Skip to content

Jacobian variety

A curve's Jacobian is the abelian variety of its degree-zero line-bundle classes, turning divisor equivalence into algebraic group geometry.

Version
v1 · 2026-10-03 · History
Domain-specific #
13352
Domain group
Formal Sciences
Origin domain
Mathematics
Subdomains
Algebraic Geometry, Curves → Mathematics
Aliases
Curve Jacobian

Core Idea

For a smooth complete geometrically connected algebraic curve C of genus g, its Jacobian variety J(C) is a g-dimensional abelian variety associated with degree-zero line bundles or divisor classes on the curve. Instead of treating each formal sum of points separately, it identifies divisors differing by a principal divisor and gives their classes algebraic group structure. This turns questions about divisors, linear equivalence and maps from the curve into questions about points and subvarieties of a group variety.[1]

If a rational base point P₀ exists, the Abel map sends P to the class [P−P₀]. For positive genus it embeds the curve in its Jacobian; for genus one with a chosen origin, curve and Jacobian become isomorphic as pointed group varieties. For genus zero, the Jacobian is trivial. Over an arbitrary field, however, it is too glib to assert J(k)=Pic⁰(C) for every curve: Milne's construction explains the descent issue when C(k) is empty and represents the appropriate relative Picard functor. The basepoint formula and identification need their hypotheses.[1]

Structural Signature

Sig role-phrases:

  • Smooth complete curve: the carrier whose genus and divisors determine the construction.
  • Degree-zero classes: divisors modulo principal divisors, equivalently suitable degree-zero line bundles.
  • Abelian variety: a complete algebraic group space of dimension g representing those classes in the correct relative sense.
  • Base-point Abel map: P ↦ [P−P₀] where a rational P₀ is available.
  • Canonical principal polarization: the theta-divisor structure needed to state Torelli recovery of the curve.[1]

The last role is not merely ornamental. Torelli's theorem says the canonically polarized Jacobian determines the curve. An isomorphism of unpolarized abelian varieties is not the theorem's hypothesis. On the other hand, elementary computations of a point in J(k) need not explicitly compute a theta divisor. Keeping the geometric invariant and the computational representation separate prevents either from being overclaimed.[1][2]

What It Is Not

The Jacobian is not the Jacobian Matrix of derivatives. It is also not just the abstract Picard group written with new notation: its abelian-variety structure permits geometry of dimension g, algebraic maps, polarization and arithmetic over fields. The live catalog's Picard Group identity is related because degree-zero Picard classes are the underlying moduli problem, but the named Jacobian is attached specifically to a curve and realized as an abelian variety. No strict parent relation is inferred from the related vocabulary alone.[1]

A genus-one curve with no k-rational point is another boundary. Its Jacobian exists over k, but identifying the curve itself with that Jacobian as a pointed group requires a rational origin. The convenient elliptic-curve slogan “the curve is its Jacobian” is therefore conditional, not a way to make all genus-one torsors disappear.[1]

Scope of Application

Milne's author-hosted treatment defines the relative degree-zero Picard problem, constructs the abelian variety and studies maps from the curve. Over the complex numbers the classical description uses holomorphic differentials and their periods to form a complex torus; the algebraic construction remains meaningful over other fields, including finite fields where divisor classes can be computed. The complex quotient is not a replacement for checking field-of-definition and descent conditions in arithmetic examples.[1]

Two first-party SageMath references expose unlike computational roles. A smooth projective cubic over F₂₉ has genus one; Sage forms its Hess-model Jacobian using the point at infinity as base divisor and adds two classes represented by rational points. A degree-five hyperelliptic curve over F₂₀₀₃ has genus two; Sage creates a Jacobian point from a pair of polynomials in the standard hyperelliptic representation and confirms that adding the identity leaves it unchanged. The second case is not “just elliptic addition with a larger group”—a genus-two point is generally a divisor class, not a single point of the curve.[3][2]

Clarity

Choose P₀ ∈ C(k). The expression P−P₀ has degree zero, so its linear-equivalence class lies in the Jacobian. Adding two classes means adding their representing divisors and reducing modulo principal divisors; it does not normally produce the ordinary sum of two points on the curve. In genus one, the chosen point P₀ makes the group law on the curve itself possible. In higher genus, the image of C in J(C) is a proper subvariety, while sums of its point classes range through the larger abelian variety.[1]

The Sage F₂₉ example starts with projective cubic x³+5z³−y²z=0, base place (0:1:0), and point places (1:8:1) and (2:10:1). Its code constructs J=C.jacobian(model='hess', base_div=B), converts those places to group elements, then prints a valid sum class. The specific output is a computed group result; the core lesson is that two point-minus-base classes add in J(F₂₉), with basepoint choice doing real work.[3]

In the Sage F₂₀₀₃ example, C is y²=x⁵+1184x³+1846x²+956x+560. It constructs J=C.jacobian(), then D from a=x²+376x+245 and b=1015x+1368. Sage prints the normalized class representation and verifies D == D + J(0). The representation uses polynomial data for a class on a two-dimensional Jacobian, not a claim that the curve has dimension two. No principal polarization is calculated by this code sample.[2]

Manages Complexity

The construction packages infinitely many formal divisor expressions into equivalence classes and then a geometric group. It lets arithmetic questions be posed on J(C) while retaining the curve's geometry through its Abel image and canonical polarization. This is why one must track which layer a claim concerns: a specific divisor representative, its degree-zero class, the abelian variety, or the curve recovered from its polarized Jacobian. Confusing these layers can produce false equalities even when the notation looks familiar.[1]

For computation, a curve model and a representation of classes are indispensable. Sage's genus-one Hess model uses a base divisor; its genus-two hyperelliptic interface uses polynomial pairs. Each is a concrete coordinate for group arithmetic, not the definition of the Jacobian as a moduli object. A different model may represent the same abstract class differently; equality is equality in the group, not literal equality of a chosen written divisor.[3][2]

Abstract Reasoning

The algebraic step is quotienting degree-zero divisors by principal divisors. Principal divisors act as the “do not count as a new class” relation because multiplication by a rational function changes a divisor representative without changing its line-bundle class. The geometric step is representing those classes by a complete group variety rather than leaving them as a bare set. Its dimension equals genus, connecting a curve invariant to the size of the classifying space.[1]

Counterfactuals isolate the concept. Remove degree zero and the Picard scheme has multiple degree components, not just this identity component. Remove the quotient by principal divisors and there is no intended class group. Remove the curve and a generic abelian variety is not its Jacobian. Remove the chosen rational basepoint and P↦[P−P₀] may not be a k-defined map of the stated form, although the Jacobian itself still exists. Forget polarization and one cannot quote Torelli as a statement about the bare abelian variety.[1]

Knowledge Transfer

The genus-one F₂₉ calculation and genus-two F₂₀₀₃ calculation share degree-zero class/group structure while requiring different representations. The transfer is the role relation—curve to divisor classes to abelian group variety—not the numerical formula for point addition. In genus one a rational origin identifies the curve with its Jacobian; in genus two a single curve point occupies only part of a two-dimensional classifying variety. This explains why a construction can recur across genera without erasing geometric differences.[1][3][2]

Picard varieties generalize the idea beyond curves, but the Jacobian's curve attachment, canonical principal polarization and Abel map remain decisive. The broader Picard Group entry is a useful comparison, not a verified strict prime parent.

Examples

  1. Sage's genus-one cubic over F₂₉. The first-party Hess-model example takes C: x³+5z³−y²z=0 with base place B=(0:1:0). Sage checks genus one, forms J=C.jacobian(model='hess', base_div=B), converts P₁=(1:8:1) and P₂=(2:10:1) into classes relative to B, then computes p₁+p₂ as another class. Mapped back: curve = that smooth projective cubic; degree-zero classes = P₁−B and P₂−B; abelian variety = J; Abel map = the chosen-baseplace conversion; canonical polarization = part of the general Jacobian theory, not calculated in this example. The computation exhibits a group operation, not an independent proof of Torelli.[3][1]

  2. Sage's genus-two hyperelliptic curve over F₂₀₀₃. The code sets f=x⁵+1184x³+1846x²+956x+560, defines C:y²=f(x) and J=C.jacobian(), then creates D=J(F₂₀₀₃)([a,b]) with a=x²+376x+245 and b=1015x+1368. It prints the normalized class and checks D+J(0)=D. Mapped back: curve = degree-five genus-two hyperelliptic C; degree-zero class = polynomial-represented D; abelian variety = two-dimensional J; Abel map = not used to construct this sample class; canonical polarization = theoretical invariant, not this calculation's output. The identity-law check makes the class-group role explicit without pretending to derive an encryption scheme or security property.[2][1]

Structural Tensions

Explicit class arithmetic versus invariant geometry. A polynomial or divisor model makes points of a Jacobian computable over a chosen field, but its formulas and reductions depend on model, base choices and representation conventions. Describing J(C) as a canonically polarized abelian variety expresses field-independent structure and supports Torelli reasoning, but does not by itself hand an implementer a cheap addition routine. Diagnostic: is the task to calculate a particular class sum, or to prove a property of the curve from its polarized Jacobian? Sage's worked examples address the former; Milne's construction and Torelli theorem address the latter.[3][2][1]

Structural–Framed Character

The class quotient and group-variety laws give the entry a strongly structural mathematical core; their evaluative weight arises from mathematicians' choice of questions about divisors, rational points or curves, not from social preference. The concept depends on human proof conventions and model choices for use, yet its identity is not institutionally created by them. The vocabulary originates in algebraic/complex geometry and travels to arithmetic over finite fields through the algebraic construction, but one cannot import a complex period formula or rational-basepoint identification without checking hypotheses. Recognition requires the curve-attached degree-zero moduli relation; calling any object with a group law a Jacobian is an import error. Its character: an abstract mathematical structure with conditional representations and field-dependent access to points.[1][2]

Structural Core vs. Domain Accent

The skeletal relation is passage from objects on a carrier to equivalence classes with a group law. The domain-bound mechanism is divisor linear equivalence on smooth curves, representability by an abelian variety of genus dimension, and a canonical polarization tied to the curve. This named entry fails the prime bar because stripping those algebraic-geometric conditions yields a generic quotient-group story that cannot recover the Abel map, basepoint caveat or Torelli claim. Algebraic Variety is the live strict genus; a different portable skeleton would need unlike-domain evidence and could not simply reuse J(C) vocabulary.[1]

This entry is a kind of Algebraic Variety.

Strict parent: Algebraic Variety. The Jacobian is a curve-attached abelian algebraic variety representing degree-zero divisor classes; most algebraic varieties are not Jacobians. Picard Group remains a related comparison, not an additional parent. Rational-basepoint and field/descent caveats remain in force.

Relationships to Other Abstractions

Local relationship map for Jacobian varietyParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.Jacobian varietyDOMAINDomain-specific abstraction: Algebraic Variety — is a kind ofAlgebraicVarietyDOMAIN

Current abstraction Jacobian variety Domain-specific

Parents (1) — more general patterns this builds on

  • Jacobian variety is a kind of Algebraic Variety Domain-specific

    Jacobian varieties are specialized algebraic varieties.

Hierarchy path (1) — routes to 1 parentless root

Neighborhood in Abstraction Space

Jacobian variety sits in a sparse region of the domain-specific corpus (68th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Sheaves & Birational Geometry (22 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With

  • The Jacobian matrix of partial derivatives.
  • A raw divisor rather than its degree-zero linear-equivalence class.
  • A genus-one curve with no rational origin automatically equal to its Jacobian over the ground field.[1]
  • An unpolarized abelian variety claimed to recover the curve by Torelli.[1]

References

[1] J. S. Milne, Jacobian Varieties, author's revised original 1986 treatment, §§1–2 and 12; especially §1's arbitrary-field caveat and §12's canonically polarized Torelli statement. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s

[2] SageMath, “Jacobian of a general hyperelliptic curve”, first-party F₂₀₀₃ polynomial-pair and identity-law example. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h

[3] SageMath, “Jacobians in Hess model”, first-party genus-one F₂₉ group example. registry ↩a ↩b ↩c ↩d ↩e ↩f