ConceptComplete

The Dualizing Sheaf Ο‰X\omega_X

The dualizing sheaf is a fundamental object in algebraic geometry that generalizes the canonical bundle from differential geometry to arbitrary schemes. It plays a central role in Serre duality and provides crucial invariants for classification of algebraic varieties.

Motivation and Definition

DefinitionDualizing Sheaf (Smooth Case)

Let XX be a smooth variety of dimension nn over an algebraically closed field kk. The dualizing sheaf (or canonical sheaf) Ο‰X\omega_X is defined as: Ο‰X=β‹€nΞ©X/k\omega_X = \bigwedge^n \Omega_{X/k} where Ξ©X/k\Omega_{X/k} is the sheaf of KΓ€hler differentials and β‹€n\bigwedge^n denotes the nn-th exterior power.

The corresponding invertible sheaf is also called the canonical bundle, and the divisor class it represents is called the canonical divisor KXK_X.

Remark

The dualizing sheaf has several equivalent characterizations:

  1. Differential Forms: As the sheaf of top differential forms
  2. Serre Duality: As the sheaf Ο‰X\omega_X such that for any coherent sheaf F\mathscr{F}: Hi(X,F)β‰…Hnβˆ’i(X,Fβˆ¨βŠ—Ο‰X)∨H^i(X, \mathscr{F}) \cong H^{n-i}(X, \mathscr{F}^\vee \otimes \omega_X)^\vee
  3. Adjunction: Via the adjunction formula for divisors
  4. Determinant: As det⁑(ΩX/k)\det(\Omega_{X/k}) when XX is smooth

For singular varieties, the definition via Serre duality is most general.

KΓ€hler Differentials and the Canonical Sheaf

DefinitionKΓ€hler Differentials

For a morphism of schemes f:Xβ†’Yf: X \to Y, the sheaf of KΓ€hler differentials Ξ©X/Y\Omega_{X/Y} is defined as the sheaf associated to the OX\mathscr{O}_X-module Ξ©R/S\Omega_{R/S} for affine opens Spec(R)βŠ†X\mathrm{Spec}(R) \subseteq X over Spec(S)βŠ†Y\mathrm{Spec}(S) \subseteq Y, where: Ξ©R/S=I/I2\Omega_{R/S} = I/I^2 with II the kernel of the multiplication map RβŠ—SRβ†’RR \otimes_S R \to R.

The universal derivation d:Rβ†’Ξ©R/Sd: R \to \Omega_{R/S} sends r↦1βŠ—rβˆ’rβŠ—1r \mapsto 1 \otimes r - r \otimes 1.

TheoremProperties of KΓ€hler Differentials

Let XX be a smooth variety of dimension nn over kk. Then:

  1. Ξ©X/k\Omega_{X/k} is locally free of rank nn
  2. For any open cover by affines Ui=Spec(Ai)U_i = \mathrm{Spec}(A_i) with local parameters t1,…,tnt_1, \ldots, t_n, we have: Ξ©Ai/kβ‰…Aiβ‹…dt1βŠ•β‹―βŠ•Aiβ‹…dtn\Omega_{A_i/k} \cong A_i \cdot dt_1 \oplus \cdots \oplus A_i \cdot dt_n
  3. If XX is a closed subvariety of YY, there is an exact sequence: I/I2β†’Ξ©Y/k∣Xβ†’Ξ©X/kβ†’0\mathscr{I}/\mathscr{I}^2 \to \Omega_{Y/k}|_X \to \Omega_{X/k} \to 0 where I\mathscr{I} is the ideal sheaf of XX in YY
ExampleCanonical Sheaf of Projective Space

For projective space Pkn\mathbb{P}^n_k, the canonical sheaf is: Ο‰Pn=OPn(βˆ’nβˆ’1)\omega_{\mathbb{P}^n} = \mathscr{O}_{\mathbb{P}^n}(-n-1)

Computation: Cover Pn\mathbb{P}^n by affine charts Ui=Spec(k[x0/xi,…,xi/xi^,…,xn/xi])U_i = \mathrm{Spec}(k[x_0/x_i, \ldots, \widehat{x_i/x_i}, \ldots, x_n/x_i]). On U0U_0, set tj=xj/x0t_j = x_j/x_0 for j=1,…,nj = 1, \ldots, n. Then: Ξ©U0/k=⨁j=1nOU0β‹…dtj\Omega_{U_0/k} = \bigoplus_{j=1}^n \mathscr{O}_{U_0} \cdot dt_j

The top form is: Ο‰βˆ£U0=OU0β‹…dt1βˆ§β‹―βˆ§dtn\omega|_{U_0} = \mathscr{O}_{U_0} \cdot dt_1 \wedge \cdots \wedge dt_n

On the overlap U0∩U1U_0 \cap U_1 with coordinates sj=xj/x1s_j = x_j/x_1, we have t1=1/s0t_1 = 1/s_0 and tj=sj/s0t_j = s_j/s_0 for jβ‰₯2j \geq 2. The Jacobian gives: dt1βˆ§β‹―βˆ§dtn=(βˆ’1)nβˆ’1β‹…s0βˆ’(n+1)β‹…ds1βˆ§β‹―βˆ§dsndt_1 \wedge \cdots \wedge dt_n = (-1)^{n-1} \cdot s_0^{-(n+1)} \cdot ds_1 \wedge \cdots \wedge ds_n

Since s0=x0/x1s_0 = x_0/x_1 has degree 00, the factor s0βˆ’(n+1)s_0^{-(n+1)} corresponds to twisting by O(βˆ’nβˆ’1)\mathscr{O}(-n-1).

Consequence: The canonical divisor KPn=(βˆ’nβˆ’1)HK_{\mathbb{P}^n} = (-n-1)H where HH is a hyperplane.

ExampleCanonical Sheaf of Smooth Curves

Let CC be a smooth projective curve of genus gg over kk. Then: Ο‰C=Ξ©C/k\omega_C = \Omega_{C/k} is a line bundle of degree 2gβˆ’22g - 2.

For P1\mathbb{P}^1: We have Ο‰P1=OP1(βˆ’2)\omega_{\mathbb{P}^1} = \mathscr{O}_{\mathbb{P}^1}(-2) and deg⁑(KP1)=βˆ’2=2(0)βˆ’2\deg(K_{\mathbb{P}^1}) = -2 = 2(0) - 2.

For an elliptic curve EE: We have Ο‰Eβ‰…OE\omega_E \cong \mathscr{O}_E and deg⁑(KE)=0=2(1)βˆ’2\deg(K_E) = 0 = 2(1) - 2.

For a curve of genus gβ‰₯2g \geq 2: The canonical divisor KCK_C is ample when gβ‰₯2g \geq 2, giving the canonical embedding: Ο•K:Cβ†’Pgβˆ’1,p↦[s0(p):β‹―:sgβˆ’1(p)]\phi_K: C \to \mathbb{P}^{g-1}, \quad p \mapsto [s_0(p) : \cdots : s_{g-1}(p)] where s0,…,sgβˆ’1s_0, \ldots, s_{g-1} is a basis of H0(C,Ο‰C)H^0(C, \omega_C).

The Adjunction Formula

TheoremAdjunction Formula

Let YY be a smooth variety and XβŠ†YX \subseteq Y a smooth divisor. Then: Ο‰X=(Ο‰YβŠ—OY(X))∣X\omega_X = (\omega_Y \otimes \mathscr{O}_Y(X))|_X

Equivalently, in terms of canonical divisors: KX=(KY+X)∣XK_X = (K_Y + X)|_X

Proof

We use the exact sequence of KΓ€hler differentials. Since XX is a smooth divisor, the ideal sheaf IX=OY(βˆ’X)\mathscr{I}_X = \mathscr{O}_Y(-X) and we have: IX/IX2β‰…OY(βˆ’X)∣Xβ‰…OX(βˆ’X)\mathscr{I}_X/\mathscr{I}_X^2 \cong \mathscr{O}_Y(-X)|_X \cong \mathscr{O}_X(-X)

The conormal sequence is: 0β†’OX(βˆ’X)β†’Ξ©Y/k∣Xβ†’Ξ©X/kβ†’00 \to \mathscr{O}_X(-X) \to \Omega_{Y/k}|_X \to \Omega_{X/k} \to 0

Taking top exterior powers (determinants): det⁑(Ξ©Y/k∣X)βŠ—OX(X)β‰…det⁑(Ξ©X/k)\det(\Omega_{Y/k}|_X) \otimes \mathscr{O}_X(X) \cong \det(\Omega_{X/k})

Since det⁑(Ξ©Y/k∣X)=Ο‰Y∣X\det(\Omega_{Y/k}|_X) = \omega_Y|_X, we obtain: Ο‰X=(Ο‰YβŠ—OY(X))∣X\omega_X = (\omega_Y \otimes \mathscr{O}_Y(X))|_X

β– 
ExampleCanonical Sheaf of Hypersurfaces

Let XβŠ†Pn+1X \subseteq \mathbb{P}^{n+1} be a smooth hypersurface of degree dd. By adjunction: Ο‰X=(Ο‰Pn+1βŠ—OPn+1(d))∣X=OPn+1(βˆ’nβˆ’2+d)∣X=OX(dβˆ’nβˆ’2)\omega_X = (\omega_{\mathbb{P}^{n+1}} \otimes \mathscr{O}_{\mathbb{P}^{n+1}}(d))|_X = \mathscr{O}_{\mathbb{P}^{n+1}}(-n-2+d)|_X = \mathscr{O}_X(d-n-2)

Special cases:

  • Plane curve (n=1n = 1): Ο‰C=OC(dβˆ’3)\omega_C = \mathscr{O}_C(d-3)
    • For d=3d = 3 (cubic): Ο‰Cβ‰…OC\omega_C \cong \mathscr{O}_C (elliptic curve)
    • For d=4d = 4 (quartic): Ο‰Cβ‰…OC(1)\omega_C \cong \mathscr{O}_C(1) (genus 3)
  • Surface in P3\mathbb{P}^3 (n=2n = 2): Ο‰S=OS(dβˆ’4)\omega_S = \mathscr{O}_S(d-4)
    • For d=4d = 4 (quartic): Ο‰Sβ‰…OS\omega_S \cong \mathscr{O}_S (K3 surface)
    • For d=5d = 5 (quintic): Ο‰Sβ‰…OS(1)\omega_S \cong \mathscr{O}_S(1)
  • Threefold in P4\mathbb{P}^4 (n=3n = 3): Ο‰X=OX(dβˆ’5)\omega_X = \mathscr{O}_X(d-5)
    • For d=5d = 5 (quintic): Ο‰Xβ‰…OX\omega_X \cong \mathscr{O}_X (Calabi-Yau threefold)
ExampleComplete Intersections

Let XβŠ†PnX \subseteq \mathbb{P}^n be a smooth complete intersection of hypersurfaces of degrees d1,…,drd_1, \ldots, d_r. Then dim⁑(X)=nβˆ’r\dim(X) = n - r and: Ο‰X=OX(βˆ‘i=1rdiβˆ’nβˆ’1)\omega_X = \mathscr{O}_X\left(\sum_{i=1}^r d_i - n - 1\right)

Derivation: Apply adjunction repeatedly. If X=H1βˆ©β‹―βˆ©HrX = H_1 \cap \cdots \cap H_r where HiH_i has degree did_i: KX=(KPn+d1+β‹―+dr)∣X=(βˆ’nβˆ’1+d1+β‹―+dr)∣XK_X = (K_{\mathbb{P}^n} + d_1 + \cdots + d_r)|_X = (-n-1 + d_1 + \cdots + d_r)|_X

Examples:

  • Curve in P3\mathbb{P}^3 as (d1,d2)(d_1, d_2) complete intersection: Ο‰C=OC(d1+d2βˆ’4)\omega_C = \mathscr{O}_C(d_1 + d_2 - 4) The genus is g=12d1d2(d1+d2βˆ’4)+1g = \frac{1}{2}d_1 d_2 (d_1 + d_2 - 4) + 1.

  • Surface in P4\mathbb{P}^4 as (d1,d2)(d_1, d_2) complete intersection: Ο‰S=OS(d1+d2βˆ’5)\omega_S = \mathscr{O}_S(d_1 + d_2 - 5)

  • Calabi-Yau threefold: (3,3)(3, 3) complete intersection in P5\mathbb{P}^5 has Ο‰Xβ‰…OX\omega_X \cong \mathscr{O}_X.

Serre Duality

TheoremSerre Duality

Let XX be a smooth projective variety of dimension nn over kk. For any coherent sheaf F\mathscr{F} on XX, there is a natural perfect pairing: Hi(X,F)Γ—Hnβˆ’i(X,Fβˆ¨βŠ—Ο‰X)β†’Hn(X,Ο‰X)β‰…kH^i(X, \mathscr{F}) \times H^{n-i}(X, \mathscr{F}^\vee \otimes \omega_X) \to H^n(X, \omega_X) \cong k

This induces isomorphisms: Hi(X,F)β‰…Hnβˆ’i(X,Fβˆ¨βŠ—Ο‰X)∨H^i(X, \mathscr{F}) \cong H^{n-i}(X, \mathscr{F}^\vee \otimes \omega_X)^\vee Hnβˆ’i(X,Fβˆ¨βŠ—Ο‰X)β‰…Hi(X,F)∨H^{n-i}(X, \mathscr{F}^\vee \otimes \omega_X) \cong H^i(X, \mathscr{F})^\vee

Remark

The isomorphism Hn(X,Ο‰X)β‰…kH^n(X, \omega_X) \cong k can be viewed as an "integration" or "trace" map. In the classical case where XX is a smooth complex manifold, this corresponds to integration of top differential forms: ∫X:Hn(X,Ξ©Xn)β†’C\int_X: H^n(X, \Omega_X^n) \to \mathbb{C}

ExampleSerre Duality for Curves

For a smooth projective curve CC of genus gg: H0(C,L)β‰…H1(C,Lβˆ¨βŠ—Ο‰C)∨H^0(C, \mathscr{L}) \cong H^1(C, \mathscr{L}^\vee \otimes \omega_C)^\vee H1(C,L)β‰…H0(C,Lβˆ¨βŠ—Ο‰C)∨H^1(C, \mathscr{L}) \cong H^0(C, \mathscr{L}^\vee \otimes \omega_C)^\vee

Application: For L=OC\mathscr{L} = \mathscr{O}_C: H0(C,OC)β‰…H1(C,Ο‰C)∨H^0(C, \mathscr{O}_C) \cong H^1(C, \omega_C)^\vee H1(C,OC)β‰…H0(C,Ο‰C)∨H^1(C, \mathscr{O}_C) \cong H^0(C, \omega_C)^\vee

Since h0(C,OC)=1h^0(C, \mathscr{O}_C) = 1 and h1(C,OC)=gh^1(C, \mathscr{O}_C) = g, we get h0(C,Ο‰C)=gh^0(C, \omega_C) = g and h1(C,Ο‰C)=1h^1(C, \omega_C) = 1.

By Riemann-Roch: deg⁑(Ο‰C)=2gβˆ’2\deg(\omega_C) = 2g - 2.

Gorenstein Varieties and the Dualizing Sheaf

DefinitionGorenstein Variety

A variety XX is called Gorenstein if it has a dualizing sheaf Ο‰X\omega_X that is locally free (i.e., an invertible sheaf).

More generally, XX is Gorenstein at a point p∈Xp \in X if the local ring OX,p\mathscr{O}_{X,p} is a Gorenstein local ring, meaning it has finite injective dimension as a module over itself.

Remark

The class of Gorenstein varieties includes:

  • All smooth varieties
  • Hypersurfaces in smooth varieties (even singular ones)
  • Complete intersections with isolated singularities
  • Many quotient singularities

But not all varieties are Gorenstein. For instance, the cone over a smooth conic has non-Gorenstein singularity at the vertex in general.

ExampleDualizing Sheaf of Singular Hypersurfaces

Let XβŠ†Pn+1X \subseteq \mathbb{P}^{n+1} be a hypersurface of degree dd (possibly singular). Then XX is Gorenstein and: Ο‰X=OX(dβˆ’nβˆ’2)\omega_X = \mathscr{O}_X(d - n - 2)

This formula holds even when XX is singular, though we cannot derive it from the adjunction formula in the same way.

Example: The cubic surface XβŠ†P3X \subseteq \mathbb{P}^3 defined by x03+x13+x23+x33=0x_0^3 + x_1^3 + x_2^3 + x_3^3 = 0 has at most isolated singularities and: Ο‰X=OX(3βˆ’4)=OX(βˆ’1)\omega_X = \mathscr{O}_X(3 - 4) = \mathscr{O}_X(-1)

Note that Ο‰X\omega_X is still invertible despite potential singularities.

ExampleCanonical and Log-Canonical Singularities

For a normal variety XX, there is a notion of canonical singularities and log-canonical singularities defined using discrepancies.

Let Ο€:X~β†’X\pi: \widetilde{X} \to X be a resolution of singularities with X~\widetilde{X} smooth. Write: KX~=Ο€βˆ—KX+βˆ‘aiEiK_{\widetilde{X}} = \pi^* K_X + \sum a_i E_i where EiE_i are exceptional divisors and ai∈Qa_i \in \mathbb{Q} are discrepancies.

  • XX has canonical singularities if aiβ‰₯0a_i \geq 0 for all ii
  • XX has terminal singularities if ai>0a_i > 0 for all ii
  • XX has log-canonical singularities if aiβ‰₯βˆ’1a_i \geq -1 for all ii

Examples:

  • Smooth points are terminal
  • Nodes (xy=z2)(xy = z^2) are canonical
  • Cusps (x2=y3)(x^2 = y^3) are log-canonical but not canonical

Kodaira Dimension and Classification

DefinitionKodaira Dimension

For a smooth projective variety XX, the plurigenera are defined as: Pm(X)=h0(X,Ο‰XβŠ—m)=dim⁑kH0(X,OX(mKX))P_m(X) = h^0(X, \omega_X^{\otimes m}) = \dim_k H^0(X, \mathscr{O}_X(mK_X))

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

  • ΞΊ(X)=βˆ’βˆž\kappa(X) = -\infty if Pm(X)=0P_m(X) = 0 for all m>0m > 0
  • ΞΊ(X)=max⁑{dimβ‘Ο•βˆ£mKX∣(X):m>0, Pm(X)>0}\kappa(X) = \max\{\dim \phi_{|mK_X|}(X) : m > 0,\, P_m(X) > 0\} otherwise

where Ο•βˆ£mKX∣:Xβ‡’PPm(X)βˆ’1\phi_{|mK_X|}: X \dashrightarrow \mathbb{P}^{P_m(X)-1} is the rational map defined by ∣mKX∣|mK_X|.

Remark

The Kodaira dimension satisfies βˆ’βˆžβ‰€ΞΊ(X)≀dim⁑(X)-\infty \leq \kappa(X) \leq \dim(X) and is a birational invariant. It provides a rough classification:

  • ΞΊ(X)=βˆ’βˆž\kappa(X) = -\infty: Uniruled varieties (covered by rational curves)
  • ΞΊ(X)=0\kappa(X) = 0: Varieties with Ο‰X∼0\omega_X \sim 0 (Calabi-Yau, abelian varieties, K3 surfaces)
  • 0<ΞΊ(X)<dim⁑(X)0 < \kappa(X) < \dim(X): Varieties of intermediate type (fibered varieties)
  • ΞΊ(X)=dim⁑(X)\kappa(X) = \dim(X): Varieties of general type (canonical model exists)
ExampleKodaira Dimension of Curves

For a smooth projective curve CC of genus gg:

  • g=0g = 0 (Cβ‰…P1C \cong \mathbb{P}^1): Ο‰C=OP1(βˆ’2)\omega_C = \mathscr{O}_{\mathbb{P}^1}(-2) is anti-ample, so Pm(C)=0P_m(C) = 0 for all m>0m > 0. Thus ΞΊ(C)=βˆ’βˆž\kappa(C) = -\infty.

  • g=1g = 1 (elliptic curve): Ο‰Cβ‰…OC\omega_C \cong \mathscr{O}_C is trivial, so Pm(C)=1P_m(C) = 1 for all m>0m > 0. The maps Ο•βˆ£mKC∣\phi_{|mK_C|} are all empty. Thus ΞΊ(C)=0\kappa(C) = 0.

  • gβ‰₯2g \geq 2: Ο‰C\omega_C is ample with deg⁑(Ο‰C)=2gβˆ’2>0\deg(\omega_C) = 2g - 2 > 0. For mβ‰₯2m \geq 2: Pm(C)=h0(C,Ο‰CβŠ—m)=(2mβˆ’1)(gβˆ’1)P_m(C) = h^0(C, \omega_C^{\otimes m}) = (2m-1)(g-1) by Riemann-Roch. The canonical map Ο•βˆ£KC∣:Cβ†’Pgβˆ’1\phi_{|K_C|}: C \to \mathbb{P}^{g-1} is an embedding (except for hyperelliptic curves where it's a degree-2 cover). Thus ΞΊ(C)=1\kappa(C) = 1.

ExampleKodaira Dimension of Surfaces

For a smooth projective surface SS:

ΞΊ(S)=βˆ’βˆž\kappa(S) = -\infty: Ruled surfaces, rational surfaces (P2\mathbb{P}^2, P1Γ—P1\mathbb{P}^1 \times \mathbb{P}^1, del Pezzo surfaces)

ΞΊ(S)=0\kappa(S) = 0: Surfaces with numerically trivial canonical class

  • K3 surfaces: Ο‰Sβ‰…OS\omega_S \cong \mathscr{O}_S, Pm(S)=1P_m(S) = 1 for all m>0m > 0
  • Abelian surfaces
  • Enriques surfaces (with 2KS∼02K_S \sim 0)
  • Hyperelliptic surfaces

ΞΊ(S)=1\kappa(S) = 1: Properly elliptic surfaces (elliptic fibrations over curves with KSK_S not nef)

ΞΊ(S)=2\kappa(S) = 2: Surfaces of general type. The canonical ring: R(S)=⨁mβ‰₯0H0(S,Ο‰SβŠ—m)R(S) = \bigoplus_{m \geq 0} H^0(S, \omega_S^{\otimes m}) is finitely generated and Proj(R(S))\mathrm{Proj}(R(S)) gives the canonical model.

ExampleK3 Surfaces

A K3 surface is a smooth projective surface SS with: ωS≅OSandH1(S,OS)=0\omega_S \cong \mathscr{O}_S \quad \text{and} \quad H^1(S, \mathscr{O}_S) = 0

Properties:

  • ΞΊ(S)=0\kappa(S) = 0
  • KS=0K_S = 0 (canonical divisor is numerically trivial)
  • h2,0(S)=1h^{2,0}(S) = 1, h1,0(S)=0h^{1,0}(S) = 0
  • Ο‡(OS)=2\chi(\mathscr{O}_S) = 2

Examples:

  1. Smooth quartic surfaces in P3\mathbb{P}^3: By adjunction, Ο‰S=OS(4βˆ’4)=OS\omega_S = \mathscr{O}_S(4-4) = \mathscr{O}_S.
  2. Complete intersection of a quadric and cubic in P4\mathbb{P}^4: Ο‰S=OS(2+3βˆ’5)=OS\omega_S = \mathscr{O}_S(2+3-5) = \mathscr{O}_S.
  3. Double cover of P2\mathbb{P}^2 branched over a smooth sextic curve.
  4. Kummer surface: Quotient of an abelian surface by involution, then blow up 16 singular points.
ExampleCalabi-Yau Threefolds

A Calabi-Yau threefold is a smooth projective threefold XX with: ωX≅OXandHi(X,OX)=0 for i=1,2\omega_X \cong \mathscr{O}_X \quad \text{and} \quad H^i(X, \mathscr{O}_X) = 0 \text{ for } i = 1, 2

Properties:

  • ΞΊ(X)=0\kappa(X) = 0
  • KX=0K_X = 0
  • Ο‡(OX)=0\chi(\mathscr{O}_X) = 0 (from Serre duality)

Examples:

  1. Smooth quintic threefold in P4\mathbb{P}^4: By adjunction, Ο‰X=OX(5βˆ’5)=OX\omega_X = \mathscr{O}_X(5-5) = \mathscr{O}_X.
  2. Complete intersection (3,3)(3,3) in P5\mathbb{P}^5: Ο‰X=OX(3+3βˆ’6)=OX\omega_X = \mathscr{O}_X(3+3-6) = \mathscr{O}_X.
  3. Complete intersection (2,4)(2,4) in P5\mathbb{P}^5: Ο‰X=OX(2+4βˆ’6)=OX\omega_X = \mathscr{O}_X(2+4-6) = \mathscr{O}_X.
  4. Complete intersection (2,2,3)(2,2,3) in P6\mathbb{P}^6: Ο‰X=OX(2+2+3βˆ’7)=OX\omega_X = \mathscr{O}_X(2+2+3-7) = \mathscr{O}_X.

The quintic threefold has Hodge numbers h1,1(X)=1h^{1,1}(X) = 1 and h2,1(X)=101h^{2,1}(X) = 101, leading to a moduli space of dimension 101.

The Canonical Ring

DefinitionCanonical Ring

For a smooth projective variety XX, the canonical ring is: R(X)=⨁mβ‰₯0H0(X,Ο‰XβŠ—m)=⨁mβ‰₯0H0(X,OX(mKX))R(X) = \bigoplus_{m \geq 0} H^0(X, \omega_X^{\otimes m}) = \bigoplus_{m \geq 0} H^0(X, \mathscr{O}_X(mK_X))

This is a graded kk-algebra with multiplication induced by tensor product.

TheoremFinite Generation of Canonical Ring

Let XX be a smooth projective variety. Then the canonical ring R(X)R(X) is finitely generated as a kk-algebra.

This fundamental result was proved by Birkar-Cascini-Hacon-McKernan (2006) as part of the minimal model program.

ExampleCanonical Ring of Curves

For a smooth projective curve CC of genus gβ‰₯2g \geq 2:

The canonical ring is: R(C)=⨁mβ‰₯0H0(C,Ο‰CβŠ—m)R(C) = \bigoplus_{m \geq 0} H^0(C, \omega_C^{\otimes m})

Structure:

  • R(C)0=kR(C)_0 = k
  • R(C)1=H0(C,Ο‰C)R(C)_1 = H^0(C, \omega_C) has dimension gg
  • For mβ‰₯2m \geq 2: dim⁑R(C)m=(2mβˆ’1)(gβˆ’1)\dim R(C)_m = (2m-1)(g-1) by Riemann-Roch

Generation: R(C)R(C) is generated in degrees 1 and 2. More precisely:

  • If CC is non-hyperelliptic: R(C)R(C) is generated in degree 1, i.e., by H0(C,Ο‰C)H^0(C, \omega_C)
  • If CC is hyperelliptic: R(C)R(C) requires generators in degree 2

Canonical embedding: The canonical map: Ο•K:Cβ†’P(H0(C,Ο‰C)∨)β‰…Pgβˆ’1\phi_K: C \to \mathbb{P}(H^0(C, \omega_C)^\vee) \cong \mathbb{P}^{g-1} embeds CC as a curve of degree 2gβˆ’22g-2 (unless CC is hyperelliptic, in which case it's a 2:1 cover of P1\mathbb{P}^1).

ExampleCanonical Ring of Surfaces of General Type

Let SS be a smooth projective surface of general type (i.e., ΞΊ(S)=2\kappa(S) = 2).

Finite generation: The canonical ring R(S)=⨁mβ‰₯0H0(S,Ο‰SβŠ—m)R(S) = \bigoplus_{m \geq 0} H^0(S, \omega_S^{\otimes m}) is finitely generated.

Canonical model: The canonical model is: Scan=Proj(R(S))S_{\text{can}} = \mathrm{Proj}(R(S)) This is a normal projective variety with at worst canonical singularities. If KSK_S is nef and big, then the canonical map: Ο•K:Sβ†’Scan\phi_K: S \to S_{\text{can}} contracts all (βˆ’2)(-2)-curves on SS.

Example - Surface of degree dβ‰₯5d \geq 5 in P3\mathbb{P}^3: For a smooth surface SβŠ†P3S \subseteq \mathbb{P}^3 of degree dβ‰₯5d \geq 5:

  • Ο‰S=OS(dβˆ’4)\omega_S = \mathscr{O}_S(d-4) is ample
  • Pm(S)=h0(S,Ο‰SβŠ—m)=h0(P3,O(m(dβˆ’4)))P_m(S) = h^0(S, \omega_S^{\otimes m}) = h^0(\mathbb{P}^3, \mathscr{O}(m(d-4))) for large mm
  • The canonical ring is generated by sections of Ο‰S\omega_S and Ο‰SβŠ—2\omega_S^{\otimes 2}

Functoriality Properties

TheoremFunctoriality of the Canonical Sheaf

Let f:X→Yf: X \to Y be a morphism of smooth varieties.

  1. Pullback: If ff is flat, there is a natural map: fβˆ—Ο‰Yβ†’Ο‰Xf^* \omega_Y \to \omega_X

  2. Ramification formula: If ff is a finite morphism of smooth varieties, then: Ο‰X=fβˆ—Ο‰YβŠ—OX(R)\omega_X = f^* \omega_Y \otimes \mathscr{O}_X(R) where RR is the ramification divisor.

  3. Relative canonical sheaf: For a smooth morphism f:Xβ†’Yf: X \to Y of relative dimension nn: Ο‰X/Y=β‹€nΞ©X/Y\omega_{X/Y} = \bigwedge^n \Omega_{X/Y} and Ο‰X=fβˆ—Ο‰YβŠ—Ο‰X/Y\omega_X = f^* \omega_Y \otimes \omega_{X/Y}.

ExampleRamification Formula

Let f:Cβ†’Df: C \to D be a finite morphism of smooth projective curves of degrees dd. Write: KC=fβˆ—KD+RK_C = f^* K_D + R where RR is the ramification divisor.

Formula: If ff has ramification index epe_p at p∈Cp \in C (meaning ff is locally t↦tept \mapsto t^{e_p}), then: R=βˆ‘p∈C(epβˆ’1)β‹…pR = \sum_{p \in C} (e_p - 1) \cdot p

Riemann-Hurwitz formula: Taking degrees: 2gCβˆ’2=d(2gDβˆ’2)+deg⁑(R)2g_C - 2 = d(2g_D - 2) + \deg(R)

Example - Double cover: If f:Cβ†’P1f: C \to \mathbb{P}^1 is a double cover branched over 2r2r points, then d=2d = 2 and deg⁑(R)=2r\deg(R) = 2r. Thus: 2gCβˆ’2=2(βˆ’2)+2rβ€…β€ŠβŸΉβ€…β€ŠgC=rβˆ’12g_C - 2 = 2(-2) + 2r \implies g_C = r - 1

For r=4r = 4: gC=3g_C = 3 (hyperelliptic curve of genus 3).

ExampleCanonical Sheaf of Products

Let XX and YY be smooth projective varieties with projections pX:XΓ—Yβ†’Xp_X: X \times Y \to X and pY:XΓ—Yβ†’Yp_Y: X \times Y \to Y. Then: Ο‰XΓ—Y=pXβˆ—Ο‰XβŠ—pYβˆ—Ο‰Y\omega_{X \times Y} = p_X^* \omega_X \otimes p_Y^* \omega_Y

Proof: Since Ξ©XΓ—Y/k=pXβˆ—Ξ©X/kβŠ•pYβˆ—Ξ©Y/k\Omega_{X \times Y/k} = p_X^* \Omega_{X/k} \oplus p_Y^* \Omega_{Y/k}, taking top exterior power: Ο‰XΓ—Y=β‹€n+mΞ©XΓ—Y/k=pXβˆ—Ο‰XβŠ—pYβˆ—Ο‰Y\omega_{X \times Y} = \bigwedge^{n+m} \Omega_{X \times Y/k} = p_X^* \omega_X \otimes p_Y^* \omega_Y

Example: For PnΓ—Pm\mathbb{P}^n \times \mathbb{P}^m: Ο‰PnΓ—Pm=O(βˆ’nβˆ’1,βˆ’mβˆ’1)\omega_{\mathbb{P}^n \times \mathbb{P}^m} = \mathscr{O}(-n-1, -m-1)

Canonical Class Computations

ExampleCanonical Class Under Blow-up

Let Ο€:X~β†’X\pi: \widetilde{X} \to X be the blow-up of a smooth variety XX at a smooth subvariety YY of codimension rr, with exceptional divisor Eβ‰…P(NY/X)E \cong \mathbb{P}(\mathscr{N}_{Y/X}). Then: KX~=Ο€βˆ—KX+(rβˆ’1)EK_{\widetilde{X}} = \pi^* K_X + (r-1) E

Proof: The conormal exact sequence gives: 0β†’OE(βˆ’E)β†’Ξ©X~/k∣Eβ†’Ξ©E/kβ†’00 \to \mathscr{O}_E(-E) \to \Omega_{\widetilde{X}/k}|_E \to \Omega_{E/k} \to 0

Computing determinants and using adjunction yields the formula.

Example - Blow-up at a point: If XX is a smooth surface and Ο€:X~β†’X\pi: \widetilde{X} \to X is the blow-up at a point pp with exceptional curve Eβ‰…P1E \cong \mathbb{P}^1: KX~=Ο€βˆ—KX+EK_{\widetilde{X}} = \pi^* K_X + E

Since E2=βˆ’1E^2 = -1, we have: KX~β‹…E=(Ο€βˆ—KX+E)β‹…E=0+(βˆ’1)=βˆ’1K_{\widetilde{X}} \cdot E = (\pi^* K_X + E) \cdot E = 0 + (-1) = -1

ExampleCanonical Divisor of Toric Varieties

Let XX be a smooth complete toric variety corresponding to a fan Ξ£\Sigma in NRN_\mathbb{R} where Nβ‰…ZnN \cong \mathbb{Z}^n. For each ray ρ∈Σ(1)\rho \in \Sigma(1) with primitive generator vρ∈Nv_\rho \in N, let DρD_\rho be the corresponding toric divisor.

The canonical divisor is: KX=βˆ’βˆ‘ΟβˆˆΞ£(1)DρK_X = -\sum_{\rho \in \Sigma(1)} D_\rho

Example - Pn\mathbb{P}^n: The fan has n+1n+1 rays generating the standard basis plus βˆ’(e1+β‹―+en)-(e_1 + \cdots + e_n). Thus: KPn=βˆ’(D0+β‹―+Dn)=βˆ’(n+1)HK_{\mathbb{P}^n} = -(D_0 + \cdots + D_n) = -(n+1)H where HH is the hyperplane class, confirming Ο‰Pn=OPn(βˆ’nβˆ’1)\omega_{\mathbb{P}^n} = \mathscr{O}_{\mathbb{P}^n}(-n-1).

Example - Hirzebruch surface Fa\mathbb{F}_a: This is a P1\mathbb{P}^1-bundle over P1\mathbb{P}^1. The canonical divisor is: KFa=βˆ’2C0βˆ’(a+2)fK_{\mathbb{F}_a} = -2C_0 - (a+2)f where C0C_0 is the zero section and ff is the fiber class.

ExampleCanonical Sheaf of Abelian Varieties

Let AA be an abelian variety of dimension gg over kk. Then: ωA≅OA\omega_A \cong \mathscr{O}_A

Proof: Since AA is a group variety, for any point a∈Aa \in A the translation map ta:Aβ†’At_a: A \to A is an automorphism. Thus: taβˆ—Ο‰Aβ‰…Ο‰At_a^* \omega_A \cong \omega_A

But taβˆ—t_a^* acts as identity on H0(A,Ο‰A)H^0(A, \omega_A), and the space of translation-invariant sections of any line bundle on AA is at most 1-dimensional. Since h0(A,Ο‰A)=1h^0(A, \omega_A) = 1 (by Serre duality and hg(A,OA)=1h^g(A, \mathscr{O}_A) = 1), we have Ο‰Aβ‰…OA\omega_A \cong \mathscr{O}_A.

Consequence: Abelian varieties have Kodaira dimension 0, and all plurigenera are Pm(A)=1P_m(A) = 1 for m>0m > 0.

Relationship with Birational Geometry

Remark

The canonical sheaf plays a central role in birational classification:

  1. Minimal Model Program: Seeks to find a "best" birational model XX where KXK_X is "as nef as possible"

  2. Mori cone: The cone of effective curves NEβ€Ύ(X)βŠ†N1(X)\overline{NE}(X) \subseteq N_1(X) is related to KXK_X via contraction morphisms

  3. Abundance conjecture: If XX is a minimal model (i.e., KXK_X nef), then KXK_X is semiample

  4. Canonical singularities: Allow singularities with "well-behaved" canonical class

These ideas lead to a complete classification in dimension 2 (Enriques-Kodaira classification) and substantial progress in dimension 3.

ExampleEnriques-Kodaira Classification

Smooth projective surfaces are classified by Kodaira dimension:

ΞΊ=βˆ’βˆž\kappa = -\infty:

  • Rational surfaces (P2\mathbb{P}^2, rational ruled surfaces)
  • Ruled surfaces over curves of genus gβ‰₯1g \geq 1

ΞΊ=0\kappa = 0:

  • K3 surfaces (Ο‰Sβ‰…OS\omega_S \cong \mathscr{O}_S, q=0q = 0)
  • Abelian surfaces
  • Enriques surfaces (2KS∼02K_S \sim 0, q=0q = 0)
  • Hyperelliptic surfaces (quotients of abelian surfaces)

ΞΊ=1\kappa = 1:

  • Properly elliptic surfaces (elliptic fibrations with Ο‡(OS)=0\chi(\mathscr{O}_S) = 0)

ΞΊ=2\kappa = 2:

  • Surfaces of general type (canonical model exists)

The minimal models in each class are characterized by KSK_S being nef.


The dualizing sheaf provides a bridge between differential geometry, algebraic geometry, and arithmetic geometry. Its study leads to deep connections with moduli theory, mirror symmetry, and string theory. Understanding Ο‰X\omega_X and its powers is essential for the classification of algebraic varieties and for computing fundamental invariants like the geometric genus, irregularity, and Hodge numbers.