I encountered a problem as folows:

Show a $3\times 3$ real matrx $A$, such that


well, this problem is not difficult, one can first find $B=\left(\begin{array}{ccc}\sqrt3&0&0\\0&\sqrt3&x\\0&0&0\end{array}\right)$ such that $B^2=\left(\begin{array}{ccc}3&0&0\\0&3&1\\0&0&0\end{array}\right)$.

My problem is:

Let $m,n$ be two positive integers. then, for what $n\times n$ real matrix $X$, there exist real matrix $A$ such that $A^m=X$? Is there a general method or theorem to calculate all the matrices $X$ and $A$?

Maybe, there does not exist a general answer. then, How about $n=3$ or $4$?

Thanks a lot!

  • 1
    $\begingroup$ Everything over $\mathbb C$? $\endgroup$ – Git Gud May 17 '14 at 15:20
  • $\begingroup$ @GitGud over $\Bbb C$, $A$ always exist $\endgroup$ – Clin May 17 '14 at 15:24
  • 1
    $\begingroup$ That is false, just take $X=\begin{bmatrix} 0 & 1\\0 & 0\end{bmatrix}$ and $m=2$. $\endgroup$ – Git Gud May 17 '14 at 15:27

If you're working over $\mathbb{C}$, then everything boils down to the Jordan canonical form. I actually needed to look this up a while ago and there's a very good paper on the topic that is completely elementary (needs only knowledge of the Jordan canonical form):



Decompose $\mathbb{C}^n$ into generalized eigenspaces. Then if you find such an X for each eigenspace, you can combine them together to give an X for $\mathbb{C}^n$.

Therefore we can assume that $\mathbb{C}^n$ is a generalized eigenspace. Then we can choose a basis so that the matrix will be upper triangular. If the eigenvalue is non-zero, it will be simple to find such a matrix, one row at a time, starting from the bottom.

The generalized eigenspace for 0 is more complicated, and X might not exist.

So a partial answer is that if the matrix is invertible, X will always exist.


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.