Sergey:
The Kohn-Sham total energy printed in the first
iteration is not variational, and therefore
cannot be trusted. It is based on occupying
atomic orbitals, which are not orthogonal.
The KS energy in the second and subsequent
iterations is variational and should never
be lower than the final energy, if the
selfconsistency iteration proceeds correctly.
The Harris energy is generally more reliable,
even in the first iteration, although
again it is not variational, and generally
increses during selfconsistency.
Jose M. Soler
