Difference in magnetic moment between structure optimization and one-point calculation

[tomoyuki_miyata]

Dear VASP community,

For Fe–Mn–C alloys, I calculate the total energy by one-point calculation after structural optimization.
However, in many cases, the magnetic moment of Mn output during structural optimization and the magnetic moment output during one-point calculation may differ significantly. Which can be trusted, the structural optimization or the magnetic moment given by the one-point calculation?

Structural optimization is performed by methfessel paxton, and one-point calculation is performed by tetrahedron with bloch correction.
Both have the same k-mesh.

Thank you.

[henrique_miranda]

The final magnetic moment might depend on the initial conditions that you choose.
These are set by the MAGMOM tag.

However, when you run a relaxation calculation, MAGMOM will affect only the first ionic step.
For the subsequent ionic steps the WFs from the previous ionic step are used.
This might effectively lead to the last electronic step of the ionic relaxation to start from different initial conditions as the one-point calculation.

Now, if you want to know which magnetic moment is the correct one, then that would be the one that yields the lowest total energy.

I suggest the following check:
Take the relaxed structure and perform a set of one-point calculations with different values for MAGMOM.
Check if there is some dependence of the final magnetic moments and total energy on the MAGMOM tag that you used to start the calculation.

If this still does not clarify the issue, please post the OUTCARS of the ionic relaxation and the one-point calculations.