TheoremComplete

Cohomology of Projective Space

The cohomology groups of twisted structure sheaves on projective space are among the most fundamental computations in algebraic geometry. These groups control the geometry of projective varieties and provide the foundation for Serre duality, Riemann-Roch, and many other essential theorems.

The Main Theorem

Theorem

Let kk be a field and Pn=Pkn\mathbb{P}^n = \mathbb{P}^n_k be projective nn-space over kk. For any integer dd, we have:

  • H0(Pn,OPn(d))=k(n+dn)H^0(\mathbb{P}^n, \mathcal{O}_{\mathbb{P}^n}(d)) = k^{\binom{n+d}{n}} for dβ‰₯0d \geq 0, and 00 for d<0d < 0
  • Hn(Pn,OPn(d))=k(∣dβˆ£βˆ’1n)H^n(\mathbb{P}^n, \mathcal{O}_{\mathbb{P}^n}(d)) = k^{\binom{|d|-1}{n}} for dβ‰€βˆ’nβˆ’1d \leq -n-1, and 00 for d>βˆ’nβˆ’1d > -n-1
  • Hi(Pn,OPn(d))=0H^i(\mathbb{P}^n, \mathcal{O}_{\mathbb{P}^n}(d)) = 0 for 0<i<n0 < i < n and all dd

In particular:

  • h0(Pn,O(d))=(n+dn)h^0(\mathbb{P}^n, \mathcal{O}(d)) = \binom{n+d}{n} for dβ‰₯0d \geq 0 and 00 for d<0d < 0
  • hi(Pn,O(d))=0h^i(\mathbb{P}^n, \mathcal{O}(d)) = 0 for 0<i<n0 < i < n and all dd
  • hn(Pn,O(d))=(∣dβˆ£βˆ’1n)h^n(\mathbb{P}^n, \mathcal{O}(d)) = \binom{|d|-1}{n} for dβ‰€βˆ’nβˆ’1d \leq -n-1 and 00 for d>βˆ’nβˆ’1d > -n-1
Remark

The "gap" phenomenon is striking: except at the extremes (i=0i=0 and i=ni=n), all cohomology groups vanish. This is a special feature of projective space not shared by general projective varieties.

Proof via Čech Cohomology

Proof

We compute using Čech cohomology with respect to the standard affine cover. Let Ui=D+(Xi)U_i = D_+(X_i) for i=0,…,ni = 0, \ldots, n be the standard affine open cover of Pn\mathbb{P}^n.

Step 1: Structure of the cover. Each Uiβ‰…AnU_i \cong \mathbb{A}^n with coordinate ring k[X0/Xi,…,Xi/Xi^,…,Xn/Xi]k[X_0/X_i, \ldots, \widehat{X_i/X_i}, \ldots, X_n/X_i]. The intersections are: Ui0βˆ©β‹―βˆ©ip=D+(Xi0β‹―Xip)β‰…Anβˆ–{Xj1β‹―Xnβˆ’p=0}U_{i_0 \cap \cdots \cap i_p} = D_+(X_{i_0} \cdots X_{i_p}) \cong \mathbb{A}^n \setminus \{X_{j_1} \cdots X_{n-p} = 0\} where {j1,…,jnβˆ’p}={0,…,n}βˆ–{i0,…,ip}\{j_1, \ldots, j_{n-p}\} = \{0, \ldots, n\} \setminus \{i_0, \ldots, i_p\}.

Step 2: Sections of O(d)\mathcal{O}(d) on opens. For O(d)\mathcal{O}(d), sections over UiU_i are: Ξ“(Ui,O(d))={fXid:f∈k[X0,…,Xn]Β homogeneousΒ ofΒ degreeΒ d}\Gamma(U_i, \mathcal{O}(d)) = \left\{ \frac{f}{X_i^d} : f \in k[X_0, \ldots, X_n] \text{ homogeneous of degree } d \right\}

For dβ‰₯0d \geq 0, this is the localization of degree dd homogeneous polynomials. For d<0d < 0, we need Laurent polynomials.

Step 3: Čech complex for dβ‰₯0d \geq 0. The Čech complex is: ∏iΞ“(Ui,O(d))β†’Ξ΄0∏i<jΞ“(Ui∩Uj,O(d))β†’Ξ΄1β‹―\prod_{i} \Gamma(U_i, \mathcal{O}(d)) \xrightarrow{\delta^0} \prod_{i<j} \Gamma(U_i \cap U_j, \mathcal{O}(d)) \xrightarrow{\delta^1} \cdots

For dβ‰₯0d \geq 0, a global section is determined by homogeneous polynomials fif_i of degree dd that agree on overlaps. Since Ui∩UjU_i \cap U_j is connected, the fif_i must glue to give a global homogeneous polynomial of degree dd.

Thus H0(Pn,O(d))=k[X0,…,Xn]dH^0(\mathbb{P}^n, \mathcal{O}(d)) = k[X_0, \ldots, X_n]_d has dimension (n+dn)\binom{n+d}{n}.

Step 4: Higher cohomology for dβ‰₯0d \geq 0. For pβ‰₯1p \geq 1, the Čech groups involve sections over intersections of (p+1)(p+1) opens. These are Laurent polynomial rings, but for dβ‰₯0d \geq 0, the cochains are determined by polynomial data. The complex is exact at each stage by the following key fact: on the punctured affine space Anβˆ–{f=0}\mathbb{A}^n \setminus \{f = 0\}, the ring k[x1,…,xn,fβˆ’1]k[x_1, \ldots, x_n, f^{-1}] has trivial higher cohomology for coherent sheaves.

By induction on pp and using the acyclicity of affine schemes, we conclude Hi(Pn,O(d))=0H^i(\mathbb{P}^n, \mathcal{O}(d)) = 0 for i>0i > 0 and dβ‰₯0d \geq 0.

Step 5: Using Serre duality. By Serre duality, Hi(Pn,O(d))β‰…Hnβˆ’i(Pn,O(βˆ’dβˆ’nβˆ’1))∨H^i(\mathbb{P}^n, \mathcal{O}(d)) \cong H^{n-i}(\mathbb{P}^n, \mathcal{O}(-d-n-1))^\vee.

For d<0d < 0, write d=βˆ’md = -m with m>0m > 0. Then: Hn(Pn,O(βˆ’m))β‰…H0(Pn,O(mβˆ’nβˆ’1))∨H^n(\mathbb{P}^n, \mathcal{O}(-m)) \cong H^0(\mathbb{P}^n, \mathcal{O}(m-n-1))^\vee

This is nonzero only when mβˆ’nβˆ’1β‰₯0m-n-1 \geq 0, i.e., mβ‰₯n+1m \geq n+1, giving dβ‰€βˆ’nβˆ’1d \leq -n-1. In this case: hn(Pn,O(βˆ’m))=h0(Pn,O(mβˆ’nβˆ’1))=(n+mβˆ’nβˆ’1n)=(mβˆ’1n)h^n(\mathbb{P}^n, \mathcal{O}(-m)) = h^0(\mathbb{P}^n, \mathcal{O}(m-n-1)) = \binom{n+m-n-1}{n} = \binom{m-1}{n}

Step 6: The middle range. For βˆ’n≀d<0-n \leq d < 0, we have d+n+1∈[1,n]d+n+1 \in [1, n], so 0≀d+n+1βˆ’1<n0 \leq d+n+1-1 < n. By step 4, H0(Pn,O(d+n+1βˆ’1))H^0(\mathbb{P}^n, \mathcal{O}(d+n+1-1)) might be nonzero, but we need d+n+1βˆ’1β‰₯0d+n+1-1 \geq 0 for this, i.e., dβ‰₯βˆ’nd \geq -n. However, for βˆ’n≀d<0-n \leq d < 0, we have 0≀d+n<n0 \leq d+n < n, and specifically:

  • If d=βˆ’nd = -n, then H0(Pn,O(βˆ’1))=0H^0(\mathbb{P}^n, \mathcal{O}(-1)) = 0, so Hn(Pn,O(βˆ’n))=0H^n(\mathbb{P}^n, \mathcal{O}(-n)) = 0.
  • For βˆ’n<d<0-n < d < 0, similar analysis shows vanishing.

Direct computation using the Čech complex confirms Hi(Pn,O(d))=0H^i(\mathbb{P}^n, \mathcal{O}(d)) = 0 for 0<i<n0 < i < n and all dd.

β– 

Complete Tables of Cohomology

ExampleComplete cohomology table for $\mathbb{P}^1$

For projective line P1\mathbb{P}^1:

| dd | h0(P1,O(d))h^0(\mathbb{P}^1, \mathcal{O}(d)) | h1(P1,O(d))h^1(\mathbb{P}^1, \mathcal{O}(d)) | |-----|--------------------------------------|--------------------------------------| | β‹―\cdots | 00 | β‹―\cdots | | βˆ’4-4 | 00 | 33 | | βˆ’3-3 | 00 | 22 | | βˆ’2-2 | 00 | 11 | | βˆ’1-1 | 00 | 00 | | 00 | 11 | 00 | | 11 | 22 | 00 | | 22 | 33 | 00 | | 33 | 44 | 00 | | 44 | 55 | 00 | | β‹―\cdots | β‹―\cdots | 00 |

Note the symmetry: h1(P1,O(βˆ’dβˆ’2))=h0(P1,O(d))h^1(\mathbb{P}^1, \mathcal{O}(-d-2)) = h^0(\mathbb{P}^1, \mathcal{O}(d)) for all dd.

ExampleComplete cohomology table for $\mathbb{P}^2$

For projective plane P2\mathbb{P}^2:

| dd | h0(P2,O(d))h^0(\mathbb{P}^2, \mathcal{O}(d)) | h1(P2,O(d))h^1(\mathbb{P}^2, \mathcal{O}(d)) | h2(P2,O(d))h^2(\mathbb{P}^2, \mathcal{O}(d)) | |-----|--------------------------------------|--------------------------------------|--------------------------------------| | β‹―\cdots | 00 | 00 | β‹―\cdots | | βˆ’5-5 | 00 | 00 | 66 | | βˆ’4-4 | 00 | 00 | 33 | | βˆ’3-3 | 00 | 00 | 11 | | βˆ’2-2 | 00 | 00 | 00 | | βˆ’1-1 | 00 | 00 | 00 | | 00 | 11 | 00 | 00 | | 11 | 33 | 00 | 00 | | 22 | 66 | 00 | 00 | | 33 | 1010 | 00 | 00 | | 44 | 1515 | 00 | 00 | | β‹―\cdots | β‹―\cdots | 00 | 00 |

The middle cohomology H1H^1 vanishes identically. The symmetry is: h2(P2,O(βˆ’dβˆ’3))=h0(P2,O(d))h^2(\mathbb{P}^2, \mathcal{O}(-d-3)) = h^0(\mathbb{P}^2, \mathcal{O}(d)).

ExampleCohomology table for $\mathbb{P}^3$

For projective 3-space P3\mathbb{P}^3:

| dd | h0h^0 | h1h^1 | h2h^2 | h3h^3 | |-----|-------|-------|-------|-------| | βˆ’6-6 | 00 | 00 | 00 | 1010 | | βˆ’5-5 | 00 | 00 | 00 | 44 | | βˆ’4-4 | 00 | 00 | 00 | 11 | | βˆ’3-3 | 00 | 00 | 00 | 00 | | βˆ’2-2 | 00 | 00 | 00 | 00 | | βˆ’1-1 | 00 | 00 | 00 | 00 | | 00 | 11 | 00 | 00 | 00 | | 11 | 44 | 00 | 00 | 00 | | 22 | 1010 | 00 | 00 | 00 | | 33 | 2020 | 00 | 00 | 00 | | 44 | 3535 | 00 | 00 | 00 |

Both H1H^1 and H2H^2 vanish identically, illustrating the gap phenomenon.

Detailed Computations

Example$H^0(\mathbb{P}^n, \mathcal{O}(d))$ for $d \geq 0$

Global sections of O(d)\mathcal{O}(d) correspond to homogeneous polynomials of degree dd in n+1n+1 variables X0,…,XnX_0, \ldots, X_n. The dimension is: h0(Pn,O(d))=(n+dn)=(n+dd)h^0(\mathbb{P}^n, \mathcal{O}(d)) = \binom{n+d}{n} = \binom{n+d}{d}

Explicit examples:

  • n=2,d=2n=2, d=2: h0=(42)=6h^0 = \binom{4}{2} = 6. Basis: {X02,X0X1,X0X2,X12,X1X2,X22}\{X_0^2, X_0X_1, X_0X_2, X_1^2, X_1X_2, X_2^2\}.
  • n=3,d=1n=3, d=1: h0=(41)=4h^0 = \binom{4}{1} = 4. Basis: {X0,X1,X2,X3}\{X_0, X_1, X_2, X_3\}.
  • n=1,d=3n=1, d=3: h0=(41)=4h^0 = \binom{4}{1} = 4. Basis: {X03,X02X1,X0X12,X13}\{X_0^3, X_0^2X_1, X_0X_1^2, X_1^3\}.

For d<0d < 0, there are no nonzero global sections since a homogeneous polynomial of negative degree in positive degree variables is trivial.

Example$H^0(\mathbb{P}^n, \mathcal{O}(d))$ for $d < 0$

For d<0d < 0, a global section of O(d)\mathcal{O}(d) would be given by a collection of Laurent polynomials on each affine chart Ui=D+(Xi)U_i = D_+(X_i) that glue compatibly. On UiU_i, sections are of the form f/Xidf/X_i^d where ff is homogeneous of degree dd.

For d<0d < 0, such a section must have poles along the hyperplane {Xi=0}\{X_i = 0\} in each chart. However, since Pn\mathbb{P}^n is covered by the UiU_i and the complement of UiU_i is the hyperplane {Xi=0}\{X_i = 0\}, a global section cannot have poles anywhere.

Therefore, H0(Pn,O(d))=0H^0(\mathbb{P}^n, \mathcal{O}(d)) = 0 for d<0d < 0.

Example$H^n(\mathbb{P}^n, \mathcal{O}(d))$ for $d \leq -n-1$

By Serre duality: Hn(Pn,O(d))β‰…H0(Pn,O(βˆ’dβˆ’nβˆ’1))∨H^n(\mathbb{P}^n, \mathcal{O}(d)) \cong H^0(\mathbb{P}^n, \mathcal{O}(-d-n-1))^\vee

For dβ‰€βˆ’nβˆ’1d \leq -n-1, we have βˆ’dβˆ’nβˆ’1β‰₯0-d-n-1 \geq 0, so: hn(Pn,O(d))=h0(Pn,O(βˆ’dβˆ’nβˆ’1))=(n+(βˆ’dβˆ’nβˆ’1)n)=(βˆ’dβˆ’1n)h^n(\mathbb{P}^n, \mathcal{O}(d)) = h^0(\mathbb{P}^n, \mathcal{O}(-d-n-1)) = \binom{n+(-d-n-1)}{n} = \binom{-d-1}{n}

Concrete cases:

  • n=1,d=βˆ’3n=1, d=-3: h1(P1,O(βˆ’3))=(3βˆ’11)=(21)=2h^1(\mathbb{P}^1, \mathcal{O}(-3)) = \binom{3-1}{1} = \binom{2}{1} = 2.
  • n=2,d=βˆ’4n=2, d=-4: h2(P2,O(βˆ’4))=(4βˆ’12)=(32)=3h^2(\mathbb{P}^2, \mathcal{O}(-4)) = \binom{4-1}{2} = \binom{3}{2} = 3.
  • n=3,d=βˆ’5n=3, d=-5: h3(P3,O(βˆ’5))=(5βˆ’13)=(43)=4h^3(\mathbb{P}^3, \mathcal{O}(-5)) = \binom{5-1}{3} = \binom{4}{3} = 4.
ExampleThe gap phenomenon

Consider P3\mathbb{P}^3 and compute H1(P3,O(d))H^1(\mathbb{P}^3, \mathcal{O}(d)) for various dd:

For any dd:

  • If dβ‰₯0d \geq 0: By the main theorem, H1(P3,O(d))=0H^1(\mathbb{P}^3, \mathcal{O}(d)) = 0.
  • If d<0d < 0: By Serre duality, H1(P3,O(d))β‰…H2(P3,O(βˆ’dβˆ’4))∨H^1(\mathbb{P}^3, \mathcal{O}(d)) \cong H^2(\mathbb{P}^3, \mathcal{O}(-d-4))^\vee.
    • For βˆ’dβˆ’4β‰₯0-d-4 \geq 0, i.e., dβ‰€βˆ’4d \leq -4, we would need H2(P3,O(βˆ’dβˆ’4))β‰ 0H^2(\mathbb{P}^3, \mathcal{O}(-d-4)) \neq 0.
    • But H2(P3,O(m))=0H^2(\mathbb{P}^3, \mathcal{O}(m)) = 0 for all mm by the main theorem.

Therefore, H1(P3,O(d))=0H^1(\mathbb{P}^3, \mathcal{O}(d)) = 0 for all d∈Zd \in \mathbb{Z}.

Similarly, H2(P3,O(d))=0H^2(\mathbb{P}^3, \mathcal{O}(d)) = 0 for all dd. The only nonzero cohomology occurs at i=0i=0 (for dβ‰₯0d \geq 0) and i=3i=3 (for dβ‰€βˆ’4d \leq -4).

Čech Cohomology Computation in Detail

ExampleČech computation for $H^1(\mathbb{P}^1, \mathcal{O}(-2))$

Consider the standard cover U0=D+(X0)U_0 = D_+(X_0) and U1=D+(X1)U_1 = D_+(X_1) of P1\mathbb{P}^1.

The Čech complex is: Ξ“(U0,O(βˆ’2))βŠ•Ξ“(U1,O(βˆ’2))→δΓ(U0∩U1,O(βˆ’2))\Gamma(U_0, \mathcal{O}(-2)) \oplus \Gamma(U_1, \mathcal{O}(-2)) \xrightarrow{\delta} \Gamma(U_0 \cap U_1, \mathcal{O}(-2))

We have:

  • Ξ“(U0,O(βˆ’2))=k[X1/X0]β‹…X02\Gamma(U_0, \mathcal{O}(-2)) = k[X_1/X_0] \cdot X_0^2 (Laurent monomials X1j/X0j+2X_1^j/X_0^{j+2})
  • Ξ“(U1,O(βˆ’2))=k[X0/X1]β‹…X12\Gamma(U_1, \mathcal{O}(-2)) = k[X_0/X_1] \cdot X_1^2 (Laurent monomials X0i/X1i+2X_0^i/X_1^{i+2})
  • Ξ“(U0∩U1,O(βˆ’2))=k[X0/X1,X1/X0]\Gamma(U_0 \cap U_1, \mathcal{O}(-2)) = k[X_0/X_1, X_1/X_0] (all Laurent monomials)

The coboundary map Ξ΄(s0,s1)=s0∣U0∩U1βˆ’s1∣U0∩U1\delta(s_0, s_1) = s_0|_{U_0 \cap U_1} - s_1|_{U_0 \cap U_1}.

A 1-cocycle is an element Ο•βˆˆΞ“(U0∩U1,O(βˆ’2))\phi \in \Gamma(U_0 \cap U_1, \mathcal{O}(-2)) such that Ο•\phi is not in the image of Ξ΄\delta.

The image of Ξ΄\delta consists of Laurent polynomials that can be written as a difference of a polynomial in X1/X0X_1/X_0 and a polynomial in X0/X1X_0/X_1 (after multiplying by X02X_0^2 or X12X_1^2).

Consider Ο•=1/(X0X1)\phi = 1/(X_0 X_1). In the U0U_0 chart, this is 1/(X02β‹…(X1/X0))=1/X02β‹…X0/X11/(X_0^2 \cdot (X_1/X_0)) = 1/X_0^2 \cdot X_0/X_1, which has a pole at X0=0X_0 = 0. Similarly, it has a pole at X1=0X_1 = 0 in the U1U_1 chart. Therefore, Ο•\phi cannot be written as Ξ΄(s0,s1)\delta(s_0, s_1).

Thus H1(P1,O(βˆ’2))β‰…kH^1(\mathbb{P}^1, \mathcal{O}(-2)) \cong k is generated by the class of 1/(X0X1)1/(X_0 X_1).

ExampleČech complex structure for $\mathbb{P}^2$

For P2\mathbb{P}^2 with cover U0,U1,U2U_0, U_1, U_2, the Čech complex for O(d)\mathcal{O}(d) is: C0β†’Ξ΄0C1β†’Ξ΄1C2C^0 \xrightarrow{\delta^0} C^1 \xrightarrow{\delta^1} C^2 where:

  • C0=Ξ“(U0,O(d))βŠ•Ξ“(U1,O(d))βŠ•Ξ“(U2,O(d))C^0 = \Gamma(U_0, \mathcal{O}(d)) \oplus \Gamma(U_1, \mathcal{O}(d)) \oplus \Gamma(U_2, \mathcal{O}(d))
  • C1=Ξ“(U0∩U1,O(d))βŠ•Ξ“(U0∩U2,O(d))βŠ•Ξ“(U1∩U2,O(d))C^1 = \Gamma(U_0 \cap U_1, \mathcal{O}(d)) \oplus \Gamma(U_0 \cap U_2, \mathcal{O}(d)) \oplus \Gamma(U_1 \cap U_2, \mathcal{O}(d))
  • C2=Ξ“(U0∩U1∩U2,O(d))C^2 = \Gamma(U_0 \cap U_1 \cap U_2, \mathcal{O}(d))

For d=βˆ’3d = -3, we compute H2(P2,O(βˆ’3))H^2(\mathbb{P}^2, \mathcal{O}(-3)):

  • C2=k[X0/X1,X0/X2,X1/X0,X1/X2,X2/X0,X2/X1]C^2 = k[X_0/X_1, X_0/X_2, X_1/X_0, X_1/X_2, X_2/X_0, X_2/X_1] (Laurent polynomials)
  • Elements of degree βˆ’3-3 in C2C^2 that are not in im(Ξ΄1)\text{im}(\delta^1) form H2H^2.

The element 1/(X0X1X2)1/(X_0 X_1 X_2) has poles everywhere and cannot be a coboundary. It generates H2(P2,O(βˆ’3))β‰…kH^2(\mathbb{P}^2, \mathcal{O}(-3)) \cong k.

Serre Duality Reflected in the Tables

ExampleSerre duality for $\mathbb{P}^1$

The canonical sheaf of P1\mathbb{P}^1 is Ο‰P1=O(βˆ’2)\omega_{\mathbb{P}^1} = \mathcal{O}(-2). Serre duality states: H1(P1,O(d))β‰…H0(P1,O(βˆ’dβˆ’2))∨H^1(\mathbb{P}^1, \mathcal{O}(d)) \cong H^0(\mathbb{P}^1, \mathcal{O}(-d-2))^\vee

Verifying from the table:

  • h1(P1,O(βˆ’3))=2=h0(P1,O(1))h^1(\mathbb{P}^1, \mathcal{O}(-3)) = 2 = h^0(\mathbb{P}^1, \mathcal{O}(1))
  • h1(P1,O(βˆ’2))=1=h0(P1,O(0))h^1(\mathbb{P}^1, \mathcal{O}(-2)) = 1 = h^0(\mathbb{P}^1, \mathcal{O}(0))
  • h1(P1,O(βˆ’1))=0=h0(P1,O(βˆ’1))h^1(\mathbb{P}^1, \mathcal{O}(-1)) = 0 = h^0(\mathbb{P}^1, \mathcal{O}(-1))
  • h1(P1,O(0))=0=h0(P1,O(βˆ’2))h^1(\mathbb{P}^1, \mathcal{O}(0)) = 0 = h^0(\mathbb{P}^1, \mathcal{O}(-2))
  • h1(P1,O(1))=0=h0(P1,O(βˆ’3))h^1(\mathbb{P}^1, \mathcal{O}(1)) = 0 = h^0(\mathbb{P}^1, \mathcal{O}(-3))

The duality is perfect.

ExampleSerre duality for $\mathbb{P}^2$

The canonical sheaf of P2\mathbb{P}^2 is Ο‰P2=O(βˆ’3)\omega_{\mathbb{P}^2} = \mathcal{O}(-3). Serre duality states: H2(P2,O(d))β‰…H0(P2,O(βˆ’dβˆ’3))∨H^2(\mathbb{P}^2, \mathcal{O}(d)) \cong H^0(\mathbb{P}^2, \mathcal{O}(-d-3))^\vee

Verifying:

  • h2(P2,O(βˆ’5))=6=h0(P2,O(2))h^2(\mathbb{P}^2, \mathcal{O}(-5)) = 6 = h^0(\mathbb{P}^2, \mathcal{O}(2))
  • h2(P2,O(βˆ’4))=3=h0(P2,O(1))h^2(\mathbb{P}^2, \mathcal{O}(-4)) = 3 = h^0(\mathbb{P}^2, \mathcal{O}(1))
  • h2(P2,O(βˆ’3))=1=h0(P2,O(0))h^2(\mathbb{P}^2, \mathcal{O}(-3)) = 1 = h^0(\mathbb{P}^2, \mathcal{O}(0))
  • h2(P2,O(βˆ’2))=0=h0(P2,O(βˆ’1))h^2(\mathbb{P}^2, \mathcal{O}(-2)) = 0 = h^0(\mathbb{P}^2, \mathcal{O}(-1))

The middle cohomology satisfies: H1(P2,O(d))β‰…H1(P2,O(βˆ’dβˆ’3))∨H^1(\mathbb{P}^2, \mathcal{O}(d)) \cong H^1(\mathbb{P}^2, \mathcal{O}(-d-3))^\vee Since both sides vanish for all dd, the duality is trivially satisfied.

Applications to Hilbert Polynomials

ExampleHilbert polynomial from cohomology

The Hilbert polynomial of Pn\mathbb{P}^n is defined as: P(d)=Ο‡(OPn(d))=βˆ‘i=0n(βˆ’1)ihi(Pn,O(d))P(d) = \chi(\mathcal{O}_{\mathbb{P}^n}(d)) = \sum_{i=0}^n (-1)^i h^i(\mathbb{P}^n, \mathcal{O}(d))

For dβ‰₯0d \geq 0: P(d)=h0(Pn,O(d))=(n+dn)P(d) = h^0(\mathbb{P}^n, \mathcal{O}(d)) = \binom{n+d}{n}

This is a polynomial in dd of degree nn with leading coefficient 1/n!1/n!.

For P1\mathbb{P}^1: P(d)=d+1P(d) = d + 1 For P2\mathbb{P}^2: P(d)=12(d+1)(d+2)=d2+3d+22P(d) = \frac{1}{2}(d+1)(d+2) = \frac{d^2 + 3d + 2}{2} For P3\mathbb{P}^3: P(d)=16(d+1)(d+2)(d+3)P(d) = \frac{1}{6}(d+1)(d+2)(d+3)

For negative dd:

  • If βˆ’n≀d<0-n \leq d < 0: Ο‡(O(d))=0\chi(\mathcal{O}(d)) = 0 (all terms vanish)
  • If d<βˆ’nd < -n: Ο‡(O(d))=(βˆ’1)n(βˆ’dβˆ’1n)\chi(\mathcal{O}(d)) = (-1)^n \binom{-d-1}{n}

For P1\mathbb{P}^1 and d=βˆ’2d = -2: Ο‡(O(βˆ’2))=h0βˆ’h1=0βˆ’1=βˆ’1=(βˆ’1)1(2βˆ’11)=βˆ’1\chi(\mathcal{O}(-2)) = h^0 - h^1 = 0 - 1 = -1 = (-1)^1 \binom{2-1}{1} = -1

ExampleHilbert series of the homogeneous coordinate ring

The Hilbert series of S=k[X0,…,Xn]S = k[X_0, \ldots, X_n] is: H(t)=βˆ‘d=0∞h0(Pn,O(d))td=βˆ‘d=0∞(n+dn)td=1(1βˆ’t)n+1H(t) = \sum_{d=0}^\infty h^0(\mathbb{P}^n, \mathcal{O}(d)) t^d = \sum_{d=0}^\infty \binom{n+d}{n} t^d = \frac{1}{(1-t)^{n+1}}

For P1\mathbb{P}^1: H(t)=1(1βˆ’t)2=1+2t+3t2+4t3+β‹―H(t) = \frac{1}{(1-t)^2} = 1 + 2t + 3t^2 + 4t^3 + \cdots For P2\mathbb{P}^2: H(t)=1(1βˆ’t)3=1+3t+6t2+10t3+β‹―H(t) = \frac{1}{(1-t)^3} = 1 + 3t + 6t^2 + 10t^3 + \cdots

The coefficients match the dimensions from the tables.

Cohomology of the Euler Sequence

Definition

The Euler sequence on Pn\mathbb{P}^n is the exact sequence: 0β†’OPnβ†’OPn(1)βŠ•(n+1)β†’TPnβ†’00 \to \mathcal{O}_{\mathbb{P}^n} \to \mathcal{O}_{\mathbb{P}^n}(1)^{\oplus (n+1)} \to T_{\mathbb{P}^n} \to 0 where TPnT_{\mathbb{P}^n} is the tangent sheaf of Pn\mathbb{P}^n.

ExampleCohomology from the Euler sequence

Taking the long exact sequence in cohomology and using Hi(Pn,O(1))=0H^i(\mathbb{P}^n, \mathcal{O}(1)) = 0 for i>0i > 0:

For P2\mathbb{P}^2, the Euler sequence gives: 0β†’H0(P2,O)β†’H0(P2,O(1))βŠ•3β†’H0(P2,TP2)β†’0 \to H^0(\mathbb{P}^2, \mathcal{O}) \to H^0(\mathbb{P}^2, \mathcal{O}(1))^{\oplus 3} \to H^0(\mathbb{P}^2, T_{\mathbb{P}^2}) \to β†’H1(P2,O)β†’H1(P2,O(1))βŠ•3β†’H1(P2,TP2)β†’β‹―\to H^1(\mathbb{P}^2, \mathcal{O}) \to H^1(\mathbb{P}^2, \mathcal{O}(1))^{\oplus 3} \to H^1(\mathbb{P}^2, T_{\mathbb{P}^2}) \to \cdots

Since H1(P2,O)=0H^1(\mathbb{P}^2, \mathcal{O}) = 0 and H1(P2,O(1))=0H^1(\mathbb{P}^2, \mathcal{O}(1)) = 0: H1(P2,TP2)=0H^1(\mathbb{P}^2, T_{\mathbb{P}^2}) = 0

Computing dimensions:

  • h0(P2,O)=1h^0(\mathbb{P}^2, \mathcal{O}) = 1
  • h0(P2,O(1))=3h^0(\mathbb{P}^2, \mathcal{O}(1)) = 3
  • Thus h0(P2,TP2)=3β‹…3βˆ’1=8h^0(\mathbb{P}^2, T_{\mathbb{P}^2}) = 3 \cdot 3 - 1 = 8

These are the infinitesimal automorphisms of P2\mathbb{P}^2, corresponding to the 88-dimensional group PGL3(k)\text{PGL}_3(k).

ExampleCohomology of $T_{\mathbb{P}^n}(d)$

Twisting the Euler sequence by O(d)\mathcal{O}(d) gives: 0β†’O(d)β†’O(d+1)βŠ•(n+1)β†’TPn(d)β†’00 \to \mathcal{O}(d) \to \mathcal{O}(d+1)^{\oplus (n+1)} \to T_{\mathbb{P}^n}(d) \to 0

For P1\mathbb{P}^1 and d=βˆ’1d = -1: 0β†’O(βˆ’1)β†’OβŠ•2β†’TP1(βˆ’1)β†’00 \to \mathcal{O}(-1) \to \mathcal{O}^{\oplus 2} \to T_{\mathbb{P}^1}(-1) \to 0

Taking cohomology: 0β†’H0(O(βˆ’1))β†’H0(O)βŠ•2β†’H0(TP1(βˆ’1))β†’H1(O(βˆ’1))β†’H1(O)βŠ•20 \to H^0(\mathcal{O}(-1)) \to H^0(\mathcal{O})^{\oplus 2} \to H^0(T_{\mathbb{P}^1}(-1)) \to H^1(\mathcal{O}(-1)) \to H^1(\mathcal{O})^{\oplus 2}

Since h0(O(βˆ’1))=0h^0(\mathcal{O}(-1)) = 0, h0(O)=1h^0(\mathcal{O}) = 1, and h1(O(βˆ’1))=0h^1(\mathcal{O}(-1)) = 0, h1(O)=0h^1(\mathcal{O}) = 0: 0β†’kβŠ•2β†’H0(TP1(βˆ’1))β†’00 \to k^{\oplus 2} \to H^0(T_{\mathbb{P}^1}(-1)) \to 0

Thus h0(TP1(βˆ’1))=2h^0(T_{\mathbb{P}^1}(-1)) = 2.

For the next cohomology: H1(O(βˆ’1))β†’H1(O)βŠ•2β†’H1(TP1(βˆ’1))β†’H2(O(βˆ’1))H^1(\mathcal{O}(-1)) \to H^1(\mathcal{O})^{\oplus 2} \to H^1(T_{\mathbb{P}^1}(-1)) \to H^2(\mathcal{O}(-1))

All terms are zero, so H1(TP1(βˆ’1))=0H^1(T_{\mathbb{P}^1}(-1)) = 0.

Vanishing Theorems and Consequences

ExampleVanishing for positive degree

A key consequence is that for any d>0d > 0 and 0<i<n0 < i < n: Hi(Pn,O(d))=0H^i(\mathbb{P}^n, \mathcal{O}(d)) = 0

This is a special case of Serre's vanishing theorem. It implies that for sufficiently positive twists, coherent sheaves on projective varieties have vanishing higher cohomology.

For example, on P2\mathbb{P}^2:

  • H1(P2,O(d))=0H^1(\mathbb{P}^2, \mathcal{O}(d)) = 0 for all dβ‰₯0d \geq 0
  • More generally, if F\mathcal{F} is coherent on P2\mathbb{P}^2, then H1(P2,F(d))=0H^1(\mathbb{P}^2, \mathcal{F}(d)) = 0 for d≫0d \gg 0
ExampleLong exact sequence application

Consider the ideal sheaf sequence of a point p∈P2p \in \mathbb{P}^2: 0β†’Ipβ†’OP2β†’Opβ†’00 \to \mathcal{I}_p \to \mathcal{O}_{\mathbb{P}^2} \to \mathcal{O}_p \to 0

Twisting by O(d)\mathcal{O}(d) and taking cohomology: 0β†’H0(Ip(d))β†’H0(O(d))β†’H0(Op(d))β†’H1(Ip(d))β†’H1(O(d))β†’β‹―0 \to H^0(\mathcal{I}_p(d)) \to H^0(\mathcal{O}(d)) \to H^0(\mathcal{O}_p(d)) \to H^1(\mathcal{I}_p(d)) \to H^1(\mathcal{O}(d)) \to \cdots

For dβ‰₯1d \geq 1:

  • H0(O(d))=k[X0,X1,X2]dH^0(\mathcal{O}(d)) = k[X_0, X_1, X_2]_d has dimension (d+22)\binom{d+2}{2}
  • H0(Op(d))=kH^0(\mathcal{O}_p(d)) = k (evaluating at pp)
  • H1(O(d))=0H^1(\mathcal{O}(d)) = 0

Therefore: h0(Ip(d))=(d+22)βˆ’1h^0(\mathcal{I}_p(d)) = \binom{d+2}{2} - 1 H1(Ip(d))=0H^1(\mathcal{I}_p(d)) = 0

For d=0d = 0:

  • h0(Ip)=0h^0(\mathcal{I}_p) = 0 (no global functions vanishing at a point)
  • From 0β†’0β†’kβ†’βˆΌkβ†’H1(Ip)β†’00 \to 0 \to k \xrightarrow{\sim} k \to H^1(\mathcal{I}_p) \to 0, we get H1(Ip)=0H^1(\mathcal{I}_p) = 0

For d=βˆ’1d = -1:

  • h0(Ip(βˆ’1))=h0(O(βˆ’1))=0h^0(\mathcal{I}_p(-1)) = h^0(\mathcal{O}(-1)) = 0
  • The sequence 0β†’0β†’0β†’H1(Ip(βˆ’1))β†’00 \to 0 \to 0 \to H^1(\mathcal{I}_p(-1)) \to 0 gives H1(Ip(βˆ’1))=0H^1(\mathcal{I}_p(-1)) = 0
ExampleArithmetic genus computation

For a curve CβŠ‚P2C \subset \mathbb{P}^2 of degree dd given by a homogeneous polynomial, the arithmetic genus is: pa(C)=(dβˆ’1)(dβˆ’2)2p_a(C) = \frac{(d-1)(d-2)}{2}

This can be computed using the exact sequence: 0β†’OP2(βˆ’d)β†’OP2β†’OCβ†’00 \to \mathcal{O}_{\mathbb{P}^2}(-d) \to \mathcal{O}_{\mathbb{P}^2} \to \mathcal{O}_C \to 0

The arithmetic genus is pa(C)=1βˆ’Ο‡(OC)p_a(C) = 1 - \chi(\mathcal{O}_C). From the long exact sequence: Ο‡(OC)=Ο‡(OP2)βˆ’Ο‡(OP2(βˆ’d))\chi(\mathcal{O}_C) = \chi(\mathcal{O}_{\mathbb{P}^2}) - \chi(\mathcal{O}_{\mathbb{P}^2}(-d))

We have:

  • Ο‡(OP2)=1\chi(\mathcal{O}_{\mathbb{P}^2}) = 1
  • Ο‡(OP2(βˆ’d))=h0(O(βˆ’d))βˆ’h1(O(βˆ’d))+h2(O(βˆ’d))\chi(\mathcal{O}_{\mathbb{P}^2}(-d)) = h^0(\mathcal{O}(-d)) - h^1(\mathcal{O}(-d)) + h^2(\mathcal{O}(-d))

For dβ‰₯1d \geq 1:

  • h0(O(βˆ’d))=0h^0(\mathcal{O}(-d)) = 0
  • h1(O(βˆ’d))=0h^1(\mathcal{O}(-d)) = 0
  • h2(O(βˆ’d))=(dβˆ’12)=d(dβˆ’1)2h^2(\mathcal{O}(-d)) = \binom{d-1}{2} = \frac{d(d-1)}{2} (by Serre duality with O(dβˆ’3)\mathcal{O}(d-3))

Wait, we need to be careful. For dβ‰₯3d \geq 3: h2(OP2(βˆ’d))=h0(OP2(dβˆ’3))=(dβˆ’3+22)=(dβˆ’12)h^2(\mathcal{O}_{\mathbb{P}^2}(-d)) = h^0(\mathcal{O}_{\mathbb{P}^2}(d-3)) = \binom{d-3+2}{2} = \binom{d-1}{2}

Thus: Ο‡(OC)=1βˆ’(0βˆ’0βˆ’(dβˆ’12))=1+(dβˆ’12)=1+(dβˆ’1)(dβˆ’2)2\chi(\mathcal{O}_C) = 1 - \left(0 - 0 - \binom{d-1}{2}\right) = 1 + \binom{d-1}{2} = 1 + \frac{(d-1)(d-2)}{2}

Therefore: pa(C)=1βˆ’Ο‡(OC)=1βˆ’1βˆ’(dβˆ’1)(dβˆ’2)2=βˆ’(dβˆ’1)(dβˆ’2)2p_a(C) = 1 - \chi(\mathcal{O}_C) = 1 - 1 - \frac{(d-1)(d-2)}{2} = -\frac{(d-1)(d-2)}{2}

Actually, there's a sign issue. Let me recalculate. For dβ‰₯3d \geq 3, βˆ’dβ‰€βˆ’3-d \leq -3, so: h2(OP2(βˆ’d))=h0(O(dβˆ’3))h^2(\mathcal{O}_{\mathbb{P}^2}(-d)) = h^0(\mathcal{O}(d-3))

For d=3d = 3: h2(O(βˆ’3))=h0(O(0))=1h^2(\mathcal{O}(-3)) = h^0(\mathcal{O}(0)) = 1 For d=4d = 4: h2(O(βˆ’4))=h0(O(1))=3h^2(\mathcal{O}(-4)) = h^0(\mathcal{O}(1)) = 3

So Ο‡(O(βˆ’d))=βˆ’h2(O(βˆ’d))\chi(\mathcal{O}(-d)) = -h^2(\mathcal{O}(-d)) for dβ‰₯3d \geq 3.

Thus: Ο‡(OC)=1βˆ’(βˆ’(dβˆ’12))=1+(dβˆ’12)\chi(\mathcal{O}_C) = 1 - \left(-\binom{d-1}{2}\right) = 1 + \binom{d-1}{2}

And: pa(C)=1βˆ’Ο‡(OC)=βˆ’(dβˆ’12)=(dβˆ’1)(dβˆ’2)2p_a(C) = 1 - \chi(\mathcal{O}_C) = -\binom{d-1}{2} = \frac{(d-1)(d-2)}{2}

For a line (d=1d=1): pa=0p_a = 0. For a conic (d=2d=2): pa=0p_a = 0. For a cubic (d=3d=3): pa=1p_a = 1.

Remark

The complete understanding of Hi(Pn,O(d))H^i(\mathbb{P}^n, \mathcal{O}(d)) is fundamental for:

  • Computing cohomology of arbitrary coherent sheaves via resolutions
  • Proving Serre's vanishing theorem
  • Understanding ampleness and embeddings
  • Computing Hilbert polynomials and dimensions of linear systems
  • Applications to curve and surface theory