detemination of ISMEAR and sigma

[luke419]

I'd like to know how ISMEAR and sigma should be deterined generally.

I am to use ISMEAR=-5 and sigma=0.2 for my sytem(metal-organic material and it may be insulator or have somewhat a semiconducting property).
Should it be O.K.?

What does decreasing of sigma=0.01 or 0.02 mean in detail and in what case should we use small sigma?

best regards,

[admin]

ISMEAR=-5 is appropriate if you have a system with a band gap, because the occupancies of the states are not a variational quantity in this method. For systems with bands crossing the Fermi level, you should only use it for one final post-processing run to get an accurate DOS. Please also note that the Gamma point should be included in your k-mesh to cover the BZ properly. the smearing does not affect the DOS if you use one of the tetrahedron methods for BZ integration.
If you use either Gaussian or Methfessel-Paxton integration, please check the largest possible SIGMA by checking the additional TS contribution to U in the free energy: the difference between U and F should not be larger than a few (1-2) meV per atom (please compare the total energy and the total energy extrapolated to sigma=0)