ABINIT Discussion Forums

Deal All

I'm learning the method to calculate the full gw band structure using the self-consistent gw + MLWFs scheme. I've used the "tests/wannier90/Input/t03.in" as a start to perform the study with Abinit 6.2.2 (MPI version, prepared for a x86_64_linux_gnu4.4.) May I ask a couple of questions.

Question-1:
I tried to separate the run into two parts: part-1 to do gw in parallel and part-2 to run WF in sequential. The Abinit input files for the gw parallel run is t03_part1.in, and those for the WF sequential run is t03_part2.in and t03o_w90.win, (see attachments). In the t03_part1.in I commented out those lines involving the calculation of the quasiparticle Wannier functions (dataset7). In the t03_part2.in I commented out those lines for the ground state and gw calculations but kept the WF part. In the WF sequential run, I used the output files t03o_DS1_DEN, t03o_DS2_WFK and t03o_DS6_QPS obtained from the gw parallel run as inputs. The files were renamed to t03i_DEN, t03i_WFK, t03i_QPS, respectively. Both runs were successfully completed. However, I still would like to ask if my input files and the steps taken are the right approach for the job?

Question-2:
In order to get a reliable band structure, I understand that convergence studies have to be properly carried out but I am not quite sure how?. Should we do the similar convergence studies as in the case of standard GW? For examples,
1. The convergence of GW corrections with respected to ecut and ngkpt used for the ground state and KSS calculations?
2. The convergence of GW corrections with respected to nband, npweps, npwwfn, used in the gw calculations?

Or maybe more, such as

3. The convergence of GW corrections with respected to the number of self-consistent gw iterations?.

All comments and suggestions are greatly appreciated.

Kind Regards,
Thanusit Burinprakhon

Dear Thanusit,

In a low-dimension system, the vacuum size is crucial in GW calculation [PRB 73, 205119(2006)]; if involving transition metal and rare-earth elements, the pseudopotential is very important too.

Sincerely,
Guangfu Luo