>> Generate a default case.inst, which will have initially these occupations: > > By the way, from general considerations I am afraid that this method with > supercell for pure Cr will give bad convergency; namely, no convergency at > all. One would possibly need to fix the position of the node of the wave, > somehow. Otherwise, the iteration cycle may jump between equal solutions.
We did that almost 10 years ago, and it did converge: http://dx.doi.org/10.1088/0953-8984/14/12/314 That was with the old mixer, and very small mixing values, many computers and a lot of patience were required ;-) Stefaan