04.05.02 · algebraic-geometry / divisors

Picard group

shipped3 tiersLean: partial

Anchor (Master): Hartshorne §II–§III; Vakil; Mumford *Abelian Varieties* Ch. III; Grothendieck FGA exposé 232

Intuition [Beginner]

The Picard group $Pic (X)$ of a scheme $X$ is the set of all line bundles on $X$ , considered up to isomorphism, with group operation given by tensor product. The identity element is the structure-sheaf bundle $O_{X}$ ; the inverse of a line bundle $L$ is its dual $L^{- 1}$ .

The Picard group is one of the most fundamental invariants of an algebraic variety. It records every way a one-dimensional vector bundle can be twisted over $X$ , and it controls the geometry of embeddings into projective space. For a smooth projective curve of genus $g$ , the connected component of $Pic$ — denoted $Pic^{0} (X)$ — is the Jacobian variety, a $g$ -dimensional abelian variety carrying the deepest structure of the curve.

Émile Picard introduced an analytic version in 1895 (the Picard variety of an algebraic surface). Grothendieck's 1962 Bourbaki seminar made the construction algebraic and functorial: the Picard scheme $Pic_{X / S}$ represents the functor of relative line bundles. The algebraic Picard group is what links divisors, line bundles, and cohomology into a single object.

Visual [Beginner]

A scheme $X$ with several local trivialisations of a line bundle, glued by transition functions on overlaps; different line bundles correspond to different homotopy classes of transition data, and the Picard group counts these classes.

Worked example [Beginner]

The Picard group of the projective line $P_{k}^{1}$ over an algebraically closed field is $Z$ , generated by the twisting sheaf $O (1)$ . Every line bundle on $P_{k}^{1}$ is isomorphic to $O (d)$ for a unique integer $d$ , the degree of the line bundle.

Concretely: $O (1)$ is the line bundle whose sections are the linear forms $a_{0} x_{0} + a_{1} x_{1}$ . Its tensor square — the line bundle $O (2)$ — has sections the homogeneous quadratics. The dual $O (- 1)$ is the tautological bundle whose fibre over a point $[v]$ is the one-dimensional subspace $k v \subset k^{2}$ .

Tensor product corresponds to addition of degrees: tensoring $O (d)$ with $O (e)$ gives $O (d + e)$ . So $Pic (P_{k}^{1}) = Z$ as an abelian group, with the integer $d$ recording the degree.

For an elliptic curve $E$ over $k = C$ : $Pic (E) = Z \oplus E (C)$ . The integer is the degree; the second factor is the elliptic curve itself, recording line bundles of degree 0. This second piece is the Jacobian, and for an elliptic curve the Jacobian equals the curve.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $X$ be a scheme. The Picard group of $X$ is the abelian group

Pic (X) := {isomorphism classes of line bundles on X},

with group operation $[L] \cdot [M] := [L \otimes_{O_{X}} M]$ , identity $[O_{X}]$ , and inverse $[L]^{- 1} := [L^{\lor}] = [H o m_{O_{X}} (L, O_{X})]$ .

Three equivalent descriptions.

(P1) Sheaf-cohomological. The exponential sequence for the multiplicative group sheaf gives

Pic (X) ≅ H^{1} (X; O_{X}^{\times}),

where $O_{X}^{\times}$ is the sheaf of invertible regular functions. Čech cohomology with respect to a trivialising cover ${U_{i}}$ : a class is a 1-cocycle of transition functions $g_{ij} \in O_{X}^{\times} (U_{i} \cap U_{j})$ satisfying $g_{ij} g_{j k} = g_{ik}$ , modulo coboundaries $g_{ij} = h_{i} h_{j}^{- 1}$ .

(P2) Cartier divisor classes. The map sending a Cartier divisor $D = {(U_{i}, f_{i})}$ to the line bundle $O (D)$ with cocycle $g_{ij} = f_{i} / f_{j}$ induces an isomorphism

CaCl (X) := CaDiv (X) / CaDiv^{princ} (X) \sim Pic (X)

between Cartier divisor classes (Cartier divisors modulo principal divisors) and the Picard group. This is the Cartier-divisor presentation of $Pic (X)$ — see [Cartier divisor]04.05.04.

(P3) Weil divisor classes (locally factorial case). On a Noetherian normal scheme that is locally factorial (every local ring a UFD — e.g., smooth varieties), Weil and Cartier divisors coincide, and

Cl (X) := Div (X) / PrincDiv (X) \sim Pic (X) .

In general (singular or non-locally-factorial), $Pic (X) \subseteq Cl (X)$ is a strict subgroup; the quotient measures the failure of local factoriality.

Functoriality. A morphism $f : X \to Y$ induces a group homomorphism $f^{*} : Pic (Y) \to Pic (X)$ via pullback of line bundles. So $Pic$ is a contravariant functor from schemes to abelian groups.

The relative Picard scheme. For a morphism $f : X \to S$ , the Picard functor $Pic_{X / S}$ assigns to a scheme $T \to S$ the group $Pic (X \times_{S} T) / Pic (T)$ . Under mild hypotheses (proper, flat, geometrically integral fibres with $f_{*} O_{X} = O_{S}$ ), this functor is representable — by an algebraic space, and for projective $X$ over a field $S$ , by a scheme — the Picard scheme $Pic_{X / S}$ (Grothendieck FGA, 1962).

Examples.

Affine space: $Pic (A_{k}^{n}) = 0$ . Every line bundle on affine space is structure-sheaf isomorphic.
Projective space: $Pic (P_{k}^{n}) = Z$ , generated by $O (1)$ .
Smooth projective curve $C$ of genus $g$ : $Pic (C) = Z \oplus Pic^{0} (C)$ , where $Pic^{0} (C)$ is the Jacobian, a $g$ -dimensional abelian variety.
Elliptic curve $E$ : $Pic^{0} (E) ≅ E$ as algebraic groups.
Spectrum of a Dedekind domain: $Pic (Spec (O_{K})) = Cl (O_{K})$ , the ideal class group.
Affine cone over a smooth conic: $Pic (X) = 0$ but $Cl (X) = Z /2$ — a textbook example where $Pic ⊊ Cl$ .

Connected component. When $Pic_{X / S}$ exists as a scheme, its connected component containing the identity is denoted $Pic_{X / S}^{0}$ . For $X$ a smooth projective variety over an algebraically closed field, $Pic_{X}^{0}$ is an abelian variety of dimension $h^{1} (X; O_{X})$ — Hodge-theoretically, the dimension of the $H^{0, 1}$ piece of $H^{1}$ .

Key theorem with proof [Intermediate+]

Theorem (Picard group as $H^{1}$ of $O^{\times}$ ). For any scheme $X$ , there is a natural isomorphism

Pic (X) ≅ H^{1} (X; O_{X}^{\times}),

where $O_{X}^{\times}$ is the sheaf of multiplicative units in $O_{X}$ .

Proof. The strategy is the standard Čech-cohomology classification of fibre bundles: line bundles, like principal $O_{X}^{\times}$ -bundles, are classified by their $H^{1}$ cocycle data.

Step 1 (Cocycle from a line bundle). Let $L$ be a line bundle on $X$ . Choose an open cover ${U_{i}}$ on which $L$ trivialises: $L ∣_{U_{i}} ≅ O_{U_{i}}$ via an isomorphism $φ_{i} : O_{U_{i}} \to L ∣_{U_{i}}$ . On overlaps $U_{i} \cap U_{j}$ , the composition

φ_{j}^{- 1} \circ φ_{i} : O_{U_{i} \cap U_{j}} \to O_{U_{i} \cap U_{j}}

is multiplication by an element $g_{ij} \in O_{X}^{\times} (U_{i} \cap U_{j})$ . The cocycle condition $g_{ij} g_{j k} = g_{ik}$ on triple overlaps $U_{i} \cap U_{j} \cap U_{k}$ follows from associativity of composition.

The class $[(g_{ij})] \in \overset{ˇ}{H}^{1} ({U_{i}}; O_{X}^{\times})$ is independent of the chosen trivialisations: a different choice $φ_{i}^{'} = h_{i} \cdot φ_{i}$ for $h_{i} \in O_{X}^{\times} (U_{i})$ shifts the cocycle by the coboundary $g_{ij}^{'} = h_{i}^{- 1} g_{ij} h_{j}$ .

Step 2 (Line bundle from a cocycle). Conversely, given a 1-cocycle $(g_{ij})$ on a cover ${U_{i}}$ , define a line bundle $L$ by gluing copies of $O_{U_{i}}$ along the overlap maps $φ_{ij} = g_{ij} : O_{U_{i} \cap U_{j}} \to O_{U_{i} \cap U_{j}}$ . The cocycle condition ensures the gluing is consistent on triple overlaps.

Step 3 (Refinement and direct limit). Passing to the direct limit over open covers gives Čech cohomology $\overset{ˇ}{H}^{1} (X; O_{X}^{\times})$ ; for any scheme this agrees with sheaf cohomology $H^{1} (X; O_{X}^{\times})$ (by Cartan's criterion or direct comparison on a basis of distinguished opens).

Step 4 (Group structure compatibility). The bijection $Pic (X) \to H^{1} (X; O_{X}^{\times})$ takes tensor product to cocycle multiplication: if $L, M$ have cocycles $(g_{ij}), (h_{ij})$ on a common refinement, then $L \otimes M$ has cocycle $(g_{ij} h_{ij})$ . The dual $L^{\lor}$ has cocycle $(g_{ij}^{- 1})$ . So the map respects the group structure.

The bijection is therefore a group isomorphism. $□$

This identification — the Picard group equals the first cohomology of the multiplicative sheaf — is the conceptual heart of the theory. It links three perspectives: line bundles (geometric), Cartier divisors (algebraic), and sheaf cohomology (homological). Every later result about $Pic (X)$ is some refinement of this trio.

Bridge. The construction here builds toward 04.05.05 (ample and very ample line bundle), where the same data is upgraded, and the symmetry side is taken up in 04.07.01 (projective space). The defining pattern appears again in those units in a sharpened form, where the local data is glued or quotiented. Putting these together, the foundational insight is that the data of this unit gives the structural signature that the rest of the strand reads off.

Exercises [Intermediate+]

Exercise 3 (medium, proof).

Show that for any morphism $f : X \to Y$ of schemes, the pullback $f^{*} : Pic (Y) \to Pic (X)$ is a group homomorphism.

Hint

Pullback commutes with tensor product and preserves the structure sheaf.

Answer

For line bundles $L, M$ on $Y$ , the natural map

f^{*} (L \otimes_{O_{Y}} M) \to f^{*} L \otimes_{O_{X}} f^{*} M

is an isomorphism (this holds for any quasi-coherent sheaves and tensor product, by the projection formula on stalks). Locally on a trivialising cover where $L, M$ are both isomorphic to $O_{Y}$ , the tensor product is $O_{Y}$ , the pullback of each side is $O_{X}$ , and the natural map is the identity.

Pullback also sends the structure sheaf to the structure sheaf: $f^{*} O_{Y} = O_{X}$ by definition of pullback of sheaves of rings.

So $f^{*}$ takes $[O_{Y}] \mapsto [O_{X}]$ and $[L] \cdot [M] \mapsto f^{*} [L] \cdot f^{*} [M]$ . This is the definition of a group homomorphism. $□$

The functoriality $Pic : Sch^{op} \to Ab$ is what makes the Picard scheme construction natural: the relative Picard functor is contravariantly representable.

Exercise 4 (medium, numeric).

Compute $Pic (P_{k}^{1} \times P_{k}^{1})$ for $k$ algebraically closed.

Hint

For products, the Künneth-type formula gives a sum of Picard groups of factors plus correspondences.

Answer

For smooth projective varieties $X, Y$ over $k$ algebraically closed with $H^{1} (X; O_{X}) = H^{1} (Y; O_{Y}) = 0$ (rationally connected, e.g., projective spaces),

Pic (X \times_{k} Y) = Pic (X) \oplus Pic (Y) .

For $X = Y = P_{k}^{1}$ , $Pic (P_{k}^{1}) = Z$ , so $Pic (P_{k}^{1} \times P_{k}^{1}) = Z^{2}$ .

The two generators are $O (1, 0)$ (pullback of $O (1)$ from the first factor) and $O (0, 1)$ (from the second factor). A general line bundle is $O (a, b) = π_{1}^{*} O (a) \otimes π_{2}^{*} O (b)$ with bidegree $(a, b) \in Z^{2}$ .

In contrast, on a product of elliptic curves $E_{1} \times E_{2}$ , $Pic (E_{1} \times E_{2}) = Pic (E_{1}) \oplus Pic (E_{2}) \oplus Hom (E_{1}, E_{2}^{\lor})$ — the third factor (correspondences) appears whenever the Picard scheme has a positive-dimensional component.

Exercise 5 (medium, conceptual).

Why is the Picard scheme $Pic_{X / S}$ a group scheme, not just a scheme?

Hint

It represents a functor with values in abelian groups.

Answer

The Picard functor $Pic_{X / S} : Sch_{S}^{op} \to Ab$ takes values in abelian groups, not merely sets: for each $T$ , $Pic_{X / S} (T) = Pic (X \times_{S} T) / Pic (T)$ is the abelian group of line bundles up to isomorphism.

When such a functor is representable by a scheme $Pic_{X / S}$ , the group structure transfers via the Yoneda lemma: the multiplication map $Pic_{X / S} \times Pic_{X / S} \to Pic_{X / S}$ on the functor side becomes a morphism of schemes $Pic_{X / S} \times_{S} Pic_{X / S} \to Pic_{X / S}$ , and the inverse map $Pic_{X / S} \to Pic_{X / S}$ becomes a scheme morphism. Together with the identity section, this realises $Pic_{X / S}$ as a commutative group scheme over $S$ .

For $X$ a smooth projective variety over an algebraically closed field, the connected component $Pic_{X}^{0}$ is a projective abelian group scheme, equivalently an abelian variety — a higher-dimensional generalisation of an elliptic curve. This is the algebraic geometry of moduli of line bundles.

Exercise 6 (hard, proof).

Show that for the affine cone $X = V (x z - y^{2}) \subset A_{k}^{3}$ over a smooth conic, $Pic (X) = 0$ but $Cl (X) = Z /2$ .

Hint

Affine schemes with rational singularities have vanishing Picard; the line through the cone point is a non-Cartier Weil divisor.

Answer

Vanishing of $Pic$ . The cone $X$ is affine and has rational singularities at the origin (it is a quotient by a finite group action on $A^{2}$ , namely $A^{2} / (\pm 1)$ ). For affine varieties with mild singularities, every line bundle is structure-sheaf isomorphic: this follows from the fact that the structure sheaf has all higher cohomology zero on an affine, and the exponential sequence reduces $Pic$ to the contribution of $π_{1}$ , which vanishes here. More directly: $X = Spec k [x, y, z] / (x z - y^{2})$ , and rank-1 projective modules over this ring are all free (computed via the local-to-global principle and the rationality of the singularity). So $Pic (X) = 0$ .

Non-vanishing of $Cl$ . The line $L = V (x, y) \cap X = {(0, 0, t) : t \in k}$ is an irreducible codimension-1 closed subscheme, hence a prime Weil divisor. Two copies of $L$ are linearly equivalent to a principal divisor: $2 [L] = div (x)$ , since the function $x$ vanishes to order 2 along $L$ (the equation $x z = y^{2}$ shows that $x = y^{2} / z$ has order 2 in $L$ ). So $L$ has order 2 in $Cl (X)$ .

To see $L$ is nonzero: if $L = div (f)$ for some rational $f$ , then $f$ would have a single simple zero along $L$ and no other zeros or poles. But this forces $f$ to lift to an element of $k [x, y, z] / (x z - y^{2})$ generating the prime $(x, y)$ — and there is no such element, because the ideal $(x, y)$ in the local ring at the cone point is not principal (the local ring at the cone point is normal but not factorial).

So $Cl (X) = Z /2$ , generated by $[L]$ , while $Pic (X) = 0$ . The quotient $Cl (X) / Pic (X) = Z /2$ is the local class group at the cone point, measuring the failure of local factoriality. $□$

Exercise 7 (hard, conceptual).

Explain the geometric meaning of $Pic^{0} (C)$ for a smooth projective curve $C$ of genus $g$ over $C$ , and connect to the analytic Jacobian.

Hint

Degree-0 line bundles on a curve correspond to points on the Jacobian via the Abel-Jacobi map.

Answer

For a smooth projective curve $C / C$ of genus $g$ :

$Pic (C) = Z \oplus Pic^{0} (C)$ , with the $Z$ recording the degree of a line bundle.
$Pic^{0} (C)$ is a $g$ -dimensional connected abelian variety, the Jacobian $J (C)$ .

Analytic description. Let $Ω_{C}^{1}$ be the sheaf of holomorphic 1-forms on $C$ , with $H^{0} (C; Ω_{C}^{1}) ≅ C^{g}$ . Choose a basis $ω_{1}, \dots, ω_{g}$ and a symplectic basis of integer cycles $α_{1}, \dots, α_{g}, β_{1}, \dots, β_{g} \in H_{1} (C; Z)$ . The period lattice $Λ \subset C^{g}$ is generated by the $2 g$ vectors $(\int_{α_{i}} ω_{j})_{j}$ and $(\int_{β_{i}} ω_{j})_{j}$ . The Jacobian is

J (C) = C^{g} /Λ,

a complex torus of dimension $g$ that carries a polarisation (from the intersection pairing on $H_{1}$ ) making it a principally polarised abelian variety.

Abel-Jacobi map. For a base point $p_{0} \in C$ , the map $C \to J (C)$ , $p \mapsto (\int_{p_{0}}^{p} ω_{1}, \dots, \int_{p_{0}}^{p} ω_{g}) mod Λ$ , extends to an isomorphism $AJ : Pic^{0} (C) \sim J (C)$ on degree-0 divisor classes. The class of a degree-0 divisor $D = \sum n_{i} [p_{i}]$ goes to $\sum n_{i} \int_{p_{0}}^{p_{i}} ω_{j} mod Λ$ .

Algebraic perspective. Grothendieck's Picard scheme construction gives $Pic_{C}^{0}$ as the connected component of $Pic_{C / C}$ containing the identity. Its tangent space at the origin is $H^{1} (C; O_{C}) ≅ C^{g}$ — the Hodge $H^{0, 1}$ piece of $H^{1}$ , dual to $H^{0} (C; Ω_{C}^{1})$ .

The two descriptions — analytic via 1-forms and periods, algebraic via Picard scheme — agree by GAGA. The Jacobian is one of the most important constructions in the geometry of curves: it encodes the moduli of degree-0 line bundles, generalises the elliptic-curve case ( $g = 1$ ), and is the receiving object for the Abel-Jacobi map. Mumford's Curves and their Jacobians and Birkenhake-Lange Complex Abelian Varieties are the canonical references.

Lean formalization [Intermediate+]

lean_status: partial — Mathlib has invertible-module API and the group structure on isomorphism classes; the full Picard-group construction for general schemes is partially formalised.

[object Promise]

Advanced results [Master]

The Picard scheme. For $f : X \to S$ a proper, flat morphism with $f_{*} O_{X} = O_{S}$ universally and geometrically integral fibres, the relative Picard functor $Pic_{X / S}$ is representable by an algebraic space (Artin), and by a scheme when $X$ is projective over $S$ (Grothendieck FGA exposé 232, 1962). The Picard scheme $Pic_{X / S}$ is locally of finite presentation, separated, smooth in characteristic 0; its connected component $Pic_{X / S}^{0}$ is a separated group scheme of finite type, an abelian scheme when $X$ has good reduction.

Néron-Severi group. The quotient $NS (X) := Pic (X) / Pic^{0} (X)$ is the Néron-Severi group, a finitely generated abelian group (Néron-Severi-Lang theorem of finiteness). Its rank is the Picard number $ρ (X)$ . For smooth projective $X / C$ , $NS (X) \otimes Q$ is the image of $Pic (X) \otimes Q$ in $H^{1, 1} (X; Q)$ — the Hodge classes of type $(1, 1)$ that are rational. The Hodge conjecture in degree 2 says all rational $(1, 1)$ -classes are algebraic, equivalently that the natural map $NS (X) \otimes Q \to H^{1, 1} (X) \cap H^{2} (X; Q)$ is an isomorphism (proved by Lefschetz for $H^{2}$ ).

Brauer group connection. The exponential exact sequence for étale topology gives a long exact sequence

\dots \to H^{1} (X; G_{m}) \to H^{2} (X; Z) \to H^{2} (X; O_{X}) \to H^{2} (X; O_{X}^{\times}) = Br (X) \to H^{3} (X; Z) \to \dots,

linking $Pic (X) = H^{1}$ to the Brauer group $Br (X) = H^{2}$ . The Brauer group classifies Azumaya algebras / projective bundles up to twist; it is the next degree of the "multiplicative cohomology" tower.

Picard variety in arithmetic. For a curve $C$ over a number field $K$ , the Mordell-Weil theorem says $J (C) (K) = Pic^{0} (C) (K)$ is a finitely generated abelian group. The rank is one of the deepest arithmetic invariants of $C$ (related to BSD for elliptic curves).

The Albanese variety. For a smooth projective variety $X / k$ , the Albanese variety $Alb (X)$ is the dual abelian variety to $Pic^{0} (X)$ . The Albanese map $X \to Alb (X)$ generalises the Abel-Jacobi map for curves and is the universal map from $X$ to an abelian variety.

Mumford's perspective. Mumford's Abelian Varieties (Ch. III) develops the Picard scheme of an abelian variety: $Pic^{0} (A) = A^{\lor}$ is the dual abelian variety, with the Poincaré bundle on $A \times A^{\lor}$ as the universal line bundle. This duality $A \leftrightarrow A^{\lor}$ is foundational for Fourier-Mukai transforms (Mukai 1981) and modern derived-category methods.

Higher Picard. $Pic (X)$ is the degree-1 case of higher line bundle stacks: $Pic (X) = π_{0}$ of the gerbe $\underline{Pic} (X)$ , which has $π_{1} = H^{0} (X; O^{\times})$ (the units of global functions). This is the start of an infinite tower related to Brauer groups, gerbes, and higher étale cohomology.

Synthesis. This construction generalises the pattern fixed in 04.05.01 (weil divisor), with the symmetric data replaced by its skew or twisted analogue. Read in the opposite direction, the construction is dual to the metric story: complements and orthogonality are taken with respect to the bilinear datum of this unit, not a metric, and the resulting category of subobjects is the one the rest of the strand classifies. The central insight is that this datum identifies algebra with geometry: functions become vector fields, subspaces become quotients, and invariants become cohomology classes — and that identification is the engine driving every theorem downstream.

Full proof set [Master]

Detailed proofs of: representability of the Picard functor (Grothendieck FGA), Néron-Severi finiteness (Néron-Severi-Lang), the structure of $Pic^{0}$ for smooth projective varieties as an abelian variety, and Mumford's Picard scheme of an abelian variety with the Poincaré bundle — these are deferred to companion units in the Hodge theory and abelian-varieties strands. The basic correspondence $Pic (X) = H^{1} (X; O_{X}^{\times})$ is proved in the formal-definition section.

Connections [Master]

Weil divisor 04.05.01 — on locally factorial schemes, $Pic (X) = Cl (X)$ ; in general, $Pic \subseteq Cl$ .
Line bundle 04.05.03 — $Pic (X)$ is by definition the abelian group of line bundles up to isomorphism.
Cartier divisor 04.05.04 — Cartier divisor classes equal the Picard group exactly.
Sheaf cohomology 04.03.01 — $Pic (X) = H^{1} (X; O_{X}^{\times})$ is a foundational sheaf-cohomology computation.
Projective space 04.07.01 — $Pic (P_{k}^{n}) = Z$ generated by $O (1)$ .
Riemann-Roch theorem for curves 04.04.01 — Riemann-Roch is naturally a statement about $Pic (C)$ for a curve.
Ample line bundle 04.05.05 — ampleness is a positivity condition on classes in $Pic (X)$ .
Coherent sheaf 04.06.02 — line bundles are rank-1 locally free coherent sheaves; the Picard group sits inside the K-theory of coherent sheaves.

Historical & philosophical context [Master]

The Picard group has two distinct origins. Émile Picard introduced an analytic prototype in his 1895 work Sur les fonctions de deux variables indépendantes (and earlier papers from 1882 onward), studying transcendental and algebraic curves on a complex algebraic surface. Picard's analytic Picard variety $Pic (S)$ of a surface $S$ recorded the moduli of algebraic curves up to linear equivalence — what we now call divisor classes. He showed that for a smooth projective surface over $C$ , this moduli space carries a complex-analytic group structure, and (with Lefschetz's later refinements) that its connected component is a complex torus of dimension $h^{1, 0} (S) = h^{0} (S; Ω^{1})$ , the analytic Picard variety. The terminology Picard variety and Picard number honour his foundational work; the contemporary phrase "the Picard group" uses Picard's name where he himself wrote of systèmes algébriques of curves.

The algebraic version was constructed by André Weil (in arithmetic, for Jacobians of curves over number fields) and made fully scheme-theoretic by Alexander Grothendieck in his 1962 Bourbaki seminar exposé 232 (Fondements de la géométrie algébrique, FGA). Grothendieck's insight was representability: the relative Picard functor $T \mapsto Pic (X_{T}) / Pic (T)$ is representable by an algebraic scheme — the Picard scheme $Pic_{X / S}$ — under mild proper-flat hypotheses. The proof (FGA exposé 232) uses Hilbert-scheme techniques and the universal-divisor construction. In Grothendieck's framing the Picard group is no longer an isolated invariant but the value at $Spec k$ of a representable functor of relative line bundles. This functorial perspective controls families: a flat family $X \to S$ of projective varieties has a flat family $Pic_{X / S}^{0} \to S$ of abelian schemes, and the geometry of moduli reduces to the geometry of these Picard schemes.

David Mumford's 1965 Geometric Invariant Theory and his 1970 Abelian Varieties established the Picard scheme as the foundation of abelian-variety theory. Mumford showed (Ch. III) that for an abelian variety $A$ , the Picard scheme $Pic^{0} (A) = A^{\lor}$ is the dual abelian variety, and the Poincaré line bundle on $A \times A^{\lor}$ is the universal degree-zero line bundle. The duality $A \leftrightarrow A^{\lor}$ is the foundation of the Fourier-Mukai transform (Mukai 1981) and the modern derived-category techniques in algebraic geometry.

In contemporary research, Picard groups remain central. The Mordell-Weil theorem gives finite generation of $Pic^{0} (C) (K)$ for a curve $C$ over a number field $K$ ; its rank is the subject of the Birch-Swinnerton-Dyer conjecture for elliptic curves. The Néron-Severi group and Picard number control the geometry of algebraic surfaces (via the Hodge index theorem and Lefschetz $(1, 1)$ -theorem). Picard schemes appear throughout moduli theory (the universal Picard scheme over $M_{g}$ ), mirror symmetry (the symplectic mirror exchanges Picard and Albanese), and arithmetic geometry (heights on Picard schemes, equidistribution). The single object $Pic (X) = H^{1} (X; O_{X}^{\times})$ thus carries an enormous arithmetic, geometric, and homological load.

Bibliography [Master]

Hartshorne, Algebraic Geometry — §II.6 (Picard group), §III.4 (cohomology and Picard scheme references).
Vakil, The Rising Sea: Foundations of Algebraic Geometry — §14 (Picard group), §28 (Picard scheme).
Grothendieck, FGA exposé 232 (Bourbaki seminar 1962) — the algebraic Picard scheme construction.
Picard, Sur les fonctions de deux variables indépendantes (1895) — origin of the analytic Picard variety.
Mumford, Abelian Varieties — Ch. III, Picard scheme of an abelian variety, Poincaré bundle.
Mumford, Curves and their Jacobians — the Jacobian as $Pic^{0}$ of a curve.
Birkenhake-Lange, Complex Abelian Varieties — analytic and algebraic perspectives on $Pic^{0}$ .
Bosch-Lütkebohmert-Raynaud, Néron Models — Picard schemes in arithmetic.
Lazarsfeld, Positivity in Algebraic Geometry I — Picard, Néron-Severi, and ample cones.
Kleiman, The Picard Scheme (FGA Explained, 2005) — modern exposition of Grothendieck's construction.

Prerequisites

04.05.01
04.05.03
04.05.04

Used in

04.05.05
04.07.01

Tier anchors

beginner: Strogatz analogous level for the group of line bundles up to isomorphism
intermediate: Hartshorne §II.6, §III.4; Vakil §14, §28; Eisenbud-Harris §III
master: Hartshorne §II–§III; Vakil; Mumford *Abelian Varieties* Ch. III; Grothendieck FGA exposé 232

References

TODO_REF
Hartshorne — Algebraic Geometry · §II.6 (Picard group), §III.4 (Picard scheme)
TODO_REF
Vakil — The Rising Sea · §14, Picard group; §28, Picard scheme
TODO_REF
Picard — Sur les fonctions de deux variables indépendantes · 1895, original analytic Picard variety
TODO_REF
Grothendieck — FGA exposé 232 (Bourbaki seminar 1962) · the algebraic Picard scheme
TODO_REF
Mumford — Abelian Varieties · Ch. III, Picard scheme of an abelian variety

Lean module

Codex.AlgebraicGeometry.Divisors.PicardGroup

Reviewer

TBD

Estimated time

beginner: 12m
intermediate: 35m
master: 55m