2

I have a question: Let $A, B, C$ be $n$-order square matrices. It is easy to prove that $AB$ and $BA$ have the same eigenvalues. Therefore, we can deduce that $ABC$ has the same eigenvalues as $BCA$ and $CAB$. My question is, if the order is not cyclic, will there be a similar result? Or are there other conclusions? Specifically, what is the relationship between the eigenvalues of $ABC$ and $BAC$? Or can we add some conditions to obtain some interesting conclusions?

William
  • 31

1 Answers1

1

You may want to check this answer out:

https://math.stackexchange.com/a/2163596/1270050

In my opinion, looking into other permutations other than the cyclic ones poses a problem because for general $n \times n$ matrices, $\mathcal{A}$ and $\mathcal{B}$, the relationship between their eigenvalues to $\mathcal{AB}$ still remains unknown according to my research:

In most cases where a certain relationship exists, matrices $\mathcal{A}$ or $\mathcal{B}$ are assumed Hermitian, or positive (semi)definite, with additional restrictions if needed (first one is not really useful for this problem):

Then there are additional cases when matrices $\mathcal{A}$ and $\mathcal{B}$ commute:

But nothing can be said about $n \times n$ matrices in general. So nothing really can be deduced for the case $ABC$ and $BAC$ in general.

If it were given $A,B,AB$ are Hermitian, then $A,B$ commute and hence $ABC = BAC$

To generalise this further (but not to any $n \times n$ matrices), if $A,B$ are simultaneously diagonalisable (of which Hermitian matrices $A,B,AB$ is a special case), then $A,B$ commute (mentioned in the Wikipedia link)

  • If you've researched this relationship between eigenvalues, it would be more helpful to share the fruits of that effort, such as an example. The present broad brush claim is not enlightening. – hardmath Mar 23 '25 at 23:32
  • @hardmath I apologise for ambiguity. I hope this answer is more acceptable – Idris Lokhande Mar 24 '25 at 01:55