My Community

May I confirm what are the proper procedures for calculation of band structure / density of states calculation when meta-GGAs are selected?

I think I read somewhere in this forum before, that the CHG/CHGCAR file does not contain the kinetic energy density information and hence these calculations' procedures are different from using say, the PBE functional.

From what I understand, the meta-GGA calculations should not read the CHGCAR, and so the charge density has to be built from scratch via a self-consistent calculation. Hence, are these steps correct for calculations with metaGGAs, e.g., SCAN?

• use a CONTCAR from previous relaxed configuration

• without reading CHGCAR/WAVECAR file, run a static calculation (NSW=0)

• k-points for band structure should be a combination of weighted k-points from IBZKPT of the regular grid plus the unweighted k-points for band path. DOS calculations use regular mesh as usual.

• the electron density will be built up during the electronic scf calculation. this can take a few steps (e.g. up to iteration 1(34) in OUTCAR, but it won't go iteration 2(1) because NSW=0)

• use the result to plot band structure/dos

Is this correct? I think metaGGA case is similar to hybrid functional band structure calculation procedure.

By the way, is it possible to read the WAVECAR from the relaxed configuration to speed up the electronic-scf calculation? (e.g. for it to converge in iteration 1 (3) instead of iteration 1 (34))
Can the WAVECAR be read without loss of kinetic energy density or other information, unlike the CHGCAR?

I think we have a wiki page for hybrid band structure calculation but none for metaGGA.
If possible, can we add a wiki page for metaGGA or at least remark that metaGGA calculations should follow the hybrid functional procedure?

Hi,

It seems that CHG/CHGCAR is not needed to restart a meta-GGA calculation if WAVECAR is present. If WAVECAR is present, then only two iterations 1(2) are done (the kinetic-energy density is built from WAVECAR). So, the steps that you described are ok, except that WAVECAR should be read (done by default) to avoid unnecessary iterations. Note that in VASP.6.3.0 there is a new feature, KPOINTS_OPT, that makes the calculation of the band structure more convenient.