Author Topic: Phonojn dispersion taking too long  (Read 2966 times)

0 Members and 1 Guest are viewing this topic.

Offline ams_nanolab

  • Supreme QuantumATK Wizard
  • *****
  • Posts: 389
  • Country: in
  • Reputation: 11
    • View Profile
Phonojn dispersion taking too long
« on: September 24, 2013, 07:19 »
Dear Sir,

I was looking to obtain phonon dispersion and phonon DOS of WS2 monolayer using ATK 13.8.b1 using DFT LDA (16x16x1 kpoints) as in PHYSICAL REVIEW B 84, 155413 (2011).

The calculation progresses okay initially then just does not seem to progress at all , last 24-hours it has been struck on the following step (screenshot attached), is anything wrong?

Offline Anders Blom

  • QuantumATK Staff
  • Supreme QuantumATK Wizard
  • *****
  • Posts: 5576
  • Country: dk
  • Reputation: 96
    • View Profile
    • QuantumATK at Synopsys
Re: Phonojn dispersion taking too long
« Reply #1 on: September 24, 2013, 13:27 »
When doing phonon calculations, it is necessary to create a larger supercell than the original system. ATK will automatically extend the system, and in your case you can see that it has determined the necessary size to be 9x9x3. Now, first of all, the value 3 indicates that the vacuum gap in the C direction is too small - there are matrix elements to the next layer in this direction, which you don't want. So you should increase the C vector to avoid this.

With 9x9x3 repetitions and 3 atoms, you actually have a system of 243 atoms. That's not so large, even for DFT - but then you have 16x16 k-points for this, and that is HUGE. Since the repetition will reduce the need for k-points, if you want 18x18 in the actual calculations, the 9x9 repetition means you should set up the calculation using only 2x2.

No matter which settings you use, however, note that a phonon calculation like this with DFT is very time-consuming and uses a lot of memory (8 GB at the least I would guess). But it should not take 24 hours per point, and this can hopefully be fixed with the hints above.

I'm actually guessing your calculation is not really running, but has exhausted the memory of the computer and stopped. When checking such things (in particular on Linux), always look in "top" first to determine if the process "atkpython" is running or not.

Offline ams_nanolab

  • Supreme QuantumATK Wizard
  • *****
  • Posts: 389
  • Country: in
  • Reputation: 11
    • View Profile
Re: Phonojn dispersion taking too long
« Reply #2 on: September 24, 2013, 14:00 »
Thanks Dr. Blom the calculation is now working...  ;D