[Wien] SDW calculation ?

[Stefaan Cottenier]

> My question is regarding the spin structure for a SDW calculation. As
> far my knowledge, the unit cell need to be extended in a particular
> direction, equivalent or greater than the wave vector, i.e. in case
> of Cr, the unit cell length has to be 50 A, which is the magnitude of
> the wave vector.

Right, with this caveat: in Cr the wave vector is incommensurate with 
the lattice. As you can create only supercells that are an integer 
multiple of the primitive cell, you cannot reproduce exactly the desired 
wave vector.

> But how to set up the spin configuration ? Whether
> the alternate atoms should be + -+ - or half of the length is  + and
> another half - ? Another thing is, symmetry may not allow to the
> desired spin states while setting up the case.

Make yourself a 1D drawing of all (e.g. 40) atoms in the supercell, with 
the moments they should have in the SDW (cosine with amplitude 2, see 
below). As far as I remember (didn't check) you will see from that that 
atoms at (0,0,+/-z) have the same moment, and therefore will be 
equivalent. Make a case.struct with these 20 inequivalent blocks. 
Generate a default case.inst, which will have initially these occupations:

Cr
Ar 3
3, 2,2.0  N
3, 2,2.0  N
3,-3,1.0  N
3,-3,0.0  N
4,-1,1.0  N
4,-1,0.0  N

This is a spin moment of 2. For each particular atom, read from your 1D 
drawing what its moment in a cosine wave would be (say 0.45) and modify 
this part of case.inst accordingly:

Cr
Ar 3
3, 2,2.0  N
3, 2,2.0  N
3,-3,0.61 N
3,-3,0.39 N
4,-1,0.61 N
4,-1,0.39 N

(Verify that this configuration has a spin moment of +0.45) Don't forget 
to give the proper atoms the opposite moment.

In this way, you have a starting density with a cosine-modulated wave. 
This is close enough to the selfconsistent SDW solution, such that 
you'll obtain that SDW at the end of the scf-cycle.

Stefaan