Defining positive-definite homogenous K-point grid for EPW calculations

[Aminul Haque]

Dear experts,

I am trying to perform EPW calculations on a system that has the following unit cell parameters:

ibrav = 4 !trigonal symmetry
celldm(1) = 8.02577068
celldm(3) = 2.43976291

First, I have performed the phonon calculation with 4x4x2 q-point grid using 'ph.x'.

Then, I have generated 576 definite kpoints using the kmesh.pl script fir a kpoint grid: 12x12x4 and gave them as input for 'nascf' calculation.

However, when running the nscf calculation with pw.x, it uses 1005 kpoints instead due to symmetry issues (I suppose).

Now, it is possible to get around this issue using

nosym = .true.

But it gives error when running the epw code after using this tag in the nscf calculation.

On the other hand, the kpoint grid in nscf calculation and the q-point grid for phonon must commensurate (i.e. the k-points must be integer multiple of the corresponding q-points as far as I have understood).

I have tried several combinations of k-point grid for nscf calculations, for some cases, the pw.x runs with specified k-points but involves odd numbered k-point grid (like 13x13x5), but this prohibits to have any other q-point grid than 13x13x5 (which is quite expensive computationally).

I have two quesitons in this regard:

• What could be a systematic approach to resolve this?

• What is meant by uniform k (or q)-point grid in EPW? Does it mean something like axaxa regardless of the crystal system? As the documentation says, for BTE solutions, an uniform grid is needed.

Thanks,
Muhammad Aminul Haque

[sponce]

Dear Muhammad Aminul Haque,

Welcome on the EPW forum !

First, it is recommanded to use the same multiplicative factor between the coarse k-point and q-point grid.
Therefore if you have a 4x4x2 q-point grid, you can have a 8x8x4 grid or 12x12x6 k-point grids.

The pw.x code in calculation mode "nscf" uses symmetry to enforce crystal symmetry.

If you want to have the exact number of points, you can change the calculation mode in your pw.x from "nscf" to "bands".

Hope this helps,
Samuel

[Aminul Haque]

Dear Prof. Poncé,

Thanks for the recommendation regarding the k and q point grid.

And using 'bands' in the calculation mode of pw.x has solved the issues with the k-points.

Thank you so much.

Regards,
Muhammad Aminul Haque