10
$\begingroup$

Let $k$ be a field, n a positive integer.

Vakil's notes, 17.4.B: Show that all the automorphisms of the projective scheme $P_k^n$ correspond to $(n+1)\times(n+1)$ invertible matrices over k, modulo scalars.

His hint is to show that $f^\star \mathcal{O}(1) \cong \mathcal{O}(1).$ (f is the automorphism. I don't if $\mathcal{O}(1)$ is the conventional notation; if it's unclear, it's an invertible sheaf over $P_k^n$) I can show what he wants assuming this, but can someone help me find a clean way to show this?

$\endgroup$
11
$\begingroup$

An automorphism of $\mathbb{P}^n_k$ induces an automorphism of the Picard group $\text{Pic}(\mathbb{P}^n_k) \cong \mathbb{Z}$. Such an automorphism must send the generator $\mathcal{O}(1)$ to a generator. Since the only two generators of $\mathbb{Z}$ are $1$ and $-1$, $f^*(\mathcal{O}(1))$ must be $\mathcal{O}(1)$ or $\mathcal{O}(-1)$. But $\mathcal{O}(-1)$ has no nonzero global sections, so it cannot be the pullback of $\mathcal{O}(1)$ (recall that $\mathcal{O}(1)$ pulls back to a line bundle together with $n+1$ global sections which have no common zero).

$\endgroup$
10
$\begingroup$

Well, $f^*(\mathcal{O}(1))$ must be a line bundle on $\mathbb{P}^n$. In fact, $f^*$ gives a group automorphism of $\text{Pic}(\mathbb{P}^n) \cong \mathbb{Z}$, with inverse $(f^{-1})^*$. Thus, $f^*(\mathcal{O}(1))$ must be a generator of $\text{Pic}(\mathbb{P}^n)$, either $\mathcal{O}(1)$ or $\mathcal{O}(-1)$. But $f^*$ is also an automorphism on the space of global sections, again with inverse $(f^{-1})^*$. Since $\mathcal{O}(1)$ has an $(n+1)$-dimensional vector space of global sections, but $\mathcal{O}(-1)$ has no non-zero global sections, it is impossible for $f^*(\mathcal{O}(1))$ to be $\mathcal{O}(-1)$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.