Continued-fraction solution of matrix equation \(AX-XB=C\) (Q584345): Difference between revisions

The matrix equation \(AX-XB=C\) has been much studied. Assuming that f(\(\lambda)\), g(\(\lambda)\) are coprime polynomials satisfying \(f(A)C=Cg(B)=0,\) the author derives another finite series solution. The new twist is that this solution involves the convergents to the finite continued fraction expansion \(g(\lambda)/f(\lambda)=g_ 0(\lambda)+\frac{1}{e_ 1(\lambda)}+\frac{1}{e_ 2(\lambda)}+...+\frac{1}{e_ r(\lambda)}.\) It is shown that these convergents are also of interest in other related contexts. For example, when \(g_ 0(\lambda)\) is a nonzero constant and the \(e_ i(\lambda)\) have degree 1, they have determinantal expressions akin to Hurwitz determinants. Again, in the case \(A=-B^*\) they intervene in a neat formula for the Lyapunov function.