Let $R$ be a commutative ring, $p >0$ prime and $G$ a finite, locally free group scheme over $R$ of rank $p^n$; $n \in \mathbb{N}_{\ge 1}$. Assume $p \in R^*$ (i.e. is a unit in $R$).

Question: Why this condition on the rank implies that $G$ is étale?

By definition etale is equivalent to flat & unramified. As $G$ is locally free it's obviously flat. Be unramified is also a local condition. Thus we can translate the problem to commutative algebra and asking why the free $R$-module $R^{p^n}$ is unramified at a prime $\mathfrak{q} \subset R$ if $p \in R^*$.


2 Answers 2


This is not so easy, but relies on a well-known structure theorem for connected group schemes over a perfect field.

Lemma 1. A finitely presented morphism $Y \to X$ of schemes is unramified if and only if $Y_x \to x$ is unramified for all $x \in X$.

Proof. See [EGA IV$_4$, Cor. 17.4.2]. $\square$

Thus, we may reduce to the case $R = k$ for $k$ a field, and by flat descent to the case where $k$ is algebraically closed. Then unramified for a finite extension means (geometrically) reduced.

Lemma 2. A finite group scheme over an algebraically closed field is a semidirect product of an étale group scheme and a connected group scheme.

Proof. See for example [Wat, §6.8]. The étale part is $\pi_0(G)$ and the connected part is $G^0$. $\square$

Theorem. If $G$ is a geometrically connected finite group scheme over a perfect field $k$ of characteristic $p > 0$, then $\Gamma(G,\mathcal O_G)$ is isomorphic to $k[X_1,\ldots,X_n]/(X_1^{p^{e_1}},\ldots,X_n^{p^{e_n}})$ for some $e_1,\ldots, e_n \in \mathbf Z_{>0}$.

Proof. See for example [Wat, §14.4]. I don't know a quick summary of why this is supposed to be true (I would be interested if someone does), but the proof is not that hard. $\square$

Theorem. If $G$ is group scheme over a field of characteristic $0$, then $G$ is geometrically reduced.

Proof. See for example [Wat, §11.4]. $\square$

In particular, if the rank of $G$ not divisible by $p$ (e.g. $p = 0$), then $G^0$ has to be trivial and $G$ is étale. $\square$


[EGA IV$_4$] A. Grothendieck, Éléments de géométrie algébrique. IV: Étude locale des schémas et des morphismes de schémas (Quatrième partie).. Publ. Math., Inst. Hautes Étud. Sci. 32, p. 1-361 (1967). ZBL0153.22301.

[Wat] W.C. Waterhouse, Introduction to affine group schemes. Graduate Texts in Mathematics 66 (1979). ZBL0442.14017.

  • $\begingroup$ Your final note I not fully understand. By your reduction steps we apply two base changes: first one we take a prime $\mathfrak{q}_x \subset R$ and do base change along $R \to R_{\mathfrak{q}_x} \to R_{\mathfrak{q}_x}/\mathfrak{q}_x=k(x)$ (can now apply Lemma 1) and then change base to alg closure via $k(x) \to \overline{k(x)}$ by the flat descent argument as you said). Set $k=k(x)$ and we are now in the setting of the Theorem, i.e. set $G=G^0$ (only con c), $\Gamma(G,\mathcal O_G)= k[X_1,\ldots,X_n]/(X_1^{q^{e_1}},\ldots,X_n^{q^{e_n}})$ where $q$ is the characteristic of $k$. Obviously $q$ is $\endgroup$
    – user267839
    Feb 11, 2020 at 21:08
  • $\begingroup$ prime to $p$ as $p \in R^*$ by assumption. If $q>0$: rank of $k[X_1,\ldots,X_n]/(X_1^{q^{e_1}},\ldots,X_n^{q^{e_n}})$ is a power of $q$ but also of $p$, a contradiction. That's fine. But what about the case $q=0$. Then the theorem isn't applicable... $\endgroup$
    – user267839
    Feb 11, 2020 at 21:19
  • $\begingroup$ @TimGrosskreutz: right, I forgot to mention the characteristic 0 situation. There everything is reduced, so a finite group scheme is always étale. $\endgroup$ Feb 11, 2020 at 21:21
  • $\begingroup$ one remark: could you give a reference for the statement that in case where $k$ is algebraically closed, then for a finite $k$-module (thus a field) unramified is equivalent to (geometrically) reduced? $\endgroup$
    – user267839
    Feb 11, 2020 at 21:25
  • $\begingroup$ See for example Tags 00U3 and 030W (but the latter deals with separable transcendental extensions as well). $\endgroup$ Feb 11, 2020 at 22:54

Since unramified-ness is local, and since $G$ is a group scheme, it is sufficient to show that, for a finite flat commutative group scheme $G$ of order $n$ over an algebraically closed field $k$ whose characteristic is prime to $n$, the structure morphism is unramified at $0$.

This is equivalent to that the tangent space of $G/k$ at $0$ is $0$.

And this is easy: [$n$], the "multiple by $n$" map on $G$ induces the one on the tangent space. Since $n$ is invertible on $k$, it is an isomorphism. But by Deligne, [$n$] is $0$ on $G$. Thus the tangent space is $0$.

  • $\begingroup$ The OP does not assume that $G$ is commutative. This is needed for Deligne's result; the same statement without that assumption is an open problem (at least over a general base). $\endgroup$ Jan 21, 2021 at 19:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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