A numerical method for computing eigenvectors of a large matrix
H. Q. Xu · Journal of Physics A Mathematical and General · 1991
Introduces a new numerical method to determine eigenvectors of a Hamiltonian matrix. The method is particularly useful for matrices of large dimension. The essence of the method is to determine each phase in the expansion of an eigenvector by computing the projection weight to the eigenvector on a trial function with an arbitrary phase. The trial function depends on the unknown phase in the expansion of the eigenvector, and it is shown that the weight takes its maximum value if the phase in the trial function takes the correct value of the phase in the expansion of the eigenvector.