weird symmetry problem with electric fields (lelfield

[chwolf]

Dear all,

I am currently trying to calculate the effect of an electric field (PWSCF: lelfield-.t., efield_cart(3)=0.002). When run without field the absorption spectrum (-o c) looks quite nice, but when the lelfield is turned on, it seems that QE messes up the kpoints and I am left with an error complaining about

Code: Select all

[03] Transferred momenta grid
 =============================


 IBZ Q-points :  7
 BZ  Q-points : 12

 Q [00001] : 0.00      0.00      0.00     (iku) * weight   0.08333
 Q [00002] :  0.08333   0.16667   0.00000 (iku) * weight  0.166667
 Q [00003] : 0.166667  0.333333  0.000000 (iku) * weight  0.166667
 Q [00004] :-0.250000 -0.500000  0.000000 (iku) * weight   0.08333
 Q [00005] : 0.250000  0.166667  0.000000 (iku) * weight  0.166667
 Q [00006] : 0.333333  0.333333  0.000000 (iku) * weight  0.166667
 Q [00007] : 0.500000  0.333333  0.000000 (iku) * weight  0.166667
  
 Q [00001] : 0.00      0.00      0.00     (cc ) * weight   0.08333
 Q [00002] :  0.04318   0.07478   0.00000 (cc ) * weight  0.166667
 Q [00003] : 0.086354  0.149569  0.000000 (cc ) * weight  0.166667
 Q [00004] :-0.129531 -0.224353  0.000000 (cc ) * weight   0.08333
 Q [00005] : 0.129531  0.074785  0.000000 (cc ) * weight  0.166667
 Q [00006] : 0.172707  0.149569  0.000000 (cc ) * weight  0.166667
 Q [00007] : 0.259061  0.149569  0.000000 (cc ) * weight  0.166667

 :: Indices: polarization function

 [ERROR] STOP signal received while in :[03] Transferred momenta grid

 [ERROR] Error search for G=G1-G2 !!

interestingly the nscf k points (manual input) are the same and so are the scf/nscf kpoints from the output. However, above section from thh run without field reads

Code: Select all

 IBZ Q-points :  7
 BZ  Q-points :  36

 Q [00001] : 0.00      0.00      0.00     (iku) * weight   0.02778
 Q [00002] :  0.08333   0.16667   0.00000 (iku) * weight  0.166667
 Q [00003] : 0.166667  0.333333  0.000000 (iku) * weight  0.166667
 Q [00004] :-0.250000 -0.500000  0.000000 (iku) * weight   0.08333
 Q [00005] : 0.250000  0.166667  0.000000 (iku) * weight  0.166667
 Q [00006] : 0.333333  0.333333  0.000000 (iku) * weight  0.333333
 Q [00007] : 0.500000  0.333333  0.000000 (iku) * weight   0.05556
  
 Q [00001] : 0.00      0.00      0.00     (cc ) * weight   0.02778
 Q [00002] :  0.04318   0.07478   0.00000 (cc ) * weight  0.166667
 Q [00003] : 0.086354  0.149569  0.000000 (cc ) * weight  0.166667
 Q [00004] :-0.129531 -0.224353  0.000000 (cc ) * weight   0.08333
 Q [00005] : 0.129531  0.074784  0.000000 (cc ) * weight  0.166667
 Q [00006] : 0.172707  0.149569  0.000000 (cc ) * weight  0.333333
 Q [00007] : 0.259061  0.149569  0.000000 (cc ) * weight   0.05556

 :: Indices: polarization function
 :: Indices: Self-Energy

i.e. the BZ points are different.

does anyone know a remedy to this behavior?

scf/nscf/yambo report files are attached, the PP is from the ONCV repository and the system is a double layer of graphene.

Any help is greatly appreciated!

Best,
Chris

[Daniele Varsano]

Dear Chris,

at first look, it seems k points in the IBZ are the same, with different weights due to the reduced symmetry properties of the system.
We will try to reproduce your error and spot the problem,

Best,
Daniele

[Daniele Varsano]

Dear Chris,
I've been checking a bit your problem and there is a problem coming from the mesh of QE and the time-reversal symmetry added by Yambo.
In order to solve it you can add in your nscf input file the flag:

Code: Select all

nosym=.true.

together with an automatic generation of k points.

Code: Select all

e.g.
K_POINTS (automatic)
3 3 1    0 0 0

If you have a reason to use a user-defined grid, you can still do that, but when adding the nosym=.true. option, you need to define a regular grid in the whole BZ, i.e. a 9 point grid in your case, which should not be difficult to derive.

Best,

Daniele