On the stalk level, the cocycle condition says that the isomorphism F g h ⋅ x ≃ F x {\displaystyle F_{gh\cdot x}\simeq F_{x}} is the same as the composition F g ⋅ h ⋅ x ≃ F h ⋅ x ≃ F x {\displaystyle F_{g\cdot h\cdot x}\simeq F_{h\cdot x}\simeq F_{x}} ; i.e., the associativity of the group action. The condition that the unit of the group acts as the identity is also a consequence: apply ( e × e × 1 ) ∗ , e : S → G {\displaystyle (e\times e\times 1)^{*},e:S\to G} to both sides to get ( e × 1 ) ∗ ϕ ∘ ( e × 1 ) ∗ ϕ = ( e × 1 ) ∗ ϕ {\displaystyle (e\times 1)^{*}\phi \circ (e\times 1)^{*}\phi =(e\times 1)^{*}\phi } and so ( e × 1 ) ∗ ϕ {\displaystyle (e\times 1)^{*}\phi } is the identity.
Note that ϕ {\displaystyle \phi } is an additional data; it is "a lift" of the action of G on X to the sheaf F. Moreover, when G is a connected algebraic group, F an invertible sheaf and X is reduced, the cocycle condition is automatic: any isomorphism σ ∗ F ≃ p 2 ∗ F {\displaystyle \sigma ^{*}F\simeq p_{2}^{*}F} automatically satisfies the cocycle condition.3
If the action of G is free, then the notion of an equivariant sheaf simplifies to a sheaf on the quotient X/G, because of the descent along torsors.
By Yoneda's lemma, to give the structure of an equivariant sheaf to an O X {\displaystyle {\mathcal {O}}_{X}} -module F is the same as to give group homomorphisms for rings R over S {\displaystyle S} ,
There is also a definition of equivariant sheaves in terms of simplicial sheaves. Alternatively, one can define an equivariant sheaf to be an equivariant object in the category of, say, coherent sheaves.
A structure of an equivariant sheaf on an invertible sheaf or a line bundle is also called a linearization.
Let X be a complete variety over an algebraically closed field acted by a connected reductive group G and L an invertible sheaf on it. If X is normal, then some tensor power L n {\displaystyle L^{n}} of L is linearizable.5
Also, if L is very ample and linearized, then there is a G-linear closed immersion from X to P N {\displaystyle \mathbf {P} ^{N}} such that O P N ( 1 ) {\displaystyle {\mathcal {O}}_{\mathbf {P} ^{N}}(1)} is linearized and the linearization on L is induced by that of O P N ( 1 ) {\displaystyle {\mathcal {O}}_{\mathbf {P} ^{N}}(1)} .6
Tensor products and the inverses of linearized invertible sheaves are again linearized in the natural way. Thus, the isomorphism classes of the linearized invertible sheaves on a scheme X form an abelian group. There is a homomorphism to the Picard group of X which forgets the linearization; this homomorphism is neither injective nor surjective in general, and its kernel can be identified with the isomorphism classes of linearizations of the trivial line bundle.
See Example 2.16 of [1] for an example of a variety for which most line bundles are not linearizable.
Given an algebraic group G and a G-equivariant sheaf F on X over a field k, let V = Γ ( X , F ) {\displaystyle V=\Gamma (X,F)} be the space of global sections. It then admits the structure of a G-module; i.e., V is a linear representation of G as follows. Writing σ : G × X → X {\displaystyle \sigma :G\times X\to X} for the group action, for each g in G and v in V, let
where σ ∗ : V → Γ ( G × X , σ ∗ F ) {\displaystyle \sigma ^{*}:V\to \Gamma (G\times X,\sigma ^{*}F)} and φ : Γ ( G × X , σ ∗ F ) → ∼ Γ ( G × X , p 2 ∗ F ) = k [ G ] ⊗ k V {\displaystyle \varphi :\Gamma (G\times X,\sigma ^{*}F){\overset {\sim }{\to }}\Gamma (G\times X,p_{2}^{*}F)=k[G]\otimes _{k}V} is the isomorphism given by the equivariant-sheaf structure on F. The cocycle condition then ensures that π : G → G L ( V ) {\displaystyle \pi :G\to GL(V)} is a group homomorphism (i.e., π {\displaystyle \pi } is a representation.)
Example: take X = G , F = O G {\displaystyle X=G,F={\mathcal {O}}_{G}} and σ = {\displaystyle \sigma =} the action of G on itself. Then V = k [ G ] {\displaystyle V=k[G]} , ( φ ∘ σ ∗ ) ( f ) ( g , h ) = f ( g h ) {\displaystyle (\varphi \circ \sigma ^{*})(f)(g,h)=f(gh)} and
meaning π {\displaystyle \pi } is the left regular representation of G.
The representation π {\displaystyle \pi } defined above is a rational representation: for each vector v in V, there is a finite-dimensional G-submodule of V that contains v.7
A definition is simpler for a vector bundle (i.e., a variety corresponding to a locally free sheaf of constant rank). We say a vector bundle E on an algebraic variety X acted by an algebraic group G is equivariant if G acts fiberwise: i.e., g : E x → E g x {\displaystyle g:E_{x}\to E_{gx}} is a "linear" isomorphism of vector spaces.8 In other words, an equivariant vector bundle is a pair consisting of a vector bundle and the lifting of the action G × X → X {\displaystyle G\times X\to X} to that of G × E → E {\displaystyle G\times E\to E} so that the projection E → X {\displaystyle E\to X} is equivariant.
Just like in the non-equivariant setting, one can define an equivariant characteristic class of an equivariant vector bundle.
MFK 1994, Ch 1. § 3. Definition 1.6. - Mumford, David; Fogarty, John; Kirwan, Frances (1994). Geometric Invariant Theory. Berlin: Springer-Verlag. ISBN 978-3-540-56963-3. MR 1304906. https://books.google.com/books?id=dFlv3zn_2-gC ↩
Gaitsgory 2005, § 6. - Gaitsgory, D. (2005). "Geometric Representation theory, Math 267y, Fall 2005" (PDF). Archived from the original (PDF) on 22 January 2015. https://web.archive.org/web/20150122111026/http://www.math.harvard.edu/~gaitsgde/267y/catO.pdf ↩
MFK 1994, end of the proof of Ch. 1, § 3., Proposition 1.5. - Mumford, David; Fogarty, John; Kirwan, Frances (1994). Geometric Invariant Theory. Berlin: Springer-Verlag. ISBN 978-3-540-56963-3. MR 1304906. https://books.google.com/books?id=dFlv3zn_2-gC ↩
Thomason 1987, § 1.2. - Thomason, R.W. (1987). "Algebraic K-theory of group scheme actions". In Browser, William (ed.). Algebraic topology and algebraic K-theory : proceedings of a conference, October 24-28, 1983 at Princeton University, dedicated to John C. Moore on his 60th birthday. Vol. 113. Princeton, N.J.: Princeton University Press. p. 539-563. ISBN 9780691084268. ↩
MFK 1994, Ch 1. § 3. Corollary 1.6. - Mumford, David; Fogarty, John; Kirwan, Frances (1994). Geometric Invariant Theory. Berlin: Springer-Verlag. ISBN 978-3-540-56963-3. MR 1304906. https://books.google.com/books?id=dFlv3zn_2-gC ↩
MFK 1994, Ch 1. § 3. Proposition 1.7. - Mumford, David; Fogarty, John; Kirwan, Frances (1994). Geometric Invariant Theory. Berlin: Springer-Verlag. ISBN 978-3-540-56963-3. MR 1304906. https://books.google.com/books?id=dFlv3zn_2-gC ↩
MFK 1994, Ch. 1. § 1. the lemma just after Definition 1.3. - Mumford, David; Fogarty, John; Kirwan, Frances (1994). Geometric Invariant Theory. Berlin: Springer-Verlag. ISBN 978-3-540-56963-3. MR 1304906. https://books.google.com/books?id=dFlv3zn_2-gC ↩
If E is viewed as a sheaf, then g needs to be replaced by g − 1 {\displaystyle g^{-1}} . ↩