From a graded ring , we defined . We also defined , the Serre twisting sheaf. (These need not be line bundles.)
We always have -module homomorphisms
Hence, by functoriality, we have maps
#TODO How do we define the first map? What this means is that these form a graded ring: we set
This is a -graded ring. There is a map of -graded rings:
Let , with grading by total degree. Then
is an isomorphism of graded rings. Let us consider a basic case: when , we have on the degree component
Write . By the sheaf condition, we have the exact sequence
For , this puts a restriction on the indices in the sums above: we need and . Therefore except and except . So
This gives the -graded -module
Let be of homogeneous degree , and . Then
Moreover, if is quasi-coherent on , i.e., is a quasi-coherent -module and , then
In particular if is generated by and (uniformly bounded), e.g. if is quasi-compact, then is a line bundle for all and
The failure of the twisting sheaf to be a line bundle has to do with not being generated in degree .
Since is an affine scheme, it suffices to prove that as (ungraded, since everything is degree ) -modules. If , we can consider
Localising, becomes a unit and is an isomorphism. Taking degree components, we have . For the case , we map the other way to obtain the same result.
If is quasi-compact, then we have that
is an adjoint pair. Moreover,
- is essentially surjective: every quasi-coherent sheaf on is the of a graded module.
- is fully faithful: is an isomorphism.
. Recall that our definition of is that a morphism of -schemes is the same as a line bundle on and a surjection .
We show that satisfies the same universal property.
Corresponding to our global sections , we get a quotient map
So there is a map (by the universal property of ). In fact, because of this, any map will have a line bundle with sections, obtained via .
Conversely, let be an -scheme together with a surjection for a line bundle on . This corresponds to such that , where
open. This means that the map
is an isomorphism for each (a nowhere-vanishing section of a line bundle). We now try to map to our natural cover of .
What follows is an argument concerning the uniqueness of a morphism .
Since is affine, a morphism is equivalent to a ring homomorphism
(By .) By the universal property of polynomial rings and quotients, this is equivalent to picking some elements , , such that . Now, if there is to be an isomorphism and (defined above), , then there must be a commutative diagram
Following this diagram around, we have
Therefore, must be the image of in under the homomorphism
It follows that is completely determined by the .
Now, suppose that are two morphisms of -schemes such that there are isomorphisms and with for all .
It suffices to check this on each . Now,
Similarly for . So it follows follows that . By the above calculations, we see that for all because any map on the opens are completely determined by the . So .
For existence of such an , for each we define a morphism
exactly as determined above: we take to be the appropriate image of . We need to show that they glue:
But a map is equivalent to a ring homomorphism . But the composition
sends for all . So the remaining question is whether in . But this boils down to the equation
which is indeed true in .