16

Let $\mathcal{F}$ be a sheaf on a scheme $X$ and $O_X(k)$ as usual. We define $\mathcal{F}(n) = \mathcal{F} \otimes_{O_X} O_X(n)$, I don't undertand this definition. What is this tensor product? Then, if we can try to find examples, if $n=1$ and $\mathcal{F}=\mathcal{O}_{\mathbb{A}^1_k}$, who is $\mathcal{O}_{\mathbb{A}^1_k}(1)$?, and $\mathcal{O}_{\mathbb{A}^1_k}(n)$?

ArthurStuart
  • 4,932

3 Answers3

31

As already noted, the twisted sheaf $\mathcal O(n)$ is defined at first on projective space $\mathbb P^r$. The basic way to think about is that the global sections of $\mathcal O(n)$ on $\mathbb P^r$ is precisely precisely the vector space of homogeneous polynomials of degree $n$. So $\mathcal O(n)$ is a way of talking about homogeneous polynomials in a more geometric way.

You ask how to think about $\mathcal F(n)$ for a coherent sheaf $\mathcal F$. One basic example of a coherent sheaf is an ideal sheaf $\mathcal I_X$ cutting out some projective variety $X$ in $\mathbb P^r$. Then $\mathcal I_X(n)$ will be the sheaf whose global sections are the homogeneous polynomials of degree $n$ that vanish on $X$.

For an example of how to use these ideas:

Suppose that $X$ is a rational curve of degree $3$ in $\mathbb P^3$ (rational curve means that $X$ is isomorphic to $\mathbb P^1$). Degree $3$ means that a generic hyperplane in $\mathbb P^3$ cuts $X$ in $3$ points, which can be rephrased by saying that $\mathcal O_X(1)$ is a degree $3$ invertible sheaf on $X$. Thus $\mathcal O_X(n)$ is a degree $3n$ invertible sheaf.

We have the short exact sequence $$0 \to \mathcal I_X \to \mathcal O_{\mathbb P^3} \to \mathcal O_X \to 0,$$ and if we twist this becomes $$0 \to \mathcal I_X(n) \to \mathcal O(n) \to \mathcal O_X(n) \to 0.$$ Taking global sections gives a left exact sequence $$0 \to \text{degree $n$ hom. polys. vanishing on } X \to \text{ all degree $n$ hom. polys.} \to \Gamma( X, \text{degree $3n$ invertible sheaf}) $$ Now the dimension of the space of all degree $n$ hom. polys. on in $4$ variables is $n+3 \choose 3$, while the number of degree $3n$ hom. polys. in $2$ variables is $3n+1$. Thus this left exact sequence can be rewritten as $$0 \to \text{degree $n$ hom. polys. vanishing on } X \to {n+3 \choose 3} \text{-dimensional vector space } \to (3n+1)~\text{-dimensional vector space}.$$

Thus we see that if $n \geq 2,$ then the space of degree $n$ hom. polys. vanishing on $X$ is non-zero, and conclude for example that any rational cubic curve in $\mathbb P^3$ is contained in a (possibly singular) quadric.

There are lots more examples of this kind worked out in Chapter IV of Hartshorne, and that is a good place to look to get an idea of how to use these tools geometrically.

Matt E
  • 123,735
  • Thank you for the very good answer! – ArthurStuart Nov 13 '13 at 10:17
  • 1
    @ArthurStuart: Dear Arthur, You're welcome. Cheers, – Matt E Nov 14 '13 at 22:15
  • @MattE This answer was helpful, but I found it by just looking for something related to the twisted sheaf. Is Hartshorne a good source for learning about twisted sheaves with an introduction, or would you recommend another source? I've been working in Vakil's book off and on for...the past year or so haha. – Tanner Strunk Jan 19 '17 at 02:51
3

It only makes sense to talk about $\mathcal F(k)$ and $\mathcal O_X(k)$ if you have an embedding $i:X \hookrightarrow \mathbb P^n$ for some $n$. Then $\mathcal O_X(k)$ is the pullback of the $k$'th twist of the structure sheaf on $\mathbb P ^n$. In symbols, $\mathcal O_X(k) := i^* \mathcal O_{\mathbb P ^n}(k)$. For details, see Hartshorne's book, Chapter II, section 5 (page 120).

Thus, it doesn't make sense to talk about $\mathcal O_{\mathbb A^1}(k)$ without specifying an embedding $i:\mathbb A^1 \hookrightarrow \mathbb P ^n$ such that $i ^* \mathcal O_{\mathbb P ^n}(k)$ is invertible (I´m not sure if this is possible).

For an example, consider the projective line $\mathbb P ^1$. It can be realized as the homogeneous zero set of $\mathsf {Proj} \, k[x,y]$. Then to give a sheaf on $\mathbb P ^1$ is equivalent to giving a graded $k[x,y]$-module $M$ and sheafify to get a sheaf $\mathcal F = \tilde M$. Then $\mathcal F(n)$ is defined as $\mathcal F \otimes \mathcal O_{\mathbb P^1}(n)$, but this is just $\tilde{ M(n)}$, the sheafification of the twisted module. (Proposition 5.12 in Hartshorne, Chapter II)

To sum up: finitely-generated graded modules correspond to coherent sheaves on projective spaces and their twist corresponds to twisting the grading of the modules. (recall that the twisted module $M(k)$ is the same module, but with grading $M(k)_l = M_{l+k}$, so the zeroth piece of $M(k)$ is $M_k$)

Fredrik Meyer
  • 20,228
2

I am sorry, but I think $O_X(n)$ can be defined when the underlying scheme is projective. So I think the proper example can be $\mathcal{O}_{\mathbb{P}^1_k}(1) $,and this is just tautological line bundle which is made from the dgree one parts of the oringinal graded ring.

Cao
  • 53