Abelian varieties — group objects in algebraic geometry

Intuition Beginner

An abelian variety is two things at once: a smooth complete shape (a projective variety) and a commutative group. You can add any two points and get a third, and this addition is given by the same kind of regular algebraic maps that define the shape itself. There is an identity point, every point has an inverse, and the group law is commutative. The defining demand is that the group operations are morphisms of varieties — addition is algebraic in the coordinates.

The one-dimensional case is an elliptic curve 04.04.03, the smooth cubic $y^2=x^3+ax+b$ whose points you add by the chord-and-tangent rule. Abelian varieties are the higher-dimensional generalisation: smooth projective shapes of any dimension $g\ge 1$ that carry a compatible group structure. Over the complex numbers an elliptic curve is a torus, and a $g$-dimensional abelian variety is a $g$-fold torus ${\mathbb{C}}^g/\Lambda$ cut out by a lattice and forced to be projective.

Two features make these objects central. First, the group law is rigid: a smooth complete variety admits at most one group structure, so the geometry fixes the arithmetic. Second, many varieties you build from other varieties — the Jacobian of a curve, the Picard variety of a surface, the Albanese — turn out to be abelian varieties, so they recur across geometry, number theory, and dynamics.

Visual Beginner

Over $\mathbb{C}$ a $g$-dimensional abelian variety is a quotient $A={\mathbb{C}}^g/\Lambda$ of complex $g$-space by a lattice $\Lambda$ of rank $2g$, subject to Riemann's positivity relations that make the quotient projective. The group law is just addition of complex vectors modulo the lattice. For $g=1$ this is an elliptic curve; for $g=2$ it is the Jacobian of a genus-2 curve, a four-real-dimensional torus folded into projective space.

The picture records the defining property: $A$ is a group object in the category of projective varieties. Addition $+$, inversion $-$, and the identity $0$ are all morphisms of varieties, so the group structure and the geometry are the same data.

Worked example Beginner

Take $E:y^2=x^3+x+1$ over the field ${\mathbb{F}}_5=\{0,1,2,3,4\}$. The points $P=(0,1)$ and $Q=(2,1)$ both lie on $E$: at $x=0$ the right side is $1$, and $1=1^2$; at $x=2$ it is $8+2+1=11\equiv 1(\mathrm{mod}5)$, again $1^2$. An elliptic curve is the one-dimensional abelian variety, so adding its points already exhibits the group-object behaviour.

To add $P$ and $Q$, draw the line through them. Its slope is $\lambda =(1-1)/(2-0)=0$, so the line is $y=1$. Substituting into the curve gives $1=x^3+x+1$, hence $x^3+x=0$, i.e. $x(x^2+1)=x(x-2)(x-3)$ over ${\mathbb{F}}_5$. The third intersection is $R^{\prime }=(3,1)$. Reflecting across the $x$-axis gives $P+Q=(3,-1)=(3,4)$.

Doubling uses the tangent. At $P=(0,1)$ implicit differentiation gives slope $\lambda =1/2\equiv 3(\mathrm{mod}5)$. The tangent $y=3x+1$ meets $E$ again at $R^{\prime }=(4,3)$, so $2P=(4,-3)=(4,2)$. Every step is a regular map — addition is algebraic in the coordinates — and this is exactly the property abstracted by an abelian variety in every dimension.

Check your understanding Beginner

Formal definition Intermediate+

Let $k$ be a field. An abelian variety over $k$ is a smooth, connected, complete group scheme over $k$; equivalently, a smooth projective geometrically connected $k$-variety $A$ equipped with morphisms

$$\mu :A{\times }_{k}A\to A(\text{addition}),\iota :A\to A(\text{inverse}),e:\mathrm{Spec}k\to A(\text{identity}),$$

satisfying the group axioms as commutative diagrams. Smooth projectivity plus the group-object property forces the group to be commutative (proved below from the rigidity lemma), and forces $\mu ,\iota ,e$ to be uniquely determined by $A$ and $e$ once they exist. The dimension of $A$ is its dimension as a $k$-variety.

Elliptic curves are the case $\dim A=1$. A one-dimensional abelian variety over $k$ is the same thing as an elliptic curve $(E,O)$: a smooth projective genus-1 curve with a marked $k$-rational point, the group law being the chord-and-tangent rule identified with ${\mathrm{Pic}}^0(E)$ via $P\mapsto [P-O]$ 04.04.03. The Weierstrass model $y^2=x^3+ax+b$ (for $\mathrm{char}k\ne 2,3$) is the explicit coordinate presentation.

The complex-analytic incarnation. Over $k=\mathbb{C}$, every abelian variety of dimension $g$ is a complex torus $A={\mathbb{C}}^g/\Lambda$ for a full lattice $\Lambda \subset {\mathbb{C}}^g$ of rank $2g$, and GAGA identifies holomorphic line bundles with algebraic ones. A complex torus is projective (hence an abelian variety) precisely when $\Lambda$ admits a Riemann form — a positive-definite Hermitian form $H$ with $\mathrm{Im}H$ integer-valued on $\Lambda$; this is Riemann's condition. The dimension-$1$ case recovers lattices $\mathbb{Z}+\tau \mathbb{Z}\subset \mathbb{C}$ with $\mathrm{Im}\tau >0$, i.e. elliptic curves.

Counterexamples and boundary cases.

• Affine group schemes. The additive group ${\mathbb{G}}_a=\mathrm{Spec}k[t]$ and the multiplicative group ${\mathbb{G}}_m=\mathrm{Spec}k[t,t^{-1}]$ are smooth connected group varieties but are affine, not complete, so they are not abelian varieties.
• Non-projective complex tori. For $g\ge 2$ a generic complex torus ${\mathbb{C}}^g/\Lambda$ admits no Riemann form and so is not projective; it is a compact complex Lie group but not an abelian variety.
• Genus-1 curves without a rational point. A smooth projective genus-1 curve $C/k$ without a $k$-rational point (e.g. Selmer's curve $3X^3+4Y^3+5Z^3=0$) is a principal homogeneous space under its Jacobian, not an abelian variety over $k$ — it carries no $k$-rational identity element.
• Non-smooth group schemes. Néron models at bad reduction and the special fibres of degenerating abelian varieties are group schemes that fail smoothness or connectedness, and the completion of a degenerating abelian variety over a disc is an open algebraic group, not an abelian variety.

Key theorem with proof Intermediate+

Lemma (theorem of the cube). Let $A,B,C$ be abelian varieties with identities $0$, and let $\mathcal{M}$ be a line bundle on $A\times B\times C$. If the restrictions $\mathcal{M}{|}_{\{0\}\times B\times C}$, $\mathcal{M}{|}_{A\times \{0\}\times C}$, $\mathcal{M}{|}_{A\times B\times \{0\}}$ are each isomorphic to $\mathcal{O}$, then $\mathcal{M}\simeq {\mathcal{O}}_{A\times B\times C}$.

The cube lemma is proved from the seesaw principle (itself a consequence of completeness of $A$: a line bundle on $A\times Y$ that is $\mathcal{O}$ on every fibre $A\times \{y\}$ descends to a pullback from $Y$) ^{[Mumford 1970]}; see the Full proof set for a sketch.

Theorem (theorem of the square). Let $A$ be an abelian variety over an algebraically closed field and $\mathcal{L}$ a line bundle on $A$. For all points $a,b\in A$,

$$t_{a+b}^{\ast }\mathcal{L}\otimes \mathcal{L}\simeq t_a^{\ast }\mathcal{L}\otimes t_b^{\ast }\mathcal{L},$$

where $t_x:A\to A$ denotes translation by $x$.

Proof. Write ${\mu }_{ij}$ and ${\mu }_{123}$ for the maps $A^3\to A$ adding the indicated coordinates, and $p_i$ for the projections. On $A\times A\times A$ form the line bundle

$$\mathcal{M}={\mu }_{123}^{\ast }\mathcal{L}\otimes {\mu }_{12}^{\ast }{\mathcal{L}}^{-1}\otimes {\mu }_{13}^{\ast }{\mathcal{L}}^{-1}\otimes {\mu }_{23}^{\ast }{\mathcal{L}}^{-1}\otimes p_1^{\ast }\mathcal{L}\otimes p_2^{\ast }\mathcal{L}\otimes p_3^{\ast }\mathcal{L}.$$

Restricting to the hyperplane $x_1=0$ sends ${\mu }_{123}\mapsto {\mu }_{23}$, ${\mu }_{12}\mapsto p_2$, ${\mu }_{13}\mapsto p_3$, and $p_1^{\ast }\mathcal{L}$ to the one-dimensional fibre $\mathcal{L}{|}_0\simeq \mathcal{O}$; the eight factors then cancel in pairs, giving $\mathcal{M}{|}_{x_1=0}\simeq \mathcal{O}$. The same cancellation holds on $x_2=0$ and $x_3=0$. By the cube lemma, $\mathcal{M}\simeq {\mathcal{O}}_{A^3}$.

Now pull this identity back along $j:A\hookrightarrow A^3$, $x\mapsto (a,b,x)$, with $a,b$ fixed. Along $j$ we have ${\mu }_{123}\mapsto (x\mapsto a+b+x)=t_{a+b}$, ${\mu }_{12}\mapsto$ the constant $a+b$, ${\mu }_{13}\mapsto t_a$, ${\mu }_{23}\mapsto t_b$, $p_1,p_2\mapsto$ the constants $a,b$, and $p_3\mapsto \mathrm{id}$. The constant factors $\mathcal{L}{|}_{a+b}$, $\mathcal{L}{|}_a$, $\mathcal{L}{|}_b$ are one-dimensional and hence isomorphic to $\mathcal{O}$, so they drop out, leaving

$${\mathcal{O}}_A\simeq j^{\ast }\mathcal{M}=t_{a+b}^{\ast }\mathcal{L}\otimes t_a^{\ast }{\mathcal{L}}^{-1}\otimes t_b^{\ast }{\mathcal{L}}^{-1}\otimes \mathcal{L}.$$

Tensoring both sides with $t_a^{\ast }\mathcal{L}\otimes t_b^{\ast }\mathcal{L}$ gives $t_{a+b}^{\ast }\mathcal{L}\otimes \mathcal{L}\simeq t_a^{\ast }\mathcal{L}\otimes t_b^{\ast }\mathcal{L}$, the theorem of the square. $\square$

Equivalently, ${\phi }_{\mathcal{L}}:A\to \mathrm{Pic}(A)$, $x\mapsto [t_x^{\ast }\mathcal{L}\otimes {\mathcal{L}}^{-1}]$, is a group homomorphism — the version in which the square is most often quoted.

Bridge. The theorem of the square builds toward the dual abelian variety $A^{\vee }={\mathrm{Pic}}^0(A)$: the homomorphism ${\phi }_{\mathcal{L}}:A\to A^{\vee }$ it constructs is exactly the map that converts a line bundle into a point of the dual, and the kernel $\ker ({\phi }_{\mathcal{L}})$ measures how far $\mathcal{L}$ sits from being ample. The same square identity appears again in the proof of Poincaré's complete reducibility theorem and in the definition of the Poincaré bundle on $A\times A^{\vee }$, where it guarantees the canonical biextension is well-defined. The central insight is that completeness forces line-bundle translation to respect the group law; putting these together, the square is the engine that makes ${\mathrm{Pic}}^0$ itself an abelian variety, and the bridge is that every structural theorem about line bundles on $A$ — ampleness, projectivity, the Riemann-Roch theorem for abelian varieties — flows from this single tensor identity.

Exercises Intermediate+

Lean formalization Intermediate+

lean_status: none. Mathlib defines elliptic curves and the chord-and-tangent group law in Mathlib.AlgebraicGeometry.EllipticCurve.Weierstrass and Mathlib.AlgebraicGeometry.EllipticCurve.Group, and the Picard-group and line-bundle infrastructure exists in the sheaf-cohomology layer. What is missing for the content of this unit is summarised in the Mathlib gap analysis: there is no Mathlib definition of an abelian variety as a smooth projective group scheme, no theorem of the square or cube, no dual abelian variety, no Poincaré bundle, no Abel-Jacobi map, and no Tate module as a Galois representation. The headline formalisation targets — the rigidity lemma, commutativity of abelian varieties, and the theorem of the square — each depend on completeness arguments for morphisms out of a product, machinery Mathlib's algebraic-geometry layer does not yet expose.

-- Target statement (not yet formalisable in Mathlib without the imports
-- listed in lean_mathlib_gap):
--
-- theorem theorem_of_the_square
--   {k : Type*} [Field k] {A : AbelianVariety k}
--   (L : LineBundle A) (a b : A) :
--   (translation (a + b))⁎L ⊗ L ≅ (translation a)⁎L ⊗ (translation b)⁎L := by
--   sorry
--
-- Missing: AbelianVariety as a smooth projective group scheme;
-- the theorem of the cube and seesaw principle; the dual A^vee = Pic^0(A);
-- the Poincaré bundle; the Tate module T_ℓ(A) as a Z_ℓ-Galois module.

Advanced results Master

Theorem of the cube. Let $\mathcal{M}$ be a line bundle on a product $A\times B\times C$ of abelian varieties. If $\mathcal{M}$ is isomorphic to $\mathcal{O}$ on each of the three coordinate slices through the identity ($\{0\}\times B\times C$, $A\times \{0\}\times C$, $A\times B\times \{0\}$), then $\mathcal{M}\simeq \mathcal{O}$. The cube is the three-variable engine behind the two-variable square; together with the seesaw principle it controls all cancellation among translates of line bundles.

The dual abelian variety. The connected component of the identity in the Picard scheme, $A^{\vee }:={\mathrm{Pic}}_{A/k}^0$, is itself an abelian variety over $k$ of the same dimension as $A$. It parametrises degree-zero line bundles on $A$, and there is a universal (Poincaré) line bundle $\mathcal{P}$ on $A\times A^{\vee }$ rigidified along $\{0\}\times A^{\vee }$ and $A\times \{0\}$. The theorem of the square makes ${\phi }_{\mathcal{L}}:A\to A^{\vee }$, $x\mapsto [t_x^{\ast }\mathcal{L}\otimes {\mathcal{L}}^{-1}]$, a homomorphism; $\mathcal{L}$ is ample iff ${\phi }_{\mathcal{L}}$ is an isogeny, and the double dual satisfies $A^{\vee \vee }\simeq A$ (reflexivity), so duality is a perfect anti-equivalence on the category of abelian varieties up to isogeny.

Jacobians of curves. For a smooth projective curve $C$ of genus $g$, the Jacobian $\mathrm{Jac}(C)={\mathrm{Pic}}^0(C)$ is an abelian variety of dimension $g$ 06.06.03. The Abel-Jacobi map $C\hookrightarrow \mathrm{Jac}(C)$, $p\mapsto [p-p_0]$, embeds $C$ into its Jacobian, and the theta divisor $\Theta =u_{g-1}({\mathrm{Sym}}^{g-1}C)$ equips $\mathrm{Jac}(C)$ with a principal polarisation ${\lambda }_{\Theta }:\mathrm{Jac}(C)\stackrel{\sim }{\to }\mathrm{Jac}(C{)}^{\vee }$. Every elliptic curve is its own Jacobian; in higher genus the Schottky problem asks which principally polarised abelian varieties arise as Jacobians of curves.

The Tate module and $\ell$-adic representations. For a prime $\ell \ne \mathrm{char}k$, the $\ell$-adic Tate module $T_{\ell }A:={\underleftarrow{\lim }}_{n}A[{\ell }^n](\bar{k})$ is a free ${\mathbb{Z}}_{\ell }$-module of rank $2g$ (the inverse limit of the ${\ell }^n$-torsion). The Galois group $\mathrm{Gal}(\bar{k}/k)$ acts continuously, producing the $\ell$-adic representation ${\rho }_{A,\ell }:\mathrm{Gal}(\bar{k}/k)\to \mathrm{GL}(T_{\ell }A)\cong {\mathrm{GL}}_{2g}({\mathbb{Z}}_{\ell })$ 21.05.01. The Weil pairing $T_{\ell }A\times T_{\ell }{A}^{\vee }\to {\mathbb{Z}}_{\ell }(1)$ is a perfect Galois-equivariant duality. For $g=1$ this recovers the $2$-dimensional representation attached to an elliptic curve; for general $g$ it is the higher-dimensional avatar at the heart of the Tate and semisimplicity conjectures.

Poincaré's complete reducibility. If $f:A\to B$ is a surjective morphism (an isogeny, or more generally a quotient), then there exists an abelian subvariety $C\subset A$ and an isogeny $C\times \ker (f{)}^0\to A$. Consequently every abelian variety is isogenous to a product of simple abelian varieties, and the decomposition is unique up to ordering and isogeny. This reduces the classification of abelian varieties up to isogeny to the simple factors together with their endomorphism rings.

Mordell-Weil and the Weil bounds. The Mordell-Weil theorem extends to arbitrary dimension: for an abelian variety $A$ over a number field $K$, the group $A(K)$ is finitely generated, $A(K)\cong {\mathbb{Z}}^r\oplus A(K{)}_{\mathrm{tors}}$, with the Néron-Tate canonical height a positive-definite quadratic form on $A(K)\otimes \mathbb{R}$. Over a finite field ${\mathbb{F}}_q$ the Frobenius endomorphism ${\pi }_A$ acts on $T_{\ell }A$ with characteristic polynomial $P_A(T)={\prod }_{i=1}^{2g}(T-{\alpha }_i)$ whose roots satisfy $|{\alpha }_i|=\sqrt{q}$ (the Riemann hypothesis for abelian varieties, proved by Weil), giving the Weil bound $|\#A({\mathbb{F}}_q)-(q^g+1)|\le (\genfrac{}{}{0ex}{}{2g}{g})q^{g-1/2}$.

Synthesis. An abelian variety is the universal recipient of integration of algebraic data: it is exactly a group object among complete varieties, and the rigidity of complete group varieties is the foundational reason its group law is commutative and unique. The theorem of the square generalises to the theorem of the cube and the dual abelian variety $A^{\vee }$, and the Poincaré bundle on $A\times A^{\vee }$ is dual to the evaluation pairing identifying $A^{\vee \vee }$ with $A$. Jacobians give the concrete realisation: $\mathrm{Jac}(C)$ is an abelian variety of dimension $g(C)$ carrying the Abel-Jacobi image of the curve, and $\mathrm{Jac}(C{)}^{\vee }\simeq \mathrm{Jac}(C)$ via the theta divisor. The central insight unifying these is that the Tate module $T_{\ell }A$ converts the geometric group law into a $2g$-dimensional $\ell$-adic Galois representation, the higher-dimensional avatar of the elliptic-curve representation. Putting these together, abelian varieties are the meeting point where complex tori, Picard varieties, Jacobians, Mordell-Weil groups, and Galois representations all turn out to be the same object viewed from different angles.

Full proof set Master

Proposition (rigidity lemma). Let $X$ be a complete connected variety over $k$, $Y,Z$ any $k$-varieties, and $f:X\times Y\to Z$ a morphism. If $f(X\times \{y_0\})=\{z_0\}$ for some $y_0\in Y$ and $z_0\in Z$, then $f$ factors through the projection ${\mathrm{pr}}_Y:X\times Y\to Y$: there is $g:Y\to Z$ with $f=g\circ {\mathrm{pr}}_Y$.

Proof. For each $y\in Y$ the fibre $X\times \{y\}\cong X$ is complete and connected, and $f{|}_{X\times \{y\}}:X\to Z$ is a morphism from a complete connected variety to a separated variety, so its image is a complete connected subvariety of $Z$ of dimension at most $\dim X$. The image is also a point: consider the graph ${\Gamma }_f\subset X\times Y\times Z$ and its projection to $Y\times Z$; the fibre over $y$ is the image $f(X\times \{y\})$. Because $f$ is constant on $X\times \{y_0\}$, the morphism $X\to Z$, $x\mapsto f(x,y)$ is constant at $y=y_0$. The set of $y$ for which this morphism is constant is closed (its complement is the image of the open locus where two points of $X$ map to distinct points of $Z$, which is constructible), and it contains $y_0$; since $X$ is complete and connected the locus where $f(\cdot ,y)$ is constant is also open, hence (connectedness of $Y$ locally) all of $Y$ in a neighbourhood, and globally on each component. Thus $f(x,y)=g(y)$ depends only on $y$, defining $g:Y\to Z$ with $f=g\circ {\mathrm{pr}}_Y$. $\square$

Corollary (abelian varieties are commutative). Every abelian variety $A$ has commutative group law.

Proof. Apply rigidity to the commutator $c(x,y)=x+y-x-y$ with $X=A$, $Y=A$, $y_0=0$. Since $c(x,0)=0$ for all $x$, the map $c$ factors through ${\mathrm{pr}}_Y$, so $c(x,y)=g(y)$ depends only on $y$; but $c(0,y)=0$ forces $g\equiv 0$, hence $c\equiv 0$ and $x+y=y+x$. $\square$

Theorem (seesaw principle). Let $X$ be complete, $\mathcal{L}$ a line bundle on $X\times Y$. Then the locus $Y_0=\{y\in Y:\mathcal{L}{|}_{X\times \{y\}}\simeq {\mathcal{O}}_X\}$ is closed, and if $Y_0=Y$ then $\mathcal{L}\simeq {\mathrm{pr}}_Y^{\ast }\mathcal{M}$ for a line bundle $\mathcal{M}$ on $Y$.

Proof sketch. For a rigidified line bundle, ${\mathrm{pr}}_{Y,\ast }\mathcal{L}$ is a coherent sheaf on $Y$ (properness of $X$) whose fibre at $y$ is $H^0(X,\mathcal{L}{|}_{X\times \{y\}})$ 04.03.01; the condition $\mathcal{L}{|}_{X\times \{y\}}\simeq {\mathcal{O}}_X$ is the vanishing of the determinant of the evaluation map on a generating set of global sections, a closed condition. When it holds globally, a nowhere-vanishing rigidified section renders $\mathcal{L}$ isomorphic to $\mathcal{O}$ on each $X$-fibre and descends to $\mathcal{M}={\mathrm{pr}}_{Y,\ast }\mathcal{L}$. $\square$

Theorem (theorem of the cube, proof sketch). A line bundle $\mathcal{M}$ on $A\times B\times C$ that is $\mathcal{O}$ on the three coordinate slices through $0$ is itself $\mathcal{O}$.

Proof sketch. View $\mathcal{M}$ as a family over $C$: by the seesaw principle it suffices to show $\mathcal{M}{|}_{A\times B\times \{c\}}\simeq {\mathcal{O}}_{A\times B}$ for all $c$, which is the locus of $c$ where the restriction vanishes — a closed condition containing $0$ by hypothesis. An open-and-closed argument on $C$ (and an inductive reduction to $A,B$ both one-dimensional, where the claim is a direct Riemann-Roch count) closes the proof. Full details in Mumford ^{[Mumford 1970]} §13 and Milne ^{[Milne 1986]} §II.6. $\square$

Theorem (Poincaré complete reducibility, proof sketch). For a surjective morphism $f:A\to B$ of abelian varieties there is an abelian subvariety $C\subset A$ such that the addition map $C{\times }_{B}A\to A$ is an isogeny.

Proof sketch. Take a suitable ample $\mathcal{L}$ on $B$, pull it back, and use ${\phi }_{f^{\ast }\mathcal{L}}$ to split the inclusion of the connected kernel; the image of a complement under the theorem of the square gives $C$. The argument is the higher-dimensional analogue of the orthogonal decomposition of a positive-definite lattice. Full proof in Mumford ^{[Mumford 1970]} §19. $\square$

Connections Master

• Elliptic curves as the dimension-one case 04.04.03. An elliptic curve is precisely a one-dimensional abelian variety: the chord-and-tangent law on the Weierstrass cubic $y^2=x^3+ax+b$ is the unique group-object structure on a smooth genus-1 curve with marked point. Every theorem in the present unit — the theorem of the square, the dual, the Tate module — specialises to the elliptic-curve constructions of the prereq unit, with $g=1$ and the $\ell$-adic representation two-dimensional.

• Sheaf cohomology and the Picard scheme 04.03.01. The dual abelian variety $A^{\vee }={\mathrm{Pic}}^0(A)$ is built from sheaf cohomology: $\mathrm{Pic}(A)=H^1(A,{\mathcal{O}}_A^{\times })$, and the connected component ${\mathrm{Pic}}^0$ is the kernel of the degree map read through the long exact cohomology sequence of $0\to {\mathcal{O}}_A^{\times }\to {\mathcal{O}}_A^{\times }\to \mathbb{Z}\to 0$. The seesaw principle itself rests on coherence of higher direct images, a theorem of sheaf cohomology.

• Jacobian variety of a curve 06.06.03. The Jacobian $\mathrm{Jac}(C)={\mathrm{Pic}}^0(C)$ is the universal abelian variety generated by a curve: it carries the Abel-Jacobi image of $C$ and satisfies the universal property that every morphism from $C$ to an abelian variety factors through it. The theta divisor gives a canonical principal polarisation, realising $\mathrm{Jac}(C)$ as a principally polarised abelian variety.

• Abel-Jacobi map and Jacobi inversion 06.06.04. The Abel-Jacobi embedding $C\hookrightarrow \mathrm{Jac}(C)$ and its extension ${\mathrm{Sym}}^{g}C\to \mathrm{Jac}(C)$ give the geometric bridge from the curve to its Jacobian; Jacobi inversion identifies ${\mathrm{Sym}}^{g}C$ birationally with $\mathrm{Jac}(C)$, the higher-genus generalisation of the genus-1 identification $E\simeq \mathrm{Jac}(E)$.

• $\ell$-adic Galois representations 21.05.01. The Tate module $T_{\ell }A$ realises an abelian variety as the source of a $2g$-dimensional $\ell$-adic Galois representation ${\rho }_{A,\ell }:\mathrm{Gal}(\bar{k}/k)\to {\mathrm{GL}}_{2g}({\mathbb{Z}}_{\ell })$, generalising the elliptic-curve representation. The Faltings isogeny theorem and the Tate conjecture for abelian varieties are statements purely about these representations, placing abelian varieties at the centre of the Langlands correspondence for ${\mathrm{GL}}_{2g}$.

Historical & philosophical context Master

Niels Henrik Abel, in his Recherches sur les fonctions elliptiques ^{[Abel 1826]} (Crelle's J. 2, 101–181), inverted the elliptic integral ${\int }^{x}dt/\sqrt{1-t^2}$ and discovered the addition theorem that bears his name — the analytic content of the genus-1 group law. Carl Gustav Jacob Jacobi independently developed the same theory in his Fundamenta nova theoriae functionum ellipticarum ^{[Jacobi 1829]} (Königsberg, 1829), introducing the theta functions that remain the computational engine of abelian-variety theory. Abel and Jacobi also generalised these integrals to arbitrary algebraic curves, producing the Abelian integrals whose inversion problem — solved by Riemann and then by Weierstrass — led directly to the Jacobian construction.

Henri Poincaré, in Sur les groupes des équations linéaires ^{[Poincaré 1902]} (Acta Math. 26), established the complete reducibility theorem for the Picard variety and recognised that the Picard variety of a curve is itself a projective variety carrying a group structure — the Jacobian as an abelian variety avant la lettre. The modern definition of an abelian variety as a complete group variety, together with the theorems of the square and cube, is due to André Weil in Variétés abéliennes et courbes algébriques ^{[Weil 1948]} (Hermann, Paris, 1948), where he also proved the Riemann hypothesis for abelian varieties over finite fields (the Weil bounds) as part of his programme culminating in the Weil conjectures.

David Mumford's Abelian Varieties ^{[Mumford 1970]} (TIFR Studies in Mathematics) systematised the theory via the seesaw principle, the theorem of the cube, and the dual variety, becoming the standard reference; James Milne's Abelian Varieties ^{[Milne 1986]} and Serge Lang's Abelian Varieties ^{[Lang 1983]} refined the arithmetic side, including the Néron-Tate canonical height and the Mordell-Weil theorem in arbitrary dimension. John Tate's $p$-divisible groups ^{[Tate 1966]} and the Serre-Tate criterion for good reduction ^{[Serre-Tate 1968]} placed the Tate module at the centre of the arithmetic theory, the bridge on which the Faltings isogeny theorem and the modularity of abelian varieties now rest.

Bibliography Master

@article{Abel1826,
  author  = {Abel, Niels Henrik},
  title   = {Recherches sur les fonctions elliptiques},
  journal = {J. Reine Angew. Math.},
  volume  = {2},
  year    = {1826},
  pages   = {101--181},
  note    = {Inversion of elliptic integrals and the addition theorem}
}

@book{Jacobi1829,
  author    = {Jacobi, Carl Gustav J.},
  title     = {Fundamenta nova theoriae functionum ellipticarum},
  publisher = {Borntr\"ager},
  address   = {K\"onigsberg},
  year      = {1829}
}

@book{Weil1948,
  author    = {Weil, Andr\'e},
  title     = {Vari\'et\'es ab\'eliennes et courbes alg\'ebriques},
  publisher = {Hermann},
  address   = {Paris},
  year      = {1948},
  note      = {The modern definition of an abelian variety; theorems of the square and cube}
}

@book{Mumford1970,
  author    = {Mumford, David},
  title     = {Abelian Varieties},
  series    = {Tata Institute of Fundamental Research Studies in Mathematics},
  publisher = {Oxford University Press},
  year      = {1970},
  note      = {Seesaw principle, theorem of the cube, dual variety}
}

@book{Milne1986,
  author    = {Milne, James S.},
  title     = {Abelian Varieties},
  publisher = {Cambridge University Press},
  year      = {1986},
  note      = {Current edition at \texttt{jmilne.org}; theorem of the square via the seesaw principle}
}

@book{Lang1983,
  author    = {Lang, Serge},
  title     = {Abelian Varieties},
  publisher = {Springer},
  year      = {1983},
  note      = {Canonical height and Mordell-Weil in arbitrary dimension}
}

@article{SerreTate1968,
  author  = {Serre, Jean-Pierre and Tate, John},
  title   = {Good reduction of abelian varieties},
  journal = {Ann. of Math.},
  volume  = {88},
  year    = {1968},
  pages   = {492--517},
  note    = {The N\'eron-Ogg-Shafarevich criterion via the Tate module}
}

@article{Poincare1902,
  author  = {Poincar\'e, Henri},
  title   = {Sur les groupes des \'equations lin\'eaires},
  journal = {Acta Math.},
  volume  = {26},
  year    = {1902},
  note    = {Complete reducibility and the Picard variety}
}

@incollection{Tate1966,
  author    = {Tate, John},
  title     = {$p$-divisible groups},
  booktitle = {Proc. Conf. Local Fields},
  publisher = {Springer},
  year      = {1966},
  note      = {The Tate module and the $p$-divisible (Barsotti-Tate) group}
}

@book{Serre1988,
  author    = {Serre, Jean-Pierre},
  title     = {Algebraic Groups and Class Fields},
  series    = {Graduate Texts in Mathematics},
  volume    = {117},
  publisher = {Springer},
  year      = {1988},
  note      = {The rigidity lemma and the theorem of the cube}
}

@book{HartshorneAG,
  author    = {Hartshorne, Robin},
  title     = {Algebraic Geometry},
  series    = {Graduate Texts in Mathematics},
  volume    = {52},
  publisher = {Springer},
  year      = {1977},
  note      = {\S III.4--III.7, sheaf cohomology underlying $\mathrm{Pic}$ and $\mathrm{Pic}^0$}
}