SCR calculation bug, EMERGENCY!

[zhooshmand]

Dear All,
I am already calculating GW calculations on a system. in 2nd step, I mean SCR calculations, I have this message:
setmesh: npwwfn = 2625; Max (m1,m2,m3) = 6 6 14
npweps/npwsigx= 2625; Max (mm1,mm2,mm3)= 7 7 14
mG0 added = 1 1 0
calculated ecutwfn = 7.995 [Ha]
calculated ecutsigx/ecuteps = 7.995 [Ha]
using method = 2 with ecuteff = 31.978 [Ha]

finding a FFT mesh compatible with all the symmetries
setmesh: divisor mesh 1 1 1
setmesh: FFT mesh size selected = 25x 27x 60
total number of points = 40500


setup_coulombian : cutoff-mode = CRYSTAL
q-points for optical limit: 1
1) 0.000010 0.000020 0.000030

setup_screening.F90:511:ERROR
Mismatch between Dtset%wtk and Kmesh%wt
-P-0000
-P-0000 leave_new : decision taken to exit ...

I dont know what is exactly the problem?! is there any variable missed? please Note that these calculations belong to the last part of Brillouine zone calculations on a system, if you need to know the exact reason for variables setting.

I really need your help as an emergency!
I have attached tgw2_1.in, tgw2_1.out,tgw2_2.in.

yours sincerely
Zahra

[bruneval]

The shifts of your kpoint grids are weird.
Try to turn to, let's say, a unique Gamma centered grid.

Fabien

[lorenzo.sponza]

I address again this subject.

I am using abinit 6.12.3 and I am making some test calculation on Silicon to compute the SCR file.
When I try to compute SCR with a non-shifted grid 3x3x1 I can not start the SCR calculation, exiting with error:
setup_screening.F90:659:ERROR
Mismatch between Dtset%wtk and Kmesh%wt
Indeed the list of k-points generated in the SCR calculation using symmetries is not correct at all and the weights of the k-points (wtk) in the echo of input variables is not the same of that computed in the K-mesh for wavefunctions.
My input file is in attachment.

I also tried to change kptopt, changing to 2,3,4 and even 0. Each time an error of the kind
m_bz_mesh.F90:399:BUG
Not allowed value for kptopt: 2
Action : contact ABINIT group.
or similar.

Just to conclude, If I try the same calculation with Graphite (cell has a different symmetry) it works.
Can you figure out what's wrong with it? Thanks a lot.

Lorenzo

[gabriel.antonius]

Dear Lorenzo,

The problem is that you are using a non-regular k-point mesh. It is a 2D mesh, while your cell has 3D symmetry. I'm not surprised it works in graphene!

Try breaking the symmetry like this:

Code: Select all

acell   10.217   10.217   10.2170001

You should get a different but manageable error message:

Not able to found umklapp G0 vector among ***** vectors
Increase mg0sh such as k1-k2 = kf+G0, present value = 50
point1 = 5 -3.33333333E-01 -3.33333333E-01 -3.33333333E-01
point2 = 3 0.00000000E+00 3.33333333E-01 0.00000000E+00

The parameter "mg0sh" is hard-coded. Is is a guess of how anisotropic your k-point grid might be.
Just go in the subroutine "70_gw/setup_screening.F90" and look for the line "mg0sh=50".
You can change it to some higher value.

Cheers!

[lorenzo.sponza]

Great Gabriel, thanks for the answer!
Cheers.