Solution for Sylvester Equation

Given n by n matrices A, B, C. I know how to solve the Sylvester equation

AX + XB + C = 0

using the matlab command >> X=lyap(A,B,C)

But how do we solve the extended Sylvester equation
AX + XB + CXD + E = 0 ?

Either numerical or analytical method I'm willing to learn.
 

chiro

Science Advisor
4,783
128
Given n by n matrices A, B, C. I know how to solve the Sylvester equation

AX + XB + C = 0

using the matlab command >> X=lyap(A,B,C)

But how do we solve the extended Sylvester equation
AX + XB + CXD + E = 0 ?

Either numerical or analytical method I'm willing to learn.
Hey matematikawan.

Have you tried just expanding out the system, collecting the terms and getting a form of AX = B?

In other words you get a matrix corresponding to ZX = F and then apply the formula X = Z^-1 x F. For the Z matrix you will need to do some algebra to get this and in this particular example, F = -E.
 
I don't think it is possible to express it as AX=B.
Even to solve the Sylvester equation you have to diagonalize the matrices.
 

chiro

Science Advisor
4,783
128
I don't think it is possible to express it as AX=B.
Even to solve the Sylvester equation you have to diagonalize the matrices.
Try pen and paper first instead of using a computer.

What will happen is that when you collect everything together you should get a linear system in terms of your X and some matrix that is premultiplied by it. Once you have separated the matrix from your X by specifying what that matrix is then you can do normal inversion techniques.

You might have to write the algorithm yourself after doing a pen and paper derivation, but the idea doesn't change.

Also when you expand out everything using algebra, I'm sure you'll find conditions for when this does not hold, possibly even as a function of A, B, and C.

Again I urge you to do the pen and paper algebraic computation if you can't use any other known results.
 

Related Threads for: Solution for Sylvester Equation

Replies
10
Views
3K
Top