Can I restart a SCF calculation job that ran out of time?

Queries about input and output files, running specific calculations, etc.


Moderators: Global Moderator, Moderator

Post Reply
Message
Author
reynaldo.putra
Newbie
Newbie
Posts: 16
Joined: Wed Jan 24, 2024 7:33 pm

Can I restart a SCF calculation job that ran out of time?

#1 Post by reynaldo.putra » Wed Jun 19, 2024 2:13 pm

As posed by the question.

Ionic relaxation can be done by taking the last structure from the CONTCAR file. How can I do the same thing for SCF calculations?

pedro_melo
Global Moderator
Global Moderator
Posts: 127
Joined: Thu Nov 03, 2022 1:03 pm

Re: Can I restart a SCF calculation job that ran out of time?

#2 Post by pedro_melo » Mon Jun 24, 2024 1:37 pm

Dear reynaldo.putra,

Currently VASP allows you to read a WAVECAR file from a previous calculation. Please see the page wiki/index.php/ISTART. You could tell VASP to stop the SCF cycle after a certain number of iterations, write the wave functions and the density to the WAVECAR, and restart from that. Note that you will need to keep a consistent parallelisation scheme, namely to ensure that the number of bands remains the same.

Kind regards,
Pedro

reynaldo.putra
Newbie
Newbie
Posts: 16
Joined: Wed Jan 24, 2024 7:33 pm

Re: Can I restart a SCF calculation job that ran out of time?

#3 Post by reynaldo.putra » Tue Jun 25, 2024 12:45 pm

Dear Pedro,

Thank you for your response. I am not too familiar with the computation schemes. Could you explain to me what a consistent parallelisation scheme is, and how I should design my input files so that I would have a consistent parallelisation scheme?

Best regards,
Reynaldo

pedro_melo
Global Moderator
Global Moderator
Posts: 127
Joined: Thu Nov 03, 2022 1:03 pm

Re: Can I restart a SCF calculation job that ran out of time?

#4 Post by pedro_melo » Fri Jun 28, 2024 1:35 pm

Dear Reynaldo,

I meant that the size of the arrays has to be consistent, specially when you set up NBANDS. As you are probably aware, VASP can adjust the value of NBANDS depending on the number of processes that run the job. So when you resubmit the new job, please make sure that this value is the same that VASP used in the previous run. The same for the cutoff variables like ENCUT and ENCUTGW.

Kind regards,
Pedro

Post Reply