3

Is it possible to solve for a matrix $A$ in an equation involving a matrix similar to $A$, of the form

$$A + P^{-1}AP = B$$?

The solution I'd be looking for would be for $A$ in terms of $P$ and $B$, ideally.

EDIT: Thanks for the existing answers! Just to add, in the particular case I'm trying to solve, P is diagonal.

Adam
  • 31

1 Answers1

3

If you pre-multiply both sides by $P$ you end up with a Sylvester equation - see the Wikipedia entry here:- $$PA+AP=PB$$

Suppose $P$ is an $n\times n$ diagonal matrix:- $$P=\left(\begin{array}{cccc} p_1 & 0 & \cdots & 0\\ 0 & p_2 & \cdots & 0 \\ \vdots & \vdots & \ddots & \vdots \\ 0 & 0 & \cdots & p_n \end{array}\right)$$ and matrices $A$ and $B$ are given by $$A=\left(\begin{array}{cccc} a_{11} & a_{12} & \cdots & a_{1n}\\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & \vdots & \ddots & \vdots \\ a_{n1} & a_{n2} & \cdots & a_{nn} \end{array}\right)$$

$$B=\left(\begin{array}{cccc} b_{11} & b_{12} & \cdots & b_{1n}\\ b_{21} & b_{22} & \cdots & b_{2n} \\ \vdots & \vdots & \ddots & \vdots \\ b_{n1} & b_{n2} & \cdots & b_{nn} \end{array}\right)$$ Then (using the Kronecker product formulation, and after some manipulation) we end up with a relationship between column $k\in\{1,2,\cdots,n\}$ of $A$ (denote as $A_{k}=[a_{1k},a_{2k},\cdots,a_{nk}]^T$) and column $k$ of $PB$ (denote as $[PB]_{k}=[p_1b_{1k},p_2b_{2k},\cdots,p_nb_{nk}]^T$) as follows:- $$(P+p_kI)A_{k}=[PB]_{k}\Rightarrow A_{k}=(P+p_kI)^{-1}[PB]_k$$ where $I$ is the $n\times n$ identity matrix, so that $P+p_kI$ is a diagonal matrix.

Alijah Ahmed
  • 11,609
  • Thanks! That's helpful. If P is diagonal, does this make solving for A easier? – Adam May 26 '15 at 16:57
  • It would simplify things considerably, as the large matrix in the rewritten Sylvester Equation (that involves Kronecker products) (in http://en.wikipedia.org/wiki/Sylvester_equation) will be diagonal too. – Alijah Ahmed May 26 '15 at 17:12
  • Thanks! Do you know of any resources for finding the solution? – Adam May 26 '15 at 18:04
  • To clarify: I'm looking for a symbolic solution, i.e., something I can write explicitly in terms of P and B. – Adam May 26 '15 at 19:36
  • @Adam: Please see further edits to my answer. – Alijah Ahmed May 27 '15 at 16:45