ROHF and semicanonicalization with pno-ccsd

[fsymalla]

Dear Turbomole experts,
i'm simulating the cations of aromatic molecules with ccsd and pno-ccsd.
If i do UHF simulations the results are practically identical for my test systems.
If i do ROHF this is not the case, because CCSD does a semi-canonicalization (according to the output) which pno-CCSD does not do.
Now, if i take the sc-orbitals from a normal rohf mp2 calculation (ricc2 module) and use them as input for an uhf pno-ccsd calculation the results are much closer to rohf-ccsd  (~50meV) but not the same.
I wonder what is happening except the semi-canonicalization in pno-ccsd compared to ccsd that does create this difference in agreement when switching from UHF to ROHF ?
Is it an oversight that pno-ccsd does not do the semi-canonicalization or is there a reason ?

Best
Franz

[christof.haettig]

As stated in the documentation, the pnoccsd code provides implementations of CCSD and CCSD(T) with PNOs for RHF and UHF references. CCSD with ROHF reference wavefunctions is not yet implemented in pnoccsd. I'm sorry that we forget to enforce that pnoccsd stops with an error message if you try to do calculations with an ROHF reference wavefunction.