I have a general question about a software dependency of band gaps.
I calculated electron states of simple materials(e.g. MgS) to rare-earth containing compounds.
elk produces the bandgap of 2.519eV in elk for MgS(Magnesium Sulfide) in the following calculation details
10x10x10 kpoints, PBE-GGA, rgkmax=8.0, (the structure is the optimized one in elk)
while wien2k shows 2.768eV in the same calculation settings.
The difference between elk and wien2k is 0.25eV.
Moreover, rare-earth containing compounds make about twice bigger difference in bandgaps.
Do you know the possible reasons of the difference in bandgap.
Both of them are FLAPW software.
Thanks in advance.
one possible reason is a different choice of basis sets: FLAPW does not uniquely describe the basis set. Have you tried to use the same basis sets? You can check for errors of the augmentation by using different muffin-tin radii. Try to vary them and see what happens. If the results depend on the radii, your augmentation within the spheres is incomplete.
Have you tried to converge the calculations with respect to smearing width? Perhaps the smearing in elk is larger. Can you try to lower swidth and see if the two results agree? Please let us know what you get.
Thank you for your replies
I confirmed that the calculations of MgS are converged with respect to smearing width trying 0.01, 0.001, and so on.
Also, varying muffin-tin radii does not change the bandgap more than 0.01eV.
Generally, what parts of the basis set are different in elk and wien2k (APW+lo)?