Large GW gap

[sdwang]

Dear developers:
I performed GW correction for my system 1D BN nanoribbons.I have the processes as follows:
>yambo -i -V 1
>yambo
>yambo -g n -p p -V 1
>yambo
attachment is my scf, nscf, log ,report and yambo.in for GW calcuation.
I have tested the convergence with FFTGvecs, NGsBlkXp,and finally I get the GW gap is very large with about 14.68 ev, the lda gap is only 4.34 eV.My system is direct gap so I only calculated the gamma point.
Why my results have so large gap? I hope you can help me find my faulse.
Thanks!

[Daniele Varsano]

Dear SdWang,

the problem of your calculation is that you are treating a 1d system,
and in this cases there are problem of divergences in the integrals of
the coulomb potential in the Bz. With your sampling you are doing
a 1d integral instead that a 3d integral.The error in this case is particularly
big for the exchange part of the self energy that tends to explode
increasing the k-point sampling. The problem can be overcome
by using the random integration method (RIM). You can activate it
by adding the yambo -c option and activating these two lines:

Code: Select all

RandQpts = 1000000		#	(RIM)	 Number of random q-points in the BZ
RandGvec	RL = 1 	#	(RIM)	Coulomb interaction RS components

Check these values, looking at the exchange part that is the less expensive.
I invite you to read the documentation about the RIM.
This part is also documented in the Yambo paper

Cheers,

Daniele

[sdwang]

Thank you for your reply so soon.
Does it mean I use >yambo -g n -p p -c instead of >yambo -g n -p p ?
Thanks!

[Daniele Varsano]

Dear Shudong,

Code: Select all

>yambo -c -g n -p p

in this way you are skipping the calculation of the exchange part with the RIM, which is the
main source of your error.
I suggest you to remove your ./SAVE/db.pp and ./SAVE/db.HF_and_locXC ./SAVE/db.QP
next select this runlevel

Code: Select all

>yambo -c -x

and then:

Code: Select all

>yambo -c -g n -p p

,
or just do the yambo -c calculation alone, and next keep the "-c" option for all your calculations.

Cheers,

Daniele

[sdwang]

for the >yambo -c,there is
% CUTBox
0.000 | 0.000 | 0.000 | # [CUT] [au] Box sides
If I chose CUTGeo= "box yz"(my 1D unitcell along x direction), the last two data are the vaccum length or the unitcell+vacuum length?
Thanks!

[Daniele Varsano]

Dear Shudong,
in this way you activate a cutoff with parallelepiped shape.
There you decide the box side. They have to be chosen in a way that
contains your system and at the same time interactions between images are avoided.
So, it is relate with your supercell size. You can have a look to this paper:

• An exact Coulomb cutoff technique for supercell calculations.
  C. A. Rozzi, D. Varsano, A. Marini, E. K. U. Gross, and A. Rubio,
  Phys. Rev. B 73, 205119 (2006)

or this one:

• Jarvis, White, Godby and Payne
  Phys. Rev. B 56,14972 (1997)

In order to calculate a box cutoff, you need the RIM first.
Anyway, using the RIM, without activating the cutoff, you solve the great part of your problem, next
the cutoff is used to speed up the convergence with the supercell size.

Cheers,

Daniele

[sdwang]

In the yambo ducumentations, it said that "For a slab periodic along x and y, set to "box z"." But for my 1D nanoribbon along x direction, should I set "box yz" or other set?
Thanks!

[sdwang]

In the yambo ducumentations, it said that "For a slab periodic along x and y, set to "box z"." But for my 1D nanoribbon along x direction, should I set "box yz" or "cylinder“?
Thanks!

[Daniele Varsano]

Dear Shudong,
you can choose the geometry you prefer.
Then, as usual, there are parameter to check!

Cheers,

Daniele

Let me remember that if you don't' use the RIM in calculating the box cutoff it will not work.

[sdwang]

Dear Shudong,
you can choose the geometry you prefer.
Then, as usual, there are parameter to check!

Cheers,

Daniele

Let me remember that if you don't' use the RIM in calculating the box cutoff it will not work.

Yes ,I use the RIM in calculating process.
ThankS!