search

gabrielasd / pyeom

GNU General Public License v3.0
2 stars 0 forks source link

Weird energies and normalzation in ph-EOM-RPA #51

Closed PaulWAyers closed 3 years ago

PaulWAyers commented 3 years ago

The 2DMs in EOM-RPA seems to be not normalized or give the right enregies for HF.

If no commutator is used in the EOM equation (CIS) and a commutator is used in the 2DM reconstruction (TDM evaluation) then normalization and energy are OK. No commutator in the TDM evaluation is really bad for both normalization and energy evaluation, and if commutators are used both palces, it looks like RPA but the correlation energy is positive instead of negative, which is unexpected.

gabrielasd commented 3 years ago

Please see branch https://github.com/gabrielasd/eomee/tree/erpa for the code updates and scripts I've been using to test eRPA

@PaulWAyers

gabrielasd commented 3 years ago

This issue was solved by commit bab6293749287f3214856f4fd676c8338f8247c1 It was related to numerical innacuracies when using the eigenvectors from the whole ph-EOM eigenvalue spectrum. Only the eigenvectors corresponding to the positive eigenvalues should be used in the transition RDMs evaluation.