bse+GW

[samaneh]

Hello developers,
I have some questions:
1- how can I use the -M with the command mpirun -np yambo to ditribute part of the memory among
the different CPU?
2- what is the difference between the yambo -o b -k sex -y d (while I use the database of yambo -g n -p p)and yambo -o b -k sex -y d -p p?
3- what is the difference between using the scissor shift (QP_E) or using the database??
(which is more costly?)
Thanks in advance
Best
S. Ataei, University of Tehran

[Daniele Varsano]

Dear Samaneh,
Here the answers:
1) Just adding the -M option when launching the executable. (mpirun -np .. yambo -M). In the next release of yambo anyway the distribution will be automatically set by the parallelization
2) The difference is not related to the use of the database of QP. In BSE you need the dielectric matrix in order to calculate the screening. This is done in static approximation (em1s). Anyway, if you did a GW calculation in Plasmon Pole before, as it also contain the static screening you can use it also for the BSE avoiding to recalculate the screening. The final outcome is the same.
3) The difference is something that you need to check. Scissor is a (rough) approximation, that anyway works for many semiconductor. Usually does not work for molecules. You need to check if it is reasonable for the system you are studying for instance by looking at the GW correction for some bands/kpoint and see if the QP correction can be simply obtained by a rigid shift of the KS conduction bands and eventually a stretching.

Hope it helps,
Daniele

[samaneh]

Hello Daniele,
Thanks a lot.
But would you please tell me if the following steps are right to do bse calculation?!
1- p2y
2- yambo -i
3- yambo
4-yambo -o b -k sex -y d -V qp
(in the input file as it is attached I use the database of running yambo -g n -p p
I would be thankful if tell me about the changes do I need in the input file which is attached)
Best
S. Ataei

[Daniele Varsano]

Sear Samaneh,

the correct syntax to include QP corrections is:

Code: Select all

KfnQPdb= "E<./SAVE/db.QP"

if you are running from the directory containing the SAVE directory.
Here you are also including the QP corrections in the calculation of the screening via XfnQPdb, this is usually not done, the RPA screening is calculated using
Kohn-Sham energies in order to be consistent. The QP energies enter in the BSE kernel only.
The rest looks me correct, even if I would set BSENGBlk to the same values of Gblk you calculate for the dielectric matrix (here NGsBlkXs= 400).
This is just because you can include all the component you are going to calculate anyway.
Next for the convergences study I invite you to have a look to the tutorials.
Best,
Daniele

[samaneh]

Dear Daniele,
Thanks a lot.
Do you mean that it is not necessary to run yambo -b and then
yambo -o b -k sex -y h?
But I read in the tutorial for LiF that first run yambo -b
and then use the SAVE/ndb.em1s database for running
yambo -o b -k sex -y h?!!
Best
S. Ataei

[Daniele Varsano]

Dear Samaneh,
this is up to you, if you want to split calculations or do all in a run.
Yambo read ./SAVE/ndb.em1s if it is found, otherwise it calculates it. I was just advising you to not include the QP correction in the calculation of the screening ( XfnQPdb) as it is not fully consistent.

Best,
Daniele