> Hi! > > 2008/4/15, Marcel Mohr <[EMAIL PROTECTED]>: >> >> The smaller energyshift, should in principle be better, and results in >> larger basis size. > > > That's not true. It only has to be better for an isolated atom, who knows > what would happen in condensed phase. However, you could use the > variational > principle: whichever basis gives you the lowest energy for a given > configuration makes the best approximation to the ideal KS solution (for a > given configuration, again).
I'd like to comment on this Vasilii's comment: in certain situations we can optimize the basis from a variational procedure, varying some constraints. However, with respect to PAO.EnergyShift this probably won't work, because the energy on less compressed orbutals will be always lower, giving no minimum - UNLESS in this search you compress the crystal, instead of orbitals, imposing a "correct" lattice constant. As usual with the basis: either you make your "reasonable choice" NOT based on variational search, or you choose a certain criterion (exp. lattice constant, or whatever) and perform a variational search subject to it. Best regards Andrei Postnikov
