About G-vectors

[paolalup]

Dear Yamboers,

I am learning how to use Yambo through the online tutorials ("Fantastic dimensions"), and I am using Pwscf for the initial calculations.
In the very first initialization step, Yambo reads how many G-vectors were used by the previous program to develop the wavefunctions and writes the actual number that it will use.
In the initialization outputs shown in the tutorials, I've noticed that the keywords G-vectors, Components and WF G-vectors in the section "[RD./SAVE//s.db1]" are always equal. This is not my case, though, when I try to do again the exercises of the tutorial: the values of Components and WF G-vectors are the same, but they are smaller than the value of the G-vectors .
I would like to know:
1) what do these keywords exactly refer to?
2) do they have to be always equal or not?

Thanks,

Paola

----------------
Paola Lupieri
PhD student
German Research School for
Simulation Sciences GmbH
Forschungszentrum Juelich
Germany

[myrta gruning]

Dear Paola

I am learning how to use Yambo through the online tutorials ("Fantastic dimensions"), and I am using Pwscf for the initial calculations.

Welcome and thank you for your interest in yambo.

In the very first initialization step, Yambo reads how many G-vectors were used by the previous program to develop the wavefunctions and writes the actual number that it will use.

In this first step, all three numbers you are referring to are read and stored in (n)s.db1. You should have these 3 lines (with different number of course) in the standard output of the p2y run

Code: Select all

<--->  :: Max WF components     :  8749
 <--->  :: RL vectors        (WF):  8749
 <--->  :: RL vectors    (CHARGE):  8749

In the initialization outputs shown in the tutorials, I've noticed that the keywords G-vectors, Components and WF G-vectors in the section "[RD./SAVE//s.db1]" are always equal. This is not my case, though, when I try to do again the exercises of the tutorial: the values of Components and WF G-vectors are the same, but they are smaller than the value of the G-vectors

At hte beginning of every run then, the files created by the interface (s.db1 and s.wf) are read. In the report under [RD./SAVE//s.db1] you have all the parameters that are read (RD) by yambo from s.db1 in the dir SAVE. So the three numbers G-vectors, Components and WF G-vectors are the same as those read above. In particular Max WF components corresponds to Components, RL vectors (WF) corresponds to WF G-vectors (RL stays for reciprocal lattice, those vectors are standardly indicated with the letter G), and RL vectors (CHARGE) corresponds to G-vectors.

I would like to know:
1) what do these keywords exactly refer to?
2) do they have to be always equal or not?

2) No they can be equal, but they do not need to. In particular G-vectors >= WF G-vectors >= Components
1) G-vectors is the maximum number of vectors used in the dft code. This can be larger than the number of reciprocal vectors used to describe the wavefunction (WF), e.g. to calculate the charge density. The other two numbers corresponds to the G vectors for the wavefunctions.In particular Components gives the maximum number of G vectors used to describe the wavefunction in reciprocal space (maximum because the number can change depending on k vector).WF G-vectors is used to handle the way of indexing the component in PWscf. (In fact when using the WF from other dft code eg. Abinit Components is set = WF G-vectors).

I hope this help you. Have fun with the tutorials!

Regards,
Myrta

[Conor Hogan]

Just to add: the p2y interface does indeed take the list of G-vectors from the charge density, so in that case the number is much greater than that of thewavefunctions. Often this is larger (or much larger, depending upon what you want to calculate) than what is needed to run the code, so you might reduce this value in the setup runlevel (yambo -i).
a2y instead takes the list of G-vectors provided in the Abinit KSS file to describe the wavefunctions. In addition, the G-vectors in that case are centred at gamma (not k-dependent, as in the Pwscf case), and ordered such that the WF components correspond exactly to the list of G-vectors needed to describe the wavefunctions, and is exactly equal to the total list of G-vectors given by Abinit (generally none of these conditions occur in PWscf).
Hope that is enough about G-vectors to satisfy you!
Conor

[paolalup]

Ok, now I have that clear
Thank you very much for your reply and for being that fast!!

Cheers,

Paola

----------------
Paola Lupieri
PhD student
German Research School for
Simulation Sciences GmbH
Forschungszentrum Juelich
Germany

[mapo]

Dear YAMBO team,

Currently, I running yambo calculations with the GPL Version 3.4.1 Revision 3187 on a semiconductor system and found that when I try to set the MaxGvecs I can not use a value larger then 74521. The maximum value of MaxGvecs for my system is 175209.
My procedure is

yambo -i -V -RL

setup # [R INI] Initialization
MaxGvecs= 175209 RL # [INI] Max number of G-vectors planned to use

Then I edit this line, lets say MaxGvecs=100000, and then I type

yambo -N

the yambo.in and r_setup files are created.

In yambo.in I get
setup # [R INI] Initialization
MaxGvecs= 100000 RL # [INI] Max number of G-vectors planned to use

However when I check r_setup I got

.-Input file : yambo.in
| setup # [R INI] Initialization
| BoseTemp= 0.000000 eV # Bosonic Temperature
| MaxGvecs= 74521 RL # [INI] Max number of G-vectors planned to use

Also, when I check the [02.03] RL shells I found that the first RL shell is 74521 (see attachement)

I do not find this problem if I use a MaxGvecs value smaller than 74521

I am wondering if I have to do something special when using fragmented files.

Many thanks in advance.

Miguel A Perez.
University of Oxford
UK

[Daniele Varsano]

Dear Miguel,
it is not vey clear to me what is going on.
MaxGvector it is meant to reduce the gvector with respect the default (the maximum).
So as you did:

Code: Select all

yambo -i -V RL

this create the input files. You can edit it, and then run:

Code: Select all

yambo -N

Anyway when changing the number of Gvector in order to create a new gvector database ((n)db.gops) you need to remove it from the SAVE directory.
Probably you have old ndb.gops database in the SAVE directory, so the code simply do nothing and report the number of the Gvector present in the old database.
From the report file you posted I presume that you cut by hand part of the report as I cannot see the electronic structure part.

Best,
Daniele

[mapo]

Dear Daniele,

Thanks for you fast reply.

Before running yambo -i -V RL I always deleted the following files contained in the SAVE directory
ndb.gops
ndb.kindx
ndb.kindx_fragment_1
ndb.kindx_fragment_2

then I follow the procedure described in my previous email.

I attach the yambo.in and r_setup files

Best,

Miguel A. Perez
Dept. of Materials
University of Oxford
UK

[mapo]

In my previous email the files were generated using the max value of MaxGvecs.

[Daniele Varsano]

Dear Miguel,
what happen if you run yambo without using the variable MaxGvecs? Of course after having deleted the databases.
Just to be sure you new input file is named yambo.in and *NOT* yambo.in.txt.
Can you also post the l_setup?

Daniele

PS1: You can fill your signature in the user profile panel, so you do not need to write it in every post.

[mapo]

Dear Daniele,

I have run a extra calculations in order to figure out what is happening with my system.
Basically, I run a SCF and NSCF calculations with different values of Ecutoff:
50, 100, 150 and 200 Ry (the last value is the one I have been using in my calculations)

The total number of G vector per electron is (data form the QE output files)
Ecutoff Ry G-vect Total( times 48 since my system has 48 e-)
50 458 21909
100 1296 61861
150 2372 113815
200 3651 175209

Once the SCF and NSCF calculation are completed, I proceed to run p2y and yambo without using the MaxGvecs flag, and found that
when I use an energy cutoff equal or larger than 150 Ry, yambo automatically set the MaxGvecs variable to 78177, as you can see in the
attached files. Moreover, with Ecutoff equal or larger than 150 Ry, yambo prints the following extra line at the end of the r_setup file:
| MaxGvecs= 78177 RL # [INI] Max number of G-vectors planned to use
It seems that the issue is related with the maximum number of G-vectors per e- that yambo can use.
I trying to track within the source code where yambo sets the MaxGvecs, but I have not succeeded yet.

I really would appreciate you comments

Best,
Miguel A. Perez
Dept. of Materials
University of Oxford