Problem in calculating optical absorption.

[Bikash]

Dear developers,

I want to calculate the optical absorption for my system. I started following the instruction given here http://www.yambo-code.org/wiki/index.ph ... tandalone).
In my case, the job quits without writing the dielectric function. I attached all the necessary files.
I would greatly appreciate any guidance. Thank you in advance.

with best regards,
Bikash Patra,
NISER, Bhubaneswar, India.

[Daniele Varsano]

Dear Bikash,

some notes:
1) the job is not finished, did it crashed?
2) the job is needing very large amount of memory:

Code: Select all

[MEMORY] Alloc kbv( 19.89103 [Gb]) TOTAL:  20.11708 [Gb] (traced)

do you have such amount of memory for CPU? Here an option is to assign a smart parallelisation in order to distribute the memory. At the moment you have the default assigned parallelisation which is not optimal. An option would be to assign in input something like:

Code: Select all

X_and_IO_CPU= "1 1 1 48 4"                 # [PARALLEL] CPUs for each role
X_and_IO_ROLEs= "q g k c v" 

3) As you can see from the report file:

Code: Select all

[WARNING] [x,Vnl] slows the Dipoles computation. To neglect it rename the ns.kb_pp file

you can also neglect the dommutatori term in the dipole calculation to speed up the calculation, and next verify if it is accurate enough. In order to do that change name to the above file e.g. as:

Code: Select all

mv ./SAVE/ns.kb_pp_pwscf   ./SAVE/ns.kb_pp_pwscf_

4) Most important point you have a non-uniform k point grid so your calculation will be not meaningful as gamma point only is considered here. Reason for that is:
a) in QE calculation you did not use automatic k point generation
b) you did not include the force_symmorphic=.true. keyword
c) Yambo was not able to recognise your grid properly.

If option c, you can try to change the algorithm for the k point grid recognition:
c1: delete all your ./SAVE/ndb* database
c2: rerun the setup adding the following keyword

Code: Select all

NoDiagSC

in the setup input file.
c3: check in the r_setup you have:

Code: Select all

X grid is uniform                                : yes

[Bikash]

Dear Sir,

I am first trying to solve point 4 as you mentioned.
In my QE calculation, I used automatic k point generation and force_symmorphic=.true. (files are attached).
I also tried point c in the following way-
1. run /home/bikashp/code/yambo/yambo-5.0.2-memory/bin/yambo -i -V all
This will produce a yambo.in file and I uncomment the line involving NoDiagSC.
2. again run /home/bikashp/code/yambo/yambo-5.0.2-memory/bin/yambo
but still I am getting X grid is uniform : No
I have attached the setup file also.
Any suggestion would be helpful.

with best regards,
Bikash Patra,
NISER, Bhubaneswer, India

[Daniele Varsano]

Dear Bikash,
in this case you should try to define the cell in QE in a different way.
Unfortunately in this moment I cannot visualise your input file, anyway I can see that you are sampling the BZ in only one direction, is this a 1D system? In this case can you arrange it in an orthorhombic cell, or in such a way that the z direction is orthogonal to the other two? It seems to me that the actual cell / sampling is rather complicated.

Best,
Daniele