size of the cell and kpoints

[jasius]

I had a question - while I understand generic kpoint number rules with the cell size, at which point do I "give up" and just work using single gamma point and gamma version of vasp which is cheaper computationally and faster than standard one? For example, I have a supercell of around 460 atoms to study doping effects. I am optimizing structure, even running frequencies and then doing PBE0 band analysis. I am running 2x2x1 mesh and standard VASP. Would I not lose any information if I just switched to a single gamma point and gamma distribution of VASP? thank you

Regular 4 x 4 x 2 mesh centered at Gamma
0
Gamma
2 2 1

P Mg O N H Fe
1.000000000000000
13.8669440065438128 -0.0029515907443530 0.0150787091048969
-0.0025883566951547 12.0702550383343734 -0.0199312800100080
0.0243645232339812 -0.0369824130200632 22.4000049526981790
P Mg O N H Fe
16 15 160 15 252 1
Direct
....

[pedro_melo]

Dear jasius,

From my experience with supercell calculations you would usually look for a parameter that indicates convergence over the cell size. I do not know the particulars of your system, but is there such a thing in your case?

For instance, in studying defects, if they introduce new, localised states, one increases the cell size until these new states show no dispersion in the bandstructure plot. You would start from a reference calculation in the unit cell that you know is converged with respect to the number of k-points before going to supercells.

However, please note that other properties might need more k-points than the ground state density. I would advise you to first do a study on a unit cell to determine which values for the number of k-points, bands, etc ensure convergence for whatever quantity you want to study. Only then would I perform supercell calculations.

Leet me know if this helps. Kind regards,
Pedro