-3

Let $A,B \in \mathcal{M}_{n\times n}(\mathbb{K})$, where $\mathbb{K}$ is a field. If $m^A(t)=m^B(t)=q(t)$ irreducible in $\mathbb{K}[t]$, then $A$ is similar to $B$?

For instance, the following two matrices

$$ A=\begin{pmatrix} 0 & 1 & 0 & 0 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 0& 0 \end{pmatrix} \qquad B=\begin{pmatrix} 0 & 1 & 0 & 0 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0& 0 \end{pmatrix}$$

have $m^A(t)=m^B(t)=t^2$ and they are not similar but the minimal polynomial is reducible in $\mathbb{Q}[t]$, so this example doen't answer to my question.


Proof. It is known that $A$ is similar to $B$ if and only if $I_{n,n}t-A$ is equivalent to $I_{n,n}t-B$ in $\mathcal{M}_{n\times n}(\mathbb{K}[t])$. We can compute Smith's canonical form of $I_{n,n}t-A$ and $I_{n,n}t-B$. If $f_1(t),\dots,f_n(t)$ are the invariant factors of $I_{n,n}t-A$, then $f_n(t)=m^A(t)$ and, since $f_i|f_{i+1}$ for all $i=1,\dots,n$, so for the irreducibility of $m^A(t)$ we have $f_i(t)=1$ for all $i=1,\dots,k$ and $f_i(t)=m^A(t)$ for all $i=k,\dots,n$, where $k$ is an integer less than $n$. The same holds for $I_{n,n}t-B$, so we denote by $g_l,\dots,g_n$ the non-costant invariant factors of $I_{n,n}t-B$, with $1\leq l\leq n$.
Since $l\deg{m^B(t)}=n=k\deg{m^A(t)}$ and $m^A(t)=m^B(t)$, so $l=k$. Hence the characteristic matrices of $A$ and $B$ have the same Smith's canonical form. Therefore $I_{n,n}t-A$ is equivalent to $I_{n,n}t-B$ in $\mathcal{M}_{n\times n}(\mathbb{K}[t])$, so $A$ is similar to $B$.

I would like to thanks Lukas Heger for the useful suggestions.

Hola
  • 151
  • You can show this via Smith canonical forms, but it's not necessarily true that $q$ appears only once as an invariant factor. But you can still know how often $q$ will appear (and hence how many $1$s) will appear because the product of all invariant factors must have degree $n$. – Lukas Heger Sep 21 '21 at 13:59
  • I follow your suggestions and I edited the proof. I thank you very much! – Hola Sep 21 '21 at 14:44

1 Answers1

3

This is true. One way to see this is to use Frobenius normal forms: all blocks must be the companion matrix to $q$ if the minimal polynomial is irreducible.

Or to phrase it in terms of $\Bbb K[T]$-modules: if the annihilator of a $\Bbb K[T]$ finite-dimensional over $\Bbb K$ is $(f)$ for $f$ irreducible, then it must be a direct sum of copies of $\Bbb K[T]/(f)$ and from the dimension we can see how many such factors there are.

Lukas Heger
  • 20,801
  • Thank you very much... I edited the question to show some arguments, which I hope they are right – Hola Sep 21 '21 at 13:55