Weil divisor

A Weil divisor on a Noetherian integral scheme is a finite integer combination of integral codimension-one closed subschemes.

A Weil divisor is Q-Cartier if some positive integer multiple is a Cartier divisor. Its Cartier index is the smallest such integer. It is ample in the Q-Cartier sense when such a multiple is an ample Cartier divisor. This need not make the original divisor itself Cartier.

A linear system of divisors is a projective family of effective Weil divisors linearly equivalent to a fixed divisor. A complete linear system of a divisor contains every effective representative of that class; a projective subspace gives a subsystem. For a Cartier divisor, the family is obtained from global sections of its associated line bundle. The algebraic curve case recovers the usual moving family of effective divisors.

On a smooth variety over a field of characteristic zero, a general member of a basepoint-free linear system of divisors is smooth. More generally it is smooth outside the base locus. Irreducibility is a separate assertion and requires suitable hypotheses on the associated morphism; a pencil composed with a map to a curve can have disconnected general divisors.

On a smooth projective curve of genus two, $D=K+P+Q$ has three sections and no base points by the Riemann-Roch theorem. Its map to ${\mathbb{P}}^2$ is nondegenerate and has mapping degree times image degree four. The only nonbirational possibility is a double cover of a conic. Every degree-two divisor with two sections is canonical, again by Riemann-Roch theorem, so that possibility forces $D\sim 2K$, or $P+Q\sim K$. If this equivalence fails, the image is a quartic and the map is birational. The quartic is singular because a smooth plane quartic has genus three.

A linear system is base-point-free if its sections have no common zero. Their ratios then define an everywhere-defined morphism to projective space whose hyperplane pullbacks are members of the original system.

A very ample linear system defines an embedding into projective space. On a smooth projective curve this means it has no base points and separates both distinct points and tangent directions; the complete system of a very ample divisor is an example.

For a finite generating space $V\subseteq H^0(X,\mathcal{L})$ on a projective scheme, its map to projective space is a closed immersion exactly when, after algebraic closure of the ground field, the displayed evaluation is surjective for every length-two closed subscheme $E$. Two distinct points test point separation; a double point tests a direction in the Zariski tangent space. For the complete section space, this characterizes a very ample line bundle.

For $D={\sum }_P{a}_{P}P$, its support is the union of the prime divisors $P$ with $a_P\ne 0$. Negative as well as positive coefficients contribute. A point outside the support has a neighbourhood on which the divisor is zero.

Two Weil divisors are linearly equivalent when their difference is a principal Weil divisor. This is the equivalence relation defining the divisor class group, and applies to normal varieties of any dimension, not just curves.

If a Weil divisor $D$ on a normal variety is not Cartier at $x$, every linearly equivalent divisor has $x$ in its support of a Weil divisor. Otherwise that representative would vanish near $x$, expressing $D$ as a principal Weil divisor there. On $V(ac-b^2)$ the prime divisor $V(a,b)$ is not Cartier at the origin: its ideal has two independent generators modulo the maximal ideal times that ideal. This gives an unavoidable support point even for representatives with negative coefficients.

For a normal variety with affine coordinate ring $A$ and a Weil divisor $D$, the fractional ideal of functions whose principal Weil divisor is at least $D$ isFor an effective prime divisor this is its height-one prime ideal. If the divisor is Cartier locally, this ideal is locally principal. A nonprincipal height-one prime ideal in a normal local ring can therefore detect a non-Cartier divisor.

On a normal integral Noetherian scheme, a nonzero rational function defines the displayed Weil divisor, with the sum over codimension-one prime divisors. The local ring at each such divisor is a discrete valuation ring, so the order is well defined. The divisor class group is the group of Weil divisors modulo principal Weil divisors. On a smooth curve this agrees with a principal divisor on an algebraic curve.

A prime Weil divisor is one integral codimension-one closed subscheme. At its generic point, regularity in codimension one supplies a discrete valuation and hence an order of vanishing.

The divisor class group is the group of Weil divisors modulo principal divisors of rational functions.

For positive $n,m$, $\mathrm{Cl}({\mathbb{P}}^n\times {\mathbb{P}}^m)={\mathbb{Z}}^2$, generated freely by the pullbacks of coordinate hyperplanes. Restrict a divisor to the standard affine chart, where unique factorization makes it principal, and subtract this principal Weil divisor to leave only the two boundary hyperplanes. Any principal relation between them would be given by a rational function whose divisor vanishes on the chart; unique factorization then makes that function a constant, proving independence. The same argument gives $\mathrm{Cl}({\mathbb{P}}^r)=\mathbb{Z}$.

If $C$ is an irreducible plane curve of degree $d$, the localization sequence for the divisor class group quotients $\mathrm{Cl}({\mathbb{P}}^2)=\mathbb{Z}[H]$ by the class $[C]=d[H]$. Nonsingularity of the removed curve is not required; irreducibility makes it a single removed prime Weil divisor.

For a smooth irreducible algebraic curve, send a divisor on an algebraic curve $D={\sum }_P{n}_{P}P$ to the line bundle associated to a divisor defined locally by ${\mathcal{O}}_X(D{)}_P=t_P^{-n_P}{\mathcal{O}}_{X,P}$. A nonzero rational section of a line bundle gives its inverse construction, and changing that section changes its divisor by a principal divisor. Thus the divisor class group is the Picard group. The quotient of the sheaf of nonzero rational functions on an irreducible variety by the sheaf of units of the structure sheaf has stalk $\mathbb{Z}$ at each closed point, via the discrete valuation. Its global sections are finite sums of points. The corresponding long exact sequence in sheaf cohomology, and the vanishing for the rational-function sheaf, identify both groups with $H^1(X,{\mathcal{O}}_X^{\ast })$.

For an elliptic curve over an algebraically closed field, ${\mathrm{Pic}}^0(E)\cong E(k)$ via $P\mapsto [P-O]$. This group is infinite and divisible by any prime $\ell$ different from the characteristic, since $[\ell ]:E\to E$ has nonzero derivative and is a surjective morphism of projective curves. A finitely generated abelian group divisible by $\ell$ has zero free rank and is finite. Hence $E(k)$, and therefore $\mathrm{Cl}(E)$, cannot be finitely generated. This proof applies to countable algebraically closed fields as well.

A Noetherian ring that is an integrally closed domain is a unique factorization domain exactly when its divisor class group vanishes. Indeed, vanishing says every prime Weil divisor, equivalently every height-one prime ideal, is principal.

Let $X$ be a Noetherian integral scheme that is regular in codimension one, and let $U=X\setminus {\bigcup }_{j=1}^r{Z}_j$, where the $Z_j$ are the codimension-one irreducible components of the complement. Restriction gives an exact sequence

If $A$ is a Noetherian normal domain and $S$ is a multiplicative set, then $\mathrm{Cl}(S^{-1}A)$ is the quotient of $\mathrm{Cl}(A)$ by the classes of height-one primes meeting $S$. Tracking the units of $S^{-1}A$ determines the relations among those prime divisors.

For the three-dimensional affine quadric conethe divisor class group is infinite cyclic. A generator is either ruling plane $V(x,z)$ or $V(x,w)$, and the principal divisor of $x$ gives the relation $[V(x,z)]+[V(x,w)]=0$.

For a Noetherian integral scheme $X$ that is regular in codimension one,where $H=X\times {\mathbb{P}}^{n-1}$ is a hyperplane divisor. Restriction to $X\times {\mathbb{A}}^n$ gives the first summand, while restriction to the generic projective-space fiber detects the coefficient of $H$.

For $n\ge 2$, the normal affine surfacehas $\mathrm{Cl}(X_n)\cong \mathbb{Z}/n\mathbb{Z}$. The class of the prime divisor $D=V(x,z)$ generates: localization at $x$ is a unique factorization domain, so the divisor-class localization sequence leaves only $D$, while $\mathrm{div}(x)=nD$ gives its exact order.