[proofplan]
We first set $C=C_G(T)$ and observe that $T\le C$ because $T$ is abelian. The centralizer connectedness theorem implies that $C$ is connected. We then compute the [Lie algebra](/page/Lie%20Algebra) of $C$ as the set of elements of $\mathfrak g$ commuting with $\mathfrak t=\operatorname{Lie}(T)$, and use maximality of $\mathfrak t$ as an abelian subalgebra to force this Lie algebra to equal $\mathfrak t$. Since $T\le C$ and both groups are connected Lie subgroups with the same Lie algebra, they coincide.
[/proofplan]
custom_env
admin
[step:Place the maximal torus inside its centralizer]
Let
\begin{align*}
C:=C_G(T)=\{g\in G:gt=tg\text{ for every }t\in T\}.
\end{align*}
Since $T$ is a torus, it is abelian. Hence every element of $T$ commutes with every element of $T$, so $T\le C$.
By [citetheorem:9721] applied to the torus $T\le G$, the subgroup $C=C_G(T)$ is connected. Since $C$ is a closed subgroup of the compact Lie group $G$, it is also a compact Lie subgroup of $G$.
[/step]
custom_env
admin
[step:Identify the Lie algebra of the centralizer]Let $\mathfrak g:=\operatorname{Lie}(G)$, let $\mathfrak t:=\operatorname{Lie}(T)$, and let $\mathfrak c:=\operatorname{Lie}(C)$. Write $\exp_G:\mathfrak g\to G$ and $\exp_T:\mathfrak t\to T$ for the Lie exponential maps; under the inclusion $T\le G$, $\exp_T(Y)=\exp_G(Y)$ for $Y\in\mathfrak t$. We claim that
\begin{align*}
\mathfrak c=\{X\in\mathfrak g:[X,Y]=0\text{ for every }Y\in\mathfrak t\}.
\end{align*}
Indeed, let $X\in\mathfrak c$. For every $s\in\mathbb R$, the element $\exp_G(sX)$ lies in $C$, so for every $t\in T$,
\begin{align*}
t\exp_G(sX)t^{-1}=\exp_G(sX).
\end{align*}
Equivalently,
\begin{align*}
\exp_G(s\operatorname{Ad}_tX)=\exp_G(sX).
\end{align*}
Differentiating at $s=0$ gives $\operatorname{Ad}_tX=X$ for every $t\in T$. Now let $Y\in\mathfrak t$. Applying this to $t=\exp_G(rY)$ for $r\in\mathbb R$ gives
\begin{align*}
\operatorname{Ad}_{\exp_G(rY)}X=X.
\end{align*}
Differentiating at $r=0$ gives $[Y,X]=0$, hence $[X,Y]=0$.
Conversely, suppose $X\in\mathfrak g$ satisfies $[X,Y]=0$ for every $Y\in\mathfrak t$. Since $T$ is a torus, the exponential map $\exp_T:\mathfrak t\to T$ is surjective. For any $t\in T$, choose $Y\in\mathfrak t$ such that $t=\exp_G(Y)$. Since $[X,Y]=0$, the one-parameter subgroups $\exp_G(sX)$ and $\exp_G(rY)$ commute for all $r,s\in\mathbb R$. In particular,
\begin{align*}
\exp_G(sX)t=t\exp_G(sX)
\end{align*}
for every $s\in\mathbb R$ and every $t\in T$. Thus $\exp_G(sX)\in C$ for every $s\in\mathbb R$, so $X\in\operatorname{Lie}(C)=\mathfrak c$.[/step]
custom_env
admin
[guided]The goal of this step is to translate the group-theoretic condition “commutes with every element of $T$” into the infinitesimal condition “commutes with every element of $\mathfrak t$.” Keep the notation $C=C_G(T)$, $\mathfrak g=\operatorname{Lie}(G)$, $\mathfrak t=\operatorname{Lie}(T)$, and $\mathfrak c=\operatorname{Lie}(C)$. Write $\exp_G:\mathfrak g\to G$ and $\exp_T:\mathfrak t\to T$ for the Lie exponential maps. We prove both inclusions.
First take $X\in\mathfrak c$. By definition of the Lie algebra of a Lie subgroup, the one-parameter subgroup
\begin{align*}
s\mapsto \exp_G(sX)
\end{align*}
has image in $C$. Since $C$ centralizes $T$, for every $t\in T$ and every $s\in\mathbb R$ we have
\begin{align*}
t\exp_G(sX)t^{-1}=\exp_G(sX).
\end{align*}
Conjugation by $t$ differentiates to the [linear map](/page/Linear%20Map) $\operatorname{Ad}_t:\mathfrak g\to\mathfrak g$, and the exponential map is equivariant under conjugation:
\begin{align*}
t\exp_G(sX)t^{-1}=\exp_G(s\operatorname{Ad}_tX).
\end{align*}
Therefore
\begin{align*}
\exp_G(s\operatorname{Ad}_tX)=\exp_G(sX)
\end{align*}
for all $s$. Differentiating this identity at $s=0$ gives
\begin{align*}
\operatorname{Ad}_tX=X.
\end{align*}
Now choose $Y\in\mathfrak t$ and put $t=\exp_G(rY)$, where $r\in\mathbb R$. The identity above becomes
\begin{align*}
\operatorname{Ad}_{\exp_G(rY)}X=X.
\end{align*}
Differentiating at $r=0$ gives
\begin{align*}
[Y,X]=0.
\end{align*}
By skew-symmetry of the Lie bracket, $[X,Y]=0$. Since $Y\in\mathfrak t$ was arbitrary, $X$ lies in the Lie algebra centralizer of $\mathfrak t$.
For the reverse inclusion, suppose $X\in\mathfrak g$ satisfies
\begin{align*}
[X,Y]=0
\end{align*}
for every $Y\in\mathfrak t$. We need to prove that the one-parameter subgroup generated by $X$ lies in $C$. Let $t\in T$. Because $T$ is a torus, its exponential map $\exp_T:\mathfrak t\to T$ is surjective, so there exists $Y\in\mathfrak t$ such that
\begin{align*}
t=\exp_G(Y).
\end{align*}
The hypothesis $[X,Y]=0$ implies that the one-parameter subgroups generated by $X$ and $Y$ commute:
\begin{align*}
\exp_G(sX)\exp_G(rY)=\exp_G(rY)\exp_G(sX)
\end{align*}
for all $r,s\in\mathbb R$. Taking $r=1$ gives
\begin{align*}
\exp_G(sX)t=t\exp_G(sX)
\end{align*}
for every $s\in\mathbb R$. Since $t\in T$ was arbitrary, $\exp_G(sX)$ centralizes all of $T$, so $\exp_G(sX)\in C$ for every $s$. Hence $X\in\operatorname{Lie}(C)=\mathfrak c$.
Combining the two inclusions proves
\begin{align*}
\mathfrak c=\{X\in\mathfrak g:[X,Y]=0\text{ for every }Y\in\mathfrak t\}.
\end{align*}[/guided]
custom_env
admin
[step:Use maximality of $\mathfrak t$ to force $\operatorname{Lie}(C)=\mathfrak t$]
By [citetheorem:9719], the Lie algebra $\mathfrak t$ of the maximal torus $T$ is a maximal abelian subalgebra of $\mathfrak g$.
Let $X\in\mathfrak c$. By the preceding step,
\begin{align*}
[X,Y]=0
\end{align*}
for every $Y\in\mathfrak t$. Define the real vector subspace
\begin{align*}
\mathfrak a:=\mathfrak t+\mathbb RX\subseteq\mathfrak g.
\end{align*}
The subspace $\mathfrak a$ is an abelian Lie subalgebra: brackets inside $\mathfrak t$ vanish because $\mathfrak t$ is abelian, brackets between $X$ and elements of $\mathfrak t$ vanish by the choice of $X$, and $[X,X]=0$ by skew-symmetry. Since $\mathfrak t$ is maximal abelian and $\mathfrak t\subseteq\mathfrak a$, we must have $\mathfrak a=\mathfrak t$. Hence $X\in\mathfrak t$.
Thus $\mathfrak c\subseteq\mathfrak t$. Since $T\le C$, differentiating the inclusion gives $\mathfrak t\subseteq\mathfrak c$. Therefore
\begin{align*}
\mathfrak c=\mathfrak t.
\end{align*}
[/step]
custom_env
admin
[step:Conclude equality from connectedness and equality of Lie algebras]
We have shown that $T\le C$, that $C$ is connected, and that
\begin{align*}
\operatorname{Lie}(C)=\mathfrak c=\mathfrak t=\operatorname{Lie}(T).
\end{align*}
For connected Lie subgroups of a Lie group, the Lie algebra determines the subgroup: if $H_1$ and $H_2$ are connected Lie subgroups with $\operatorname{Lie}(H_1)=\operatorname{Lie}(H_2)$, then $H_1=H_2$. Applying this to $T$ and $C$ gives
\begin{align*}
C=T.
\end{align*}
Since $C=C_G(T)$ by definition, this proves
\begin{align*}
C_G(T)=T.
\end{align*}
[/step]