TheoremComplete

Riemann–Roch for Surfaces

The Riemann–Roch theorem for surfaces expresses the Euler characteristic of a line bundle in terms of intersection numbers. Combined with vanishing theorems, it becomes a powerful tool for computing dimensions of linear systems.


Statement

Theorem5.1Riemann–Roch for surfaces

Let XX be a smooth projective surface over an algebraically closed field, and let DD be a divisor on XX. Then:

χ(OX(D))=12D(DKX)+χ(OX),\chi(\mathcal{O}_X(D)) = \frac{1}{2}D \cdot (D - K_X) + \chi(\mathcal{O}_X),

where KXK_X is the canonical divisor and χ(OX)=1q+pg\chi(\mathcal{O}_X) = 1 - q + p_g with q=h1(OX)q = h^1(\mathcal{O}_X) (the irregularity) and pg=h0(KX)p_g = h^0(K_X) (the geometric genus).

Expanding: χ(OX(D))=12(D2DKX)+χ(OX)\chi(\mathcal{O}_X(D)) = \frac{1}{2}(D^2 - D \cdot K_X) + \chi(\mathcal{O}_X).

RemarkRelation to Hirzebruch–Riemann–Roch

This is the Hirzebruch–Riemann–Roch theorem for surfaces: χ(L)=Xch(L)td(TX)\chi(\mathcal{L}) = \int_X \operatorname{ch}(\mathcal{L}) \cdot \operatorname{td}(T_X).

Writing L=O(D)\mathcal{L} = \mathcal{O}(D): ch(L)=1+c1(D)+12c1(D)2\operatorname{ch}(\mathcal{L}) = 1 + c_1(D) + \frac{1}{2}c_1(D)^2 and td(TX)=1+12c1(TX)+112(c1(TX)2+c2(TX))\operatorname{td}(T_X) = 1 + \frac{1}{2}c_1(T_X) + \frac{1}{12}(c_1(T_X)^2 + c_2(T_X)).

Multiplying and integrating over XX: χ(O(D))=12D2+12Dc1(TX)+112(c12+c2).\chi(\mathcal{O}(D)) = \frac{1}{2}D^2 + \frac{1}{2}D \cdot c_1(T_X) + \frac{1}{12}(c_1^2 + c_2).

Since c1(TX)=KXc_1(T_X) = -K_X and χ(OX)=112(KX2+c2(X))\chi(\mathcal{O}_X) = \frac{1}{12}(K_X^2 + c_2(X)) (Noether's formula), this gives χ(O(D))=12D212DKX+χ(OX)\chi(\mathcal{O}(D)) = \frac{1}{2}D^2 - \frac{1}{2}D \cdot K_X + \chi(\mathcal{O}_X) ✓.


Verification: basic cases

ExampleD = 0

χ(OX(0))=120(0KX)+χ(OX)=χ(OX)\chi(\mathcal{O}_X(0)) = \frac{1}{2} \cdot 0 \cdot (0 - K_X) + \chi(\mathcal{O}_X) = \chi(\mathcal{O}_X) ✓. The formula is consistent.

ExampleD = K_X

χ(O(KX))=12KX(KXKX)+χ(OX)=0+χ(OX)=χ(OX)\chi(\mathcal{O}(K_X)) = \frac{1}{2}K_X \cdot (K_X - K_X) + \chi(\mathcal{O}_X) = 0 + \chi(\mathcal{O}_X) = \chi(\mathcal{O}_X).

By Serre duality, χ(ωX)=χ(OX)\chi(\omega_X) = \chi(\mathcal{O}_X) (since hi(ωX)=h2i(OX)h^i(\omega_X) = h^{2-i}(\mathcal{O}_X), the alternating sum is preserved). ✓

ExampleSymmetry: D ↔ K - D

The Serre duality pairing gives hi(O(D))=h2i(O(KD))h^i(\mathcal{O}(D)) = h^{2-i}(\mathcal{O}(K - D)), so χ(O(D))=χ(O(KD))\chi(\mathcal{O}(D)) = \chi(\mathcal{O}(K - D)).

From the formula: 12D(DK)+χ(OX)\frac{1}{2}D(D - K) + \chi(\mathcal{O}_X) vs. 12(KD)(KDK)+χ(OX)=12(KD)(D)+χ(OX)=12KD+12D2+χ(OX)\frac{1}{2}(K-D)(K-D-K) + \chi(\mathcal{O}_X) = \frac{1}{2}(K-D)(-D) + \chi(\mathcal{O}_X) = -\frac{1}{2}KD + \frac{1}{2}D^2 + \chi(\mathcal{O}_X).

Both equal 12(D2KD)+χ(OX)\frac{1}{2}(D^2 - KD) + \chi(\mathcal{O}_X) ✓. The formula is automatically compatible with Serre duality.


Applications to ℙ²

ExampleLine bundles on ℙ²

On P2\mathbb{P}^2: K=3HK = -3H, χ(O)=1\chi(\mathcal{O}) = 1 (since q=pg=0q = p_g = 0). For D=dHD = dH:

χ(O(d))=12d(d+3)+1=d2+3d+22=(d+1)(d+2)2=(d+22).\chi(\mathcal{O}(d)) = \frac{1}{2}d(d + 3) + 1 = \frac{d^2 + 3d + 2}{2} = \frac{(d+1)(d+2)}{2} = \binom{d+2}{2}.

This recovers:

  • d=0d = 0: χ=1=h0(O)=1\chi = 1 = h^0(\mathcal{O}) = 1 ✓.
  • d=1d = 1: χ=3=h0(O(1))=3\chi = 3 = h^0(\mathcal{O}(1)) = 3 (linear forms x,y,zx, y, z) ✓.
  • d=2d = 2: χ=6\chi = 6 (monomials x2,xy,xz,y2,yz,z2x^2, xy, xz, y^2, yz, z^2) ✓.
  • d=1d = -1: χ=0\chi = 0. Indeed h0(O(1))=0h^0(\mathcal{O}(-1)) = 0, h1(O(1))=0h^1(\mathcal{O}(-1)) = 0, h2(O(1))=0h^2(\mathcal{O}(-1)) = 0 ✓.
  • d=2d = -2: χ=0\chi = 0. Same vanishing ✓.
  • d=3d = -3: χ=1\chi = 1. We have O(3)=ωP2\mathcal{O}(-3) = \omega_{\mathbb{P}^2}, so h2(O(3))=h0(O)=1h^2(\mathcal{O}(-3)) = h^0(\mathcal{O}) = 1, h0=h1=0h^0 = h^1 = 0, giving χ=00+1=1\chi = 0 - 0 + 1 = 1 ✓.
  • d=4d = -4: χ=3\chi = 3. h2(O(4))=h0(O(1))=3h^2(\mathcal{O}(-4)) = h^0(\mathcal{O}(1)) = 3 ✓.

Applications to ruled surfaces

ExampleHirzebruch surfaces 𝔽ₙ

The Hirzebruch surface Fn=P(OP1OP1(n))\mathbb{F}_n = \mathbb{P}(\mathcal{O}_{\mathbb{P}^1} \oplus \mathcal{O}_{\mathbb{P}^1}(-n)) has:

  • Pic(Fn)=ZC0ZF\operatorname{Pic}(\mathbb{F}_n) = \mathbb{Z} C_0 \oplus \mathbb{Z} F where C0C_0 is the negative section (C02=nC_0^2 = -n) and FF is a fiber (F2=0F^2 = 0, C0F=1C_0 \cdot F = 1).
  • KFn2C0(n+2)FK_{\mathbb{F}_n} \sim -2C_0 - (n + 2)F.
  • χ(OFn)=1\chi(\mathcal{O}_{\mathbb{F}_n}) = 1 (rational surface, so q=pg=0q = p_g = 0).

For DaC0+bFD \sim aC_0 + bF: D2=na2+2abD^2 = -na^2 + 2ab, DK=2na2b(n+2)a=(n2)a2bD \cdot K = 2na - 2b - (n+2)a = (n-2)a - 2b. So:

χ(O(D))=12(na2+2ab(n2)a+2b)+1.\chi(\mathcal{O}(D)) = \frac{1}{2}(-na^2 + 2ab - (n-2)a + 2b) + 1.

Check: D=C0D = C_0 (a=1,b=0a = 1, b = 0): χ=12(nn+2)+1=12(22n)+1=2n\chi = \frac{1}{2}(-n - n + 2) + 1 = \frac{1}{2}(2 - 2n) + 1 = 2 - n. For n=0n = 0 (P1×P1\mathbb{P}^1 \times \mathbb{P}^1): χ(O(C0))=2\chi(\mathcal{O}(C_0)) = 2, and h0(O(1,0))=2h^0(\mathcal{O}(1,0)) = 2 ✓.


Applications to K3 surfaces

ExampleK3 surfaces

A K3 surface has KX=0K_X = 0, χ(OX)=2\chi(\mathcal{O}_X) = 2 (since h0=1h^0 = 1, h1=0h^1 = 0, h2=1h^2 = 1). Riemann–Roch gives:

χ(O(D))=12D2+2.\chi(\mathcal{O}(D)) = \frac{1}{2}D^2 + 2.

  • If DD is a primitive ample class with D2=2dD^2 = 2d, then χ(O(D))=d+2\chi(\mathcal{O}(D)) = d + 2. By Kodaira vanishing, h1=h2=0h^1 = h^2 = 0 for ample DD, so h0(D)=d+2h^0(D) = d + 2. The linear system D|D| has dimension d+1d + 1.

  • Degree-22 K3 (D2=2D^2 = 2): h0(D)=3h^0(D) = 3, so D|D| is a pencil in P2\mathbb{P}^2... rather, D|D| gives a 2:12:1 map to P2\mathbb{P}^2.

  • Degree-44 K3 (D2=4D^2 = 4): h0(D)=4h^0(D) = 4, embedding as quartic in P3\mathbb{P}^3.

  • Degree-66 K3 (D2=6D^2 = 6): h0(D)=5h^0(D) = 5, complete intersection of a quadric and cubic in P4\mathbb{P}^4.

  • Degree-88 K3 (D2=8D^2 = 8): h0(D)=6h^0(D) = 6, complete intersection of three quadrics in P5\mathbb{P}^5.


Applications to surfaces of general type

ExampleSurfaces of general type

For a minimal surface of general type (KXK_X nef and big, KX2>0K_X^2 > 0):

χ(O(nKX))=n(n1)2KX2+χ(OX).\chi(\mathcal{O}(nK_X)) = \frac{n(n-1)}{2}K_X^2 + \chi(\mathcal{O}_X).

Pluricanonical dimensions:

  • n=1n = 1: χ(ωX)=χ(OX)=1+pgq\chi(\omega_X) = \chi(\mathcal{O}_X) = 1 + p_g - q.
  • n=2n = 2: χ(O(2K))=K2+χ(OX)\chi(\mathcal{O}(2K)) = K^2 + \chi(\mathcal{O}_X). By Kawamata–Viehweg vanishing (2KK=K2K - K = K is nef and big), hi(O(2K))=0h^i(\mathcal{O}(2K)) = 0 for i>0i > 0, so P2=h0(2K)=K2+χ(OX)P_2 = h^0(2K) = K^2 + \chi(\mathcal{O}_X).
  • n=3n = 3: P3=3K2+χ(OX)P_3 = 3K^2 + \chi(\mathcal{O}_X).
  • nn \to \infty: Pn12n2K2P_n \sim \frac{1}{2}n^2 K^2 (quadratic growth — this is the hallmark of a surface of general type).

Bombieri's theorem: 5KX|5K_X| gives a birational map for all surfaces of general type.


Applications to abelian surfaces

ExampleAbelian surfaces

An abelian surface AA has KA=0K_A = 0, χ(OA)=0\chi(\mathcal{O}_A) = 0 (since h0=1h^0 = 1, h1=2h^1 = 2, h2=1h^2 = 1). So:

χ(O(D))=12D2.\chi(\mathcal{O}(D)) = \frac{1}{2}D^2.

For an ample DD with D2=2dD^2 = 2d: χ(O(D))=d\chi(\mathcal{O}(D)) = d. By the Mumford vanishing theorem, hi(D)=0h^i(D) = 0 for i>0i > 0 when DD is ample, so h0(D)=dh^0(D) = d.

  • Principal polarization (D2=2D^2 = 2): h0(D)=1h^0(D) = 1. The theta divisor Θ\Theta has a unique section.
  • (1,2)(1,2)-polarization (D2=4D^2 = 4): h0(D)=2h^0(D) = 2, giving a 2:12:1 map AP1A \to \mathbb{P}^1... actually, D|D| maps to P1\mathbb{P}^1 — but abelian surfaces don't admit maps to P1\mathbb{P}^1 (no rational curves). More precisely, D|D| has base locus and ϕD\phi_D maps AA to P1\mathbb{P}^1 as a genus-22 fibration after resolving.
  • (1,3)(1,3)-polarization (D2=6D^2 = 6): h0(D)=3h^0(D) = 3. Maps AP2A \to \mathbb{P}^2 as 3:13:1.

The inequality h0(D)χ(D)h^0(D) \geq \chi(D) and effective bounds

TheoremLower bound from Riemann–Roch

For any nef divisor DD on a surface:

h0(D)χ(O(D))=12D(DK)+χ(OX),h^0(D) \geq \chi(\mathcal{O}(D)) = \frac{1}{2}D \cdot (D - K) + \chi(\mathcal{O}_X),

since h1(O(D))h^1(\mathcal{O}(D)) appears with a minus sign in χ\chi and h2(O(D))=h0(KD)0h^2(\mathcal{O}(D)) = h^0(K - D) \geq 0.

When DD is ample (or nef and big) and satisfies certain positivity conditions, Kodaira vanishing (Hi(X,O(D+K))=0H^i(X, \mathcal{O}(D + K)) = 0 for i>0i > 0 and DD ample) or Kawamata–Viehweg vanishing gives:

h0(D)=χ(O(D))(exact computation, no correction terms).h^0(D) = \chi(\mathcal{O}(D)) \quad \text{(exact computation, no correction terms)}.

ExampleWhen Riemann–Roch gives exact answers

Complete list of common vanishing situations:

  • DD ample on P2\mathbb{P}^2: hi(O(d))=0h^i(\mathcal{O}(d)) = 0 for 0<i<20 < i < 2 and d2d \geq -2 (from the explicit computation of cohomology of Pn\mathbb{P}^n).
  • DD ample on K3: h1=h2=0h^1 = h^2 = 0 by Kodaira vanishing (since D=(DK)+K=D+0D = (D - K) + K = D + 0, and DD ample implies vanishing). So h0(D)=12D2+2h^0(D) = \frac{1}{2}D^2 + 2.
  • DD ample on an Enriques surface (K2=0K^2 = 0, χ(O)=1\chi(\mathcal{O}) = 1, 2K02K \sim 0): h0(D)=12D2+1h^0(D) = \frac{1}{2}D^2 + 1.
  • nKXnK_X with n2n \geq 2 on a surface of general type: h0(nK)=n(n1)2K2+χ(OX)h^0(nK) = \frac{n(n-1)}{2}K^2 + \chi(\mathcal{O}_X).

Riemann–Roch and curve counting

ExampleCounting curves on surfaces

Riemann–Roch determines the expected dimension of families of curves:

  • Curves of degree dd in P2\mathbb{P}^2: The linear system dH|dH| has dimension (d+22)1\binom{d+2}{2} - 1. A curve passes through nn general points iff (d+22)1n\binom{d+2}{2} - 1 \geq n. Through (d+22)1\binom{d+2}{2} - 1 general points, there is a unique degree-dd curve.

  • Lines through 2 points: (32)1=2\binom{3}{2} - 1 = 2; imposing 22 points gives 00-dimensional family, i.e., a unique line. ✓

  • Conics through 5 points: (42)1=5\binom{4}{2} - 1 = 5; imposing 55 conditions gives a unique conic. ✓

  • Cubics through 9 points: (52)1=9\binom{5}{2} - 1 = 9; but 99 points may not impose independent conditions. Cayley–Bacharach: if 88 of the 99 base points of two cubics are specified, the 99th is determined.


Summary

RemarkThe power of Riemann–Roch for surfaces

Riemann–Roch for surfaces:

χ(O(D))=12D(DK)+χ(OX)\chi(\mathcal{O}(D)) = \frac{1}{2}D \cdot (D - K) + \chi(\mathcal{O}_X)

is the starting point for all dimension computations on surfaces. Combined with:

  • Serre duality (h2(O(D))=h0(KD)h^2(\mathcal{O}(D)) = h^0(K - D)): reduces to bounding h1h^1.
  • Kodaira/Kawamata–Viehweg vanishing: often kills h1h^1 and h2h^2, giving exact formulas.
  • Noether's formula (χ(OX)=112(K2+e(X))\chi(\mathcal{O}_X) = \frac{1}{12}(K^2 + e(X))): constrains the invariants of XX itself.

It is a special case of the Hirzebruch–Riemann–Roch theorem and generalizes the classical Riemann–Roch for curves.