The Stacks project

Lemma 59.89.1. Let $K/k$ be an extension of fields. Let $X$ be a smooth affine curve over $k$ with a rational point $x \in X(k)$. Let $\mathcal{F}$ be an abelian sheaf on $\mathop{\mathrm{Spec}}(K)$ annihilated by an integer $n$ invertible in $k$. Let $q > 0$ and

\[ \xi \in H^ q(X_ K, (X_ K \to \mathop{\mathrm{Spec}}(K))^{-1}\mathcal{F}) \]

There exist

  1. finite extensions $K'/K$ and $k'/k$ with $k' \subset K'$,

  2. a finite étale Galois cover $Z \to X_{k'}$ with group $G$

such that the order of $G$ divides a power of $n$, such that $Z \to X_{k'}$ is split over $x_{k'}$, and such that $\xi $ dies in $H^ q(Z_{K'}, (Z_{K'} \to \mathop{\mathrm{Spec}}(K))^{-1}\mathcal{F})$.

Proof. For $q > 1$ we know that $\xi $ dies in $H^ q(X_{\overline{K}}, (X_{\overline{K}} \to \mathop{\mathrm{Spec}}(K))^{-1}\mathcal{F})$ (Theorem 59.83.10). By Lemma 59.51.5 we see that this means there is a finite extension $K'/K$ such that $\xi $ dies in $H^ q(X_{K'}, (X_{K'} \to \mathop{\mathrm{Spec}}(K))^{-1}\mathcal{F})$. Thus we can take $k' = k$ and $Z = X$ in this case.

Assume $q = 1$. Recall that $\mathcal{F}$ corresponds to a discrete module $M$ with continuous $\text{Gal}_ K$-action, see Lemma 59.59.1. Since $M$ is $n$-torsion, it is the uninon of finite $\text{Gal}_ K$-stable subgroups. Thus we reduce to the case where $M$ is a finite abelian group annihilated by $n$, see Lemma 59.51.4. After replacing $K$ by a finite extension we may assume that the action of $\text{Gal}_ K$ on $M$ is trivial. Thus we may assume $\mathcal{F} = \underline{M}$ is the constant sheaf with value a finite abelian group $M$ annihilated by $n$.

We can write $M$ as a direct sum of cyclic groups. Any two finite étale Galois coverings whose Galois groups have order invertible in $k$, can be dominated by a third one whose Galois group has order invertible in $k$ (Fundamental Groups, Section 58.7). Thus it suffices to prove the lemma when $M = \mathbf{Z}/d\mathbf{Z}$ where $d | n$.

Assume $M = \mathbf{Z}/d\mathbf{Z}$ where $d | n$. In this case $\overline{\xi } = \xi |_{X_{\overline{K}}}$ is an element of

\[ H^1(X_{\overline{k}}, \mathbf{Z}/d\mathbf{Z}) = H^1(X_{\overline{K}}, \mathbf{Z}/d\mathbf{Z}) \]

See Theorem 59.83.10. This group classifies $\mathbf{Z}/d\mathbf{Z}$-torsors, see Cohomology on Sites, Lemma 21.4.3. The torsor corresponding to $\overline{\xi }$ (viewed as a sheaf on $X_{\overline{k}, {\acute{e}tale}}$) in turn gives rise to a finite étale morphism $T \to X_{\overline{k}}$ endowed an action of $\mathbf{Z}/d\mathbf{Z}$ transitive on the fibre of $T$ over $x_{\overline{k}}$, see Lemma 59.64.4. Choose a connected component $T' \subset T$ (if $\overline{\xi }$ has order $d$, then $T$ is already connected). Then $T' \to X_{\overline{k}}$ is a finite étale Galois cover whose Galois group is a subgroup $G \subset \mathbf{Z}/d\mathbf{Z}$ (small detail omitted). Moreover the element $\overline{\xi }$ maps to zero under the map $H^1(X_{\overline{k}}, \mathbf{Z}/d\mathbf{Z}) \to H^1(T', \mathbf{Z}/d\mathbf{Z})$ as this is one of the defining properties of $T$.

Next, we use a limit argument to choose a finite extension $k'/k$ contained in $\overline{k}$ such that $T' \to X_{\overline{k}}$ descends to a finite étale Galois cover $Z \to X_{k'}$ with group $G$. See Limits, Lemmas 32.10.1, 32.8.3, and 32.8.10. After increasing $k'$ we may assume that $Z$ splits over $x_{k'}$. The image of $\xi $ in $H^1(Z_{\overline{K}}, \mathbf{Z}/d\mathbf{Z})$ is zero by construction. Thus by Lemma 59.51.5 we can find a finite subextension $\overline{K}/K'/K$ containing $k'$ such that $\xi $ dies in $H^1(Z_{K'}, \mathbf{Z}/d\mathbf{Z})$ and this finishes the proof. $\square$


Comments (2)

Comment #3591 by shanbei on

In your condition (1), maybe you meant ?

There are also:

  • 6 comment(s) on Section 59.89: Smooth base change

Post a comment

Your email address will not be published. Required fields are marked.

In your comment you can use Markdown and LaTeX style mathematics (enclose it like $\pi$). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.




In order to prevent bots from posting comments, we would like you to prove that you are human. You can do this by filling in the name of the current tag in the following input field. As a reminder, this is tag 0EYT. Beware of the difference between the letter 'O' and the digit '0'.