GW+ELPH+HF in YAMBO 5.1

[LHuang]

Dear developers,
I found when I performed GW+ELPH calculations with HF correction, the calculations will be closed by some error. Could you help me, please?

[LHuang]

The calculation will be stuck in HF_and_locXC step. However, if I used yambo 4.5, it wouldn't happen.

[Daniele Varsano]

Dear Liushneg,

it can be a memory issue. In order to improve the memory distribution, try to put all the CPUs on the Gbands by setting in input:

Code: Select all

SE_CPU= " 1 1 32"       # [PARALLEL] CPUs for each role
SE_ROLEs= "q qp b" 

Anyway, please nots you are using RIM_W, I'm not sure if it has been tested in elph calculations. Note also that RIM_W was not present in yambo 4.5
I would also try to remove the RIM_W keyword.

Best,

Daniele

[LHuang]

Dear Daniele,
Thanks for your reply. Though I have tried to put all the CPUs on the Gbands, it still doesn't work. The calculation will be stuck in the same step. Could you help me, please?

Best,

Liusheng

[LHuang]

Dear Daniele,
Can I just remove HF correction? Is it possible to get a reasonable BSE spectrum by using a scissor operator?

Best,

Liusheng

[Daniele Varsano]

Dear Liusheng,
after inspecting your input, I doubt it is a memory issue.
The RIM-W method as far as I know has never been tested/used together with el_ph_corr, can you try to do a standard GW calculation?

Also, please note that your value VXCRLvcs is very low and this will introduce an error when subtracting the <Vxc> component.

Is it possible to get a reasonable BSE spectrum by using a scissor operator?

Yes, if the scissor approximation applies, in order to validate it usually some GW corrections are needed.

Best,
Daniele

[LHuang]

Dear Daniele,
Thanks for your reply. I have tried to increase the value of VXCRLvcs. But the result is same. What should I do? Could you help me, please?
Best,

Liusheng

[Daniele Varsano]

Dear Liusheng,

my comment on Vxc component was a general remark not related with your problem.

Let's do some tests step by step to spot the problem.
1) Reduce your QP calculations as you had in the first report you sent:
%QPkrange # [GW] QP generalized Kpoint/Band indices
1|1|26|27|
%
This will exclude memory issues.
2) Try to do a calculation without the el-ph contribution: i.e. remove el_ph_corr from your input.
As stated before, I'm sure that W_RIM and el-ph clash in some part of the code
3) If the error persists, try to recompile the code using the --enable-int-linalg option.
> make disctlean
> configure *your options* --enable-int-linalg
> make yambo

and see if this solves the problem.

Best,
Daniele

[LHuang]

Dear Daniele,
Thanks for your reply. I will try it later. But I have performed the calculation of GW+ELPH+HF in YAMBO 4.5. However, the qp gap correction is -0.08663, which is wrong. Could you help me, please?
Best,
Liusheng

[claudio]

Dear Liusheng,

I never calculated GW and ELHP correction in the same run, I advice you to make a test and try to do
two separate runs one for GW and one for ELPH

yambo -F yambo.in_gw -J GWQP
yambo -F yambo.in_elph -J ELPHQP

in such a way you can check the different contribution to the gap correction
o-GWQP.QP and o-ELPH.QP

you can make a simple script to sum them up and
ff you want you can join them in a single database using

ypp -qpdb g

let me know what you get and if this is ok

best
Claudio

ps I will check if yambo can reads two quasi-particle databases in the same run, otherwise create a new one with ypp