- The calculation of the orbital shifts was not computed with a consistent
Hamiltonian. The Hamiltonian is now updated at each OEP step ad show a much better convergence.
- The Hartree potential in the Hamiltonian used to compute the orbital shifts is now consistent with
the right-hand side.
- The OEP potential is not saved between different SCF iterations, as this lead to an inconsistent mix of it. Instead, we solve first KLI after each SCF step, and iterate from it to get the OEP potential.
- The constant shift is not mixed anymore but it only added at the end.

These changes show nice and quick convergence compare to the extremely slow behavior we had before.