More on hexagonal cells

[abotello]

Dear YAMBO Developers,

I have recently installed YAMBO 3.2.1 r.392 in a linux PC.
details of installation are:
./configure --without-mpi --with-iotk=/opt/pwscf/iotk --with-p2y=4.0 --with-netcdf_include=/usr/local/include/ --with-netcdf_lib=/usr/local/lib --enable-largedb

I went through most of the tutorials with no problems. Now I'm trying to work with my system of interest: ZnO. I tried a simple bulk wurtzite calculation. I have read in previous posts that shifted hexagonal cells generated with the automatic option in pwscf leads to an error because it generates a non uniform k-grid. However, I tried initializating the wurtzite ZnO system with YAMBO using a 10x10x10 grid with no shift and leads the following:

<---> [01] Job Setup
<---> [02] Input variables setup
<---> [02.01] K-grid lattice
<---> [02.02] RL shells
<15s> Shells finder |####################| [100%] 15s(E) 15s(X)
<15s> [02.03] Input (E)nergies[ev] & Occupations
<15s> [03] Transferred momenta grid
<15s> [RL indx] X grid is not uniform. Gamma point only.
<15s> [04] Game Over & Game summary

Am I missing something? I attached the pwscf imputs.

Thank you and best wishes,
Andres.

[andrea marini]

Dear Andres,

I have correctly reproduced your error. I will work on it in the next days.

Thanks for pointing out this problem !

[andrea marini]

Am I missing something? I attached the pwscf inputs.

Thank you and best wishes,
Andres.

Dear Andres,

using the calculation mode 'bands' I indeed get your error with the 10x10x10 grid. Nevertheless if I perform a 'scf' or an 'ncsf' calculation with the same grid Yambo does not complain.

By comparing the output of pw.x in the two calculation modes (attached to this post) I noticed that in the case of calculation='bands' pw.x generates 84 k-points, that are 116 when calculation='scf' or 'nscf'.

To my understanding in the 'bands' mode pw.x does not expand the k-points in the whole irreducible wedge of the brillouin zone. I am not an expert of PW, so I do not know why it does so.

The easy solution is to run pw.x with calculation='nscf' instead of 'bands'.

Andrea