MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $A$ and $B$ be two $n \times n$ full-rank matrices.

Let $XAY = B$ be the given equation where $X$ and $Y$ are unknown $n \times n$ matrices. We know that $Vec(B) = (Y^{T} \otimes X)Vec(A)$. Under what conditions can we determine $X$ and $Y$ and what would be the procedure to determine $X$ and $Y$?

Not looking for the obvious solutions such as $Y^{-1}=A$ and $X=B$, $A$ is the eigenvalue (or some multiple of eigenvalue) diagonal matrix of $B$ where $X$ and $Y$ diagonalizes $B$ and $A$ and $B$ are similar to $X$ and $Y$. In general are there any other possible choices that $Y^{T} \otimes X$ can take? These may be the only choices for general $A$ and $B$. However, given $A$ and $B$ are some other structured matrices, has such an equation been studied before?

share|cite|improve this question
$X=I$ and $Y=A^{-1}B$? or am I missing something? – Suvrit Aug 31 '11 at 0:18

Since $A$ and $B$ are invertible, replace $X$ by $X' = XA$ and replace $Y$ by $Y' = YB^{-1}$. Then your equation reduces to solving $X'Y' = I$, i.e. the space of solutions is parametrized by $GL(n)$.

share|cite|improve this answer
I was actually curious of a meaning behind Theorem 9.3.5 here:… – user16007 Aug 31 '11 at 4:23
No idea about that, sorry. I can't see the second half of the theorem since Google is not showing me page 311. – MTS Aug 31 '11 at 4:48
I think you need only pages 309 and 310! – user16007 Aug 31 '11 at 11:27

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.