Page 1 of 1

em1d and NGsBlkXp in BSE input file

Posted: Thu Mar 05, 2020 2:38 am
by jiejiang
Dear Yambo Community,

I did a BSE calculation after a G0W0 calculation.
I set KfnQPdb= "E < SAVE/ndb.QP" to read quasiparticle energies from G0W0.

In BSE input, I see em1d (Dynamical Inverse Dielectric Matrix) and NGsBlkXp (Response block size).
The LOG files in the BSE output directory show calculation of Xo@q,
and the Timing shows Xo (procedure) : 3h 4m CPU and Xo (REDUX) : 34m27.00s CPU.

My first question is,

do the above information indicate that BSE repeats the G0W0 calculation of dynamical dielectric matrix?
If yes? why BSE does NOT read the data from G0W0? Is it because BSE could NOT find the data?
Also, BSE needs static, NOT dynamical dielectric.

My second question is,

In G0W0, I set NGsBlkXp= 12 Ry

So, in BSE input, can NGsBlkXp be a number <=12 Ry?

I attach my BSE input and output files.

Thank you very much,

Jie Jiang
AFRL
Dayton OH

Re: em1d and NGsBlkXp in BSE input file

Posted: Thu Mar 05, 2020 9:19 am
by Daniele Varsano
Dear Jie Jiang,
do the above information indicate that BSE repeats the G0W0 calculation of dynamical dielectric matrix?
Yes, you are repeating the calculation.
If yes? why BSE does NOT read the data from G0W0? Is it because BSE could NOT find the data?
Yes, this is because the ndb.pp file it is neither in the SAVE directory nor in the output directory (in your case BSE_21k), probably it was stored in another directory.
If you move or link the file in the SAVE or BSE_21k directory the database will be read.
Also, BSE needs static, NOT dynamical dielectric.
Sure, but in your input, you indicated to calculate the dynamical one (em1d/ppa). Note that ndb.pp contain the dielectric matrix also for the static limit and this part is the one which is read for BSE calculation.
In G0W0, I set NGsBlkXp= 12 Ry
So, in BSE input, can NGsBlkXp be a number <=12 Ry?
Yes, as if you put a larger value this will be incompatible with the screened database and it will be recalculated by scratch.

Best,
Daniele

PS:
note that it is not needed you add your signature by hand as you can fill it in your user profile once for all.

Re: em1d and NGsBlkXp in BSE input file

Posted: Thu Mar 05, 2020 6:16 pm
by jiejiang
Dear Daniele,

Thank you very much for your answers.

Thing is much clearer now.

One more point needed you further help.

After I put ndb.pp file in a right place that BSE can read, can I delete all parameters related to dynamic matrix
(e.g. em1d, NGsBlkXp etc created by -p p in input generation) in BSE input file.

Thanks,

Jie

Re: em1d and NGsBlkXp in BSE input file

Posted: Thu Mar 05, 2020 7:29 pm
by Daniele Varsano
Dear Jie,
I think so, in this case, Yambo should not complain, you need also to delete the corresponding runlevel (em1d, ppa) you can make a try, in general, it is anyway suggested to keep them as yambo when reading a database also read the parameters used to calculate the database and compare them with the parameters in the input file. If they are not compatible, depending on the seriousness of the incompatibility yambo can emit a warning in the report file and continue or emit an error and start to calculate the database again with the parameter in the input (Note inconsistency in NGsBlkXp is considered serious). If databases are present when generating the input file, you should find the right parameter already in the input.

Best,
Daniele

PS: Please, fill you signature in the user profile

Re: em1d and NGsBlkXp in BSE input file

Posted: Thu Mar 05, 2020 7:52 pm
by jiejiang
Dear Daniele,

Than you very much for your answers.

Things are very clear.

Thanks again,

Jie