Noncollinear relaxation

[faerie]

Good afternoon,

I'm studying finite systems (clusters) and I'm planning to study the possibility of non-collinear magnetism in them. I expect the non-collinear magnetic moments to strongly affect the geometry of the system. The VASP manual suggest doing a nonmagnetic calculation first and after that enabling the non-collinear magnetism in a separate run. I'm wondering how this applies to structural relaxations.

Do I first relax the structure with ISPIN = 1 and then enable the LNONCOLLINEAR option and relax again? Can I enable collinear magnetism with ISPIN = 2 in the first calculation? Are there some other aspects to consider when applying LNONCOLLINEAR in finite systems?

I'd appreciate any help .

Mari

[admin]

Usually, the geometry changes due to magnetic effects are rather small (for bulk systems and slabs) , so, to save computing time, one pre-converges the non-magnetic structure (which is relatively 'cheap') and uses the demanding ncl-relaxation for the last steps only. However, for small clusters you might be correct with your assumpition that a preliminary non-magnetic run anticipates a structure which is only metastable in the ncl-structure. Therefore, if you expect that the final geometry of the noncollinear structure of the cluster differs significantly from the collinear one, it may be safer not to proceed from a non-magnetic structure.
You might be interested in the following 2 papers:
Stable structural and magnetic isomers of small transition-metal clusters from the Ni group: an ab initio density-functional study. Futschek-T; Hafner-J; Marsman-M, Journal-of-Physics:-Condensed-Matter. 18(42): 9703-48, (2006)
Structural and magnetic isomers of small Pd and Rh clusters: an ab initio density functional study. Futschek-T; Marsman-M; Hafner-J, Journal-of-Physics:-Condensed-Matter. 17(38): 5927-63 (2005)