Band structure of GW

[davood]

Hi
Dear all
I calculated the Gw level of two-dimensional AlN monolayer, and obtained band structure (GW and LDA).
In Figure of band strucure, the valence bandwidth found with GW is smaller the with the LDA, contrary to what is generally found and expected.
The results presented look very strange.why?
So, could you please tell me from where the problem originate? Thanks a lot!

[Daniele Varsano]

Dear Davood,
in order to see if there is something odd, can you please also post the report file of the GW calculation?
Best,
Daniele

[davood]

Hi
Dear Daniele
I sent for you.

Best
Davood

[Daniele Varsano]

Dear Davood,
please post the report r_setup* , r_HF* etc., as from the standard output (l_settup*) cannot see much.
Best,
Daniele

[davood]

Hi
I sent for you.
Best

[Daniele Varsano]

Dear Davood,
first of all I invite you to check , convergence parameter, for instances you are using 4000 Gvector for the exchange are you sure they are enough? The total charge has (191907 Gvectors!!). The same for the Blk in the calculation of the screening,
Beside that, assuming that you checked the convergences it looks there are problems with the interpolation.
Have a look to the report:
The calculations for instances indicates at Gamma a gap of 5.27eV while in your bands you have 5.53eV.
It looks that the interpolation did not work, may be the k-point sampling you have and the input circuit are not optimal.
My suggestions:

1) Check from the *report* , and not from the interpolation if your calculations are meaningful.

Starting from line 1915:

Code: Select all

 [10.02] QP properties and I/O
  =============================

  Legend (energies in eV):
  - B  : Band       - Eo  : bare energy
  - E  : QP energy  - Z   : Renormalization factor
  - So : Sc(Eo)     - S   : Sc(E)
  - dSp: Sc derivative precision

  - lXC: Starting Local XC (DFT)
  -nlXC: Starting non-Local XC (HF)

  QP [eV] @ K [1] (iku): 0.00      0.00      0.00
   B=1 Eo=-13.51 E=-12.36 E-Eo=  1.15 Z=-.30 nlXC=-29.43323 lXC=-15.34517 So= 10.08324
   B=2 Eo= -2.33 E= -5.25 E-Eo= -2.92 Z=0.69 nlXC=-20.69772 lXC=-11.74623 So=  4.72557
   B=3 Eo= -0.73 E= -3.27 E-Eo= -2.54 Z=0.72 nlXC=-24.80763 lXC=-16.71330 So=  4.55399
   B=4 Eo= -0.73 E= -3.26 E-Eo= -2.54 Z=0.72 nlXC=-24.80389 lXC=-16.71024 So=  4.55162
   B=5 Eo=  2.97 E=  2.01 E-Eo= -0.96 Z=0.77 nlXC=-3.487685 lXC=-7.165117 So=-4.931324
   B=6 Eo=  5.13 E=  4.33 E-Eo= -0.79 Z=0.74 nlXC=-1.576256 lXC=-4.096622 So=-3.596116

you have the detailed output of your calculations, and you can build partially the band structure.
For instance the first valence band, having a so small Z values does not make much sense.

2) ) You can think to switch to 3.4 version of the code. Many bugs have been fixed, and the 3.3 it is not developed anymore. This strongly suggested.
Other stuff to check:
a) convergence paramenters in general if you not have already done
b) Please check also the q--> direction:
% LongDrXp
1.000000 | 0.000000 | 0.000000 | # [Xp] [cc] Electric Field
%

may be a 1 | 1 | 0 | is more correct as you have a 2D system, but may this will not have much effect.
c) The cutoff has been built using 7985 Gvectors, they could be very few. This is because you set up the MaxGvecs variable.


Hope it helps
Daniele

[Daniele Varsano]

Dear Davood,

I also noticed you are using a cutoff distance which is nearly the size of the supercell, a reasonable condition in order to cut spurious interactions is to set L_cut about half of the supercell size (see PRB 73, 205119, 2006).

Best,
Daniele

[davood]

Dear Daniele
Thank you for your quick reply.
We calculated Gw level using yambo 3.4 and result is good (better than result of yambo 3.3).
But we have a qustion for you:
The size of the my supercell is
-----------------------------
... with scaling factors [a.u.]: 5.09019 5.09019 34.01507.
-----------------------------------------
Is correct input file of GW level?
--------------------------------------------------
CUTGeo= "box z" # [CUT] Coulomb Cutoff geometry: box/cylinder/sphere
% CUTBox
0.00000 | 0.00000 | 18.00000 | # [CUT] [au] Box sides
-------------------------------------------------

Best regard.
Davood

[Daniele Varsano]

Dear Davood,
In line of principle yes, in practice a cutoff side of 18 au, is equivalent to cut the interaction beyond 9 au. This is related to the subroutine computing the modified potential,
where there is a factor two to be taken into account.
For a sheet it could be enough, anyway if you consider to set 0.0 | 0.0 | 30 (which means cutting above 15au) it should be safer.

Best,
Daniele

[davood]

Dear Daniele
Thank you for your quick reply.

Best
Davood