4
$\begingroup$

I got these questions while reading Chapter 3 of "Representations of Finite Groups of Lie Type" by Digne-Michel.

Let $T$ be a torus defined over $\mathbb{F}_q$. Then $\text{Gal}(\overline{\mathbb{F}_q}/{\mathbb{F}_q})$ acts on $X(T)$ as it acts on the algebras of $\mathbb{G}_m$ and on $\overline{\mathbb{F}_q}[T]$ via the arithmetic Frobenius. Let $\tau$ be the action of the element $x \mapsto x^q$ on $X(T)$.

First Question (Mainly to make sure I think about this correctly):

If I consider $X(T) \subseteq \overline{\mathbb{F}_q}[T]$, then is this $\tau$ the restriction of the arithmetic Frobenius of $\overline{\mathbb{F}_q}[T]$ to $X(T)$? I mean the isomorphism

$$\text{Hom}(\overline{\mathbb{F}_q}[X,X^{-1}], \overline{\mathbb{F}_q}[T]) \cong \text{Hom}(T,\mathbb{G}_m) $$

is given by $\varphi \mapsto \varphi(X)$, right? And the action of $x \mapsto x^q$ of the Galois group on the left is given by $\varphi \mapsto F_0 \circ \varphi \circ F_0^{-1}$ where $F_0$ denotes the arithmetic Frobenius of both algebras. Then under this isomorphism, the action of $\tau$ on $\chi = \varphi(X)$ would be $\tau(\chi) = F_0 (\varphi (F_0^{-1}(X))) = F_0(\varphi(X)) = F_0(\chi)$, because $F_0(X) = X$ in $\overline{\mathbb{F}_q}[X,X^{-1}]$, right?

Second Question (what I really would like to understand):

The following can be found on p. 40 of Digne-Michel:

"If $T$ is a rational maximal torus of a reductive connected algebraic group $G$, it is clear that $\tau$ permutes the root system of $G$ relative to $T$. Similarly the transpose automorphism $\tau^*$ of $Y (T)$ permutes the coroots."

Unfortunately, this is not clear to me at all.

My Thoughts:

What I know is that if $F: G \to G$ is the Frobenius morphism associated with the $\mathbb{F}_q$-structure of $G$, being an isogeny, we get an induced $p$-morphism (in the sense of Springer), that is, a bijection $b : \Phi \to \Phi$ of the roots and a $p$-power $q(\alpha)$ for every $\alpha \in \Phi$ such that $F^*(b(\alpha)) = q(\alpha) \alpha$ holds for all $\alpha \in \Phi$ (as well as a similar equation for coroots). Here $F^*$ denotes the comorphism of $F$ on $\overline{\mathbb{F}_q}[G]$, $\overline{\mathbb{F}_q}[T]$ or $X(T)$.

From this and $\tau (F^*(\alpha)) = q \alpha$ we get $q b(\alpha) = q(\alpha) \tau(\alpha)$ for all $\alpha \in \Phi$ and if $\tau(\alpha)$ were a root this would imply $b(\alpha) = \tau(\alpha)$ and $q = q(\alpha)$ for all $\alpha \in \Phi$. On the other hand, if I could prove that $q(\alpha) = q$ for all $\alpha \in \Phi$, I would be done, but I do not know if this is the correct approach.

$\endgroup$
3
$\begingroup$

The important thing to note is, that the comorphism $F^* : \mathbb{F}[G] \to \mathbb{F}[G]$ has the property $F^*(\mathbb{F}[G]) = \mathbb{F}[G]^q$ and this implies $F^*(\mathbb{F}[U_{b(\alpha)}]) = \mathbb{F}[U_\alpha]^q$ for all $\alpha \in \Phi$. But from $F(u_\alpha(x)) = u_{b(\alpha)}(c_\alpha x^{q(\alpha)})$ we obtain $F^*(\mathbb{F}[U_{b(\alpha)}]) = \mathbb{F}[U_\alpha]^{q(\alpha)}$. As $\mathbb{F}[U_\alpha]$ is isomorphic to the polynomial ring in one variable, it follows that $q(\alpha) = q$ for all $\alpha$. Then $\tau(\alpha) = b(\alpha)$ for all $\alpha \in \Phi$, so $\tau$ maps $\Phi$ to itself.

$\endgroup$
  • $\begingroup$ Could you please say that how the relation $F(u_{\alpha}(x))=u_{b(\alpha)}(c_{\alpha} x^{q_{\alpha}})$ will imply that $F^*(\mathbb{F}[U_{b(\alpha)}])={\mathbb{F}[U_{\alpha}]}^{q_{\alpha}}$? $\endgroup$ – user97635 May 16 '17 at 12:53
  • 1
    $\begingroup$ @user97635 : Up to the isomorphisms $u_\alpha$ and $u_{b(\alpha)}$ (or, more precisely, their comorphisms), the comorphism of $F$ is an endomorphism of the polynomial ring in one variable given by $\mathbb{F}[T] \to \mathbb{F}[T], T \mapsto c_\alpha T^{q(\alpha)}$. The image of this map is precisely $\mathbb{F}[T]^{q(\alpha)}$. $\endgroup$ – Matthias Klupsch May 17 '17 at 6:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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