[Columbus] ciudg to do single reference CI (CIS)

[Ron Shepard]

> On Nov 6, 2017, at 11:42 AM, SPIRIDOULA MATSIKA <smatsika at temple.edu> wrote:
> 
> Dear Hans and everyone else who sent some suggestions,
> 
> Thank you all! My systems is very simple: closed shell H2 +CIS.  So I only have 1 reference (the closed shell HF) and I add singles at the CI level.

I expect a problem with this calculation will be that the HF reference CSF will end up with a coefficient of zero. That will cause several problems during the calculation, such as C0=0 in the Davdison calculations, and the initial wave function will have zero overlap with the final converged wave function. A workaround might be to add another ag orbital into the internal space when specifying the CI drt. With the appropriate occmin+occmax restrictions, that will give you two internal configurations (one of which will be a single excitation), and those two internal configurations will be the z-walks (I think you must have at least one z-walk for the ciudg program to proceed). I expect the HF wave function will end up coming out at the end as the ground state, and then the other states will have just single excitation CSF coefficients. I also remember doing calculations like this where it was difficult to keep some of the double excitations out of the expansion. You need to use occmin(:) to force the first orbital to be at least singly occupied.

If you are computing wave functions not of Ag symmetry, then the internal orbital space will need to include the ag orbital along with an orbital of the desired symmetry. You will end up with only 1 z-walk in this case, rather than 2 as in the Ag case.

-Ron