Is there any good way to do this in psi4?
With DF-CCSD the scf_type needs to be df as well, but there is no aug-cc-pv6z-JKFIT basis… My idea was to run conventional scf and read it in to a DF-CCSD calculation with scf maxiter set to 0, but I don’t know density fit coupled cluster well enough to know if that’s good idea or not… in particular I don’t know what happens to e.g. the <OV|VV> integrals.
in my experience the RI basis sets are ok for JKFIT purposes:
As you’re using a 6-zeta basis I’m going to assume that high-accuracy is important to your investigation. Because of that, I’d suggest that you at least benchmark how large the density fitting error (at the HF level) is, perhaps for a smaller but related system. At this point the DF error (without optimal auxiliary sets) may start to approach basis set incompleteness errors.
For reference, it looks like the DF errors under these circumstances are still completely negligible. On the order of 10^-2 kcal/mol for absolute energies of small molecules, and 10^-4-10^-5 kcal/mol for difference properties like atomization energies.