KROHF not working

[efertitta]

I am encountering problems when running KROHF. It appear (error below) that there is a problem of indexing of the orbitals? The same script with KUHF runs seamlessly

File "/cluster/projects/nn8040k/SEP-QUEST/LiNiO2/C2m_KROHF/Embedding.py", line 76, in
emb = vayesta.ewf.EWF(mf, solver='CCSD', bath_options=dict(bathtype='dmet'))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/Vayesta/vayesta/ewf/init.py", line 23, in EWF
return UEWF(mf.to_uhf(), *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/Vayesta/vayesta/ewf/ewf.py", line 70, in init
super().init(mf, solver=solver, log=log, **kwargs)
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/Vayesta/vayesta/core/qemb/qemb.py", line 294, in init
self.init_mf(mf)
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/Vayesta/vayesta/core/qemb/qemb.py", line 361, in init_mf
self._veff_orig = self.mf.get_veff()
^^^^^^^^^^^^^^^^^^
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/Vayesta/vayesta/core/foldscf.py", line 126, in get_veff
vk = self.kmf.get_veff(self.kmf.mol, dm, *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/pyscf/pbc/scf/kuhf.py", line 442, in get_veff
vj, vk = self.get_jk(cell, dm_kpts, hermi, kpts, kpts_band)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/pyscf/pbc/scf/khf.py", line 529, in get_jk
vj, vk = self.with_df.get_jk(dm_kpts, hermi, kpts, kpts_band,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/pyscf/pbc/df/mdf.py", line 174, in get_jk
vk = mdf_jk.get_k_kpts(self, dm, hermi, kpts, kpts_band, exxdiv)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/pyscf/pbc/df/mdf_jk.py", line 83, in get_k_kpts
vk_kpts = df_jk.get_k_kpts(mydf, dm_kpts, hermi, kpts, kpts_band, None)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/pyscf/pbc/df/df_jk.py", line 308, in get_k_kpts
skmoR, skmoI = _format_mo(mo_coeff, mo_occ, shape=(nset,nkpts), order='F',
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/pyscf/pbc/df/df_jk.py", line 1336, in _format_mo
mos = [mo[:,mocc>precision]*mocc[mocc>precision]**0.5
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/cluster/home/edoardof/.local/lib/python3.11/site-packages/pyscf/pbc/df/df_jk.py", line 1336, in
mos = [mo[:,mocc>precision]*mocc[mocc>precision]**0.5
~~^^^^^^^^^^^^^^^^^^

[maxnus]

Hi Edoardo,
We do not support restricted open-shell mean-field objects; you will have to convert it to a UHF class (after converging to the ROHF solution).
Since UHF is able to represent any ROHF solution, I believe that "CCSD@(ROHF->UHF)" should give the same result as "CCSD@ROHF", just perhaps be a bit less efficient. I think even PySCF does not have a RO-CCSD solver and just defers to UCCSD?

[efertitta]

Hi Max,
Understood, however in the molecular case EWF@ROHF works. EWF does convert it to UHF as you say and invokes the UCCSD solver. This is not the case with the periodic code though.
In the error above, mf.to_uhf() is invoked (same as it happens in the molecular code) but I am not sure why it won't work
I will try to convert KROHF->KUHF before passing it to emb

[basilib]

#156 adds automatic conversion to UHF