Some query in updated yambo 4.3.0

[sitangshu]

Dear Yambo group: Kudos!! Excellent job this time...

Meanwhile I did a small double grid computation that generated 3600 k points. I find that there is a line
DbGdQsize= 1.000000 # [X,DbGd][o/o] Percentual of the total DbGd transitions to be used

what this line means and what I should put here? Initially when I ran with the default values, it was telling k-->k-q transitions, but it terminated after a while. I am using 16.1.8.1.1 cpu structure.
I found this to be also a new term:
#BSSPertWidth # [BSS] Include QPs lifetime in a perturbative way

Please suggest...
Regards,
Sitangshu

[Davide Sangalli]

Dear Sitangshu,

the first variable

Code: Select all

DbGdQsize

is active when you compute the response function using
the double-grid technique either in G-space or with the inversion solver in BSE:
http://www.yambo-code.org/input_file/ypp/ypp_rim.php
In the past the technique was called "RIM" since the second grid was meant to be a random grid.
We renamed it "Double Grid" since the second grid in principle can be anything.

You can ignore it if you are not using a double grid.
The value of "DbGdQsize" goes from 0 to 1, where 1 corresponds to the most accurate calculation, i.e. yambo fully uses the double grid.
Lowering its value only a part of the double grid is used.

The second variable

Code: Select all

BSSPertWidth

refers to the perturbative approach to BSE when quasi-particles lifetimes are used.
The equations are the same as the ones for perturbative SOC used in the supplementary of this PRL: https://journals.aps.org/prl/abstract/1 ... 111.216805
However now the perturbation is not the SOC.
Thus, if a BSE calculations loading the QP lifetimes is performed (as in this PRL: https://journals.aps.org/prl/abstract/1 ... 101.106405), yambo can
(a) solve the BSE for the non hermitian hamiltonian
(b) solve the hermitian problem without the QP lifetimes and then add a width to each pole with a perturbative approach
Case (a) is the default, case (b) is activated by the flag BSSPertWidth

[sitangshu]

Dear Daniele,

Thanks for your reply. But it seems that there could a bug....
All I did is first a pwscf on 8 8 8 k points resulting in a 65 total k points. I compute the GW also to get ndb.QP file. Then I did a double nscf at 12 12 12 resulting in 468 k points. But when I ypp BZ map fine grid to course, it says:

<---> :: K points [DB #1]: 468
<---> :: [imposed]: 468
<---> :: Bands : 100
<---> :: Symmetries : 12
<---> :: Time reversal not included
<---> :: Gap indirect [ev]: 3.443296
<---> :: direct [ev]: 4.924504
<---> [WARNING] Internal check failed. A consistent X grid k-point was not found among the Double Grid list
<---> [07] BZ Double Grid K-points expansion ...found 5184 (468) points in the BZ (IBZ)
<24s> BZ Double Grid Map |########################################| [100%] 24s(E) 24s(X)
<24s> :: Blocks filling range : 6 25
<24s> Built-in grid |########################################| [100%] --(E) --(X)
<24s> [WR./SAVE//ndb.Double_Grid]---------------------------------
<24s> Bands : 100
<24s> Fine grid number of points : 533
<24s> BZ ASSOCIATED k-points : 512
<24s> BZ Blocks max filling : 25
<24s> Points expanded in the BZ :yes

This number 533 is reflecting in the yambo BSE calculation:
[02.05] Double K-grid
=====================

K-points : 533
Bands : 100
Is the grid expanded :yes

I cross checked the wfc.dat in the SAVE file of the double nscf and I found there are 468 such wfc.dat.
The net effect is that lines like this:
<02m-47s> P0001: DbGd@q[33] k->k-q transitions included [%]: 4.378816
are appearing, but sudden termination is occurring. Am I supposed to get this line?

As far as I remember, in my earlier calculations with yambo 4.2.3, with double grid I did not find this problem.

Please suggest..
Regards,
Sitangshu

[Davide Sangalli]

Dear Sitangshu,
the numbers seem fine to me.

I guess your 12x12x12 grid is not gamma centered.
Then pwscf uses 3 shifted grids according to the supercell kind.
5184 = 12*12*12*3 (BZ), with 468 in the IBZ
Instead the 8x8x8 is probably gamma centered.
512=8*8*8 (BZ), with 65 in the IBZ.

For the message "Fine grid number of points", it is just the text which is not correct.
It should have been "Double grid number of points (IBZ)".
Indeed 533=468+65
We will fix the message

Finally the fact that the program stops with the message

Code: Select all

<02m-47s> P0001: DbGd@q[33] k->k-q transitions included [%]: 4.378816

is to be checked. I understood you are running in parallel.
Can you check all the log files if there is any error message ?
You can just do

Code: Select all

grep -i error LOG*

Best,
Davide

[sitangshu]

Dear Davide,

Many thanks for your justification. Yes the grid is off-gamma. But, I am still getting the error. But when I do it without the grid, it runs smoothly. Thus I realized it might be due to the double grid. I think I am missing something here. I am pasting the input file which is giving me the error.

Can you please look into it and tell me where I am wrong?
P.S: The BSE is using GW (on 8 8 8) result here.


Regards,
Sitangshu

[Davide Sangalli]

Dear Sitangshu,
it's not easy to tell from just the output and input files.
I see just 1 log file, don't you have one per core ?

Best,
Davide