How do we choose SIGMA

[jianmin]

Hi guys,

For adsorptions of small molecules on metal slabs, I have always been using ISMEAR=1 and SIGMA=0.2 (SIGMA=0.2 is recommended for metals in VASP the Guide). The other day, I tried SIGMA=0.1 and SIGMA=0.05 just out of curiosity, and got different energies from those of SIGMA=0.2. As we know, these energies are energies extropolated from SIGMA=0.2 or 0.1 or0.05 to SIGMA=0.

My question comes: Which energy should we trust the most, the energy from a smaller SIGMA?

[graeme]

This is not a vtst question, but anyways, for converged energies, you should look for the sigma->0 extrapolated energy. As you make sigma lower, this value should converge. (so yes, smaller sigma is better for accuracy)