6

I know an Hermitian matrix is diagonalizable, and similarly a real symmetric matrix is diagonalizable, but what's wrong in a complex symmetric matrix.

Why does the Gram-Schmidt process fail?

Dac0
  • 9,504
  • I think, a complex symmetric matrix may or may not be diagonalizable. The title of your question suggests that a complex symmetric matrix is never diagonalizable. – Solidification Nov 14 '19 at 17:11

2 Answers2

9

First of all, there is an easy counterexample. The complex symmetric matrix $$\begin{pmatrix} 1 & i \\ i & -1 \end{pmatrix}$$ is not diagonalizable, because trace and determinant are zero, but the matrix is not zero. Now try the Gram-Schmidt process in this example.

Dietrich Burde
  • 140,055
  • Thank you for the counterexample but I still don't totaly get what wrong in the process. Indeed the process is fine since the two vectors are orthogonal and correctly gives the same two vectors – Dac0 Feb 16 '16 at 16:23
  • 3
    If $x$ is a complex vector, then $x^Tx=0$ does not imply that $x=0$. – Chris Godsil Feb 16 '16 at 16:27
  • 2
    With $x=(1,i)^T$ we have $x^Tx=0$, but not $x=0$. – Dietrich Burde Feb 16 '16 at 16:28
  • 1
    So the Gram Schmidt process itself would work, but the matrix would not be diagonal anyway? – Dac0 Feb 16 '16 at 16:29
6

As Chris Godsil and Dietrich Burde pointed out it's because $\langle x,y \rangle =x^*y=0$ which is the orthogonality condition on complex vectors does not imply that $x^Ty=0$ which is the complex symmetry condition.

So the Gram Schmidt process actually will produce orthogonal vectors, but they will not be able to diagonalize the matrix.

Dac0
  • 9,504