For what conditions on $A$, $B$ and $C$ (square matrices of size $n$) would there be a unique solution to $$ ABX + AXC + XBC = D, $$ for any $D$? Can one expect a characterization similar to the Sylvester Theorem, which states that there always exists a unique solution to $AX + XB = C$, for any $C$, if and only if $A$ and $-B$ do not share an eigenvalue? And then, can this be extended to equations of the form $ABCX + ABXD + AXCD + XBCD = E$, and beyond?

  • 7
    $\begingroup$ I didn't even know Sylvester was in the Army, much less that he held the rank of a General. $\endgroup$ – Gerry Myerson Jan 26 at 3:57
  • 1
    $\begingroup$ @Gerry Myerson may be he meant the concept of generalization in mathematics. $\endgroup$ – Ahmad Jamil Ahmad Masad Jan 26 at 6:48
  • 1
    $\begingroup$ @GerryMyerson: of course, he was: en.wikipedia.org/wiki/John_B._Sylvester $\endgroup$ – M.G. Jan 26 at 11:50
  • 1
    $\begingroup$ @AhmadJamilAhmadMasad: no need to apologize, it was a joke :) The general is a different Sylvester. Or is he?... $\endgroup$ – M.G. Jan 26 at 12:14
  • 2
    $\begingroup$ @GerryMyerson. I heard this joke 20 years ago, from Efimov, à propos of his proof of the General Burnside conjecture. $\endgroup$ – Denis Serre Jan 26 at 14:35

Can one expect a characterization similar to the Sylvester Theorem

As far as I know, no, apart from very special cases where the coefficients can be triangularized simultaneously. There is a big gap in difficulty between the 2-term case (where there is a canonical form for pairs of matrices, an $O(n^3)$ algorithm, etc.) and the 3-term case, when there is basically nothing apart from turning it into a $n^2 \times n^2$ linear system.

Even in your case where your coefficients have a special form no particular simplifications spring to mind. As far as I know, a nice characterization is an unsolved (and very likely unsolvable) problem.

  • 1
    $\begingroup$ Maybe a more definite negative answer (for fixed $n\ge 2$) would be to show that the characterization does not only depend on the conjugacy classes of $A$, $B$, and $C$. $\endgroup$ – YCor Jan 26 at 8:20

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.