8].
This is the main result of this paper, which we prove in Section 2.
Theorem 1.
Take genus g curves X and Y. Fix integers r ≥ 2 and c. If g = 1 assume c ≥ 2 . If g ≥ 2 assume c ≥ 2 g − 1 . Let E and F be rank r vector bundles on X and Y, respectively, of the same degree and such that μ − ( E ) ≥ c + 2 g − 1 and μ − ( F ) > c + 2 g − 2 . We have h 0 ( X , E ) = h 0 ( X , F ) = deg ( F ) + r ( 1 − g ) . Use any linear isomorphism H 0 ( X , E ) ≅ H 0 ( X , F ) to give an isomorphism between G 2 ( H 0 ( X , E ) ) and G 2 ( H 0 ( X , F ) ) . If the algebraic varieties G c ( E ) and G c ( F ) are birational, then Pic 0 ( X ) ≅ Pic 0 ( Y ) .
For g = 1 , Theorem 1 implies X ≅ Y , but for g > 1 , not quite so. We recall that for any smooth curve C of genus g the Abelian variety, Pic 0 ( C ) is isomorphic to the Jacobian J ( C ) . The Jacobian is a principally polarized Abelian variety, its principal polarization being called the theta-divisor Θ C of J ( C ) . The theorem of Torelli says that if ( J ( X ) , Θ X ) ≅ ( J ( Y ) , Θ Y ) as polarized Abelian varieties, then X ≅ Y . But Theorem 1 only says that J ( X ) ≅ J ( Y ) as Abelian varieties. It has been known since the late 1800s that there are Jacobian twins, non-isomorphic curves with isomorphic Jacobians ([9,10,11] also containing references to the previously known examples). For all g ≥ 3 a general genus g smooth curve X has the property that if J ( X ) ≅ J ( Y ) for some smooth curve Y, then X ≅ Y (Remark 1). Moreover, for each smooth curve X the number of smooth curves Y such that J ( Y ) ≅ J ( X ) is always finite (Remark 2).
Then we consider the properties of E, which we may detect knowing some W c ( E ) or the information on X, e.g., its gonality, that we may get testing suitable bundles E. Sometimes, we also use G c ( E ) , not just W c ( E ) . It is easy to see that if W c ( E ) = { L } and h 0 ( X , Hom ( L , E ) ) = 1 , then G c ( E ) is a projective space of dimension h 0 ( X , L ) − 1 (see [2], Prop. 3.8, for much more).
Obviously, W c ( E ) = ∅ if c ≤ 0 or if c = 1 and g = 1 . For g ≥ 2 the first integer c such that W c ( F ) ≠ ∅ for some bundle F on X is the gonality of X (see Proposition 2 for the higher gonalities). For a fixed bundle E we have W c ( E ) = ∅ for c ≫ 0 ([2], §4.1).
In Section 3 we describe the case g = 0 (Example 1) and the case g = 1 started by C. Spirodon in [3] (Example 2 and Proposition 1). Then we give ways to check the higher gonalities of the curve X.
In Section 4 we describe the case in which X is a singular curve of arithmetic genus 1. Thus X has a unique singular point and the singular point is an ordinary node or an ordinary cusps. Two nodal (resp. cuspidal) curves of arithmetic genus 1 are isomorphic. None of the result we prove distinguish between the nodal curve and the cuspidal one (see part (c) of Section 5 for a discussion). We also use pencils associated to the rank 1 sheaves that are not locally free (they are unique, up to isomorphisms, for each degree ([12], Lemma 0.2)).
Then there is a Section 5 in which we discuss possible extensions to the interested reader quoting several useful papers.
We thank the referees for several useful remarks.

2. Proof of Theorem 1

Lemma 1.
Let X be a smooth curve of genus g ≥ 0 . Fix an integer c. If g = 1 , we assume c ≥ 2 . If g ≥ 2 , we assume c ≥ 2 g − 1 . Let E be a rank r vector bundle on X satisfying the inequality μ − ( E ) ≥ c + 2 g − 1 . Then for all L ∈ Pic c ( X ) we have h 0 ( X , L ) = c + 1 − g ≥ 2 and h 0 ( X , H o m ( L , E ) ) = r μ ( E ) − r c + r − r g .
Proof. 
Fix L ∈ Pic c ( X ) . Since d ≥ 2 g − 1 , Riemann–Roch gives h 0 ( X , L ) = c + 1 − g . Since c ≥ g + 1 , we have h 0 ( X , L ) ≥ 2 and hence any non-zero map f : L → E contributes to a pencil, the saturation of f ( L ) . Since μ − ( E ) > c + 2 g − 2 , we have
μ − ( H o m ( L , E ) ) = μ − ( E ) − c > 2 g − 2 .
Thus h 1 ( X , H o m ( L , E ) ) = 0 . Thus Riemann–Roch for vector bundles gives the equality h 1 ( X , H o m ( L , E ) ) = deg ( E ) − r c + r − r g = r μ ( E ) − r c + r − r g . □
Lemma 2.
Let X be a smooth curve of genus g. Fix c ∈ Z and L ∈ Pic c ( X ) . Let E be a rank r ≥ 2 vector bundle such that μ − ( E ) > 2 g − 1 + c . Then
h 0 ( X , H o m ( L , E ) ) = r μ ( E ) − r c + r ( 1 − g ) > 0
and a general f ∈ H 0 ( X , H o m ( L , E ) ) has saturated image f ( L ) .
Proof. 
Lemma 1 gives h 0 ( X , H o m ( L , E ) ) = r μ ( E ) − r c + r ( 1 − g ) > 0 . To prove the last assertion we give an upper bound of the dimension of the maps f such that f ( L ) is not saturated. Take f ∈ H 0 ( X , H o m ( L , E ) ) , f ≠ 0 , which has a non-saturated image. Hence there is p ∈ X such that the line bundle f ( L ) contains a subsheaf isomorphic to L ( p ) . Lemma 1 gives h 0 ( X , H o m ( L ( p ) , E ) ) = h 0 ( X , H o m ( L , E ) ) − r . Since r > dim X , we get a contradiction for a general f. □
Proof of Theorem 1: 
By Lemmas 1 and 2 each L ∈ Pic c ( X ) contributes to a subpencil, i.e., W c ( E ) = Pic c ( X ) , and the surjection R c ( E ) → Pic c ( X ) has as fibers rational varieties of the same dimension. Let T be the desingularization of the dominating irreducible component of R c ( E ) ¯ . The Albanese mapping induces a surjection α : T → Pic c ( X ) ([13], Ch. IV, §9). Hence Pic c ( X ) is uniquely determined by R c ( E ) . For the same reason Pic c ( Y ) is uniquely determined by R c ( F ) . By the property of the Albanese mapping we get Pic c ( X ) ≅ Pic c ( Y ) . Since Pic c ( X ) ≅ Pic 0 ( X ) and the latter is isomorphic to the Jacobian of X and the same is true for Y, we get that X and Y have isomorphic Jacobians. □
Remark 1.
Fix an integer g ≥ 3 and let X be a general smooth curve of genus g. We have End ( J ( X ) ) ≅ Z [14] and we may even be more precise. Remember that the moduli space M g of all smooth genus g curve has dimension 3 g − 3 . Let M be an irreducible subvariety of M g such that a general C ∈ M has End ( J ( X ) ) ≅ Z . Then dim M ≤ 2 g − 2 .
Take any X ∈ M g such that End ( J ( X ) ) ≅ Z and take Y ∈ M g such that J ( X ) ≅ J ( Y ) . Hence End ( J ( Y ) ) ≅ Z . Hence every isomorphism between J ( X ) and J ( Y ) preserves their principal polarization. Thus the theorem of Torelli gives X ≅ Y . Hence for a general X ∈ M g , g ≥ 3 , Theorem 1 gives X ≅ Y .
Remark 2.
Let X be a smooth curve of genus g ≥ 2 . The set of all smooth curves Y such that J ( Y ) ≅ J ( X ) is finite [15].

3. Examples

First, we do the case X = P 1 . Aprodu and Spiridon described the case r = 2 ([2], §4.2).
Example 1.
We call c : = deg ( L ) and of course we need c > 0 . Take a rank r vector bundle E : = O P 1 ( a 1 ) ⊕ ⋯ ⊕ O P 1 ( a r ) . Let F be the direct sum of the factors of E with positive degree. The pencils of E and F are the same for each degree (and also the associated subsets on the Grassmannians, although the Grassmannians are different if a i = 0 for some i. Thus we describe the pencils of F (none of them if F = 0 , i.e., a i ≤ 0 for all i). Assume F ≠ 0 , say F = O P 1 ( b 1 ) ⊕ ⋯ ⊕ O P 1 ( b s ) with b 1 ≥ ⋯ ≥ b s . We have W c ( F ) = ∅ for c > b 1 and W b 1 ( F ) ≠ ∅ . We need the flattening stratification of [2], §3 and §4. The last stratum (the closed stratum) arises from O P 1 ( b 1 ) and it is P b 1 − 1 × P k − 1 . Then we use O P 1 ( b 1 − 1 ) keeping in mind that the large part comes from O P 1 ( b 1 ) (it is an open subset of a vector space of dimension 2 k ). Then we use b 1 − 2 and so on.
Example 2.
Let X be an elliptic curve, i.e., g = 1 . We have X ≅ Pic 0 ( X ) . Note the equality h 0 ( X , L ) = max { 0 , deg ( L ) } for all line bundles. So it is possible to recover the curve if we twist the bundle high enough. Theorem 2 gives the reconstruction of all indecomposable bundles with degree ≡ 0 mod r . As in Example 1 we reconstruct all direct sums of line bundles (see [3], Corollaries 3.4 and 3.5, for the case r = 2 ). For the reconstruction of of stable bundles, see Proposition 1.
Proposition 1.
Let X be an elliptic curve and E a globally generated stable vector bundle of rank r > 1 . Set d : = deg ( E ) . We have ( d , r ) = 1 . Set a : = ⌊ d / r ⌋ . Take L ∈ Pic c ( X ) . L contribute a saturated pencil if and only if c ≥ 2 and either c < a (with h 0 ( X , H o m ( L , E ) ) = d − r c or c = a (with h 0 ( X , H o m ( L , E ) ) = d − r c ).
Proof. 
Since E is stable of slope d / r ∉ Z , for all line bundles R either h 0 ( X , E ⊗ R ) = 0 or h 1 ( X , E ⊗ R ) = 0 ([16], Part. II). Hence h 0 ( X , E ⊗ R ) > 0 if and only if deg ( R ) ≥ − a . We have h 0 ( X , L ) = c , h 0 ( X , H o m ( L , E ) ) = d − r c if c ≤ a . Hence each pencil is saturated. □
If a = 2 in Proposition 1 we have a unique stratum. Thus this stratum is closed.
Definition 1.
Let X be a smooth projective curve of genus g ≥ 3 . For each integer r such that 1 ≤ r ≤ g − 2 the r-gonality of X is the first integer d such that h 0 ( X , L ) ≥ r + 1 for some line bundle L ∈ Pic d ( X ) .
Remark 3.
Let X be a smooth projective curve of genus g ≥ 3 . Obviously, the 1-gonality of X is the gonality of X. If a line bundle L compute the r-gonality, then h 0 ( X , L ) = r + 1 . Hence the gonalities of X are strictly increasing. By Clifford’s theorem, the r-gonality of X is at least 2 r and equality holds for a single r ≤ g − 2 if and only if X is hyperelliptic and hence all i-gonalities of X are 2 i . The r-gonalities are also defined for all r ≥ g − 1 , but they are obvious. Indeed, by Riemann–Roch and duality the (g − 1)-gonality is 2 g − 2 evinced only by the canonical bundle K X and for all r ≥ g the r-gonality of X is r − g . Now assume g ≥ 4 . The 2-gonality of X is at most twice the gonality of X. For more results on the higher gonalities, see [17,18,19].
Proposition 2.
Let X be a smooth curve of genus g ≥ 3 . Fix an integer i ∈ { 1 , … , g } and an integer r ∈ { 2 , … , g } and let x be the i-gonality of X. Then there is a stable rank r vector bundle on X such that G x − 1 ( E ) = ∅ , G x ( E ) ≠ ∅ and W x ( E ) contains a unique line bundle L with h 0 ( X , L ) = i + 1 and no other degree x rank 1 subbundle L 1 of E satisfies h 0 ( X , L 1 ) > i .
Proof. 
Take L ∈ Pic x ( X ) computing the i-gonality. We need to construct a rank r stable vector E such that G x − 1 ( E ) = ∅ , L ∈ W ( E ) x and if R ∈ W x ( E ) either L = R or h 0 ( X , R ) ≤ i . For all integers d let U ( r , d ) be the moduli space of the stable vector bundles on X of rank r and degree d. By [20], Prop. 1.4, [21], there is an integer d such that a general F ∈ U ( r , d ) has a degree x line subbundle, but no higher degree line subbundle. By [20], Prop. 1.4, the dimension of the set of all degree x line subbundles of F is at most r − 1 < g . Thus there is R ∈ Pic 0 ( X ) such that E : = F ⊗ R has L as a subbundle, but no other degree x subbundle evincing the i-gonality of X. Since dim Pic 0 ( X ) = g > x , we may also find E such that all its other degree x saturated line subbundles L 1 , if any, satisfies h 0 ( X , L 1 ) ≤ i . □
Proposition 3.
Let X be a smooth curve of genus g ≥ 3 and let x be the gonality of g. Fix an integer d such that 2 x + 1 ≤ d ≤ g + x − 1 . Then there is a rank 2 stable vector bundle E on X such that h 0 ( X , E ) = 2 , W x − 1 ( E ) = ∅ , W x ( E ) = { L } with L evincing the gonality of X and W y ( E ) = ∅ for all y > x .
Proof. 
Fix L ∈ Pic x ( X ) such that h 0 ( X , L ) = 2 . Fix a general R ∈ Pic d − x ( X ) . Let E be a general extension of R by L. Since d ≥ 2 x + 1 , deg ( R ) > deg ( L ) . Since deg ( R ) > deg ( L ) , g ≥ 2 and E is given by a general extension, E is stable. Since d − x ≤ g − 1 and R is general, h 0 ( X , R ) = 0 . Thus h 0 ( X , E ) = h 0 ( X , L ) = 2 . Hence L is the only line bundle contributing to G ( E ) . □
Proposition 4.
Let X be a smooth curve of genus g ≥ 3 and let x be the gonality of g. Fix an integer d such that 2 x + 1 ≤ d ≤ 2 g + 1 . Then there is a rank 2 stable vector bundle E on X such that h 0 ( X , E ) = 3 , W x − 1 ( E ) = ∅ and W x ( E ) = { L } with L evincing the gonality of X.
Proof. 
Since x is the gonality of a smooth genus g curve, x ≤ ⌈ ( g + 2 ) / 2 ⌉ . Fix L ∈ Pic x ( X ) such that h 0 ( X , L ) = 2 . Note that h 1 ( X , L ) = g + 1 − x . Let E be the rank 2 vector bundle of degree d obtained from L ⊕ O X making d − x general positive elementary transformations. Since these positive elementary transformations are general, L is a saturated subbundle of X. Since h 1 ( X , L ⊕ O X ) = 2 g + 1 − x and the elementary transformations are general, h 1 ( X , E ) = h 1 ( X , L ⊕ O X ) − d + x . Thus h 0 ( X , E ) = 3 . Counting the dimensions of the unstable bundles and using that d − x > x , we get the stability of E. □
By Riemann–Roch we have h 0 ( X , F ) ≥ 4 for all rank 2 vector bundles F on X such that deg ( F ) ≥ 2 g + 2 . Hence the upper bound in Proposition 4 is a necessary one. Stability and the existence of a rank 1 subsheaf of degree x require the inequality d > 2 x .

4. Singular Curves of Arithmetic Genus 1

In this section X is an integral projective curve of arithmetic genus 1. We assume that X is singular, leaving to Remark 7 the case X smooth and E an indecomposable bundle of degree d ≡ 0 ( mod r ) . The curve X has a unique singular point, o, which is either an ordinary node or an ordinary cusp. The affine algebraic set X ∖ { o } is isomorphic to Pic 0 ( X ) . This algebraic group acts on X and the restriction of this action to X ∖ { o } is simply transitive. The algebraic group Pic 0 ( X ) is isomorphic to the multiplicative group if X is nodal and to the additive group if X has a cusp. Since X is a Gorenstein curve, a torsion free sheaf is reflexive ([22], Lemma 1.1). Hence we use the classification of rank 1 torsion free sheaves on X ([12], Lemma 0.2). There is a unique rank 1 reflexive sheaf F on X of degree 0, which is not locally free. For all R , L ∈ Pic c ( X ) we have F ⊗ L ≅ F ⊗ R and F is the unique degree c reflexive, but not locally free, sheaf on X.
Lemma 3.
Fix integers c and a. Take L c ∈ Pic c ( X ) and L a ∈ Pic a ( X ) .
(a) If c ≤ a , then h 0 ( X , H o m ( F ⊗ L c , L a ) ) = 0 .
(b) If c > a , then h 0 ( X , H o m ( F ⊗ L c , L a ) ) = a − c .
(c) If c > a , then h 0 ( X , H o m ( F ⊗ L c , F ⊗ L a ) ) = a − c , h 0 ( X , H o m ( F ⊗ L c , F ⊗ L c ) ) = 1 and h 0 ( X , H o m ( F ⊗ L c , F ⊗ L a ) ) = 0 if c > 0 .
(d) We have h 0 ( X , F ⊗ L c ) = 0 if c ≤ 0 and h 0 ( X , F ⊗ L c ) = c if c > 0 .
Proof. 
Let C o denote the skyscraper sheaf on X with o its support and h 0 ( X , C o ) = 1 . For any line bundle R the unique surjection (up to a non-zero scalar) R → C o has kernel isomorphic to F ⊗ L , where L is any line bundle of degree deg ( R ) − 1 . This observation gives parts (a) and (b) and that h 0 ( X , H o m ( F ⊗ L c , L a ) ) = h 0 ( X , H o m ( F , L a ⊗ L c ∨ ) ) .
The case c ≤ 0 of (c) follows from the fact that F ≠ O X . The case c > 0 follows from duality and Riemann–Roch. The case c = 0 of part (d) is true, because F ⊗ L c ≠ 0 . It gives the case c < 0 , while the case c > 0 follows from duality. □
Proposition 5.
Fix integers c, a and b such that a ≤ b . We have h 0 ( F ⊗ L c ) ≥ 2 if and only if c ≥ 2 . There is non-zero f ∈ H 0 ( X , H o m ( F ⊗ L c , L a ⊕ L b ) with f ( F ⊗ L c ) saturated in L a ⊕ L b if and only if c < a . If c < a , then h 0 ( X , H o m ( F ⊗ L c , L a ⊕ L b ) ) = a + b − 2 c and the set of all f ∈ H 0 ( X , H o m ( F ⊗ L c , L a ⊕ L b ) ) such that f ( F ⊗ L c ) is saturated is a nonempty Zariski open subset of C a + b − 2 c .
Proof. 
It only remains to prove that if c < a , then f ( F ⊗ L c ) is saturated for a general map f ∈ H 0 ( X , H o m ( F ⊗ L c , L a ⊕ L b ) ) . Call B the set of all f ∈ H 0 ( X , H o m ( F ⊗ L c , L a ⊕ L b ) ) such that f ≠ 0 and f ( F ⊗ L c ) is not saturated. Take f ∈ B . Let G be the saturation of f ( F ⊗ L c ) . Since f ( F ⊗ L c ) , there is a degree c + 1 torsion free sheaf A such that f ( F ⊗ L c ) ⊆ A ⊆ G and deg ( A ) = c + 1 . Use that the set of all rank 1 torsion free sheaves of degree c + 1 has dimension 1 and h 0 ( X , H o m ( A , L a ⊕ L b ) ) = a + b − 2 c − 2 . □
Remark 4.
Let A and B be non-isomorphic rank 1 torsion free sheaves of the same degree. The vector space E x t 1 ( A , B ) is trivial ([12], Th. 1.1).
Remark 5.
Let E be a rank r > 1 indecomposable and semistable rank r vector bundle on X of degree a r for some integer a. Fix L a ∈ Pic a ( X ) such that L a ⊗ r ≅ det ( E ) . Note that L a exists, because deg ( det ( E ) ) = a r and Pic 0 ( X ) is r-divisible. Set F : = E ( − L a ) . Since F is semistable, indecomposable and det ( F ) ≅ O X , there is a unique a rank 1 torsion free sheaf B such that deg ( B ) = 0 and there is a filtration 0 = F 0 ⊂ F 1 ⊂ ⋯ ⊂ F r − 1 ⊂ F r = F such that F i / F i − 1 ≅ B for all i = 1 , … , r ([12], Th. 1.1). Obviously, each F i , i ≠ 0 , is semistable. Let A be a rank 1 torsion free sheaf such that deg ( A ) = 0 and A ≠ B . The proof of [12], Theorem 1.1, gives h 0 ( X , H o m ( A , E ) ) = 0 . Thus B is uniquely determined by F. In this case (but not for elliptic curves) roots of unity are unique. We say that B ⊗ L a the degree a sheaf associated to E.
Lemma 3 gives the following result.
Proposition 6.
Fix an integer r > 1 , a degree c rank 1 torsion free sheaf B and a semistable rank r vector bundle E on X such that deg ( E ) = a r with a ≠ c . Then E has a pencil associated to B if and only if c < a . If c < a , then the set of all saturated pencils is a nonempty Zariski open subset of a vector space of dimension r ( a − c ) .
Theorem 2.
Let E be a rank r > 1 indecomposable and semistable rank r vector bundle on X of degree a r for some integer a. Call B the rank 1 degree a sheaf associated to E. Fix an integer c and take L c ∈ Pic c ( X ) . Take as A either L c or F ⊗ L c .
(a) A is a pencil and there is a saturated f : A → E if and only if either 2 ≤ c < a (and in this case the set of saturated pencils is a nonempty Zariski open subset of a vector space of dimension r ( a − c ) ) or a = c and A ≅ B .
(b) Assume a = c and A ≅ B . We have h 0 ( X , H o m ( B , E ) ) ≥ 1 . We have h 0 ( X , H o m ( B , E ) ) = 1 if and only if either B is locally free or B is not locally free and E is strongly indecomposable, i.e., h 0 ( X , H o m ( E , F ) ) = 1 , and in this case h 0 ( X , H o m ( B , E ) ) = 1 .
Proof. 
By Proposition 6 we only need to check the case a = c . If B is locally free, then h 0 ( X , H o m ( B , E ) ) = 1 by [12], Lemma 1.7. If B is not locally free, then this is the definition of strongly indecomposable bundle ([12] p. 17) and the observation that taking F as in Remark 5 we have F ≅ F ∨ . □
Remark 6.
Obviously, h 0 ( X , H o m ( A , E ) ) > 1 if A ≅ B , A is not locally free and E is not strongly indecomposable.
Remark 7.
Obviously, Theorem 2 is true also if X is smooth.

5. Suggestions for Further Work

In this section we give for the interested reader several possible extensions of this paper.
(a) X is a smooth curve of genus g ≥ 2 . Let U ( r , d ) denote the smooth and integral variety of all rank r stable vector bundles of degree d (we only allow the stable ones in this moduli space).
  • Continue Section 3 trying to identify other properties, e.g., the Clifford index, of X. Much more important, to prove Proposition 2 we needed to know x. In principle, at least for the usual gonality it is obvious, but it is cheating if one needs to know the pencils of X. Only in the Oracle model of this query is this a real solution.
  • Describe all cases in which E is a direct sum of line bundles (assuming everything you need about the Brill–Noether theory of line bundles of X).
  • Use the Brill–Noether theory of stable vector bundles [23,24].
  • Do the rank 2 case. Let E be a rank 2 indecomposable bundle. Set d : = deg ( E ) . Let L be a subbundle of E with c : = deg ( E ) maximal. If c > d / 2 , then E is unstable, L is unique and, since E is indecomposable, h 1 ( X , ( E / L ) ⊗ L ∨ ) ≠ 0 . Thus 2 c ≤ d + g . If E is properly semistable, then d = 2 c , L is unique and we need just L and E / L . Now assume that E is stable, i.e., 2 c < d . The integer e : = d − 2 c is called the index of stability of E. We have e ≡ d ( mod 2 ) and e ≤ g ([25], Ex. V.2.5). The variety U ( 2 , d ) is stratified by the index of stability, the open stratum being the one with largest e, with e = g − 1 or e = g according to the parities of d and g. For each 0 < e ≤ g − 2 , e ≡ d ( mod 2 ) the set U ( 2 , d , e ) is irreducible and a general E ∈ U ( 2 , d , e ) has a unique line subbundle of maximal degree. For more, see for instance [26].
(b) Give a detailed study of the resonance of all vector bundles on low genus curves, say genus 2 or genus 3 dividing between hyperelliptic curves and the other genus 3 curves. E. W. Howe gave examples of twin Jacobian pairs in genus 3, one of the twins being hyperelliptic [10]. The paper [11] gives infinitely many genus 2 twins defined over the rational numbers. C. Ciliberto and G. Van der Geer gave the first example of Jacobian twins for genus 4 curves [9].
(c) Consider a singular curve X, maybe with only ordinary nodes [8,27,28] or with only ordinary nodes and cusps, with help from the arithmetic genus 1 case considered in [12] and in full generality in [29]. Note that even for the arithmetic genus 1 indecomposability, even in the rank 2 case, does not imply semistability ([28], Lemma 2.1). To the best of our knowledge the equivalent of Theorem 1 would only give the isomorphic type of the Jacobian of the nomalization of X. This is the reason we are unable to distinguish cuspidal and nodal curves of genus 1: they have isomorphic normalization. For smooth curves we use that Pic 0 ( X ) is isomorphic to the Albanese variety of X, which is false for singular curves.
(d) Take X of higher dimension. The case in which X is either P 2 or P 3 should be doable. For X smooth it is useful that reflexive rank 1 sheaves are locally free ([30], Proposition 1.9). Hence for r = 2 we are just looking at bundles E fitting in an exact sequence
0 → L 1 → E → I Z ⊗ L 2 → 0
with L 1 , L 2 line bundles and either Z = ∅ or Z ⊂ X a locally complete intersection subscheme of pure codimension 2 satisfying some conditions (Cayley–Bacharach conditions) that need to have E locally free. If X is a smooth surface, then each reflexive sheaf is locally free ([30], Corollary 1.4), but for dim X > 2 reflexive sheaves are important [30] and may be studied instead of the locally free E. Multilinear algebra can be done for these sheaves ([25], Ex. II.5.16), but the tensor product of two torsion-free sheaves may have torsion.
(e) Consider the case of an algebraically closed field of positive characteristic.
(f) Consider the case of complex projective varieties defined over the real number field.

6. Discussion

Let V be a finite-dimensional complex vector space and K ⊂ ⋀ 2 V a linear subspace. Let G 2 ( V ∨ ) denote the Grassmannian of all two-dimensional linear subspace of V ∨ embedded in P ⋀ 2 ( V ∨ ) by the Plücker embedding. Set K ⊥ : = ( ⋀ 2 V ) / K ⊆ ⋀ 2 V ∨ . Given the pair ( V , K ) Papadima and Suciu defined an algebraic subset of the Grassmannian of the rank 2 linear subpaces of V ∨ [1]. This notion was used to get an impressing number of major results in several areas, from group theory to Commutative Algebra, from Koszul cohomology to the Hodge theory of projective manifolds and the study of algebraic vector bundles on projective manifolds [2,4,5,6,7]. Let X be a projective variety and E an algebraic vector bundle on X. The resonance variety R ( E ) of the vector bundle E is an algebraic subset of the Grassmannian of two-dimensional linear subspaces of the vector space H 0 ( X , E ) . Aprodu and Spiridu gave the foundational work, e.g., the flattening stratification, to apply these concepts for an algebraic vector bundle E on X and define the resonance R ( E ) of the vector bundle E [2]. If X is a curve, as in our paper, they proved that R ( E ) is the union of finitely many strata R c ( E ) , c ∈ Z , each of them associated to degree c line bundles on X. In our paper we presume to know a specific slice, say R c ( E ) , of the resonance variety R ( E ) and show that this knowledge gives very strong information on X and on E. Fix an integer c ≥ g + 1 . Take smooth genus g curves X and Y and vector bundles E on X and F on Y with the same ranks and degrees. Our main result is that if E and F are sufficiently positive and they have birational resonance in degree c, then X and Y have isomorphic Jacobians. Hence if either g = 1 or X is general, we get X ≅ Y . With no assumption for g ≥ 2 there are only finitely many genus g curves Y such that J ( Y ) ≅ J ( X ) . We also study the resonance for a singular curve with arithmetic genus 1. One section of the paper is devoted to suggestions for further work.

Funding

This research received no external funding.

Data Availability Statement

No new data were created or analyzed in this study and no database was constructed.

Conflicts of Interest

The author declares no conflicts of interest.

References

  1. Papadima, S.; Suciu, A.I. Vanishing resonance and representations of Lie algebras. J. Reine Angew. Math. 2014, 706, 83–101. [Google Scholar] [CrossRef] [Scilit]
  2. Aprodu, M.; Spiridon, C. Linear sections of Grassmannians and resonance of vector bundles. Int. Math. Res. Not. 2025, 2025, rnaf328. [Google Scholar] [CrossRef] [Scilit]
  3. Spiridon, C. Resonance of rank-two vector bundles over elliptic curves. arXiv 2026, arXiv:2604.23291v1. [Google Scholar] [CrossRef] [Scilit]
  4. Aprodu, M.; Farkas, G.; Papadima, S.; Raicu, C.; Weyman, J. Topological invariants of groups and Koszul modules. Duke Math. J. 2022, 171, 2013–2046. [Google Scholar] [CrossRef] [Scilit]
  5. Aprodu, M.; Farkas, G.; Raicu, C.; Suciu, A.I. Reduced resonant schemes and Chen ranks. J. Reine Angew. Math. 2024, 814, 205–240. [Google Scholar]
  6. Aprodu, M.; Farkas, G.; Raicu, C.; Weyman, J. Koszul modules with vanishing resonance in algebraic geometry. Sel. Math. 2024, 30, 24. [Google Scholar]
  7. Aprodu, M.; Farkas, G.; Raicu, C.; Sammartano, A.; Suciu, A.I. Higher resonance schemes and Koszul modules of simplicial complexes. J. Algebr. Comb. 2024, 59, 787–805. [Google Scholar] [CrossRef] [Scilit]
  8. Seshadri, C.S. Fibrés vectoriels sur les courbes algébriques. In Astérisque; Soc. Math. de France: Paris, France, 1982; Volume 96, 209p. [Google Scholar]
  9. Ciliberto, C.; van der Geer, G. Non-isomorphic curves of genus four with isomorphic (nonpolarized) jacobians. In Classification of Algebraic Varieties; Ciliberto, C., Livorni, E.L., Sommese, A.J., Eds.; Contemp. Math.; American Mathematical Society: Providence, RI, USA, 1994; Volume 162, pp. 129–133. [Google Scholar]
  10. Howe, E.W. Plane quartics with Jacobians isomorphic to a hyperelliptic Jacobian. Proc. Am. Math. Soc. 2001, 129, 1647–1657. [Google Scholar] [CrossRef] [Scilit]
  11. Howe, E.W. Infinite families of pairs of curves over Q with isomorphic Jacobians. J. Lond. Math. Soc. 2005, 72, 327–350. [Google Scholar] [CrossRef] [Scilit]
  12. Friedman, R.; Morgan, J.; Witten, E. Vector bundles over elliptic fibrations. J. Algebr. Geom. 1999, 2, 279–401. [Google Scholar]
  13. Ueno, K. Classification Theory of Algebraic Varieties and Compact Complex Spaces; Lecture Notes in Math; Springer: Berlin/Heidelberg, Germany; New York, NY, USA, 1975; Volume 439, xix+278p. [Google Scholar]
  14. Ciliberto, C.; van der Geer, G.; Teixidor i Bigas, M. On the number of parameters of curves whose Jacobians possess nontrivial endomorphisms. J. Algebr. Geom. 1992, 1, 215–229. [Google Scholar]
  15. Narasimhan, M.; Nori, M. Polarizations on an abelian variety. Proc. Indian Acad. Sci. Math. Sci. 1981, 90, 125–128. [Google Scholar] [CrossRef] [Scilit]
  16. Atiyah, M. Vector bundles over an elliptic curve. Proc. Lond. Math. Soc. 1957, 7, 414–452. [Google Scholar] [CrossRef] [Scilit]
  17. Lange, H.; Martens, G. On the gonality sequence of an algebraic curve. Manuscripta Math. 2012, 137, 457–473. [Google Scholar] [CrossRef] [Scilit]
  18. Kato, T.; Martens, G. The gonality sequence of a curve with an involution. Arch. Math. 2014, 103, 111–116. [Google Scholar] [CrossRef] [Scilit]
  19. Keem, C.; Martens, G. The gonality sequence of covering curves. Arch. Math. 2015, 105, 33–43. [Google Scholar] [CrossRef] [Scilit]
  20. Oxbury, W.M. Varieties of maximal line subbles. Math. Proc. Camb. Philos. Soc. 2000, 129, 9–18. [Google Scholar] [CrossRef] [Scilit]
  21. Holla, Y.I. Counting maximal subbundles via Gromov-Witten invariants. Math. Ann. 2004, 328, 121–133. [Google Scholar] [CrossRef] [Scilit]
  22. Hartshorne, R. Generalized divisors on Gorenstein curves and a theorem of Noether. J. Math. Kyoto Univ. 1986, 26-3, 375–386. [Google Scholar]
  23. Russo, B.; Teixidor i Bigas, M. On a conjecture of Lange. J. Algebr. Geom. 1999, 8, 483–496. [Google Scholar]
  24. Newstead, P.E. Higher rank Brill-Noether theory and coherent systems: Open questions. Proyecciones 2022, 41, 449–480. [Google Scholar] [CrossRef] [Scilit]
  25. Hartshorne, R. Algebraic Geometry; Springer: Berlin/Heidelberg, Germany, 1977. [Google Scholar]
  26. Choe, I.; Choy, J.; Park, S. Maximal line subbundles of stable bundles of rank 2 over an algebraic curve. Geom. Dedicata 2007, 125, 191–202. [Google Scholar] [CrossRef] [Scilit]
  27. Bhosle, U.N. Maximal subsheaves of torsionfree sheaves on nodal curves. J. Lond. Math. Soc. 2006, 74, 59–74. [Google Scholar] [CrossRef] [Scilit]
  28. Bhosle, U.; Biswas, I. Torsionfree sheave over a nodal curve of arithmetic genus one. Proc. Indian Acad. Sci. (Math. Sci.) 2008, 118, 81–98. [Google Scholar] [CrossRef] [Scilit]
  29. Teodorescu, T. Semistable Torsion-Free Sheaves over Curves of Arithmetic Genus One. Ph.D. Thesis, Columbia University, New York, NY, USA, 1999. [Google Scholar]
  30. Hartshorne, R. Stable reflexive sheaves. Math. Ann. 1980, 254, 121–176. [Google Scholar] [CrossRef] [Scilit]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.