1
$\begingroup$

$\DeclareMathOperator\GL{GL}\DeclareMathOperator\SO{SO}\DeclareMathOperator\Sp{Sp}\DeclareMathOperator\ad{ad}\DeclareMathOperator\Ad{Ad}$Let $G\subset \GL_n(\mathbb{C})$ be a complex matrix Lie group, (e.g., $\SO_n$, $\Sp_{n}$), and let $x\in G$ with Jordan form $u$ ($u$ may not belong to $G$) given, I wonder how to compute the centralizer $Z_{G}(x)$.

I know the Lie algebra of $Z_{G}(x)$ is $\{Y\in \mathfrak{g}\mid \Ad(x)Y=Y\}$. But it's hard for me to find a matrix $t\in \GL_n(\mathbb{C})$ such that $tut^{-1}\in G$. I tried to find $U\in \mathfrak{gl}_n(\mathbb{C})$ such that $\exp(U)=u$ and then find a $T\in \mathfrak{gl}_n(\mathbb{C})$ such that $\ad(T)(U)\in \mathfrak{g}$, but I don't think that means $\exp(T)u(\exp(T))^{-1}\in G$.

Any help will be appreciated.

$\endgroup$

1 Answer 1

4
$\begingroup$

You’re question talks abut the centraliser but you then seem instead to be unsure how to relate an arbitrary Jordan normal to an element of $G$. I’ll try to answer both.

Firstly, the centraliser. A matrix of the form: $$ \begin{pmatrix}\lambda & 1 & & \\ & \ddots &\ddots \\ & & \ddots &1 \\ & & & \lambda \end{pmatrix}$$ has centraliser given by upper triangular Toeplitz matrices e.g. : $$ \begin{pmatrix}a & b & c & d\\ & a &b & c \\ & & a &b \\ & & & a \end{pmatrix} $$

So the centraliser of a matrix in Jordan normal form comprises block diagonal matrices with these Toeplitz blocks where the generalised eigenspaces are and generic blocks where the honest eigenspaces are.

Now if we have a $v$ conjugate to $u$, their centralisers are conjugate as well by the same element so we just need fo find a way to relate $u$ to an element of $G$. In practice this is a bit tedious so I'll be brief. We just need to pick $v$ with the right eigenvalues, dimension of eigenspaces and generalised eigenspaces, etc. We then find a basis of eigenvectors and generalised eigenvectors (arranged in Jordan chains) and then our conjugacy matrix has these as columns.

Of course there will be many such $v$ in general so it would usually make more sense to start with $v$ find its Jordan normal form and use that to compute the centraliser.

$\endgroup$
2
  • $\begingroup$ I know for conjugate matrices, their centralizers are conjugate; but here the Jordan form u may not belong to G, so the centralizer in G of the element it represents may not be conjugate to the centralizer of u. Also, I think what you did is for $G=GL_n$, but general matrix Lie group can be more complicated. $\endgroup$
    – mhahthhh
    May 19, 2022 at 0:56
  • 2
    $\begingroup$ Sorry, I didn't make that part clear. The centraliser of $x \in G$ is just the intersection of its centraliser in $GL_n$ with $G$. So take the conjugation of the centraliser of $u$ and then intersect it with $G$ $\endgroup$
    – Callum
    May 19, 2022 at 9:08

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.