Remarks on large-scale matrix diagonalization using a Lagrange–Newton–Raphson minimization in a subspace

Abstract.

We present a matrix diagonalization method where the diagonalization is carried out through a normal Lagrange–Newton–Raphson method solved in a subspace. The subspace is generated using the correction vector that predicts the standard Lagrange–Newton–Raphson formula in the full space. Some numerical examples and the performance of the algorithm are given.