3
$\begingroup$

I am planning to use HSEO6 to calculate the band structure and DOS of a perovskite super cell. My cell has about 80 atoms. Furthermore I have also used dispersion correction and I will also have to use SOC for heavy metals.

I have tried playing with ALGO = All/Damped and TIME, but still the calculations are very slow, due to some reason the calculations tend to slow down after 6-7 iterations. So much so that one iteration takes atleast 5-6 hours.

Are there any other things that I can try out which could help me speed up the calculations.

P.S I have tried running calculations without dispersion correction and they tend to move a lot faster, but even with dispersion correction the DOS calculation took days to end, even after that it did not converge. What should I do ? Should I change the Exchange correlational functional ??

Any advice is greatly appreciated.

Improve this question
$\endgroup$
5
  • $\begingroup$ How many k-points are you using? Should the gamma point represent a valid approximation in your system, the gamma-only binary is noticably faster. $\endgroup$
    – A-V Labs
    Jun 28, 2022 at 21:36
  • $\begingroup$ Switching off spin polarization if it is not needed can also help $\endgroup$
    – A-V Labs
    Jun 28, 2022 at 21:37
  • 4
    $\begingroup$ Generally speaking, a substantial increase in computational time as you describe has to be expected when using exact exchange $\endgroup$
    – A-V Labs
    Jun 28, 2022 at 21:38
  • $\begingroup$ Sorry for the late reply @A-VLabs I have been a bit sick. Thanks for taking the time out. I can not turn of the spin polarization as my system has unpaired electrons. And other papers which have used similar systems have actually been using same K points and similar input parameters. So I am gonna start the calculations again hoping they will converge $\endgroup$
    – Parmeet Singh EP 066
    Jul 15, 2022 at 18:53
  • $\begingroup$ How did it go when you started the calculations again? Got convergence? Hybrid functionals will usually take much more CPU time than simpler functionals. $\endgroup$
    – Nike Dattani
    Feb 6 at 18:47

0

Reset to default

Browse other questions tagged .