MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

The structure of a commutative affine algebraic group $G$ over a field $k$ is understood (SGA 3): $G$ has a maximal subgroup of multiplicative type $M$ and the quotient $G/M$ is unipotent.

I am looking for any generalization of this structure theory for commutative group schemes over an integral base scheme, $X$. I'd appreciate any general comments/references but for sake of an explicit question:

By a unipotent group scheme I mean a group scheme over $X$ that has a normal series with $\mathbb G_a$-factors. Then the question is: let $G$ be a smooth commutative group scheme of finite type $G$ over an integral $\mathbb C$-scheme $X$ admitting a maximal torus $T$. Does there exist a Zariski open $U \subset X$ over which $G/T$ is unipotent in this sense?

share|cite|improve this question
Dear Pooya, It might help to clarify exactly what you mean by "maximal torus" in the relative context. Regards, – Emerton Sep 4 '12 at 3:08
Are there smooth commutative unipotent groups that are not vector spaces? – Will Sawin Sep 4 '12 at 3:18
Dear Emerton: yes! A maximal torus is a subgroup scheme $T$ of $G$ which is a torus (i.e. fpqc locally isomorphic to $\mathbb G_m^r$) and point-wise maximal: for any point $x \in X$, $T_{\bar x}$ is the maximal torus of $G_{\bar x}$ where $\bar x$ is the spectrum of the algebraic closure of $\kappa (x)$ [SGA 3-IIX, Definition 1.3]. Thanks, – PRon Sep 4 '12 at 3:38
Dear Will: I think they are vector spaces (i.e. Zariski locally trivial) by my definition [Kambayashi-Miyanishi, on flat fibrations by the affine line]. – PRon Sep 4 '12 at 3:48
The quotient $G/T$ exists as a smooth affine group, and by design it is unipotent on geometric fibers over the integral base in char. 0. The generic fiber is unipotent over the char. 0 function field of the base. Your definition of unipotence is wrong in char. $> 0$ (where there are many smooth connected commutative unipotent groups containing no $\mathbf{G}_a$), but in char. 0 it is equivalent to the "right" definition (the filtration condition on geometric fibers). Such a composition series on the actual (not just geometric) generic fiber spreads out over some dense open of the base. – grp Sep 4 '12 at 6:19

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.