16
$\begingroup$

The Riemann-Hurwitz formula implies that the projective line $\mathbb{P}^1_K$ over any algebraically closed field $K$ is simply connected (i.e., $\pi_1^{et}(\mathbb{P}^1_K) = 1$; equivalently, if $\phi\colon C\to \mathbb{P}^1_{K}$ is finite etale, then $\deg\phi=1$). For $K=\mathbb{C}$, this follows from the connection with topology and from the fact that the complex plane $\mathbb{A}^1_{\mathbb{C}}$ is contractable. In positive characteristic the affine line is not simply connected due to Artin-Schreier covers.

My question is whether there is a short proof for this fact in positive characteristic?

$\endgroup$
16
$\begingroup$

You can deduce this from the classification of vector bundles on $\mathbf{P}^1$. Say $f:C \to \mathbf{P}^1$ is a connected finite etale Galois cover of degree $n$. We must show $n=1$.

The sheaf $E := f_* \mathcal{O}_C$ is a rank $n$ vector bundle on $\mathbf{P}^1$, so we can write it as $E \simeq \oplus_{i=1}^n \mathcal{O}(a_i)$ for some integers $a_i$. As $C$ is connected, we have $h^0(\mathbf{P}^1,E) = 1$, so we must have $a_1 = 0$ and $a_i < 0$ for $i > 1$ (after rearrangement). It then follows that after pullback along any finite cover $g:D \to \mathbf{P}^1$ of smooth connected curves, we still have $h^0(D, g^* E) = 1$ as negative line bundles remain negative after pullback along finite maps, and thus cannot acquire sections.

But $f$ was a finite etale cover, so there is some $g:D \to \mathbf{P}^1$ as above with $g^* C \to D$ is just a disjoint union of $n$ copies of $D$ mapping down (in fact, one may take $g = f$ as $f$ was Galois). But then $h^0(D, g^*E) = h^0(D, \oplus_{i=1}^n \mathcal{O}_D) = n$. The only way this can happen is if $n=1$.

$\endgroup$
7
$\begingroup$

You can apply the following statement to $X = \mathbb{P}^1_K$ and $L = O(1)$ when $K$ is a separably closed field.

Let $L$ be a line bundle on a reduced connected scheme $X$ such that $H^{0}(X,\mathcal{O}_X)$ is a separably closed field. Assume that any vector bundle on $X$ is of the form $\oplus_{i=1}^{n} L^{\otimes \lambda_i}$ for some integers $\lambda_1 \geq \dots \geq \lambda_n$, and that $L^{\lambda}$ has no nonzero global section for $\lambda < 0$. Then $X$ is simply connected.

Indeed, let $\mathcal{A}$ be a finite etale $\mathcal{O}_X$-algebra. Then $\mathcal{A} = \oplus_{i=1}^{n} L^{\otimes \lambda_i}$ as an $\mathcal{O}_X$-module for some integers $\lambda_1 \geq \dots \geq \lambda_n$. Because $\mathcal{A}$ has a global section (the unit) one must have $\lambda_1 \geq 0$.

If $\lambda_1 > 0$ then the morphism $L^{2 \lambda_1} \rightarrow \mathcal{A} \otimes_{\mathcal{O}_X} \mathcal{A} \rightarrow \mathcal{A} \rightarrow L^{\lambda_i}$ must be zero for each $i$ since $2 \lambda_1 > \lambda_i$. The the multiplication of $\mathcal{A}$ is $0$ on the $L^{ \lambda_1}$-factor, contradicting the reducedness of $\mathcal{A}$. Thus $\lambda_1 = 0$.

Since $\mathcal{A}$ is a finite etale $\mathcal{O}_X$-algebra the discriminant morphism $$ \mathrm{det}(\mathcal{A})^{\otimes 2} \rightarrow \mathcal{O}_X $$ is an isomorphism, so that $\sum_i \lambda_i =0$. Since $\lambda_{i} \leq 0$ for all $i$ we get that $\lambda_i = 0$ and thus that $\mathcal{A}$ is a trivial vector bundle. In particular the morphism $$ H^0(X,\mathcal{A}) \otimes_{H^0(X,\mathcal{O}_X)} \mathcal{O}_X \rightarrow \mathcal{A} $$ is an isomorphism. In particular $H^0(X,\mathcal{A})$ is a finite etale $H^0(X,\mathcal{O}_X)$-algebra, hence a trivial one, so that $\mathcal{A}$ is a trivial finite etale $\mathcal{O}_X$-algebra.

$\endgroup$

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.