betaFeedback

WorksMathematics

Rational Torsion and Ramification Constraints on Odd Superelliptic Jacobians

Abstract

Let C: y^d = f(x) be a smooth projective superelliptic curve over Q, where d >= 3 is odd, f(x) in Q[x] is separable of degree n >= 3, and gcd(d,n) = 1. Let J = Jac(C) be its Jacobian. We study the rational torsion subgroup J(Q)_tors through the interaction of ramification geometry, Galois actions, reduction at primes of good reduction, and the geometry of divisors. First, we describe the subgroup R_d contained in J[d] generated by the ramification-point differences and determine its Galois-invariant subgroup explicitly in terms of the sizes of the Galois orbits of the roots of f. This yields a precise description of the rational d-torsion classes arising from ramification, without assuming that they account for all of J(Q)[d]. Second, we establish criteria based on the orders of reduced Jacobians that can prove the vanishing of the full rational torsion subgroup. Finally, we derive geometric constraints on divisor classes represented by effective divisors of degree smaller than the genus, including a restriction arising when distinct effective divisors represent the same class. Explicit examples illustrate how the Galois-theoretic and reduction-theoretic methods can be combined to study rational torsion on odd superelliptic Jacobians.

Introduction

Let

C: yd=f(x)C:\ y^{d}=f(x)

be a smooth projective superelliptic curve over Q\mathbb{Q}, where d≥3d \ge3 is odd, f(x)∈Q[x]f(x) \in\mathbb{Q}[x] is separable of degree n≥3n \ge3, and gcd⁡(d,n)=1\gcd(d,n)=1. Let J=Jac⁡(C)J=\operatorname{Jac}(C) be its Jacobian, and write J(Q)torsJ(\mathbb{Q})_{\mathrm{tors}} for the subgroup of rational torsion points. The study of rational torsion on Jacobians is a central problem in arithmetic geometry: it asks which divisor classes of finite order are defined over the ground field, and how the geometry and arithmetic of the underlying curve constrain their possible orders and structure.

For elliptic curves, the Nagell–Lutz theorem gives an explicit integrality condition for rational torsion, and Mazur’s theorem classifies the possible rational torsion subgroups over Q\mathbb{Q} [10]. In higher genus, there is no comparable general classification for Jacobians of arbitrary curves. One therefore seeks effective structural information and criteria that can be applied to specific families or individual curves. The geometry of algebraic curves, including special divisors and Weierstrass points, provides important tools in this setting [1, 5]. The arithmetic study of Jacobians also draws on the theory of Néron models and reduction, which provides a framework for relating torsion over number fields to torsion over finite residue fields [3, 12].

For superelliptic curves, the covering structure provides a natural starting point. The map

π:C⟶P1,(x,y)⟼x\pi:C\longrightarrow\mathbb{P}^{1},\qquad(x,y)\longmapsto x

is a cyclic covering of degree dd, and its ramification points yield distinguished divisor classes on the Jacobian. The structure of torsion subgroups associated with superelliptic Jacobians has been studied from several perspectives; see, for example, Jedrzejak [7] and Wawrów [14]. The cyclic symmetries of curves and their induced actions on Jacobians also fit into the broader study of group actions and decompositions of Jacobians [8].

Related Diophantine questions involving superelliptic curves have also been investigated. In particular, our earlier work on a generalized fruit Diophantine equation connects a Diophantine problem with the study of superelliptic curves [2]. This provides part of the broader motivation for examining how arithmetic conditions on a curve interact with its geometric structure. The present paper focuses specifically on rational torsion on Jacobians, using ramification classes, Galois actions, reduction at primes of good reduction, and restrictions arising from low-degree divisors. For further background on the arithmetic of curves and their Jacobians, we refer to Hindry and Silverman [6].

The aim of this paper is to combine three complementary approaches to rational torsion: the Galois action on ramification classes, reduction at primes of good reduction, and the geometry of low-degree divisors. The first approach identifies a distinguished subgroup of the geometric dd-torsion and computes its rational Galois invariants. The second gives criteria that can force the entire rational torsion subgroup to vanish, without assuming that all rational torsion comes from ramification. The third describes geometric restrictions on divisor classes that admit effective representatives of degree smaller than the genus. These approaches answer different questions, and keeping them separate is essential: the ramification calculation alone does not classify all of J(Q)[d]J(\mathbb{Q})[d].

Ramification classes and Galois invariants. Let α1,…,αn\alpha_1,\ldots,\alpha_n be the roots of ff in Q‾\overline{\mathbb{Q}}, and let

Ri=(αi,0),1≤i≤n,R_i=(\alpha_i,0),\qquad1\le i\le n,

be the corresponding geometric ramification points. Since gcd⁡(d,n)=1\gcd(d,n)=1, the curve has a unique point ∞\infty above infinity. The divisor relations

div⁡(x−αi)=dRi−d∞,div⁡(y)=∑i=1nRi−n∞\operatorname{div}(x-\alpha_i)=dR_i-d\infty,\qquad\operatorname{div}(y)=\sum_{i=1}^{n}R_i-n\infty

show that the classes

ri=[Ri−∞]∈J[d](Q‾)r_i=[R_i-\infty]\in J[d](\overline{\mathbb{Q}})

are dd-torsion and satisfy ∑iri=0\sum_i r_i=0. We define the ramification-supported subgroup

Rd=⟨r1,…,rn⟩⊂J[d](Q‾).R_d=\langle r_1,\ldots,r_n\rangle\subset J[d](\overline{\mathbb{Q}}).

Although RdR_d is stable under the absolute Galois group GQ=Gal⁡(Q‾/Q)G_{\mathbb{Q}}=\operatorname{Gal}(\overline{\mathbb{Q}}/\mathbb{Q}), its generators need not be rational. Consequently,

RdGQ⊆J(Q)[d],R_d^{G_{\mathbb{Q}}}\subseteq J(\mathbb{Q})[d],

and equality is not automatic.

The first structural result identifies the abstract group generated by these classes and gives a concrete description of its Galois invariants. The following statements are proved in Section 4.

Lemma 1.1. [Structure of the ramification subgroup] The homomorphism

Φ:(Z/dZ)n⟶Rd,(a1,…,an)⟼∑i=1nairi\Phi:(\mathbb{Z}/d\mathbb{Z})^n\longrightarrow R_d,\qquad(a_1,\ldots,a_n)\longmapsto\sum_{i=1}^{n}a_i r_i

is surjective and has kernel equal to the diagonal subgroup

Δ={(t,…,t):t∈Z/dZ}.\Delta=\{(t,\ldots,t):t\in\mathbb{Z}/d\mathbb{Z}\}.

In particular,

Rd≅(Z/dZ)n/Δ≅(Z/dZ)n−1.R_d\cong(\mathbb{Z}/d\mathbb{Z})^n/\Delta\cong(\mathbb{Z}/d\mathbb{Z})^{n-1}.

Theorem 1.2. [Galois invariants of the ramification subgroup] Suppose that the roots of ff decompose into GQG_{\mathbb{Q}}-orbits

{α1,…,αn}=O1⊔⋯⊔Os,mj=∣Oj∣.\{\alpha_1,\ldots,\alpha_n\}=O_1\sqcup\cdots\sqcup O_s,\qquad m_j=|O_j|.

Then there is an isomorphism

RdGQ≅{(b1,…,bs)∈(Z/dZ)s:∑j=1smjbj=0 in Z/dZ}.R_d^{G_{\mathbb{Q}}}\cong\left\{(b_1,\ldots,b_s)\in(\mathbb{Z}/d\mathbb{Z})^s:\sum_{j=1}^{s}m_jb_j=0\text{ in }\mathbb{Z}/d\mathbb{Z}\right\}.

Under this isomorphism, (b1,…,bs)(b_1,\ldots,b_s) corresponds to

∑j=1sbj∑αi∈Oj(Ri−∞).\sum_{j=1}^{s}b_j\sum_{\alpha_i\in O_j}(R_i-\infty).

This result reduces the computation of the rational classes generated by ramification differences to a single congruence involving the sizes of the Galois orbits. In particular, it shows precisely how the arithmetic symmetry of the branch locus affects this distinguished subgroup.

Corollary 1.3. [The transitive case] If GQG_{\mathbb{Q}} acts transitively on the roots of ff, then

RdGQ≅{b∈Z/dZ:nb=0}.R_d^{G_{\mathbb{Q}}} \cong\{b \in\mathbb{Z}/d\mathbb{Z} : nb = 0\}.

In particular, if gcd⁡(d,n)=1\gcd(d,n) = 1, then

RdGQ=0.R_d^{G_{\mathbb{Q}}} = 0.

This is a statement about the ramification-supported subgroup only; it does not, by itself, imply that J(Q)[d]=0J(\mathbb{Q})[d] = 0.

Good reduction and rational torsion. The structural results above describe a distinguished part of the rational dd-torsion, but they do not control every rational torsion class. To obtain information about the full group J(Q)torsJ(\mathbb{Q})_{\mathrm{tors}}, we use reduction at primes of good reduction. The underlying principle is standard: if pp is a prime of good reduction, reduction is injective on the torsion subgroup of order prime to pp; see, for example, Hindry–Silverman [6 Theorem C.1.4]. Thus a rational torsion point of prime order ℓ≠p\ell\ne p forces ℓ\ell to divide #J(Fp)\#J(\mathbb{F}_p). Comparing the orders of reduced Jacobians at several primes can therefore eliminate possible prime divisors of rational torsion.

The precise criterion and a convenient sufficient condition are proved in § 7.

Theorem 1.4. [Good-reduction criterion for torsion vanishing] Let p1,…,prp_1,\ldots,p_r be primes of good reduction for JJ. Suppose that, for every prime number ℓ\ell, there is an index i∈{1,…,r}i \in\{1,\ldots,r\} such that

pi≠ℓandℓ∤#J(Fpi).p_i \ne\ell\qquad\text{and}\qquad\ell\nmid\#J(\mathbb{F}_{p_i}).

Then

J(Q)tors=0.J(\mathbb{Q})_{\mathrm{tors}} = 0.

Corollary 1.5. [A gcd criterion for torsion vanishing] Let p1,…,prp_1,\ldots,p_r be distinct primes of good reduction for JJ, where r≥2r \ge2. Suppose that

gcd⁡(#J(Fp1),…,#J(Fpr))=1\gcd\bigl(\#J(\mathbb{F}_{p_1}),\ldots,\#J(\mathbb{F}_{p_r})\bigr) = 1

and

pi∤#J(Fpj)whenever i≠j.p_i \nmid\#J(\mathbb{F}_{p_j}) \qquad\text{whenever } i \ne j.

Then

J(Q)tors=0.J(\mathbb{Q})_{\mathrm{tors}} = 0.

The theorem gives a prime-by-prime elimination criterion, while the corollary packages it into conditions on finitely many group orders. These results are independent of any assumption that the rational dd-torsion is generated by ramification classes. When the orders #J(Fp)\#J(\mathbb{F}_p) are obtained from point counts or Frobenius polynomials, the criterion becomes an explicit method for proving torsion vanishing.

Geometric restrictions from divisors. A complementary question is how a divisor class can be represented on the curve. Fix the rational point ∞\infty. Riemann–Roch gives effective representatives of degree at most gg for classes in Pic⁡0(C)\operatorname{Pic}^{0}(C), while effective divisors of degree r<gr < g map under the Abel–Jacobi map from Sym⁡r(C)\operatorname{Sym}^{r}(C) into a proper subvariety of JJ. These observations do not classify torsion points, but they restrict the geometry of classes with small representatives. The relevant results are established in § 8.

Proposition 1.6. [Effective representatives of degree at most the genus] Every class ξ∈Pic⁡0(C)\xi\in\operatorname{Pic}^{0}(C) can be written as

ξ=[E−r∞].\xi= [E-r\infty].

where EE is an effective divisor of degree rr for some 0≤r≤g0 \le r \le g. More generally, if ξ\xi is represented by a divisor DD of degree zero defined over Q\mathbb{Q}, then there exist an integer rr, with 0≤r≤g0 \le r \le g, and an effective divisor EE of degree rr defined over Q\mathbb{Q} such that

ξ=[E−r∞].\xi= [E-r\infty].

Proposition 1.7. [Finiteness of rational points mapping to torsion] Assume g(C)≥2g(C) \ge2. There are only finitely many points P∈C(Q)P \in C(\mathbb{Q}) for which

[P−∞]∈J(Q)[P-\infty] \in J(\mathbb{Q})

is a torsion point. This follows from the Manin–Mumford theorem [11].

Theorem 1.8. [Low-degree divisor constraint] Let 0<r<g0 < r < g, and let E1,E2E_1, E_2 be distinct effective divisors of degree rr. If

[E1−r∞]=[E2−r∞]in J,[E_1-r\infty]=[E_2-r\infty]\quad\text{in }J,

then the complete linear system ∣E1∣|E_1| has positive dimension. Equivalently,

ℓ(E1)≥2.\ell(E_1)\ge2.

The dimension bound for the Abel–Jacobi image and this condition on linear systems are geometric restrictions rather than a complete classification of torsion classes. They complement the reduction criteria by addressing the shape of divisor representatives instead of the prime divisors of their orders.

Examples and organization. Section 7 illustrates how the results interact in explicit examples.

For

C:y3=x5−2,C:\quad y^{3}=x^{5}-2,

the polynomial x5−2x^{5}-2 is irreducible over Q\mathbb{Q}, so its roots form a single Galois orbit. Corollary 1.3 gives R3GQ=0R_{3}^{G_{\mathbb{Q}}}=0. This alone does not settle the full rational torsion, but the finite-field computations at the good primes 7 and 11, together with Corollary 1.5, give

J(Q)tors=0.J(\mathbb{Q})_{\mathrm{tors}}=0.

For the curve

C:y3=(x2−2)(x2−3),C:\quad y^{3}=(x^{2}-2)(x^{2}-3),

the roots form two Galois orbits of size 2. By Theorem 1.2, the Galois-invariant part of the ramification subgroup is

R3GQ≅Z/3Z.R_{3}^{G_{\mathbb{Q}}}\cong\mathbb{Z}/3\mathbb{Z}.

To determine the full rational 3-torsion in this example, we use the Prym variety associated with the involution (x,y)↦(−x,y)(x,y)\mapsto(-x,y). The resulting decomposition of the Jacobian, together with the Galois-module description of Prym 3-torsion, yields

J(Q)[3]≅Z/3Z;J(\mathbb{Q})[3]\cong\mathbb{Z}/3\mathbb{Z};

see Section 7.2 and [9 §7.2].

The paper is organized as follows. § 4 fixes notation and conventions. § 5 recalls the background on superelliptic curves, their ramification, principal divisors, good reduction, and Weierstrass points. § 6 proves the structural results for the ramification subgroup and computes its Galois invariants. § 7 establishes the reduction-theoretic criteria for rational torsion vanishing. § 8 develops the geometric constraints from effective divisors and low-degree linear systems. § 9 gives the two examples, and § 10 concludes with further directions.

Acknowledgements

The author gratefully acknowledges the University Grants Commission (UGC), Government of India, for financial support. The author thanks Prof. Kalyan Banerjee for valuable guidance and constructive comments. The author also thanks the Department of Mathematics, SRM University-AP, for providing a supportive research environment.

AI Declaration

ChatGPT was used solely to assist with the literature review and to check grammatical errors in the manuscript. All mathematical ideas, proofs, calculations, and the writing of the manuscript were carried out independently by the author. The author takes full responsibility for the mathematical correctness, originality, and content of the paper.

Relevant Notations and Conventions

Throughout the paper, we use the following notation and conventions.

  • Base field and Galois groups. The base field is Q\mathbb{Q}. We fix an algebraic closure Q‾\overline{\mathbb{Q}} and write

GQ=Gal⁡(Q‾/Q)G_{\mathbb{Q}}=\operatorname{Gal}(\overline{\mathbb{Q}}/\mathbb{Q})

for the absolute Galois group. If f(x)∈Z[x]f(x) \in\mathbb{Z}[x] is separable, then Gal⁡(f/Q)\operatorname{Gal}(f/\mathbb{Q}) denotes the Galois group of its splitting field over Q\mathbb{Q}, viewed as a permutation group on the roots of ff. For a prime pp, we write Ip⊆GQI_p \subseteq G_{\mathbb{Q}} for an inertia subgroup at pp.

  • Group invariants. If a group GG acts on an abelian group MM, we write

MG={m∈M:σ(m)=m for every σ∈G}M^G=\{m \in M:\sigma(m)=m\text{ for every }\sigma\in G\}

for the subgroup of GG-invariant elements.

  • The superelliptic curve. Let

C:yd=f(x)C:y^d=f(x)

be a smooth projective curve over Q\mathbb{Q}, where d≥3d \ge3 is odd and f(x)∈Z[x]f(x) \in\mathbb{Z}[x] is separable of degree n≥3n \ge3. We assume gcd⁡(d,n)=1\gcd(d,n)=1 and write g=g(C)≥2g=g(C) \ge2 for the genus. The natural cyclic covering is denoted by

π:C⟶P1,(x,y)⟼x.\pi:C\longrightarrow\mathbb{P}^1,\qquad(x,y)\longmapsto x.

Under these hypotheses, the covering has degree dd and is totally ramified above the roots of ff and at infinity.

  • The cyclic action. We write μd\mu_d for the group of dd-th roots of unity in Q‾\overline{\mathbb{Q}}. The automorphism

(x,y)⟼(x,ζy),ζ∈μd,(x,y)\longmapsto(x,\zeta y),\qquad\zeta\in\mu_d,

defines an action of μd\mu_d on CQ‾C_{\overline{\mathbb{Q}}} and hence on its Jacobian. This action is understood over Q‾\overline{\mathbb{Q}}; it need not be defined over Q\mathbb{Q} for every ζ∈μd\zeta\in\mu_d.

  • The Jacobian and its torsion. We denote the Jacobian of CC by

J=Jac⁡(C).J = \operatorname{Jac}(C).

For a positive integer mm, the geometric mm-torsion subgroup is

J[m]={P∈J(Q‾):mP=0}.J[m] = \{P \in J(\overline{\mathbb{Q}}) : mP = 0\}.

The rational mm-torsion subgroup is

J(Q)[m]={P∈J(Q):mP=0}=J[m]GQ.J(\mathbb{Q})[m] = \{P \in J(\mathbb{Q}) : mP = 0\} = J[m]^{G_{\mathbb{Q}}}.

We write J(Q)torsJ(\mathbb{Q})_{\mathrm{tors}} for the torsion subgroup of J(Q)J(\mathbb{Q}). For a prime ℓ\ell, we denote its ℓ\ell-primary subgroup by J(Q)tors(ℓ)J(\mathbb{Q})_{\mathrm{tors}}(\ell).

  • Ramification points and the ramification subgroup. Let α1,…,αn∈Q‾\alpha_1,\ldots,\alpha_n \in\overline{\mathbb{Q}} be the distinct roots of f(x)f(x). For each ii, let

Ri=(αi,0)∈C(Q‾)R_i = (\alpha_i,0) \in C(\overline{\mathbb{Q}})

be the corresponding ramification point. Since gcd⁡(d,n)=1\gcd(d,n)=1, there is a unique point ∞\infty above infinity, and it is defined over Q\mathbb{Q}. We write

ri=[Ri−∞]∈J(Q‾).r_i = [R_i-\infty] \in J(\overline{\mathbb{Q}}).

The principal-divisor relation

div⁡(x−αi)=dRi−d∞\operatorname{div}(x-\alpha_i)=dR_i-d\infty

implies that ri∈J[d]r_i \in J[d]. We define the geometric ramification subgroup by

Rd=⟨r1,…,rn⟩⊆J[d].R_d=\langle r_1,\ldots,r_n\rangle\subseteq J[d].

The absolute Galois group acts on RdR_d by permuting the ramification classes. In particular,

RdGQ⊆J(Q)[d].R_d^{G_{\mathbb{Q}}} \subseteq J(\mathbb{Q})[d].

The inclusion is not assumed to be an equality: the ramification subgroup need not account for all rational dd-torsion.

  • Divisors and divisor classes. For a nonzero rational function hh on a curve, div⁡(h)\operatorname{div}(h) denotes its principal divisor. We write D1∼D2D_1 \sim D_2 when two divisors are linearly equivalent. Whenever a divisor class is regarded as an element of JJ, it is understood to have degree zero. If a divisor involves points defined only over Q‾\overline{\mathbb{Q}}, its divisor class is initially regarded as a geometric class in J(Q‾)J(\overline{\mathbb{Q}}).

  • Good reduction. A prime pp is a prime of good reduction for CC if CC has a smooth proper model over Zp\mathbb{Z}_p. In this case, JJ has good reduction at pp, and its special fiber is the Jacobian of the reduced curve over Fp\mathbb{F}_p. We denote the reduction map on rational points by

ρp:J(Q)⟶J(Fp).\rho_p : J(\mathbb{Q}) \longrightarrow J(\mathbb{F}_p).

We use this map to constrain rational torsion by comparing the orders of Jacobians over finite fields.

  • Weierstrass points. For a smooth projective curve of genus g≥2g \ge2, a geometric point P∈C(Q‾)P \in C(\overline{\mathbb{Q}}) is a Weierstrass point if its gap sequence differs from the generic sequence {1,2,…,g}\{1,2,\ldots,g\}. We write W(C)W(C) for the set of geometric Weierstrass points. This set is finite and satisfies the classical bound

#W(C)≤g3−g;\#W(C) \le g^3-g;

see [1].

  • Genus formula. Since gcd⁡(d,n)=1\gcd(d,n)=1, the Riemann–Hurwitz formula gives

g(C)=(d−1)(n−1)2.g(C)=\frac{(d-1)(n-1)}{2}.
  • Symmetric powers and the Abel–Jacobi map. For r≥1r \ge1, we write Sym⁡r(C)\operatorname{Sym}^r(C) for the rr-th symmetric power of CC, which parametrizes effective divisors of degree rr. Using the rational point ∞\infty, we define the Abel–Jacobi map

ι:C⟶J,P⟼[P−∞].\iota:C\longrightarrow J,\qquad P\longmapsto[P-\infty].

More generally, an effective divisor EE of degree rr determines the class [E−r∞]∈J[E-r\infty]\in J.

Background theory

In this section, we collect the geometric and arithmetic facts used in the subsequent analysis. We describe the ramification of the cyclic covering and the resulting genus formula, record the principal divisor relations associated with the covering, and recall the reduction property of Jacobians and the classical bound for Weierstrass points.

Superelliptic curves and ramification

Let

C:yd=f(x),C:\quad y^d=f(x),

where f∈Q[x]f\in\mathbb{Q}[x] is separable of degree nn, d≥3d\ge3 is odd, and gcd⁡(d,n)=1\gcd(d,n)=1. We write CC for the smooth projective model of this curve and set J=Jac⁡(C)J=\operatorname{Jac}(C). The natural projection

π:C⟶P1,(x,y)⟼x\pi:C\longrightarrow\mathbb{P}^1,\qquad(x,y)\longmapsto x

is a morphism of degree dd.

Proposition 3.1 (Ramification structure). Let α1,…,αn\alpha_1,\ldots,\alpha_n be the roots of ff in Q‾\overline{\mathbb{Q}}, and put

Ri=(αi,0).R_i=(\alpha_i,0).

The map π\pi is totally ramified with ramification index dd above each αi\alpha_i. Since gcd⁡(d,n)=1\gcd(d,n)=1, there is a unique point ∞∈C(Q)\infty\in C(\mathbb{Q}) above infinity, and the map is also totally ramified at that point.

Proof. At a finite point of the affine curve, the equation is

F(x,y)=yd−f(x)=0.F(x,y)=y^d-f(x)=0.

Since the curve is smooth in characteristic zero, the ramification points of the projection to the xx-line occur precisely where ∂F/∂y=dyd−1=0\partial F/\partial y=dy^{d-1}=0. Thus y=0y=0, and the defining equation then gives f(x)=0f(x)=0. The finite ramification points are therefore exactly the points RiR_i. Because ff is separable, near RiR_i we can write

f(x)=(x−αi)u(x),u(αi)≠0.f(x) = (x-\alpha_i)u(x), \qquad u(\alpha_i) \ne0.

Consequently,

yd=(x−αi)u(x),y^d = (x-\alpha_i)u(x),

so yy is a local parameter at RiR_i and

vRi(x−αi)=d.v_{R_i}(x-\alpha_i) = d.

The ramification index at RiR_i is therefore dd.

At infinity, the pole orders of xx and yy are determined by the equation yd=f(x)y^d=f(x). Since gcd⁡(d,n)=1\gcd(d,n)=1, there is a single point above infinity, with pole orders dd for xx and nn for yy. The ramification index there is dd. These statements also follow from the standard ramification theory of superelliptic curves; see [13 Chapter III].

Proposition 3.2 (Genus formula). Under the hypotheses above, the genus of CC is

g(C)=(d−1)(n−1)2.g(C) = \frac{(d-1)(n-1)}{2}.

Proof. The morphism π\pi has degree dd and is totally ramified at the nn finite branch points and at infinity. Applying the Riemann–Hurwitz formula gives

2g(C)−2=d(2g(P1)−2)+(n+1)(d−1).2g(C)-2 = d(2g(\mathbb{P}^1)-2) + (n+1)(d-1).

Since g(P1)=0g(\mathbb{P}^1)=0, this becomes

2g(C)−2=−2d+(n+1)(d−1),2g(C)-2 = -2d + (n+1)(d-1),

and hence

g(C)=(d−1)(n−1)2.g(C) = \frac{(d-1)(n-1)}{2}.

See also [13 Chapter III].

Lemma 3.3 (Good reduction criterion). Suppose that f∈Z[x]f \in\mathbb{Z}[x] has degree nn, is separable over Q\mathbb{Q}, and satisfies gcd⁡(d,n)=1\gcd(d,n)=1. Let pp be a prime such that

p∤d,p∤lc⁡(f),p∤disc⁡(f).p \nmid d,\qquad p \nmid\operatorname{lc}(f),\qquad p \nmid\operatorname{disc}(f).

Then the smooth projective model of yd=f(x)y^d=f(x) has good reduction at pp, and the reduced curve has genus

(d−1)(n−1)2.\frac{(d-1)(n-1)}{2}.

Proof. The discriminant condition implies that the reduction of ff modulo pp is separable and has the same degree nn. Since p∤dp \nmid d, the affine equation

yd=f(x)y^d=f(x)

has no singular points over F‾p\overline{\mathbb{F}}_p: a singular point would satisfy

yd=f(x),f′(x)=0,dyd−1=0.y^d=f(x),\qquad f'(x)=0,\qquad dy^{d-1}=0.

The last equation forces y=0y=0, which would give f(x)=f′(x)=0f(x)=f'(x)=0, contradicting separability.

The leading-coefficient condition preserves the degree of the equation after reduction. Moreover, p∤dp \nmid d ensures that the cover is tame, and gcd⁡(d,n)=1\gcd(d,n)=1 gives the same single point above infinity as in characteristic zero. The smooth projective model therefore has the same ramification structure and genus. These conditions give the standard good-reduction criterion for this superelliptic model; see [4].

Principal divisors from the covering

The ramification points give natural torsion classes in the geometric Jacobian. The relevant relations follow directly from the divisors of x−αix-\alpha_i and yy.

Proposition 3.4. For each root αi\alpha_i of ff, the following principal-divisor identity holds on CQ‾C_{\overline{\mathbb{Q}}}:

div⁡(x−αi)=dRi−d∞.\operatorname{div}(x-\alpha_i)=dR_i-d\infty.

Consequently,

d[Ri−∞]=0in J(Q‾).d[R_i-\infty]=0 \quad\text{in } J(\overline{\mathbb{Q}}).

Moreover,

div⁡(y)=∑i=1nRi−n∞.\operatorname{div}(y)=\sum_{i=1}^{n}R_i-n\infty.

In particular,

∑i=1n[Ri−∞]=0in J(Q‾).\sum_{i=1}^{n}[R_i-\infty]=0 \quad\text{in } J(\overline{\mathbb{Q}}).

Proof. On P1\mathbb{P}^1, the divisor of x−αix-\alpha_i is

div⁡P1(x−αi)=(αi)−(∞).\operatorname{div}_{\mathbb{P}^1}(x-\alpha_i)=(\alpha_i)-(\infty).

Pulling this divisor back under π\pi and using total ramification at RiR_i and ∞\infty gives

div⁡C(x−αi)=π∗((αi)−(∞))=dRi−d∞.\begin{aligned} \operatorname{div}_{C}(x-\alpha_i) &=\pi^*((\alpha_i)-(\infty))\\ &=dR_i-d\infty. \end{aligned}

This proves the first identity and the associated dd-torsion relation.

At each RiR_i, the function yy vanishes to order one. Its only pole is at infinity, where it has pole order nn. Hence

div⁡(y)=∑i=1nRi−n∞.\operatorname{div}(y)=\sum_{i=1}^{n}R_i-n\infty.

Passing to divisor classes proves the final relation.

Néron models and reduction of torsion

We recall the reduction property used in Section 5.

Theorem 3.5 (Injectivity on prime-to-pp torsion). Let pp be a prime of good reduction for CC. The reduction map

ρp:J(Q)⟶J(Fp)\rho_p:J(\mathbb{Q})\longrightarrow J(\mathbb{F}_p)

is injective on the subgroup of rational torsion points whose orders are relatively prime to pp.

Proof. Since CC has good reduction at pp, its Jacobian has an abelian scheme as its good integral model over Zp\mathbb{Z}_p. The kernel of reduction on Qp\mathbb{Q}_p-points has no nontrivial torsion of order prime to pp. Thus a torsion point of order prime to pp that reduces to zero must itself be zero. Restricting this property to J(Q)J(\mathbb{Q}) proves the assertion. See [6 Theorem C.1.4].

Weierstrass points

We finish by recalling the classical bound for Weierstrass points. This result is background and is not specific to superelliptic curves.

Theorem 3.6 (Weierstrass-point bound). Let CC be a smooth projective curve of genus g≥2g \ge2 over a field of characteristic zero. If W(C)W(C) denotes its set of Weierstrass points over an algebraic closure, then

#W(C)≤g3−g.\#W(C) \le g^3 - g.

Proof. The bound follows from the classical theory of Weierstrass gap sequences and the total weight formula for Weierstrass points; see [1 Chapter IV].

For comparison, a hyperelliptic curve of genus gg has exactly 2g+22g + 2 Weierstrass points. The bound above applies more generally and will only be used as background in the present work.

The divisor identities in Proposition 3.4 identify the classes [Ri−∞][R_i-\infty] as natural geometric dd-torsion classes. The reduction theorem supplies an independent arithmetic constraint on rational torsion. In the next section, we study the subgroup generated by the ramification classes and its Galois invariants, without assuming that this subgroup exhausts the rational dd-torsion of JJ.

Ramification Classes and Galois Invariants

Let

C:yd=f(x),f(x)=c∏i=1n(x−αi).C:\quad y^d=f(x),\qquad f(x)=c\prod_{i=1}^{n}(x-\alpha_i).

where f∈Q[x]f \in\mathbb{Q}[x] is separable of degree nn, d≥3d \ge3 is odd, and gcd⁡(d,n)=1\gcd(d,n)=1. We work with the smooth projective model of CC and write J=Jac⁡(C)J=\operatorname{Jac}(C). Let Ri=(αi,0)R_i=(\alpha_i,0) be the finite ramification points, and let ∞\infty denote the unique point above infinity. Throughout this section, the points RiR_i and their divisor classes are considered over Q‾\overline{\mathbb{Q}}.

Put

ri=[Ri−∞]∈J[d](Q‾)r_i=[R_i-\infty]\in J[d](\overline{\mathbb{Q}})

and define the ramification subgroup

Rd=⟨r1,…,rn⟩⊆J[d](Q‾).R_d=\langle r_1,\ldots,r_n\rangle\subseteq J[d](\overline{\mathbb{Q}}).

The purpose of this section is to determine the structure of RdR_d and describe its subgroup of Galois-invariant elements.

Principal Divisors and the Ramification Subgroup

The functions x−αix-\alpha_i and yy have divisors

div⁡(x−αi)=dRi−d∞(1≤i≤n)\operatorname{div}(x-\alpha_i)=dR_i-d\infty\qquad(1\le i\le n)

and

div⁡(y)=∑i=1nRi−n∞.\operatorname{div}(y)=\sum_{i=1}^{n}R_i-n\infty.

Consequently,

dri=0(1≤i≤n),∑i=1nri=0.dr_i=0\qquad(1\le i\le n),\qquad\sum_{i=1}^{n}r_i=0.

Proof of Lemma 1.1. Since ∑i=1nri=0\sum_{i=1}^{n} r_i = 0, we have Δ⊆ker⁡Φ\Delta\subseteq\ker\Phi. We prove the reverse inclusion.

Let (a1,…,an)∈ker⁡Φ(a_1,\ldots,a_n) \in\ker\Phi, and choose representatives ai∈{0,…,d−1}a_i \in\{0,\ldots,d-1\}. Then the divisor

A=∑i=1naiRi−(∑i=1nai)∞A = \sum_{i=1}^{n} a_iR_i-\left(\sum_{i=1}^{n}a_i\right)\infty

is principal. Hence there exists h∈Q‾(C)×h \in\overline{\mathbb{Q}}(C)^{\times} such that

div⁡(h)=A.\operatorname{div}(h)=A.

We first observe that the covering π:C→P1\pi:C\to\mathbb{P}^{1} is geometrically cyclic of degree dd. Indeed, since ff is separable, it has a simple root, say α1\alpha_1. Writing f(x)=(x−α1)g(x)f(x)=(x-\alpha_1)g(x), with g(α1)≠0g(\alpha_1)\ne0, we see that Td−f(x)T^d-f(x) is Eisenstein at x−α1x-\alpha_1 in Q‾[x][T]\overline{\mathbb{Q}}[x][T]. Consequently,

[Q‾(C):Q‾(x)]=d.[\overline{\mathbb{Q}}(C):\overline{\mathbb{Q}}(x)]=d.

Moreover, fixing a primitive dd-th root of unity ζd∈Q‾\zeta_d\in\overline{\mathbb{Q}}, the automorphism

σ(x,y)=(x,ζdy)\sigma(x,y)=(x,\zeta_dy)

generates the Galois group

Gal⁡(Q‾(C)/Q‾(x))≅Z/dZ.\operatorname{Gal}(\overline{\mathbb{Q}}(C)/\overline{\mathbb{Q}}(x))\cong\mathbb{Z}/d\mathbb{Z}.

In particular, we have the eigenspace decomposition

Q‾(C)=⨁k=0d−1Q‾(x)yk,\overline{\mathbb{Q}}(C)=\bigoplus_{k=0}^{d-1}\overline{\mathbb{Q}}(x)y^k,

where σ\sigma acts on the kk-th eigenspace by multiplication by ζdk\zeta_d^k. Since σ\sigma fixes each of the points R1,…,Rn,∞R_1,\ldots,R_n,\infty, we obtain

div⁡(σ(h))=σ(div⁡(h))=A=div⁡(h).\operatorname{div}(\sigma(h))=\sigma(\operatorname{div}(h))=A=\operatorname{div}(h).

Thus σ(h)/h\sigma(h)/h has trivial divisor and is therefore a nonzero constant. Hence hh is an eigenfunction for σ\sigma. By the above eigenspace decomposition, there exist u(x)∈Q‾(x)×u(x)\in\overline{\mathbb{Q}}(x)^{\times} and k∈{0,…,d−1}k\in\{0,\ldots,d-1\} such that

h=u(x)yk.h=u(x)y^k.

For each ii, the point RiR_i is totally ramified under π\pi, and hence

vRi(u(x))≡0(modd)v_{R_i}(u(x))\equiv0\pmod d

for every u(x)∈Q‾(x)×u(x)\in\overline{\mathbb{Q}}(x)^{\times}. Since vRi(y)=1v_{R_i}(y)=1, it follows that

ai=vRi(h)=vRi(u(x))+k≡k(modd).a_i=v_{R_i}(h)=v_{R_i}(u(x))+k\equiv k\pmod d.

As this holds for every ii, we conclude that

a1≡a2≡⋯≡an(modd).a_1\equiv a_2\equiv\cdots\equiv a_n\pmod d.

Therefore (a1,…,an)∈Δ(a_1,\ldots,a_n)\in\Delta, proving that ker⁡Φ⊆Δ\ker\Phi\subseteq\Delta. Hence ker⁡Φ=Δ\ker\Phi=\Delta, and the first isomorphism theorem yields

Rd≅(Z/dZ)n/Δ≅(Z/dZ)n−1.R_d\cong(\mathbb{Z}/d\mathbb{Z})^n/\Delta\cong(\mathbb{Z}/d\mathbb{Z})^{n-1}.

Remark 4.1. Because gcd⁡(d,n)=1\gcd(d,n)=1, every class in (Z/dZ)n/Δ(\mathbb{Z}/d\mathbb{Z})^n/\Delta has a unique representative (a1,…,an)(a_1,\ldots,a_n) satisfying

∑i=1nai=0in Z/dZ.\sum_{i=1}^{n}a_i=0\quad\text{in }\mathbb{Z}/d\mathbb{Z}.

Indeed, adding the diagonal vector (t,…,t)(t,\ldots,t) changes the coordinate sum by ntnt, and multiplication by nn is invertible modulo dd. In particular, the subgroup RdR_d can be represented uniquely by sum-zero coefficient vectors.

Galois Invariants of the Ramification Subgroup

The absolute Galois group

GQ=Gal⁡(Q‾/Q)G_{\mathbb{Q}}=\operatorname{Gal}(\overline{\mathbb{Q}}/\mathbb{Q})

permutes the roots of ff and hence the ramification points RiR_i. It therefore acts on RdR_d. Write the decomposition of the roots into Galois orbits as

{α1,…,αn}=O1⊔⋯⊔Os,mj=∣Oj∣.\{\alpha_1,\ldots,\alpha_n\}=O_1\sqcup\cdots\sqcup O_s,\qquad m_j=|O_j|.

Proof of Theorem 1.2. By Lemma 1.1 and Remark 4.1, every element of RdR_d has a unique coefficient-vector representative whose coordinate sum is zero modulo dd. The Galois group permutes the coordinates of this representative.

Suppose that a class in RdR_d is Galois invariant. For every γ∈GQ\gamma\in G_{\mathbb{Q}}, its coefficient vector and the vector obtained by applying γ\gamma represent the same class. Both have coordinate sum zero, so uniqueness implies that the two vectors are equal. Hence the coefficients are constant on each Galois orbit OjO_j. Write the common coefficient on OjO_j as bjb_j. The sum-zero condition becomes

∑j=1smjbj=0in Z/dZ.\sum_{j=1}^{s}m_jb_j=0\quad\text{in }\mathbb{Z}/d\mathbb{Z}.

Conversely, any vector (b1,…,bs)(b_1,\ldots,b_s) satisfying this congruence defines a coefficient vector that is constant on each Galois orbit and whose coordinate sum is zero. It therefore defines a Galois-invariant class in RdR_d. The uniqueness of the sum-zero representative proves that the correspondence is bijective, and it is plainly a homomorphism.

Proof of Corollary 1.3. There is a single Galois orbit, of size nn. Theorem 1.2 therefore reduces the invariant condition to nb=0nb=0 modulo dd. Since nn is invertible modulo dd, this forces b=0b=0.

Relation with Rational dd-Torsion

The subgroup RdR_d is stable under the Galois action, but its individual generators need not be rational. Since

J(Q)[d]=J[d](Q‾)GQ,J(\mathbb{Q})[d]=J[d](\overline{\mathbb{Q}})^{G_{\mathbb{Q}}},

we have the inclusion

RdGQ⊆J(Q)[d].R_d^{G_{\mathbb{Q}}}\subseteq J(\mathbb{Q})[d].

Theorem 1.2 thus gives an explicit description of the rational dd-torsion classes that are generated by the ramification differences.

It does not, by itself, describe all of J(Q)[d]J(\mathbb{Q})[d]. A rational dd-torsion class may lie outside RdR_d. Consequently, the vanishing of RdGQR_d^{G_{\mathbb{Q}}} does not imply the vanishing of J(Q)[d]J(\mathbb{Q})[d] without an additional argument controlling the quotient J[d](Q‾)/RdJ[d](\overline{\mathbb{Q}})/R_d.

The Covering Automorphism and the Base Field

The automorphism

(x,y)⟼(x,ζdy)(x,y)\longmapsto(x,\zeta_dy)

is generally defined over Q(ζd)\mathbb{Q}(\zeta_d), rather than over Q\mathbb{Q}. Accordingly, a decomposition into character spaces for the μd\mu_d-action must be formulated over a field containing the dd-th roots of unity, and the action of GQG_{\mathbb{Q}} on those character spaces must be treated separately.

In particular, individual nontrivial character spaces cannot be assumed to be GQG_{\mathbb{Q}}-stable without proof. No isotypic-vanishing hypothesis is needed for the results established in this section: the conclusions above concern only the ramification-generated subgroup and its Galois invariants.

Reduction-theoretic constraints on rational torsion

The ramification and Galois-theoretic analysis of the preceding section describes a subgroup of the rational torsion of the Jacobian. We now complement that analysis with reduction-theoretic constraints on the full rational torsion subgroup Jac⁡(C)(Q)tors\operatorname{Jac}(C)(\mathbb{Q})_{\mathrm{tors}}.

The basic principle is that, at a prime of good reduction, the reduction map is injective on torsion of order prime to the residue characteristic. Consequently, the order of a rational torsion point imposes divisibility conditions on the finite groups Jac⁡(C)(Fp)\operatorname{Jac}(C)(\mathbb{F}_p). Comparing these groups at several primes can therefore eliminate possible prime divisors of rational torsion.

In Section 4, we obtained information about the Galois-invariant part of the ramification subgroup. The results below are complementary: they constrain the full rational torsion subgroup and do not require a hypothesis identifying rational dd-torsion with the ramification subgroup.

Good reduction and torsion constraints

Let J=Jac⁡(C)J=\operatorname{Jac}(C), and let pp be a prime of good reduction for JJ. Write

J(Q)tors,p′={P∈J(Q)tors:p∤ord⁡(P)}.J(\mathbb{Q})_{\mathrm{tors},p'}=\{P\in J(\mathbb{Q})_{\mathrm{tors}}:p\nmid\operatorname{ord}(P)\}.

The following standard result will be used repeatedly.

Proposition 5.1 (Good reduction and prime-to-pp torsion). Let pp be a prime of good reduction for JJ. The reduction map

red⁡p:J(Q)tors,p′⟶J(Fp)\operatorname{red}_p:J(\mathbb{Q})_{\mathrm{tors},p'}\longrightarrow J(\mathbb{F}_p)

is injective.

Proof. This is the standard injectivity of reduction on torsion of order prime to the residue characteristic; see Hindry–Silverman [6 Theorem C.1.4].

We next record the numerical consequence of this proposition. Let P∈J(Q)torsP\in J(\mathbb{Q})_{\mathrm{tors}} have order N=ord⁡(P)N=\operatorname{ord}(P). Write

N=pvp(N)Np′,p∤Np′,N=p^{v_p(N)}N_{p'},\qquad p\nmid N_{p'},

where Np′N_{p'} is the prime-to-pp part of NN. The point

Pp′=pvp(N)PP_{p'}=p^{v_p(N)}P

has order Np′N_{p'}. By Proposition 5.1, its reduction has the same order. It follows that

Np′∣#J(Fp).N_{p'}\mid\#J(\mathbb{F}_p).

In particular, if ℓ\ell is a prime divisor of ord⁡(P)\operatorname{ord}(P) and ℓ≠p\ell\ne p, then

ℓ∣#J(Fp).\ell\mid\#J(\mathbb{F}_p).

Thus every prime divisor of the order of a rational torsion point must divide the order of the reduced Jacobian at every chosen prime of good reduction different from that torsion prime.

Elimination of torsion primes by good reduction

The preceding divisibility condition yields a criterion for proving that the rational torsion subgroup is trivial. Proof of Theorem 1.4. Suppose, to the contrary, that J(Q)tors≠0J(\mathbb{Q})_{\mathrm{tors}} \ne0. Choose a nonzero torsion point P∈J(Q)torsP \in J(\mathbb{Q})_{\mathrm{tors}}, and let ℓ\ell be a prime divisor of ord⁡(P)\operatorname{ord}(P). The point

Q=ord⁡(P)ℓPQ = \frac{\operatorname{ord}(P)}{\ell}P

is nonzero and has order ℓ\ell.

By hypothesis, there is an index ii such that

pi≠ℓandℓ∤#J(Fpi).p_i \ne\ell\qquad\text{and} \qquad\ell\nmid\#J(\mathbb{F}_{p_i}).

Since QQ has order ℓ≠pi\ell\ne p_i, its reduction modulo pip_i has the same order ℓ\ell, by Proposition 5.1. Therefore

ℓ∣#J(Fpi),\ell\mid\#J(\mathbb{F}_{p_i}),

a contradiction. Hence J(Q)tors=0J(\mathbb{Q})_{\mathrm{tors}} = 0.

The hypothesis of Theorem 1.4 can often be verified by computing the orders of the reduced Jacobians. The following convenient sufficient condition accounts explicitly for the possibility that a candidate torsion prime equals one of the chosen residue characteristics.

Proof of Corollary 1.5. We verify the hypothesis of Theorem 1.4 for every prime ℓ\ell. First suppose that ℓ∉{p1,…,pr}\ell\notin\{p_1,\ldots,p_r\}. If ℓ\ell divided #J(Fpi)\#J(\mathbb{F}_{p_i}) for every ii, then it would divide their gcd, contradicting the assumption that this gcd is 1. Hence there is an index ii such that

ℓ∤#J(Fpi).\ell\nmid\#J(\mathbb{F}_{p_i}).

Since ℓ≠pi\ell\ne p_i, this is the required index.

Now suppose that ℓ=pk\ell= p_k for some kk. Since r≥2r \ge2, choose j≠kj \ne k. The second hypothesis gives

ℓ=pk∤#J(Fpj),\ell= p_k \nmid\#J(\mathbb{F}_{p_j}),

and pj≠ℓp_j \ne\ell. Thus the required index exists in this case as well. Theorem 1.4 now gives the result.

Application to the superelliptic family

Consider

C:yd=f(x),C : y^d = f(x),

where d≥3d \ge3 is odd, f(x)∈Q[x]f(x) \in\mathbb{Q}[x] is separable of degree n≥3n \ge3, and

gcd⁡(d,n)=1.\gcd(d,n) = 1.

Let J=Jac⁡(C)J = \operatorname{Jac}(C). Under these hypotheses, the smooth projective curve has a unique point at infinity, and its genus is

g=(d−1)(n−1)2.g = \frac{(d-1)(n-1)}{2}.

To apply the reduction criteria, choose an integral model of the curve and a set Pgood\mathcal{P}_{\mathrm{good}} of primes at which that model has good reduction. The set of excluded primes depends on the chosen integral model. In particular, for a model yd=f(x)y^d = f(x) with f(x)∈Z[x]f(x) \in\mathbb{Z}[x], primes dividing dd, the leading coefficient of ff, or the discriminant of ff must be excluded when using the usual smooth-model criterion. More generally, denominators introduced when passing from a rational equation to an integral model must also be taken into account. The good-reduction hypothesis should be verified for the model actually used.

For p∈Pgoodp \in\mathcal{P}_{\mathrm{good}}, let CpC_p and JpJ_p denote the reductions of CC and JJ, respectively. Set

Np=#Jp(Fp)=#J(Fp).N_p = \#J_p(\mathbb{F}_p) = \#J(\mathbb{F}_p).

These finite-field cardinalities are the numerical input for Theorem 1.4 and Corollary 1.3. Indeed, suppose that P∈J(Q)torsP \in J(\mathbb{Q})_{\mathrm{tors}} is nonzero, and let ℓ\ell be a prime divisor of ord⁡(P)\operatorname{ord}(P). Then

Q=ord⁡(P)ℓPQ = \frac{\operatorname{ord}(P)}{\ell}P

has order ℓ\ell. For any p∈Pgoodp \in\mathcal{P}_{\mathrm{good}} with p≠ℓp \ne\ell, the reduction of QQ has order ℓ\ell, so

ℓ∣Np.\ell\mid N_p.

Consequently, to prove that the rational torsion subgroup is trivial, it is enough to choose good primes p1,…,prp_1,\ldots,p_r and show that for every prime ℓ\ell there is an index ii such that

pi≠ℓandℓ∤Npi.p_i \ne\ell\qquad\text{and}\qquad\ell\nmid N_{p_i}.

In practice, one can factor the integers NpiN_{p_i} and use these factorizations to eliminate candidate torsion primes.

The cardinalities NpN_p can be computed from Frobenius data. With the standard normalization

Pp(T)=∏j=12g(1−αjT),P_p(T) = \prod_{j=1}^{2g}(1-\alpha_jT),

where the αj\alpha_j are the Frobenius eigenvalues on the first étale cohomology of CpC_p, the characteristic polynomial Pp(T)P_p(T) satisfies

#Jp(Fp)=Pp(1).\#J_p(\mathbb{F}_p) = P_p(1).

Equivalently, the required cardinality can be obtained from the Frobenius polynomial of the curve or Jacobian over Fp\mathbb{F}_p. The precise convention for arithmetic versus geometric Frobenius must be kept consistent with the polynomial normalization.

The reduction-theoretic information developed here complements the ramification analysis in § 6. That section computes the Galois-invariant part of the subgroup generated by the ramification classes and places constraints on rational dd-torsion arising from the geometry of the covering. The present section instead constrains the full rational torsion subgroup by using reductions at suitable good primes. It does not rely on an equality between the ramification subgroup and all of J[d]J[d].

The two approaches can therefore be used together. Structural information about J(Q)[d]J(\mathbb{Q})[d], when available, can be combined with the reduction constraints to eliminate possible torsion primes. If the resulting arguments rule out every prime divisor of the order of a rational torsion point, then

J(Q)tors=0.J(\mathbb{Q})_{\mathrm{tors}} = 0.

The examples below illustrate how finite-field computations can supply the reduction data needed for these criteria.

Geometric constraints on exceptional torsion

The reduction-theoretic results of § 7 constrain the possible prime divisors of J(Q)torsJ(\mathbb{Q})_{\mathrm{tors}} through good reduction. We now consider complementary geometric restrictions arising from the degrees of divisors representing classes in the Jacobian.

The key observation is that classes admitting effective representatives of small degree are controlled by the symmetric powers of CC and their images under the Abel–Jacobi map. Since the rr-th symmetric power has dimension rr, these images have dimension at most rr. In particular, when r<g=dim⁡Jr < g = \dim J, they lie in proper subvarieties of JJ. We also explain how equality between classes represented by distinct effective divisors imposes a condition on the associated complete linear system.

Throughout this section, CC is a smooth projective curve over Q\mathbb{Q} of genus gg, J=Jac⁡(C)J = \operatorname{Jac}(C), and ∞∈C(Q)\infty\in C(\mathbb{Q}) is a fixed rational point. In our superelliptic setting, the coprimality condition gcd⁡(d,deg⁡f)=1\gcd(d,\deg f) = 1 supplies such a unique point at infinity.

Reduced representatives and Abel–Jacobi constraints

We first recall the existence of effective representatives of degree at most gg.

Proof of Proposition 1.6. Let DD be a divisor of degree zero over Q\mathbb{Q} representing ξ\xi. By the Riemann–Roch theorem,

ℓ(D+g∞)−ℓ(K−D−g∞)=deg⁡(D+g∞)−g+1=1,\ell(D + g\infty) - \ell(K - D - g\infty) = \deg(D + g\infty) - g + 1 = 1,

where KK is a canonical divisor. Since ℓ(K−D−g∞)≥0\ell(K - D - g\infty) \ge0, we obtain

ℓ(D+g∞)≥1.\ell(D + g\infty) \ge1.

Thus there exists a nonzero rational function h∈Q(C)h \in\mathbb{Q}(C) such that

E=D+g∞+div⁡(h)E = D + g\infty+ \operatorname{div}(h)

is effective. The divisor EE is defined over Q\mathbb{Q} and has degree gg. Therefore,

[D]=[E−g∞],[D] = [E - g\infty],

which proves the first assertion.

For the more general statement, let rr be the smallest nonnegative integer such that

ℓ(D+r∞)>0.\ell(D + r\infty) > 0.

Such an integer exists by the first part, and r≤gr \le g. Choose a nonzero h∈Q(C)h \in\mathbb{Q}(C) such that

E=D+r∞+div⁡(h)E = D + r\infty+ \operatorname{div}(h)

is effective. By the minimality of rr, the effective divisor EE cannot contain ∞\infty; otherwise, subtracting ∞\infty would give a nonzero section of D+(r−1)∞D + (r - 1)\infty when r>0r > 0. Hence EE has degree rr, and

[D]=[E−r∞].[D] = [E - r\infty].

This completes the proof.

The proposition gives an effective representative of degree gg. It does not assert that every class has a representative of degree strictly less than gg. Whenever a class has a representative [E−r∞][E - r\infty] with r<gr < g, the Abel–Jacobi map places that class in a proper subvariety of JJ, as explained below.

Proof of Proposition 1.7. Consider the Abel–Jacobi embedding

ι:C⟶J,P⟼[P−∞].\iota: C \longrightarrow J,\qquad P \longmapsto[P - \infty].

Since g≥2g \ge2, the image ι(C)\iota(C) is not a translate of a positive-dimensional abelian subvariety of JJ. By the Manin–Mumford theorem [11], the intersection of ι(C)\iota(C) with the set J(Q‾)torsJ(\overline{\mathbb{Q}})_{\mathrm{tors}} of geometric torsion points is finite. Consequently, its subset consisting of images of rational points is also finite. This proves the proposition.

Low-degree linear systems

The next result gives a necessary condition for two distinct effective divisors of small degree to determine the same class in the Jacobian.

Proof of Theorem 1.8. The equality of divisor classes gives

E1−r∞∼E2−r∞,E_1-r\infty\sim E_2-r\infty,

and hence E1∼E2E_1\sim E_2. Thus E1E_1 and E2E_2 are distinct effective members of the complete linear system ∣E1∣\lvert E_1\rvert. Therefore ∣E1∣\lvert E_1\rvert contains at least two distinct members, so

dim⁡∣E1∣≥1.\dim\lvert E_1\rvert\ge1.

Since dim⁡∣E1∣=ℓ(E1)−1\dim\lvert E_1\rvert=\ell(E_1)-1, we conclude that ℓ(E1)≥2\ell(E_1)\ge2.

Remark 6.1. For 0<r<g0<r<g, effective divisors of degree rr are parametrized by the symmetric power Sym⁡r(C)\operatorname{Sym}^r(C), which has dimension rr. The Abel–Jacobi map

ur:Sym⁡r(CQ‾)⟶JQ‾,E⟼[E−r∞]u_r:\operatorname{Sym}^r(C_{\overline{\mathbb{Q}}})\longrightarrow J_{\overline{\mathbb{Q}}},\qquad E\longmapsto[E-r\infty]

therefore has image of dimension at most r<g=dim⁡Jr<g=\dim J. Consequently, the geometric classes admitting effective representatives of degree r<gr<g lie in a proper subvariety of JQ‾J_{\overline{\mathbb{Q}}}.

This dimension bound is a geometric restriction; it does not imply that the image contains only finitely many torsion points. Theorem 1.8 provides an additional restriction: if two distinct effective divisors of the same degree r<gr<g represent the same divisor class, then their common complete linear system has positive dimension. These observations constrain the geometry of divisor representatives but do not, by themselves, classify the rational torsion subgroup J(Q)torsJ(\mathbb{Q})_{\mathrm{tors}}.

The geometric constraints in this section complement the reduction-theoretic criteria of § 7. Good reduction restricts possible prime divisors of rational torsion through the finite groups J(Fp)J(\mathbb{F}_p), whereas the Abel–Jacobi map and the geometry of linear systems constrain classes admitting representatives of small degree. Neither method alone establishes a general classification of J(Q)torsJ(\mathbb{Q})_{\mathrm{tors}}, but they provide distinct restrictions that may be combined when studying a specific curve.

Examples

We illustrate the structural, reduction-theoretic, and geometric constraints developed in the preceding sections through two explicit superelliptic curves. The first example has a transitive Galois action on the roots of the defining polynomial and admits a reduction-theoretic argument for the vanishing of the full rational torsion subgroup. The second example has two Galois orbits of roots and illustrates how the orbit decomposition determines the Galois-invariant part of the ramification subgroup.

A transitive genus-4 example

Consider the superelliptic curve

C:y3=x5−2C:\quad y^3=x^5-2

over Q\mathbb{Q}, and let J=Jac⁡(C)J=\operatorname{Jac}(C).

Since gcd⁡(3,5)=1\gcd(3,5)=1, the genus formula gives

g(C)=(3−1)(5−1)2=4.g(C)=\frac{(3-1)(5-1)}{2}=4.

Moreover, f(x)=x5−2f(x) = x^5 - 2 is Eisenstein at 2, and hence irreducible over Q\mathbb{Q}. Thus the Galois group acts transitively on its five roots. There is a single Galois orbit of size m1=5m_1 = 5.

By Theorem 1.2, the Galois-invariant part of the ramification subgroup is described by

R3GQ≅{a∈Z/3Z:5a≡0(mod3)}.R_3^{G_{\mathbb{Q}}} \cong\{a \in\mathbb{Z}/3\mathbb{Z}: 5a \equiv0 \pmod{3}\}.

Since 5≡2(mod3)5 \equiv2 \pmod{3}, the only solution is a=0a = 0. Hence

R3GQ=0.R_3^{G_{\mathbb{Q}}} = 0.

This conclusion concerns the ramification subgroup only; it does not, by itself, imply that J(Q)[3]=0J(\mathbb{Q})[3] = 0.

We next use good reduction to constrain the full rational torsion subgroup. The discriminant of f(x)=x5−2f(x) = x^5 - 2 is

disc⁡(f)=(−1)5(5−1)/255(−2)4=2455.\operatorname{disc}(f) = (-1)^{5(5-1)/2}5^5(-2)^4 = 2^4 5^5.

The leading coefficient of ff is 1. Therefore, the primes p=7p = 7 and p=11p = 11 do not divide 3lc⁡(f)disc⁡(f)3\operatorname{lc}(f)\operatorname{disc}(f). Both are primes of good reduction for the curve.

For p∈{7,11}p \in\{7, 11\} and k≥1k \ge1, put

Nk(p)=#C(Fpk).N_k(p) = \#C(\mathbb{F}_{p^k}).

Since gcd⁡(3,5)=1\gcd(3,5) = 1, the smooth projective model has a unique point at infinity. Thus the projective point count is obtained by counting the affine solutions of

y3=x5−2y^3 = x^5 - 2

over Fpk\mathbb{F}_{p^k} and adding this point at infinity.

The point counts reported in the manuscript are

ppN1(p)N_1(p)N2(p)N_2(p)N3(p)N_3(p)N4(p)N_4(p)
78503442766
1112154133215274

Table 1.

These counts are stated to have been obtained by exhaustive finite-field enumeration and checked by the same procedure. The SageMath verification is included in Appendix § 11.

We now recover the Frobenius polynomials from these counts. For a prime pp of good reduction, let

Pp(T)=∏i=18(1−αiT),P_p(T) = \prod_{i=1}^{8}(1-\alpha_iT),

where the αi\alpha_i are the Frobenius eigenvalues, with the normalization for which

#J(Fp)=Pp(1).\#J(\mathbb{F}_p) = P_p(1).

The point-counting formula gives

Nk(p)=pk+1−∑i=18αik.N_k(p) = p^k + 1 - \sum_{i=1}^{8}\alpha_i^k.

Writing

Sk(p)=pk+1−Nk(p),S_k(p) = p^k + 1 - N_k(p),

the values Sk(p)S_k(p) are the power sums of the Frobenius eigenvalues. Newton’s identities determine the first four coefficients of Pp(T)P_p(T), and the functional equation determines the remaining coefficients.

For p=7p = 7, the point counts give

S1(7)=0,S2(7)=0,S3(7)=0,S4(7)=−364.S_1(7) = 0,\qquad S_2(7) = 0,\qquad S_3(7) = 0,\qquad S_4(7) = -364.

Newton’s identities give

e1=e2=e3=0,e4=91,e_1 = e_2 = e_3 = 0,\qquad e_4 = 91,

where eie_i denotes the ii-th elementary symmetric function of the Frobenius eigenvalues. The functional equation then yields

P7(T)=1+91T4+2401T8.P_7(T) = 1 + 91T^4 + 2401T^8.

Consequently,

#J(F7)=P7(1)=1+91+2401=2493.\#J(\mathbb{F}_7) = P_7(1) = 1 + 91 + 2401 = 2493.

Similarly, for p=11p = 11,

S1(11)=0,S2(11)=−32,S3(11)=0,S4(11)=−632.S_1(11) = 0,\qquad S_2(11) = -32,\qquad S_3(11) = 0,\qquad S_4(11) = -632.

Newton’s identities give

e1=0,e2=16,e3=0,e4=286.e_1 = 0,\qquad e_2 = 16,\qquad e_3 = 0,\qquad e_4 = 286.

Using the functional equation, we obtain

P11(T)=1+16T2+286T4+1936T6+14641T8.P_{11}(T) = 1 + 16T^2 + 286T^4 + 1936T^6 + 14641T^8.

It follows that

#J(F11)=P11(1)=1+16+286+1936+14641=16880.\begin{aligned} \#J(\mathbb{F}_{11}) &= P_{11}(1) \\ &= 1 + 16 + 286 + 1936 + 14641 \\ &= 16880. \end{aligned}

We now verify the hypotheses of Corollary 1.5. First,

gcd⁡(#J(F7),#J(F11))=gcd⁡(2493,16880)=1.\gcd(\#J(\mathbb{F}_7), \#J(\mathbb{F}_{11})) = \gcd(2493, 16880) = 1.

Moreover,

7∤16880and11∤2493.7 \nmid16880 \qquad\text{and}\qquad11 \nmid2493.

Thus the primes 7 and 11 satisfy both the gcd and the cross-divisibility conditions of the corollary. Consequently,

J(Q)tors=0.J(\mathbb{Q})_{\mathrm{tors}} = 0.

This conclusion follows from the reduction-theoretic calculation and does not depend on any additional hypothesis concerning the ramification subgroup.

A non-transitive example

Consider the curve

C:y3=(x2−2)(x2−3)=x4−5x2+6C:\quad y^3 = (x^2 - 2)(x^2 - 3) = x^4 - 5x^2 + 6

over Q\mathbb{Q}, and let JJ denote its Jacobian. The polynomial

f(x)=(x2−2)(x2−3)f(x) = (x^2 - 2)(x^2 - 3)

has two Galois orbits of roots, namely

{2,−2}and{3,−3}.\{\sqrt{2},-\sqrt{2}\}\qquad\text{and}\qquad\{\sqrt{3},-\sqrt{3}\}.

Consequently, Theorem 1.2 gives

R3Q≅{(a,b)∈(Z/3Z)2:2a+2b=0}≅Z/3Z.R_3^{\mathbb{Q}} \cong\{(a,b) \in(\mathbb{Z}/3\mathbb{Z})^2 : 2a+2b=0\} \cong\mathbb{Z}/3\mathbb{Z}.

Thus the ramification subgroup supplies a nontrivial rational subgroup of order 3. We now determine the full rational 3-torsion of JJ.

Consider the involution

ι:C⟶C,(x,y)⟼(−x,y).\iota:C\longrightarrow C,\qquad(x,y)\longmapsto(-x,y).

Its quotient is the genus-one curve

D:y3=u2−5u+6,u=x2.D:\quad y^3=u^2-5u+6,\qquad u=x^2.

The curve DD has a rational point, so its Jacobian is an elliptic curve over Q\mathbb{Q}. The change of variables

X=4y,Y=8u−20X=4y,\qquad Y=8u-20

transforms the equation of DD into the Weierstrass equation

E:Y2=X3+16.E:\quad Y^2=X^3+16.

Consequently, the Jacobian of DD is isomorphic over Q\mathbb{Q} to EE. The point (0,4)∈E(Q)(0,4)\in E(\mathbb{Q}) has order 3. Moreover, the 3-division polynomial of EE is

ψ3(X)=3X(X3+64).\psi_3(X)=3X(X^3+64).

Its only rational roots are X=0X=0 and X=−4X=-4. The latter does not give a rational point on EE, since it yields Y2=−48Y^2=-48. For X=0X=0, we obtain precisely the points (0,±4)(0,\pm4). It follows that

E(Q)[3]≅Z/3Z.E(\mathbb{Q})[3]\cong\mathbb{Z}/3\mathbb{Z}.

We next study the Prym variety associated with ι\iota. Write PP for the Prym variety of the double cover C→DC\to D, so that

P=ker⁡(1+ι∗:J⟶J)0.P=\ker(1+\iota^*:J\longrightarrow J)^0.

The curve is a member of the bielliptic Picard family

Ca,b:y3=x4+ax2+bC_{a,b}:\quad y^3=x^4+ax^2+b

with a=−5a=-5 and b=6b=6. The smoothness condition in this family is satisfied, since

16b(a2−4b)=16⋅6⋅(25−24)=96≠0.16b(a^2-4b)=16\cdot6\cdot(25-24)=96\ne0.

Thus the Prym results of Laga–Shnidman [9 §2.5, §7.2] apply to this curve. In particular, the natural map

E×P⟶JE\times P\longrightarrow J

is an isogeny with kernel of order a power of 2. It therefore induces an isomorphism on rational 3-primary torsion, and

J(Q)[3]≅E(Q)[3]⊕P(Q)[3].J(\mathbb{Q})[3]\cong E(\mathbb{Q})[3]\oplus P(\mathbb{Q})[3].

It remains to show that P(Q)[3]=0P(\mathbb{Q})[3]=0.

Let ω\omega be a primitive cube root of unity and put p=(1−ω)\mathfrak{p}=(1-\omega) in Z[ω]\mathbb{Z}[\omega]. By Lemma 7.8 of [9], the group P[p]P[\mathfrak{p}] is a two-dimensional vector space over F3\mathbb{F}_3, with basis

D1=(2,0)−(−2,0),D2=(3,0)−(−3,0).D_1=(\sqrt{2},0)-(-\sqrt{2},0),\qquad D_2=(\sqrt{3},0)-(-\sqrt{3},0).

Here the points are viewed as points on the smooth projective model of CC, and the differences denote their corresponding divisor classes in the Prym.

By Proposition 7.9 of [9], the Galois action on this basis is induced by the action on the root pairs. More explicitly, changing the sign of 2\sqrt{2} while fixing 3\sqrt{3} sends D1D_1 to −D1-D_1 and fixes D2D_2; changing the sign of 3\sqrt{3} while fixing 2\sqrt{2} fixes D1D_1 and sends D2D_2 to −D2-D_2. Consequently, the representation is

ρ(σ)=(χ2(σ)00χ3(σ)),\rho(\sigma) = \begin{pmatrix} \chi_2(\sigma) & 0 \\ 0 & \chi_3(\sigma) \end{pmatrix},

where χ2\chi_2 and χ3\chi_3 are the quadratic characters associated with Q(2)\mathbb{Q}(\sqrt{2}) and Q(3)\mathbb{Q}(\sqrt{3}), respectively. Since both characters are nontrivial, there are no nonzero Galois-invariant vectors, and hence

P[p](Q)=0.P[\mathfrak{p}](\mathbb{Q}) = 0.

Equation (7.3) of [9] gives an exact sequence of Galois modules

0⟶P[p]⟶P[3]⟶P−27[p]⟶0,0 \longrightarrow P[\mathfrak{p}] \longrightarrow P[3] \longrightarrow P_{-27}[\mathfrak{p}] \longrightarrow0,

where P−27P_{-27} is the quadratic twist of PP by Q(−3)/Q\mathbb{Q}(\sqrt{-3})/\mathbb{Q}. Under this twist, the two characters describing the action on the corresponding p\mathfrak{p}-torsion become

χ2χ−3=χ−6,χ3χ−3=χ−9=χ−1.\chi_2\chi_{-3} = \chi_{-6}, \qquad\chi_3\chi_{-3} = \chi_{-9} = \chi_{-1}.

Both characters are nontrivial. Therefore

P−27[p](Q)=0.P_{-27}[\mathfrak{p}](\mathbb{Q}) = 0.

Taking Galois invariants in the exact sequence, we conclude that

P[3](Q)=0,P[3](\mathbb{Q}) = 0,

or equivalently P(Q)[3]=0P(\mathbb{Q})[3] = 0. It follows that

J(Q)[3]≅Z/3Z.J(\mathbb{Q})[3] \cong\mathbb{Z}/3\mathbb{Z}.

In this example, the rational 33-torsion is therefore exactly the subgroup supplied by the Galois-invariant ramification classes.

Concluding remarks and further directions

In this paper, we have studied rational torsion on the Jacobians of odd superelliptic curves

C: yd=f(x),C : \ y^d = f(x),

where f∈Q[x]f \in\mathbb{Q}[x] is separable, d≥3d \ge3 is odd, and gcd⁡(d,deg⁡f)=1\gcd(d,\deg f) = 1. Our approach combines the geometry of the ramification points of the covering C→P1C \to\mathbb{P}^1, the Galois action on these points, and reduction-theoretic constraints on rational torsion.

The structural results of § 6 describe the subgroup Rd⊆J[d](Q‾)R_d \subseteq J[d](\overline{\mathbb{Q}}) generated by the classes of the ramification points relative to the point at infinity. They also give an explicit description of its Galois-invariant subgroup in terms of the Galois orbits of the roots of ff. In particular,

RdGQ⊆J(Q)[d],R_d^{G_{\mathbb{Q}}} \subseteq J(\mathbb{Q})[d],

and the orbit structure can therefore yield concrete information about rational dd-torsion. This calculation concerns the ramification-generated subgroup; without additional arguments, it does not determine all of J(Q)[d]J(\mathbb{Q})[d].

The reduction-theoretic results of § 7 provide a complementary method. At primes of good reduction, the prime-to-pp torsion of J(Q)J(\mathbb{Q}) injects into J(Fp)J(\mathbb{F}_p). Consequently, suitable finite-field point counts can eliminate individual prime-primary components of rational torsion. Under the hypotheses of Corollary 1.5, these constraints imply

J(Q)tors=0.J(\mathbb{Q})_{\mathrm{tors}}=0.

Thus, the reduction method can establish complete torsion vanishing without requiring a separate hypothesis about the ramification-generated subgroup.

The examples in § 9 illustrate how the Galois-invariant ramification subgroup and reduction-theoretic methods complement one another. For the curve

C:y3=x5−2,C:\quad y^3=x^5-2,

the transitive Galois action on the five roots implies

R3GQ=0.R_3^{G_{\mathbb{Q}}}=0.

The point counts at the good primes 7 and 11, together with the required cross-divisibility conditions, yield

J(Q)tors=0J(\mathbb{Q})_{\mathrm{tors}}=0

by Corollary 1.5.

For the curve

C:y3=(x2−2)(x2−3),C:\quad y^3=(x^2-2)(x^2-3),

the two Galois orbits of roots give

R3GQ≅Z/3Z.R_3^{G_{\mathbb{Q}}}\cong\mathbb{Z}/3\mathbb{Z}.

Thus the ramification classes provide a nontrivial rational subgroup of order 3. The Prym variety associated with the involution (x,y)↦(−x,y)(x,y)\mapsto(-x,y) allows us to go further: the quotient curve has Jacobian isomorphic to

E:Y2=X3+16,E:\quad Y^2=X^3+16,

and the Prym 3-torsion calculation shows that the Prym contributes no nontrivial rational 3-torsion. Consequently,

J(Q)[3]≅Z/3Z.J(\mathbb{Q})[3]\cong\mathbb{Z}/3\mathbb{Z}.

This example therefore illustrates how combining the orbit description of ramification classes with the arithmetic of a Prym variety can determine the full rational 3-torsion, rather than only the contribution from the ramification subgroup; see [9 §7.2].

Several questions merit further investigation.

(1) Torsion outside the ramification subgroup. The quotient J[d]/RdJ[d]/R_d records the part of the geometric dd-torsion not accounted for by the ramification classes. Understanding its Galois action, together with the cohomological obstruction associated with

0⟶Rd⟶J[d]⟶J[d]/Rd⟶0,0\longrightarrow R_d\longrightarrow J[d]\longrightarrow J[d]/R_d\longrightarrow0,

may yield additional criteria for determining J(Q)[d]J(\mathbb{Q})[d]. In particular, one should identify hypotheses that genuinely imply the desired description of rational torsion, rather than infer it from the vanishing of quotient invariants alone.

(2) Non-coprime degrees. It would be useful to extend the structural analysis to gcd⁡(d,deg⁡f)≠1\gcd(d,\deg f)\ne1. In this setting, the points above infinity and their divisor classes require separate treatment, and the ramification relations may differ from those considered here.

(3) Curves over general number fields. The structural and reduction-theoretic arguments can be investigated over a general number field KK. Such an extension would require accounting for the GKG_K-action, the residue fields at primes of good reduction, and the corresponding reduction maps.

(4) Finite-field computations. Systematic computation of Frobenius polynomials and Jacobian orders for explicit superelliptic families may provide effective ways to select good primes that eliminate rational torsion. The interaction between these computations and the automorphisms of the covering may also be worth exploring.

(5) Low-degree divisor geometry. The results of § 8 give necessary geometric conditions for distinct effective divisors of degree less than the genus to represent the same divisor class. A more detailed study of the Abel–Jacobi images of Sym⁡r(C)\operatorname{Sym}^r(C), for r<gr < g, may provide further restrictions on rational torsion classes admitting effective representatives of small degree.

Overall, the results show that ramification geometry, Galois symmetry, and arithmetic reduction provide complementary tools for studying rational torsion on odd superelliptic Jacobians. Determining when these tools can be combined into general classification theorems remains an interesting problem.

Finite-field verification for Example 7.1

We verify the finite-field point counts for the curve

C:y3=x5−2C :\quad y^3 = x^5 - 2

over Fpk\mathbb{F}_{p^k}, where p∈{7,11}p \in\{7,11\} and 1≤k≤41 \le k \le4. Since gcd⁡(3,5)=1\gcd(3,5) = 1, the smooth projective model has a unique point at infinity. Thus, if NkN_k denotes the number of Fpk\mathbb{F}_{p^k}-rational points on the projective curve, then

Nk=1+#{(x,y)∈Fpk2:y3=x5−2}.N_k = 1 + \#\{(x,y) \in\mathbb{F}_{p^k}^2 : y^3 = x^5 - 2\}.

The following SageMath code counts the cube roots of each field element once and then uses these counts to enumerate the affine points. The assertions check the values used in Subsection 7.1.

def point_count(p, k):
    F = GF(p^k, name=’a’)

    # Count the cube roots of each element of F.
    cube_counts = {}
    for y in F:
        z = y^3
        cube_counts[z] = cube_counts.get(z, 0) + 1

    # Count affine points satisfying y^3 = x^5 - 2.
    count_affine = 0
    for x in F:
        rhs = x^5 - F(2)
        count_affine += cube_counts.get(rhs, 0)

    # There is exactly one point at infinity.
    return count_affine + 1
expected = {
    7: [8, 50, 344, 2766],
    11: [12, 154, 1332, 15274],
}

for p in [7, 11]:
    computed = [point_count(p, k) for k in [1, 2, 3, 4]]
    print("p =", p, "point counts =", computed)
    assert computed == expected[p]

The expected output is

p = 7 point counts = [8, 50, 344, 2766]
p = 11 point counts = [12, 154, 1332, 15274]

References

  1. [1]Enrico Arbarello, Maurizio Cornalba, Phillip A. Griffiths, and Joe Harris. Geometry of Algebraic Curves. Volume I. Vol. 267. Grundlehren der mathematischen Wissenschaften. New York: Springer, 1985.
  2. [2]Kalyan Banerjee, Kalyan Chakraborty, and Ankita Das. “Generalized fruit Diophantine equation and super elliptic curves”. In: Monatshefte für Mathematik (2026). DOI: 10.1007/s00605-026-02201-6.DOI
  3. [3]Siegfried Bosch, Werner Lütkebohmert, and Michel Raynaud. Néron Models. Vol. 21. Ergebnisse der Mathematik und ihrer Grenzgebiete (3). Berlin: Springer, 1990.
  4. [4]Irene I. Bouw and Stefan Wewers. “Computing L-functions and semistable reduction of superelliptic curves”. In: Glasgow Mathematical Journal 59.1 (2017), pp. 77–108. DOI: 10.1017/S0017089516000057.DOI
  5. [5]Marc Coppens. “Weierstrass points of curves of genus 3”. In: Manuscripta Mathematica 67 (1990), pp. 193–213.
  6. [6]Marc Hindry and Joseph H. Silverman. Diophantine Geometry: An Introduction. Vol. 201. Graduate Texts in Mathematics. New York: Springer-Verlag, 2000. DOI: 10.1007/978-1-4612-1210-2.DOI
  7. [7]Tomasz Jȩdrzejak. “On the Torsion of the Jacobians of Superelliptic Curves”. In: Journal of Number Theory 145 (2014), pp. 402–425. DOI: 10.1016/j.jnt.2014.06.013.DOI
  8. [8]Ernst Kani and Michael Rosen. “Idempotent relations and factors of Jacobians”. In: Mathematische Annalen 284 (1989), pp. 307–327.DOI
  9. [9]Jef Laga and Ari Shnidman. “The Geometry and Arithmetic of Bielliptic Picard Curves”. In: Journal of the London Mathematical Society 112.5 (2025), e70347. DOI: 10.1112/jlms.70347.DOI
  10. [10]Barry Mazur. “Modular curves and the Eisenstein ideal”. In: Publications Mathématiques de l’IHÉS 47 (1977), pp. 33–186.DOI
  11. [11]David Mumford. Tata Lectures on Theta II: Jacobian Theta Functions and Differential Equations. Boston: Birkhäuser, 1984.
  12. [12]Joseph H. Silverman. Advanced Topics in the Arithmetic of Elliptic Curves. Vol. 151. Graduate Texts in Mathematics. New York: Springer, 1994.DOI
  13. [13]Henning Stichtenoth. Algebraic Function Fields and Codes. 2nd ed. Vol. 254. Graduate Texts in Mathematics. Berlin: Springer, 2009.DOI
  14. [14]Wojciech Wawró​w. “On torsion of superelliptic Jacobians”. In: Journal de Théorie des Nombres de Bordeaux 33.1 (2021), pp. 223–235. DOI: 10.5802/jtnb.1158.DOI

Paper details

vlnc:W_3z8kqf8m
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents