Divisor (Algebraic Geometry)¶

An algebraic-geometric divisor records codimension-one zero-and-pole data as a Weil cycle or compatible local Cartier equations, with class and line-bundle relations set by the space.

Version: v1 · 2026-10-03 · History
Domain-specific #: 13162
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomain: Algebraic Geometry → Mathematics
Aliases: Algebraic-geometric divisor

Core Idea¶

An algebraic-geometric divisor records codimension-one data on a geometric space, often the places where a rational function vanishes or has a pole. There are two related but not identical constructions. A Weil divisor is an integer-weighted formal sum of integral codimension-one subspaces. A Cartier divisor is described by compatible local rational equations, up to multiplication by invertible regular functions; an effective Cartier divisor is locally cut out by a single nonzerodivisor. Both can encode hypersurface-like information, but the passage between them depends on local properties of the ambient space.^[1]^[2]^[3]^[4]

For a locally Noetherian integral scheme, a rational function f gives a principal Weil divisor div(f)=Σ_Z ord_Z(f)[Z]: positive coefficients mark zeros and negative ones mark poles along prime divisors. Modding out all such principal divisors gives the Weil class group Cl(X). Cartier data instead connect to invertible sheaves and the Picard group. On locally factorial spaces—including regular locally Noetherian varieties—the class comparison is an isomorphism; on a merely normal singular space, one must not assume that every Weil class is Cartier.^[1]^[4]^[5]^[6]

Structural Signature¶

Sig role-phrases:

Ambient integral space: the variety or scheme supplies codimension, local rings and a field of rational functions.
Codimension-one support: prime divisors are integral closed subschemes of codimension one, such as points on a curve or irreducible curves on a surface.
Valuation or local equation: orders of vanishing assign Weil coefficients; locally principal equations define Cartier data.
Equivalence and class: adding div(f) changes a representative but not its Weil class; Cartier divisors connect to invertible sheaves.
Regularity boundary: local factoriality decides whether all Weil classes come from Cartier/Pic data.^[1]^[3]^[4]

Condensed: space + codimension-one locus + integer vanishing order/local equation → divisor representative → class after principal changes. Each role is mathematical; the word “divisor” here does not mean a positive integer factor of another integer.

What It Is Not¶

Weil, Cartier, effective Cartier and divisorial sheaf are not interchangeable aliases merely because their Wikipedia titles redirect to one article. Weil is a codimension-one cycle; effective Cartier requires a locally invertible ideal, equivalently a local nonzerodivisor equation; a rank-one reflexive divisorial sheaf can correspond to a Weil class on a normal space without being invertible at singularities. A smooth setting can hide these differences, but a broad entry must preserve them.^[1]^[3]^[4]^[7]

Nor is every subset of a surface a divisor. A closed point on an algebraic surface ordinarily has codimension two, not the required codimension one for a Weil prime. An arbitrary hypersurface written without multiplicity also lacks the full zero/pole ledger. On a non-quasi-compact locally Noetherian scheme, Stacks allows locally finite Weil sums; saying “finite sum” without the quasi-compact qualification would be false at that generality.^[1]

Scope of Application¶

On a smooth algebraic curve, closed points are prime divisors. A rational function's zeros and poles become a signed point sum; principal divisors expose relations among such sums. On projective space, line-bundle classes carry degree information: the Stacks Project gives Pic(P^n_k)≅ℤ, with m represented by O(m). The projective-line coordinate example below makes those definitions calculable rather than treating “zeros minus poles” as a slogan.^[1]^[8]

In commutative algebra, X=Spec(k[x,y]) has codimension-one primes such as (x) and (y). Since k[x,y] is a Noetherian UFD, its height-one primes are principal. The rational function x/y yields a signed difference of the coordinate-line divisors, with trivial class. This is a second, affine setting for the same mechanism. The examples are author-worked consequences of Stacks definitions and the UFD result, not quotations of a single precomputed worked example.^[1]^[9]^[10]

Clarity¶

On P^1_k, let t be the usual affine coordinate. At the point 0, t has a simple zero, so ord_0(t)=+1. At infinity the local coordinate is s=1/t; hence t=1/s has a simple pole and ord_∞(t)=−1. Therefore div(t)=[0]−[∞]. Because this is principal, [0] and [∞] represent the same divisor class. Their equality is class equality, not equality of points or of divisor representatives. The projective-space Picard calculation locates this in the integer degree class system.^[1]^[8]

On Spec(k[x,y]), the axes V(x) and V(y) are distinct height-one loci. The rational function x/y vanishes once along V(x) and has a simple pole along V(y), giving div(x/y)=[V(x)]−[V(y)]. These loci are not physically the same line, but their difference is principal and hence zero in Cl(X). If one forgot the negative coefficient at V(y), one would incorrectly report a zero class for a one-sided zero ledger without the compensating pole.^[1]^[9]^[10]

Manages Complexity¶

Divisors turn distributed local zero/pole behavior into an algebraic object that can be added, compared and quotiented. The Weil sum records where codimension-one multiplicity occurs; principal divisors identify changes caused by multiplying rational functions; Cartier data supply local equations that can be packaged as line bundles. On P^1, this lets two different points represent the same degree-one class even though their representatives differ. On the affine plane, a rational quotient explicitly mediates the equivalence between two lines.^[1]^[6]^[8]

The distinction also prevents overcompression. The class group and Picard group agree under local factoriality, not on every scheme. A rank-one reflexive sheaf on a normal scheme may extend codimension-one information across a singular set yet fail to be a line bundle there. Treating the useful regular-case isomorphism as unconditional would discard exactly the singular obstruction that divisor theory can detect.^[4]^[7]

Abstract Reasoning¶

To analyze a proposed divisor, first specify the ambient integral space and whether the question is Weil-cycle or Cartier-equation based. For Weil data, list codimension-one prime supports and compute integer orders. For a rational function, positive and negative orders together form div(f). To compare two Weil divisors, ask whether their difference is principal; that is a class-group question. For Cartier data, check whether local equations differ by units on overlaps and whether effective equations are nonzerodivisors.^[1]^[3]

Then inspect the ambient local rings before replacing one language with the other. The Stacks Project proves that local UFDs give Pic(X)≅Cl(X) for locally Noetherian integral X, and regular local rings are UFDs. Normality alone gives an injection of Pic into Cl, not general surjectivity. Thus a smooth curve is a safe place to identify the familiar point-divisor and line-bundle languages, while a singular non-factorial space demands separate tests.^[4]^[5]

Knowledge Transfer¶

The transfer from P^1 to Spec(k[x,y]) is the valuation ledger plus principal-equivalence test. Both identify codimension-one supports and assign +1 to a simple zero and −1 to a simple pole. But the geometry changes: points are codimension one on a curve; coordinate lines are codimension one on the affine plane. Copying a curve's statement “every closed point is a prime divisor” to a surface would fail the codimension test.^[1]^[9]

The Cartier/Weil comparison transfers only with a hypothesis. Regularity gives local UFD behavior, so one can pass to invertible-sheaf classes. In singular settings, codimension-one cycle information may still exist while local principality fails. A later user of divisor classes for projective maps, intersections or canonical classes must check the theorem's precise hypotheses rather than infer every such construction from the bare word “divisor.”^[4]^[6]

Examples¶

A coordinate on the projective line¶

Let X=P^1_k and take t on its affine chart. It has a simple zero at 0 and, because 1/t is a local parameter at infinity, a simple pole at ∞. The computed Weil divisor is [0]−[∞]. This is principal, so the two degree-one point divisors are linearly equivalent. Stacks separately proves Pic(P^1_k)=ℤ; the class distinction retains degree while forgetting which particular rational coordinate representative was used. The calculation is derived explicitly from the cited valuation definition.^[1]^[8]

Mapped back: P^1_k is the ambient integral space; its points 0,∞ are codimension-one support; local parameters give valuation/local equation orders +1,−1; principal div(t) proves equivalence and class of the points; smooth regularity permits the Weil/Cartier and Picard comparison. If infinity were omitted, the computed divisor would not be the divisor of t on the projective curve.

Two coordinate lines in the affine plane¶

Let X=Spec(k[x,y]), and use the rational function x/y. The height-one primes (x) and (y) define the coordinate lines; valuations give a simple zero along V(x) and simple pole along V(y). Thus div(x/y)=[V(x)]−[V(y)], which is zero as a class. k[x,y] is a UFD, so height-one prime ideals are principal and the local Cartier comparison is available. This affine calculation uses the same mechanism without pretending that surface points are the prime supports.^[1]^[9]^[10]

Mapped back: the affine plane is the ambient integral space; its axes, not closed points, are codimension-one support; x and y give local equations/vanishing orders; their quotient produces a principal equivalence; UFD regularity boundary explains why these Weil supports admit Cartier representatives. In a non-factorial singular ambient ring, that last step would require separate proof.

Structural Tensions¶

Effective locus versus subtraction. Restricting to nonnegative coefficients makes a divisor an effective geometric condition: its support can be read as zeros or a closed hypersurface. But effective divisors form only a monoid; the group operation needed for a rational function's signed zero–pole ledger forces negative coefficients and therefore objects that are not effective closed subschemes. For div(t)=[0]−[∞] on P^1, retaining only the zero at 0 would falsely erase the pole that makes the principal relation work. The tradeoff is representational, not a claim that negative divisors are bad: use the effective cone when an actual zero locus matters, and the full divisor group when cancellation or linear equivalence matters. Diagnostic: would forbidding negative multiplicities preserve the relation being asserted, or discard an essential pole/subtraction term?^[1]^[3]^[8]

Concrete representative versus invariant class. A rational function gives a precise signed zero/pole divisor, useful for calculation; quotienting by principal divisors forgets that coordinate-dependent representative to retain an equivalence class. 0 and ∞ on P^1 remain distinct places even though their difference is principal. Diagnostic: does a proposed distinction survive adding div(f), or is it only a choice of rational function representative?^[1]^[8]

Structural–Framed Character¶

The entry is strongly structural: codimension, valuations, local equations and group quotients are formally defined, with little evaluative weight. Human mathematical practice chooses the ambient category and hypotheses—integral, locally Noetherian, regular or singular—which determine which comparisons are legal. Its institutional origin is algebraic geometry's developed language of cycles and line bundles, not a social ranking of “important” hypersurfaces. The vocabulary travels from curves to affine schemes because codimension-one support and order of vanishing can be re-identified; importing “points are divisors” from a curve to a surface without codimension checking is false transfer. Its character: a formal geometric-algebraic structure with hypothesis-sensitive variants.

Structural Core vs. Domain Accent¶

The skeletal relation is codimension-one support/local defining data → integer multiplicity → quotient by principal change, with an optional Cartier-to-line-bundle passage when hypotheses permit. Projective points and affine axes are accents. The domain mechanism depends on schemes, rational functions, local rings and sheaves; “record weighted boundaries” would lose the valuation and class-group claims. Thus the named entry fails the prime bar. A portable weighted-boundary prime would need independently admitted unlike systems, not an unproved analogy to arithmetic divisibility or general bookkeeping. No strict parent edge is asserted.

No strict parent is assigned; a broader codimension-one geometric-data genus remains to be established. A live Algebraic Cycle relation may cover Weil divisors, but does not automatically subsume Cartier local-equation or invertible-sheaf data across regular, locally factorial and merely normal singular spaces. Five source titles redirect to the same Wikipedia article but denote distinct variants, not synonyms. Arithmetic Divisor Function and Zero Divisor are homonyms, not parents.

Neighborhood in Abstraction Space¶

Divisor (Algebraic Geometry) sits in a sparse region of the domain-specific corpus (78^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Sheaves & Birational Geometry (22 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

An effective Cartier divisor is more than an arbitrary locally principal closed subscheme on a general scheme: the local equation must be a nonzerodivisor. A Weil sum on a non-quasi-compact locally Noetherian scheme need only be locally finite, not globally finite. Cl(X) is the Weil-divisor quotient by principal divisors; Pic(X) classifies invertible sheaves, and their equality needs local factoriality rather than mere normality. A divisorial reflexive sheaf need not be invertible at a singular point. None of these objects is an integer factor in elementary arithmetic.^[3]^[1]^[4]^[7]