Say I have a reductive complex algebraic group $G$ with maximal torus $T$ and associated Weyl group $W$. I would like to be able to say that the characters of $G$ are in bijection with the $W$-invariant characters of $T$, and the bijection is given by restriction. Is this correct? If so, how could it be proven?
2 Answers
I thought at first that you meant "trace character of a finite-dimensional, irreducible representation of $G$" by "character of $G$". Then this is false; for example, if $G$ is $\operatorname{SL}_2$, then $G$ certainly has non-trivial, finite-dimensional, irreducible representations (e.g., the defining representation), but the only $W(G, T)$-fixed character of a maximal torus $T$ in $G$ is the trivial one.
If you mean literally "homomorphism $G \to \mathbb C^\times$", then certainly such a character gives a $W(G, T)$-fixed character of $T$. Conversely, a $W(G, T)$-invariant character of $T$ is trivial on $[W(G, T), T] = T \cap [G, G]$ (where the bracket notation means the group generated by commutators, not the set of commutators), and $T/[W(G, T), T] \to G/[G, G]$ is an isomorphism, so we obtain a map $G/[G, G] \to \mathbb C^\times$, as desired.
In case you're curious about that equality and isomorphism, they use the structure of reductive groups:
Since $N_G(T) \to W(G, T)$ is surjective, certainly $[W(G, T), T]$ is contained in $T \cap [G, G]$. Conversely, $T \cap [G, G]$ is the smallest subtorus of $T$ containing the image of $\alpha^\vee$ for every root $\alpha$ of $T$ in $G$, but $[t, w_\alpha]$ equals $\alpha^\vee(\alpha(t))$ for every such root and every $t \in T$.
We have just shown that $T/[W(G, T), T] \to G/[G, G]$ is an embedding. Since $G$ is the almost direct product of its maximal central torus $Z$ with $[G, G]$, already $Z \to G/[G, G]$ is surjective; but $Z$ (being a central torus) is contained in $T$, so $T/[W(G, T), G] \to G/[G, G]$ is also surjective.
Incidentally, I also used the surjectivity of $N_G(T) \to W(G, T)$ to assert that the restriction of a homomorphism $G \to \mathbb C^\times$ to $T$ is $W(G, T)$-fixed.
-
$\begingroup$ I have posted an alternative answer. $\endgroup$ Commented Jul 3, 2023 at 19:44
Let $G$ be a connected reductive group over an algebraically closed field $k$. Write $G^{\rm ss}=[G,G]$ (which is semisimple) and let $G^{\rm sc}$ denote the universal cover of the semisimple group $G^{\rm ss}$; then $G^{\rm sc}$ is simply connected. Consider the composite homomorphism $$\rho\colon G^{\rm sc}\twoheadrightarrow G^{\rm sc}\hookrightarrow G.$$
Let $T\subseteq G$ be a maximal torus. Consider the maximal tori $T^{\rm ss}=T\cap G^{\rm ss}\subset G^{\rm ss}$, $T^{\rm sc}=\rho^{-1}(T)\subset G^{\rm sc}$. We write $X^*$ for the character group, and $X_*$ for the cocharacter group. Clearly, \begin{multline*} X^*(G)\cong\{x\in X^*(T)\ |\ x|_{T^{\rm ss}}=1\}\\ =\{x\in X^*(T)\ |\ \langle x,u\rangle=0\ \ \forall u\in X_*(T^{\rm ss})\} =:X_*(T^{\rm ss})^\bot \end{multline*}
Proposition. $X^*(G)\cong X^*(T)^W$.
Proof. Write $X=X^*(T)$, $X^\vee=X_*(T)$. Let $R=R(G,T)\subset X$ denote the root system, and $R^\vee=R^\vee(G,T)\subset X^\vee$ denote the coroot system. Let $B\subset G$ be a Borel subgroup containing $T$. Let $S=S(G,T,B)\subset R\subset X$ denote the system of simple roots, and $S^\vee=S^\vee(G,T,B)\subset R^\vee\subset X^\vee$ denote the system of simple coroots. Then $S^\vee$ is a basis of $X_*(T^{\rm sc})$ embedded in $X$.
Since the homomorphism $T^{\rm sc}\to T^{\rm ss}$ is surjective with finite kernel, the induced homomorphism $X_*(T^{\rm sc})\to X_*(T^{\rm ss})$ is injective with finite cokernel. We conclude that $$ X^*(G)\cong X_*(T^{\rm ss})^\bot =X_*(T^{\rm sc})^\bot= (S^\vee)^\bot.$$
We know that the Weyl group $W=W(G,T)$ is generated by the reflections $s_\alpha$ for $\alpha\in S$. These reflections act on $X=X^*(T)$ by $$s_\alpha(x)= x-\langle x,\alpha^\vee\rangle \alpha\quad\ \text{for}\ \, x\in X, \,\alpha\in S;$$ see, for instance, Subsection 1.1 in Tonny A. Springer, Reductive groups. Proc. Sympos. Pure Math., XXXIII, Amer. Math. Soc., Providence, R.I., 1979, pp. 3–27. We write $X^W$ for the subgroup of $W$-fixed points of $X$; then \begin{multline*} X^W=\{x\in X\ |\ s_\alpha(x)=x\ \ \forall \alpha\in S\}\\ =\{x\in X\ |\ \langle x,\alpha^\vee\rangle =0 \ \ \forall \alpha^\vee\in S^\vee\}= (S^\vee)^\bot. \end{multline*} Thus $X^*(G)\cong X_*(T^{\rm ss})^\bot = (S^\vee)^\bot=X^W$, as required.
-
$\begingroup$ Could you say something about how this answer differs from mine? There's nothing clever about mine—it's just sort of the natural approach—but the reasoning seems quite similar to yours, except that I don't pass explicitly to the simply connected cover. $\endgroup$– LSpiceCommented Jul 3, 2023 at 21:33