TURBOMOLE Users Forum
TURBOMOLE Modules => Molecular Properties, Wavefunction Analysis, and Interfaces to Visualization Tools => Topic started by: jfurness on October 07, 2020, 08:49:44 PM

I am interested in studying a 2 electron system, therefore CCSD stands as very useful "basis set exact" energies for me.
I would like to compare the electron densities from other methods to a reference accurate electron density. I am hoping CCSD can provide electron densities as well.
So far I have tried using the "proper" program to generate plots of the electron density along the z axis. This works well for HF and DFT calculations, but when I run the same for CCSD I appear to get the density for the reference Slater determinant, rather than a correlated density. This is apparent as comparing the HF reference density and the CCSD density show they are exactly the same.
I would like to ask if it is possible to plot the correlated density from the CCSD solution?
I admit my knowledge of coupled cluster methods is not well developed, so maybe what I am asking is not meaningful. It seems to me that CCSD solves to find a correlated wavefunction as a combination of the reference ground state with weighted excitations, I would like to plot the electron density of that correlated wavefunction, if possible.
Any advice around this is appreciated.
EDIT:
It occurs to me this might not be the correct place to ask this. Please move the post if it is in the wrong section.

Further to this,
From my explorations, it seems I can only get densities from the approximate CC2 method.
If I include
$ricc2
geoopt model=cc2
$pointval
in the control file and run ricc2
then proper
finds the density as hoped.
However, if I try
$ricc2
geoopt model=ccsd
$pointval
then ccsdf12
reports "No gradient available for the requested model." and terminates.
I feel like this may be abusing the geoopt option. I do not want the forces on the nuclei (which I presume are the "gradients" in the program), I only want to plot the relaxed densities from the CCSD calculation.
I wonder then if there is a flag I am missing that will evaluate the relaxed density without worrying about the forces on the nuclei.
While using CC2 presents something of a solution, I am not familiar with the method and therefore reluctant to take it as a reference.

Hi,
I think the problem is that CCSD gradient has not been implemented and without it, you cannot get the relaxed density.
The "geoopt trick" you used to get the densities for CC2 is fine. Other option would be to use the $response keyword (subkeyword gradient).
Best,
Antti

Ok, thanks Antti!
It's great to get a firm answer to this. I was unaware that the gradients were a necessary step for calculating the relaxed density.
I will focus some time on studying CC2 to see if it can fit my needs. My suspicion is that it will be sufficient for qualitatively comparing electron densities.
Thanks,
James