EPW Forum

Hi Dear All
I am doing el-phonon calculations using EPW-4.1 I am running a 70 atom el-ph calculation with following details
coarse grid:
2x2x2 qpoints (8 qpoints)
4x4x4 kpoints (62 points) in nscf calculation
fine grid:
2x2x2 qpoints
4x4x4 k-points
nbndsub=136
nbndskip=0 and another with 50
etf_mem = .false.

I am running on 40 cpu in 5 nodes and memory of each node is 64G. It gives following error

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Error in routine davcio (116):
error while writing from file "./terph.epmatwe1"
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

When I do the same calculation with nbndsub=50 it gives error related called "Ef cannot bracket" and by giving Fermi energy (from scf calculation) in input It goes well but it output 0 DOS at fermi level and zero values for el-ph coefficients. From that what I understood is that I am giving very few wannier function in input so it can transform only very less number of bands and they are well below fermi level and hence can't bracket fermi level.

So I increased nbndsub to 136 (same as number of bands in scf & nscf calculations) but then thi error is occuring. I have read other posts and what I understood is that this error comes because of the memory requirement becomes much higher than available memory. How can I fix this issue?

I am too naive in el-ph calculation and wannier basis so please comment on my understanding and correct it as well. Thank you

Best Regards
Nadeem

Dear Nadeem,

It is generally good to try and reduce nbndsub if you don't need all the bands from the nscf calculation; however, have you checked that the band structure around the Fermi level is correctly reproduced when you use nbndsub=50?

Best,
Carla

Thank you Carla

Now it is working fine when I give fermi energy in input and my band structure is reproduced around fermi level. I have one question that how can I know the number of kpoints along k-points path that i gave for band structure plot?
Thank you again for lot of help from you guys

Dear Nadeem,

The interpolation of the band structure is very cheap so you can easily put 30-50 k points between each high-symmetry point.

Carla