Oliver Backhouse
Oliver Backhouse
Hi Peter, thanks for spotting this - I have addressed it in #120. You were correct that too many vectors were being returned from `davidson` - though `v_ip[:, 0]` should...
I have tried different combinations of integrals and Davidson solver from `pyscf` and `psi4` with the `psi4numpy` ADC equations, and all are in good agreement with the eigenvectors from `pyscf`'s...
> I'll have to check my spin-summation it's probably why I get incorrect results. I'm happy from your results that all is OK with psi4numpy now. @pwborthwick no worries -...
@pwborthwick yes, `v_ip` has shape `(nocc + nocc*nocc*nvir, nroots)`. I projected that into the singles space for a particular root `i`, i.e. `v_ip[:nocc, i]`, and then took the norms over...
Yes, this look like the Davidson solver in `psi4numpy` is converging such that the character of the ionisation potential is completely due to that MO and no other excitation -...
@pwborthwick good catch! That is another bug. Seems like I wrote a really buggy solver...
Are there any plans and/or updates regarding the implementation of this?
Great - thanks for your response
Great - I will check this out in the week.
> I don't really _need_ it for anything urgent, moving the gradients forward seems more important right now. (EOM)-CC is more of a fun playground project smile I actually implemented...