WIEN2k-FAQ: Setup of antiferromagnetic calculations?

©2001 by P. Blaha, K. Schwarz and J. Luitz

Back to:

Setup of antiferromagnetic calculations requires some tricks which will be described below:
  1. You need to construct a unit cell which allows for the desired AF ordering. E.g. for bcc Cr you have to put "P" lattice and 2 NON-equivalent atoms at (0,0,0) and (0.5,0.5,0.5). In order to make nn working properly, you should also specify different labels in the "name(3:10)"-field for the two AF-atoms (Eg. "Cr1" and "Cr2").
  2. Run init_lapw. nn and/or sgroup determines equivalent atoms. Accept eventually the new struct_nn file (You may have to use the same procedure as described in faq:supercells).
  3. Edit case.inst and flip the spin of the second AF atom. This means you must invert the spin-up and dn occupation of this atom. Atoms which carry no spin (e.g. an O between two AF metal atoms) should be made non-spinpolarized by setting equal numbers of up and dn-electrons.
    For AFM NiO case.inst looks like: 
    Ni
Ar 3 5
3, 2,2.0  N
3, 2,2.0  N
3,-3,3.0  N
3,-3,1.0  N
4,-1,1.0  N
4,-1,1.0  N
Ni
Ar 3 5
3, 2,2.0  N
3, 2,2.0  N
3,-3,1.0  N   <=== spin flipped
3,-3,3.0  N   <=== spin flipped
4,-1,1.0  N
4,-1,1.0  N
O
He 3 5
2,-1,1.0  N
2,-1,1.0  N
2, 1,1.0  N
2, 1,1.0  N
2,-2,1.0  N  <=== equal occup for up/dn
2,-2,1.0  N  <=== equal occup for up/dn
****     End of Input
****     End of Input
    This case.inst file can also be prepared by the new: instgen_lapw -ask , where you can specify for each atom spin-up, dn or non-magnetic configurations.
  4. Continue with init_lapw (rerunning lstart) and select "spin-polarized" calculations.
  5. You may now run an scf cycle using runsp_lapw; or
  6. You may try to select "AFM-calculations", generate "case.inclmcopy" and use "runafm_lapw". For simplest use of "afminput" you should provide the nonmagnetic (or ferromagnetic) supergroup in "case.struct_supergroup". Please read the usersguide for more details. (help_lapw and search for AFM).

    If you succeed to generate proper input files, you can do the scf-cycle by "runafm_lapw", which will save at least half the computer time, since only one spin-direction is calculated and the other one is generated by symmetry. However, since these programs are little tested, I recommend to check the final results of "runafm_lapw" by

  7. "runsp_lapw". If the charge distance (grep :DIS case.scf) remains small and the magnetic moments do not change, runafm_lapw was successful, otherwise continue with runsp_lapw until convergence.

    Peter Blaha,