[Wien] How to determine magnetic configuration

[Stefaan Cottenier]

> My aim is to calculated case of MnF2 which has antiferromagnetic 
> configuration. I have defined struct file using Cmmm space group to 
> distinguish Mn atoms and set initial configuration in case.inst file (I 
> have flipped spins for one Mn to get antiferromagnetism) but my results 
> strongly depends on this initial configuration (if I set e.g. the same 
> spin for both Mn atoms, at the end I will get the results as it will be 
> ferromagnetic, and total energy is non higher than for afm). So my 
> question is: should I know exactly magnetic configuration of my case at 
> the beginning or it may be determine during scf cycle (and I am doing 
> something wrong)?

In the majority of cases, the *type* of magnetic configuration from 
which you start will be maintained in the self-consistent result: if you 
start from ferromagnetic, the result will be ferromagnetic. Only the 
*value* of the moments will change.

You say your energies for FM and AFM are identical. Are they really 
identical, or just very close? (energy differences of say 0.01 Ry are 
meaningful) Make sure that your k-mesh and basis set size are sufficient 
to obtain energy differences up to an accuracy better than the energy 
difference between AFM and FM.

Before doing anything else, check in your outpout (grep :MMI case.scf) 
whether the AFM and FM converged results are really AFM and FM (i.e. 
check the signs of the moments). If you did something wrong in the 
initialization, or if that material really doesn't "like" AFM, it could 
have given you twice the FM solution. With identical energies, obviously.

Stefaan