# finite etale group scheme over a field

Could I have an example of a finite etale group scheme over a field k which is not a constant group scheme?

I just know that the category of etale group schemes over a k is equivalent to the category of abstract groups with a galois action. So I am wondering what the objects would be like in the left category. In fact any finite etale scheme over k should be a finite union of spectrums of fields which are finite separable extensions of k, and if it is a group scheme then it should have a rational point, but I don't see how one can translate the rational point to other points, so I guess there might be some non-rational point in the group scheme. is that right? an example?

-
How about $\mu_n$, the $n$-th roots of $1$, over a field $k$ in which $n$ is invertible but which does not contain a primitive $n$-root of $1$ ? – Chandan Singh Dalawat Nov 4 '12 at 15:27
For $k=\mathbf{Q}$, it works for every $n>2$. – Chandan Singh Dalawat Nov 4 '12 at 16:08
Yes. Let $x$ be an element which is not an $n$th power in $k$. Then the set of $n$th roots of $x$ is a torsor for the $n$th roots of unity in an obvious way, and has no $k$-valued points, thus is nontrivial. – Will Sawin Nov 4 '12 at 16:25
If you have a surjective morphism of commutative groups $G\to H$ with kernel $N$, then every fibre is a torsor under $N$. – Chandan Singh Dalawat Nov 4 '12 at 16:25
@stefan: every torsor under an étale group scheme is étale as a scheme (or algebraic space), because étaleness is an fppf-local property. Hence it is trivialized by an étale covering (namely itself). – Laurent Moret-Bailly Nov 5 '12 at 7:19