My Community

Dear users and developers,

I am running one-shot GW calculation (ALGO=EVGW0) on a HPC. But due to time limitation, it might get killed (after some NQ points).

Is it possible to restart the calculation from the same NQ where it stopped or does it always start from the beginning ?

Regards

Dear asif_ali28,

At the moment, a GW calculation cannot be restarted and all NQ points have to be recalculated from the beginning.
However, there are some ways one could potentially speed up the calculation.
If you haven't set KPAR, you can try that, as it should noticeably speed up the calculation.
Be aware that setting KPAR distributes the work over NQ points, so it increases the memory requirements for the calculation.
For big systems, it is usually beneficial to use the low-scaling algorithm GWR (see wiki).

Dear alexey.tal,

I am already using KPAR tag for it. I will try the low-scaling GWR.
Thank you for the help.

I always wanted to know if GWR can be used for BSE calculation?

Thankyou

A BSE calculation can be performed starting from the low-scaling GW (GWR). For that you need to set NOMEGA_DUMP = 1 to write the screened potential WFULLXXXX.tmp see the wiki. Also, please update the code to version 6.4.1 as the older versions had a problem with the conjugation of the potential in the GWR, which could lead to incorrect results in the BSE calculation. As of VASP 6.4.1 this issue is resolved.

I found the error which says that GWR is not yet implemented with spin-orbit coupling.