How to choose smearing method

[yuxiao_zihan]

Dear VASP users,

I have a question of the smearing method. As I know, In the INCAR file, we should use ISMEAR=-5 for semiconductor and 1 or 2 for metal. Sometimes if the system is very big , we could also use ISMEAR=0.

The question is,
i notice also in the vasp guide, to calculate the groundstate energy of atoms with spin , it suggests to use 'Gaussian smearing, otherwise negative occupancies'. So it means even for metal, to obtain the the accurate magnetic groundstate, we should use Gaussian smearing or not?

So I guess the right answer is to use Methfessel-Paxton to relax the metal, and then switch to Gaussion method to get the groundstate of a magnetic system?

Thanks for your answer.
<span class='smallblacktext'>[ Edited ]</span>

[yuxiao_zihan]

I notice for single atom calculation, the smearing method does not change the magnetic moment. But for large system, a molecule on metal surface, the smearing method does influence the results a lot. Up to now, it is still not clear to me.