5
$\begingroup$

I am performing phonon calculations ph.x of CsPbBr3 primitive cell in QE v7.2 and it is taking a long time on 6 6 6q-points and 8 8 8 k-points.

Can we make just the gamma point phonon calculations instead of the entire grid?

Improve this question
$\endgroup$
2
  • 1
    $\begingroup$ Wannier90 has nothing to do with phonons. Phonon calculations are just very computationally intensive; but if you are doing phonons and Wannier90, the computation that you do after that will be much more expensive. You had better make sure your QE computation and phonon computation and Wannierisation have converged to the correct values before trying that next step! $\endgroup$
    – naturallyInconsistent
    Commented May 10 at 10:00
  • $\begingroup$ Yikes. Does that mean I have to do convergence tests on the q-point grid in ph.x as well? $\endgroup$
    – Misbah
    Commented May 10 at 10:39

1 Answer 1

Reset to default
2
$\begingroup$

Replying to your comment above: of course you have to converge wrt. q-points!

To answer your original question, of course you can calculate at Gamma only with DFPT as long as you only care about energies, eigenvectors, etc. at Gamma. This is perfectly reasonable for studying e.g. Raman spectra.

If you want to do e.g. electron-phonon stuff at q != 0, you must converge wrt to q-points. Note that converge of electron-phonon stuff is slow which is why e.g. EPW interpolates over q-points (and k-points). You can get by with a coarse q-point grid and if the interpolation is good, but you still have to check convergence wrt to the coarse q-point grid. For a detailed discussion, see the Abinit documentation here. This won't help you with QE, but should help explain about convergence!

If the ph.x calculation takes too long, you could parallelize over q-points and run each q-point calculation separately. I have done this and it works great. See the example here: /PHonon/examples/GRID_example/

I hope this helps! :)

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thank you @Tyler Sterling. It helped alot and i hope it will save my time. But i am still confused in what value should be assigned to start_q, last_q, start_irr and last_irr. $\endgroup$
    – Misbah
    Commented May 12 at 14:34
  • 1
    $\begingroup$ Did you look at the other ph.x documentation? Go read it for those variables if not. I don’t parallelize over irreducible reps so I only use start_q and last_q. These tell each job you run which irreducible q-points to do DFPT for. Suppose you want to do a 2x2x2 mesh or smth and there are 4 irreducible q-points. You can do the first one only with start_q=1 and last_q=1 etc or do qpts 2 and 3 with start_q=2 and last_q=3 etc. This way, you can split up many q -points into smaller independent jobs. $\endgroup$
    – Tyler Sterling
    Commented May 13 at 3:07

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .