ConceptComplete

Ext Sheaves and Local-to-Global Spectral Sequence

The Ext functors provide a fundamental tool for studying extensions of sheaves and computing global cohomology through local data. This concept bridges homological algebra with sheaf cohomology theory.

Ext as Derived Functors

DefinitionExt Functors for Modules

Let RR be a ring and let M,NM, N be RR-modules. The Ext functors are defined as: ExtRi(M,N)=RiHomR(βˆ’,N)(M)\text{Ext}^i_R(M, N) = R^i \text{Hom}_R(-, N)(M) where RiHomR(βˆ’,N)R^i \text{Hom}_R(-, N) is the ii-th right derived functor of HomR(βˆ’,N)\text{Hom}_R(-, N).

Equivalently, taking a projective resolution Pβˆ™β†’MP_\bullet \to M: ExtRi(M,N)=Hi(HomR(Pβˆ™,N))\text{Ext}^i_R(M, N) = H^i(\text{Hom}_R(P_\bullet, N))

Remark

The functor HomR(M,βˆ’)\text{Hom}_R(M, -) is left exact but not right exact in general. The Ext functors measure the failure of right exactness. We have ExtR0(M,N)=HomR(M,N)\text{Ext}^0_R(M, N) = \text{Hom}_R(M, N).

ExampleComputing Ext for Modules

Let R=ZR = \mathbb{Z} and consider M=Z/nZM = \mathbb{Z}/n\mathbb{Z} and N=Z/mZN = \mathbb{Z}/m\mathbb{Z}.

The projective resolution of MM is: 0→Z→nZ→Z/nZ→00 \to \mathbb{Z} \xrightarrow{n} \mathbb{Z} \to \mathbb{Z}/n\mathbb{Z} \to 0

Applying HomZ(βˆ’,Z/mZ)\text{Hom}_\mathbb{Z}(-, \mathbb{Z}/m\mathbb{Z}): 0β†’Hom(Z/nZ,Z/mZ)β†’Hom(Z,Z/mZ)β†’nHom(Z,Z/mZ)0 \to \text{Hom}(\mathbb{Z}/n\mathbb{Z}, \mathbb{Z}/m\mathbb{Z}) \to \text{Hom}(\mathbb{Z}, \mathbb{Z}/m\mathbb{Z}) \xrightarrow{n} \text{Hom}(\mathbb{Z}, \mathbb{Z}/m\mathbb{Z})

This gives: 0β†’Z/gcd⁑(n,m)Zβ†’Z/mZβ†’nZ/mZ0 \to \mathbb{Z}/\gcd(n,m)\mathbb{Z} \to \mathbb{Z}/m\mathbb{Z} \xrightarrow{n} \mathbb{Z}/m\mathbb{Z}

Therefore:

  • ExtZ0(Z/nZ,Z/mZ)=Z/gcd⁑(n,m)Z\text{Ext}^0_\mathbb{Z}(\mathbb{Z}/n\mathbb{Z}, \mathbb{Z}/m\mathbb{Z}) = \mathbb{Z}/\gcd(n,m)\mathbb{Z}
  • ExtZ1(Z/nZ,Z/mZ)=Z/gcd⁑(n,m)Z\text{Ext}^1_\mathbb{Z}(\mathbb{Z}/n\mathbb{Z}, \mathbb{Z}/m\mathbb{Z}) = \mathbb{Z}/\gcd(n,m)\mathbb{Z}
  • ExtZi(Z/nZ,Z/mZ)=0\text{Ext}^i_\mathbb{Z}(\mathbb{Z}/n\mathbb{Z}, \mathbb{Z}/m\mathbb{Z}) = 0 for iβ‰₯2i \geq 2

Ext Sheaves

DefinitionExt Sheaves

Let (X,OX)(X, \mathcal{O}_X) be a ringed space and let F,G\mathcal{F}, \mathcal{G} be OX\mathcal{O}_X-modules. The Ext sheaves are defined as: ExtOXi(F,G)=RiHomOX(βˆ’,G)(F)\mathcal{E}xt^i_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G}) = R^i \mathcal{H}om_{\mathcal{O}_X}(-, \mathcal{G})(\mathcal{F}) where RiHomOXR^i \mathcal{H}om_{\mathcal{O}_X} denotes the ii-th right derived functor of the sheaf Hom functor.

On stalks, we have: ExtOXi(F,G)x=ExtOX,xi(Fx,Gx)\mathcal{E}xt^i_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G})_x = \text{Ext}^i_{\mathcal{O}_{X,x}}(\mathcal{F}_x, \mathcal{G}_x)

Remark

The sheaf HomOX(F,G)\mathcal{H}om_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G}) is defined by: HomOX(F,G)(U)=HomOX∣U(F∣U,G∣U)\mathcal{H}om_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G})(U) = \text{Hom}_{\mathcal{O}_X|_U}(\mathcal{F}|_U, \mathcal{G}|_U)

This is a sheaf, and ExtOX0(F,G)=HomOX(F,G)\mathcal{E}xt^0_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G}) = \mathcal{H}om_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G}).

ExampleExt Sheaves with Locally Free Sheaves

Let XX be a scheme and let E\mathcal{E} be a locally free sheaf of finite rank on XX. For any OX\mathcal{O}_X-module F\mathcal{F}: ExtOXi(E,F)=0forΒ i>0\mathcal{E}xt^i_{\mathcal{O}_X}(\mathcal{E}, \mathcal{F}) = 0 \quad \text{for } i > 0

This is because locally free sheaves are locally isomorphic to OXβŠ•n\mathcal{O}_X^{\oplus n}, which has projective dimension zero. Moreover: HomOX(E,F)β‰…Eβˆ¨βŠ—OXF\mathcal{H}om_{\mathcal{O}_X}(\mathcal{E}, \mathcal{F}) \cong \mathcal{E}^\vee \otimes_{\mathcal{O}_X} \mathcal{F} where E∨=HomOX(E,OX)\mathcal{E}^\vee = \mathcal{H}om_{\mathcal{O}_X}(\mathcal{E}, \mathcal{O}_X) is the dual sheaf.

ExampleExt with Ideal Sheaves

Let X=SpecΒ AX = \text{Spec } A be an affine scheme and ZβŠ‚XZ \subset X a closed subscheme defined by an ideal IβŠ‚AI \subset A. Let I=I~\mathcal{I} = \tilde{I} be the corresponding ideal sheaf.

For the structure sheaf OZ=OX/I\mathcal{O}_Z = \mathcal{O}_X/\mathcal{I}: ExtOX0(OZ,OX)=I\mathcal{E}xt^0_{\mathcal{O}_X}(\mathcal{O}_Z, \mathcal{O}_X) = \mathcal{I} ExtOX1(OZ,OX)=OX/I=OZ\mathcal{E}xt^1_{\mathcal{O}_X}(\mathcal{O}_Z, \mathcal{O}_X) = \mathcal{O}_X/\mathcal{I} = \mathcal{O}_Z

These come from the short exact sequence: 0→I→OX→OZ→00 \to \mathcal{I} \to \mathcal{O}_X \to \mathcal{O}_Z \to 0

The Local-to-Global Spectral Sequence

TheoremLocal-to-Global Ext Spectral Sequence

Let XX be a noetherian scheme and let F,G\mathcal{F}, \mathcal{G} be coherent OX\mathcal{O}_X-modules. There exists a spectral sequence: E2p,q=Hp(X,ExtOXq(F,G))β‡’ExtOXp+q(F,G)E_2^{p,q} = H^p(X, \mathcal{E}xt^q_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G})) \Rightarrow \text{Ext}^{p+q}_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G}) where ExtOXn(F,G)=Hn(X,HomOX(F,Gβ€²))\text{Ext}^n_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G}) = H^n(X, \mathcal{H}om_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G}')) for any injective resolution Gβ†’Gβˆ™β€²\mathcal{G} \to \mathcal{G}'_\bullet.

Proof

This is the Grothendieck spectral sequence for the composition of functors: ExtOXn(F,βˆ’)=H0(X,RnHomOX(F,βˆ’))\text{Ext}^n_{\mathcal{O}_X}(\mathcal{F}, -) = H^0(X, R^n \mathcal{H}om_{\mathcal{O}_X}(\mathcal{F}, -))

We compose the functors:

  • G↦HomOX(F,G)\mathcal{G} \mapsto \mathcal{H}om_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G}) (from OX\mathcal{O}_X-modules to OX\mathcal{O}_X-modules)
  • H↦H0(X,H)\mathcal{H} \mapsto H^0(X, \mathcal{H}) (from OX\mathcal{O}_X-modules to abelian groups)

The first functor has right derived functors ExtOXq(F,βˆ’)\mathcal{E}xt^q_{\mathcal{O}_X}(\mathcal{F}, -), and the second has right derived functors Hp(X,βˆ’)H^p(X, -). The Grothendieck spectral sequence gives: E2p,q=RpH0(X,RqHom(F,βˆ’)(G))=Hp(X,Extq(F,G))E_2^{p,q} = R^p H^0(X, R^q \mathcal{H}om(\mathcal{F}, -)(\mathcal{G})) = H^p(X, \mathcal{E}xt^q(\mathcal{F}, \mathcal{G})) converging to: Rp+q(H0∘Hom(F,βˆ’))(G)=ExtOXp+q(F,G)R^{p+q}(H^0 \circ \mathcal{H}om(\mathcal{F}, -))(\mathcal{G}) = \text{Ext}^{p+q}_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G})

β– 
Remark

This spectral sequence shows how to compute global Ext groups from:

  1. Local Ext sheaves ExtOXq(F,G)\mathcal{E}xt^q_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G})
  2. Global cohomology Hp(X,βˆ’)H^p(X, -)

It relates local homological algebra to global topology.

ExampleSpectral Sequence with Locally Free Sheaves

Let XX be a smooth projective variety and E\mathcal{E} a locally free sheaf. For any coherent sheaf F\mathcal{F}: ExtOXq(E,F)=0forΒ q>0\mathcal{E}xt^q_{\mathcal{O}_X}(\mathcal{E}, \mathcal{F}) = 0 \quad \text{for } q > 0

The spectral sequence degenerates at E2E_2:

  • E2p,0=Hp(X,Eβˆ¨βŠ—F)E_2^{p,0} = H^p(X, \mathcal{E}^\vee \otimes \mathcal{F})
  • E2p,q=0E_2^{p,q} = 0 for q>0q > 0

Therefore: ExtOXn(E,F)=Hn(X,Eβˆ¨βŠ—F)\text{Ext}^n_{\mathcal{O}_X}(\mathcal{E}, \mathcal{F}) = H^n(X, \mathcal{E}^\vee \otimes \mathcal{F})

This is a fundamental computation technique.

Extensions of Sheaves and ExtΒΉ

DefinitionExtensions of Sheaves

Let F\mathcal{F} and G\mathcal{G} be OX\mathcal{O}_X-modules. An extension of G\mathcal{G} by F\mathcal{F} is a short exact sequence: 0→F→E→G→00 \to \mathcal{F} \to \mathcal{E} \to \mathcal{G} \to 0

Two extensions E\mathcal{E} and Eβ€²\mathcal{E}' are equivalent if there exists an isomorphism Eβ†’Eβ€²\mathcal{E} \to \mathcal{E}' making the diagram commute: \requireAMScd\require{AMScd} 0β†’Fβ†’Eβ†’Gβ†’00 \to \mathcal{F} \to \mathcal{E} \to \mathcal{G} \to 0 0β†’F↓≅Gβ†’0\phantom{0 \to \mathcal{F}} \downarrow \cong \phantom{\mathcal{G} \to 0} 0β†’Fβ†’Eβ€²β†’Gβ†’00 \to \mathcal{F} \to \mathcal{E}' \to \mathcal{G} \to 0

TheoremExtensions and ExtΒΉ

There is a natural bijection: ExtOX1(G,F)β‰…{extensionsΒ ofΒ GΒ byΒ F}/∼\text{Ext}^1_{\mathcal{O}_X}(\mathcal{G}, \mathcal{F}) \cong \{\text{extensions of } \mathcal{G} \text{ by } \mathcal{F}\}/\sim

The trivial extension 0β†’Fβ†’FβŠ•Gβ†’Gβ†’00 \to \mathcal{F} \to \mathcal{F} \oplus \mathcal{G} \to \mathcal{G} \to 0 corresponds to the zero element.

Proof

Given an extension 0β†’Fβ†’Eβ†’Gβ†’00 \to \mathcal{F} \to \mathcal{E} \to \mathcal{G} \to 0, take an injective resolution Fβ†’Iβˆ™\mathcal{F} \to \mathcal{I}^\bullet. The extension gives a map Eβ†’I0\mathcal{E} \to \mathcal{I}^0 which lifts partially to I1\mathcal{I}^1, producing an element of: Ext1(G,F)=H1(HomOX(G,Iβˆ™))\text{Ext}^1(\mathcal{G}, \mathcal{F}) = H^1(\text{Hom}_{\mathcal{O}_X}(\mathcal{G}, \mathcal{I}^\bullet))

Conversely, an element of Ext1(G,F)\text{Ext}^1(\mathcal{G}, \mathcal{F}) is represented by a cocycle which can be used to construct an extension via pushout.

The correspondence is natural and respects the group structure where addition of extensions corresponds to Baer sum.

β– 
ExampleExtensions of Line Bundles

Let X=Pk1X = \mathbb{P}^1_k and consider extensions: 0→O(a)→E→O(b)→00 \to \mathcal{O}(a) \to \mathcal{E} \to \mathcal{O}(b) \to 0

Using the spectral sequence: ExtOX1(O(b),O(a))=H1(P1,O(aβˆ’b))\text{Ext}^1_{\mathcal{O}_X}(\mathcal{O}(b), \mathcal{O}(a)) = H^1(\mathbb{P}^1, \mathcal{O}(a-b))

For aβˆ’bβ‰₯βˆ’1a - b \geq -1, we have H1(P1,O(aβˆ’b))=0H^1(\mathbb{P}^1, \mathcal{O}(a-b)) = 0, so all extensions split.

For aβˆ’bβ‰€βˆ’2a - b \leq -2, we have: dim⁑Ext1(O(b),O(a))=bβˆ’aβˆ’1\dim \text{Ext}^1(\mathcal{O}(b), \mathcal{O}(a)) = b - a - 1

These non-split extensions give rank 2 vector bundles on P1\mathbb{P}^1.

ExampleExtensions and Tangent Bundle

On Pn\mathbb{P}^n, the tangent bundle TPnT_{\mathbb{P}^n} fits in the Euler 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

This represents an element of: ExtO1(TPn,OPn)=H1(Pn,TPn∨)\text{Ext}^1_{\mathcal{O}}(T_{\mathbb{P}^n}, \mathcal{O}_{\mathbb{P}^n}) = H^1(\mathbb{P}^n, T_{\mathbb{P}^n}^\vee)

Using TPnβˆ¨β‰…Ξ©Pn1β‰…O(βˆ’2)βŠ•nT_{\mathbb{P}^n}^\vee \cong \Omega^1_{\mathbb{P}^n} \cong \mathcal{O}(-2)^{\oplus n} (from the cotangent sequence), this shows: Ext1(TPn,O)=H1(Pn,Ξ©Pn1)\text{Ext}^1(T_{\mathbb{P}^n}, \mathcal{O}) = H^1(\mathbb{P}^n, \Omega^1_{\mathbb{P}^n}) which vanishes for nβ‰₯2n \geq 2 by cohomology computations.

Examples on Curves

ExampleExt on Curves with Structure Sheaves

Let CC be a smooth projective curve of genus gg and let p∈Cp \in C be a closed point. Consider: ExtOCi(k(p),OC)\text{Ext}^i_{\mathcal{O}_C}(k(p), \mathcal{O}_C) where k(p)=OC/mpk(p) = \mathcal{O}_C/\mathfrak{m}_p is the skyscraper sheaf at pp.

The local-to-global spectral sequence gives: E2p,q=Hp(C,ExtOCq(k(p),OC))E_2^{p,q} = H^p(C, \mathcal{E}xt^q_{\mathcal{O}_C}(k(p), \mathcal{O}_C))

Locally at pp, we have ExtOC1(k(p),OC)p=k(p)\mathcal{E}xt^1_{\mathcal{O}_C}(k(p), \mathcal{O}_C)_p = k(p) and the sheaf Ext1\mathcal{E}xt^1 is supported at pp: ExtOC1(k(p),OC)=k(p)\mathcal{E}xt^1_{\mathcal{O}_C}(k(p), \mathcal{O}_C) = k(p)

Therefore:

  • ExtOC0(k(p),OC)=0\text{Ext}^0_{\mathcal{O}_C}(k(p), \mathcal{O}_C) = 0 (no global sections)
  • ExtOC1(k(p),OC)=H0(C,k(p))βŠ•H1(C,0)=k\text{Ext}^1_{\mathcal{O}_C}(k(p), \mathcal{O}_C) = H^0(C, k(p)) \oplus H^1(C, 0) = k
  • ExtOC2(k(p),OC)=H1(C,k(p))=0\text{Ext}^2_{\mathcal{O}_C}(k(p), \mathcal{O}_C) = H^1(C, k(p)) = 0
ExampleExt with Ideal Sheaf of a Point

Let CC be a smooth projective curve and Ip\mathcal{I}_p the ideal sheaf of a point pp. We have the exact sequence: 0→Ip→OC→k(p)→00 \to \mathcal{I}_p \to \mathcal{O}_C \to k(p) \to 0

For any coherent sheaf F\mathcal{F}, applying Hom(βˆ’,F)\text{Hom}(-, \mathcal{F}): 0β†’Hom(k(p),F)β†’Hom(OC,F)β†’Hom(Ip,F)0 \to \text{Hom}(k(p), \mathcal{F}) \to \text{Hom}(\mathcal{O}_C, \mathcal{F}) \to \text{Hom}(\mathcal{I}_p, \mathcal{F}) β†’Ext1(k(p),F)β†’Ext1(OC,F)β†’Ext1(Ip,F)β†’β‹―\to \text{Ext}^1(k(p), \mathcal{F}) \to \text{Ext}^1(\mathcal{O}_C, \mathcal{F}) \to \text{Ext}^1(\mathcal{I}_p, \mathcal{F}) \to \cdots

Since Hom(k(p),F)=Fp/mpFp\text{Hom}(k(p), \mathcal{F}) = \mathcal{F}_p/\mathfrak{m}_p \mathcal{F}_p and Hom(OC,F)=H0(C,F)\text{Hom}(\mathcal{O}_C, \mathcal{F}) = H^0(C, \mathcal{F}), this sequence computes Ext groups with ideal sheaves.

ExampleExt between Line Bundles on Curves

Let CC be a smooth projective curve and L,M\mathcal{L}, \mathcal{M} line bundles on CC. Then: ExtOCi(L,M)=Hi(C,Lβˆ¨βŠ—M)\text{Ext}^i_{\mathcal{O}_C}(\mathcal{L}, \mathcal{M}) = H^i(C, \mathcal{L}^\vee \otimes \mathcal{M})

By Serre duality: ExtOC1(L,M)=H1(C,Lβˆ¨βŠ—M)β‰…H0(C,LβŠ—Mβˆ¨βŠ—Ο‰C)∨\text{Ext}^1_{\mathcal{O}_C}(\mathcal{L}, \mathcal{M}) = H^1(C, \mathcal{L}^\vee \otimes \mathcal{M}) \cong H^0(C, \mathcal{L} \otimes \mathcal{M}^\vee \otimes \omega_C)^\vee

In particular, for the structure sheaf: ExtOC1(OC,OC)=H1(C,OC)=kg\text{Ext}^1_{\mathcal{O}_C}(\mathcal{O}_C, \mathcal{O}_C) = H^1(C, \mathcal{O}_C) = k^g where gg is the genus of CC.

Examples on Projective Space

ExampleExt on Projective Space

Let X=PnX = \mathbb{P}^n over a field kk. For integers a,ba, b: ExtOXi(O(a),O(b))=Hi(Pn,O(bβˆ’a))\text{Ext}^i_{\mathcal{O}_X}(\mathcal{O}(a), \mathcal{O}(b)) = H^i(\mathbb{P}^n, \mathcal{O}(b-a))

Using the cohomology of O(m)\mathcal{O}(m) on Pn\mathbb{P}^n:

  • H0(Pn,O(m))=k(m+nn)H^0(\mathbb{P}^n, \mathcal{O}(m)) = k^{\binom{m+n}{n}} for mβ‰₯0m \geq 0
  • Hn(Pn,O(m))=k(βˆ’mβˆ’1n)H^n(\mathbb{P}^n, \mathcal{O}(m)) = k^{\binom{-m-1}{n}} for mβ‰€βˆ’nβˆ’1m \leq -n-1
  • Hi(Pn,O(m))=0H^i(\mathbb{P}^n, \mathcal{O}(m)) = 0 otherwise

Therefore:

  • Ext0(O(a),O(b))=k(bβˆ’a+nn)\text{Ext}^0(\mathcal{O}(a), \mathcal{O}(b)) = k^{\binom{b-a+n}{n}} for bβ‰₯ab \geq a
  • Extn(O(a),O(b))=k(aβˆ’bβˆ’1n)\text{Ext}^n(\mathcal{O}(a), \mathcal{O}(b)) = k^{\binom{a-b-1}{n}} for b≀aβˆ’nβˆ’1b \leq a - n - 1
  • Exti(O(a),O(b))=0\text{Ext}^i(\mathcal{O}(a), \mathcal{O}(b)) = 0 otherwise
ExampleExt with Coherent Sheaves on Projective Space

Let F\mathcal{F} be a coherent sheaf on Pn\mathbb{P}^n with a locally free resolution: 0→Er→⋯→E1→E0→F→00 \to \mathcal{E}_r \to \cdots \to \mathcal{E}_1 \to \mathcal{E}_0 \to \mathcal{F} \to 0

For any coherent sheaf G\mathcal{G}: Exti(F,G)=Hi(Homβˆ™(Eβˆ™,G))\text{Ext}^i(\mathcal{F}, \mathcal{G}) = H^i(\text{Hom}^\bullet(\mathcal{E}_\bullet, \mathcal{G}))

Since each Ej\mathcal{E}_j is locally free: Hom(Ej,G)=Ejβˆ¨βŠ—G\text{Hom}(\mathcal{E}_j, \mathcal{G}) = \mathcal{E}_j^\vee \otimes \mathcal{G}

This reduces Ext computations to cohomology of vector bundles.

ExampleExt with Quotient Sheaves

On P2\mathbb{P}^2, consider a twisted cubic curve CβŠ‚P3C \subset \mathbb{P}^3 embedded by OP1(3)\mathcal{O}_{\mathbb{P}^1}(3). The structure sheaf OC\mathcal{O}_C has a resolution: 0β†’OP3(βˆ’3)βŠ•2β†’OP3(βˆ’2)βŠ•3β†’OP3β†’OCβ†’00 \to \mathcal{O}_{\mathbb{P}^3}(-3)^{\oplus 2} \to \mathcal{O}_{\mathbb{P}^3}(-2)^{\oplus 3} \to \mathcal{O}_{\mathbb{P}^3} \to \mathcal{O}_C \to 0

For G=OP3(n)\mathcal{G} = \mathcal{O}_{\mathbb{P}^3}(n): Ext1(OC,O(n))=H1(Homβˆ™(Eβˆ™,O(n)))\text{Ext}^1(\mathcal{O}_C, \mathcal{O}(n)) = H^1(\text{Hom}^\bullet(\mathcal{E}_\bullet, \mathcal{O}(n)))

The complex Homβˆ™(Eβˆ™,O(n))\text{Hom}^\bullet(\mathcal{E}_\bullet, \mathcal{O}(n)) is: 0β†’O(n)β†’O(n+2)βŠ•3β†’O(n+3)βŠ•2β†’00 \to \mathcal{O}(n) \to \mathcal{O}(n+2)^{\oplus 3} \to \mathcal{O}(n+3)^{\oplus 2} \to 0

Computing cohomology determines the Ext groups.

Connection to Serre Duality

TheoremSerre Duality via Ext

Let XX be a smooth projective variety of dimension nn over a field kk with dualizing sheaf Ο‰X\omega_X. For coherent sheaves F,G\mathcal{F}, \mathcal{G} on XX: ExtOXi(F,G)βˆ¨β‰…ExtOXnβˆ’i(G,FβŠ—Ο‰X)\text{Ext}^i_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G})^\vee \cong \text{Ext}^{n-i}_{\mathcal{O}_X}(\mathcal{G}, \mathcal{F} \otimes \omega_X)

In particular, for G=OX\mathcal{G} = \mathcal{O}_X: Exti(F,OX)βˆ¨β‰…Extnβˆ’i(OX,FβŠ—Ο‰X)=Hnβˆ’i(X,FβŠ—Ο‰X)\text{Ext}^i(\mathcal{F}, \mathcal{O}_X)^\vee \cong \text{Ext}^{n-i}(\mathcal{O}_X, \mathcal{F} \otimes \omega_X) = H^{n-i}(X, \mathcal{F} \otimes \omega_X)

Remark

This formulation of Serre duality via Ext highlights the role of the dualizing sheaf in homological algebra. It generalizes the classical Serre duality: Hi(X,F)βˆ¨β‰…Hnβˆ’i(X,Fβˆ¨βŠ—Ο‰X)H^i(X, \mathcal{F})^\vee \cong H^{n-i}(X, \mathcal{F}^\vee \otimes \omega_X) for locally free sheaves.

ExampleSerre Duality on Curves via Ext

Let CC be a smooth projective curve of genus gg with canonical bundle Ο‰C\omega_C. For line bundles L,M\mathcal{L}, \mathcal{M}: Ext0(L,M)∨=Hom(L,M)βˆ¨β‰…Ext1(M,LβŠ—Ο‰C)\text{Ext}^0(\mathcal{L}, \mathcal{M})^\vee = \text{Hom}(\mathcal{L}, \mathcal{M})^\vee \cong \text{Ext}^1(\mathcal{M}, \mathcal{L} \otimes \omega_C)

Explicitly: H0(C,Lβˆ¨βŠ—M)βˆ¨β‰…H1(C,Mβˆ¨βŠ—LβŠ—Ο‰C)H^0(C, \mathcal{L}^\vee \otimes \mathcal{M})^\vee \cong H^1(C, \mathcal{M}^\vee \otimes \mathcal{L} \otimes \omega_C)

This is the classical Serre duality for curves, expressed in Ext language.

ExampleSelf-Duality for Ext

On a smooth projective surface SS with canonical bundle Ο‰S\omega_S, consider a coherent sheaf F\mathcal{F}. Then: Exti(F,F)βˆ¨β‰…Ext2βˆ’i(F,FβŠ—Ο‰S)\text{Ext}^i(\mathcal{F}, \mathcal{F})^\vee \cong \text{Ext}^{2-i}(\mathcal{F}, \mathcal{F} \otimes \omega_S)

For i=1i = 1: Ext1(F,F)βˆ¨β‰…Ext1(F,FβŠ—Ο‰S)\text{Ext}^1(\mathcal{F}, \mathcal{F})^\vee \cong \text{Ext}^1(\mathcal{F}, \mathcal{F} \otimes \omega_S)

This duality is important in deformation theory, where Ext1(F,F)\text{Ext}^1(\mathcal{F}, \mathcal{F}) parametrizes first-order deformations of F\mathcal{F}.

Vanishing Theorems via Ext

TheoremLocal Vanishing of Ext Sheaves

Let XX be a regular scheme and F,G\mathcal{F}, \mathcal{G} coherent OX\mathcal{O}_X-modules. Then: ExtOXi(F,G)=0for i>dim⁑X\mathcal{E}xt^i_{\mathcal{O}_X}(\mathcal{F}, \mathcal{G}) = 0 \quad \text{for } i > \dim X

This follows because every point has a regular local ring, which has finite global dimension equal to its dimension.

ExampleExt Vanishing on Curves

Let CC be a smooth curve (regular of dimension 1). For any coherent sheaves F,G\mathcal{F}, \mathcal{G}: Exti(F,G)=0forΒ iβ‰₯2\mathcal{E}xt^i(\mathcal{F}, \mathcal{G}) = 0 \quad \text{for } i \geq 2

This implies the local-to-global spectral sequence has only two rows: E2p,q=Hp(C,Extq(F,G))with q∈{0,1}E_2^{p,q} = H^p(C, \mathcal{E}xt^q(\mathcal{F}, \mathcal{G})) \quad \text{with } q \in \{0, 1\}

The spectral sequence gives exact sequences: 0β†’H1(C,Hom(F,G))β†’Ext1(F,G)β†’H0(C,Ext1(F,G))β†’00 \to H^1(C, \mathcal{H}om(\mathcal{F}, \mathcal{G})) \to \text{Ext}^1(\mathcal{F}, \mathcal{G}) \to H^0(C, \mathcal{E}xt^1(\mathcal{F}, \mathcal{G})) \to 0

ExampleExt on Higher Dimensional Varieties

Let XX be a smooth projective variety of dimension nn. For a locally free sheaf E\mathcal{E} and coherent sheaf F\mathcal{F}: Exti(E,F)=Hi(X,Eβˆ¨βŠ—F)\text{Ext}^i(\mathcal{E}, \mathcal{F}) = H^i(X, \mathcal{E}^\vee \otimes \mathcal{F})

By standard vanishing (e.g., for ample line bundles L\mathcal{L} and large mm): Hi(X,Eβˆ¨βŠ—FβŠ—Lm)=0forΒ i>0,m≫0H^i(X, \mathcal{E}^\vee \otimes \mathcal{F} \otimes \mathcal{L}^m) = 0 \quad \text{for } i > 0, m \gg 0

This gives: Exti(EβŠ—Lβˆ’m,F)=0forΒ i>0,m≫0\text{Ext}^i(\mathcal{E} \otimes \mathcal{L}^{-m}, \mathcal{F}) = 0 \quad \text{for } i > 0, m \gg 0

Computational Techniques

Remark

The primary methods for computing Ext groups are:

  1. Spectral Sequence Method: Use the local-to-global spectral sequence when Ext\mathcal{E}xt sheaves are known.

  2. Resolution Method: Take a locally free resolution of F\mathcal{F} and compute cohomology of the Hom complex.

  3. Long Exact Sequence: From a short exact sequence of sheaves, derive long exact sequences of Ext groups.

  4. Reduction to Cohomology: When one sheaf is locally free, reduce to cohomology computations via Exti(E,F)=Hi(X,Eβˆ¨βŠ—F)\text{Ext}^i(\mathcal{E}, \mathcal{F}) = H^i(X, \mathcal{E}^\vee \otimes \mathcal{F}).

ExampleLong Exact Sequence for Ext

Given a short exact sequence of coherent sheaves: 0→F′→F→F′′→00 \to \mathcal{F}' \to \mathcal{F} \to \mathcal{F}'' \to 0

For any coherent sheaf G\mathcal{G}, we obtain long exact sequences: 0β†’Hom(Fβ€²β€²,G)β†’Hom(F,G)β†’Hom(Fβ€²,G)0 \to \text{Hom}(\mathcal{F}'', \mathcal{G}) \to \text{Hom}(\mathcal{F}, \mathcal{G}) \to \text{Hom}(\mathcal{F}', \mathcal{G}) β†’Ext1(Fβ€²β€²,G)β†’Ext1(F,G)β†’Ext1(Fβ€²,G)β†’β‹―\to \text{Ext}^1(\mathcal{F}'', \mathcal{G}) \to \text{Ext}^1(\mathcal{F}, \mathcal{G}) \to \text{Ext}^1(\mathcal{F}', \mathcal{G}) \to \cdots

And in the other variable: 0β†’Hom(G,Fβ€²)β†’Hom(G,F)β†’Hom(G,Fβ€²β€²)0 \to \text{Hom}(\mathcal{G}, \mathcal{F}') \to \text{Hom}(\mathcal{G}, \mathcal{F}) \to \text{Hom}(\mathcal{G}, \mathcal{F}'') β†’Ext1(G,Fβ€²)β†’Ext1(G,F)β†’Ext1(G,Fβ€²β€²)β†’β‹―\to \text{Ext}^1(\mathcal{G}, \mathcal{F}') \to \text{Ext}^1(\mathcal{G}, \mathcal{F}) \to \text{Ext}^1(\mathcal{G}, \mathcal{F}'') \to \cdots

These sequences are essential for inductive computations.


The theory of Ext sheaves provides a powerful framework for understanding both local and global aspects of coherent sheaves. The local-to-global spectral sequence serves as a bridge between pointwise homological algebra and global cohomology, while the connection to extensions and Serre duality reveals deep structural properties of algebraic varieties.