I see in the release notes that it looks like DF-Lambda-CCSD(T) is available in psi4. Two questions:
How do I use this feature? I can’t seem to find it in the documentation.
Does it have support for densities? Similar to DF-CCSD(T), I think this means just having the gradient implemented.
They keyword and available implementations are listed at the very bottom:
This should be enough:
set cc_type df
No gradients. And a quick glance at the code suggests no density. But @bozkaya would know best.
The following two options are necessary:
set cc_type df
set qc_module occ
Thanks, @hokru and @bozkaya!
@bozkaya, are you saying that calling
energy('ccsd(at)') will give me a DF-Lambda-CCSD(T) density matrix? I know for other coupled cluster methods in psi4, you have to call
gradient to get a density.
@joshrackers @hokru For DF-Lambda-CCSD(T) we have only energies, hence we do not have densities.
@bozkaya that’s what I figured. I’m sure implementing the gradient is a significant amount of work!
That being said, is there any reason that I couldn’t just construct the density matrix from the returned wavefunction from
energy using psi4numpy? I don’t need the gradient, just the density itself.
In order to form PDMs you need to T- and Lambda-amplitudes, which are written to disk, not saved to wfn. Hence, you cannot get them from returning wfn. However, if you disable psi.clean() function you may try to read them in from the DFOCC files. @joshrackers
I’ll warn you that to get the same density matrix as from gradient theory, at the very least, you would also need to solve the orbital Z-vector equations. An unrelaxed density matrix might be doable, though.
Anyways, this topic seems answered now. It’s good practice to mark a post as “solving” the topic when it’s answered, so that way we know there’s nothing else we need to do here. There should be a checkbox on each post to mark it as “solving” the topic.
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.