0

How can we get an invertible matrix $W \in$ $\mathbb{R}$$^{3\times3}$ to make $W\hspace{-1mm}AW^{T}$ a diagonal matrix? Here,

$$ A = \begin{bmatrix} 2 & 1 & 3 \\ 1 & 0 & 1 \\ 3 & 1 & 3 \end{bmatrix} $$

I try to compute the eigenvalues of $A$, but it is difficult to factor the polynomial $t^3 - 5t^2 - 5t - 1 = 0$. I've also attempted to compute its SVD, but the polynomial becomes more complicated. Do we have other ways to solve it?

Darius
  • 1,111
Jonny
  • 433
  • 2
  • 9
  • As Lord Shark the Unknown mentioned, you don’t need the eigenvalues to diagonalize this matrix. See https://math.stackexchange.com/q/1388421/265466. – amd Apr 24 '19 at 06:00

1 Answers1

0

You only need to find the eigenvectors if you insist that $W$ must be orthogonal. Basically here we are completing the square. First we find a matrix $$M=\pmatrix{1&0&0\\*&1&0\\*&0&1}$$ such that $$MAM^T=B=\pmatrix{2&0&0\\0&*&*\\0&*&*}.$$ We can take $$M=\pmatrix{1&0&0\\-1/2&1&0\\-3/2&0&1}$$ and then $$B=\pmatrix{2&0&0\\0&-1/2&-1/2\\0&-1/2&-3/2}.$$ Now, we seek a matrix $$N=\pmatrix{1&0&0\\0&1&0\\0&*&1}$$ to make $NBN^T$ diagonal. Then $W=NM$ suffices.

Angina Seng
  • 158,341