HKUST1 MOF Convergence Issue

[Vincent Luo]

Hello all,

I am having the following issue when trying to converge to the AFM (open shell singlet) ground state of the HKUST-1 MOF with Cu as the metal center (HKUST1_primitive.xyz).

I arrive at a converged state using UKS+OT (EPS_SCF=1.0E-06) and proper initialization. According to the spin moment in the Mulliken population analysis, the neighboring Cu atoms do seem to have opposing spins, therefore suggesting the correct state. The relevant files are OT.inp and OT.out.

However, when I feed the OT wavefunction (OT-RESTART.wfn) as initial guess to a direct diagnolization SCF for subsequent calculations, it does not converge in 1 iteration, but gradually becomes unstable (ending with "KS energy is an abnormal value (NaN/Inf) error). Changing the mixing parameters seems to have no effect at all. The relevant files are diag.inp and diag.out

How can this be? Does this mean that the OT did not converge to the correct state? Or is the charge mixing (and possibly other scf tricks in diagonalization) doing something weird?

I also tried doing one diagonalization step on the OT result, and then feeding:; that back to OT, and then diagonalizing again, etc... The energy (and possibly the state) barely changed throughout this whole process. But still, the diagonalization SCF just does not converge... Could degeneracies have some effect on this?

Thanks,

Vincent