ConceptComplete

Canonical Class KXK_X

The canonical class is the single most important invariant of a smooth projective surface. It governs the classification of surfaces via Kodaira dimension, determines the birational geometry through pluricanonical maps, and constrains the numerical invariants through deep inequalities like Noether's formula and the Bogomolov--Miyaoka--Yau inequality.


Definition of the canonical class

Definition5.10Canonical class

Let XX be a smooth projective surface over an algebraically closed field kk. The canonical sheaf (or dualizing sheaf) is Ο‰X=det⁑ΩX/k1=∧2Ξ©X/k1\omega_X = \det \Omega^1_{X/k} = \wedge^2 \Omega^1_{X/k}, the determinant of the cotangent bundle.

The canonical class is the corresponding divisor class:

KX=c1(Ο‰X)=c1(det⁑ΩX/k1)∈Pic⁑(X).K_X = c_1(\omega_X) = c_1(\det \Omega^1_{X/k}) \in \operatorname{Pic}(X).

Any divisor in this class is called a canonical divisor. The canonical class is defined up to linear equivalence: if Ο‰\omega is a nonzero rational 22-form on XX, then KX=div⁑(Ο‰)K_X = \operatorname{div}(\omega).

Definition5.11Plurigenera and Kodaira dimension

For a smooth projective surface XX, the nn-th plurigenus is

Pn(X)=h0(X,Ο‰XβŠ—n)=h0(X,OX(nKX)),nβ‰₯1.P_n(X) = h^0(X, \omega_X^{\otimes n}) = h^0(X, \mathcal{O}_X(nK_X)), \quad n \geq 1.

In particular, P1=pg=h0(KX)P_1 = p_g = h^0(K_X) is the geometric genus.

The Kodaira dimension ΞΊ(X)\kappa(X) is defined as:

  • ΞΊ(X)=βˆ’βˆž\kappa(X) = -\infty if Pn=0P_n = 0 for all nβ‰₯1n \geq 1,
  • ΞΊ(X)=0\kappa(X) = 0 if the PnP_n are bounded but not all zero,
  • ΞΊ(X)=1\kappa(X) = 1 if PnP_n grows linearly: Pn∼cnP_n \sim cn for some c>0c > 0,
  • ΞΊ(X)=2\kappa(X) = 2 if PnP_n grows quadratically: Pn∼cn2P_n \sim cn^2 for some c>0c > 0.

Equivalently, ΞΊ(X)=lim sup⁑nβ†’βˆžlog⁑Pnlog⁑n\kappa(X) = \limsup_{n \to \infty} \frac{\log P_n}{\log n} when Pnβ‰ 0P_n \neq 0 for some nn.

RemarkBirational invariance of plurigenera

The plurigenera PnP_n are birational invariants of smooth projective surfaces. If f:Xβ‡’Yf: X \dashrightarrow Y is a birational map between smooth projective surfaces, then Pn(X)=Pn(Y)P_n(X) = P_n(Y) for all nβ‰₯1n \geq 1. In particular, the Kodaira dimension ΞΊ(X)\kappa(X) is a birational invariant. This is because rational maps induce isomorphisms on pluricanonical forms away from codimension-22 loci, and sections extend by Hartogs' theorem.


Canonical class of basic surfaces

ExampleCanonical class of P^2

On P2\mathbb{P}^2 with hyperplane class HH, the Euler sequence 0β†’Ξ©P21β†’O(βˆ’1)3β†’Oβ†’00 \to \Omega^1_{\mathbb{P}^2} \to \mathcal{O}(-1)^3 \to \mathcal{O} \to 0 gives:

KP2=βˆ’3H.K_{\mathbb{P}^2} = -3H.

Since KP2K_{\mathbb{P}^2} is anti-ample, Pn=h0(nK)=h0(O(βˆ’3n))=0P_n = h^0(nK) = h^0(\mathcal{O}(-3n)) = 0 for all nβ‰₯1n \geq 1. Thus ΞΊ(P2)=βˆ’βˆž\kappa(\mathbb{P}^2) = -\infty.

The self-intersection is KP22=(βˆ’3H)2=9K_{\mathbb{P}^2}^2 = (-3H)^2 = 9.

ExampleCanonical class of P^1 x P^1

On X=P1Γ—P1X = \mathbb{P}^1 \times \mathbb{P}^1 with fiber classes F1,F2F_1, F_2:

KX=βˆ’2F1βˆ’2F2.K_X = -2F_1 - 2F_2.

This follows from the product formula Ο‰X=Ο‰P1βŠ Ο‰P1=O(βˆ’2)⊠O(βˆ’2)\omega_X = \omega_{\mathbb{P}^1} \boxtimes \omega_{\mathbb{P}^1} = \mathcal{O}(-2) \boxtimes \mathcal{O}(-2).

Again KXK_X is anti-ample, so ΞΊ=βˆ’βˆž\kappa = -\infty. The self-intersection is KX2=(βˆ’2F1βˆ’2F2)2=2β‹…4=8K_X^2 = (-2F_1 - 2F_2)^2 = 2 \cdot 4 = 8.

ExampleCanonical class of Hirzebruch surfaces

On the Hirzebruch surface Fn=P(OβŠ•O(βˆ’n))\mathbb{F}_n = \mathbb{P}(\mathcal{O} \oplus \mathcal{O}(-n)) over P1\mathbb{P}^1, with negative section C0C_0 (C02=βˆ’nC_0^2 = -n) and fiber FF:

KFn=βˆ’2C0βˆ’(n+2)F.K_{\mathbb{F}_n} = -2C_0 - (n+2)F.

All Hirzebruch surfaces are rational, so ΞΊ(Fn)=βˆ’βˆž\kappa(\mathbb{F}_n) = -\infty and K2=8K^2 = 8 (for nβ‰₯0n \geq 0). Note K2=(βˆ’2C0βˆ’(n+2)F)2=4(βˆ’n)+2β‹…2(n+2)=βˆ’4n+4n+8=8K^2 = (-2C_0 - (n+2)F)^2 = 4(-n) + 2 \cdot 2(n+2) = -4n + 4n + 8 = 8.


Adjunction formula and canonical class of hypersurfaces

TheoremAdjunction formula for hypersurfaces

Let XX be a smooth projective variety and HβŠ‚XH \subset X a smooth divisor. The adjunction formula states:

KH=(KX+H)∣H.K_H = (K_X + H)|_H.

Equivalently, Ο‰Hβ‰…Ο‰XβŠ—OX(H)∣H\omega_H \cong \omega_X \otimes \mathcal{O}_X(H)|_H. On surfaces, this specializes to: if CC is a smooth curve on a surface XX, then 2g(C)βˆ’2=C2+KXβ‹…C2g(C) - 2 = C^2 + K_X \cdot C.

ExampleCanonical class of smooth hypersurfaces in P^3

Let SβŠ‚P3S \subset \mathbb{P}^3 be a smooth surface of degree dd, so S∼dHS \sim dH where HH is the hyperplane class. By adjunction with KP3=βˆ’4HK_{\mathbb{P}^3} = -4H:

KS=(KP3+S)∣S=(βˆ’4H+dH)∣S=(dβˆ’4)H∣S.K_S = (K_{\mathbb{P}^3} + S)|_S = (-4H + dH)|_S = (d - 4)H|_S.

  • d=1d = 1 (plane P2\mathbb{P}^2): K=βˆ’3H∣SK = -3H|_S, anti-ample. ΞΊ=βˆ’βˆž\kappa = -\infty.
  • d=2d = 2 (quadric β‰…P1Γ—P1\cong \mathbb{P}^1 \times \mathbb{P}^1): K=βˆ’2H∣SK = -2H|_S, anti-ample. ΞΊ=βˆ’βˆž\kappa = -\infty.
  • d=3d = 3 (cubic surface): K=βˆ’H∣SK = -H|_S, anti-ample. ΞΊ=βˆ’βˆž\kappa = -\infty. Rational (del Pezzo).
  • d=4d = 4 (quartic surface): K=0K = 0. This is a K3 surface. ΞΊ=0\kappa = 0.
  • d=5d = 5 (quintic surface): K=H∣SK = H|_S, ample. ΞΊ=2\kappa = 2. Surface of general type.
  • dβ‰₯5d \geq 5: K=(dβˆ’4)H∣SK = (d-4)H|_S is ample, so ΞΊ=2\kappa = 2.

The self-intersection: KS2=(dβˆ’4)2dK_S^2 = (d-4)^2 d (since H∣S2=dH|_S^2 = d). For d=5d = 5: K2=5K^2 = 5. For d=6d = 6: K2=24K^2 = 24.

ExampleCanonical class of complete intersections

For a smooth complete intersection surface S=V(F1,F2)βŠ‚P4S = V(F_1, F_2) \subset \mathbb{P}^4 of degrees d1,d2d_1, d_2:

KS=(d1+d2βˆ’5)H∣S.K_S = (d_1 + d_2 - 5)H|_S.

  • (2,2)(2, 2): K=βˆ’H∣SK = -H|_S, anti-ample. This is a del Pezzo surface of degree 44. ΞΊ=βˆ’βˆž\kappa = -\infty.
  • (2,3)(2, 3): K=0K = 0. A K3 surface. ΞΊ=0\kappa = 0.
  • (2,4)(2, 4): K=H∣SK = H|_S, ample. K2=8K^2 = 8. Surface of general type.
  • (3,3)(3, 3): K=H∣SK = H|_S, ample. K2=9K^2 = 9. Surface of general type.

Kodaira dimension βˆ’βˆž-\infty: rational and ruled surfaces

Definition5.12Rational and ruled surfaces

A surface XX is rational if it is birational to P2\mathbb{P}^2. It is ruled if it is birational to CΓ—P1C \times \mathbb{P}^1 for some smooth curve CC. Ruled surfaces with C=P1C = \mathbb{P}^1 are exactly the rational surfaces.

A surface has ΞΊ(X)=βˆ’βˆž\kappa(X) = -\infty if and only if Pn(X)=0P_n(X) = 0 for all nβ‰₯1n \geq 1, which happens if and only if XX is ruled (Enriques--Kodaira classification).

ExampleSurfaces with kappa = -infty
  • P2\mathbb{P}^2: rational, K=βˆ’3HK = -3H, K2=9K^2 = 9. The prototype.
  • P1Γ—P1\mathbb{P}^1 \times \mathbb{P}^1: rational, K=βˆ’2F1βˆ’2F2K = -2F_1 - 2F_2, K2=8K^2 = 8. Birational to P2\mathbb{P}^2 via blowup/blowdown.
  • Fn\mathbb{F}_n: rational, K2=8K^2 = 8. For nβ‰ 1n \neq 1, these are minimal; F1β‰…Bl⁑P(P2)\mathbb{F}_1 \cong \operatorname{Bl}_P(\mathbb{P}^2).
  • Del Pezzo surfaces SdS_d (degree dd): KSd=βˆ’HK_{S_d} = -H (anticanonical embedding). K2=dK^2 = d for d=1,…,9d = 1, \ldots, 9. These are P2\mathbb{P}^2 blown up at 9βˆ’d9 - d general points (d≀8d \leq 8) or P1Γ—P1\mathbb{P}^1 \times \mathbb{P}^1 (d=8d = 8) or P2\mathbb{P}^2 (d=9d = 9).
  • Ruled surfaces over curves of genus gβ‰₯1g \geq 1: P(E)β†’C\mathbb{P}(\mathcal{E}) \to C for a rank-22 bundle E\mathcal{E} on a curve CC of genus gg. Still ΞΊ=βˆ’βˆž\kappa = -\infty, but not rational when gβ‰₯1g \geq 1.

The key invariant: P12(X)=0P_{12}(X) = 0 characterizes ΞΊ=βˆ’βˆž\kappa = -\infty (Castelnuovo's rationality criterion says P2=q=0P_2 = q = 0 implies rationality).


Kodaira dimension 00: surfaces with torsion canonical class

Definition5.13Surfaces with kappa = 0

A minimal surface has ΞΊ(X)=0\kappa(X) = 0 if and only if nKX∼0nK_X \sim 0 for some nβ‰₯1n \geq 1 (i.e., KXK_X is a torsion element in Pic⁑(X)\operatorname{Pic}(X)). The plurigenera satisfy Pn∈{0,1}P_n \in \{0, 1\} for all nn, and Pn=1P_n = 1 whenever nn is a multiple of the order of KXK_X in Pic⁑(X)\operatorname{Pic}(X).

ExampleK3 surfaces (kappa = 0)

A K3 surface XX is a smooth projective surface with KX∼0K_X \sim 0 (trivial canonical class) and H1(X,OX)=0H^1(X, \mathcal{O}_X) = 0 (simply connected in the algebraic sense: q=0q = 0).

Numerical invariants: K2=0K^2 = 0, pg=1p_g = 1, q=0q = 0, Ο‡(OX)=2\chi(\mathcal{O}_X) = 2, c2(X)=24c_2(X) = 24, b2=22b_2 = 22.

Examples of K3 surfaces:

  • Quartic surface in P3\mathbb{P}^3 (e.g., the Fermat quartic x4+y4+z4+w4=0x^4 + y^4 + z^4 + w^4 = 0).
  • Complete intersection (2,3)βŠ‚P4(2, 3) \subset \mathbb{P}^4.
  • Complete intersection (2,2,2)βŠ‚P5(2, 2, 2) \subset \mathbb{P}^5.
  • Double cover of P2\mathbb{P}^2 branched along a smooth sextic.
  • Kummer surface: Km⁑(A)\operatorname{Km}(A) for an abelian surface AA (minimal resolution of A/{Β±1}A/\{{\pm 1}\}).

All K3 surfaces are deformation-equivalent (over C\mathbb{C}), forming a 2020-dimensional moduli space.

ExampleAbelian surfaces (kappa = 0)

An abelian surface AA is a 22-dimensional abelian variety: a smooth projective group variety of dimension 22. Equivalently, over C\mathbb{C}, A≅C2/ΛA \cong \mathbb{C}^2 / \Lambda for a lattice Λ≅Z4\Lambda \cong \mathbb{Z}^4.

The canonical class is trivial: KA=0K_A = 0 (the group structure provides a nowhere-vanishing 22-form dz1∧dz2dz_1 \wedge dz_2).

Numerical invariants: K2=0K^2 = 0, pg=1p_g = 1, q=2q = 2, Ο‡(OA)=0\chi(\mathcal{O}_A) = 0, c2(A)=0c_2(A) = 0, b1=4b_1 = 4, b2=6b_2 = 6.

Examples: products E1Γ—E2E_1 \times E_2 of elliptic curves, Jacobians Jac⁑(C)\operatorname{Jac}(C) of genus-22 curves.

ExampleEnriques surfaces (kappa = 0)

An Enriques surface XX is a smooth projective surface with KX≁0K_X \not\sim 0 but 2KX∼02K_X \sim 0, and q=0q = 0.

The canonical class has order exactly 22 in Pic⁑(X)\operatorname{Pic}(X). The plurigenera alternate: Pn=0P_n = 0 for nn odd, Pn=1P_n = 1 for nn even.

Numerical invariants: K2=0K^2 = 0, pg=0p_g = 0, q=0q = 0, Ο‡(OX)=1\chi(\mathcal{O}_X) = 1, c2(X)=12c_2(X) = 12, b2=10b_2 = 10.

Every Enriques surface admits an unramified double cover that is a K3 surface. The moduli space of Enriques surfaces (over C\mathbb{C}) has dimension 1010.

ExampleBielliptic (hyperelliptic) surfaces (kappa = 0)

A bielliptic surface (also called hyperelliptic surface) is a surface of the form X=(E1Γ—E2)/GX = (E_1 \times E_2)/G where E1,E2E_1, E_2 are elliptic curves and GG is a finite group acting by translations on E1E_1 and faithfully on E2E_2.

The canonical class satisfies nKX∼0nK_X \sim 0 where n∈{2,3,4,6}n \in \{2, 3, 4, 6\} depending on GG.

Numerical invariants: K2=0K^2 = 0, pg=0p_g = 0, q=1q = 1, Ο‡(OX)=0\chi(\mathcal{O}_X) = 0, c2(X)=0c_2(X) = 0.

These are the only minimal surfaces with ΞΊ=0\kappa = 0 and q=1q = 1.

RemarkComplete classification of kappa = 0

Over C\mathbb{C}, the four classes of minimal surfaces with ΞΊ=0\kappa = 0 are distinguished by their invariants:

  • K3: K∼0K \sim 0, q=0q = 0, pg=1p_g = 1, Ο‡(O)=2\chi(\mathcal{O}) = 2.
  • Abelian: K∼0K \sim 0, q=2q = 2, pg=1p_g = 1, Ο‡(O)=0\chi(\mathcal{O}) = 0.
  • Enriques: 2K∼02K \sim 0, K≁0K \not\sim 0, q=0q = 0, pg=0p_g = 0, Ο‡(O)=1\chi(\mathcal{O}) = 1.
  • Bielliptic: nK∼0nK \sim 0 (n=2,3,4,6n = 2, 3, 4, 6), q=1q = 1, pg=0p_g = 0, Ο‡(O)=0\chi(\mathcal{O}) = 0.

In all four cases, K2=0K^2 = 0 and c2=12Ο‡(OX)c_2 = 12\chi(\mathcal{O}_X) (by Noether's formula).


Kodaira dimension 11: elliptic fibrations

Definition5.14Elliptic surfaces

A smooth projective surface XX has κ(X)=1\kappa(X) = 1 if and only if it admits an elliptic fibration f:X→Cf: X \to C (a morphism to a curve whose general fiber is an elliptic curve) and the canonical class KXK_X is not torsion.

For such surfaces, KXK_X is numerically equivalent to a positive rational multiple of the fiber class. The plurigenera grow linearly: Pn∼cnP_n \sim cn for some positive constant cc as nβ†’βˆžn \to \infty.

ExampleSurfaces with kappa = 1
  • Elliptic surface over P1\mathbb{P}^1 with Ο‡(O)β‰₯3\chi(\mathcal{O}) \geq 3: For instance, the surface y2=x3+t12+1y^2 = x^3 + t^{12} + 1 in a suitable weighted projective bundle over Pt1\mathbb{P}^1_t gives ΞΊ=1\kappa = 1.
  • Product EΓ—CE \times C where EE is elliptic and g(C)β‰₯2g(C) \geq 2: K=Ο€1βˆ—KE+Ο€2βˆ—KC=Ο€2βˆ—KCK = \pi_1^* K_E + \pi_2^* K_C = \pi_2^* K_C. Then Pn=h0(C,nKC)=(2nβˆ’1)(gβˆ’1)P_n = h^0(C, nK_C) = (2n - 1)(g - 1) for nβ‰₯2n \geq 2, growing linearly. So ΞΊ=1\kappa = 1.
  • Dolgachev surfaces: certain logarithmic transforms of rational elliptic surfaces, giving pg=0p_g = 0, q=0q = 0, ΞΊ=1\kappa = 1.

Note: surfaces with ΞΊ=1\kappa = 1 always have K2=0K^2 = 0 (since KK is nef and K2=0K^2 = 0 for a numerically fiber-type class).


Kodaira dimension 22: surfaces of general type

Definition5.15Surface of general type

A smooth projective surface XX has κ(X)=2\kappa(X) = 2 (is of general type) if and only if, for the minimal model Xmin⁑X_{\min}, the canonical class KXmin⁑K_{X_{\min}} is nef and big:

  • Nef: Kβ‹…Cβ‰₯0K \cdot C \geq 0 for every curve CβŠ‚Xmin⁑C \subset X_{\min}.
  • Big: K2>0K^2 > 0.

Equivalently, PnP_n grows quadratically: Pn∼12K2n2P_n \sim \frac{1}{2}K^2 n^2 as nβ†’βˆžn \to \infty (by Riemann--Roch and vanishing). These are the "generic" surfaces, analogous to curves of genus β‰₯2\geq 2.

ExampleSurfaces of general type
  • Quintic in P3\mathbb{P}^3 (d=5d = 5): K=H∣SK = H|_S, K2=5K^2 = 5, Ο‡(O)=5\chi(\mathcal{O}) = 5, pg=4p_g = 4, q=0q = 0.
  • Sextic in P3\mathbb{P}^3 (d=6d = 6): K=2H∣SK = 2H|_S, K2=24K^2 = 24, pg=20p_g = 20.
  • Complete intersection (2,4)βŠ‚P4(2, 4) \subset \mathbb{P}^4: K=H∣SK = H|_S, K2=8K^2 = 8.
  • Complete intersection (3,3)βŠ‚P4(3, 3) \subset \mathbb{P}^4: K=H∣SK = H|_S, K2=9K^2 = 9.
  • Product C1Γ—C2C_1 \times C_2 with g(Ci)β‰₯2g(C_i) \geq 2: K=Ο€1βˆ—KC1+Ο€2βˆ—KC2K = \pi_1^*K_{C_1} + \pi_2^*K_{C_2}, K2=2(2g1βˆ’2)(2g2βˆ’2)=8(g1βˆ’1)(g2βˆ’1)K^2 = 2(2g_1 - 2)(2g_2 - 2) = 8(g_1 - 1)(g_2 - 1).
  • Godeaux surface: pg=q=0p_g = q = 0, K2=1K^2 = 1. The simplest surface of general type. Classically constructed as Fermatquintic⁑/(Z/5)\operatorname{Fermat quintic}/(\mathbb{Z}/5).
  • Barlow surface: pg=q=0p_g = q = 0, K2=1K^2 = 1. A simply connected surface of general type -- showing that Ο€1\pi_1 cannot distinguish rational surfaces from general type in all cases.
  • Beauville surface: pg=q=0p_g = q = 0, K2=8K^2 = 8. A quotient (C1Γ—C2)/G(C_1 \times C_2)/G for suitable curves and group GG.

Abundance for surfaces

TheoremAbundance theorem for surfaces

Let XX be a smooth projective minimal surface with ΞΊ(X)β‰₯0\kappa(X) \geq 0. Then KXK_X is semiample: there exists n>0n > 0 such that the linear system ∣nKX∣|nK_X| is base-point-free.

In particular, ∣nKX∣|nK_X| defines a morphism Ο•βˆ£nKX∣:Xβ†’PPnβˆ’1\phi_{|nK_X|}: X \to \mathbb{P}^{P_n - 1}, called the pluricanonical map. The image has dimension equal to ΞΊ(X)\kappa(X).

Concretely:

  • ΞΊ=0\kappa = 0: Ο•βˆ£nK∣\phi_{|nK|} maps XX to a point (since Pn≀1P_n \leq 1).
  • ΞΊ=1\kappa = 1: Ο•βˆ£nK∣\phi_{|nK|} factors through the elliptic fibration Xβ†’CX \to C.
  • ΞΊ=2\kappa = 2: Ο•βˆ£nK∣\phi_{|nK|} is birational onto its image for n≫0n \gg 0 (in fact, nβ‰₯5n \geq 5 suffices by Bombieri's theorem).
RemarkAbundance in higher dimensions

The abundance conjecture -- that nef canonical divisors on minimal varieties are always semiample -- is proven for surfaces (Mumford, Kawamata, and others) and threefolds (Miyaoka, Kawamata), but remains open in dimension β‰₯4\geq 4. It is one of the central conjectures in birational geometry, closely related to the Minimal Model Program.


Numerical invariants and Noether's formula

Theorem5.2Noether's formula

For any smooth projective surface XX:

Ο‡(OX)=112(KX2+c2(X)),\chi(\mathcal{O}_X) = \frac{1}{12}(K_X^2 + c_2(X)),

where c2(X)=e(X)c_2(X) = e(X) is the topological Euler characteristic (the second Chern number of TXT_X).

Writing out: 1βˆ’q+pg=112(K2+e(X))1 - q + p_g = \frac{1}{12}(K^2 + e(X)), so K2+e(X)=12Ο‡(OX)K^2 + e(X) = 12\chi(\mathcal{O}_X).

ExampleNoether's formula: verifications
  • P2\mathbb{P}^2: K2=9K^2 = 9, e=3e = 3 (Betti numbers 1,0,1,0,11, 0, 1, 0, 1), Ο‡(O)=1\chi(\mathcal{O}) = 1. Check: 9+3=129 + 3 = 12 βœ“.
  • P1Γ—P1\mathbb{P}^1 \times \mathbb{P}^1: K2=8K^2 = 8, e=4e = 4 (Betti numbers 1,0,2,0,11, 0, 2, 0, 1), Ο‡(O)=1\chi(\mathcal{O}) = 1. Check: 8+4=128 + 4 = 12 βœ“.
  • K3 surface: K2=0K^2 = 0, Ο‡(O)=2\chi(\mathcal{O}) = 2. Noether gives e=24e = 24, so b2=22b_2 = 22.
  • Abelian surface: K2=0K^2 = 0, Ο‡(O)=0\chi(\mathcal{O}) = 0. Noether gives e=0e = 0, consistent with b0=1,b1=4,b2=6b_0 = 1, b_1 = 4, b_2 = 6 giving e=1βˆ’4+6βˆ’4+1=0e = 1 - 4 + 6 - 4 + 1 = 0.
  • Enriques surface: K2=0K^2 = 0, Ο‡(O)=1\chi(\mathcal{O}) = 1. Noether gives e=12e = 12, so b2=10b_2 = 10.
  • Quintic in P3\mathbb{P}^3: K2=5K^2 = 5, Ο‡(O)=5\chi(\mathcal{O}) = 5 (since pg=4,q=0p_g = 4, q = 0). Noether gives e=55e = 55.

Geography of surfaces: the BMY inequality

TheoremBogomolov--Miyaoka--Yau inequality

For any smooth projective surface XX of general type (over C\mathbb{C}):

KX2≀3c2(X)=3e(X).K_X^2 \leq 3c_2(X) = 3e(X).

Using Noether's formula K2+e=12Ο‡(O)K^2 + e = 12\chi(\mathcal{O}), this is equivalent to:

KX2≀9Ο‡(OX).K_X^2 \leq 9\chi(\mathcal{O}_X).

Equality K2=3c2K^2 = 3c_2 (equivalently K2=9Ο‡K^2 = 9\chi) holds if and only if the universal cover of XX is the complex 22-ball B2βŠ‚C2\mathbb{B}^2 \subset \mathbb{C}^2, i.e., XX is a ball quotient.

TheoremNoether's inequality

For a minimal surface of general type:

KX2β‰₯2Ο‡(OX)βˆ’6.K_X^2 \geq 2\chi(\mathcal{O}_X) - 6.

Equivalently, K2β‰₯2pgβˆ’4K^2 \geq 2p_g - 4 when q=0q = 0. This provides a lower bound complementing the BMY upper bound.

ExampleGeography of surfaces of general type

The geography problem asks: which pairs (K2,Ο‡(O))(K^2, \chi(\mathcal{O})) are realized by minimal surfaces of general type? The constraints are:

  • K2>0K^2 > 0 and Ο‡(O)>0\chi(\mathcal{O}) > 0 (general type, minimal).
  • BMY inequality: K2≀9Ο‡(O)K^2 \leq 9\chi(\mathcal{O}) (the "BMY line").
  • Noether inequality: K2β‰₯2Ο‡(O)βˆ’6K^2 \geq 2\chi(\mathcal{O}) - 6 (the "Noether line").
  • Bogomolov--Miyaoka--Yau: c12≀3c2c_1^2 \leq 3c_2.
  • K2K^2 and Ο‡\chi are positive integers with some congruence conditions.

The "geography" of surfaces is the region in the (K2,Ο‡)(K^2, \chi)-plane between the Noether line and the BMY line:

2Ο‡βˆ’6≀K2≀9Ο‡.2\chi - 6 \leq K^2 \leq 9\chi.

  • On the BMY line (K2=9Ο‡K^2 = 9\chi): ball quotients. The simplest is the fake projective plane with K2=9,Ο‡=1K^2 = 9, \chi = 1 (pg=q=0p_g = q = 0). Exactly 100100 fake projective planes exist (Prasad--Yeung, Cartwright--Steger).
  • On the Noether line (K2=2Ο‡βˆ’6K^2 = 2\chi - 6): these include many families of fibered surfaces.
  • K2=1,Ο‡=1K^2 = 1, \chi = 1: Godeaux surfaces (pg=q=0p_g = q = 0).
  • K2=2,Ο‡=1K^2 = 2, \chi = 1: Campedelli surfaces (pg=q=0p_g = q = 0).
  • K2=1,Ο‡=2K^2 = 1, \chi = 2: numerical Godeaux with pg=1p_g = 1, no known examples with q=0q = 0 -- an open problem.
  • Most lattice points in the region are realized by some surface.

The canonical class under birational transformations

TheoremCanonical class of a blowup

Let Ο€:X~β†’X\pi: \tilde{X} \to X be the blowup of a smooth surface XX at a point PP, with exceptional divisor EE. Then:

KX~=Ο€βˆ—KX+E.K_{\tilde{X}} = \pi^* K_X + E.

Consequently: KX~2=KX2βˆ’1K_{\tilde{X}}^2 = K_X^2 - 1 (since E2=βˆ’1E^2 = -1 and Ο€βˆ—KXβ‹…E=0\pi^*K_X \cdot E = 0). Each blowup decreases K2K^2 by 11.

ExampleBlowups of P^2

Starting from P2\mathbb{P}^2 with K2=9K^2 = 9 and blowing up rr points in general position:

K2=9βˆ’r,KX~=βˆ’3H+E1+β‹―+Er.K^2 = 9 - r, \quad K_{\tilde{X}} = -3H + E_1 + \cdots + E_r.

  • r=0r = 0: K2=9K^2 = 9, this is P2\mathbb{P}^2.
  • r=1r = 1: K2=8K^2 = 8, this is F1\mathbb{F}_1.
  • r≀8r \leq 8: del Pezzo surfaces S9βˆ’rS_{9-r}, with βˆ’K-K ample.
  • r=9r = 9: K2=0K^2 = 0, a rational elliptic surface. βˆ’K-K defines an elliptic pencil βˆ£βˆ’K∣|-K|.
  • r=10r = 10: K2=βˆ’1K^2 = -1. No longer del Pezzo; the surface becomes increasingly complicated.

For del Pezzo surfaces (r≀8r \leq 8), KK is anti-ample and βˆ’K-K embeds the surface as a degree-(9βˆ’r)(9-r) surface in P9βˆ’r\mathbb{P}^{9-r}.


Invariants derived from KXK_X

Definition5.16Irregularity and geometric genus

For a smooth projective surface XX:

  • Geometric genus: pg=h0(X,KX)=h0,2=h2,0p_g = h^0(X, K_X) = h^{0,2} = h^{2,0} (over C\mathbb{C}, by Hodge theory).
  • Irregularity: q=h1(X,OX)=h0,1q = h^1(X, \mathcal{O}_X) = h^{0,1} (dimension of Pic⁑0(X)=Alb⁑(X)\operatorname{Pic}^0(X) = \operatorname{Alb}(X)).
  • Holomorphic Euler characteristic: Ο‡(OX)=1βˆ’q+pg\chi(\mathcal{O}_X) = 1 - q + p_g.
  • Chern numbers: c12=KX2c_1^2 = K_X^2 and c2=e(X)=c_2 = e(X) = topological Euler number.
  • Relation: K2+c2=12Ο‡(OX)K^2 + c_2 = 12\chi(\mathcal{O}_X) (Noether's formula).

These satisfy further constraints: pgβ‰₯0p_g \geq 0, qβ‰₯0q \geq 0, and for surfaces of general type, Ο‡(OX)>0\chi(\mathcal{O}_X) > 0.

ExampleTable of invariants for classical surfaces

Summary of the canonical class and invariants:

Rational surfaces: K2=9βˆ’rK^2 = 9 - r (for Bl⁑rP2\operatorname{Bl}_r \mathbb{P}^2), q=0q = 0, pg=0p_g = 0, Ο‡(O)=1\chi(\mathcal{O}) = 1.

K3 surfaces: K∼0K \sim 0, K2=0K^2 = 0, q=0q = 0, pg=1p_g = 1, Ο‡(O)=2\chi(\mathcal{O}) = 2, c2=24c_2 = 24.

Abelian surfaces: K∼0K \sim 0, K2=0K^2 = 0, q=2q = 2, pg=1p_g = 1, Ο‡(O)=0\chi(\mathcal{O}) = 0, c2=0c_2 = 0.

Enriques surfaces: 2K∼02K \sim 0, K2=0K^2 = 0, q=0q = 0, pg=0p_g = 0, Ο‡(O)=1\chi(\mathcal{O}) = 1, c2=12c_2 = 12.

Bielliptic surfaces: nK∼0nK \sim 0, K2=0K^2 = 0, q=1q = 1, pg=0p_g = 0, Ο‡(O)=0\chi(\mathcal{O}) = 0, c2=0c_2 = 0.

Quintic surface (P3\mathbb{P}^3): K=H∣SK = H|_S, K2=5K^2 = 5, q=0q = 0, pg=4p_g = 4, Ο‡(O)=5\chi(\mathcal{O}) = 5, c2=55c_2 = 55.

Product Cg1Γ—Cg2C_{g_1} \times C_{g_2} (giβ‰₯2g_i \geq 2): K2=8(g1βˆ’1)(g2βˆ’1)K^2 = 8(g_1-1)(g_2-1), q=g1+g2q = g_1 + g_2, pg=g1g2p_g = g_1 g_2, c2=4(g1βˆ’1)(g2βˆ’1)c_2 = 4(g_1-1)(g_2-1). BMY ratio: K2/c2=2K^2/c_2 = 2.


Pluricanonical maps and Bombieri's theorem

TheoremBombieri's theorem

Let XX be a minimal surface of general type. Then the nn-th pluricanonical map Ο•βˆ£nKX∣:Xβ†’PPnβˆ’1\phi_{|nK_X|}: X \to \mathbb{P}^{P_n - 1} is a birational morphism onto its image for all nβ‰₯5n \geq 5.

Moreover:

  • ∣2KX∣|2K_X| is base-point-free (for K2β‰₯2K^2 \geq 2; if K2=1K^2 = 1 then ∣3K∣|3K| is base-point-free).
  • ∣5KX∣|5K_X| is very ample (gives a birational embedding) in all cases.
  • The image Ο•βˆ£nK∣(X)\phi_{|nK|}(X) is a canonical model with at worst rational double point singularities.
ExampleGrowth of plurigenera

By Riemann--Roch and Kawamata--Viehweg vanishing, for nβ‰₯2n \geq 2 on a minimal surface of general type:

Pn=h0(nKX)=n(nβˆ’1)2K2+Ο‡(OX).P_n = h^0(nK_X) = \frac{n(n-1)}{2}K^2 + \chi(\mathcal{O}_X).

Explicit values for a quintic surface (K2=5K^2 = 5, Ο‡=5\chi = 5):

  • P1=pg=4P_1 = p_g = 4.
  • P2=5+5=10P_2 = 5 + 5 = 10.
  • P3=15+5=20P_3 = 15 + 5 = 20.
  • P4=30+5=35P_4 = 30 + 5 = 35.
  • P5=50+5=55P_5 = 50 + 5 = 55.

For a Godeaux surface (K2=1K^2 = 1, Ο‡=1\chi = 1):

  • P1=0P_1 = 0 (since pg=0p_g = 0, q=0q = 0).
  • P2=1+1=2P_2 = 1 + 1 = 2.
  • P3=3+1=4P_3 = 3 + 1 = 4.
  • P4=6+1=7P_4 = 6 + 1 = 7.
  • P5=10+1=11P_5 = 10 + 1 = 11.

Note that P1P_1 can differ from the Riemann--Roch prediction because vanishing may fail for n=1n = 1.


The canonical class and the minimal model program

RemarkMinimal models and canonical class

The minimal model program (MMP) for surfaces proceeds by contracting (βˆ’1)(-1)-curves (smooth rational curves EE with E2=βˆ’1E^2 = -1, equivalently Kβ‹…E=βˆ’1K \cdot E = -1 by adjunction). Each contraction increases K2K^2 by 11.

A smooth surface is minimal if it contains no (βˆ’1)(-1)-curves. For a minimal surface:

  • ΞΊ=βˆ’βˆž\kappa = -\infty: XX is P2\mathbb{P}^2 or a ruled surface P(E)β†’C\mathbb{P}(\mathcal{E}) \to C (Castelnuovo--Enriques).
  • ΞΊβ‰₯0\kappa \geq 0: KXK_X is nef, and the canonical model exists by the abundance theorem.

The canonical class governs the entire MMP: (βˆ’1)(-1)-curves are exactly the irreducible curves EE with Kβ‹…E<0K \cdot E < 0 and E2<0E^2 < 0, and their contraction "improves" KK toward nefness.

ExampleRunning the MMP on a blowup

Let X=Bl⁑P1,P2(P2)X = \operatorname{Bl}_{P_1, P_2}(\mathbb{P}^2), the blowup of P2\mathbb{P}^2 at two points, with KX=βˆ’3H+E1+E2K_X = -3H + E_1 + E_2, K2=7K^2 = 7.

The (βˆ’1)(-1)-curves are E1,E2E_1, E_2, and the strict transform LL of the line through P1,P2P_1, P_2 (with L∼Hβˆ’E1βˆ’E2L \sim H - E_1 - E_2, L2=βˆ’1L^2 = -1).

  • Contract E1E_1: get Bl⁑P2(P2)=F1\operatorname{Bl}_{P_2}(\mathbb{P}^2) = \mathbb{F}_1, K2=8K^2 = 8.
  • Contract the remaining (βˆ’1)(-1)-curve on F1\mathbb{F}_1: get P2\mathbb{P}^2, K2=9K^2 = 9.

Alternatively, contract LL first: get a surface with K2=8K^2 = 8, which is P1Γ—P1\mathbb{P}^1 \times \mathbb{P}^1 (since contracting LL resolves to the other ruling). Then P1Γ—P1\mathbb{P}^1 \times \mathbb{P}^1 is already minimal.

The MMP is not unique, but the minimal model for ΞΊ=βˆ’βˆž\kappa = -\infty is always P2\mathbb{P}^2 or a Hirzebruch surface.


Summary

RemarkThe canonical class as the master invariant

The canonical class KXK_X organizes the entire theory of algebraic surfaces:

Classification: The Kodaira dimension ΞΊ(X)∈{βˆ’βˆž,0,1,2}\kappa(X) \in \{-\infty, 0, 1, 2\} determines the coarse type: rational/ruled, K3/abelian/Enriques/bielliptic, elliptic, or general type.

Numerical constraints: Noether's formula K2+c2=12Ο‡K^2 + c_2 = 12\chi and the BMY inequality K2≀3c2K^2 \leq 3c_2 bound the geography of surfaces.

Birational geometry: The MMP contracts curves with Kβ‹…E<0K \cdot E < 0 to reach a minimal model where KK is nef, and abundance ensures that ∣nK∣|nK| eventually defines the Iitaka fibration.

Pluricanonical maps: For general type surfaces, ∣5K∣|5K| is birational (Bombieri), and Pn=n(nβˆ’1)2K2+Ο‡(O)P_n = \frac{n(n-1)}{2}K^2 + \chi(\mathcal{O}) for nβ‰₯2n \geq 2.