Abelian Varieties and the Jacobian

⚡ Key Takeaways

One-line statement: An abelian variety is a projective algebraic group; the Jacobian of a smooth projective curve is the canonical abelian variety attached to it, parametrizing degree-zero divisor classes.

Key result in words: Every smooth projective curve $C$ of genus $g$ has a $g$ -dimensional abelian variety $\mathrm{Jac}(C) = \mathrm{Pic}^0(C)$ whose group law geometrizes linear equivalence of divisors, and over $\mathbb{C}$ this is the complex torus $H^0(C,\Omega^1)^\ast / H_1(C,\mathbb{Z})$ .

Where it applies: Number theory (rational points, Mordell–Weil), algebraic geometry (moduli, Torelli), complex geometry, cryptography (elliptic/hyperelliptic curve groups), and the theory of theta functions.

Who/when: Abel, Jacobi, Riemann (19th c.); Lefschetz, Weil, Chow, Mumford, Tate (20th c.); Faltings (1983).

💡 The Core Insight

A smooth projective curve is a one-dimensional object, but its linear equivalence classes of divisors assemble into a higher-dimensional group that is itself a projective variety. Linearizing the nonlinear data of a curve into a torus-like group is what makes integration, the Abel–Jacobi map, and the geometry of theta divisors possible. The Jacobian converts geometric statements about $C$ into group-theoretic and lattice-theoretic statements.

DefinitionAbelian Variety

An abelian variety over a field $k$ is a group object in the category of varieties over $k$ that is complete (proper) and connected. Concretely it is a smooth projective variety $A/k$ equipped with morphisms

1

\quad m: A\times A \to A,\qquad i: A\to A,\qquad e:\operatorname{Spec}k \to A

Derivation hint: $m$ is multiplication, $i$ is inversion, $e$ picks the identity; the group axioms are expressed as commuting diagrams of these morphisms. satisfying associativity, identity, and inverse axioms diagrammatically. The dimension $g=\dim A$ is the dimension of the abelian variety. Over $\mathbb{C}$ every abelian variety is a complex torus $\mathbb{C}^g/\Lambda$ admitting a polarization (see below); not every complex torus is algebraic.

TheoremRigidity and Commutativity

Any morphism of abelian varieties $f:A\to B$ that sends $e_A\mapsto e_B$ is automatically a group homomorphism. Consequently the group law on an abelian variety is commutative, and is uniquely determined by the choice of identity element.

Proof

Rigidity Lemma. Let $X$ be complete, $Y,Z$ varieties, and $f:X\times Y\to Z$ a morphism such that $f(X\times\{y_0\})$ is a single point $z_0$ . Then $f$ factors through the projection $\mathrm{pr}_Y: X\times Y\to Y$ .

Step 1. Pick an open affine $U\ni z_0$ in $Z$ . Let $W=f^{-1}(Z\setminus U)$ , a closed set. Its image under $\mathrm{pr}_Y$ is closed (since $X$ is complete, $\mathrm{pr}_Y$ is a closed map) and avoids $y_0$ , giving an open $V\ni y_0$ with $X\times V$ mapping into $U$ .

Step 2. For each $y\in V$ , $f(X\times\{y\})$ is a complete connected subvariety of the affine $U$ , hence a point. Define $g(y)=f(x_0,y)$ ; then $f=g\circ\mathrm{pr}_Y$ on $X\times V$ , and by density of $V$ on all of $X\times Y$ .

Step 3 (commutativity). Apply the lemma to the commutator map

2

\quad \phi: A\times A \to A,\qquad \phi(x,y)=m\big(m(x,y),\, i(m(y,x))\big)

Derivation hint: $\phi(x,y)=xyx^{-1}y^{-1}$ in group notation; show it is constant. Since $\phi(x,e)=e$ for all $x$ , rigidity gives $\phi(x,y)=\phi(e,y)=e$ , so $xy=yx$ .

Step 4 (homomorphism property). Given $f:A\to B$ with $f(e_A)=e_B$ , the map $(x,y)\mapsto f(xy)f(y)^{-1}f(x)^{-1}$ is constant in $x$ for fixed setups by the same rigidity argument, forcing $f(xy)=f(x)f(y)$ . $\blacksquare$

∎

DefinitionJacobian Variety

Let $C$ be a smooth projective geometrically connected curve of genus $g$ over $k$ with a rational point. The Jacobian $J=\mathrm{Jac}(C)$ is the connected component of the identity of the Picard scheme:

3

\quad \mathrm{Jac}(C) = \mathrm{Pic}^0(C) = \big\{ \text{degree-0 line bundles on } C \big\}/\cong

Derivation hint: $\mathrm{Pic}(C)=\bigoplus_d \mathrm{Pic}^d(C)$ graded by degree; the degree-0 part is the connected identity component, an abelian variety of dimension $g$ . It is an abelian variety of dimension $g$ , representing the functor

4

\quad \mathrm{Pic}^0_{C/k}(T) = \big\{ \mathcal{L}\in\mathrm{Pic}(C\times T) : \deg(\mathcal{L}|_{C\times\{t\}})=0 \big\}/\,\mathrm{pr}_T^\ast\mathrm{Pic}(T)

Derivation hint: rigidify by quotienting out line bundles pulled back from the base $T$ ; representability is Grothendieck's theorem on Picard functors.

DefinitionAbel–Jacobi Map

Fixing a base point $p_0\in C(k)$ , the Abel–Jacobi map embeds the curve into its Jacobian:

5

\quad \alpha: C \longrightarrow \mathrm{Jac}(C),\qquad p \longmapsto \mathcal{O}_C(p - p_0)

Derivation hint: a point $p$ gives the degree-0 divisor $p-p_0$ ; its class lives in $\mathrm{Pic}^0$ . For $g\geq 1$ this is a closed immersion. Over $\mathbb{C}$ it has the analytic description, with $\{\omega_1,\dots,\omega_g\}$ a basis of holomorphic 1-forms,

6

\quad \alpha(p) = \left(\int_{p_0}^p \omega_1,\ \dots,\ \int_{p_0}^p \omega_g\right) \bmod \Lambda

Derivation hint: integrals are path-dependent only up to periods $\int_\gamma\omega_j$ for $\gamma\in H_1(C,\mathbb{Z})$ , which form the lattice $\Lambda$ ; reducing mod $\Lambda$ makes $\alpha$ well-defined.

Proof

The complex-analytic construction of $\mathrm{Jac}(C)$ . Let $C$ be a compact Riemann surface of genus $g$ .

Step 1 — The period lattice. The space $V=H^0(C,\Omega^1)^\ast \cong \mathbb{C}^g$ is dual to holomorphic 1-forms. Integration of forms over 1-cycles defines a map

7

\quad H_1(C,\mathbb{Z}) \longrightarrow V,\qquad \gamma \longmapsto \left(\omega\mapsto \int_\gamma \omega\right)

Derivation hint: $H_1(C,\mathbb{Z})\cong\mathbb{Z}^{2g}$ ; the image is a full lattice $\Lambda$ of rank $2g$ in $\mathbb{C}^g$ by the nondegeneracy of the period matrix.

Step 2 — The torus. Define the complex torus

8

\quad \mathrm{Jac}(C) = V/\Lambda = H^0(C,\Omega^1)^\ast / H_1(C,\mathbb{Z})

Derivation hint: $\dim_\mathbb{R} V = 2g = \operatorname{rank}\Lambda$ , so the quotient is a compact $2g$ -real-dimensional torus with a complex structure inherited from $V$ .

Step 3 — Choosing a basis and the period matrix. Choose a symplectic basis $\{a_1,\dots,a_g,b_1,\dots,b_g\}$ of $H_1(C,\mathbb{Z})$ with intersection pairing $a_i\cdot b_j=\delta_{ij}$ , and a normalized basis $\{\omega_j\}$ of 1-forms with $\int_{a_i}\omega_j=\delta_{ij}$ . Then $\tau_{ij}=\int_{b_i}\omega_j$ defines the period matrix $\tau$ , and $\Lambda = \mathbb{Z}^g + \tau\mathbb{Z}^g$ .

Step 4 — Riemann's bilinear relations. The period matrix satisfies

9

\quad \tau = \tau^{T}\quad(\text{symmetric}),\qquad \operatorname{Im}\tau > 0\quad(\text{positive definite})

Derivation hint: apply Stokes to $\int_C \omega_i\wedge\omega_j=0$ and $\int_C \omega_i\wedge\overline{\omega_j}>0$ , expanding over the polygon presentation of $C$ ; the first gives symmetry, the second positivity.

Step 5 — Polarization and algebraicity. The matrix $\tau\in\mathfrak{H}_g$ (Siegel upper half-space) defines a positive-definite Hermitian form whose imaginary part is the integral symplectic intersection form. This Riemann form is a principal polarization, and by the Kodaira embedding / Lefschetz theorem it makes $V/\Lambda$ projective — an abelian variety. $\blacksquare$

∎

Figure

The Abel–Jacobi map: a genus-2 curve maps into its 2-dimensional Jacobian torus

mathematician · Carl Gustav Jacob Jacobi | 1804–1851

Jacobi pioneered the inversion of elliptic and hyperelliptic integrals. Where Abel established the addition theorem for sums of integrals, Jacobi posed and solved the inversion problem: given the multivalued integral functions, invert them via theta functions on a $g$ -dimensional torus. His Fundamenta Nova Theoriae Functionum Ellipticarum (1829) and later work on Abelian functions laid the analytic foundation for what Riemann would geometrize and Weil would algebraicize. The variety now bears his name precisely because the Jacobi inversion problem is solved on it.

🧠 Intuition

Think of the Jacobian as the abelianization of the curve's geometry. A single point of $C$ is rigid, but formal sums of points modulo linear equivalence form a group: you can add divisors, and two divisors are "the same" if their difference is the divisor of a meromorphic function. The genus $g$ counts the independent holomorphic 1-forms; each gives an independent "coordinate" obtained by integration. The path-dependence of integration produces the period lattice — the torus arises because going around a loop in $C$ shifts the integral by a fixed period vector. So $\mathrm{Jac}(C)$ is literally the space of values of integrals modulo their ambiguities. The principal polarization (theta divisor) is the geometric shadow of the symplectic intersection form on cycles — it remembers which loops link.

WA

Why a physicist cares

Jacobians and theta functions encode the integrable structure of classical mechanical systems: the spectral curve of a Lax pair has a Jacobian, and the linearized flow of the integrable system is straight-line motion on this torus (Liouville–Arnold meets algebraic geometry). Solutions to KdV, KP, and Toda hierarchies are written with Riemann theta functions on Jacobians, and the Krichever construction reverses the Abel–Jacobi map to build solutions. In string theory, period matrices of Riemann surfaces are the moduli integrated in genus- $g$ amplitudes.

ExampleElliptic curve: genus 1,

J\cong C

Given: The elliptic curve $E: y^2 = x^3 - x$ over $\mathbb{C}$ , with the holomorphic 1-form $\omega = dx/y$ .

Steps. 1. Genus $g=1$ , so $\mathrm{Jac}(E)$ is a 1-dimensional torus $\mathbb{C}/\Lambda$ . 2. The period lattice is $\Lambda = \mathbb{Z}\,\omega_1 + \mathbb{Z}\,\omega_2$ where the periods are real integrals. For $y^2=x^3-x$ the real period is

10

\quad \omega_1 = 2\int_1^\infty \frac{dx}{\sqrt{x^3-x}} = \frac{\Gamma(1/4)^2}{\sqrt{2\pi}} \approx 5.2441

Derivation hint: substitute $x=1/\sin^2\theta$ or use the lemniscatic Beta-function identity $\int_1^\infty (x^3-x)^{-1/2}dx = \tfrac12 B(\tfrac14,\tfrac12)$ . 3. By symmetry $j$ -invariant $=1728$ gives $\tau = i$ , so $\Lambda = \omega_1(\mathbb{Z}+i\mathbb{Z})$ — a square lattice (lemniscatic case). 4. The Abel–Jacobi map $\alpha(P)=\int_{O}^P \omega \bmod \Lambda$ is an isomorphism $E\xrightarrow{\sim}\mathrm{Jac}(E)$ , with the chord-tangent group law on $E$ matching addition on $\mathbb{C}/\Lambda$ .

Boxed answer:

\boxed{\ \mathrm{Jac}(E)\cong \mathbb{C}/\big(\omega_1(\mathbb{Z}+i\mathbb{Z})\big),\qquad \omega_1=\Gamma(1/4)^2/\sqrt{2\pi}\approx 5.2441\ }

ExampleGenus 2: dimension of the Jacobian and the theta divisor

Given: A genus-2 hyperelliptic curve $C: y^2 = f(x)$ , $\deg f = 6$ , with the two 1-forms $\omega_1 = dx/y$ , $\omega_2 = x\,dx/y$ .

Steps. 1. $g=2 \Rightarrow \dim\mathrm{Jac}(C)=2$ , a 2-dimensional complex torus $\mathbb{C}^2/\Lambda$ , $\Lambda\cong\mathbb{Z}^4$ . 2. The period matrix $\tau\in\mathfrak{H}_2$ is a symmetric $2\times2$ matrix with $\operatorname{Im}\tau>0$ : $\tau=\begin{pmatrix}\tau_{11}&\tau_{12}\\\tau_{12}&\tau_{22}\end{pmatrix}$ , a 3-complex-dimensional choice. 3. The theta divisor $\Theta = \{\theta(z,\tau)=0\}$ is the image of $C$ under a degree-1 Abel–Jacobi map: $\Theta\cong C$ as a subvariety. Its self-intersection numbers obey

11

\quad (\Theta^g) = g! \quad\Rightarrow\quad (\Theta^2) = 2! = 2

Derivation hint: for a principally polarized abelian variety the top self-intersection of the theta divisor equals $g!$ , computed from the Riemann–Roch / Pfaffian of the polarization form, here $g=2$ . 4. Every genus-2 curve is a Jacobian (the Schottky problem is vacuous for $g\le 3$ ): $\mathcal{M}_2$ has dimension $3g-3=3$ , matching $\dim\mathcal{A}_2 = g(g+1)/2 = 3$ .

Boxed answer:

\boxed{\ \dim\mathrm{Jac}(C)=2,\quad \tau\in\mathfrak{H}_2\ (3\text{ params}),\quad (\Theta^2)=2\ }

⚠️ Common mistake

"Every complex torus is an abelian variety." False. A generic complex torus $\mathbb{C}^g/\Lambda$ for $g\ge2$ admits no positive line bundle and is not projective — it has no nonconstant meromorphic functions and cannot be embedded in $\mathbb{P}^N$ . Algebraicity requires the existence of a Riemann form (polarization): a positive-definite Hermitian form $H$ with $\operatorname{Im}H(\Lambda,\Lambda)\subseteq\mathbb{Z}$ . For $g=1$ every torus admits one (the Riemann relations are automatically satisfiable), so all 1-tori are elliptic curves — but this is special to dimension 1.

⚠️ Common mistake

"Every abelian variety is a Jacobian." False for $g\ge 4$ . The Jacobian locus $\mathcal{J}_g$ has dimension $3g-3$ (= $\dim\mathcal{M}_g$ ), while $\mathcal{A}_g$ has dimension $g(g+1)/2$ . These agree for $g\le 3$ but $3g-3 < g(g+1)/2$ once $g\ge4$ . The problem of characterizing which ppavs are Jacobians is the Schottky problem.

Counterexample

A ppav that is not a Jacobian. Take a generic principally polarized abelian variety $A$ of dimension $4$ . Since $\dim\mathcal{A}_4 = 10$ but $\dim\mathcal{J}_4 = 9$ , a generic point of $\mathcal{A}_4$ lies off the Jacobian locus. Concretely, a generic abelian fourfold has a smooth, irreducible theta divisor $\Theta$ , whereas the theta divisor of a Jacobian is singular in codimension $\ge1$ when $g\ge4$ (Andreotti–Mayer: $\dim\operatorname{Sing}\Theta \ge g-4$ , with equality characterizing Jacobians generically). The singular structure of $\Theta$ is the diagnostic distinguishing Jacobians.

connection

The Jacobian unifies several neighbouring theories. For $g=1$ it reduces to Elliptic Curves, where curve and Jacobian coincide. The construction via $\mathrm{Pic}^0$ is a special case of the Picard Scheme and generalizes to the Albanese and Picard varieties of higher-dimensional varieties (which are dual). The duality $\widehat{\mathrm{Jac}(C)}\cong\mathrm{Jac}(C)$ is an instance of Dual Abelian Variety theory, made canonical by the principal polarization. The lattice description connects to Hodge Theory (the Jacobian is the intermediate Jacobian $H^{0,1}/H^1(\mathbb{Z})$ ) and to Theta Functions. Arithmetically, the rational points $\mathrm{Jac}(C)(k)$ are governed by the Mordell–Weil Theorem, and the $\ell$ -adic Tate module links to Galois representations and the Tate Conjecture.

🔬 Where this appears

⏳ History

- 1826–1829 — Abel & Jacobi: Abel's addition theorem and Jacobi's inversion problem for elliptic and hyperelliptic integrals set the analytic stage. Jacobi proposes inverting

g

integrals on a

g

-torus.

- 1857 — Riemann: Theorie der Abel'schen Functionen introduces Riemann surfaces, the period matrix, the Riemann bilinear relations, and theta functions; the Jacobian appears as the torus solving Jacobi inversion.

- 1860s — Weierstrass, Clebsch–Gordan: Systematic theory of abelian functions and theta divisors; the addition theorem given geometric form.

- 1880s–1910s — Picard, Poincaré, Lefschetz: Picard and Albanese varieties; Lefschetz's theorem (1921) characterizing which complex tori are projective (existence of a polarization), proving algebraicity via theta line bundles.

- 1948–1954 — Weil: Variétés abéliennes et courbes algébriques gives the first purely algebraic construction of the Jacobian over arbitrary fields, foundational for abelian varieties as algebraic groups; proves the Riemann hypothesis for curves over finite fields.

- 1950s — Chow, Matsusaka, Igusa: Chow's construction of Jacobians; Matsusaka–Hoyt and the algebraic Torelli theorem.

- 1960s — Grothendieck, Mumford: Representability of the Picard functor (Grothendieck); Mumford's Abelian Varieties (1970) recasts the theory scheme-theoretically, introduces GIT and theta-group methods, the Mumford bundle, and the structure of

\mathcal{A}_g

.

- 1968 — Andreotti–Mayer: Singularities of the theta divisor and approach to the Schottky problem.

- 1983 — Faltings: Proof of the Mordell conjecture using heights on Jacobians and the Tate/Shafarevich/Mordell–Weil machinery; finiteness of rational points on higher-genus curves.

- 1986 — Shiota: Solution of the Schottky problem via the KP equation (Novikov's conjecture), characterizing Jacobians among ppavs.

open-problem

The Schottky problem asks for explicit equations cutting out the Jacobian locus $\mathcal{J}_g$ inside the Siegel modular variety $\mathcal{A}_g$ (equivalently inside Siegel upper half-space modulo $\mathrm{Sp}(2g,\mathbb{Z})$ ). Shiota (1986) proved Novikov's conjecture: a ppav is a Jacobian iff its theta function gives a solution of the KP hierarchy (the Krichever–Novikov characterization), but a clean, finite, geometric set of defining equations valid for all $g$ remains elusive. Related open directions: the Andreotti–Mayer characterization via $\dim\operatorname{Sing}\Theta$ has known excess components; precise stratification is unresolved. Over number fields, the Birch–Swinnerton-Dyer conjecture for Jacobians (relating $L$ -functions to Mordell–Weil rank and Tate–Shafarevich groups) is open in general, as is effective Mordell (effective bounds on the height of rational points, not just finiteness). The Coleman–Oort conjecture predicts only finitely many CM Jacobians exist for large genus — still open.

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also