LRC

Deals with issues related to computation of optical spectra, in RPA (-o c) or by solving the Bethe-Salpeter equation (-o b). Includes local field effects, excitons, etc.

Moderators: Davide Sangalli, Daniele Varsano, andrea.ferretti, andrea marini, Conor Hogan, myrta gruning

User avatar
Daniele Varsano
Posts: 2655
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: LRC

Post by Daniele Varsano » Mon Feb 24, 2020 5:14 pm

Try to reduce further the parameters in order to have an idea of the needed memory.
Reduce the FFTGvecs parameter, this should help, set it lower than the cutoff on the wfs.
Actually I should look to this variable: EMStpsXd I do not remember now the meaning, I will do as soon as I have time.
In general, note that you are handling a grid of 976 k points which is rather big, are you sure that this is needed?

If none of this will help I would need to reproduce your error, but today I will be travelling back to Europe and I will be busy in the next days.
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

haseebphysics1
Posts: 148
Joined: Sat Aug 17, 2019 2:48 pm

Re: LRC

Post by haseebphysics1 » Mon Feb 24, 2020 6:10 pm

Dear Daniele,
Reduce the FFTGvecs parameter
.

This has been proved to be the reason for memory overflow! I always forget to add this in input, and this time I even not realized after getting errors constantly!

By the way, I used ecutwfc = 100 Ry (via convergence testing) in DFT and I am going with FFTGvecs = 20 Ry to remain in the manageable limit of resources! Hoping it will be a converging number of G-vectors in the sum for my system!

Many thanks,
Haseeb Ahmad
MS - Physics,
LUMS - Pakistan

User avatar
Daniele Varsano
Posts: 2655
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: LRC

Post by Daniele Varsano » Mon Feb 24, 2020 6:14 pm

Dear Haseeb,
20y over 100Ry is a quite dramatic reduction and should be checked a bit.
Anyway, I did notice this:
It is interesting that static LRC TDDFT is working with 118 GB RAM!
This sounds strange and adding the dynamical part should not require more memory: I will investigate on this issue in the next days when I find the time.

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

haseebphysics1
Posts: 148
Joined: Sat Aug 17, 2019 2:48 pm

Re: LRC

Post by haseebphysics1 » Mon Feb 24, 2020 6:37 pm

Dear Daniele,
It is interesting that static LRC TDDFT is working with 118 GB RAM!
Kindly see the screenshot.

I think a large number of k-points might be the reason for a huge amount of memory, for 36 atoms big cell, I used 18 x 6 x 18 k-mesh (# of k-points were converged using convergence tests using RPA spectra of Q.E. epsilon.x).
I will investigate on this issue in the next days when I find the time.
Thanks, if you need input files, then I will send you.

Regards,
You do not have the required permissions to view the files attached to this post.
Haseeb Ahmad
MS - Physics,
LUMS - Pakistan

haseebphysics1
Posts: 148
Joined: Sat Aug 17, 2019 2:48 pm

Re: LRC

Post by haseebphysics1 » Sun Mar 01, 2020 11:59 am

Dear Daniele,

I had computed the dipoles at the very start using yambo -o c and later did a lot of things like IPA + LFE corrections and even TDDFT (LRC) using the same computed dipole database -J directory (in the same directory). So, is it okay to use the same dipole database for later different calculations?


Thanks,
Haseeb Ahmad
MS - Physics,
LUMS - Pakistan

User avatar
Daniele Varsano
Posts: 2655
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: LRC

Post by Daniele Varsano » Sun Mar 01, 2020 12:32 pm

Sure it is okay, provided that in the database you have the transition dipoles for all the bands you need in the following calculations. In case you need more dipoles that are not present in the database, yambo will complain and recalculate the database for all the needed transitions.
Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

haseebphysics1
Posts: 148
Joined: Sat Aug 17, 2019 2:48 pm

Re: LRC

Post by haseebphysics1 » Tue Mar 17, 2020 12:09 am

Dear all, hopefully, everyone will be fine and doing well,

I want to know that in the following tutorial:

http://www.yambo-code.org/wiki/index.ph ... ed_kernel)

A negative value of alpha was chosen:

LRC_alpha= -1.200000 # [TDDFT] LRC alpha factor

So, what is the physical significance of negative alpha? Should not be it always positive? I know this is a material-dependent parameter but why the negative value was selected?

Thanks,
Haseeb Ahmad
MS - Physics,
LUMS - Pakistan

User avatar
Daniele Varsano
Posts: 2655
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: LRC

Post by Daniele Varsano » Tue Mar 17, 2020 9:43 am

Dear Haseeb,
I think this is just a convention on how the kernel is defined e.g. (alfa/q^2) or (-alfa/q^2).
In both case what it is important is that the interaction has to be attractive.
You can have a look to the phD thesis of Dr. Sottile, here the correct link.
see e.g. section 7.3.3

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

haseebphysics1
Posts: 148
Joined: Sat Aug 17, 2019 2:48 pm

Re: LRC

Post by haseebphysics1 » Tue Mar 17, 2020 5:36 pm

Dear Daniele,

Thanks for sharing the useful resource.

Now, in Yambo, how the LRC kernel is defined? Since f_xc should be negative so I guess, f_xc= (alfa/q^2) will be the definition in Yambo, if we are providing negative alpha! Am I right or missing the interpretation?




Thanking and wishing you best health,
Haseeb Ahmad
MS - Physics,
LUMS - Pakistan

User avatar
Daniele Varsano
Posts: 2655
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: LRC

Post by Daniele Varsano » Wed Mar 18, 2020 9:23 am

Dear Haseeb,
Am I right or missing the interpretation?
Right!

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

Post Reply