My Community

Dear Colleagues!

I am running bulk calculations in G0W0 approximation for different structures using the instruction at vaspwiki pages/tutorials.
I get a full success only for the unit cells with at most 4 atoms.

When I take the structure with 8 atoms or more, I get the Segmentation fault error.
The structures are fully relaxed, non-metallic, the atoms are of one type.

I have read all the topics on GW calculations here and at vaspwiki+manual. I tried:
1) to decrease k-point density up to 3 1 1 or even Gamma-point only, to take MP or G-centered meshes, to take even and odd meshes.
2) to decrease ENCUT to the smallest possible value.
3) to make calculations with different number of parallel processes (=1, =number of atoms, =2*number of atoms, ...)
4) to change NEDOS, NBANDS, CSHIFT, NOMEGA ...
I have checked that my calculation do not overfull the memory of the cores when running.

The both 2 executables -- VASP 5.3.5 and VASP 5.4.1 -- give me this Segmentation fault error.
It arises always after the output of HEAD OF MICROSCOPIC DIELECTRIC TENSOR for all frequency points for NQ = 1.

Can you propose any solution?

This is very likely the memory problem. You need to distribute the task over more CPUs.