My Community

Dear Admin,
I am using VASP-5.4.4. I am having problem generating appropriate KPOINTS file for tetragonal body-centered lattice. Using the KPOINTS file ( 'Generalized regular grid' method) as is explained in vasp wiki (wiki/index.php/KPOINTS#generalized_regular_grids), VASP is generating the IBZKPT file which contains a completely different list of k-points in comparison to when I generate the KPOINTS file using 'autoGR'-software (https://github.com/msg-byu/autoGR). I am not sure which one should be the right one (or both)? May ask for some hints, suggestions, or insights?
Note:
1. I used primitive cells in both cases.
2. I first ran the VASP and then used that total no. of reducible k-points from IBZKPT (in this example 32) to generate the KPONTS file in autoGR. This ensures the total number of reducible k-points is the same in both processes.
3. I checked different symmetry options for autoGR as well. No change.

The Gamma point weight in VASP IBZKPT file is 4, which I don't understand. In autoGR KPOINTS file some of the k-point coordinates are greater than 0.5 in the reciprocal lattice, which again I don't understand because there are no k-point coordinates greater than 0.5 in the body-centered tetragonal lattice Brillouin zone high symmetry path.
I have also tried to use the autoGR.gen file as the KPOINTS file in VASP calculation. But it generates another different set of k-points.

Please see the attachment for zip file with all the files. Sincerely,
Badal Mondal

Some comments:

1) VASP uses Monkhorst-Pack grids, i.e. "regular grids" (each recip lattice vector is divided into an integer number of parts). But autoGR uses *generalized* regular grids. The grid generating vectors create a kpoint lattice that is *commensurate* with the reciprocal cell, but the generating vectors are not necessarily in the same directions as the recip lattice vectors. So there is no expectation that VASP's grids and autoGR grids will be the same. Typically, they won't be. (See attachment)

2) Even though VASP weights the gamma point more than 1 (in some cases, depending on the crystal structure) the relative weights are correct. I'm not sure what the convention is, but when I have checked the grids by hand, the relative weighting was correct, even if the gamma point was not weighted as 1.

3) It's possible for kpoints to have lattice coordinates with magnitude greater that 0.5 depending on the choice of reciprocal lattice vectors (which depend on the choice of real space lattice vectors) and still be inside the first BZ. Points will necessarily have lattice coordinates with magnitude less than 0.5 only when the recip lattice vectors are the so-called Minkowski-reduced vectors---the shortest basis vectors possible. autoGR uses a Minkowski-reduced basis internally but expresses the final KPOINT set in terms of the recip lattice vectors that the user chose (via their choice of real-space vectors).

4) Different choices of real-space vectors lead to different reciprocal lattice vectors and thus to different representations of the grids generated by autoGR. It should make no difference in your calculations since internally the best grid is chosen irrespective of the basis choice.

5) Vasp may fold the kpoints differently (but equivalently) to autoGR so the list generated by vasp from autoGR.gen may *look* different from the autoGR list, but they should be equivalent.

I hope this helps.