How can we explain that the QR iteration for approximating to the eigenvalues of a symmetric matrix A annihilates the off-diagonals?

Similar questions and discussions