smearing and the energy

[konglt]

It is said in the manual that ISMEAR = -5 can give a more accurate total energy, and that ISMEAR = 1 or 2 can also give very accurate total energy, given the SIGMA is appropriate.

I am working with some metal surfaces, during the relaxation of the surface, ISMEAR =1 and SIGMA = 0.1 are used, the resultant entropy term |F - E0| is found to be less than 0.5E-3 eV/atom.
I continued with the relaxed structure to do a single point calculation by using ISMEAR = 1 and SIGMA = 0.1, ISMEAR = -5 and SIGMA = 0.2, respectively. The obtained total energy from the two procedures are different from each other by about 6 meV/atom.
At this stage, I'd like to know, which one is more reliable? In my case I need a very accurate description of the total energy.

[Graciani]

I have the same problem and I'd like admin (or someone) reply to this matter.
Thank you for your attention.
J. Graciani

[admin]

in general, the tetrahedron method (with the Bloechl corrections
included, ISMEAR=-5) should be used to get total energies very
accurately.
Please note a few points:
1) always compare the total energies without entropy contribution
(or the sigma --> 0 values)
2) also for using the MP method, a representative sample of
k-points should be used.
3) when using the tetrahedron method, check
-- if the Gamma point is included (the tetrahedra should fill the
volume of the BZ efficiently)
-- if the k-point mesh is dense enough (please always make
k-point convergence tests if you have to calculate the total
energies very accurately)

[dkanan]

hello, i would like to know why the gamma point must be included. isn't it the case that if the gamma point is at the conduction band min or valence band maximum there are oscillations that are introduced in the DOS (i.e. at van Hove singularities)?

[admin]

the volumes of the tetrahedra have to cover the complete BZ