ConceptComplete

Divisors on Curves

Divisors are the fundamental tool for studying functions and line bundles on algebraic curves. They encode the zeros and poles of meromorphic functions and provide the language for the Riemann–Roch theorem.


Weil divisors

Definition4.1Divisor on a curve

Let CC be a smooth projective curve over a field kk. A (Weil) divisor on CC is a formal Z\mathbb{Z}-linear combination of closed points:

D=βˆ‘P∈CnPβ‹…P,nP∈Z,Β almostΒ allΒ nP=0.D = \sum_{P \in C} n_P \cdot P, \quad n_P \in \mathbb{Z}, \text{ almost all } n_P = 0.

The divisor group Div⁑(C)\operatorname{Div}(C) is the free abelian group on the closed points of CC.

The degree of DD is deg⁑D=βˆ‘PnPβ‹…[ΞΊ(P):k]\deg D = \sum_P n_P \cdot [\kappa(P) : k], where ΞΊ(P)\kappa(P) is the residue field. Over k=kΛ‰k = \bar{k}, this is simply βˆ‘nP\sum n_P.

A divisor DD is effective (written Dβ‰₯0D \geq 0) if nPβ‰₯0n_P \geq 0 for all PP. The support of DD is Supp⁑(D)={P∣nPβ‰ 0}\operatorname{Supp}(D) = \{P \mid n_P \neq 0\}.

ExampleDivisors on β„™ΒΉ

On Pk1\mathbb{P}^1_k with k=kΛ‰k = \bar{k}, closed points correspond to a∈ka \in k (the point [a:1][a:1]) plus ∞=[1:0]\infty = [1:0]. Typical divisors:

  • D=3β‹…[0]βˆ’2β‹…[∞]D = 3 \cdot [0] - 2 \cdot [\infty]: degree 11.
  • D=[0]+[1]+[∞]D = [0] + [1] + [\infty]: degree 33, effective.
  • D=βˆ’[0]D = -[0]: degree βˆ’1-1, not effective.

Over R\mathbb{R}: the point [i:1]∈PC1[i:1] \in \mathbb{P}^1_\mathbb{C} is not a point of PR1\mathbb{P}^1_\mathbb{R}, but the closed point corresponding to x2+1∈R[x]x^2 + 1 \in \mathbb{R}[x] has degree [C:R]=2[\mathbb{C}:\mathbb{R}] = 2.


Principal divisors

Definition4.2Valuation and principal divisor

For each closed point P∈CP \in C, the valuation vP:k(C)βˆ—β†’Zv_P : k(C)^* \to \mathbb{Z} measures the order of vanishing (or pole) at PP:

  • vP(f)>0v_P(f) > 0: ff has a zero of order vP(f)v_P(f) at PP.
  • vP(f)<0v_P(f) < 0: ff has a pole of order ∣vP(f)∣|v_P(f)| at PP.
  • vP(f)=0v_P(f) = 0: ff is regular and nonvanishing at PP.

The principal divisor of f∈k(C)βˆ—f \in k(C)^* is

div⁑(f)=βˆ‘PvP(f)β‹…P.\operatorname{div}(f) = \sum_P v_P(f) \cdot P.

This is a finite sum since ff has only finitely many zeros and poles.

ExamplePrincipal divisors on β„™ΒΉ

On P1\mathbb{P}^1 with coordinate t=x/yt = x/y:

  • div⁑(t)=[0]βˆ’[∞]\operatorname{div}(t) = [0] - [\infty]: a simple zero at 00, simple pole at ∞\infty.
  • div⁑(t2)=2[0]βˆ’2[∞]\operatorname{div}(t^2) = 2[0] - 2[\infty]: double zero at 00, double pole at ∞\infty.
  • div⁑(tβˆ’1)=[1]βˆ’[∞]\operatorname{div}(t - 1) = [1] - [\infty].
  • div⁑(ttβˆ’1)=[0]βˆ’[1]\operatorname{div}\left(\frac{t}{t-1}\right) = [0] - [1]: zero at 00, pole at 11, regular at ∞\infty.
  • div⁑(t2βˆ’1)=[1]+[βˆ’1]βˆ’2[∞]\operatorname{div}(t^2 - 1) = [1] + [-1] - 2[\infty].

Notice: deg⁑(div⁑(f))=0\deg(\operatorname{div}(f)) = 0 always. This is a theorem!

TheoremDegree of a principal divisor

For any f∈k(C)βˆ—f \in k(C)^* on a smooth projective curve CC:

deg⁑(div⁑(f))=0.\deg(\operatorname{div}(f)) = 0.

Equivalently, ff has the same number of zeros as poles (counted with multiplicity and degree).

ExamplePrincipal divisors on an elliptic curve

Let E:y2=x3βˆ’xE: y^2 = x^3 - x over C\mathbb{C} with identity O=[0:1:0]O = [0:1:0].

  • div⁑(x)=(0,0)+P1+Pβˆ’1βˆ’3O\operatorname{div}(x) = (0,0) + P_1 + P_{-1} - 3O where PΒ±1=(Β±1,0)P_{\pm 1} = (\pm 1, 0)... but wait, we need deg⁑=0\deg = 0. Actually div⁑(x)\operatorname{div}(x) is the divisor of xx viewed as a function Eβ†’P1E \to \mathbb{P}^1. The function xx has zeros at (0,0)(0,0) (where x=0x = 0, and y2=0y^2 = 0 so y=0y = 0 too β€” but we must account for the intersection multiplicity). More carefully:

div⁑(x)=2β‹…(0,0)βˆ’2O\operatorname{div}(x) = 2 \cdot (0,0) - 2O

since xx has a double zero at (0,0)(0,0) (the tangent to EE at (0,0)(0,0) is vertical) and a double pole at OO.

  • div⁑(y)=(0,0)+(1,0)+(βˆ’1,0)βˆ’3O\operatorname{div}(y) = (0,0) + (1,0) + (-1,0) - 3O: degree 00. βœ“

Linear equivalence and the Picard group

Definition4.3Linear equivalence

Two divisors DD and Dβ€²D' are linearly equivalent (written D∼Dβ€²D \sim D') if Dβˆ’Dβ€²=div⁑(f)D - D' = \operatorname{div}(f) for some f∈k(C)βˆ—f \in k(C)^*.

The divisor class group (or Picard group) is

Pic⁑(C)=Div⁑(C)/{principal divisors}.\operatorname{Pic}(C) = \operatorname{Div}(C) / \{\text{principal divisors}\}.

Since principal divisors have degree 00, the degree map factors through Pic⁑\operatorname{Pic}:

0β†’Pic⁑0(C)β†’Pic⁑(C)β†’deg⁑Zβ†’00 \to \operatorname{Pic}^0(C) \to \operatorname{Pic}(C) \xrightarrow{\deg} \mathbb{Z} \to 0

where Pic⁑0(C)\operatorname{Pic}^0(C) is the group of degree-00 divisor classes.

ExamplePic(β„™ΒΉ)

On P1\mathbb{P}^1: any two points are linearly equivalent ([a]βˆ’[b]=div⁑(tβˆ’atβˆ’b)[a] - [b] = \operatorname{div}\left(\frac{t - a}{t - b}\right)). So every degree-dd divisor is equivalent to dβ‹…[∞]d \cdot [\infty], and:

Pic⁑(P1)β‰…Z,Pic⁑0(P1)=0.\operatorname{Pic}(\mathbb{P}^1) \cong \mathbb{Z}, \quad \operatorname{Pic}^0(\mathbb{P}^1) = 0.

ExamplePic⁰(E) for an elliptic curve

For an elliptic curve EE with origin OO, the Abel–Jacobi map gives:

E(k)β†’βˆΌPic⁑0(E),P↦[Pβˆ’O].E(k) \xrightarrow{\sim} \operatorname{Pic}^0(E), \quad P \mapsto [P - O].

This is an isomorphism of groups! The group law on EE (chord-tangent construction) corresponds to addition of divisor classes. So Pic⁑0(E)β‰…E\operatorname{Pic}^0(E) \cong E as a group variety.

Over C\mathbb{C}: Pic⁑0(E)β‰…C/Ξ›\operatorname{Pic}^0(E) \cong \mathbb{C}/\Lambda (a complex torus), and Pic⁑(E)β‰…ZβŠ•E(C)\operatorname{Pic}(E) \cong \mathbb{Z} \oplus E(\mathbb{C}).

ExamplePic⁰ for higher genus

For a smooth curve CC of genus gg:

Pic⁑0(C)β‰…Jac⁑(C)\operatorname{Pic}^0(C) \cong \operatorname{Jac}(C)

the Jacobian variety, an abelian variety of dimension gg. Over C\mathbb{C}, Jac⁑(C)β‰…Cg/Ξ›\operatorname{Jac}(C) \cong \mathbb{C}^g / \Lambda where Ξ›\Lambda is a lattice determined by the period matrix.

  • g=0g = 0: Jac⁑=0\operatorname{Jac} = 0 (a point).
  • g=1g = 1: Jac⁑(E)β‰…E\operatorname{Jac}(E) \cong E (the curve is its own Jacobian).
  • g=2g = 2: Jac⁑(C)\operatorname{Jac}(C) is a 22-dimensional abelian surface. The curve CC embeds in Jac⁑(C)\operatorname{Jac}(C) via P↦[Pβˆ’P0]P \mapsto [P - P_0].
  • g=3g = 3: Jac⁑(C)\operatorname{Jac}(C) is a 33-fold, and the theta divisor Ξ˜βŠ†Jac⁑(C)\Theta \subseteq \operatorname{Jac}(C) is a surface isomorphic to Sym⁑2(C)\operatorname{Sym}^2(C) (for non-hyperelliptic CC).

Divisors and line bundles

Definition4.4Line bundle associated to a divisor

To a divisor DD on CC, we associate the invertible sheaf (line bundle):

OC(D)(U)={f∈k(C)βˆ—βˆ£div⁑(f)∣U+D∣Uβ‰₯0}βˆͺ{0}.\mathcal{O}_C(D)(U) = \{f \in k(C)^* \mid \operatorname{div}(f)|_U + D|_U \geq 0\} \cup \{0\}.

This gives an isomorphism of groups:

Div⁑(C)/βˆΌβ€…β€Šβ†’βˆΌβ€…β€ŠPic⁑(C)β‰…H1(C,OCβˆ—)\operatorname{Div}(C)/\sim \;\xrightarrow{\sim}\; \operatorname{Pic}(C) \cong H^1(C, \mathcal{O}_C^*)

sending D↦O(D)D \mapsto \mathcal{O}(D). Key properties:

  • O(D+Dβ€²)β‰…O(D)βŠ—O(Dβ€²)\mathcal{O}(D + D') \cong \mathcal{O}(D) \otimes \mathcal{O}(D').
  • O(βˆ’D)β‰…O(D)βˆ’1=O(D)∨\mathcal{O}(-D) \cong \mathcal{O}(D)^{-1} = \mathcal{O}(D)^\vee.
  • O(div⁑(f))β‰…OC\mathcal{O}(\operatorname{div}(f)) \cong \mathcal{O}_C (principal divisors give trivial line bundles).
  • deg⁑O(D)=deg⁑D\deg \mathcal{O}(D) = \deg D.
ExampleLine bundles and divisors on β„™ΒΉ

On P1\mathbb{P}^1: O(dβ‹…[∞])=OP1(d)\mathcal{O}(d \cdot [\infty]) = \mathcal{O}_{\mathbb{P}^1}(d), the usual twisting sheaf. So Pic⁑(P1)={O(d)∣d∈Z}β‰…Z\operatorname{Pic}(\mathbb{P}^1) = \{\mathcal{O}(d) \mid d \in \mathbb{Z}\} \cong \mathbb{Z}, matching the divisor picture.

On Pn\mathbb{P}^n: a hypersurface H=V(F)H = V(F) of degree dd gives a divisor with O(H)=O(d)\mathcal{O}(H) = \mathcal{O}(d). Different degree-dd hypersurfaces give linearly equivalent divisors.


The canonical divisor

Definition4.5Canonical divisor

The canonical divisor KCK_C (or KK) is any divisor with O(KC)β‰…Ο‰C=Ξ©C/k1\mathcal{O}(K_C) \cong \omega_C = \Omega^1_{C/k} (the sheaf of differentials). It is unique up to linear equivalence.

Key facts:

  • deg⁑KC=2gβˆ’2\deg K_C = 2g - 2 where gg is the genus.
  • h0(KC)=gh^0(K_C) = g (the space of regular differentials has dimension gg).
ExampleCanonical divisors
  • P1\mathbb{P}^1 (g=0g = 0): K=βˆ’2[∞]K = -2[\infty], deg⁑K=βˆ’2\deg K = -2. The differential dtdt has a double pole at ∞\infty: dt=d(1/s)=βˆ’sβˆ’2dsdt = d(1/s) = -s^{-2}ds. So div⁑(dt)=βˆ’2[∞]\operatorname{div}(dt) = -2[\infty].

  • Elliptic curve (g=1g = 1): K∼0K \sim 0 (trivial). The regular differential dx/ydx/y has no zeros or poles. So div⁑(dx/y)=0\operatorname{div}(dx/y) = 0.

  • Genus-2 curve y2=f(x)y^2 = f(x), deg⁑f=5\deg f = 5 or 66: the differential dx/ydx/y has div⁑(dx/y)=βˆ‘iWiβˆ’polarΒ terms\operatorname{div}(dx/y) = \sum_i W_i - \text{polar terms} where WiW_i are the Weierstrass points. deg⁑K=2\deg K = 2.

  • Smooth plane curve of degree dd: KC∼(dβˆ’3)H∣CK_C \sim (d - 3)H|_C by adjunction. deg⁑K=d(dβˆ’3)\deg K = d(d-3), g=(dβˆ’1)(dβˆ’2)2g = \frac{(d-1)(d-2)}{2}.


Effective divisors and linear systems

RemarkFrom divisors to geometry

The Riemann–Roch space L(D)=H0(C,O(D))L(D) = H^0(C, \mathcal{O}(D)) is the set of rational functions with poles bounded by DD:

L(D)={f∈k(C)βˆ—βˆ£div⁑(f)+Dβ‰₯0}βˆͺ{0}.L(D) = \{f \in k(C)^* \mid \operatorname{div}(f) + D \geq 0\} \cup \{0\}.

With β„“(D)=dim⁑L(D)\ell(D) = \dim L(D):

  • β„“(D)=0\ell(D) = 0 if deg⁑D<0\deg D < 0.
  • β„“(0)=1\ell(0) = 1 (only constants).
  • β„“(D)≀deg⁑D+1\ell(D) \leq \deg D + 1 (trivial bound, sharp for g=0g = 0).
  • Riemann–Roch: β„“(D)βˆ’β„“(Kβˆ’D)=deg⁑Dβˆ’g+1\ell(D) - \ell(K - D) = \deg D - g + 1.

Summary: the divisor–line bundle dictionary

RemarkDictionary
DivisorsLine bundles
D∈Div⁑(C)D \in \operatorname{Div}(C)O(D)∈Pic⁑(C)\mathcal{O}(D) \in \operatorname{Pic}(C)
D+Dβ€²D + D'O(D)βŠ—O(Dβ€²)\mathcal{O}(D) \otimes \mathcal{O}(D')
βˆ’D-DO(D)∨\mathcal{O}(D)^\vee
D∼Dβ€²D \sim D'O(D)β‰…O(Dβ€²)\mathcal{O}(D) \cong \mathcal{O}(D')
div⁑(f)\operatorname{div}(f) (principal)OC\mathcal{O}_C (trivial)
deg⁑D\deg Ddeg⁑O(D)=c1(O(D))\deg \mathcal{O}(D) = c_1(\mathcal{O}(D))
L(D)=H0(O(D))L(D) = H^0(\mathcal{O}(D))Global sections
KCK_C (canonical)Ο‰C=Ξ©C/k1\omega_C = \Omega^1_{C/k}
Pic⁑0(C)\operatorname{Pic}^0(C)Jac⁑(C)\operatorname{Jac}(C) (Jacobian)