10
$\begingroup$

Let $R$ be a complete local $\mathbf{Z}_p$-algebra, for some prime $p$. In the 1970 paper Group schemes of prime order by Oort and Tate, they write down an explicit finite flat group scheme $G_R(a, b)$ of rank $p$, for each pair of elements $a,b \in R$ satisfying $ab = p$, and show that every FFGS over $R$ of rank $p$ is isomorphic to one of these.

What are the homomorphisms (of $R$-group schemes) from $G_R(a, b)$ to $G_R(a', b')$?

The underlying scheme of $G_R(a, b)$ is $\{ X : X^p - aX = 0\}$. If we look for homomorphisms which are "linear", $X \mapsto \lambda X$ for some $\lambda$, then (after a bit of unravelling) we conclude that $\lambda$ has to be a point of the $R$-scheme $$\{ Z : Z(aZ^{p-1} - a') = Z(b - b'Z^{p-1}) = 0\}.$$ (This is a group scheme, but neither finite nor flat over $R$ in general, although it is a FFGS if $\{a, b'\}$ generate the unit ideal of $R$.)

Are these all the homomorphisms $G_R(a, b) \to G_R(a', b')$?

(The answer is clearly "yes" for $p = 2$. It is also "yes" for $p = 3$ via a deeply nasty polynomial computation.)

$\endgroup$

1 Answer 1

8
$\begingroup$

Yes. The Tate-Oort description is an equivalence of categories between finite flat group schemes (over a $\Lambda$-scheme $S$ where $\Lambda$ is a certain ring decribed in the paper) and the category of triples $(L , a, b)$ where $L$ is an invertible sheaf over $S$ and $a\in \Gamma(S,L^{\otimes (p−1)})$, $b\in\Gamma(S,L^{\otimes (1− p)})$ satisfy $a\otimes b=w_p\cdot 1$. The morphisms between $(L,a,b)$ and $(L',a',b')$ are the morphisms of invertible sheaves $f:L\to L'$, viewed as global sections of $L^{\otimes -1}\otimes L'$, such that $a\otimes f^{\otimes p}=f\otimes a'$ and $b'\otimes f^{\otimes p}=f\otimes b$. In the case where the base is a local scheme, this is exactly what you wrote. All I wrote is almost verbatim from the Tate-Oort paper.

$\endgroup$
5
  • $\begingroup$ Oort and Tate just write this down as a bijection on automorphism classes of objects; they don't explicitly state that $G \to (L, a, b)$ is a functor. I'm sure you're correct and it's functorial, but do you know a reference where this is written down explicitly? $\endgroup$ Jan 20, 2022 at 20:09
  • 1
    $\begingroup$ The classification of the group schemes in Oort-Tate uses the decomposition of the Hopf algebra (more precisely the augmentation ideal of this) of the group scheme by eigenspaces for the natural action of $\mathbf{F}_p^*$. So of course any morphism $G \to G'$ gives a corresponding map between the eigenspaces which is what $f$ in the answer above is (by picking the correct eigenspace). This is pretty clear from Lemma 2 of their paper. So I agree with Matthieu that you can't say this isn't in Oort-Tate IMHO. $\endgroup$
    – Johan
    Jan 21, 2022 at 1:38
  • 1
    $\begingroup$ In "Moduli of Galois p-covers in mixed characteristics" with Dan Abramovich (Algebra Number Theory 2012) we write this in Definition A.5 + comments after it. But we say hardly more than that! $\endgroup$ Jan 21, 2022 at 15:14
  • 1
    $\begingroup$ Also even though it doesn't matter much, note that the Tate-Oort paper is authored by John Tate and Frans Oort in that order. $\endgroup$ Jan 21, 2022 at 15:14
  • $\begingroup$ Yes, very good Tate-Oort indeed. $\endgroup$
    – Johan
    Jan 21, 2022 at 15:31

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.