7
$\begingroup$

I would like to simulate $\ce{Bi_2Se_3}$ using a DFT calculation on VASP. However, it seems I am not able to get a result; the electronic steps take $\sim \infty$ of time to complete. Here are the details of my input.

INCAR file:

System = bi2se3 soc calculation -- SCF

ISMEAR = 0
SIGMA = 0.2


EDIFF = 0.00000001

PREC = Accurate

LSORBIT = .TRUE.



NCORE = 16

ICHARG = 2


ENCUT = 500

KPOINTS file:

Automatic Mesh generation 
0
M
15 15 15
0 0 0

I am pretty sure the structure file is correct. However, I did not relax the bulk structures. I suspect the problem could be due to the density of the Brillouin zone grid mesh / the ENCUT parameter value, but I am not sure. How can I approach this problem? Currently, the calculation seems to be frozen even before the output shows the table of the electronic steps.

$\endgroup$
3
  • 3
    $\begingroup$ Hi ! have you tried using a smaller k point grid (say 7x7x7) and a lower cutoff close to the minimum requirement (specified in your pseudopotential files) so as to know whether the calculation actually runs. After that you could increase the k point grid and ECUT values so that it gives you a good enough accuracy for a reasonable computation time. It might be because your computational resources may not be able to handle a SOC calculation (which tend to be expensive) on a 15x15x15 grid. Im not an avid user of VASP, but these suggestions are from my experience in using Quantum ESPRESSO. :) $\endgroup$ Jul 26, 2021 at 21:42
  • $\begingroup$ Hi Anoop. Thanks! I'm using 256 CPU cores to do this. I will try to reduce the ENCUT and the density of the k-grid and see what I can get. I $\endgroup$
    – PHy
    Jul 26, 2021 at 22:00
  • $\begingroup$ +1. Please look at my edit to see some small changes I made such as using ChemJax for the chemical formula and a code block for the input parameter. $\endgroup$ Jul 26, 2021 at 22:58

1 Answer 1

3
$\begingroup$

Here are some possible changes to your input that may resolve the problem:

  1. Your EDIFF is very small; use the default and then increase precision from there.
  2. Use a smaller SIGMA, start with 0.15 or 0.1.
  3. Decrease the number of k-points, start with 5x5x5, then increase to reach the desired precision.

Good luck!

$\endgroup$

You must log in to answer this question.

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