Picard group ($\mathrm{Pic}(X)$)

The Picard group is the group of isomorphism classes of line bundles under tensor product.

For any field $k$, degree identifies the Picard group of the projective line with $\mathbb{Z}$, generated by $\mathcal{O}(1)$. A closed point on the affine chart is cut out by an irreducible polynomial $f(t)$, whose divisor is that point minus $\mathrm{deg}(f)$ times infinity. Hence every divisor is linearly equivalent to its degree times infinity. Principal divisors have degree zero, so the degree map is injective as well as surjective. Thus a line bundle of degree $d$ is $\mathcal{O}(d)$.

For a smooth projective curve, the degree-zero Picard group consists of line bundles of degree zero up to isomorphism, with tensor product as addition. A nonzero rational section of a line bundle identifies each line bundle with $\mathcal{O}(D)$ for a divisor on an algebraic curve; degree is the sum of multiplicities weighted by residue-field degrees. Over an algebraically closed field, every degree-zero divisor is a sum of differences of points.

Let $C$ be the smooth projective model over ${\mathbb{F}}_2$ of $y^2+y=x^5+x+1$. Its only rational point is its unique point $p_{\infty }$ at infinity. At that point $x,y$ have pole orders two and five. The degree-two closed point $D$ defined by $y=0$, $x^2+x+1=0$ gives a nonzero class $[D-2p_{\infty }]$. Indeed a function with poles bounded by $2p_{\infty }$ has the form $a+bx$: in the affine coordinate ring every function is uniquely $A(x)+B(x)y$, and the two possible pole orders have opposite parity. No nonzero $a+bx$ vanishes at both roots of $x^2+x+1$. Thus the Abel map of a pointed smooth projective curve on rational points has trivial image here, but the degree-zero Picard group of a curve is nontrivial. Generation by point differences needs an algebraically closed field, or closed points with their residue degrees instead of just rational points.

Given a base point $p_0$, the Abel map takes points of a smooth projective curve to its degree-zero Picard group of a curve. For positive genus it is injective: equality at distinct $p,q$ would give a principal divisor with one simple zero and one simple pole, forcing a degree-one map to the projective line and genus zero. For genus one the Riemann-Roch theorem gives a unique effective degree-one representative of $L(p_0)$ for every degree-zero line bundle $L$, making the map bijective and inducing an abelian group law on the points.

For a smooth projective geometrically connected algebraic curve $C/k$, this functor assigns to a scheme $T$ the line bundles on $C\times T$ having degree $d$ on each geometric fibre, modulo bundles pulled back from $T$. One generally takes the associated sheaf for faithfully flat covers. A rational point of $C$ permits normalization of every line bundle along that point; normalized bundles have no automorphisms preserving their normalization and satisfy effective descent. The resulting functor is represented by the degree-$d$ component of the Picard scheme of a curve.

For a smooth projective algebraic curve of genus $g$, this smooth proper $g$-dimensional scheme represents degree-$d$ line bundles in families. Charts come from nonspecial degree-$g$ bundles: their first sheaf cohomology vanishes, so the Riemann-Roch theorem gives exactly one independent global section, whose zero Cartier divisor varies algebraically. Twisting by fixed line bundles gives charts covering every degree-$d$ class. For degree zero the tensor product gives the Jacobian variety; other components are its torsors.

A degree-$g$ line bundle with vanishing first sheaf cohomology has exactly one section up to scale, by Riemann-Roch theorem. Its zero divisor is unique and varies in families by cohomology and base change for line bundles on a curve. For a fixed divisor $B$ of degree $N-g$, the open subset of $C^{(g)}$ where $\mathcal{O}(E)$ has vanishing first cohomology gives a chart in degree $N$, by $E\mapsto \mathcal{O}(B+E)$. Different charts are glued by the unique residual divisor of $\mathcal{O}(B+E-B^{\prime })$ wherever that line bundle is again nonspecial. Normalizing the universal bundles along a fixed rational point gives their canonical compatible gluing and a representing scheme for the Picard functor of a curve.

The Picard group of a commutative ring $A$ consists of isomorphism classes of finitely generated projective modules that have rank one at every prime ideal. Tensor product is the group law, $[A]$ is the identity, and the module dual supplies the inverse. For an integral domain, every such module can be represented by an invertible fractional ideal after choosing a basis over the fraction field.

The holomorphic Picard group consists of isomorphism classes of holomorphic line bundles, with tensor product as its group law. The trivial bundle is the identity and dualization gives inverses. Transition functions identify it with $H^1(X,{\mathcal{O}}_X^{\ast })$. On the complex projective line every class is $\mathcal{O}(n)$, so degree gives an isomorphism with $\mathbb{Z}$.

For a rank-two holomorphic vector bundle on a nonempty complex manifold, this map from ${\mathrm{Pic}}_{\mathrm{hol}}(X)\times \mathbb{Z}$ to ${\mathrm{Pic}}_{\mathrm{hol}}(\mathbb{P}(E))$ is injective. Fibre degree detects $n$. If $p^{\ast }M$ is trivial, a nonvanishing holomorphic section is constant along every compact projective fibre in a local bundle chart, so it descends to a nonvanishing section of $M$. The proof requires no global section of $p$.

For a smooth integral variety, every Weil divisor is a Cartier divisor, so the Picard group equals the divisor class group. If $U$ is obtained by removing irreducible Weil divisors $D_j$, the localization sequence for the divisor class group gives the displayed exact sequence. A rational trivialization of a line bundle on $U$ has divisor supported on the removed divisors, describing the kernel concretely.

On an irreducible variety, a Cartier divisor is a section of ${\mathcal{K}}_X^{\ast }/{\mathcal{O}}_X^{\ast }$: its local rational equations differ by regular units. Their ratios define an invertible sheaf, and a global rational equation gives the trivial class. Conversely a rational trivialization of an invertible sheaf gives local equations. The rational-function sheaf is flasque, so its long exact sequence in sheaf cohomology also identifies the quotient with $H^1(X,{\mathcal{O}}_X^{\ast })$.

The Néron-Severi group is the quotient $\mathrm{Pic}(X)/{\mathrm{Pic}}^0(X)$ of the Picard group by line bundles algebraically equivalent to zero.