XYG3 or XYGJ-OS?


#1

Dear Psi4 developers,

I’ve been using Psi4 for almost two years, mainly for SAPT calculations (one published work) and PR-ADC(2) calculations (work-in-progress). I must say great work here as I have thoroughly enjoyed the features.

I’m just wondering if anyone is bringing XYG3 or XYGJ-OS xDH functionals into Psi4?

Best Regards,
Desmond Toa


#2

We are dependant on https://gitlab.com/libxc/libxc for functionals. It does not appear that they have these functionals yet, best to ask there when they will get them in.

Edit: I vaguely remember something about these, do you have paper links for these functionals? We may be able to stitch them together from existing functionals.


#3

Good point. I will check with the developers behind libxc.

Papers links:



#4

You should be able to build them using our Python interface as we have all of the necessary components. You can see other double hybrids here.

It might be a bit before I will have a chance to try this out, but if you want to give it a go please join our slack and we would be happy to help!


#5

I want to point out that these functionals are a little different in that they are made of a density functional and an energy functional. The density functional (which is B3LYP in both of them) generates the orbitals and densities, which are then fed into the energy functional to obtain the energy of the electronic state.

So, I guess unlike the other double hybrids, the procedure would be to generate orbitals and densities from B3LYP and feed into the energy expression given in the two papers.

I will try to give it a shot over the the next few weeks.


#6

Interesting, that should not be too hard to do. If you can make a PR with the functional energy expressions and make a test for this class of functional similar to tests/dft-custom-dhdf/input.dat. I should be able to help hook in the separate orbital optimization/energy evaluation technology.