0
$\begingroup$

This is the form of matrices I am concerned about recently,

Denote by $A=(a_{ij})_{e \times e}$, where $a_{ij}$ is block matrix of dimension $n_i \times n_j$. Now given $n, 0 \leq n \leq e-1 $. Let $a_{ij}=0$, except for $j-i=n$ mod $e$. For example, $e=3,n=2$, then

$A= \left( \begin{matrix} 0 & 0 & a_{13} \\ a_{21} & 0 & 0 \\ 0 & a_{32} & 0 \\ \end{matrix} \right) $

My question is, what kind of Jordan decomposition can we get here, especially for its nilpotent part? For another example, $e=2, n=1$ case, we cannot directly use the construction way, because the characteristic polynomial for $ A= \left( \begin{matrix} 0 & a_{12} \\ a_{21} & 0 \\ \end{matrix} \right) $ is $x^{n_1-n_2} \times |xI_{n_2}-a_{21}a_{12}|$.

Since there is no further restriction on $A$, I think the answer should be general.

$\endgroup$
2
$\begingroup$

Even the conjugacy classes with respect of block diagonal matrices $\text{diag}\,(a_{11},\ldots,a_{ee})$ are known. More precisely:

The matrix $A$ can be regarded as a cycle of homomorphisms $$ k^{n_1}\to k^{n_2}\to\ldots\to k^{n_e}\to k^{n_1}. $$ So it is a representation of the cyclic quiver of type $A_{e-1}^{(1)}$ whose representation theory is well known.

The indecomposable nilpotent blocks look as follows: pick $i$ between $1$ and $e$ and a nilpotence degree $N$. Then a basis of the vector space is $v_1,\ldots, v_N$ and $A$ acts as $Av_\nu=v_{\nu+1}$ and $Av_N=0$. Moreover $k^{n_j}$ is spanned by all $v_\nu$ with $j\equiv \nu+i-1 \text{ mod }e$ (so the $v_\nu$ circle around with period $e$ and starting point $i$). From this it is a purely combinatorial problem to find all possible Jordan forms. For example, a single Jordan block is only possible if the dimension vectors are of the form $(a+1,\ldots,a+1,a,\ldots,a)$ or a cyclic permutation thereof.

You can find more, e.g., in Kempken's thesis "Eine Darstellung$\ldots$" MR0666395 (84a:14022) of 1982.

$\endgroup$
1
  • $\begingroup$ Can you recommend a good elementary source for the explicit representation theory of quivers? $\endgroup$ – LSpice Oct 21 '16 at 16:16

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.