non-magnetic ground state?

[duncan2]

Dear VASP

I want it to be clarified specifically what 'non-magnetic ground state' exactly means.
According to the manual, non-magnetic ground state CHGCAR file is needed to do some non-collinear calculations.

What does the exact setting in INCAR file?
Is it right that non-magnetic ground state is not only LNONCOLLINEAR=F, but also ISPIN=1?

If it is right, I have another question about scf.

Resulting 'length of vectors line' in OUTCAR is different in both cases, right? (magnetic and non-magnetic)

So, some stress or forces can be induced from difference between magnetic and non-magnetic structure.

In this situation, is the optimized structure determined by non-magnetic calculation reliable? Also the CHGCAR file?

Or, I have to do some trick not to induce stress or force in obtaining non-magnetic CHGCAR file such as:

1. structural optimization with ISPIN=2, LNONCOLLINEAR=T to get optimized magnetic structure.

2. static calculation with the structure obtained from above to get non-magnetic CHGCAR file. (NSW=0, ISPIN=1, LNONCOLLINEAR=F).

Well, in summary, how can I set INCAR file explicitly to get a non-magnetic ground state CHGCAR file?

and, is the difference in lattice parameters between magnetic and non-magnetic state negligible? If not, is the trick suggested by me reasonable?

best regards,

Kisung

[admin]

1) non-magnetic means that ISPIN=1, you are correct
2) usually, the equilibrium structures of magnetic and non-magnetic states are slightly different. However, the changes due to magnetization usually are rather small, so one can safely assume that the non-magnetic structure is a good starting guess for the optimization of the magnetic strucutre.
3) ISPIN=2 is not appropriate for non-collinear magnetic structures, just for FM and AFM spin orientations.
4) you will have to do further relaxation steps if you turn on/off magnetization, NSW must not be set to 0. How many steps are needed, depends on the diffrence of the ground state structures.

[duncan2]

in your apply 3), you mean ISPIN=1 and LNONCOLLINEAR=T yields the same result as ISPIN=2 and LNONCOLLINEAR=T??

OR

switching on both ISPIN=2 and LNONCOLLINEAR=% has to be avoided?

which one??