joaquin peralta wrote: > The eigenvalues are not calculated, but why?
I know close to nothing on this specific subject, but, as a rule, in CP the eigenvalues are obtained as a byproduct of iterative orthonormalization (from diagonalization of the "Lambda" matrices, i.e. Lagrange multipliers enforcing orthonormalization constraints). If you use Gram-Schmidt orthonormalization, you do not have these matrices. Nothing would prevent the calculation of eigenvalues by some other algorithms (e.g. Davidson diagonalization), but as a matter of facts, this is not currently implemented in CP. It is a problem of "tradition", so to speak. Gram-Schmidt is inefficient and should nbot be used anyway, except in the first few time steps. Paolo -- Paolo Giannozzi, Democritos and University of Udine, Italy
