2

My lecture notes write that two matrices are said to be similar if there exists an $n \times n$ invertible matrix $P$ such that $B = P^{-1} AP$.

But suppose I am given two matrices, $A$ and $B$, how do you 'show'/'determine' whether they are similar? Do you just have think up a matrix $P$ or is there a systematic way of finding it?

A second question I have if I have a matrix $A$ and find its eigenvectors, and place these in my similar matrix $P$, do I have to normalise my eigenvectors so that the matrix $P^{-1} AP$ equals a diagonal matrix, whose diagonals are the eigenvalues of $A$?

Thanks.

PhysicsMathsLove
  • 3,142
  • 1
  • 21
  • 39

1 Answers1

1

Question 1: The answers to this question answers it a lot better than I ever could.

Question 2: You do not have to normalise the eigenvectors. If you, say, keep one eigenvector long (which makes the corresponding column in $P$ have large values), then that is made up for in $P^{-1}$, so the diagonal values in the diagonal matrix are the eigenvalues of $A$ regardless.

Arthur
  • 199,419
  • Why do textbooks normalise the eigenvectors? Is there ever a reason? – PhysicsMathsLove May 18 '17 at 10:49
  • 1
    @PhysicsMathsLove Some don't like the fact that there are infinitely many eigenvectors to choose from. If you normalise it, there are only two (unless your elements are complex). If $A$ is symmetric, and you are diagonalising it, then making sure $P$ has normalised columns makes $P$ an orthogonal matrix (unitary in the complex case), with correspondingly nice properties (like its (conjugate) transpose being its inverse). Mostly it's just esthetic reasons like these. – Arthur May 18 '17 at 10:53