Hackmd Day 3 ## Density-functional perturbation theory: phonons * Zeyuan Tang: for the born effective charge calculation in quantum espresso, is the symmetry used to reduce the number of linear response calculation (e.g. only computing on irreducible atoms, and applying symmetry operation for other atoms)? * The symmetry is used. But it's more convenient to use it for reducing the number of k-points, so all perturbations are computed and then symmetryzed. Have a look at formulas A29 and A30 in [J. Phys.: Condens. Matter 21 395502](https://iopscience.iop.org/article/10.1088/0953-8984/21/39/395502) ### Hands-on: Phonons * Hello, I have a question. I am trying to calculate phonons in bilayer hBN using the B3LYP functional and phonopy (since PHonons is not an option), and I am having trouble with the SCF calculation for the displacements. Specifically, my calculation stops at:... ! total energy = -321.14910912 Ry estimated scf accuracy < 0.00000098 Ry convergence has been achieved in 10 iterations Using ACE for calculation of exact exchange EXX grid: 105021 G-vectors FFT dimensions: ( 60, 60, 75) ACE projected onto 32 (nbndproj) and applied to 32 (nbnd) bands At first, I thought the issue was computational cost, but when I checked the memory, it showed: "Estimated total dynamical RAM > 3.79 GB," so this does not seem to be the problem. Could you provide some guidance on how to resolve this issue? Thank you in advance! * > It seems that the issue is that your calculation is too heavy and gets stuck computing the Fock potential. You may try different things to make it less expensive. * >first reduce `nq1x` and `nq2x` values (if you have not already done it) * >reduce `ecutfock` to something like $2\times$ `ecutwfc` * >One other thing that you may try if it supported by `phonopy` is to avoid single atom displacements but use more symmetric patterns, this would reduce the number of k-point in each displaced calculation. * >Otherwise use more MPI ranks when you start the calculation and use the pool parallelism `-nk ` option of `pw.x` * Can I use `asr='simple'` for molecules and surfaces? * > yes, `simple` works in most of the cases. Not always though. It doesn't enforce the particular requirements of 2D and 0D systems, so you may end up with some immaginary frequencies. For 2D you need to be sure that the layers are far apart enough. The other asr options instead have been implemented for removing such interactions even if they are significant. * Is there a plan to support hybrid functional in the `ph.x` module? * > For the electric field responses we have some plans. For the phonons no definite plans at the moment.