MO energies difference vs PySCF for triplet systems

[mmutugi]

Hey @scemama,

I have been running Hartree-Fock calculations on PySCF, QP2, and Orca, and I realized that while they end up at the same energy, QP2 orbital energies are different from the other two for triplet systems. However, the orbital energies match perfectly for closed-shell systems and open-shell singlets. Are there any HF solver options or optimizations in QP2 that would closely track the PySCF answers for the triplets?

Thanks

[scemama]

Hello @mmutugi ,

the Fock matrix in ROHF calculations is not uniquely defined. We chose the definition of Plakhutin, Gorelik and Breslavskaya (https://doi.org/10.1063/1.2393223) because it restores Koopman's theorem. Therefore, MP2 energies and initial guesses of coupled cluster amplitudes with ROHF orbitals are more consistent with RHF.

It is not complicated to change these parameters. I will try to add an option quickly.

[scemama]

Look at the dev-stable branch: I have added an option rohf_parameters in scf_utils. The default is Roothaan. You can try the other parameterizations to find which one is consistent with PySCF and Orca. You can have a look at the parameters here:

qp2/src/scf_utils/fock_matrix.irp.f

Line 1 in fa28043

BEGIN_PROVIDER [ double precision, Fock_matrix_param, (2,3) ]

[mmutugi]

Thank you so much @scemama, problem solved. Guest parameterization matches PySCF.
qp set scf_utils rohf_parameters Guest